Abstract Concepts and the Embodied Mind 9780190061999, 2022008559, 9780190061975

Table of contents :
Cover
Abstract Concepts and the Embodied Mind
Copyright
Dedication
Contents
Acknowledgments
1 Introduction
1.1 Embodied Cognition
1.2 Abstract Concepts
1.3 The Road Ahead
2 The Conceptual Brain
2.1 Two Dogmas of Concept Research
2.2 Questioning These Dogmas
2.3 Something We Do with the Mind
2.4 Elasticity and 4E Cognition
2.5 Staying Grounded
3 Body in Mind
3.1 Perception Systems
3.2 Action Systems
3.3 Simulation and Its Discontents
3.4 The Symbol Grounding Problem
3.5 Weak Versus Strong Embodiment
4 Three Problems
4.1 The Problem of Generalization
4.2 The Problem of Disembodiment
4.3 The Problem of Flexibility
4.4 Abstract Concepts Reconsidered
5 Hierarchies and Hubs
5.1 Beyond Unimodality
5.2 Hierarchical Organization
5.3 Higher-​Level Representations
5.4 Hubs and Spokes
5.5 Multimodal and Multilevel Representations
6 Language Is a Neuroenhancement
6.1 A Symbolic Medium
6.2 Thinking Without Words
6.3 Word and Thought
6.4 Labels
6.5 Word Associations
6.6 Syntax
6.7 Conversations
6.8 Abstract Concepts in the Brain
6.9 Language and the Embodied Mind
7 Heterogeneity
7.1 Affective Embodiment
7.2 Dimensions of Variation
7.3 Multiple Mechanisms
7.4 Grounding Reconsidered
8 Growth and Development
8.1 The Acquisition of Abstract Words
8.2 Relational Categories
8.3 Theory of Mind
8.4 Developmental Language Disorder
8.5 Multiple Cues
9 Metaphor
9.1 Conceptual Metaphor Theory
9.2 Language, Discourse, and Culture
9.3 The Living and the Dead (or Sleeping)
9.4 Development
9.5 Metaphors Are Elastic
10 The Elastic Mind
10.1 Causal Relevance
10.2 Theoretical Challenges
10.3 Hierarchical Representations
10.4 The Influence of Language
10.5 Synchronic and Diachronic Flexibility
10.6 Metaphor: A Case Study
10.7 The Elasticity of Abstract Concepts
10.8 The Future Is Now
References
Index

Citation preview

Abstract Concepts and the Embodied Mind

Abstract Concepts and the Embodied Mind Rethinking Grounded Cognition G U Y D OV E

Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America. © Oxford University Press 2022 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. Library of Congress Control Number number: 2022008559 ISBN 978–​0–​19–​006197–​5 DOI: 10.1093/​oso/​9780190061975.001.0001 1 3 5 7 9 8 6 4 2 Printed by Integrated Books International, United States of America

To Jenny, for all the encouragement, patience, and humor.

World is crazier and more of it than we think, Incorrigibly plural. I peel and portion A tangerine and spit the pips and feel The drunkedness of things being various. —​Louis MacNeice

Contents Acknowledgments 

ix

1 Introduction 

1

1.1 Embodied Cognition  1.2 Abstract Concepts  1.3 The Road Ahead 

2 4 6

2 The Conceptual Brain 

10

3 Body in Mind 

25

4 Three Problems 

50

5 Hierarchies and Hubs 

64

6 Language Is a Neuroenhancement 

97



2.1 Two Dogmas of Concept Research  2.2 Questioning These Dogmas  2.3 Something We Do with the Mind  2.4 Elasticity and 4E Cognition  2.5 Staying Grounded  3.1 Perception Systems  3.2 Action Systems  3.3 Simulation and Its Discontents  3.4 The Symbol Grounding Problem  3.5 Weak Versus Strong Embodiment  4.1 The Problem of Generalization  4.2 The Problem of Disembodiment  4.3 The Problem of Flexibility  4.4 Abstract Concepts Reconsidered  5.1 Beyond Unimodality  5.2 Hierarchical Organization  5.3 Higher-​Level Representations  5.4 Hubs and Spokes  5.5 Multimodal and Multilevel Representations  6.1 A Symbolic Medium  6.2 Thinking Without Words  6.3 Word and Thought  6.4 Labels 

10 12 15 18 22 26 31 36 44 47 52 54 58 62 64 72 77 88 94

98 107 110 117

viii Contents

6.5 Word Associations  6.6 Syntax  6.7 Conversations  6.8 Abstract Concepts in the Brain  6.9 Language and the Embodied Mind 

121 123 126 130 132

7 Heterogeneity 

134

8 Growth and Development 

154

9 Metaphor 

175

10 The Elastic Mind 

199

References  Index 

211 255



7.1 Affective Embodiment  7.2 Dimensions of Variation  7.3 Multiple Mechanisms  7.4 Grounding Reconsidered  8.1 The Acquisition of Abstract Words  8.2 Relational Categories  8.3 Theory of Mind  8.4 Developmental Language Disorder  8.5 Multiple Cues  9.1 Conceptual Metaphor Theory  9.2 Language, Discourse, and Culture  9.3 The Living and the Dead (or Sleeping)  9.4 Development  9.5 Metaphors Are Elastic 

10.1 Causal Relevance  10.2 Theoretical Challenges  10.3 Hierarchical Representations  10.4 The Influence of Language  10.5 Synchronic and Diachronic Flexibility  10.6 Metaphor: A Case Study  10.7 The Elasticity of Abstract Concepts  10.8 The Future Is Now 

135 138 147 151 155 164 168 172 173 176 184 188 190 197 199 201 201 202 203 204 205 208

Acknowledgments I have been working on challenges posed by abstract concepts to embodied cognition for many years. My first article on this topic, entitled “Beyond Perceptual Symbols: A Call for Representational Pluralism,” was published in 2009. Since then, I have had several bites at the apple. Naturally enough, my theoretical position has evolved over time, and this book is the culmination of that process. I have used substantially revised portions of my previously published work with permission. Chapter 4 contains elements of “Three Symbol Ungrounding Problems: Abstract Concepts and Future of Embodied Cognition,” in Psychonomic Bulletin & Review (2016) and “The Challenges of Abstract Concepts,” in M. D. Robinson and L. E. Thomas (Eds.), Handbook of Embodied Psychology: Thinking, Feeling, and Acting (2021). Chapter 6 is derived in part from three articles: “Thinking in Words: Language as an Embodied Medium of Thought,” in Topics in Cognitive Science (2014); “Language as a Disruptive Technology: Abstract Concepts, Embodiment, and the Flexible Mind,” in Transactions of the Royal Society B (2018); and “More Than a Scaffold: Language Is a Neuroenhancement,” in Cognitive Neuropsychology (2019). Chapter 7 contains material from “The Challenges of Abstract Concepts.” A revised portion of “More Than a Scaffold: Language Is a Neuroenhancement” appears in Chapter 8. Many people have helped me along the way. To begin at the beginning, I need to thank Jesse Prinz for dragging me to Larry Barsolou’s classes in graduate school. In typical fashion, Jesse saw the importance of embodied cognition much earlier than I did. He continues to be a valuable guide. Although I doubt that Larry even remembers me from that time, I have been fortunate enough to have received constructive feedback from him over the years. I have also presented my ideas on these topics at various interdisciplinary meetings and forums, and I owe a debt of gratitude to everyone who participated. I want to personally thank Fred Adams, Colin Allen, Michael Anderson, Michael Arbib, Ken Aizawa, Kristin Andrews, David Barack, Leda Berio, John Bickle, Cameron Buckner, Cara Cashon, Tony Chemero, Hayley Clatterback, Louise Connell, Rutvik Desai, Felipe De Brigard, Léo Dutriaux, Luis Favela, Carrie Figdor, Martin Fischer, Arthur Glenberg,

x Acknowledgments Sean Hermanson, Zoe Jenkin, Amy Kind, Nikola Kompa, David Landy, Peter Langland-​Hassan, Gary Lupyan, Edouard Machery, Mandy Maguire, Corey Maley, Carolyn Mervis, John Pani, Penny Pexman, Martin Pickering, Tom Polger, Emily Prychitko, Friedemann Pulvermüller, Joanna Raczaszek-​ Leonardi, Chris Rickels, Sarah Robins, Robert Rupert, Elizabeth Schechter, Larry Shapiro, Shannon Spaulding, Charles Starkey, Evan Thompson, Gabriella Vigliocco, Daniel Weiskopf, Bodo Winter, and Tadeusz Zawidzki for helpful conversations about abstract concepts and grounded cognition. I also want to thank my present and former colleagues in the department of philosophy at the University of Louisville, Asaf Angermann, Julianne Chung, Andreas Elpidorou, Lauren Freeman, John Gibson, Stephen Hanson, Avery Kolers, Tom Maloney, Dismas Masolo, Cecilea Mun, David Owen, Bailey Thomas, and Boomer Trujillo, for putting up with me as I struggled to finish this project. I also thank Mark Wilson for letting me use his figures in the final chapter. This book would not have happened without the direct help of a few key individuals. I thank my editor, Martin Baum, for his patience and hard work. I also thank Joan Bossert for her initial help with this project and Emily Benitez for helping me with its completion. Shelby Roberts helped me look over the manuscript. Jaime Reilly provided invaluable constructive advice. I am especially indebted to Laura Barca, Anna Borghi, and Luca Tummolini for their generous feedback and collaboration on the role of language in the acquisition of abstract concepts. Finally, I need to thank David Kemmerer for providing critical feedback throughout this process, from its inception to the final draft.

1 Introduction Our thoughts depend on knowledge about objects, people, properties, and events. In order to think about where we left our keys, what we are going to make for dinner, when we last fed the dogs, and how we are going to survive our next visit with our family, we need to know something about locations, keys, cooking, dogs, survival, families, and so on. This book examines how our brains can store and access such general knowledge about the world and our place in it. More specifically, it investigates the neurocognitive mechanisms responsible for encoding our concepts. Sorting objects, events, and experiences into categories enables us to recall and use information that we have gathered over time. Stored information about these categories helps us make decisions, communicate, and respond in an intelligent way to changing circumstances. Our concepts serve as the building blocks for many forms of thinking. We use them to recognize patterns, draw inferences, make decisions, and understand word meanings. Concepts are bodies of knowledge that are quickly accessed in various situations (Machery, 2009; Margolis & Laurence, 1999; Murphy, 2002). Consider an everyday concept such as that of a pug.1 This concept aggregates information gathered through physically interacting with pugs, talking to other people about them, reading about them, and perhaps watching Tik-​Tok videos of them. It gives us access to information about their shape, fur, movements, general disposition, propensity to overheat, and much more. It may contain language-​related knowledge such as the fact that a cobby pug exhibits the short, stocky body type favored by breeders and the fact that a group of pugs is called a grumble. This concept can be used to imagine what a new pug might look like or to anticipate how a particular pug might behave.

1 In the interest of clarity, I am going to follow a convention that has been adopted by cognitive scientists in which concepts are presented in uppercase letters in order to distinguish them from words (which I will often italicize). Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0001

2  Abstract Concepts and the Embodied Mind Language can play a crucial role in concept acquisition. For example, a couple of years ago, my wife and I read a novel by a British author in which the narrator talked about her father retreating to his growlery. We had never run across this word before, so we looked it up and discovered that it was Briticism for a room—​often a den or study—​that is used as a place of refuge. We have been so taken with this term that we now use it regularly. Indeed, I am currently writing these words in my growlery. Some caution is warranted before we draw broad conclusions from this example, though. My wife and I were able to learn the meaning of this term so quickly because we already had access to a great deal of related background knowledge: we were familiar with dens and studies, we understood the sort of events that might lead one to engage in the visceral act of growling, and we recognized the solace that can be provided in a shared household by having a room of one’s own. Significantly, we can have concepts without words or labels. Consider the once ubiquitous hip-​hop dance move, the dab. This move involves a sort of sneeze-​like movement of the head into the crook of one arm while simultaneously throwing out the other arm. It is likely that the first instances of dabbing occurred in the absence of a label. It is also possible to have the concept without knowing the term. This unlabeled concept might precede questions of the following sort: What was that? Does it have a name? It is also easy to imagine being familiar with the term without having the concept. This situation might lead one to google it and watch video tutorials. Maybe you are doing that now. While our understanding of word meanings relies on our concepts, we have access to concepts that are nonlinguistic. Indeed, our capacity to acquire new word meanings likely depends on such access. Learning new words often involves affixing a label to a preexisting concept. The independence of concepts from language also explains how word meanings can shift over time.

1.1  Embodied Cognition Traditionally, researchers in the psychological and brain sciences have assumed that concepts are handled by a representational system separate from and independent of the modality-​specific systems associated with our experiences of the world (Anderson, 1983; Fodor, 1975; Pylyshyn, 1984). For

Introduction  3 this reason, it is often referred to it as the amodal view of concepts. Once dominant, this orthodox view is now under threat. Over the past three decades, researchers of various stripes have gathered a large body of evidence suggesting that many of our concepts are at least partially grounded in action, emotion, and perception systems (for reviews, see Barsalou, 2016; Borghesani & Piazza, 2017; Conca & Tettamanti, 2018; Fischer & Zwaan, 2008; Gallese & Lakoff, 2005; Kemmerer, 2010, 2015a; Kiefer & Pulvermüller, 2012; Meteyard et al., 2012; Pulvermüller, 2017). Grounded concepts emerge from the identification of regularities in our experiences. These regularities encompass objects, actions, events, internal perceptions, and social interactions. The relevant information is stored and accessed through the neural systems associated with our experiences of them. In other words, we often think about the world by means of the very same neural mechanisms that we use to experience it. Embodied cognition owes an intellectual debt to the British Empiricists such as Locke, Berkeley, and Hume. Each of these philosophers proposed that conceptual knowledge depends crucially on sense experience (Markie, 2017). Given this debt, it is not surprising that some philosophers refer to embodied or grounded cognition as new concept empiricism (Prinz, 2002), neo-​empiricism (Machery, 2007), or simply concept empiricism (Löhr, 2019). Such terminology is misleading. Contemporary views of embodied cognition go well beyond the traditional parameters of empiricism in several important ways. For instance, most theories hold that grounding is not limited to conscious mental imagery but explicitly includes unconscious mental processes (Kemmerer, 2015b). In addition, grounding leaves room for rich innate predispositions (Barsalou, 2008). Many theories of grounding emphasize the importance of external physical factors and the dynamic contribution of body–​world couplings (Chemero, 2011). In sum, embodied cognition should not be characterized as a form of empiricism because doing so distorts contemporary theories of grounding and undermines our ability to assess its merits. The importance of this point is demonstrated by the fact that criticisms of neo-​empiricism in the philosophical literature often fail to acknowledge important elements of current theories of grounding. These criticisms regularly fail to consider the possibility of the sort of multimodal, hierarchical, and flexible account of grounding—​often referred to in the experimental literature as weak embodiment—​that will be the central focus of this book. While the proximal cause of this failure may be an unfamiliarity with the

4  Abstract Concepts and the Embodied Mind current empirical literature, it may also stem from the false assumption that grounding is (or perhaps should be) an updated form of empiricism.

1.2  Abstract Concepts Even though embodied cognition and empiricism are distinct, both struggle with an important feature of human cognition: our ability to think about things and events that go beyond our immediate experience. Until recently, most of the research supporting embodied cognition focused on concepts that relate to objects and events that we can directly perceive and manipulate. In other words, it focused on concepts that are highly concrete or imageable. Simply put, it has generally concentrated on the very sort of concepts that seem most likely to be grounded in experience. This raises an obvious question: Are concepts that are further removed from our immediate experience processed in the same way as those associated with more directly perceivable and manipulable referents? Our ability to acquire and use abstract concepts creates empirical and theoretical challenges for approaches to concepts that appeal to sensorimotor and affective grounding (Dove, 2016; Myachykov & Fischer, 2019). First, many abstract concepts have referents that we do not directly perceive or manipulate (Borghi & Binkofski 2014; Borghi et al., 2017). Given this, it is not immediately clear how they can be grounded in action, emotion, and perception systems (Chatterjee, 2010; Mahon, 2015; Mahon & Caramazza, 2008). Second, it is not clear that all concepts are grounded all the time (Pecher, 2018; Pexman, 2019). The fact that most of the relevant research involves concrete concepts makes the question of scope particularly pressing with abstract concepts (Dove, 2016). Third, a diverse body of behavioral and neuropsychological research suggests that abstract and concrete concepts are at least partially handled by different neuroanatomical structures (Del Maschio et al., 2021; Wang et al., 2018). All in all, our facility with abstract concepts suggests that we may need to rethink the structure and dynamics of grounded cognition. A difficulty facing any discussion of abstract concepts is that it is not exactly clear how they should be defined. This problem begins with the observation that all concepts involve a significant degree of abstraction. Concepts need to abstract away from experiential particulars to enable us draw inferences and categorize new exemplars. Simply put, abstraction is a

Introduction  5 fundamental design feature of concepts. Thus, any account of our concepts—​ and indeed the concepts of other intelligent organisms—​must offer some explanation of how this abstraction occurs. Having conceded this point, it remains the case that some concepts (e.g., absence, fun, and idea) appear to be more abstract than others (e.g., chair, hammer, and pear). What to make of this difference in degree—​and how to operationalize it in a reliable and reproducible way—​remains an important research question. Several different characterizations of the abstract–​concrete distinction have been offered. Barr and Caplan (1987) propose that we should distinguish concepts that are categorized by extrinsic features (which are associated with relations between two or more entities) from those that are represented by intrinsic features (which are associated with individual entities). Crutch and Warrington (2005) propose that we should draw a qualitative distinction between concepts that are organized around semantic association and those that are organized around similarity. Wiemer-​Hastings and Xu (2005) propose that abstract concepts are both less contextually specific and predominately associated with the social aspects of situations. Shea (2018) proposes that socially oriented metacognition might be especially important for abstract concepts. Researchers have also employed several empirical measures to trace the contours of abstractness. These measures include body–​object interaction (Siakaluk et al., 2008), concreteness (Marschark & Paivio, 1977), context-​ availability (Schwanenflugel & Shoben, 1983), emotional valence (Kousta et al., 2011), imageability (Paivio, 1986), interoceptive strength (Connell, Lynott, & Banks, 2018), semantic richness (Recchia & Jones, 2012), and strength of perceptual experience (Connell & Lynott, 2012). Research using these measures leaves us with a mixed picture because, while they correlate up to a point, they are not equivalent (Kousta et al., 2011). Admittedly, this could simply be the result of a failure of collective imagination. It could simply be the case that we have just failed to come up with the right characterization of abstractness. However, it could also be the case that abstract concepts are heterogeneous (Barsalou, Dutriaux, & Scheepers, 2018). After all, they can range over several different areas of human activity and interest (Borghi et al., 2018). Some abstract concepts involve moral and aesthetic notions such as beauty, justice, piety, and sin (Fingerhut & Prinz, 2018). Others directly involve emotions such as desolation, gratitude, pleasure, resentment, and schadenfreude (Winkielman, Coulson, & Niedenthal, 2018). Many appear to be dependent on social and

6  Abstract Concepts and the Embodied Mind cultural factors, such as celebrity, demagogue, martyr, scapegoat, and xenophobe (Rice et al., 2018). Yet another group consists of mathematical and scientific concepts, such as force, infinity, median, quark, and zero (Desai, Reilly, & van Dam, 2018). A different group involves concepts associated with language use, such as assert, cajole, lie, and promise (Dove, 2018). Languages often rely on grammatical elements and morphemes associated with abstract concepts such as modal verbs (can, may, must, shall, and will), negatives (not), and quantifiers (few, most, some, and all). They can differ with respect to which of these elements are lexicalized (Kemmer, 2019). The apparent heterogeneity of abstract concepts suggests that the search for a unified, single-​factor definition of them may be a fool’s errand. It may be better to think of abstract concepts in terms of a multidimensional space (Borghi et al., 2019). Although some have suggested that the heterogeneity of abstract concepts means that they no longer pose a threat to grounded cognition, this heterogeneity may lead to a metastasis of the theoretical cancer rather than a cure (Dove, 2021).

1.3  The Road Ahead I am going to argue that abstract concepts pose serious challenges for grounded cognition. An inclusive account of all concepts is possible, but it requires updating and transforming our understanding of what it means for concepts to be grounded. Theories designed to fit the data collected from concrete concepts are simply not up to the task. My argument proceeds as follows. Chapter 2 proposes that we need an elastic conception of embodiment and grounding in which experiential resources are stretched to handle categories that are not tied to the direct experience of our immediate physical environment. This approach has three main elements. The first is a commitment to multimodality. Concepts are handled in part by representations that are distributed across multiple experiential modalities, including those associated with action, emotion, exteroception, and interoception. The second is a commitment to hierarchical representations. Conceptual grounding is not limited to representations located within primary sensorimotor areas. The third is a commitment to flexibility. Concepts are realized in a context-​and task-​sensitive fashion. Each of these elements facilitates our capacity to use abstract concepts. The multimodal nature of our concepts enables us to use

Introduction  7 inner experiences associated with social cognition to capture aspects of abstract concepts that are not accessible through external perception and action systems. Higher-​level and heteromodal representations help us capture the generalizations needed to understand conceptual hierarchies. Flexibility, in conjunction with the cognitive scaffolding provided by language, enables us to deploy abstract concepts in a dynamic and contextually appropriate fashion. Chapter 3 critically examines the empirical support for embodied and grounded cognition. The evidence typically cited in this context characteristically involves concrete concepts and implicates perception and action systems in their processing. Rehearsing this evidence has become something of a touchstone in the field of embodied cognition—​papers regularly begin with a quick summary of it. This chapter offers a more comprehensive critical assessment of the relevant research and defends two novel claims: one that concerns the character of the available evidence and another that concerns our theoretical interpretation of it. The empirical claim is that the available evidence implicates both sensory and motor representations in conceptual processing but falls short of establishing that they are necessary for such processing. The theoretical claim is that what really matters is the causal influence of grounded representations, and the scope of this influence remains an open and pressing question. Chapter 4 examines the ways in which abstract concepts threaten grounded cognition. I argue that abstract concepts pose at least three distinct theoretical challenges: the problems of generalization, disembodiment, and flexibility. The problem of generalization emerges in the context of conceptual hierarchies where higher-​level concepts are more abstract than lower-​level concepts. It is hard to see how such abstractions can be captured solely by means of lower-​level sensorimotor representations. The problem of disembodiment arises because some abstract concepts lack any clear connections to our immediate experiences. Such concepts are not likely to arise from merely abstracting away from experiential particulars. Finally, the problem of flexibility arises because the content of abstract concepts tends to vary more in response to context than the content of concrete concepts. Each of these imperils the hypothesis that conceptual processing depends entirely on simulations of experience. Any successful account of grounded cognition must address all of them. Chapter 5 focuses on the problem generalization. After demonstrating that multimodal and multilevel structure are common design features of

8  Abstract Concepts and the Embodied Mind human action, emotion, and perception systems, I propose that higher-​level representations associated with these systems help our brains generalize and abstract away from experiential particulars. This hypothesis is supported by neuroimaging and neuropsychological evidence implicating these sorts of representations in conceptual processing. Evidence from neuropsychological case studies also suggests that conceptual content is encoded through the interaction of higher-​level heteromodal and supramodal “hubs” with lower-​ level experiential “spokes.” I conclude that an account of grounded cognition that posits hierarchical representations is well-​positioned to address the problem of generalization. Chapter 6 argues that language is a neuroenhancement for grounded minds. The fact that much of our experience is mediated by language raises the question of whether language can itself serve as a source of grounding. I propose that its symbolic properties facilitate our capacity to encode abstract semantic content in several important ways: having labels for our concepts facilitates our ability to link together our diverse experiences, word-​ to-​word associations help us draw inferences that go beyond our immediate experience, and knowledge linked to conversations and narratives enables us to tailor concepts to specific contexts and tasks. In sum, language helps our brains encode fundamentally disembodied content. Flexibility occurs at both synchronic and diachronic timescales. Chapter 7 defends the claim that abstract concepts are synchronically heterogeneous. It reviews the emerging body of evidence suggesting that different types of abstract concepts are grounded in different ways. Abstract concepts can be distinguished in terms of the degree to which they evoke action, emotion, exteroceptive, interoceptive, and linguistic systems. Some appear to rely more on sensorimotor features, others appear to rely more on emotions or inner experiences, and others still appear to rely more on verbal associations. Our ability to process abstract concepts of various stripes depends on the multidimensional nature of our conceptual system. Chapter 8 reviews recent research on abstract word learning. What emerges from this review is that children rely on multiple grounded cues to learn the meanings of abstract words—​including cues associated with emotion, iconicity, interoception, and language. The relative importance of these cues varies at different stages of development. Iconicity, for instance, appears to be very important at early stages of word acquisition. Affective information appears to play a central role during a period of rapid growth of abstract word learning that occurs from 8 to 9 years of age. Language-​based

Introduction  9 information appears to be important throughout the learning process but exerts greater influence later in development when children tend to acquire emotionally neutral abstract words. Chapter 9 argues that metaphor is a complex phenomenon shaped by grounding, culture, language, and discourse factors. Some supporters of embodied cognition hold that most abstract concepts are indirectly grounded by means of conceptual metaphors linking concrete source domains with abstract target domains. Unfortunately, this is unlikely to be a complete solution to the challenges posed by abstract concepts. The same abstract domain can be linked to several different concrete domains, and the same concrete domain can be mapped onto by several different abstract domains. Critics of conceptual metaphor theory generally argue that metaphor is largely a discourse phenomenon and point to the fact that much of the evidence for grounded metaphors is circumscribed and plagued by variability. Acknowledging these criticisms, I argue that metaphors are elastic. Different metaphors rely on different resources. Novel metaphors tend to rely more on sensorimotor grounding, and conventionalized metaphors tend to rely more on language. All metaphors are context-​sensitive and task-​ dependent to some degree. Chapter 10 chapter offers a unified assessment of the elasticity hypothesis and ties together the arguments from the previous chapters. Our facility with abstract concepts reveals that our conceptual system is multimodal, hierarchical, scaffolded, and flexible. These design features suggest that abstract concepts are more conditional than is often assumed. The concept of hardness in the material sciences is used to show that even technical concepts are provisional cognitive tools that are adapted to specific tasks and interpreted against the background of shifting practical and theoretical concerns. The chapter concludes with a discussion of the ways in which the elasticity of our concepts transforms the research program of grounded cognition. New research needs to explore the diverse sources of grounding, the influence of higher-​level representations, and the ways in which concepts are shaped by social factors and situations.

2 The Conceptual Brain This book focuses on the neural mechanisms responsible for conceptual processing. More specifically, it addresses the degree to which our concepts are grounded in action, emotion, and perception systems. My aim in this chapter is descriptive rather than argumentative. I hope to provide a brief sketch of the sort of view I am going to defend and situate it with respect to previous theories. My view holds that concepts are handled by neural representations that are distributed across multiple experiential modalities, include higher-​level representations located within heteromodal, supramodal, or even amodal areas,1 and these vary in how they are realized in different contexts and tasks.

2.1  Two Dogmas of Concept Research Research into the structure of our concepts has played—​and indeed continues to play—​an important role in the development of cognitive science (Margolis & Laurence, 1999, 2015; Murphy, 2002). Much of the early research on concepts was behavioral and focused on what sort of properties they encode. Contrary to the view inherited from a long-​standing philosophical tradition, which held that concepts involve knowledge of necessary and sufficient conditions, categorization and priming experiments suggested that concepts regularly encode the typical properties of category members (Murphy, 2002). Several theories emerged in response to this data. Prototype theories proposed that concepts involved representations that capture the central tendencies of categories (Hampton, 1979; Medin & Smith, 1981; Rosch & Mervis, 1975). Exemplar theories proposed that our concepts rely on stored information about specific category exemplars (Medin & Schaffer, 1978; 1 While there is no agreed upon definition for these terms, they are generally thought to form a hierarchy defined relative to the immediacy of their connection to modality-​specific representations. Heteromodal areas link together information from different modalities. Supramodal areas capture information that goes beyond these associative links, often by connecting different heteromodal areas. Amodal areas are distantly removed from affective and sensorimotor systems. For a discussion of multilevel models of concepts, see Chapter 5. Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0002

The Conceptual Brain  11 Smith & Medin, 1999). So-​called theory-​theories proposed that our concepts rely on implicit theories (Carey, 1985, 1991; Murphy & Medin, 1985; Wellman, 1990). Of course, researchers are interested in more than just the character of the properties contained within our concepts; they are also interested in the brain mechanisms responsible for them. Despite the controversies just outlined, most of these accounts agree about the basic features of these mechanisms. In particular, they agree on two core dogmas: (i) that concepts are handled by an independent amodal representational system (i.e., one that is not indigenous to any affective or sensorimotor modality) and (ii) that concepts are realized in a largely invariant fashion. The first dogma involves a commitment to what Hurley (2008) colorfully identifies as the “classical sandwich.”2 This view holds that our concepts are encoded by a cognitive system that is functionally and anatomically distinct from action and perception systems.3 The classical sandwich views cognition as the meat that sits between two slices of bread (action and perception). Importantly, the representations employed by this autonomous conceptual system are independent of those employed by action and perception systems. This independence is thought to have two main benefits: first, it enables these representations to collate information gathered from distinct sensorimotor modalities, and, second, it underwrites our capacity to integrate and abstract away from sensorimotor particulars (Pinker, 1997). The second dogma involves a commitment to what might be referred to as invariantism—​that is, the hypothesis that concepts are realized in roughly the same form in different contexts. Machery (2009, 2010, 2015) characterizes the standard view in terms of default knowledge. He suggests that, as repositories of default knowledge, concepts exhibit three characteristic properties (2015, p. 570): • Speed. Default knowledge is quickly retrieved from long-​term memory. • Automaticity. Default knowledge is automatically retrieved from long-​ term memory. 2 Using less colorful language, Barsalou, Brezeal, and Smith (2007) link the orthodox amodal approach to the view that cognition is divided into discrete steps of “sense-​think-​act.” 3 I am consciously leaving out emotion systems here because that is what most people do in the context of the sandwich metaphor. One might be tempted to think of emotions as a subclass of perceptions, but this ignores the action-​tendencies associated with most emotions. The truth of the matter is that emotion remains something of a red-​headed stepchild in these contexts. I will have more to say about this stepchild later in Chapter 5.

12  Abstract Concepts and the Embodied Mind • Context-​independence. Default knowledge is retrieved from long-​term memory in every context. Orthodox theories of concepts seek to explain these properties. For example, the disagreements outlined earlier between prototype, exemplar, and theory-​ based theories generally concern the character of the default knowledge deployed in working memory during the relevant cognitive tasks. A corollary of the second dogma is that concepts are often seen as static objects that we possess. They are thought of as bodies of information that are stored in long-​term memory and retrieved as a bolus for the purposes of carrying out various cognitive tasks. Experimental paradigms are explicitly designed to uncover the characteristics of this default knowledge.

2.2  Questioning These Dogmas Both dogmas have been questioned within the past few decades by a growing number of people working within the theoretical framework that has come to be known as embodied cognition. While this term encompasses a variety of theoretical positions (Binder & Desai, 2011; Meteyard et al., 2012), most of its advocates share the conviction that cognitive science needs to recognize the degree to which cognition is fundamentally tied to our bodily interactions with the world. The idea is that we cannot hope to understand the functioning of the brain without appreciating the central role it plays in guiding perception and action (Clark, 2008). This view has led to a robust and diverse research program in which investigators examine the possible ways in which thinking, remembering, and understanding language are shaped by the fact that we dynamically interact with our complex physical and social environment by means of perceptual and motor capacities. Most supporters of embodied cognition unequivocally reject the classical sandwich. This rejection is supported by mounting evidence implicating modality-​specific representations in conceptual tasks (for reviews, see Fischer & Zwaan, 2008; Kemmerer, 2010; Kiefer & Pulvermüller, 2012). As I argue in detail in the next chapter, there are good reasons to think that the neural mechanisms we use to perceive and act on objects and events in the world are often deployed when we think about them. This has led to the development of a grounded view of concepts in which affective and sensorimotor representations play a central role (see Chapter 3).

The Conceptual Brain  13 Although early versions of embodied cognition generally accepted invariantism, over the past two decades there has been a gradual shift toward contextualism. Many now hold that concepts are realized in a context-​and task-​dependent fashion. The impetus for this contextualist turn has two elements: a growing body of evidence suggesting that concepts are realized in different ways in different contexts and an emerging theoretical consensus that cognition should be seen as a fundamentally dynamical phenomenon (Lebois et al., 2015b; Yee & Thompson-​Schill, 2016). Significantly, there are reasons to question invariantism that are independent of embodiment and grounding. For one, the fact that distinct collections of evidence favor the existence of prototypes, exemplars, and theories suggests that concepts may vary in how they are organized and the properties that they encode. Indeed, this situation has led to pluralistic or hybrid accounts. At one extreme, Machery (2009, 2010) argues that we should abandon the very notion of a concept as a scientific category because the evidence favors distinct cognitive mechanisms associated with exemplars, prototypes, and theories. Others acknowledge the usefulness of concepts as theoretical posits and treat them as complex hybrids of one variety or another (e.g., Edwards, 2011; Prinz, 2002). Further support for context-​dependence emerges with ad hoc categories. These are categories that are structured to achieve certain goals such as things to sell at a garage sale (Barsalou, 1983) or tourist activities to perform in Beijing (Barsalou, 2010). A surprising thing about ad hoc categories is that, once formed, they tend to behave very much like established concepts. To give some examples: they exhibit typicality gradients (Barsalou, 1983), these gradients are structured around the relevant goals (Barsalou, 1991), and their internal structure is stable and robust (Barsalou, 1987). The fact that we behave with respect to spontaneously constructed categories much the way that we do with established ones throws the assumption of conceptual invariance into question. Given that most researchers have abandoned a filing cabinet conception of long-​term memory in favor of a constructive approach, it may well be time for those working on concepts to do the same. Barsalou (1983) initially proposed that concepts encompass both context-​ independent and context-​dependent properties. More recently, Casasanto and Lupyan (2015) have suggested that all concepts are ad hoc.4 They note

4 Barsalou himself has recently embraced contextualism (Barsalou, 2016a, 2017; Barsalou, Duriaux, & Scheepers, 2018). Given the central role attributed to situatedness in his perceptual symbol system theory (Barsalou, 1999), this may not reflect a dramatic theoretical shift.

14  Abstract Concepts and the Embodied Mind that this position fits with at least some of Barsalou’s statements. For example, they point out that Barsalou (1987, p. 121) acknowledges It may be extremely difficult, if not impossible, to identify where the knowledge for a particular category in long-​term memory begins and ends. To the extant that this is true, it is hard to imagine how there could be invariant representations for categories stored in long-​term memory.

Casasanto and Lupyan run with this idea and contend that the stability of our concepts is an illusion. They propose that (2015, p. 543), “all words are infinitely polysemous, all communication is ‘good enough,’ and no idea is ever the same idea twice.” In short, they offer a full-​throated contextualism. This perspective fits well with the growing evidence that conceptual representations may vary with stimulus (Sidhu et al., 2014; Taikh et al., 2015), task (Connell & Lynott, 2014a; Pexman et al., 2008; Sidhu, Heard, & Pexman, 2016), and context (Lebois, Wilson-​Mendenhall, & Barsalou, 2015a, 2015b; Yee, Ahmed, & Thompson-​Schill, 2012). We should acknowledge from the start that context sensitivity is likely to be a degree property and that a successful theory will likely need to posit a circumscribed sensitivity to context that strikes the right balance between automaticity and flexibility. If individuals are too influenced by the details of their current circumstances, then they are not going to be able to successfully leverage the knowledge that they have gained from past experience. If their conceptual system is too inflexible, then they are not going to be able to use that information in a timely and context-​appropriate manner. This need for a Goldilocks balance between automaticity and flexibility is not an unfamiliar one. All we need to do is consider well-​rehearsed yet flexible physical actions. After all, a top-​spin forehand tennis swing or a euro-​step basketball layup both need to be highly consistent and yet readily adaptable to specific situations. Barsalou, Casasanto, and Lupyan are not alone in emphasizing both embodiment/​grounding and context sensitivity. Wilson and Golonka (2013) propose that task-​dependence is a central component of embodied cognition. Connell and Lynott (2014b) contend that the dynamic influences that the body, the environment, the relevant goals, and the task have on our conceptual representations imply that “you can’t represent the same concept twice.” Lambon Ralph and colleagues (2017) posit a semantic control system that actively interacts with a partially grounded semantic representation

The Conceptual Brain  15 system. Winkielman, Coulson, and Niedenthal (2018) defend a view that they refer to as content-​dependent embodied simulation (CODES). While the conjunction of a commitment to embodiment and flexibility has become somewhat common, it is also the case that the word has not completely gotten out. Too many criticisms of embodied or grounded cognition fail to even address the possibility of flexible approaches that rely on multilevel representations (Barsalou, 2016a).5

2.3  Something We Do with the Mind As researchers focus on the nature of the neurological mechanisms responsible for our concepts, the question becomes not merely what information is encoded by our concepts, but also how that information is captured and deployed. In these pages, I am going to outline and defend an approach to concepts that emphasizes that they are something we do rather than something we possess. This book is thus not about concepts as they are often conceived—​that is, as static entities contained within or grasped by the mind. Thinking is a form of action and, as such, exhibits a dynamic flexibility adapted to the situation at hand. Casasanto and Lupyan (2015, p. 546) offer the following characterization of such an approach: We will use the term concept to mean a dynamic pattern of information that is made active in memory transiently, as needed, in response to internally generated or external cues. Activating a “concept of X” is thinking about X; this should be uncontroversial. More controversially perhaps, we suggest that concepts exist only when they are being used. Concepts are not something we have in the mind, they are something we do with the mind.

5 My personal experience provides some anecdotal evidence for this claim. I am often asked to review manuscripts that are critical of embodied or grounded cognition. With alarming frequency, these essays fail to consider what I perceive to be a broad class of mainstream views. Don’t take my word for it, though, and have a look at recent surveys of approaches to embodied cognition (such as Kiefer & Pulvermüller, 2012; Meteyard et al., 2012). What you will find in these surveys is an acknowledgment that many researchers favor some version of “weak embodiment” which emphasizes both the flexibility of concepts and the importance of intermediate, non–​ modality-​ specific representations (be they intramodal, heteromodal, or supramodal). Next, examine extant criticisms of embodied or grounded cognition. You will find that far too many of these rely on the assumption of invariantism and ignore the potential role of hierarchical representations.

16  Abstract Concepts and the Embodied Mind From this perspective, concepts are fleeting constructions that are adapted to the situation at hand. Conceptual representations are produced dynamically and are consistently variable (Connell & Lynott, 2014b; Kemmerer, 2015a). By these lights, having a concept is having the ability to produce contextually appropriate representations that can support the cognitive task at hand (Barsalou, 1983, 1987, 2012). This approach fits with the idea that concepts are skills rather than mental objects (Clark, 1996; Barsalou, 2012; Glock, 2010; Michel, 2020). None of this is to say that our concepts do not depend on information stored in long-​term memory. Indeed, the approach that I envision (which admittedly may differ from the approach that other supporters of the idea that concepts are something we do with the mind might envision) posits distributed networks of representations associated with categories. These networks influence how concepts are realized in specific situations (see Chapter 5 for details). Importantly, they change with experience. In other words, no permanent and fixed representation of the category exists even in long-​term memory.6 To see how this approach revamps our understanding of concepts, let’s return to the sort of prosaic examples I used earlier to introduce the theoretical notion of a concept. As I mentioned earlier, concepts are seen as repositories for acquired knowledge about categories. Embodied approaches propose that this stored information is not simply transduced from action, emotion, and perception systems. Instead, it retains an intimate connection to these systems. Lynott and Connell (2010, p. 2) outline how we might understand the nature of our concepts from a grounded perspective. A concept is an aggregated memory of aspects of experience that have repeatedly received attention in the past, and incorporates perceptual, motor, affective, introspective, social, linguistic and other information. For instance a concept of dog could potentially include a host of perceptual-​ motor information, possibly including visual information of the color and shape of a dog, tactile information regarding the feel of a dog’s coat, olfactory information of the smell of a dog, auditory information of a dog’s bark, motor information about patting a dog, social information about the status of dogs in human households, along with positive or negative affective 6 Perhaps this should not be seen as a radical assumption. The idea that long-​term memory is reconstructive has been around for quite some time (Bartlett, 1932; Loftus, 1979).

The Conceptual Brain  17 valence depending one’s experience with dogs in the past. Any time the word “dog” is encountered, a subset of these aspects will be retrieved to suit the task at hand.

This passage places great emphasis on the sources of the information. Strictly speaking, we could envision an account in which this modality-​specific information is encoded solely by amodal representations. An embodied or grounded approach, however, proposes that much of this information remains captured in the relevant sensorimotor systems. Access to this information is gained by selectively reactivating the representations contained within these distinct modalities. The dynamic pattern of information elicited in the performance of a certain task is thus distributed across various modalities. This selective reactivation of neural activity is often characterized as a kind of neural simulation. The idea is that conceptual thinking involves simulating our affective and sensorimotor experiences with category members. For example, categorizing someone’s pet as a dog might involve comparing our ongoing experiences with a suite of simulations that encompass auditory, olfactory, touch, and visual features. The notion of simulation provides an intuitive means of grasping how neural reactivation might enable conceptual processing, but it can also be misleading. While we normally think of our experiences—​and consequently any simulations of them—​in terms of their conscious properties, a great deal of the processing associated with the posited reactivations is likely to be unconscious (Kemmerer, 2015b). Moreover, these reactivations are purported to be selective to a degree that our experiences are not (Barsalou, 1999). For instance, it may be possible to selectively reengage visual shape representations without reengaging those associated with color processing. Clearly, this is not something that can be achieved under normal circumstances in our everyday conscious perceptual experience. An important feature of this approach to concepts is that it makes specific predictions concerning how they are neurologically realized. Barsalou offers another example that highlights the proposed connection to modality-​ specific representations (2015, p. 84; citation in the original). As people experience hammers, brain areas that process their multimodal aspects become active and associated together (Martin, 2007). Specifically, distributed associative patterns become established across fusiform gyrus (shape), premotor cortex (action), inferior parietal cortex (spatial

18  Abstract Concepts and the Embodied Mind trajectory), and posterior temporal gyrus (visual motion). Following many learning episodes with hammers, an increasingly entrenched associative network reflects the aggregate effects of neural processing in these areas.

According to Barsalou, we can treat these distributed associative patterns as representations because they carry information that can be used in the future. This information can be used to reactivate a subset of this network in working memory in a situation-​specific manner. In other words, depending on the occasion, different portions of the distributed network associated with hammers will be deployed (also Binder et al., 2016; Kemmerer, 2015a).

2.4  Elasticity and 4E Cognition This book explores the degree to which conceptualization—​the something that we do with the mind—​is grounded in experience and what that might mean with respect to how it is realized in the brain. To put it simply, this book is about how concepts are both embodied and embrained. One of the challenges facing this project is that embodiment means different things to different people. I suggest that embodied cognition is best thought of as an active and diverse research program that emphasizes the central role that the body and its interaction with the environment play in cognition. As a research program, it is neither theoretically homogeneous nor free from disagreement. Indeed, the proper way to characterize the very notion of embodiment remains a contentious topic among advocates of embodied cognition (Anderson, 2003; Meteyard et al., 2012; Shapiro, 2019; Wilson, 2002). Questions also circulate around how embodiment relates to other types of what has come to be called 4E cognition (Newman, De Bruin, & Gallagher, 2018). This acronym refers to the combination of Embodied, Embedded, Enactive, and Extended cognition. While I do not have the space to delve deeply into each of these approaches here, a sketch of each may be sufficient for my current purposes. Embedded cognition holds that our thoughts are often scaffolded by the external resources that are present in our cognitive and developmental niches (Rupert, 2010; Sterelny, 2010; Stotz, 2010). Enactive cognition views cognition as something that arises from the interaction of living autonomous agents and their environments (Di Paolo, 2009; Thompson, 2007; Varela, Thompson, & Rosch, 1991). Extended cognition suggests that external aspects of our environment can be constituents

The Conceptual Brain  19 of cognitive processes or, if you prefer, integrated cognitive systems (Clark, 2008; Menary, 2007; Sutton, 2010). Given the thematic similarity of embodied, embedded, enactive, and extended cognition, it has become somewhat fashionable to speak of 4E cognition as a research program of its own.7 This designation has its use, but it is important to recognize that tensions exist between its subtypes. For instance, there is a clear tension between extended cognition, which requires dissolving the distinction between agent and environment, and enactive cognition, which relies on that very distinction (Menary, 2010; Thompson & Stapleton, 2009). Embedded cognition is often offered as a means of preserving cognitive theories from the perceived excesses of the other E approaches (Rupert, 2010). Although the position that I ultimately defend contains elements of each of these approaches, this book is not a manifesto for either embodied or 4E cognition. Instead, its aim is to uncover the neurocognitive mechanisms responsible for our conceptual system. Importantly, this focus provides a certain theoretical freedom. Given that I am concerned with developing a theory of concepts rather than a general defense of the 4E research program, I am free to pick and choose from a menu of different 4E ideas. Furthermore, nothing I write in this book excludes 4E theories that go beyond, or abstract away from, the neural underpinnings that are my concern. Broadly speaking, theories of 4E cognition can be divided into one of two major categories: those that emphasize the influence of the body on the mind and those that emphasize the importance of body–​world couplings. My account integrates ideas that focus on body–​mind connections with those that focus on body–​world connections. At a minimum, I propose (i) that flexible neural reuse is a central feature of the brain mechanisms responsible for cognition and (ii) that the manipulation of external resources often serves to scaffold our cognitive endeavors. Both proposals will make important contributions to the elastic view of concepts that I defend in this book. Because my position focuses on the deep entwinement that the conceptual system enjoys with action, emotion, and perception systems, it falls within the broad research program of 4E cognition. However, as is so often the case, the devil is in the details. And several of the details of my account 7 One could argue that the tent should be made even larger to include affective, ecological, and social cognition. This larger research program would be defined by its opposition to the internalism associated with orthodox forms of computationalism. Of course, it wouldn’t fit as neatly with the convenient E designation.

20  Abstract Concepts and the Embodied Mind require rejecting some of the more radical ideas of 4E cognition—​or at least some of its prominent forms. Many of these details concern how concepts are embrained. Perhaps the most significant of these is the fact that I remain committed to the theoretical importance of the notion of neural representations. As the direct quotes provided earlier clearly demonstrate, I am not alone in taking this position. Many researchers working on semantic memory share my commitment. Nevertheless, it is also true that a significant number of researchers believe that 4E cognition, properly construed, should involve a rejection of the very idea of neural representation (e.g., Chemero, 2011; Hutto & Myin, 2013; Ramsey, 2007). Although a full discussion of the importance of representations to cognitive neuroscience would require its own book, I can offer a brief—​largely pragmatic—​defense of representations. Several considerations suggest that the outright rejection of representations throws the baby out with the bathwater. First, there is the simple fact that explanations of cognitive phenomena in terms of computations over neural representations have proved successful in terms of their explanatory power, reach, and robustness (Bechtel, 2008; Boone & Piccinini, 2016). Rejecting representations thus requires rejecting the products of a very productive research program. Of course, it may be that the time is right for a catastrophic paradigm shift within the psychological and brain sciences (Kuhn, 1970). I would suggest, though, that such scientific revolutions are fewer and farther between than many would like to acknowledge. More to the point, given the theoretical upheaval that such revolutions involve, they require significant justification, and I maintain that it is not clear that this threshold has been met in cognitive science. Second, the notion of computation attacked by those who oppose neural representations is often unfairly circumscribed and impoverished. For example, it is not uncommon to assume that this notion involves a commitment to discrete, language-​like symbols. Maley (2017) has recently argued that neural representations are much more likely to be analog in nature. The notion of neural representation thus needs to be extended to include the possibility of analog computation (Bechtel, 2008; Markman & Dietrich, 2000). To be fair, critics of a restrictive notion of representation are not simply attacking a straw version of computationalism but rather have their sights set on a traditional view that encompasses a cluster of theses (Clark, 2003). Even so, they often reject representations outright. Ultimately, this rejection is unsupported because the notion of computation is more flexible than its critics typically recognize.

The Conceptual Brain  21 Third, arguments for abandoning representational explanations often turn on questionable projections about the outcomes of future research. They often assert that nonrepresentational explanations for cognitive phenomena are waiting in the wings and that representational explanations are going to ultimately prove to be unnecessary (Chemero, 2011; Hutto & Myin, 2013; Ramsey, 2007). A common weakness of these efforts is that the proposed exemplars of how cognitive science should proceed without representations rarely include rich explanations of the neural mechanisms involved. Of course, this could change. Chemero (2011), for example, proposes that dynamical systems theory is up to the job of providing both brain-​only and brain-​body-​environment explanations. While I embrace explanatory pluralism and think that nonrepresentational explanations should be explored, I also contend that it is far from clear that they will supplant representational explanations.8 Finally, it seems plausible that some aspects of cognition might be “representation hungry” (Clark & Toribo-​Mateas, 1994). Barsalou (2015, p. 81), for instance, argues that representations are needed to explain basic processes that “mediate between stimuli and responses” such as memory, decision-​ making, language, and problem-​solving. At a minimum, conceptual processing seems to be a good candidate for representation-​based explanation. Having offered this defense of representations, I need to offer a couple of caveats. To begin with, the notion of a representation that I am appealing to is the one that is used by cognitive neuroscientists and does not always line up well with the notion of representation employed by philosophers. In some ways it is looser, and in other ways it is more restrictive. It is looser in the sense that it characterizes representations broadly in terms of a notion of carrying information. This leaves room for us to construe diverse phenomena as representational. It also frees representationalism from strong commitments with respect to the nature of mental content. For instance, it may inoculate it from some of the philosophically inspired criticisms (e.g., Hutto & Myin, 2013). It is more restrictive in the sense that it remains tied to explanations of neural mechanisms. What we are interested in is explaining the role that neural circuits, networks, and systems play in explaining intelligent behavior. 8 Several people have argued in support of representational accounts of dynamical systems (Bechtel, 2008; Markman & Dietrich, 2000). Skeptics of representation often argue that this distorts our view of the relevant phenomena by underplaying external factors (e.g., Chemero, 2011). However, supporters of representation can similarly argue that nonrepresentational explanations distort our view of internal factors and cause us to miss important generalizations. What will ultimately settle these issues is explanatory success.

22  Abstract Concepts and the Embodied Mind Forms of functionalism that ignore such neural details may be incompatible with this notion of representation. Another important feature of this notion of representation is that it does not involve a commitment to methodological solipsism or internalism. This should come as no surprise. After all, any explanation of context-​sensitivity requires some explanation of the dynamic role played by context. Less obviously, the account of representations on offer leaves room for the sort of complex environmental factors favored by many supporters of 4E cognition. In other words, it does not presume that explanations of cognitive behavior only involve the inner states of the organism. Some defenders of computationalism (Adams & Aizawa, 2008; Aizawa, 2017) accuse supporters of 4E cognition of focusing on intelligent behavior rather than cognition properly construed. I reject this formulation of the issues and embrace intelligent behavior as the ultimate explanandum. Researchers in the psychological and brain sciences seek to understand how evolved creatures make their way in the world. Representations prove their worth with respect to the degree that they lead to explanations that exhibit both ecological and external validity. While it may be possible, and even expedient, to draw on the competence/​performance distinction when examining neural mechanisms (perhaps even at several levels of analysis; Marr, 1982), we cannot fully grasp how our concepts are embrained without an understanding of how the dynamic interactions between elements in our environment, our body, and our nervous system generate the sorts of intelligent behavior exhibited by us and other organisms (particularly the animal models employed by the many branches of neuroscience). Part of the motivation for the criticism of the focus on intelligent behavior is the worry that 4E approaches amount to a kind of gussied up behaviorism. While I am not convinced that this criticism hits its mark, it is worth pointing out that it certainly does not apply to the approach advocated in these pages because it is committed to the importance of internal states.

2.5  Staying Grounded From here on out, I am going to draw a terminological distinction between embodied and grounded cognition. I am going to reserve the term embodied cognition for the broad, somewhat loosely defined collection of theoretical approaches to cognition that emphasize various ways in which the

The Conceptual Brain  23 body contributes to cognition. I am going to use the term grounded cognition to identify a more circumscribed class of theories that focuses on the roles played regularities in our physical experiences and the neural mechanisms associated with them. Drawing this distinction in this way may be somewhat idiosyncratic, but it is largely compatible with current usage, and it will help me identify the relevant empirical and theoretical issues. Abstract concepts are often viewed in relation to a polemical debate concerning the promise of grounded cognition. In this context, they are treated as a kind of litmus test. Critics point to them as indicators of the inadequacy and implausibility of grounded cognition. Supporters seek to undermine their status as a potential counterexample by uncovering evidence that they are grounded to some degree. I adopt a different stance in this book. Rather than treat abstract concepts as simply a test case—​one that will result in a simple up or down judgment with respect to conceptual grounding—​I use them as a springboard for thinking about grounded cognition. First, in keeping with current conceptions of grounding, I contend that our concepts often depend on the multimodal reengagement of affective and sensorimotor areas. In other words, hard-​won experiential knowledge is leveraged through the dynamic reuse of experiential systems. Abstract concepts may rely more heavily on certain sorts of experiential representations than concrete concepts do—​in particular, they may rely more on affective and introspective representations. Second, I propose that our semantic memory exhibits hierarchical neuroanatomical organization, both internal to and across specific modalities (Barsalou, 2016b; Binder, 2016). Our access to higher-​level representations underwrites our capacity to generalize away from experiential particulars. Our concepts involve mechanisms associated with the ongoing evaluation of incoming sensory input relative to the predictions generated by the motor system (Pickering & Clark, 2014). While the possible relevance of heteromodal and supramodal areas has been acknowledged by some supporters of grounding, the account developed in these pages identifies a clear functional role for the representations contained in these areas. Third, I maintain that the language system not only provides a means of communicating our thoughts, but also makes important contributions to the internal representation of our concepts (Borghi & Binkofski, 2014; Dove, 2011). While these contributions are not exclusive to abstract concepts, they are particularly important for them. To put it simply, language helps us go beyond experience because it is a form of symbolic cognition. Thus, my account

24  Abstract Concepts and the Embodied Mind can be considered a kind of grounded/​ symbolic hybrid (Pulvermüller & Garagnani, 2014; Reilly et al., 2014; Shallice & Cooper, 2013; Watson & Chatterjee, 2011; Zwaan, 2014). Fourth, I propose that conceptual representations may vary with stimulus and context. The degree, form, and character of conceptual representations is likely to change over the course of development (Kontra, Goldin-​Meadow, & Beilock, 2012; Pexman, 2019). This flexibility tends to be more pronounced with abstract concepts (Yee, 2019). To conclude, our facility with abstract concepts demonstrates that our conceptual system is multimodal, hierarchical, scaffolded, and flexible. In short, abstract concepts demonstrate the elasticity of our conceptual system. The rest of the book is a defense of this thesis.

3 Body in Mind Concepts are collections of stored knowledge that are used in cognitive activities such as categorization, decision-​making, and language processing. They aggregate information about category exemplars that has proved to be relevant or useful across different situations. Consider again our parade case: the concept pug. Our experiences with pugs often have perceptual, motor, affective, social, and linguistic dimensions. We see, touch, smell, love, fear, and talk to (and about) pugs. The traditional and grounded views diverge with respect to how the information drawn from these experiences is represented. According to the traditional view, this information is encoded in symbols that are part of a self-​contained conceptual system (e.g., Anderson, 1983). These symbols are often described as amodal because they are not indigenous to any sensorimotor modality. According to the grounded view, this information is encoded in representations indigenous to the systems used to experience and interact with pugs. This might include visual representations of size, shape, color, and movement; tactile representations of touching fur; auditory representations of barking and panting; and affective representations associated with interacting with pugs (Barsalou, 1999, 2008; Lynott & Connell, 2010; Prinz, 2002). This chapter examines some of the reasons that researchers think that we should shift from a traditional view to an embodied or grounded one. I begin with a survey of the empirical evidence for sensorimotor grounding, outlining some of the representative behavioral, brain imaging, and neuropsychological studies that seem to indicate that our concepts are anchored in sensorimotor systems. For the sake of clarity and ease of exposition, I focus on perceptual and action systems separately (the role of affective systems will be taken up in later chapters). While I do not have the space to be comprehensive (extensive reviews can be found in Borghesani & Piazza, 2017; Conca & Tettamanti, 2018; Hauk, 2016; Kemmerer, 2010; 2015c; Pulvermüller, 2017), I will consider some of the evidence in detail to shine a light on the reasoning behind the grounded approach. This will enable me to highlight features of this evidence that are occasionally overlooked or misrepresented. Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0003

26  Abstract Concepts and the Embodied Mind

3.1  Perception Systems Several behavioral studies implicate perceptual systems in semantic tasks. Let’s begin with an oft-​cited example that has contributed to the development of grounded cognition. This example builds on the established finding that detecting a signal in a perceptual task is often slower when it is preceded by a task carried out in a different perceptual modality than when it is preceded by a task in the same perceptual modality (Spence & Driver, 1998; Spence, Nicholls, & Driver 2000). Pecher, Zeelenberg, and Barsalou (2003) found a similar modality-​switching cost associated with a conceptual task: participants verified verbally expressed facts involving one modality (such as the fact that leaves rustle) more rapidly after verifying a fact involving the same modality (such as the fact that blenders make noise) than after verifying a fact involving a different modality (such as the fact that cranberries are tart). This evidence fits neatly with the proposal that conceptual processing involves the deployment of perceptual representations. A follow-​up study found a similar modality-​switching cost when participants alternated between a perceptual detection task and a property verification task (van Dantzig et al., 2008). Whereas the original study found a modality-​switching cost that was analogous to one found with perceptual processes (leading the authors to infer that the same mechanisms are involved), this study found a clear interaction between perceptual and conceptual processes. It is worth pausing to examine why these results are surprising and potentially informative. The language comprehension task at hand is clearly a conceptual one: it involves the sort of general information about the relevant categories that should be encoded within semantic memory. Given the assumptions of the classical view of conceptual processing, there is little reason to expect the sort of interference observed in these experiments. If concepts are represented and processed within a semantic system that is separate from and independent of perceptual systems, then there should be no conceptually elicited modality-​switching cost. On a grounded view, however, this sort of interference is to be expected. To be more precise, because this view holds that conceptual processing involves the same neural systems that are used to perceive, act upon, and emotionally respond to category exemplars, previous activity in these systems is likely to either enhance or interfere with related conceptual tasks. Inspired by the success of this experimental paradigm, researchers often look for behavioral evidence of an interaction between conceptual and

Body in Mind  27 perceptual processing. Both facilitation and interference effects have been found. For example, Stanfield and Zwaan (2001) asked participants to affirm that pictures depicted the actions described in previously presented sentences. The actions had either a vertical or horizontal orientation (such as driving a nail into a floor or into a wall). Participants responded more quickly to the pictures that had the same orientation as the action described. Other evidence suggests that hearing motion-​related verbs interferes with visual motion processing (Meteyard, Bahrami, & Vigliocco, 2007) and visual motion processing interferes with the processing of motion-​related verbs (Meteyard et al., 2008). Whether the interaction between perceptual and semantic processing leads to interference or facilitation seems to be determined by task and context. Kaschak and colleagues (2005), for example, found that watching upward or downward motion slowed the semantic processing of sentences describing motion in the same direction, but Kaschak and colleagues (2006) found that participants were faster in processing spoken sentences describing motion when they concurrently heard an auditory stimulus conveying motion in a matching direction. The presence of both interference and facilitation effects creates a challenge for the hypothesis that conceptual processes regularly depend on perceptual mechanisms. Responding to this challenge, Connell and Lynott (2012) propose that the tension created by this discrepancy can be resolved if the attentional demands that each task places on modality-​specific processing are considered. In a study exploring the possible task-​specific role of attention in grounding, Connell and Lynott (2014a) compared the performance of participants on two standard tasks: lexical decision (in which participants indicate whether a presented stimulus is a word or not) and reading aloud. Lexical decision was facilitated by how strongly the referent of a word is visually experienced, and reading aloud was facilitated by how strongly the referent is both visually and auditorily experienced. Importantly, grounded cognition suggests that attentional mechanisms associated with perceptual processing should influence conceptual processing. This provides some reason to think that the explanation of the discrepancy on offer is not simply an ad hoc attempt to deal with recalcitrant data. Neuroimaging provides further evidence of perceptual grounding. For example, a group of studies provides some reason to think that a particular area of the temporal cortex, the lateral posterior fusiform gyrus (pFG), is involved in both the perceptual processing of the shapes of objects and the

28  Abstract Concepts and the Embodied Mind semantic processing of object words. The lateral pFG showed increased, category-​specific activation both when participants viewed pictures of animals and when participants read the names of and answered questions about animals (Chao, Haxby, & Martin, 1999). A similar pattern was observed when participants in another study judged whether written words denoted animals or manmade artifacts (Devlin, Rushworth, & Matthews, 2005). In a study examining the degree to which task-​related effects might explain the variability found in the neuroimaging results associated with these broad categories, participants answered yes-​no questions about visual or nonvisual characteristics of living and nonliving things (Thompson-​Schill et al., 1999). Questions about the visual characteristics of nonliving things yielded increased activation in the left pFG when compared to questions about nonvisual characteristics. No such differences were observed with living things. This pattern is consistent with the hypothesis that visual features are retrieved when individuals think about living things whether the task involves visual characteristics or not. Simmons and colleagues (2007) provide evidence of a shared neural substrate associated with the perception of color and a property verification task involving color (e.g., taxi–​yellow). Because this substrate involves an anterior region of the left fusiform gyrus rather than a posterior region of it (which has been associated with low-​level color perception), there is some dispute concerning the degree to which it supports sensorimotor grounding. Critics of grounding have proposed that what the researchers found was shared activation in an amodal area associated with conceptual processing that is near to but independent of color perception areas (Chatterjee, 2010; Mahon, 2015; Mahon & Caramazza, 2008). Simmons and colleagues acknowledge this possibility but argue that other evidence shows that the more anterior region plays a role in higher-​level color perception. This dispute exemplifies what has become a core issue in the debate surrounding perceptual grounding: the question of whether the observed activity is epiphenomenal or causally relevant to the conceptual processing elicited by the task. Most participants in this debate grant that researchers have provided compelling—​even surprising—​evidence that some components of perceptual systems become active during or perhaps in response to conceptual processing. What remains in dispute is the claim that this activity is part of the ongoing conceptual processing. Those who are skeptical of perceptual grounding suggest that the activity identified in these neuroimaging studies might instead be by-​products or side effects of conceptual activity (Goldinger et al., 2016; Mahon & Caramazza, 2008; Weiskopf, 2007).

Body in Mind  29 Researchers have responded to this dispute by exploring the nature of the relationship between activity in perceptual areas and conceptual processing. Earlier we reviewed some of the neuroimaging studies that implicate the lateral pFG in processing the visual features of object concepts. A carefully constructed event-​related functional magnetic resonance imaging (fMRI) experiment provides evidence that this area is involved in the processing of object concepts (Wheatley et al., 2005). Participants silently read rapidly presented word pairs while they were in the scanner. These pairs consisted of words that were semantically related (fork-​cup), semantically unrelated (ankle-​carrot), or identical (crow-​crow). The semantically related pairs elicited less activity in the lateral pFG than the unrelated pairs but more activity than the identical pairs. Outside of the scanner participants carried out a behavioral version of the experiment in which they read aloud the second word in each pair. Their response times exhibited the sort of semantic priming effects that one would expect under these conditions: slowest for the unrelated pairs, faster for the related ones, and fastest for the identical ones. It seems likely that the observed reduction in the activity in the lateral pFG is the result of some form of neural adaptation caused by the engagement of the same population of neurons by the first element of the pair. If so, then this reduction would be a neurophysiological correlate of semantic priming. All together this evidence suggests that the lateral pFG is directly involved in semantic processing. A series of studies by Kiefer and colleagues (2008) examines the possible contribution of auditory areas to the semantic processing of concepts with significant associations with sound. Two of these studies involved fMRI and one involved recordings of event-​related potentials (ERP). In the first fMRI study, the participants carried out a lexical decision task on a set of visually presented words and pseudo-​words. The words used in these experiments varied to the degree that their semantic content was associated with acoustic features. Some of the words were highly associated with acoustic features (e.g., bell) while some were not (e.g., table). In the second fMRI experiment, participants listened to actual sounds produced by animals and tools. Kiefer and colleagues found that the conceptual and perceptual processing of acoustic features elicited overlapping activation within the auditory association cortex—​particularly in posterior regions of the left temporal cortex that handle high-​level auditory processing. This activation was not found with words primarily associated with visual or motor features. ERPs recorded with the same conceptual task found that the relevant increase in activity

30  Abstract Concepts and the Embodied Mind begins around 150 milliseconds after word onset. Taken together, the fMRI and ERP data suggest that the response is both rapid and selective, neither of which fits well with the proposal that it is the result of post-​conceptual imagery. A recent magnetocephalography (MEG) study using a single-​word reading task similarly found evidence of the automatic engagement (~200 ms) of high-​level auditory areas by words with strong sound associations (Borghesani et al., 2018).1 Several neuroimaging studies have found links between conceptual processing and activation in other sensory areas. For instance, textual metaphors have been found to activate sensorimotor cortex (Lacey, Silla, & Sathian, 2012). Reading odor-​related words (e.g., cinnamon, garlic, and jasmine) elicits increased activation in the primary olfactory cortex relative to neutral control words (Gonzalez et al., 2006). In addition, reading taste-​related words has been found to activate primary gustatory cortices more than words that are not taste-​related (Barros-​Loscertales et al., 2012). Other studies, however, have found that such words activate only secondary gustatory regions rather than primary ones (Goldberg, Perfetti, & Schneider, 2006). The purported role of touch, taste, and smell in semantic memory has been questioned recently (Speed & Majid, 2020). Speed and Majid acknowledge that there is compelling evidence for conceptual grounding in general (particularly when it comes to the influence of vision and audition) but question whether these specific senses contribute in a meaningful way to conceptual grounding. They begin by conceding that some behavioral evidence suggests that sensory processing in these modalities can affect conceptual processing (Connell, Lynott, & Dreyer, 2012) and that conceptual processing can affect sensory processing in them (Brunyé et al., 2012; Olofsson et al., 2012). They argue, though, that the evidence for grounded simulations involving these modalities is weak for several reasons. First, the relevant studies generally fail to exclude the possible influence of other, more established sources of grounding. Second, the observed activations often involve secondary rather primary sensory cortex. Third, many of the studies seem more likely to involve the strategic use of mental imagery rather than the use of mental simulation for the purposes of language comprehension. 1 I should note that this MEG study also found evidence of conceptual processing that was not located within a particular sensorimotor modality. The authors interpret this in the context of a hybrid model of concepts. For the moment, though, I am focusing on the role of perceptual representations. I discuss the possible relevance of hybrid models in later chapters.

Body in Mind  31 Neuropsychological cases studies provide further evidence of the causal relevance of the auditory system to conceptual processes. Bonner and Grossman (2012) hypothesized that individuals with damage to the auditory association cortex should have difficulty processing concepts with strong auditory associations. They had patients with the logopenic variant of primary progressive aphasia (lvPPA) perform a word recognition task on a set of words that were strongly associated with either sights, sounds, or manipulations. The participants exhibited a selective deficit for those words strongly associated with sounds. A structural MRI of the brains of the lvPPA participants revealed gray matter atrophy in a sector of the left auditory association cortex that was shown to respond in the processing of the same sound words by healthy participants in a separate fMRI study. Suggestively, the extent of the lvPPA patients’ difficulties with the sound words correlated with the degree of atrophy found in the auditory association areas. A recent single case study of a patient with a focal lesion in the left superior and middle temporal gyrus found that he was consistently impaired in the conceptual processing of concepts related to everyday objects with strong sound associations and the perceptual recognition of the corresponding sounds (Trumpp et al., 2013). Taken together, the behavioral, neuroimaging, and neuropsychological evidence provides a compelling case for the involvement of perceptual systems in our concepts. Questions about scope and variability remain, but perceptual representations appear to play a causal role in our concepts.

3.2  Action Systems Let’s now turn our attention to the contribution of action systems to our concepts. One of the more striking examples of research implicating the motor system in semantic processing involves effector-​specific activations in motor areas during language tasks involving words that refer to actions that involve specific body parts and movements (Kemmerer, 2010; Willems & Casasanto, 2011). Scorolli and Borghi (2007) found a facilitation effect when they asked their participants to judge the sensibility of simple sentences containing a verb that referred to an action typically performed with the mouth, hands, or the feet. Participants responded either by pressing a pedal or speaking into a microphone. Response times with the microphone were fastest with “mouth-​sentences” and response times with the pedal were fastest with “foot-​sentences” (see also Scorolli, Borghi, & Glenberg, 2009).

32  Abstract Concepts and the Embodied Mind This somatopic specificity fits with the findings of several brain imaging studies. Hauk, Johnsrude, and Pulvermüller (2004) used fMRI to map motor areas in individual research participants by having them move their feet, fingers, or tongues. They then had participants read action verbs associated with these body parts (e.g., kick, pick, or lick). Reading the verbs associated with the hands and the feet led to a similar pattern of activation in the premotor cortex to that elicited in the movement tasks. Similar somatopic specificity has been found in other imaging experiments involving different language comprehension tasks (Tettamanti et al., 2005; Rueschmeyer et al., 2010). The specificity of the modulated activity can be quite fine-​grained. For instance, right-​and left-​handers exhibit increased activation in the premotor areas that were contralateral to their dominant hands (Willems, Hagoort, & Casasanto, 2010). The degree to which expert hockey players comprehend hockey-​action sentences better than controls correlates positively with activity in the left dorsal premotor cortex (Beilock et al., 2008). Buccino and colleagues (2005) found that listening to action-​related sentences modulated activity in the motor system. Motor evoked potentials (MEPs) recorded from hand and foot muscles were selectively modulated by hand-​related and foot-​related action sentences, respectively. Pulvermüller and colleagues (2005) carried out a transcranial magnetic stimulation (TMS) study in which they weakly stimulated different parts of the motor system shortly (150 ms) before participants performed a lexical decision task on arm-​and leg-​related action words. Response times were faster with the arm-​ related words when there was stimulation of left hemisphere areas associated with arm movement and with the leg-​related words when there was stimulation of left hemisphere areas associated with leg movement, but response times were not modulated by the stimulation of right hemisphere motor areas or in a control condition with sham stimulation. Studies employing repetitive TMS have uncovered interference effects. In one, repetitive TMS (rTMS) applied to the left primary motor cortex was found to delay the processing of action nouns and verbs involving the hands but not state words of either grammatical variety (Gerfo et al., 2008). In another, rTMS was applied separately over the left and right primary motor cortex prior to a semantic comprehension task involving hand-​related action verbs and abstract verbs describing intellectual and symbolic activities that were matched for number of letters, number of syllables, and frequency (Repetto et al., 2013). The main finding of this experiment was that the stimulation of the left hemisphere primary motor

Body in Mind  33 cortex selectively delayed the reaction times to the hand action verbs. No such delay was found with the abstract verbs. Given that all the participants were right-​handed, the absence of a similar effect with the right hemisphere rTMS seems to be in keeping with the effector-​specificity discussed earlier. In a third study, rTMS was applied to the primary motor cortex 200 ms before word onset in lexical decision and concreteness judgment tasks involving action-​related and abstract words (Vukovic et al., 2017). A selective delay in reaction times for the action words was found in the concreteness judgment task but not the lexical decision task, with the stimulation of the left primary motor cortex. One explanation for the task-​related difference could be the fact that the concreteness judgment is likely to require a greater depth of semantic processing than the lexical decision task. Unlike the previous study, the left lateralized stimulation produced shorter reaction times with the abstract words. The TMS evidence is revealing for a couple of reasons. A primary objection to the body of evidence implicating sensorimotor systems in cognition is that it could be the result of epiphenomenal byproducts of conceptual processes. This objection builds on the observation that correlation does not establish causation. The TMS research, however, provides direct evidence of functional relevance. Furthermore, it provides a tantalizing introduction to the possibility of an important neurophysiological abstract–​concrete concept distinction. Neuropsychological studies provide further evidence of the functional relevance of the motor system to action concepts. A recent study examined the relationship between manual abilities and conceptual abilities of 41 chronic stroke patients and found that the degree to which the participants were impaired with processing action words in implicit and explicit semantic processing tasks was predicted by their degree of impairment measured in a reaching task (Desai et al., 2015). Another study compared the processing of words and pictures representing actions and objects by 21 aphasic and 20 control participants (Arévalo et al., 2007). The aphasic participants were less accurate at processing the test items involving concepts associated with manipulability (such as the noun accordion or the verb to squeeze) than concepts not associated with it (such as the noun airplane or the verb to snow). Among the patients who exhibited the manipulability effect, 60% were found to have a lesion in a hand-​related motor region. In keeping with the hypothesis that motor areas contribute to the processing of action concepts, several group studies of stroke patients have found that impairments of action concepts

34  Abstract Concepts and the Embodied Mind are often associated with damage to left motor cortices (Akinina et al., 2019; Gleichgerrcht et al., 2016; Kemmerer et al., 2012). Another growing body of research examines the effects of movement disorders associated with neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) or Parkinson’s disease (PD) and finds that action concepts can be selectively affected. Let’s begin with ALS. ALS (also known as Lou Gehrig’s disease or motor neuron disease) is a movement disorder associated with the degeneration of motor neurons in the brain and spinal cord. It often degrades the motor cortex and nearby regions (Chiò et al., 2014). Several studies of ALS patients have found a selective deficit in action-​concept processing (Bak et al., 2001; Bak & Hodges, 2004; Cousins, Ash, & Grossman, 2018; York et al., 2014). For example, one study found that patients with ALS were selectively impaired with action verbs when their performance was compared to age-​matched healthy controls (York et al., 2014). The degree to which action verb performance was impaired correlated with the extent of atrophy in motor-​associated cortex. Another study separated ALS patients into two groups: those with high motor impairment and those with low motor impairment (Cousins et al., 2018). This study examined their performance on a picture-​elicited verb production task that involved verbs where the body is the agent of the action (e.g., The boy is stealing the cookies) and verbs where the body is the subject of the action (e.g., The boy is falling). When the two groups were compared, those patients with high motor impairment showed an impairment for body agent verbs but not body theme verbs. Regression analyses revealed a correlation between the extent of this impairment with gray matter atrophy in premotor areas. Because an impairment for action concepts has been found in both ALS and PD, they are often linked in reviews of the evidence favoring the importance of motor features to action concepts. Nevertheless, there are important neurophysiological and cognitive differences between these diseases. PD is a progressive movement disorder that begins with substantial cell death within the dopaminergic neurons of the basal ganglia (Jankovic, 2008). Voluntary motor control and ultimately high-​order cognition are progressively impaired. In its later stage, PD can lead to widespread cortical atrophy and thinning, including frontal, parietal, and temporal regions, but in patients without dementia there appears to be limited cortical atrophy (González-​ Redondo et al., 2014; Mak et al., 2015). While PD may not impact the motor cortex directly, it generally leads to a reduction in projections from the basal

Body in Mind  35 ganglia to frontal motor regions and a hypoactivation of motor and premotor cortex (Jankovic, 2008). Several studies using a variety of tasks and experimental paradigms have found a selective impairment for action concepts in patients with early-​stage PD (Bocanegra et al., 2015, 2017; Fernandino et al., 2013a, 2013b; García et al., 2018; Herrera & Cuetos, 2012; Herrera, Rodriguez-​Fereiro, & Cueto, 2012; Ibáñez et al., 2013; Melloni et al., 2015; Roberts et al., 2017; Salmazo-​ Silva et al., 2017). To get an idea of the robustness of this research, consider some examples. Fernandino and colleagues (2013a) investigated the performance of PD patients and age-​matched controls with action and abstract verbs. They used tasks that imposed both implicit semantic processing (lexical decision and priming) and explicit semantic processing (sematic similarity judgment). The PD patients were selectively impaired for the action verbs on both types of tasks. A more recent study finds evidence of the sort of effector-​specificity previously found in the behavioral and neuroimaging research (Roberts et al., 2017). This study compared the action verb processing of PD patients with upper and lower limb motor impairments who did not have dementia. Three classes of verbs were used: upper limb verbs (reach), lower limb verbs (kick), and psychological verbs (think). PD patients with upper limb impairments took longer to process upper limb verbs than lower limb verbs, while those with lower limb impairments exhibited the reverse pattern. Other studies have examined the degree to which the action-​sentence compatibility effect is extinguished in PD patients and connected this absence to specific neurophysiological changes (Cardona et al., 2014; Ibáñez et al., 2013; Melloni et al., 2015). This research suggests that action concepts are handled by a cortico-​subcortical network (Cardona et al., 2014). It has been proposed that the extinction of the action-​sentence compatibility effect can be used diagnostically to help with the early detection and assessment of PD (Garcia & Ibáñez, 2014). In sum, studies using different experimental paradigms and techniques implicate both perceptual and action systems in various conceptual tasks. Positing that components of these systems play a constitutive role in conceptual processing provides an economical and robust explanation for a diverse set of observed phenomena, including reaction times, neural activation patterns, the functional character of neuropathologies, and the effects of TMS interventions. With respect to the question of causal relevance, the rapidity and apparently automatic nature of many of the observed effects throws into

36  Abstract Concepts and the Embodied Mind doubt the claim that they are epiphenomenal (although, for a contrary view, see Mahon, 2015). In addition, neuropsychological case studies and TMS interventions provide further, more direct evidence that the areas in question are functionally relevant to the conceptual tasks at hand.

3.3  Simulation and Its Discontents We can use the term experiential systems as shorthand for action, emotion, and perceptual systems. The upshot of our review of the available evidence is that there are good reasons to think that experiential systems are often active during cognitive tasks (Barsalou, 2008; Gallese, 2005; Hesslow, 2012). This activity can be described as a kind of neural reuse (Anderson, 2010, 2014) because the same mechanisms used to perceive and interact with category exemplars are reused when we think about them. We should however remain circumspect about the scope of conceptual grounding for several reasons: First, the available evidence falls short of showing that affective and sensorimotor representations are necessary for conceptual processing. Second, it does not rule out the possibility that nonexperiential representations are also important components of our concepts. Finally, in keeping with the focus of this book, the relevant studies overwhelmingly rely on concrete or highly imageable concepts, which means that the status of abstract concepts remains very much in question (Chatterjee, 2010; Dove, 2009, 2011; Mahon & Caramazza, 2008). Now that I have outlined the case for grounding and introduced the hypothesis that neural reuse is a possible mechanism for taking advantage of hard-​won experiential knowledge, I am going to consider some of the main counterarguments offered by critics of grounded cognition.

3.3.1  Arguments from Dissociation As we have seen, part of the case for grounding rests on neuropsychological case studies that implicate experiential systems in cognition. Unfortunately, there is a sense in which the neuropsychological evidence, taken as a whole, remains equivocal: while some studies appear to implicate experiential systems, others do not appear to do so (Mahon, 2015; McCaffrey & Machery, 2012). For example, it is not uncommon to find preserved

Body in Mind  37 conceptual processing in the face of damage to action-​related sensorimotor areas (Johnson-​Frey, 2004; Mahon & Caramazza, 2005). While it appears to be true that some apraxics experience difficulties with action concepts, it also appears to be true that others do not experience such difficulties (Halsband et al., 2001; Negri et al., 2007). Even when the relevant conceptual deficits are found, they often fail to be catastrophic (Binder & Desai, 2011). Critics of grounding often cite these weaknesses in the neuropsychological literature as evidence of the functional irrelevance of the activity in experiential systems (e.g., Mahon, 2015; McCaffrey, 2015; McCaffrey & Machery, 2012; Weiskopf, 2007)—​but what do they actually show? Clearly, the absence of a profound deficit in the action concepts of apraxics demonstrates that certain action systems and movement abilities are not necessary for those concepts and undermines any theory that posits one-​to-​one links between actions and action concepts. Fortunately, few theories of grounding posit such links. Most view concepts as distributed across multiple experiential modalities. Knocking out one modality should have real consequences, but it need not lead to a catastrophic deficit. As Yee and Thompson-​Schill (2016) put it, losing a finger does not entail losing the entire use of a hand. Typically, the critics are after more than a demonstration of a lack of necessity. For them, the real goal is to show that the activity in experiential systems observed in the various brain imaging experiments is epiphenomenal. As discussed earlier in this chapter, there are good reasons (including, but not limited to, neuropsychological case studies) to think that this activity is functional. This suggests that the effort to leverage the equivocal nature of neuropsychological literature into a case for epiphenomenalism is unlikely to succeed. Supporters of grounded cognition, however, should not celebrate too quickly. The failure of universal epiphenomenalism turns on the fact that compelling evidence supports the existentially quantified claim that specific experiential systems are functionally relevant on some occasions. This falls well short of establishing that all concepts are grounded. In addition, critics of grounded cognition can point to studies that implicate amodal areas in conceptual processing (McCaffrey & Machery, 2012). A good example of this sort of argument involves neuropsychological patients with disorders such as semantic dementia (SD).2 SD is characterized by a gradual bilateral atrophy of the anterior temporal lobes and 2 I will have more to say about SD in upcoming chapters, but, for the moment, I am simply going to use it as a prima facie case for the functional relevance of amodal representations to our concepts.

38  Abstract Concepts and the Embodied Mind a concomitant progressive loss of semantic memory for many semantic domains (Hodges et al., 2000; Lambon Ralph et al., 2001, 2010; Patterson, Graham, & Hodges, 1994). SD patients often exhibit degraded knowledge of several items within a larger category but have preserved knowledge for others. For example, a patient who was assessed longitudinally initially had trouble in a picture-​naming task with the concepts eagle and ostrich, but not with chicken, duck, and swan (Hodges, Graham, & Patterson, 1995; Patterson, Nestor, & Rogers, 2007). Remarkably, the degradation of semantic knowledge in SD can proceed in a hierarchical fashion. For example, as the patient just mentioned gradually lost the ability to identify more and more bird species, they remained able to identify most of them as birds or animals (Hodges et al., 1995; Patterson et al., 2007). One of the reasons that SD appears to undermine strong versions of grounded cognition is that the deficit associated with these items is typically cross-​modal (Bozeat et al., 2000; Garrard & Carroll, 2006; McCaffrey & Machery, 2012). At this stage, the neuropsychological research suggests that the scope of conceptual grounding remains very much in question. While there is support for the claim that experiential systems are functionally relevant some of the time, the extent to which they are relevant is not clear. In addition, some of this research suggests that some aspects of our concepts are handled by amodal systems.

3.3.2  Alternative Hypotheses A common strategy adopted by many critics of grounded cognition is to offer alternative explanations for the relevant data. We have already discussed efforts to explain away the brain imaging results as the downstream byproducts of amodal conceptual processing. These efforts could not be sustained in the face of evidence that the activity in experiential systems is causally relevant at least some of the time. Here, I consider a different sort of alternative explanation that leaves room for the functional use of experiential systems. Recognizing that byproduct views struggle to account for the positive evidence implicating grounded representations in conceptual processes, Machery (2016) offers a different proposal that he refers to as the offloading hypothesis. This proposal holds the line with respect to the amodal nature of concepts but allows that the deployment of sensorimotor representations

Body in Mind  39 may be useful on certain occasions. Machery (p. 1094) summarizes the core idea: “While concepts themselves are amodal, we often manipulate perceptual and motor representations to solve tasks.” He suggests that this proposal promises to invigorate the empirical investigation of the task-​specific deployment of sensorimotor systems. Machery proposes that his hypothesis supports three research questions. The first question concerns the use of perceptual cues as reliable, yet not infallible, heuristics. Machery cites the work of Landy and Goldstone (2007a, 2007b) demonstrating that people employ visual cues associated with perceptual grouping (such as spatial relation and proximity) when they structure and solve arithmetical equations. Whereas Landy and Goldstone propose that their evidence fits best with the hypothesis that mathematical symbols are grounded in visual and spatial structures, Machery offers a more deflationary assessment, suggesting that this is an example of heuristic offloading. The second question concerns the conditions under which it may be possible to intentionally offload. Machery offers no specific examples of extant research to this effect, but one can imagine appealing to the effects of conscious mental imagery on certain tasks. The third question concerns whether offloading is sometimes done in a flexible, context-​sensitive way. Machery acknowledges that the offloading hypothesis invites a certain rejoinder from those who support embodied cognition. He follows the quote just given with the following (p. 1094): “One may wonder why we have amodal concepts at all if we can offload cognitive tasks onto perceptual systems. A plausible response is that not all tasks can [be] offloaded.” This is a plausible response, but it has limited consequences. Even if it is true that some tasks cannot be handled by means of experiential systems, this does not require an outright rejection of conceptual grounding. Machery appears to presume that the only two live options are an amodal approach and a universal form of grounding in which all concepts are made up of modality-​specific representations and nothing else (what he refers to as neo-​empiricism). While some supporters of embodied cognition advocate the latter (which is commonly referred to in the experimental literature as strong embodiment), not all do. In fact, many don’t (Binder & Desai, 2011; Meteyard et al., 2012). There is an obvious flaw with Machery’s reasoning: just because some cognitive tasks are not carried out by means perceptual systems does not mean that all concepts are amodal. Certainly, intermediate positions are available. Given this rich middle ground, the offloading hypothesis may not be a stable position. It attempts to strike a balance between evidence

40  Abstract Concepts and the Embodied Mind implicating amodal representations and evidence implicating experiential representations by excluding offloading from conceptual processing. It is not immediately clear that this exclusion can be justified. After all, we are not working with an a priori notion of concepts. Instead, concepts are theoretical posits intended to explain a broad class of cognitive activities. Because the offloading hypothesis grants that some of these cognitive activities can be carried out by means of the manipulation of sensorimotor representations, we can reasonably ask: Why isn’t offloading just a form of conceptual processing? Pointing out that some concepts may require amodal representations does not answer to this question. Supporters of the offloading hypothesis need to provide principled reasons for claiming that, among all the cases where affective or sensorimotor processes are causally implicated, none is conceptual. It is far from clear that such reasons are forthcoming. The offloading hypothesis faces a related challenge that is perhaps even more damning. What our review has demonstrated is that the engagement of relevant sensorimotor areas is rapid and selective with many cognitive tasks. Often it occurs within 200 ms of the introduction of the relevant stimulus. Supporters of grounded cognition often point out that this fits well with the hypothesis that this engagement is the result of semantic processing. Acknowledging this data, some critics of grounded cognition propose that this engagement might be the result of automatic but nonconceptual processes such as spreading activation (e.g., Mahon & Caramazza, 2008; Weiskopf, 2007).3 In contrast to the spreading activation proposal, the offloading hypothesis grants causal relevance to sensorimotor representations. Unfortunately, it fails to account for the rapid and selective responses found in many of the studies. This means that the offloading hypothesis fails to explain the very sort of data that seem to support grounded cognition. To put this point in a slightly different way, the offloading hypothesis fails to be a compelling alternate hypothesis because it does not account for the target phenomena. None of this is to say that offloading does not occur. Most supporters of grounded cognition would happily agree that it does. Offloading seems like a perfectly reasonable description of instances in which we deploy conscious imagery to carry out some cognitive or metacognitive tasks. It 3 Deflationary proposals such as the spreading activation hypothesis struggle to account for the neuropsychological and TMS data.

Body in Mind  41 just does not provide a compelling means of explaining the growing body of evidence that experiential systems actively contribute to conceptual processing.

3.3.3  Arguments for Amodal Representations In one of my earliest articles on grounded cognition (Dove, 2009), I developed a positive argument for the causal relevance of amodal representations from the existing experimental literature on number sense. After making my case in a manuscript, I made the sort of dispiriting discovery that most academics have experienced at some point in their careers: I was not the first to connect these dots. Machery (2007) had made a very similar case in an earlier paper. Acknowledging that his argument preceded mine, I took my belated arrival at the same conclusion as an indication that I was on the right track. In an interesting plot twist, Machery and I drew very different conclusions from this body of evidence: for him, number sense shows the amodal nature of concepts, and, for me, it indicates the need to posit a theory of concepts that includes both modal and amodal representations. Let’s get to the argument. Several different species have been examined for their ability to make numerical judgments, including dolphins, pigeons, raccoons, rats, and monkeys (Flombaum, 2002; Gallistel, 1990). Sensitivity to numerical properties has been measured in the wild and in the lab using several different research paradigms. In many experiments, researchers have been careful to distinguish between sensitivity to the relative cardinality of sets from sensitivity to other scalar physical attributes of the stimuli. One reason to think that a form of numerical competence is involved is the ease with which animals can transfer numerosity between modalities. A striking example of this is that rats trained to respond to numerical sequences in one modality can generalize to novel sequences involving stimuli in other modalities or even two modalities (Meck & Church, 1983). A common feature of this research is that number discrimination varies with the ratio of the two numerosities in accordance with Weber’s law: that is, in order to obtain the same level of performance with larger numerical quantities that is obtained with smaller quantities, the difference between the compared quantities must be greater. In other words, there is typically a numerical distance effect. Behavioral research on adult humans suggests a similar capacity. In a common experimental paradigm, adults are given brief presentations of

42  Abstract Concepts and the Embodied Mind stimuli containing large cardinalities in order to prevent them from using explicit verbal counting. Another paradigm involves responses (such as sequential button presses) at rates that exclude the possibility of vocal or subvocal counting. These studies indicate that adults are able to represent approximate cardinality despite not having access to verbal representations. Their judgments are at above chance and vary in accuracy in proportion to the size of the compared sets (Cordes, Gelman, & Gallistel, 2002; Hauser & Spelke, 2004). As was the case with the animal studies, there are several reasons to believe that amodal representations underlie this capacity. One indication is that similar performance has been found using various types of stimuli presented in various different modalities (Whalen, Gallistel, & Gelman, 1999). Another is that adults are just as successful when comparing sets across modalities as they are within a modality (Barth, Kanwisher, & Spelke, 2003; Barth et al., 2006). A body of research indicates that preverbal infants have a similar capacity for number approximation. Several studies have shown that infants as young as 6 months have the ability to distinguish sets involving 8 versus 16 or even 16 versus 32 elements when other continuous variables such as element size and total filled area are controlled for (Lipton & Spelke, 2003; Xu, Spelke, & Goddard, 2005). It appears that this ability requires large ratios at first but becomes more precise over development (Lipton & Spelke, 2003; Xu, 2003; Xu & Arriaga, 2007). As was true with the adult studies, these abilities have been shown in experiments involving different modalities. Evidence from neuropsychology and cognitive neuroscience provides further support for a number approximation system. Lesion studies exhibit a double dissociation between number processing and semantic processing. In general, cases in which there is preserved linguistic and semantic processing but impaired number processing involve damage to specific areas of the parietal cortex (Dehaene et al., 2003). Piazza and Dehaene (2004) argue that previous research indicates that one area of the parietal cortex in particular, the horizontal segment of the intraparietal sulcus (HIPS), is the best candidate for a domain-​specific numerical estimation system. Some of their reasons for this claim are the following: the HIPS is more active during estimation tasks than those areas involving accurate computation; the observed activation in the HIPS correlates with numerical distance between compared sets; the HIPS shows higher activation when processing numbers than when processing other continuous categories, such as colors or letters; and stimuli presented in different modalities can activate the HIPS in number-​related

Body in Mind  43 tasks. While it is too soon to definitively identify the HIPS or any other part of the parietal cortex as where number approximation is localized, the results of the imaging studies are suggestive and fit well with the behavioral results described earlier. The existence of amodal symbols for approximate quantities is thus buttressed by a diverse, multidisciplinary, and convergent body of research using various research measurements and methodologies. This shows that there is at least one area of cognition where research supports the causal relevance of amodal representations. At least two responses to this argument are possible. One is that alternative explanations may be available. By these lights, we should not rush to conclude that the representations employed in number approximation are amodal because one could explain this cognitive ability without appealing to extrasensory representations. One possibility is that number approximation is handled by mapping operations between representations in different modalities (Prinz, 2002). The evidence outlined earlier, however, suggests that we can estimate number both within and across modalities. Given this, number estimation cannot be fully explained in terms of a mapping operation between modalities. Another weakness of this proposal is that it does not specify the mechanism by which the mapping is carried out. This is problematic because an effective means of carrying out such a mapping is to have an amodal symbol system tracking approximate numerosity (Meck & Church, 1983). A second possible response is to adopt a deflationary stance. After all, the evidence only shows that amodal symbols are used in a single cognitive domain. Perhaps concepts generally contain perceptual symbols, and number approximation is just an exception to the rule. This deflationary move faces three main challenges. First, it sets foot on a slippery slope because the difference between it and representational pluralism is a matter of degree. Second, the extant data provide no clear reason to prefer this deflationary proposal over representational pluralism since the positive evidence for perceptual symbols is itself limited in scope. Given the limited evidence, why should we think that most concepts are couched in perceptual representations? Third, there are other cognitive domains that seem to involve amodal representations. For example, psycholinguists have long argued that many linguistic representations are amodal. If other domains involve amodal representations, then we should be open to the possibility of amodal semantic representations (Dove, 2009).

44  Abstract Concepts and the Embodied Mind Two intriguing speculations emerge from this case. The first is that amodal codes may be a solution to the problem of information integration within cognitive systems that receive input from diverse sources. This proposal is in keeping with the increasing evidence of cross-​modal effects (Shimojo & Shams, 2001). The second is that grounding might be, in the end, a degree property. Some codes may be more closely tied to a particular modality than others. What remains unresolved is the scope of representational pluralism. Are amodal representations only part of highly specialized systems such as number sense or are they a general feature of our concepts? This is a question that I discuss in detail in subsequent chapters.

3.4  The Symbol Grounding Problem Now that I have reviewed the evidence for grounded cognition and evaluated the arguments against it, I want to turn to an important theoretical benefit that grounding purportedly offers—​a benefit implied by the term itself. One of the compelling reasons to adopt a grounded approach is that holds out the promise of addressing what has come to be known as the symbol grounding problem (Barsalou, 1999, 2008, 2016a; Glenberg & Kaschak, 2002; Glenberg & Robertson, 2000; Zwaan & Madden, 2005). Unfortunately, the case for grounding based on the symbol grounding problem is not as clear-​cut as many take it to be. The symbol grounding problem was first identified by Steven Harnad (1990) in a widely cited paper.4 It is a problem incurred by computationalism. Computationalism requires physical symbol manipulation governed by syntactic rules (Fodor, 1975; Pylyshyn, 1984). For it to work as an account of thinking in general and concepts in specific these symbols must be semantically interpretable. Roughly put, they must represent objects and events in the world. The central question at the heart of the symbol grounding problem is how these symbols accomplish this trick—​how they manage to be about objects and events in the world. The problem arises because computationalism characterizes cognition merely in terms of the relation 4 Harnad (1990) explicitly acknowledges that the symbol grounding problem is inspired by Searle’s (1980) attack on computationalism by means of his “Chinese Room Argument.” Although there is a tendency to lump these arguments together, there are important differences. A good indication that they are distinct is that Harnad’s proposed solution to the symbol grounding problem would not be a solution to the Chinese Room Argument. I am going to focus on Harnad’s formulation because this is the one that has had the greatest influence on embodied/​grounded cognition.

Body in Mind  45 of symbols to other symbols. Harnad (1990, p. 335) summarizes the symbol grounding problem with the question, “How can the meanings of the meaningless symbol tokens, manipulated solely on the basis of their (arbitrary) shapes, be grounded in anything but other meaningless symbols?” He uses a thought experiment to illustrate the problem. He asks us to suppose that we had to learn Chinese as a second language only using a Chinese-​to-​Chinese dictionary. Going through the dictionary would lead to a kind of symbolic “merry-​go-​round” that bounces from one meaningless symbol to another. Harnad proposes that this situation is the same for purely symbolic accounts of cognitive systems because they do not have access to anything but other symbols. This is a general problem for computational approaches to cognition. What it shows is that a system entirely defined in terms the relations between symbols struggles to maintain contact with the world. The obvious solution is to find some alternate way to connect neural representations to their content. Almost everyone agrees that experiential systems are the most likely candidates to provide this connection. Where the disagreement lies is the relationship between these systems and our concepts. Many standard amodal views adopt a modular approach to the mind and posit links between a central autonomous functional module that handles concepts and peripheral modules associated with our experience of the world. Harnad (1990, p. 340) suggests that “this radically underestimates the difficulty of picking out the objects, events, states of affairs in the world that symbols refer to, i.e., it trivializes the symbol grounding problem.” According to him, the assumption inherent in many forms of computationalism, that theories of cognition developed in a top-​down manner will eventually make contact with theories of peripheral experiential systems developed in a bottom-​up manner, is theoretically suspect. He suggests that a more promising approach involves intrinsically dedicated symbols that retain their connection to the world. Supporters of grounded cognition have followed Harnad’s lead and argued that representations borrowed from experiential systems provide a means of overcoming the symbol grounding problem. The idea itself is straightforward: neural reuse provides the cognitive system access to neural representations that retain an appropriate connection to the world and its contents. This is perhaps easiest to see with sensorimotor representations. Consider again our parade case: the concept pug. The hypothesis under consideration is that thinking about pugs involves the selective reenactment of

46  Abstract Concepts and the Embodied Mind various visual, auditory, olfactory, and tactile experiences of pugs. The modal specificity of the deployed neural activity (i.e., the fact that it occurs in the relevant perceptual or motor areas of the cortex) ensures that the selected neural representations retain their pragmatic connection to their referents. In other words, neural simulation enables cognitive processes to build on hard-​won experiential connections to the world that have been gained through sensorimotor experience. It is important to recognize, though, that experiential grounding is likely to extend beyond the prototypical action and motor systems. Exteroception extends beyond the so-​called five senses and includes sensations such of those of balance, itch, pain, and temperature, while interoception includes sensations such as those of hunger and thirst (Barsalou, 2020; Stapleton, 2013). A consequence of the fact that the symbol grounding problem has become part of the origin story of embodied and 4E cognition (Embodied, Embedded, Enactive, and Extended cognition) is that there has been a tendency to overlook the tenuousness of the connection between this very general problem and the specifics of any proposed solution. Several caveats need to be kept in mind. The first is that, at least as it is initially formed, the problem itself only rules out purely symbolic accounts. In other words, it leaves room for hybrid nonsymbolic/​symbolic accounts of various types. While it may turn out that fully grounded concepts provide the most satisfying solution to the problem, other solutions remain viable. The second caveat is that resources are available for supporters of amodal theories. Shapiro (2019) argues that any evaluation of a solution to the symbol grounding problem must be made against of the background of an accepted naturalistic theory of meaning. The trouble is that there are theories of meaning that are compatible with traditional symbolic accounts but seem to sidestep the problem. For instance, information-​theoretic accounts of meaning (Dretske, 1981) appeal to causal connections between symbols and the world without relying on the sort of symbol-​to-​symbol relations assumed by the symbol grounding problem. Some use the purported failure of the philosophical research project to find a naturalistic theory of meaning as a reason to adopt an embodied or grounded approach (e.g., Hutto & Myin, 2013), but this move clearly involves extra premises. The third caveat is that the details of how experiential grounding connects neural representations to objects, events, and states of affairs are not always provided. Admittedly, it is true that some specific proposals have been offered. Barsalou (1999) and Prinz (2002), for example,

Body in Mind  47 adopt an information-​theoretic perspective and contend that grounded representations provide an important leg up with respect to establishing the appropriate causal connections between neural symbols and their referents. Cuccio and Gallese (2018) maintain that grounded simulations act as icons and serve as the input for abductive inferences. Stapleton (2013) proposes that a result of grounding is that cognition depends on fine-​grained physiological processes in such a way that the distinction between algorithmic and implementation levels breaks down. The diversity of these proposals demonstrates that much remains to be determined. Given these caveats, I hesitate to treat the symbol grounding problem as fully determinative. Instead, I view it as a component of a broader case for grounded cognition. Throughout this chapter, we have reviewed evidence that cognition relies in part on experiential neural systems. This evidence suggests that thinking about the world involves the very same neural mechanisms that are used to experience it. Importantly, this reuse appears to be functional and causally relevant rather than epiphenomenal. What the symbol grounding problem highlights is a possible reason for this reliance on experiential representations: they may help us connect to the world.

3.5  Weak Versus Strong Embodiment In this chapter, I have surveyed some of the positive evidence for grounding and considered some of the arguments against it. While this survey falls short of a complete review, it goes into enough depth to highlight some of the core issues. Theories of embodied/​grounded cognition come in many forms. Recently, researchers have begun to distinguish theories in terms of the degree to which they are committed to strong or weak embodiment. There are at least two ways of defining the difference between weaker and stronger grounding. The first, and by far more prevalent, involves the types of representations that are reused during conceptual processing (Meteyard et al., 2012). On this construal, stronger theories are characterized by the degree to which they limit reuse to primary experiential representations and weaker theories are characterized by the degree to which they allow higher-​ level experiential representations and multi-​or cross-​modal representations. By these lights, the evidence implicating higher-​ level representations during the processing of action verbs, which I outlined earlier, favors weak grounding (Kemmerer, 2015c).

48  Abstract Concepts and the Embodied Mind A second way of defining the spectrum involves the relative centrality of reuse to our concepts. Kompa (2019) provides a clear example of this sort of approach. She focuses on language comprehension rather than conceptual processing, broadly construed, and embodiment rather than grounding (a term which I prefer because it manages to simultaneously be more specific and leave room for other aspects of 4E cognition), but her approach will apply mutatis mutandis to concepts and grounding. Kompa finds differing views of the importance of simulation (and enactment) in the literature. She begins with what she takes to be a minimal commitment of embodied cognition and proposes that weaker theories are at least committed to the following claim (Kompa, 2019, p. 2): E-​Claimweak: Simulation (by means of modality-​specific, embodied representations) is necessary for language comprehension.

Those favoring a stronger version of embodiment are committed to one of the following claims: E-​Claimstrong-​1: Simulation is necessary and sufficient for language comprehension. E-​Claimstrong-​2: Language comprehension is simulation.

Kompa then goes on to offer several reasons to find the evidence marshaled in support of embodiment inconclusive, including the inconsistency of the activation of experiential areas (Watson et al., 2013), the lack of precise overlap in the sites of activation between cognition and experience (Fargier, 2016), the absence of relevant conceptual deficits in patients with damage to the relevant experiential modalities (Mahon & Caramazza, 2005, 2008), the fact that conceptual losses in SD are not modality-​specific (McCaffrey, 2015; McCaffrey & Machery, 2012), and the existence of similar patterns of activity in the processing of visual action verbs by blind and sighted individuals (Bedny et al., 2012). In other words, she argues that even weak embodiment is not supported by the extant evidence. In her discussion of these challenges to grounded cognition, Kompa (2019, p. 3) makes a striking aside in which she acknowledges that simulations might actively contribute to linguistic processing but not be strictly necessary. Later in the essay, she also grants that (p. 16), “It is immensely plausible

Body in Mind  49 that symbols acquire meaning only by being grounded in perception and action—​in phylogeny as well as ontogeny.” She goes on to propose that grounded representations may well be deployed in a context-​sensitive and task-​dependent way. The apparent ambivalence of Kompa’s stance—​which dismisses weak embodiment and yet acknowledges its potential importance—​demonstrates that we need to rethink how we construe and evaluate the commitments of grounded cognition. I suggest that what cognitive scientists and neuroscientists are really after is an understanding of the physical processes (both internal and external to the body) that are causally responsible for our conceptual abilities. Of particular interest are the neuromechanisms involved in semantic memory. Against the background of the research program of cognitive neuroscience, we can offer the following reformulation of a core claim of grounded cognition: G-​Claim: Neural reuse (simulation or reenactment) is causally relevant to conceptual processing.

The central question then becomes: What is the nature and extent of this causal influence? We can define conceptions of grounding in terms of the centrality and scope of the causal influence of neural reuse. At one end of the spectrum are accounts of grounding, such as the offloading hypothesis, that view simulation or reenactment as functional only under very limited and circumscribed conditions. At the other end are accounts that view simulation or reenactment as central to all conceptual processing. With this reformulation on hand, it is possible to offer a straightforward assessment of the evidence surveyed in this chapter. This evidence provides compelling support for the claim that the neural systems we use to experience objects, actions, and events in the world can be reused to internally simulate those objects, actions, and events at later points in time. Questions remain about the scope of grounding. The body of evidence discussed so far is incomplete because it primarily involves concrete or highly imageable concepts (Dove, 2009; Louwerse & Jeuniaux, 2008; Pezzulo & Castelfranchi, 2007). It also fails to exclude the possibility that amodal representations contribute to our concepts. Indeed, as we saw in our discussion of criticisms of grounding, there is even some evidence that conceptual processing is handled to some extent by cortical areas that are not modality specific.

4 Three Problems As discussed in Chapter 3, one of the purported benefits of conceptual embodiment or grounding (Barsalou, 1999; Glenberg & Robertson, 2000; Prinz, 2002) is its ability to overcome the symbol grounding problem (Harnad, 1990). This problem arises because a system containing only abstract symbols and their interrelations struggles to explain how representations come to be associated with things and events in the world. Supporters of grounding propose that the links between the representations of modality-​specific sensorimotor systems and dynamic interactions of our body with our external environment (both physical and social) provide a means of overcoming the symbol grounding problem. Presumably, given their etiology and function, affective and sensorimotor representations manage to hook up to the world with some degree of success. A conceptual system selectively employing these representations would seem to have a leg up with respect to capturing any semantic content associated with those aspects of our environment that we can perceive and act on in some sort of dynamic and immediate way. In other words, grounded representations are advantageous because they leverage their experiential origins. However, this benefit may also have a down side. After all, many of our concepts refer to entities or properties that are not closely linked to entities or properties that we can directly perceive or act upon. Representational systems containing modality-​specific symbols would thus appear to face a kind of corresponding symbol ungrounding problem: How can conceptual representations that are grounded in sensorimotor systems capture abstract conceptual content? This is not merely a theoretical challenge. Consider the findings from a recent imaging study (Fernandino et al., 2015). Using the functional magnetic resonance imaging (fMRI) data from an initial set of 820 words, the researchers developed an encoding model based on the relevance of five sensorimotor attributes: color, manipulation, shape, sound, and visual motion. They tested the model on a set of 80 novel words. The model performed better than chance at predicting the distributed fMRI activation pattern elicited by Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0004

Three Problems  51 the concrete words in the novel set but failed to perform better than chance with the abstract words. There is an important sense in which it is not an entirely new problem. The British Empiricists—​most prominently John Locke, George Berkeley, and David Hume—​were a group of philosophers in the seventeenth and eighteenth centuries who argued that most, if not all, knowledge depended on sense experience. Each of the major figures of British Empiricism recognized that abstract ideas (which presumably involve abstract concepts) represent an important theoretical challenge for their approach, and each proposed his own theory of how abstract ideas are acquired from sense experience. While there are several important differences between empiricism and embodied cognition, both must deal with the distal connection that many of our concepts have with respect to our (sensory) experiences. Indeed, I would suggest that any theory that posits an intimate connection between conception on the one hand and action/​perception on the other owes some explanation of our capacity to use and understand abstract concepts. Given this historical precedent, it should not be surprising that several people have argued that our facility with abstract concepts poses a serious theoretical challenge to grounded cognition (Chatterjee, 2010; Dove, 2009; Machery, 2007; Mahon & Caramazza, 2008; Weiskopf, 2007). Initially, this worry was strengthened by the fact that most of the evidence involved concrete or highly imageable concepts. With time, though, things have become more complicated. First, more and more research from an embodied and grounded perspective has been carried out on abstract concepts (e.g., Barsalou & Wiemer-​Hastings, 2005; Borghi & Binkofski, 2014; Glenberg et al., 2008; Kousta et al., 2011; Scorolli et al., 2011, 2012; Vigliocco et al., 2014; Watson et al., 2013). Second, there is a growing consensus that some kind of hybrid account is needed (e.g., Dove, 2009, 2011; Pulvermüller & Garagnani, 2014; Reilly et al., 2014; Shallice & Cooper, 2013; Watson & Chatterjee, 2011; Zwaan, 2014). Both factors suggest that there is a real need to reexamine abstract concepts and how they fit within a comprehensive theory of concepts. We need to move beyond our preconceived notions of abstract concepts and actively rethink the questions that they raise for grounded views of concepts. My aim in this chapter is to show that, when viewed against the background of current research and theory, abstract concepts do not pose a single cohesive problem for embodied cognition but rather several distinct problems (for a similar acknowledgment, see Barsalou, 2010). In particular,

52  Abstract Concepts and the Embodied Mind they raise what I refer to as the problems of generalization, disembodiment, and flexibility.

4.1  The Problem of Generalization Generalization would seem to be a fundamental design feature of our concepts. Asking why we have concepts to begin with, Yee writes (2019, p. 1258; emphasis in the original): Most people who study concepts would agree that concepts allow us to make sense of and organize the world—​they allow us, when encountering something new, not to have to learn from scratch what it does, how we should interact with it, and how it might change. In this sense, having concepts gives us the ability to generalize—​which we define here as the ability to apply knowledge that we have learned to new situations.

By these lights, supporting generalization is a core function of our concepts. How is generalization achieved? Yee (2019) suggests that it depends crucially on processes of abstraction that enable us to pick up on statistical regularities embedded in our experiences of the world. Through abstraction we are able to aggregate commonalities that extend across affective and sensorimotor episodes and extract information that goes beyond direct experience. Abstraction is a matter of degree—​that is, some concepts appear to require more abstraction than others. A good example are conceptual hierarchies in which superordinate concepts appear to require greater abstraction than subordinate ones (Rosch, 1978). For example, we can situate the basic-​level concept dog within a conceptual hierarchy in which pug is more specific and mammal is more general. Among cognitive scientists, it is not uncommon to define abstract concepts in terms of generalization understood in this vertical sense. Researchers in epigenetic robotics, for instance, explicitly characterize them as higher order concepts (Cangelosi & Riga, 2006; Stramandinoli, Marocco, & Cangelosi, 2012; Thill, Pado, & Ziemke, 2014). After acknowledging that abstract concepts have not received enough attention by researchers in cognitive robotics who are committed to the importance of sensorimotor grounding, Thill, Pado, and Ziemke (2014, p. 547) point to recent work that attempts to address this lacuna.

Three Problems  53 Nonetheless, some work in the field of cognitive robotics has explicitly tackled the grounding of abstract (higher order) concepts. . . . Here, the core idea is that, while some concepts can be directly grounded in sensorimotor experience, others are grounded in other concepts, which may or may not in turn be directly grounded. For instance, an agent might have a direct grounding of concepts like “grasp,” “push,” “release,” and “frown” and from this constructs a higher-​order concept such as “give” (grounded in “grasp,” “push,” and “release”) or the even higher “reject” (grounded in “give” and “frown” . . . ).

For the moment, let’s put aside the question of whether or not the proposed conceptual groundings are likely to work and instead focus on the notion of abstraction on offer. The idea is that higher-​order concepts are built up from lower-​order concepts. Often these hierarchies bottom out in concepts grounded in sensorimotor (and affective) experience (see Chapter 5 for a discussion of multilevel models of semantic memory that explore this idea in more detail). Presumably some process of generalization facilitates the establishment of these hierarchies. The theoretical challenge posed by abstract concepts then becomes to explain how this generalization works. What neural mechanisms are involved? How do they fit with embodied cognition? Some researchers contend that fully grounded theories are not up to the task of addressing this problem. Patterson, Nestor, and Rogers (2007, p. 977), for instance, write, If semantic memory consisted only of the modality-​specific content of objects (and the links between them), it is doubtful that we could ever achieve the higher-​order generalizations on which so much of our semantic processing relies.

The idea is that the representations involved in neural simulations are simply not well suited for the task of capturing higher-​level generalizations. At the very least, supporters of conceptual grounding owe us an explanation of how such generalizations can be achieved and put to use. Because amodal representations provide an effective means of integrating information from multiple sources (Dove, 2009; Machery, 2007), a controversy has arisen concerning whether or not theories of embodied cognition should make room for them. Meteyard and colleagues (2012) suggest that we can place different variants of embodied cognition along a continuum ranging

54  Abstract Concepts and the Embodied Mind from the strongly to the weakly embodied: at the strongly embodied end are theories that fully limit cognition to action, emotion, and perception systems (e.g., Gallese & Lakoff, 2005; Glenberg & Gallese, 2012; Pulvermüller, 2001); at the weakly embodied end are theories that see the activation of sensorimotor areas as a secondary consequence of cognitive processing handled by amodal areas (e.g., Chatterjee, 2010; Mahon, 2015; Mahon & Caramazza, 2008); and in between are theories that advocate for some sort of intermediate position at which concepts are at least partially grounded in action and perception systems (e.g., Binder & Desai, 2011; Simmons & Barsalou, 2003; Vigliocco et al., 2004).

4.2  The Problem of Disembodiment Not everyone agrees that generalization is the core issue associated with abstract concepts. Borghi and Binkofski (2014), for instance, suggest that we need to distinguish abstractness from mere abstraction. Some examples may help to elucidate their distinction. A concept such as mammal may sit at the top of a conceptual hierarchy and thus be more abstract than related lower-​order concepts, but its referents are nevertheless concrete, perceivable objects. A concept like odd number, on the other hand, is abstract because its referents are not concrete, perceivable objects. Odd numbers are, instead, something more ephemeral. Borghi and Binkofski contend that abstractness, not abstraction, is the real problem posed by abstract concepts. I offer a friendly amendment to their view and suggest that both abstraction and abstractness are problems posed by abstract concepts. In my terminology, “abstraction” is associated with the problem of generalization and “abstractness” is associated with the problem of disembodiment. Both are important, and both need to be addressed. As was the case with the problem of generalization, there are reasons to think that people are aware of the problem of disembodiment. Mahon and Caramazza (2008, p. 60) offer the following parsimony argument in support of an amodal approach to concepts: Given that an embodied theory of cognition would have to admit “disembodied” cognitive processes in order to account for the representation of abstract concepts, why have a special theory just for concepts of concrete objects and actions?

Three Problems  55 Such parsimony arguments are far from convincing. After all, the history of psychology is rich with highly economical failed theories. What I am interested in is a core premise of this argument (i.e., that abstract concepts require disembodied cognitive processes). No matter how abstract concepts are realized in the brain, the categories associated with many of them are in some important sense divorced from experiential factors. Abstract concepts tend to have referents that are not temporally or physically bounded entities or events that we can easily perceive with our senses or manipulate with our actions (Borghi & Binkofski 2014; Borghi et al., 2017). They also tend to involve more complex relations and associations, introspective features, and social interactions than concrete concepts (Barsalou, 1999, 2008). A robust body of evidence suggests that abstract concepts are processed in a functionally and neuroanatomically different way than other concepts. Recent meta-​analyses of neuroimaging data find that abstract concepts are processed, at least in part, by different brain areas than are concrete concepts (Del Maschio et al., 2022; Wang et al., 2018). Concreteness or imageability effects provided an early indication of this difference. Concreteness is typically defined as the extent to which an item or event can be experienced by the senses, and imageability is typically defined in terms of the subjective ease with which a word gives rise to sensorimotor mental imagery. Because these two measures overlap to a great degree, researchers have tended to treat them as interchangeable (although see Vigliocco et al., 2011 for a discussion of their differences). Reliable imageability ratings on number scales have been gathered for linguistically encoded concepts by several researchers (Bird, Franklin, & Howard, 2001; Paivio, Yuille, & Madigan, 1968; Toglia & Battig, 1978). Much of the original research on imageability was behavioral and demonstrated a processing advantage for highly imageable concepts: in several cognitive tasks, concrete/​ high-​imageable words exhibit a number of processing advantages over abstract/​low-​imageable words (Paivio, 1986; Wattennmaker & Shoben, 1987). For instance, lexical access has been shown to be quicker for highly imageable words than for abstract ones (Coltheart, Patterson, & Marshall, 1980). Highly imageable words are also recalled more quickly in memory tasks than are abstract words (Paivio, 1971, 1986; Wattenmaker & Shoben, 1987). A similar advantage is also found in word comprehension tasks (Schwanenflugel, Harnishfeger, & Stowe, 1988).

56  Abstract Concepts and the Embodied Mind More recently, a motor-​related measure has been used to similar effect. The dimension of body–​object interaction (BOI) is meant to capture the ease with which a human body can physically interact with category exemplars. A number of studies have indicated that words with higher BOI ratings are processed more efficiently than words with lower BOI ratings (Siakaluk et al., 2008; Wellsby et al., 2011; Yap et al., 2012). Lynott and Connell (2009) hypothesize that words that express strongly perceptual concepts should be processed more efficiently than those that express less perceptual concepts. They develop a different new measure: perceptual strength. Participants are asked to rate the extent to which they experience the referent of a word through each of the five senses (hearing, seeing, smelling, tasting, and touching). Evidence suggests that the maximum perceptual strength rating for a word (i.e., the strength rating found in the dominant sensory modality for that word) provides a better predictor of word processing advantages than either concreteness or imageability (Connell & Lynott, 2012). The behavioral differences associated with these measures suggests that abstract concepts may be processed in different ways than other concepts. Evidence from cognitive neuroscience broadly supports a neurophysiological distinction between abstract/​ low-​ imageable and concrete/​ high-​ imageable concepts. For instance, Adorni and Proverbio (2012) applied LORETA source reconstruction to an event-​related potential (ERP) experiment involving a lexical decision task and found evidence of increased activation of the left medial frontal gyrus and the left temporal cortex, as well as decreased activation of extrastriate visual areas, with abstract relative to concrete words (see Lehmann et al., 2010 for similar findings using different tasks). In another ERP study, distinct concreteness effects were elicited in the left and right hemispheres (Huang, Lee, & Federmeier, 2010). Neuropsychological research provides further reason to suppose that abstract/​ low-​ imageable and concrete/​ high-​ imageable concepts are partially supported by distinct neurological systems. A greater impairment for the processing of abstract words has been correlated with left hemisphere damage, including patients who present with aphasia (Goodglass, Hyde, & Blumsten, 1969), deep dyslexia (Coltheart et al., 1980; Franklin, Howard, & Patterson, 1995; Shallice & Warrington, 1975), and deep dysphasia (Katz & Goodglass, 1990; Martin & Saffran, 1992). Some patients with semantic dementia (SD) can exhibit a contrasting impairment for concrete/​high-​imageable words (Reilly & Peelle, 2008; Yi,

Three Problems  57 Moore, & Grossman, 2007). There is an active discussion among researchers concerning whether this reversal is a typical feature of SD (e.g., Bonner et al., 2009) or not (e.g., Hoffman & Lambon Ralph, 2011). Whatever the ultimate answer to this theoretical question, the relative preservation of abstract conceptual knowledge in the face of a processing deficit for concrete concepts has been attested in several case studies. For example, the patient AB was able to produce definitions for abstract concepts much more effectively than concrete concepts (Warrington, 1975). The following two examples of spontaneous speech from another patient, identified by the initials SC, manifest a dissociation between concrete and abstract word processing (Macoir, 2009): “When there is a lot of snow on the roof of my house and my driveway, I remove the snow. I go on the roof of my house with my legs. I cannot tell you how I go up there. Also, I cannot tell you what I use to remove the snow. Also for my driveway, I use another object that I cannot name. To defrost also, I do it but I do not remember the word for it.” “I progressively discover that my personal thoughts develop my well-​ being. I do not wish to prove that these thoughts are absolute. On the other hand, I wish to try them out daily to check if they can maintain this well-​being. What emerges gradually is the awareness of the symptom. What happens is that I am not able any more to increase my personal well-​being. When I wake in the morning, my anxiety emerges immediately. The probability that I will make a mistake is huge. I constantly live in danger . . . ”

In the first example, SC exhibits clear difficulties with accessing specific words for familiar concrete objects while describing everyday actions. In the second, SC’s speech is more fluent and exhibits few difficulties with abstract mental terms. A study of healthy participants also supports a functional and neuroanatomic abstract/​concrete asymmetry. Papagno and colleagues (2009) found that accuracy on a lexical decision task decreased with abstract concepts when repetitive transcranial magnetic stimulation (rTMS) was applied over the left inferior frontal gyrus and the left superior temporal gyrus, but accuracy decreased with concrete concepts when rTMS was applied over the right superior temporal gyrus. The problem of disembodiment arises because there appears to be a qualitative difference between the contents of abstract and concrete concepts.

58  Abstract Concepts and the Embodied Mind A diverse body of evidence suggests that abstract concepts are processed differently than concrete concepts.

4.3  The Problem of Flexibility The third problem has a significant historical precedent: Aristotle’s well-​ known criticism of Plato. Aristotle argues that Plato is too quick to assume that abstract concepts are univocal (Shields, 2020). For example, he challenges Plato’s claim that a universal goodness is shared by all good things (Aristotle, 1995). Aristotle contends that, if one looks at different exemplars of goodness, one finds evidence that the term is multivocal. In other words, the factors that determine goodness depend on the type of thing being evaluated and the context of evaluation. The conditions that make something a good pop song are different from those that make something a good apple, a good form of government, or a good proof in mathematics. In other words, the relevant concept of goodness is determined in part by what category we are assessing. The relevant background conditions can be quite specific. Are we talking about pop songs from the 1940s or the 1980s? Fruits in the context of a snack or an ingredient for a pie? Without getting into difficult questions of translation and scholarship, Aristotle speaks of these issues in largely linguistic terms (even though he thinks that the multivocality of goodness has metaphysical consequences). Given this linguistic focus, we need to decide what his examples show about the underlying concepts. One possible interpretation of the situation is fairly deflationary. On this reading the multivocality of goodness is simply the result of garden variety polysemy. Typically, when philosophers discuss ambiguity or polysemy, they do so in terms of concrete object nouns. The overused parade case is the English word bank. This word can be used to refer to a side of a river, a slope, or a financial institution (other readings are also possible). The stark differences between these interpretations suggest that this is a true case of polysemy and that we can treat its distinct uses as arising from homonyms. A problem with this focus on object nouns is that it makes polysemy seem rare, circumscribed, and straightforward. Even a cursory consideration of verbs and adjectives reveals that their meaning is often more open-​ended and complex. We don’t need to look any further than children’s books and nursery rhymes. Consider the beloved character Amelia Bedelia. In a long

Three Problems  59 running series of books, she repeatedly screws up instructions given to her by her employers by misinterpreting them. More often than not, her failures involve verbs or adjectives. In the first book (Parish, 1963), Amelia changes the towels by cutting them up, dusts the furniture but spreading dusting powder on them, draws the drapes with a paper and pencil, and measures two cups of rice by means of a tape measure. While some of these misinterpretations clearly arise from lexical ambiguity or polysemy (as is the case with the verb draw in the preceding example), others are harder to place. For instance, the measuring failure seems likely to be the result of pragmatic considerations and discourse factors. What the series of Amelia Bedelia books reveals is that verbs and adjectives—​which tend to be more abstract than concrete nouns—​ often exhibit a semantic flexibility. This flexibility can often go unnoticed because of a failure to recognize the implicit influence of context. Consider the famous nursery rhyme line Mary had a little lamb. Pinker (1994, p. 208) points out that several surprising interpretations are possible. To get these, imagine adding the phrases with mint jelly and as a part of an interspecies breeding experiment. There is even an x-​rated reading that we do not need to make explicit (please excuse the pun). In keeping with this example, we could give the verb have an Aristotelian treatment and consider different uses of this term to see whether or not there is an underlying universal concept. As was the case with goodness, this effort faces significant challenges. It is far from clear that a single notion underlies having a dog, a migraine, a PhD, a bank account, a desire for the sweet embrace of death, a spouse, a neurodegenerative disease, and a fondness for donuts. Certainly, the fact that we use the same word in the different cases is not sufficient evidence of an underlying universal concept. A possible response to this flexibility is the deflationary one offered earlier. We might treat the verb have as highly polysemous and posit a set of determinate semantic contents that are all stored separately in our mental lexicon. By these lights, our job as interpreters is to figure out which of these distinct meanings is intended by the speaker in the relevant context. Another response is possible, though. On this version, the word is associated with a smaller number of meaning schemas whose content is then filled through discourse processes associated with aspects of the context of utterance. Within the philosophy of language, this latter approach to semantic content is known as contextualism (Carston, 2002; Recanati, 2004). I propose that we should adopt a similar view of conceptual content in which the same concept

60  Abstract Concepts and the Embodied Mind is realized in neurologically different ways in different contexts and with different tasks (see also Barsalou, 2016b; Connell & Lynott, 2014b). To be clear, concrete concepts are also flexible. Flexibility is a degree property. It is just that abstract concepts exhibit greater flexibility than concrete concepts (Hoffman, 2016). This difference in degree arises from the fact that they tend to be about objects, events, and relations that are more heterogeneous and more distributed across time and space (Davis, Altmann, & Yee, 2020; Wilson-​Mendenhall et al., 2013). Their referents tend to be more dispersed across contexts because they are less directly tied to our immediate experiences of category members or situations. As was the case with the other two problems posed by abstract concepts, researchers have developed explanations and measures of the abstract–​ concrete distinction that address the problem of flexibility. One of the early accounts of concreteness effects sought to explain them in terms of the differing degree of context availability (Schwanenflugel et al., 1988). This measure is typically formulated in terms of the subjective ease with which participants can think of a specific context for the use of a word. On this view, we have more difficulty generating plausible contexts in which abstract concepts might be used than we do for concrete concepts. More recently, situated simulation theory (Barsalou, 2016; Barsalou, Dutriaux, & Scheepers, 2018; Barsalou et al., 2008; Barsalou & Weimar-​ Hastings, 2005) has adopted some aspects of context availability theory. This theory claims that both concrete and abstract concepts are represented by means of the selective reenactment of the neural activity elicited by our experiences of specific situations. What distinguishes them is the particular aspects of situations that they focus on: abstract concepts tend to focus on social aspects, while concrete concepts tend to focus on physical entities and actions. In a preliminary study (Barsalou & Wiemer-​Hastings, 2005), participants listed typical properties for three abstract concepts (truth, freedom, and invention), three concrete concepts (bird, car, and sofa) and three intermediate concepts (cooking, farming, and carpeting). They generated situational properties with both concrete and abstract concepts, but they tended to generate more event and introspective properties with abstract concepts. A more fully realized experiment employing similar methodology found that participants tended to generate fewer entity properties, more introspective properties, and more relational properties with abstract concepts than with concrete concepts (Wiemer-​Hastings & Xu, 2005).

Three Problems  61 Other researchers have developed different measures that capture aspects of flexibility. Some of these measures focus on the linguistic contexts in which abstract words appear. For example, semantic diversity is a measure of the degree to which a word is used in different contexts (Hoffman, 2016). Working with large corpora of real language data, researchers have calculated the average similarity of the contexts in which a given word was used (Hoffman, Lambon Ralph, & Rogers, 2013). Words encoding abstract concepts tend to rate higher on semantic diversity than those encoding concrete concepts. For example, the word spinach tends to occur in contexts relating to cooking and eating and thus receives a relatively low semantic diversity rating while the word life can occur in a number of different contexts (e.g., life on earth, life sentence, shelf life, life of the party, and stage of life) and thus receives a higher rating (Hoffman, 2016). Semantic relatedness tasks appear to take longer with words that have high semantic diversity irrespective of imageability (Hoffman & Woollams, 2015). Neuropsychological evidence provides further reason to think that semantic diversity is more than simply a negative correlate of imageability. Although patients with semantic aphasia (SA) and patients with semantic dementia (SD) both experience processing difficulties with abstract words, semantic diversity was only a significant predictor for word processing performance with SA patients (Hoffman, 2016). Semantic diversity has a potential weakness as a measure of abstract concepts. As defined, the measure applies merely to words and their linguistic contexts. It is not immediately clear, though, that semantic diversity captures the relevant underlying concepts. It might well be the case that semantic diversity provides an operationalized indirect measure of an important property of abstract concepts, but it is important to keep in mind that concepts may contain nonlinguistic information and need not be linguistically encoded. One can have a concept of schadenfreude or umami before learning the relevant terms. Presumably, part of the explanation of the emergence and spread of these terms is to be explained by the fact that they capture accessible conceptual content. Recently, some researchers have tried to develop more direct measures of what I have been calling flexibility. For instance, Davis, Altmann, and Yee (2020) propose that abstract concepts exhibit less situational systematicity than concrete concepts. Rather than focus on linguistic contexts, this measure focuses on “the real-​world contexts in which a concept might be acquired and recognized” (p. 3; emphasis in the original). The idea is that

62  Abstract Concepts and the Embodied Mind situations to which abstract concepts apply tend to be more diverse than those to which concrete concepts apply. This approach fits well with theories of concepts that emphasize their dynamic nature and view them as schemas which capture information about how objects and events interact within real-​world situations (Barsalou, 2009; Barsalou, Dutriaux, & Scheepers, 2018; Gilboa & Matlatte, 2017). It also fits with evidence suggesting that abstract concepts contain some situation-​based perceptual knowledge. In one recent study, pictures of situations, such as two girls sharing a corn cob, were found to facilitate the processing of abstract words such as the verb share in a lexical decision task, and the presentation of the word share 250 ms before the presentation of the picture facilitated the judgment that the picture depicted a normal situation (McRae et al., 2018). As discussed in Chapter 2, many researchers in grounded cognition have moved beyond the traditional commitment to invariantism. This shift has occurred in large part because of evidence that grounded representations can be engaged to a greater or lesser degree depending on context, goals, and task. From this perspective, all concepts are flexible. The problem of flexibility posed by abstract concepts arises because they tend to exhibit greater flexibility than do concrete concepts. This difference is revealed by measures such as context availability, semantic diversity, and situational systematicity.

4.4  Abstract Concepts Reconsidered Traditionally, cognitive science has examined our concepts from a computational perspective that views cognition as being functionally independent from perception and action (Hurley, 2008). Embodied cognition offers an alternative framework, one that views cognition as being fundamentally grounded in action, emotion, and perception systems. In this heterogeneous and constantly evolving research program (Barsalou, 2008, 2010, 2012; Gibbs, 2006; Wilson, 2002), abstract concepts have emerged as an important challenge. In this chapter, I have argued that abstract concepts pose three separate problems for grounded cognition: the problems of generalization, disembodiment, and flexibility. Once we move beyond a pre-​theoretical notion of abstractness and look at the specific issues engendered by the diverse bodies of research on abstract concepts, we are in a better position to assess and develop our

Three Problems  63 theories. The rest of the book will focus on developing theoretically a unified approach to grounded cognition that neither assumes that abstract concepts are homogeneous nor conflates these three problems. I argue that a multimodal, hierarchical, and flexible view of the human conceptual system provides a compelling account of how we are able to handle concepts of all stripes.

5 Hierarchies and Hubs My aim in this book is to show that a clear-​eyed consideration of the challenges posed by our facility with the various types of abstract concepts forces us to reconsider the nature of grounded cognition. In the preceding chapter, I outlined three distinct theoretical problems posed by abstract concepts: the problems of generalization, disembodiment, and flexibility. In this chapter, I focus on the problem of generalization. Generalization lies at the heart of conceptualization. This is true for both abstract and concrete concepts. This chapter examines what we know about the neural mechanisms responsible for generalization both within and between experiential modalities. Hierarchical structure appears to be a common feature of sensorimotor systems. Similarly, action–​perception feedback loops help guide and update action plans. Both suggest that higher-​level representations might play an important role in grounded concepts. In what follows, I outline and defend an account of flexible grounded neural reuse that recognizes the importance of hierarchical representations.

5.1  Beyond Unimodality At the heart of the problem of generalization is the question of how our brains can capture semantic knowledge that extends beyond the particulars of our individual experience. This question remains especially pressing for grounded theories of concepts. Such theories must provide some explanation of how experiential representations can pull this off. Otherwise, grounded cognition would seem unable to meet a minimal explanatory aim of a theory of concepts. We have reviewed the evidence implicating experiential systems in our concepts. This evidence suggests that cognition often involves the selective reactivation of experiential representations during cognition. This selective reactivation is often referred to by the more intuitive term simulation. The general idea is that we simulate aspects of our experience when we use our Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0005

Hierarchies and Hubs  65 concepts. A purported benefit of simulation is that it helps overcome the symbol grounding problem. It is supposed to do this by leveraging the connection between experiential representations and their referents. What is the nature of this connection, though? Does it support the claims offered in support of grounded cognition? In order to see the importance of these questions, consider a counterargument to grounding offered by Hickok (2014, p. 128). To say that a cognitive operation is accomplished via simulation doesn’t simplify the problem, it just hands it off to another domain of inquiry, in this case sensory and motor information processing.

The idea is that supporters of grounded cognition have simply shifted the problem to another subdiscipline rather than answered it directly. Responding to concerns of this sort, some supporters of grounding have pointed out that any successful account of action and perception systems will have to overcome the symbol grounding problem (e.g., Barsalou, 1999; Prinz, 2002). After all, there is little question that we successfully act on the world and perceive objects and events. No theory of cognition should be asked to settle all theoretical issues at once, and it seems reasonable to appeal to a broadly mainstream conception of the experiential systems—​particularly ones as well-​studied as the visual and motor systems. While Hickock’s criticism may not be as devastating as initially advertised, he still has a point: namely, there is a sense in which supporters of grounded cognition have not done their theoretical homework and thought carefully enough about the nature of action and perception. This is particularly evident when we look at the central notion of a modality-​specific representation. Given that this notion purportedly separates grounded approaches to cognition from others, one would think that supporters would have a firm grasp of what is and is not an experiential modality. I suggest that they don’t. This turns out to be a significant oversight because individuating the experiential modalities turns out to be a difficult and thorny problem. We can see this clearly when we try to figure out just exactly what qualifies as a sensory modality. Ever since Aristotle, it has been common to talk of five senses: hearing, sight, smell, taste, and touch. Psychologists and brain scientists have long recognized that this list is incomplete. At a minimum, it seems likely that we need to add interoception (which is associated with our awareness of the internal states of our bodies—​e.g., thirst and hunger),

66  Abstract Concepts and the Embodied Mind proprioception (which is associated with our awareness of the position of our bodies and limbs), and vestibular sense (which is associated with our sense of the orientation of our bodies with respect to gravity). Questions have also been raised about the prototypical senses. For example, some researchers have argued that touch should be subdivided into distinct senses that are associated with sensing mechanical pressure, pain, and temperature (Ratcliffe, 2012). Rivlin and Gravelle (1984) go as far as to suggest that humans have around 17 distinct senses. Keeley (2002, 2009, 2015) examines the question of what makes something a sense. He finds three main theoretical approaches in the literature. The first individuates the senses by means of the characteristic experiences that they provide. By these lights, researchers can use the introspection-​ based techniques of psychophysics to map out the senses (Keeley, 2015). Right off the bat, this approach has some weaknesses. Clearly, it limits the range of senses to those we as humans can consciously experience. As a result, it faces what has been dubbed the other-​species-​of-​mind problem (Allen & Bekoff, 1997) because many organisms have senses that are foreign to our experience. For instance, sharks can sense magnetic fields (Kalmijn, 1982) and several species of fish can sense electricity (Keeley, 2000). Aside from the fact that this approach is ineffectual for the purposes of comparative neurophysiology and neuroethology, its reliance on introspection does not seem to be fine-​grained enough to settle important issues about human perception. Consider, for instance, the current debate surrounding the existence of a vomeronasal system in humans that detects the presence of pheromones (Meredith, 2001). Although we have no direct introspective awareness of pheromones, the functional status of this system seems to be a legitimate empirical question (Keeley, 2002). The second theoretical approach individuates the senses by means of their intentional objects. In other words, it individuates them by what they reveal about the world: colors, shapes, smells, textures, etc. On this approach, vision is the sense responsible for the perception of visual qualities, and audition is the sense responsible for the perception of auditory qualities. Although it relies less on introspection, this approach also faces significant challenges. The first concerns the metaphysical status of these objects. Whether or not secondary qualities like color are real features of the world is a matter of great philosophical debate (Chirimuuta, 2020; Cohen, 2009; Matthen, 2005). If colors are not real features of the world, then this approach threatens to be viciously circular—​defining color vision as the system that senses color and

Hierarchies and Hubs  67 color as that which is sensed by the color vision system. Another apparent weakness of this approach is that some of the proposed sensory objects, such as shape and texture, can be perceived by more than one modality. Thus, only certain qualities individuate the senses. Again, a noncircular justification is needed, and without one this approach remains in doubt. Furthermore, there is the question of how we individuate sensory objects in general. Synesthesia, a condition in which normally distinct sensory experiences are reliably and involuntarily linked with particular types of sensory stimuli such as seeing colors when hearing sounds or reading words, raises the prospect of novel multisensory objects (Keeley, 2015). The sensory object approach also appears to falter when it comes to individuating senses in neurotypical individuals. Consider the perception of flavor when eating food (Fulkerson, 2014). The experience of flavors appears to depend on aroma, taste, texture, and even temperature. Phenomenologically, flavors seem to be distinct sensory objects, and yet they seem to rely on input from multiple sensory modalities. The third approach treats the senses as neurophysiological organs (Keeley, 2002, 2015). This approach treats the senses as peripheral components of the nervous system that react to the presence or absence of physical changes in some part of the external world or in some internal tissue or organ. They are relatively self-​contained systems that at a minimum are responsible for transducing some particular form of physical energy into neural signals. Often, they can be characterized by a set of specific receptors that respond selectively to specific types of proximal stimuli. For example, the visual system processes the neural signals that—​under normal conditions—​emerge from the absorption of photons by the rods and cones located within the retina. The caveat with respect to normal conditions is required because these receptors may also be activated by means of other forms of stimulation. Gentle pressure on the eyeball of a sighted person with closed eyes will cause that person to perceive spots of lights referred to as phosphenes. A consequence of this flexibility is that, when individuating sensory systems, it is common to take into account their developmental and evolutionary history in order to determine their proper function. The sensory organ approach faces three immediate challenges. First, given the variety of human sensory receptors, this approach tends to offer a fine-​grained accounting of what a sense is. For instance, there are several distinct types of touch receptors that respond selectively to distinct types of stimuli. Do we want to identify distinct senses of touch? If so, does this

68  Abstract Concepts and the Embodied Mind cause us to miss important generalizations about touch construed more broadly? Second, the sensory organ approach tends to be permissive with respect to what counts as a sense (Noe, 2004). The bar is set quite low: all that is required is the existence of specialized receptors that provide the nervous system with information about the internal milieux or the external world. Third, the sensory organ approach is less clear when it comes to identifying which higher subcortical and cortical systems are sensory (Keeley, 2015). Why should supporters of grounded cognition worry about the problem of how to differentiate the senses? Well, first, grounded simulations are generally thought to consist of modality-​specific representations. For this claim to be testable, we need to have some idea of what is and is not an experiential modality. The trouble is that, at least with respect to the senses, the three approaches just outlined often give different answers concerning what is and is not a sense. Supporters of grounding may be tempted to adopt a provisional stance that ties grounding to whatever account of the sense turns out to be the most empirically robust. Rather than choose between the current candidates, they might defer to a future theory of the senses that overcomes the challenges facing the extant theories and provides a unified account of them. Unfortunately, such a theory may not be forthcoming. Indeed, Fulkerson (2014, p. 1) has recently defended a form of sensory pluralism that he describes as “the view there are many forms of sensory interaction and unity, and no single category that classifies them all.” Olfaction is a useful case study. On an intuitive level there appears to be a single sense of smell. Smells strike us as a single type of a sensory object. However, it is also possible to distinguish two separate olfactory systems. The first, the orthonasal system, responds to chemical stimulants from the external environment that enter through the nasal cavity. The second, the retronasal system, responds to chemical stimulants that arise from the mouth. Although both systems involve the same olfactory receptors contained within the olfactory epithelium, the experiences that arise from them are distinct: the orthonasal system helps us identify odors and their sources in the environment, and the retronasal system contributes to our experience of flavor. So, is there a single sense of smell? Are there two? The sensory pluralist suggests that both answers are right: that is, how we classify olfaction may depend in part on our explanatory purposes and multiple legitimate classification schemes are possible (Fulkerson, 2014).

Hierarchies and Hubs  69 Given the difficulties surrounding the effort to provide a satisfying means of individuating the senses, supporters of grounded cognition might adopt a more inclusive approach that leaves room for sensory pluralism. On this approach, we need only limit the sensory grounding to representations that are indigenous to systems that can reasonably be classified as sensory on any classification scheme that has proved useful. This sort of inclusiveness inoculates supporters from having to take a side in what is clearly an active controversy, but it also undermines the theoretical bite of the claim of modal specificity. Even this broadly inclusive approach may not be comprehensive enough to rescue the commitment to modal specificity, however. Consider our perception of space. Many theories of grounded cognition view spatial representations and processes as fundamental elements of grounded simulations. Indeed, spatial processing has often been presented as the key to understanding how abstract concepts might be embodied (Lakoff & Núñez, 2000; Tversky, 2019). And yet there are good reasons to doubt that spatial processing is contained within a specific sensory modality. Spatial processing is not handled by a specific sensory organ. While it is true that certain regions appear to process spatial information within the context of individual sensory systems such as audition, touch, or vision, many spatial processes are inherently multisensory. Some of the evidence supporting this can be found in single-​unit recordings of neurons in the monkey brain. For example, bimodal neurons that respond to visual-​tactile input have been implicated in the representation of extrapersonal space (Graziano, 2018; Graziano & Gross, 1994). The apparent importance of spatial processing to our concepts suggests that an appeal to sensory pluralism won’t fully rescue the commitment to modal specificity. Worse, the problem of multisensory processes is not limited to spatial processing. To see why, we need to consider the very source of the idea of modal specificity: a traditional view of the organization of the cortex in which early sensory areas are considered unimodal because they predominantly, if not exclusively, respond to input from a single sensory input (Mesulam, 1998). Multisensory areas are thought to be limited to intermediate and higher-​ order regions. Recent evidence has thrown this traditional view of cortical organization into question. Regions previously thought to be modality-​specific have been shown to respond to stimuli associated with different modalities. To give an example, a robust and diverse body of evidence shows that lower-​level visual

70  Abstract Concepts and the Embodied Mind cortex is multisensory in function (Ghazanfar & Schroeder, 2006; Murray et al., 2016). Auditory signals, for instance, have been shown to modulate activity in early visual areas (Iurilli et al., 2012). Admittedly, this influence may be the result of lateral and top-​down connections rather than the combination of feedforward audio and visual signals that occurs in other, nonprimary multisensory areas (Petro, Paton, & Muckli, 2017). Even so, this would not undermine the central point that visual cortex may be functionally multisensory. For instance, it is possible to induce a visual illusion through the presence of sound. When a single visual flash is accompanied by multiple beeps, participants incorrectly perceived multiple flashes (Shams, Kamitani, & Shimojo, 2000). In addition, multisensory phenomena are not limited to visual areas. For example, specific regions of the early auditory cortex have been found to respond to somatosensory stimuli (Foxe et al., 2002). Observations of cross-​modal plasticity also threaten the traditional conception of cortical organization. Evidence suggests that in cases of early damage to sensory pathways purportedly modality-​specific cortical areas may respond in a functionally similar ways to sensory input from atypical sensory pathways (Ricciardi et al., 2014). For example, studies of early blind individuals show that some occipital areas continue to process the same type of information in response to input from auditory and touch pathways (Amedi et al., 2017; Cecchetti et al., 2016; Striem-​Amit et al., 2012). Visual cortex associated with visuospatial discrimination has been shown to respond in texture perception in blind individuals (Sadato et al., 1996; Sathian & Stilla, 2010). Auditory cortex of early-​onset deaf participants has been found to respond in a more expanded fashion than that of hearing controls to vibrotactile stimulation (Auer et al., 2007). Building on the evidence of early multisensory areas and cross-​modal plasticity, Ghazanfar and Schroeder (2006) famously conjecture that the entire neocortex is multisensory. Whether or not this conjecture proves accurate, current evidence clearly suggests that early perceptual areas contain multisensory circuits, and seemingly “modality-​specific” areas can be cross-​modal in special populations. Because much of the evidence for conceptual grounding depends on finding activation in early perceptual areas, these observations undermine the case for the unimodality of conceptual representations (Calzavarini, 2021). Should this lead us to reject the very idea of grounded cognition? I suggest that this would be an overreaction. After all, the problem is not with the core idea that cognition relies in part on action, emotion, and perception

Hierarchies and Hubs  71 systems. Instead, it arises from a mischaracterization of the neurological mechanisms that underpin these systems. We have already seen in the previous chapters that it is common to identify a cline of theories of grounded cognition defined in terms of their relevant commitment to the importance of modality-​specific sensorimotor areas (Meteyard et al., 2012). Theories committed to strong grounding exclude all other areas from their view of grounding, and weaker theories leave room for the contribution of cross-​ modal and heteromodal areas. As I noted in earlier chapters, many criticisms of grounded cognition are targeted at strong forms of grounding despite the fact most current researchers support weaker forms. Part of the reason for the persistence of this sort of criticism may be the fact that supporters of grounded cognition still tend to talk in terms of modal specificity. This appears to be a terminological holdover from an anachronistic view of how our action and perception systems work. In the end, there are a number of inherent weaknesses associated with connecting the notion of grounding to the idea of modality specificity: not only is it the case that it is difficult to nail down just what is and what is not an experiential modality, but it is also true that many phenomena that have already been characterized as grounded rely on multisensory or cross-​modal processes. Strong grounding is simply implausible because it is incompatible with the way action and perception works. More importantly, weak grounding is not merely an ad hoc extension of the grounding thesis in response to troublesome data but is, instead, the version that best fits with our current understanding of experiential mechanisms.1 The terminology of modality specificity has become part of the origin story of grounded cognition. It is a shorthand than enables researchers to contrast their views with traditional views. This shorthand appears to have outlived its usefulness. As researchers have progressed to a more flexible, multimodal, and indeed hierarchical view of concepts, the notion of modal specificity is no longer central to the theories themselves. I propose that the notion of grounding should be decoupled from this notion and the outdated view of cortical organization upon which it rests. Instead, grounding should be tied to our dynamic experience of the world—​to action and perception 1 It is important to keep in mind that grounded cognition is not the product of a priori reasoning. Given that the movement to flexibility and hierarchy has been driven by the emergence of evidence that does not fit with early versions of grounding, we can reasonably worry that some aspects of our updated theories are ad hoc. Of course, this is a worry that should attach to any theoretical changes that emerge in response to recalcitrant data. The proper way to address this worry is to develop and test detailed theories that offer robust explanations and make clear predictions.

72  Abstract Concepts and the Embodied Mind understood in interactive terms. Our conception of grounding should incorporate elements of our physical niche and should encompass body–​ world dependencies. It should enable us to connect grounding to research involving other conceptions of 4E (Embodied, Embedded, Enactive, and Extended) cognition. From this perspective, the association of grounding with multisensory processing should not been seen as an addendum to the claim of embodiment; it should be seen as a fundamental design feature of the mechanisms by which we experience the world.

5.2  Hierarchical Organization Hickock accuses supporters of grounded cognition of offloading significant theoretical challenges by invoking simulation while adopting a black box approach to sensorimotor modalities. In keeping with this criticism, when we examine the neurological mechanisms involved in perceptual phenomena, we find a great deal of complexity, including multilevel circuits, top-​down pathways, and cross-​modal processes. Rather than simply treat this as counterevidence to grounding, we should look to our understanding of the workings of perception and action systems for guidance with respect to addressing important theoretical challenges facing grounded cognition. When we do this, our approach to grounding shifts. In particular, we begin to recognize the potential importance of hierarchical, multilevel representation. My argument in support of hierarchical representations unfolds in two steps. First, I briefly consider their theoretical promise both within and across modalities. Next, I review some recent evidence that supports their relevance to our concepts.

5.2.1  Opening the Black Box The myth of modal specificity exposes a theoretical weakness of “strong” versions of grounded cognition. The early focus on sensorimotor areas was likely driven in part by understandable pragmatic considerations. Experimental results indicating that semantic tasks could modulate activity in such areas were striking and clearly incompatible with the traditional approach to concepts. They led to a research program in which experimenters would look for evidence implicating these areas during conceptual tasks. This focus was always questionable, though. Once we open the black box of a

Hierarchies and Hubs  73 particular sensory modality, we find the same sort of complexity—​the same hierarchical structure characterized by complex feedforward, lateral, and top-​down dynamics—​that we find in other brain areas. Consider for example the visual system, the most studied and best understood sensory modality in the human brain. Even a cursory review of some of its basic elements reveals the inadequacy of the commitment to unimodality and the importance of hierarchical representation. Much of our visual perception depends on two major pathways that travel from our retinas through the structures in the lateral geniculate nucleus of the thalamus to the primary visual cortex (Milner & Goodale, 1995). One of these pathways, the ventral pathway, extends throughout the temporal lobe and is primarily associated with object recognition and identification, and the other, the dorsal pathway, extends throughout the parietal cortex and is associated with guiding movement. Importantly, these pathways are not fully segregated; there are significant interconnections between them. Vision scientists often distinguish lower-​level, intermediate-​level, and higher-​ level processing in the ventral pathway: lower-​ level visual processing involves features such as color, contrast, direction, and orientation; intermediate-​level visual processing is associated with aspects of contour integration, object motion, shape discrimination and surface properties; and higher-​level visual processing carries out the final stages of object identification. The core insight is that a functional hierarchy is supported by an anatomical one. As we ascend the anatomical hierarchy of the ventral stream, we find populations of neurons that exhibit both a selectivity to complex object-​ related properties and a tolerance of change or variation in context-​specific properties (Rust & DiCarlo, 2010). In other words, this functional hierarchy generates increasingly rich representations of the visual scene (Marr, 1982). An active debate persists over the precise role of top-​down circuits and the functional dynamics of particular regions (Gilbert & Li, 2013; Zipser, Lamme, & Schiller, 1996). Nevertheless, there are good reasons to think that higher-​level visual areas respond to increasingly abstract perceptual properties. The capacity to construct representations that are simultaneously complex and viewer invariant is a core design feature of the visual system (Rolls, 2012). A body of research exploring the nature of the receptive fields of cells in the anterior inferior cortex provides an indication of the sort of abstract representations that may be deployed in service of object recognition (Lehky & Tanaka, 2016). This research examines the types of stimuli that most activate the neurons within particular areas. Figure 5.1 displays

74  Abstract Concepts and the Embodied Mind

Figure 5.1  Examples of reductive determination of optimal features for 12 TE cells (Tanaka, 2003). The images to the left of the arrows represent the most effective object stimulus; those to the right represent the critical features determined by the reduction.

how researchers employ stimulus reduction to uncover the critical sensitivities of the inferotemporal cells in area TE (Tanaka, 2003). The most effective original stimuli are given to the left of the arrows, and the simplified stimulus that produces almost the same response is given to the right. This evidence suggests that inferotemporal neurons are sensitive to the presence of partial features of objects rather than objects as a whole. Objects are then represented in terms of populations of neurons capturing such higher-​level visual features. Researchers in artificial intelligence have created neurologically inspired models that are able to extract perceptual information from visual scenes. Some of the most promising of these involve so-​called deep learning networks that depend crucially on hierarchical representation (Goodfellow, Bengio, & Courville, 2016). LeCun, Bengio, and Hinton (2015, p. 436) offer the following broad summary of deep learning methods: Deep-​learning methods are representation-​learning methods with multiple levels of representation, obtained by composing simple but non-​linear

Hierarchies and Hubs  75 modules that each transform the representation at one level (starting with the raw input) into a representation at a higher, slightly more abstract level.

In keeping with this description, Buckner (2018) identifies a form of hierarchical processing in deep convolutional neural networks that he refers to as “transformational abstraction,” in which sensory-​based representations of category exemplars are iteratively converted into new formats that are more tolerant of variation and noise. These models serve as a demonstration of the possible importance of hierarchical representation to perceptual learning. We have just barely begun to sketch the complexities of the human visual system. Even so, this rudimentary picture is rich enough to supply prima facie reasons to question the assumptions of strong grounding. First, it shows that visual perception depends on the dynamic interaction of representations at multiple levels. Limiting our focus to an essentially flat conception of perceptual areas—​one where all that matters is that they are unimodal—​seems counterproductive and unprincipled. Second, many of the representations that would seem to be candidates for the grounding of our concepts—​and certainly those that are lexically encoded—​would likely be contained within hierarchically organized extrastriate areas. Consider, for example, object concepts. It is difficult to imagine how these concepts could be captured without relatively abstract inferotemporal representations. Third, a cogent argument could be made that we should divide vision into separate sensory modalities. For example, bilateral damage to the lateral occipitotemporal cortex can lead to a functionally specific deficit in the ability to process movement referred to as motion agnosia (Zihl & von Cramon, 1983; Zihl et al., 1991). I am not claiming that a lot is riding on the question of whether, say, color and motion vision should be considered separate modalities. Indeed, the fact that this is an unimportant question is my very point. Much of the same sort of hierarchically organized integration that occurs between modalities also occurs within modalities. The notions of neural reenactment or neural simulation provided a theoretical underpinning for understanding the grounding of concepts in action, emotion, and perception systems. Nothing about these notions, though, requires exclusively unimodal representations. Indeed, as discussed earlier, a wealth of evidence suggests that multisensory processes are important contributors to experiential processing. In the end, there are good reasons to think that they are central to our capacity to act on and respond to events in

76  Abstract Concepts and the Embodied Mind the world. Given this, they would seem to be central components of any reenactment or simulation.

5.2.2  Generalization and Grounding To some extent, the problem of generalization has been hiding in plain sight because of the way in which researchers typically study conceptual grounding. Many, if not most, experimental paradigms focus on the categorical question of whether concepts are grounded or not. This research typically involved behavioral experiments looking for sensorimotor facilitation or inhibition effects, neuroimaging experiments looking for the modulation of activity in sensorimotor areas, or neuropsychological experiments looking for conceptual deficits related to damage to sensorimotor areas. Far too often, it has failed to examine the ways in which concepts may differ with respect to their generality or degree of abstractness (Borghi & Binkofski, 2014). Many categories identified by lexical concepts, though, are hierarchically organized. A pug is a type of dog, which is type of mammal, which is a type of living thing. Pugs are also a type of brachycephalic breed. These hierarchically organized category memberships are used to draw inferences and to reason about individual pugs and pugs as a group. In other words, hierarchical relationships are a fundamental and pervasive design feature of our concepts. If grounding is going to serve as a robust theory of concepts rather than simply a useful counterweight to traditional amodal theories, then it is going to have to provide a meaningful account of how such hierarchies are neurologically encoded by means of experiential systems. The suggestion on offer here is that hierarchical representation may provide a solution to this problem and thus may prove to be an important component of grounding. I have already shown that hierarchical representations are thought to play a fundamental role within a single modality such as vision. In this section, I want to go further and suggest that hierarchical representations are likely to be important to multisensory processing. There are good reasons to think that cross-​modal and heteromodal representations are important for our capacity to perceive and act on the world. Graziano (2018, p. 56) makes this point clearly. Visual information by itself is pretty much useless. An image is projected onto the back of the eyes, but the eyes are constantly moving around, the

Hierarchies and Hubs  77 head is moving, the body is moving. It’s like putting a video camera on a bucking bronco and trying to analyze that wheeling, shaking image to figure out where anything is in the world.

Our experience of the world relies on more than transformations of a static image projected on the retina. Incoming visual information must be incorporated and integrated with information taken in by other sensory modalities and with ongoing actions taken by an individual within their physical environment. Not only must eye movement be considered, but so, too, must body position and the ways in which body movement transforms the incoming visual information (Gibson, 1986). Most contemporary versions of weak grounding hold that concepts rely on a hierarchy of neural circuits that extend from modality-​specific areas up to multimodal areas located within association cortices (Binder, 2016; Fernandino et al., 2016; Garagnani & Pulvermüller, 2016; Simmons & Barsalou, 2003). On these views, heteromodal convergence zones (Meyer & Damasio, 2009) or network hubs (van den Heuvel & Sporns, 2013) make important contributions to our concepts. This hierarchical structure may provide an explanation of how we are able to generalize or abstract away from experience.

5.3  Higher-​Level Representations The shift to a broader conception of grounding has several consequences. The first and most obvious consequence is that it raises the possibility that higher-​level sensorimotor areas are causally relevant to concepts. In this section, I outline some recent evidence that implicates such representations in our concepts. As I did when I outlined the case for grounding in general, I first consider perception and action systems separately for the sake of clarity and ease of exposition. Later in the chapter, I consider evidence that implicates hierarchical representations in generalization more broadly construed.

5.3.1  Perception Systems A growing body of research implicates higher-​level perceptual areas in the processing of some action concepts. In this section, I concentrate on research that focuses on a particular portion of the extrastriate cortex. This enables

78  Abstract Concepts and the Embodied Mind me to provide a more detailed review and highlight the particular issues that arise in response to such hypotheses. Kemmerer (2019) makes a compelling case for the importance of higher-​level visual features to the processing action concepts. He focuses on the possible contribution of a swath of visual areas known as the lateral occipitotemporal cortex (LOTC). This region encompasses the motion-​ selective human middle temporal cortex (hMT+​), which responds to a number of complex properties associated with visual movement (Lingnau & Downing, 2015; Wall et al., 2008),2 and the extrastriate body area (EBA), which responds selectively to images of human bodies and body parts when compared to other visual stimuli (Downing & Peelen, 2011). The hMT+​and the EBA project forward to other important subregions of the LOTC, the posterior superior temporal sulcus (pSTS) and the posterior middle temporal gyrus (pMTG; Kemmerer, 2022a). This region has been implicated in the perception, understanding, and even production of action. Figure 5.2 outlines the approximate borders of the LOTC and collates the observed activation peaks associated with various action, perception, and cognition tasks (Lingnau & Downing, 2015). The LOTC has also recently been found to respond to the perceptual components of social actions (Wurm & Caramazza, 2019). Several neuroimaging studies implicate portions of the LOTC in the processing of action concepts. Conceptual tasks involving action words have been found to activate the hMT+​and nearby areas (Kable, Lease-​ Spellmeyer, & Chatterjee, 2002; Kable et al., 2005). Further evidence implicating the LOTC comes from experiments that examine the processing of motion concepts in the context of stories. In one (Deen & McCarthy, 2010), participants in a functional magnetic resonance imaging (fMRI) scanner read brief stories about a single character. Half of the stories involved an incidental description of biological motion (such as walking or grasping) and half did not. The former activated the pSTS bilaterally and the latter did not. In another experiment (Wallentin et al., 2011), Danish participants heard a recording of the story, “The Ugly Duckling.” When clauses containing motion verbs were compared to the rest of the text, the left pMTG was found to be selectively activated.

2 Subregions of the hMT+​have been found to be also sensitive to auditory (Strnad et al., 2013) and tactile (van Kenemade et al., 2014) motion. We might therefore be better off viewing these subregions as multimodal motion perception areas.

Hierarchies and Hubs  79 (a)

(b) Key:

Basic motion Biological motion Tool viewing Body parts Hands Action observation Action imitation Action planning Actions crossmodal Action concepts Verbs

Figure 5.2  Visual representation of the approximate borders of the lateral occipitotemporal cortex (LOTC) and the observed activation peaks within it (Lingnau & Downing, 2015). (a) Outline of the LOTC that has been superimposed on the segmented and inflated left hemisphere of a single research participant. (b) Outline of the LOTC together with the activation peaks reported in several studies.

It remains possible that the responses found in these studies are byproducts of the relevant conceptual processing rather than components of it. Again, we can turn to neuropsychology for defeasible help in evaluating these competing hypotheses. Several studies suggest that the LOTC plays a causal role in action concepts (Aggujaro et al., 2006; Kalénine, Buxbaum, & Coslett, 2010; Kemmerer et al., 2012; Taylor et al., 2017; Tranel et al., 2003, 2008; Urgesi, Candidi, & Avenanti, 2014). To give an example, Kemmerer

80  Abstract Concepts and the Embodied Mind and colleagues (2012) gave a battery of six tasks involving action concepts to 226 brain-​damaged participants. One patient who had a highly focal lesion that was primarily confined to the left pMTG turned out to be one of the few who failed all six tasks. More broadly, the researchers found that group analyses of the patients implicated this region in all but one of the tasks. An upshot of this evidence is that we have good reason to think that the LOTC participates in conceptual processing. This serves as a demonstration that some higher-​level visual areas may indeed be causally relevant components of semantic memory (at least under certain conditions). The question then becomes whether the LOTC underwrites generalization in some important way. A collection of recent studies suggest that it does. Some reviews of the available brain imaging evidence infer that the LOTC might be involved in generalizing away from the perceptual details of specific actions (Oosterhof, Tipper, & Downing, 2012; Watson et al., 2013). Wurm and Lingnau (2015) sought to investigate the neural representation of actions at different levels of abstraction. Participants viewed three exemplars of eight actions, which involved the opening or closing of two distinct bottles and two distinct boxes. The researchers examined three levels of representation: a concrete level that distinguishes specific actions (e.g., opening or closing a specific bottle), an intermediate level that generalizes across movement kinematics (e.g., opening different bottles with cork or screw cap), and an abstract level that generalizes across objects (e.g., opening bottles or boxes). Region of interest and searchlight multivoxel pattern analysis (MVPA) indicate that the inferior parietal cortex and the LOTC encode actions at the intermediate and abstract levels. A similar study involving the observation two different types of actions (cutting and peeling) performed on two different types of objects (apples and potatoes) also found evidence that the LOTC encodes actions independently of the particular objects on which the actions are performed (Wurm et al., 2016). A third study finds evidence that concrete and abstract representations are organized along a posterior–​anterior gradient in the LOTC (Wurm, Caramazza, & Lingnau, 2017). The evidence just outlined is important for two reasons. First, it demonstrates how higher-​level perceptual areas can play a causally relevant role in our concepts. Second, it shows how such higher-​level representations may help address the problem of generalization and help encode abstract concepts.

Hierarchies and Hubs  81

5.3.2  Action Systems As I outlined in Chapter 3, some of the most compelling evidence for conceptual grounding involves the motor system. A common criticism of this evidence is that it often implicates premotor areas (e.g., Chatterjee, 2010; Mahon, 2015; Weiskopf, 2007). While this would pose a serious problem for many traditional views of grounding, it is compatible with more inclusive views.3 In this section, I draw a connection between this premotor activity and action generalization. The motor cortex is organized somatotopically: effectors for bodily movements are arranged in an inverted bodily map that extends in a vertical strip from the top of the brain downward. This organization was famously uncovered by a series of experiments on conscious preoperative patients in which brief, low-​intensity electrical stimulation of the motor cortex caused muscle twitches (Penfield & Boldrey, 1937). Using this technique repeatedly, Penfield and his collaborators were able to map out the somatotopic layout of the motor strip. Figure 5.3 provides a graphic representation of the body map, often referred to as a homunculus, that overlays this cortical area (Penfield & Rasmussen, 1950). The basic contours of this rough map have been confirmed by subsequent stimulation and neuroimaging experiments (Catani, 2017). One of the more striking features of the evidence for motor involvement in semantic memory is the fact that several studies find effector-​specific activation patterns. Verbs associated with leg/​foot actions, arm/​hand actions, and face/​mouth actions have been found to elicit peak activations in areas of the central motor cortex that correspond to those involved in producing those actions. Figure 5.4 provides a summary of the findings reported by several fMRI studies (Kemmerer et al., 2012). This figure shows that these studies recapitulate the rough body map initially outlined by Wilder and those who followed him, but it also reveals the degree to which premotor areas are implicated. This distributed pattern fits well with the hierarchical structure that I discussed earlier in the context of the visual system. A question remains though: What is the nature of the premotor representations that are implicated? 3 Chatterjee (2010) makes the intriguing claim that grounding should be thought of as a graded phenomenon. I consider this to be a variant of an inclusive “weak” approach.

es

Li t Ri tle M ng id dl e

To

Ha nd

Trunk Hip ee e Kn nkl A

Shoulder Elbow Wris t

82  Abstract Concepts and the Embodied Mind

x de In mb u Th eck N w Broball e e d y d an i l Face e y

ti o Sal n iv a t

Jaw Sw Tong u all ow e ing

io n

Lips

Vocalization

E

s ti

ca

Ma

1948

Figure 5.3  The motor homunculus (Penfield & Rasmussen, 1950).

A recent body of evidence suggests that the traditional emphasis on somatotopic organization might have been a bit misplaced. Fittingly, this evidence emerged from a series of stimulation experiments involving monkeys that were very much in line with Wilder’s original experiments. In an autobiographical discussion of this research, Graziano (2018) explains how these experiments, which explore the consequences of longer, more intense stimulations, were the result of the accidental use of a stimulation protocol normally reserved for studying eye movements on the motor cortex. Rather than eliciting a mere muscle twitch, the longer stimulation elicited a coherent reach action. Building on this serendipitous mistake, Graziano’s lab carried out many years of experiments (for a review, see Graziano, 2008). These experiments found that the motor and premotor cortices were organized in

Hierarchies and Hubs  83 Premotor Cortex Primary Motor Cortex Leg/Foot Actions Arm/Hand Actions Mouth Actions

Figure 5.4  Activation peaks in left primary motor and premotor cortices reported by some of the functional magnetic resonance imaging (fMRI) studies that have probed the neural substrates of the motor features of verbs and sentences encoding leg/​foot actions, arm/​hand actions, and mouth actions (Kemmerer et al., 2012).

a series of zones associated with coordinated actions. Figure 5.5 provides a sketch of the results of these experiments (Graziano & Aflalo, 2007). This figure suggests that much of the motor system is not organized in terms of a map of the body, but rather as a map of actions. In keeping with this, some have suggested that while some portions of the motor cortex may exhibit somatotopy, others might exhibit actotopy (Fernandino & Iacoboni, 2010; Kemmerer & Gonzalez-​Castillo, 2010). All of this suggests that premotor areas may encode aspects of actions that generalize away from the various motoric particulars associated with realized actions. A study by Umiltà and colleagues (2008) provides further evidence of abstraction. The researchers compared the responses of neurons in the premotor areas F1 and F5 in monkeys who were trained to pick up objects either using normal pliers (which require the monkey to close its hand) or using reverse pliers (which require the monkey to open its hand). Many neurons in F5 and some in F1 responded similarly under both conditions. This suggests that their activity was modulated by abstract action properties

84  Abstract Concepts and the Embodied Mind Climbing/leaping Hand in lower space

Reach to grasp

Manipulate in central space

Defense

Chewing/ licking

Hand to mouth

Figure 5.5  Action zones in the motor cortex of the monkey (Graziano & Aflalo, 2007).

such as the goal of the action (picking up the object) rather than the specific hand movements required to achieve that goal. Additional evidence of the importance of premotor representations for our concepts is provided by research on so-​called mirror neurons. Mirror neurons are neurons that fire both when an individual executes an action and when they observe another carrying out that same action. They were initially found in area F5 of the Macaque monkey ventral premotor cortex (Di Pellegrino et al., 1992; Gallese et al., 1996; Rizzolati et al., 1996) but have also been found in the dorsal premotor cortex (Cisek & Kalaska, 2004; Raos, Evangeliou, & Savaki, 2007), the inferior parietal cortex (Fogassi et al., 2005; Gallese et al., 2002; Evangeliou et al., 2009), and even the primary motor cortex (Raos, Evangeliou, & Savaki, 2004, 2007; Tkach, Reimer, & Hatsopoulos, 2007). For the moment, though, let’s focus on mirror populations of the premotor cortex contained within area F5. This area has

Hierarchies and Hubs  85 been found to encode actions carried out with hands and the mouth (Ferrari et al., 2003; Kurata & Tanji, 1986). It contains two major types of visuomotor neurons associated with grasping actions: canonical neurons that respond to the observation of objects and mirror neurons that respond to the observation of actions (Bonini et al., 2014; Rizzolatti & Fogassi, 2014). While the former are thought to help guide motor actions, the latter are thought by many to be involved in action understanding. The discovery of mirror mechanisms in the ventral premotor cortex of the monkey played a formative and influential role in the development of embodied cognition. Mirror neurons provide an explanation of how motor simulations might be employed in action understanding (Gallese, 2009; Rizzolatti & Fogassi, 2014). The idea is straightforward enough: the automatic activation of an internal motor representation of an action by the observation of that action helps render that action comprehensible and meaningful. Motor simulation contributes to understanding. The literature on the human mirror system is large and contains a great deal of intriguing yet ambitious speculation. Evidence suggests that multiple task-​and context-​dependent factors modulate the response of the mirror neuron system during action observation (Kemmerer, 2021). To address the functioning of the mirror neuron system properly would require a whole other book.4 Rather than offer a review of this literature, I focus on the evidence supporting the claim that this system involves various types of generalization or abstraction. Before outlining this evidence, I offer two caveats. First, I consider mirror neurons to be just one of the mechanisms by which grounding occurs. They are an important part of the grounding story of social cognition, but they are not the whole story. Second, I remain committed to a broad notion of grounding that includes action, emotion, and perception systems. Given this, it seems likely that neurons with mirror-​like properties exist in several experiential regions of the cortex (Caspers et al., 2010; Kilner & Lemon, 2013). Careful and deliberate single-​cell recordings of ventral premotor mirror neurons have produced strong evidence that they are able to encode abstract features of actions. For example, some F5 mirror neurons fire in response to specific types of actions (e.g., precision grip vs. whole hand) without responding to specific features of those actions (e.g., grip size), while others respond to more than one type of grasping action (Di Pellegrino et al., 4 A recent PubMed search using the phrase “mirror neurons” generated more than 1,900 results.

86  Abstract Concepts and the Embodied Mind 1992; Gallese et al., 1996; Rizzolati et al., 1996). Additionally, a subset of F5 mirror neurons have been found to respond during the observation of hand actions when the whole action is visible and when the final part of the action is visually obscured (Umiltà et al., 2001). This is often taken to be evidence that these neurons encode the goal-​state of the action. Some F5 mirror neurons respond both when the monkey sees an action and when it hears a sound that is typically produced by that action (Kohler et al., 2002). Such auditory-​visual mirror neurons would seem to represent the relevant action at a fairly high level of abstraction. A group of F5 mirror neurons have been found to respond preferentially to actions that take place in the monkey’s extrapersonal space, while another group has been found to respond preferentially to actions that take place in their peripersonal space (Caggiano et al., 2009). Again, this sort of space-​selective coding would seem to indicate a fair degree of abstraction. One recent study provides clear evidence of abstraction (Caggiano et al., 2016). The researchers measured the responses of mirror neurons located in F5 to observations of naturalistic stimuli depicting hand actions and abstract stimuli depicting the same causal relationships. The naturalistic stimuli consisted of filmed hand–​object actions, and the abstract stimuli consisted of brief videos of interacting discs. The spatiotemporal similarity of the responses under these conditions suggests a “high degree of generalization between these two stimulus classes” (p. 3077). The researchers were able to reduce the responses to the abstract stimuli by introducing manipulations known to eliminate the perception of causality, such as changing the trajectories of the actions or introducing spatial gaps. The authors propose that their results support the inference that these mirror neurons are involved in the perception of causality. I have argued that a misguided view of the nature of sensorimotor systems has led supporters and critics of grounded cognition alike to overemphasize the importance of modality-​specific representations. This emphasis can be clearly seen in the criticism of mirror neuron research offered by Caramazza and colleagues (2014, p. 11). They summarize their position as follows (citations in the original): “in the field of action understanding, studies on mirror neurons have shown that areas thought to carry relatively low-​level representations contain neurons that show surprisingly high levels of abstraction (Caggiano et al., 2011, 2012; Ferrari et al., 2005; Gallese et al., 1996; Umlità et al., 2001, 2008) that, we argue, cannot plausibly be considered to be modal.” In other words, the fact that the premotor neurons in question encode

Hierarchies and Hubs  87 abstract action properties rather than low-​level motor properties excludes them from being considered modality-​specific representations. Caramazza and colleagues concede that action observation and understanding often involve hierarchically organized representations and multimodal processing, but they maintain that this fits better with a modified amodal approach to concepts that allows for some hybrid (i.e., amodal/​sensorimotor) processing than it does with an approach that appeals to neural simulation. In contrast to the assumptions underlying this criticism, hierarchical models of the motor system have played a role in the understanding of the mirror neuron system for quite some time (e.g., Grafton & Hamilton, 2007). It is thus possible to offer an approach to conceptual grounding that relies on neural simulation but nevertheless predicts that hierarchical and multimodal processing should underwrite our capacity to conceptualize abstract action properties. The question then becomes whether this updated notion of grounding provides a richer account of concepts than does the sort of hybrid approach offered by defenders of amodal cognition. I suggest that the diverse and robust evidence of the role played by different populations of neurons with mirror properties in different areas of the cortex provides support for a rich and active causal role for simulation in semantic memory (Kemmerer, 2015a). Indeed, this proposal offers a potential solution to the problem of generalization by giving us access to abstract representations that are grounded in sensorimotor systems. The evidence implicating mirror neurons in action understanding remains controversial because it has been offered as an explanation for many aspects of social cognition (Gallese, 2009; Rizzolatti & Fogassi, 2014). Putting these claims aside for future investigation, we can treat mirroring phenomena as being one of the ways in which cognition is grounded in hierarchically organized and multimodal sensorimotor systems. This leads to the prediction that such phenomena are likely to be common and widespread—​something supported by a growing body of research (Caspers et al., 2010). Returning to the evidence involving the mirror neurons found in ventral premotor cortex, it is now possible to acknowledge a specific functional link between their capacity to represent general features of actions and their relevance to action understanding. This relevance remains unexplained by traditional amodal approaches. Even hybrid approaches do little to offer a positive account of why mirroring is so pervasive and well-​attested. Accounts that turn on a notion of neural simulation or reenactment, though, provide a compelling explanation of the functional role of mirror systems because the very fact that

88  Abstract Concepts and the Embodied Mind these premotor neurons retain their connection to action is important for understanding their capacity to encode aspects of action concepts. A further prediction of our updated approach to grounding is that different populations of mirror neurons should encode actions at different levels of abstraction (Kemmerer & Gonzalez-​Castillo, 2010). Again, some recent evidence supports this prediction: while many mirror neurons in ventral premotor cortex have been found to encode view-​dependent aspects of actions (Caggiano et al., 2011), a study of the human mirror neuron system reveals that posterior areas in the parietal and occipitotemporal cortex contain cross-​modal representations that encode view-​independent action features (Oosterhof et al., 2012). A broader notion of sensorimotor grounding—​one that encompasses multimodal processing—​enables us to formulate a compelling simulation-​based approach that includes action representations at different levels of generality.

5.4  Hubs and Spokes Generalization involves abstracting away from the particulars of our everyday experience. It lies at the heart of even concrete concepts. We have seen that positing hierarchical representations can provide some explanation for our capacity to generalize. Are they enough? Certainly, this is an empirical issue, but amodal representations appear to have a great deal of theoretical promise with respect to representing semantic contents that abstract away from the details of experience. Against this background, it is not surprising that there has been an active discussion concerning whether concepts might depend, at least in part, on amodal representations (Dove, 2009, 2011; Shallice & Cooper, 2013).

5.4.1  The Anterior Temporal Lobes Some of the strongest empirical evidence for amodal conceptual representations involves the anterior temporal lobes (ATL). The ATL have been implicated in the representation of semantic knowledge using fMRI (Binney et al., 2010; Coutanche & Thompson-​Schill, 2015; Visser et al., 2012) and magnetoencephalography (MEG) (Marinkovic et al., 2003; van Akeren et al., 2014). Striking evidence for the involvement of the ATL in our

Hierarchies and Hubs  89 concepts can be found in research on the syndrome of semantic dementia (SD). SD is characterized by an atrophy of the ATL and a progressive impairment of conceptual knowledge (Patterson, Nestor, & Rogers, 2007). Because this impairment cuts across modalities and involves both linguistic and nonlinguistic semantic knowledge, some researchers propose that the ATL serves as a semantic “hub” (Sporns, Honey, & Kotter, 2007) containing conceptual representations that arise from the integration of information from modality-​specific areas (Lambon Ralph et al., 2010; Patterson et al., 2007). Further support for this idea is provided by studies that elicit a generalized slowing of conceptual processing through the previous application of repetitive transcranial magnetic stimulation (rTMS) to areas within the ATL (Pobric, Jefferies, & Lambon Ralph, 2007, 2010; Pobric, Lambon Ralph, & Jefferies, 2009). Critics of grounded cognition often point to the research on SD as providing support for an amodal approach to concepts (e.g., Mahon, 2015; McCaffrey & Machery, 2012). These arguments have two primary weaknesses: first, they generally assume a very strong form of embodiment that requires full grounding in sensorimotor areas. While such an industrial strength form of embodiment is falsified by the mere presence of any amodal representations in our concepts, inclusive versions are not. Second, these arguments fail to recognize that there are two competing accounts of the role played by the ATL: the hub-​only view and the hub-​and-​spoke view (Pobric et al., 2010). On the hub-​only view (which fits well with the amodal approach advocated by the critics of embodiment), the ATL is the primary locus of conceptual representation and sensorimotor areas are excluded from our concepts. On the hub-​and-​spoke view, the ATL and modality-​specific regions both contribute to conceptual representation. Clearly, the hub-​and-​spoke view fits best with the evidence outlined in this chapter and previous chapters that implicates modality-​specific experiential systems in our concepts. Somewhat ironically, abstract concepts themselves pose a specific challenge for hub-​and-​spoke theories of the ATL. Given the proposed functional role of amodal hubs as sources of generalization, one would expect that they would play a central role in the processing of abstract concepts. Unfortunately, while several studies have found that the ATL respond to semantic processing in concrete concepts (Marinkovic et al., 2003; Spitsyna et al., 2006; Visser, Jeffferies, & Lambon Ralph, 2010), similar results with abstract concepts have been elusive (for a discussion, see Hoffman et al., 2015). The inconsistent imaging data have led some researchers to question

90  Abstract Concepts and the Embodied Mind whether the ATL play an important role abstract concepts (Bonner et al., 2009; Shallice & Cooper, 2013). Some recent evidence suggests that there may be a way out of this theoretical bind. A distortion-​corrected fMRI study found a concreteness effect in the ATL, with preferential activation for concrete words in ventromedial portions of this area and preferential activation for abstract concepts in dorsolateral portions (Hoffman, Binney, & Lambon Ralph, 2015). This is consistent with the idea that different portions of the ATL might serve more specialized functions, with the ventromedial regions more closely linked to visual experiences and dorsolateral regions being more closely linked to auditory-​verbal experiences and (Binney, Parker, & Lambon Ralph, 2012; Hung et al., 2020; Rice et al., 2015). Hoffman (2016) proposes that there is a ventral-​to-​dorsal gradient along the left ATL that responds to increasing degrees of abstractness (see also Hoffman & Lambon Ralph, 2018). A study of congenitally blind participants found evidence of such a gradient and that both abstract words that lack clear sensorimotor features (e.g., freedom) and words that depict predominately visual phenomena that could not be directly experienced by the participants (e.g., rainbow) elicited preferential activity in the left dorsal superior ATL (Striem-​Amit et al., 2018). Amodal representations seem well-​suited to encoding abstract content. Indeed, their apparent functionality in this regard is one of the reasons for the popularity of amodal approaches to cognition (Mahon & Caramazza, 2008). Once we abandon the all too common a priori commitment to strong embodiment, we can look to empirical evidence to see whether transmodal or amodal representations also make important contributions. When we do this, it is not hard to find such evidence. Neuropsychological case studies involving individuals with SD provide some of the strongest support for amodal representations. Admittedly, the anatomical characteristics of the ATL create some challenges for neuroimaging techniques, and the results have been somewhat equivocal (Hoffman, Binney, & Lambon Ralph, 2015). While some of the recent evidence is suggestive, clearly more research is needed. What remains interesting, though, is that the emerging picture suggests that amodal representations contained within the ATL may serve to bind together the various grounded properties associated with individual concepts. Research on the SD thus provides further support for a broad and inclusive approach to grounding that allows for representational pluralism.

Hierarchies and Hubs  91

5.4.2  Multiple Hubs The ATLs are not the only heteromodal or supramodal areas that have been associated with conceptual processing. A meta-​analysis of 120 fMRI studies implicates a left-​lateralized semantic system that includes the posterior inferior parietal lobe (the anterior gyrus and portions of the supramarginal gyrus), lateral temporal cortex (the middle temporal gyrus and portions of the inferior temporal gyrus), ventral temporal cortex (the mid-​fusiform and parahippocampal gyri), dorsomedial prefrontal cortex, inferior frontal gyrus, ventromedial prefrontal cortex, and posterior cingulate gyrus (Binder et al., 2009; see also Binder et al., 2016). Given the wide distribution of these areas, it seems likely that multiple hubs contribute to semantic memory. Two prominent theories explicitly adopt a multiple hub approach. The first is the embodied abstraction theory (Binder & Desai, 2011). According to this theory, concepts are made up of experiential representations at multiple levels of abstraction. These levels are activated in a flexible and task-​sensitive manner. The highest levels contain schematic representations that are highly abstracted from experiential particulars. Intermediate and lower-​level representations are engaged in an as-​needed fashion: that is, with certain tasks, the schematic representations may be sufficient, but when deeper processing is required, more fine-​grained experiential simulations are likely to be engaged. This theory posits more complex structure than a simple hybrid account because it envisions (p. 532) “an interactive continuum of hierarchically ordered neural assemblies, supporting progressively more combinatorial and idealized representations.” Modal convergence zones (Damasio, 1989) contained in intermediate levels converge on several distinct higher-​ level hubs located in the inferior parietal cortex and much of the ventral and lateral temporal cortex. These supramodal hubs bind representations from multiple modalities and encode similarity structures associated with categories. The other prominent multi-​hub theory is the dynamic multilevel reactivation framework (Reilly & Peele, 2008; Reilly et al., 2014, 2016). This framework hypothesizes that our concepts are encoded by a system of low-​and high-​order hubs. The low-​order hubs are heteromodal in the sense that they bind together different sensory and motor features. In other words, they carry out a process of sensorimotor feature integration (Bonner et al., 2013).

92  Abstract Concepts and the Embodied Mind Ultimately, they converge on high-​order hubs in the temporal pole. Reilly and colleagues explain (2016, pp. 1006–​1007): Activity within low-​order hubs can be characterized as heteromodal in that sensory features are bound within these regions. . . . We hypothesize that high-​order hubs situated primarily within the anterolateral temporal lobes conduct symbolic transformations upon these bound representations. During the transformation process, conceptual knowledge is abstracted from its sensorimotor roots via a series of successive processing stages whereby perceptual and linguistic knowledge ultimate converge (sensory → heteromodal → amodal).

The amodal representations contained within the anterolateral lobes require enrichment through sensorimotor simulations that are engaged in accordance with task demands. Both these theories are variants of the general approach advocated in these pages. They combine representational pluralism with a commitment to a flexible multimodal and multilevel conceptual system. They differ from each other, though, with respect to important details. For instance, they differ with respect to the function of hubs in the anterior gyrus and the temporal pole. The embodied abstraction theory denies that a central hub exists in the temporal pole and holds that an independent higher-​level hub exists in or near to the angular gyrus. The dynamic multilevel reactivation framework, on the other hand, situates a central hub in the temporal pole and sees the angular gyrus as the location of a low-​order heteromodal hub or hubs. Both models are consistent with the observation that the AG acts as complex cross-​modal hub, one which integrates multimodal information during semantic comprehension and reasoning (Seghier, 2013). While there is a palpable need for further research, we have good reasons to suspect that our concepts rely on the dynamic interaction between lower-​level and higher-​level hubs. A recent meta-​analysis of 33 functional neuroimaging studies examined the activations of abstract and concrete concepts (Del Maschio et al., 2021). This study found that (i) both types of concepts activated temporo-​ fronto-​parietal circuits often associated with sensorimotor experiences, (ii) concrete concepts activated fronto-​ parietal areas better than did abstract concepts, and (iii) abstract concepts activated Broca’s area more than did concrete concepts. This evidence fits with a multilevel approach that sees language as one source of grounding among many (Dove, 2014, 2018, 2019).

Hierarchies and Hubs  93

5.4.3  Semantic Pointers I have been defending a version of representational pluralism that invokes hierarchically organized representations. Such representations fit with neuroimaging data and neuropsychological case studies and offer a possible solution to the problem of generalization. In this subsection, I consider a theoretical framework for understanding how concepts might rely on hierarchical representations. This framework is based on the idea that representation and computation should be understood in terms of the activation patterns of large populations of neurons and seeks to explain the functioning of concepts in terms of semantic pointers (Blouw et al., 2016; Eliasmith, 2013). Semantic pointers are a consequence of hierarchical organization. They are compressed higher-​level representations that carry information about certain domains. Consider our perception of a visual object. Initially, the object will be encoded by a very large population of neurons. This information becomes transformed and compressed as it is passed along through a series of layers of neural populations of decreasing size. Each successive layer generates increasingly abstract statistical summaries of the information contained within the image. Semantic pointers are the highly compressed representations at the top of such hierarchies. It is important to keep in mind that the information flow can be bidirectional: the activity in lower-​level populations can be recovered through downward processes of decompression from the relevant semantic pointer. Structured representations containing information from different experiential systems can be achieved by binding together collections of semantic pointers. These higher-​level representations themselves are semantic pointers. Nevertheless, Blouw and colleagues (2016) warn against viewing semantic pointers as amodal symbols that capture concepts. In keeping with the idea that concepts are something we do with the mind, they suggest that we should understand concepts as dispositions. Having a concept is having the ability to activate appropriate sensorimotor simulations and expressions in natural language to carry out cognitive tasks. Different tasks will activate different neural populations. Figure 5.6 provides a graphic representation of the different sorts of activations and transformations that might occur in relation to the concept dog. While I am not fully committed to the semantic pointer framework, it serves as a useful proof of concept. This framework has several important features. First, it shows how generalization can arise in a hierarchical and

94  Abstract Concepts and the Embodied Mind Linguistic Inference Conceptual SP

“Has incisors” “Can bite”

“Is a vertabrae” “Is a mammal”

DOG

“Has teeth”

“Breathes”

Visual SP

Tactile SP

Auditory SP

Neural pop. in hierarchy Neural transformation

BITE

Neural representation

SP Semantic Pointer

Related Concept

Definition of Symbols

Perceptual Simulation

Figure 5.6  Simplified diagram of possible transformations of a semantic pointer for the concept dog. Transforming the semantic pointer can result in perceptual simulation, linguistic inferences, and consideration of related concepts (Blouw et al., 2016).

multimodal conceptual system. Second, it shows that this can be achieved without identifying concepts with amodal representations. Third, it provides an account of the causal relevance of multimodal sensorimotor simulations. Finally, it does all of this within the context of a neurobiologically plausible computational framework.

5.5  Multimodal and Multilevel Representations In this chapter, I examined evidence implicating specific heteromodal, supramodal, and even amodal areas in abstract concepts. This evidence fits poorly with versions of embodiment that limit grounding to modality-​ specific sensorimotor areas. More importantly, it suggests that higher-​level representations may help us generalize and abstract away from experience. Given this, we need ask how more general concepts might differ in terms of their overall representation from more specific ones. A recent study comparing the neural representations of intermediate basic-​ level concepts and more specific subordinate-​level concepts provides some indication of what these differences might be (Bauer & Just, 2017). Categorization

Hierarchies and Hubs  95 experiments have shown that participants are more likely to respond to an image of a robin with the basic-​level term bird than with either the more specific term robin or the more general term animal (Rosch et al., 1976; Tversky & Hemenway, 1984). There has been a great deal of discussion in the literature concerning the possible reasons for the processing advantage associated with basic-​level concepts (Murphy & Wisniewski, 1989). Significantly, this processing advantage can be modulated by experience. For example, people from traditional societies who possess greater ethnobiological knowledge than those from Western, educated, industrialized, rich, and democratic societies (so-​ called WEIRD societies; Henrich, 2020) tend to categorize species at more subordinate levels (Kemmerer, 2019). What is important for my purposes, though, is that basic-​level concepts are situated in an intermediate level of a conceptual hierarchy defined relative to abstractness (Murphy & Lassaline, 1997). As such, they seem likely to combine concrete and abstract elements (Goldstone, Feng, & Rogosky, 2005). Some previous brain imaging evidence suggests that naming pictures of objects at the subordinate level elicits greater activation in sensorimotor areas (Gauthier et al., 1997; Kosslyn, Alpert, & Thompson, 1995; Rogers et al., 2006). Superordinate concepts often elicit greater activation in higher-​level areas central to the language system, such as the left inferior frontal gyrus (Binder et al., 2005; Wang et al., 2010; Wang, Baucom, & Shinkavera, 2013). It is important to keep in mind that this difference is a matter of degree and that there are good reasons to think that superordinate-​level concepts rely in part on grounded representations associated with category exemplars (De Wilde et al., 2003; Heit & Barsalou, 1996). Bauer and Just (2017) used MVPA in a fMRI study to compare the neural representations of basic-​level and subordinate-​level concepts. They report several intriguing findings: the first was that basic-​level concepts were encoded by a combination of sensorimotor areas, heteromodal areas (temporal and parietal), and frontal and temporal language areas,5 whereas subordinate concepts tended to rely more heavily on sensorimotor areas. In other words, this study links basic-​level concepts with both sensorimotor grounding and areas that have been associated with the processing of abstract semantic content (i.e., heteromodal and language areas). The second finding was that subordinate-​level concepts elicited greater activation in the anterior temporal areas thought to be critical for binding the 5 I discuss the importance of language areas in Chapter 6.

96  Abstract Concepts and the Embodied Mind various sensorimotor features associated with a category. The third finding was that the neural representation of basic-​level concepts (e.g., bird) was more like the neural representation of subordinate-​level concepts associated with typical exemplars (e.g., robin) than it was with subordinate-​level concepts associated with atypical exemplars (e.g., woodpecker). This fits well with the notion that sensorimotor representations play a significant role in basic-​level concepts. Even though some prominent researchers have recently adopted a multimodal and multilevel approach to grounded cognition, critics of grounding often see strong embodiment as their primary target. The perception that so-​called weak embodiment represents a potentially ad hoc retreat from a purer embodiment thesis may be unintentionally reinforced by the tendency of advocates of embodied and grounded cognition to characterize their position in terms of modal specificity. In contrast, I have defended a multimodal and multilevel view of grounding. First, I have shown that the attempt to define embodiment in terms of modal specificity is counterproductive: not only do grounded theories regularly appeal to multisensory representations such as those involved in taste and spatial processing, but many areas of the cortex that have traditionally been thought to be unimodal also have been revealed to be multimodal. Second, I have argued that hierarchical representations hold some promise with respect to addressing the problem of generalization. Recent evidence suggests that higher-​level perceptual and motor representations help encode our concepts. This shows that multilevel accounts of grounding enjoy advantages over ones that rely entirely on modality-​specific sensorimotor areas. Third, I have argued that there are reasons to think that heteromodal, supramodal, or amodal representations may help bind together sensorimotor features associated with our concepts. Finally, I have argued that the evidence differentiating the neural representation of basic-​level and subordinate-​level concepts brings each of these elements together.

6 Language Is a Neuroenhancement In the preceding chapter, I outlined how hierarchical and multimodal representations may help us generalize away from the particulars of our everyday experiences. Our concepts leverage the hierarchical structure of action, emotion, and perception systems. Is such a system powerful enough to handle abstract concepts across the board, though? In other words, does moving to a more inclusive conception of grounding—​one that recognizes the importance of higher-​level representations—​provide a full solution to the challenge posed by abstract concepts? As I discussed in Chapter 4, there are good reasons to think that it does not. In particular, the problem of disembodiment arises because grounded mechanisms seem insufficient to explain our capacity to employ concepts that are distantly tied to affective, motor, and perceptual experience. Grounded cognition posits an intimate connection between how concepts are acquired and how they are represented in the brain. Given that we learn concepts through both direct sensorimotor experience and language, we might expect some conceptual content to be encoded in representations indigenous to the language system. My aim in this chapter is to defend the idea that semantic representations include linguistic information. Within cognitive science, theoretical positions concerning the role of language in our thoughts have traditionally been polemical: they have tended to treat language either as having almost no direct influence on our conceptual system, serving primarily as a means of gaining access to culturally derived information, or as having almost magical powers, underwriting our capacity for abstract thought. To put it less colorfully, theories have tended to favor either communicative or cognitive conceptions of the role of language in thought. Early advocates of grounded cognition generally favored communicative views (Borghi, 2020). More recently, though, there has been growing interest in the idea that language might play a significant cognitive role in semantic memory (Dove, 2018). Here, I defend a view of the contribution made by language to our concepts that is intimately connected to grounded cognition. I propose that the widely distributed language system amounts to Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0006

98  Abstract Concepts and the Embodied Mind a neurologically realized neuroenhancement that supports new forms of intelligent behavior. Communicating through language—​ even at the most rudimentary level—​enables children to act on the world and gather information about it in a powerful and effective way. At a minimum, it provides access to a highly effective channel of cultural transmission. While becoming fluent, the child learns how to manipulate grounded symbols in a systematic and productive fashion. This linguistic competence enhances cognition by means of providing a distinctly effective medium of thought. It is a symbol technology that enhances the way our brain functions. The language system is a useful piece of neural/​cognitive hardware that extends our conceptual reach by enabling us to formulate thoughts that we would struggle to formulate otherwise. Over time, children become linguistic cyborgs (Clark, 2003). Language is both an external cognitive scaffold and an internal neuroenhancement. When an individual acquires a natural language, they acquire a symbol system that has different computational properties than the embodied codes that exist independently of language (Clark, 2006, 2008). In other words, the distributed and grounded language system transforms our ability, not only to glean information from our external and social environment, but also to formulate thoughts about the world and our place in it. Conceptual content is captured in part by the associative or inferential relationships of partially grounded linguistic representations with other partially grounded linguistic representations (Borghi & Cimatti, 2009; Dove, 2011; Dove et al., 2020). A concept will not only be represented on a given occasion by multimodal simulations associated with interacting with its referents, but also by multimodal simulations and action schemas associated with talking or listening to talk about them—​that is, experiences of words, phrases, and conversations. The neurologically realized language system is an important subcomponent of a flexible, multimodal, and multilevel conceptual system.

6.1  A Symbolic Medium To see how language enhances our capacity for abstract concepts—​beyond providing access to socially derived knowledge—​we need to understand its structural properties. These structural properties amount to a suite of design

Language Is a Neuroenhancement  99 features that enable language to encode abstract content that goes beyond our direct experience. Although I do not have the space to discuss these in detail, I hope that a brief sketch will allow me to highlight core symbolic properties of language. Linguists have identified and studied the structural properties of language at several levels of analysis. For example, natural languages tend to have circumscribed speech sound (or sign) inventories consisting of distinct basic elements (phonemes) that can be combined to form larger constituents. The recognition of these basic elements often depends on categorical perception (in which a continuous acoustic or visual dimension is perceived in a discontinuous manner; Goldstone & Hendrickson, 2009). Our capacity to produce and categorize them depends on rich implicit knowledge. For example, English speakers implicitly know to aspirate the consonant /​p/​(i.e., produce a strong burst of breath at the end of the articulatory gesture) when saying the word pit but not when saying the word spit. They have also learned to ignore the presence/​absence of this aspiration when categorizing these sounds; that is, they are implicitly able to treat the aspirated and unaspirated forms as allophones. Speakers of Hindi, though, recognize aspiration as a contrastive distinction (Ohala & Ohala, 1972). This divergence shows that infants are not only born with the capacity to become attuned to this sound difference but also to ignore it in the context of speech perception. The phonemes of a natural language can be combined to form larger constituents referred to as morphemes. These include word stems as well as the affixes that modify their meaning. English is not a morphologically rich language, but it does contain prefixes such as the modifiers un-​and anti-​ and suffixes such as the past tense -​ed and the plural -​s. Arguably, English slang also contains a rare group of infixes that are used for dramatic and colorful effect. Consider the insertion of expletives into other words such as abso-​ fucking-​lutely and fan-​fucking-​tastic (McCarthy, 1982). Although this morphological process is generally not taught explicitly, native speakers are sensitive to implicit patterns governing how and where this infix can be inserted into words. For example, they generally do not produce the possible analog of the first example, *fanta-​fucking-​stic.1 1 The “*” indicates that the sentence would typically be judged as unacceptable by a native speaker of English or at least some dialect of it. To my ears at least, the analog of the second example,?ab-​ fucking-​solutely, sounds much less awkward. This shows that the pattern might be complicated and that our acceptability judgments are often graded rather than categorical. Linguists often represent this graded quality by means of a crude scale that extends from acceptable through “?” and “??” to “*”.

100  Abstract Concepts and the Embodied Mind There is an obvious but nevertheless important disconnect between the phonological and morphological levels: at any given point in time, there are more acceptable groupings of phonemes than there are meaningful morphemes. Psycholinguists often take advantage of this mismatch and use unfamiliar “nonce” words such as blick to enable them to test aspects of language and concept learning. There is a clear sense in which this asymmetry is functional because it leaves room for the emergence of new lexical items and neologisms. Recent examples in English are words such as google (used as a verb) or puggle. Borrowed words from other languages such as emoji and hygge also become anglicized in their pronunciation. An important symbolic feature of language is that it is generative and productive in the mathematical sense. Speakers can produce a virtually unbounded number of sentences because sentences are made of words and grammatical elements that can be recursively combined to form new constituents (Pinker, 1994). Syntax concerns the rules or constraints that govern how words are combined to form phrases and sentences. The combinatorial properties of syntax not only explain our capacity to produce a virtually unbounded set of sentences but also explain our capacity to understand novel phrases or sentences. A story about an expedition searching for a spotted zebra makes sense to an English-​speaking child who has never seen or heard of such a creature because of their implicit understanding of how adjectives and nouns combine to form a noun phrase. An understanding of syntax also enables us to assign thematic roles to elements of the discourse. For example, the striking difference between A pug bit a large man and A large man bit a pug is the result of the fact that the former identifies a pug as the agent of the action with a human as the patient of the action and the latter identifies a human as the agent of the action with a pug as the patient of the action. Importantly, syntactic operations are often recursive—​that is, it is often possible to construct a constituent out of smaller constituents that are constructed in the same manner. Consider the following group of sentences: Mary thinks that Tom is rude. Jules knows that Mary thinks that Tom is rude. Hilary believes that Jules knows that Mary thinks that Tom is rude.

The processes by which the embedded sentences are constructed are productive because they can be repeated indefinitely. Our capacity to infer the

Language Is a Neuroenhancement  101 meaning of a phrase or sentence from the meaning of its parts rests on two important elements: a collection of meaningful symbols that are stored in long-​term memory and a grammar for combining these symbols to form larger, interpretable symbols. In other words, a combinatorial syntax is an important design feature of natural language. Within linguistics, it is common to assume that conceptual knowledge is handled by an independent cognitive system. Concepts are connected to words or morphemes in what has come to be known as the mental lexicon. The mental lexicon is a component of semantic memory. It is a repository for all that is idiosyncratic within a particular language, dialect, or idiolect. On this standard view, the mental lexicon connects the separate brain systems associated with semantic and syntactic processing. Content words such as nouns, verbs, and adjectives (or their equivalents) are key elements of the mental lexicon. These words are often referred to as “open-​class” words because they are the type of words that are more likely to fall out of use or to be added through lexical innovation, contact with another language, or other forms of language change. Open-​class words are often contrasted with closed-​class words, which are more resistant to loss and less likely to be added through innovation, contact, or other forms of language change. Closed-​class words tend to be function words linked to grammatical processes such as auxiliaries, determiners, prepositions, and pronouns. Open-​class words are a central focus of research on concepts because they encode our general knowledge about the world. The question at hand in this chapter is the degree to which having a label for a concept—​and the connections to other words that the label brings—​extends our capacity to capture abstract content. A branch of linguistics often referred to as cognitive linguistics questions the idea that there is a clean separation of conceptual knowledge and grammar. Croft and Cruse (2004) propose that this theoretical approach is characterized by three major hypotheses: (i) that the processes governing language use are the same as—​or at least highly similar to—​those that govern other cognitive activities, (ii) that conceptualization is shaped by social and communicative goals, and (iii) that our knowledge of phonology, morphosyntax, semantics, and discourse is derived from our experiences of specific instances of linguistic communication. Some of the strongest arguments in support of cognitive linguistics involve analyses of grammatical and semantic phenomena that other linguists view as peripheral but are clearly relevant to an understanding of how language

102  Abstract Concepts and the Embodied Mind encodes conceptual content. A good starting point is the structure and organization of idioms. Idioms are collections of morphemes or words that are idiosyncratic with respect to their morphosyntax, semantics, or both. Croft and Cruse (2004, p. 230) offer the following examples: It takes one to know one Pull a fast one Bring down the house Wide awake Sight unseen All of a sudden (X) blow X’s nose Once upon a time . . .

A central feature of idioms is that their meanings often cannot be fully predicted from the conventional meanings of their parts. In other words, understanding idioms often requires idiosyncratic semantic conventions. This conventionality can be a matter of degree. Nunberg, Sag, and Wasow (1994), for instance, draw a distinction between idiomatically combining expressions and idiomatic phrases. The former contain elements that can be put in correspondence with literal counterparts. For example, the word answer in answer the door can be treated as corresponding to the action of opening the door. The latter are not compatible with this sort of analysis. For example, no such correspondences exist for the word kick or the phrase the bucket in kick the bucket (in the sense of “to die”). Instead, the idiom needs to be interpreted holistically. Idioms exhibit several other characteristic features (Croft & Cruse, 2004): they often express a figurative meaning (e.g., give a hoot, lend a hand, and take the bull by the horns), they often refer to social rather than physical activities (e.g., chew the fat, get on board, and let the cat out of the bag), and they often have an affective or evaluative character (e.g., cut corners, miss the boat, and stuck in the mud). Many are used in specific pragmatic or discourse-​related contexts (e.g., good morning, once upon a time, and see you later). Another important distinction concerns the relationship of idioms to the general grammatical patterns of a language (Fillmore, Kay, & O’Connor, 1988): while some idioms—​such as in the same boat, kick the bucket, and spill the beans—​appear straightforwardly grammatical, others appear to be extra-​ grammatical—​such as all of a sudden, believe you me, and no can do (for more

Language Is a Neuroenhancement  103 examples, see Fillmore et al., 1988; Nunberg et al., 1994). Idioms can also be lexically open or closed; that is, they may or may not have parts that can be filled by syntactically and semantically appropriate elements. Idioms represent a theoretical challenge to the standard conception of the relationship between grammar and meaning because (1) their meaning typically cannot be determined in accordance with general semantic interpretation rules, and (2) they often participate in grammatical processes known as constructions. Consider, for instance, the use of let alone as a form of conjunction with respect to items that are the foci of discourse. This use can be seen with the following pairs of sentences (Fillmore et al., 1988, p. 514):2 1. a. Max won’t eat shrimp, let alone squid. b. We’ll need shrimp and squid. 2. a. Max won’t touch the shrimp, let alone clean the squid. b. I want you to cook the shrimp and clean the squid. While let alone can act semantically as a conjunction, it does not behave syntactically like the English conjunction and in all contexts. Consider how these two elements interact differently with the grammatical process known as topicalization (Fillmore et al., 1988, p. 515): 3. a. Shrimp and squid Moishe won’t eat. b. *Shrimp let alone squid Moishe won’t eat. c. *Shrimp Moishe won’t eat and squid. d. Shrimp Moishe won’t eat, let alone squid. What a careful analysis of this construction demonstrates is that idioms can be both productive and grammatically complex. Construction grammar embraces the fact that there are a wide class of structures that make up a language which are not considered to part of the “core” grammatical phenomena recognized by the linguistic mainstream. Constructions are regular form-​meaning pairings that are not completely predictable from other aspects of grammar (Goldberg, 1995). In a sense, the appeal to constructions represents the recognition that localized 2 In the original version, Fillmore and colleagues use uppercase letters to indicate prosodic prominence. Since I am using uppercase to indicate concepts, I have put the relevant words in bold to avoid confusion.

104  Abstract Concepts and the Embodied Mind grammatical phenomena—​which at one time would have been part of pedagogical or descriptive grammars—​play an important role in our linguistic competence (Fillmore & Kay, 1993). An advantage of a construction-​based approach is that it can capture generalizations concerning the relationship between the symbolic properties of grammar and aspects of semantic meaning that are ignored by other approaches. Consider the relationship of a verb to its complements. On a standard approach, the verb is thought to determine the number of complements it takes in virtue of the fact that it acts as an n-​place semantic predicate. The trouble is that a garden-​variety action verb such as kick participates in several different argument structures (Goldberg, 1995, p. 11): 4. a. Pat kicked the wall. b. Pat kicked Bob black and blue. c. Pat kicked the football into the stadium. d. Pat kicked at the football. e. Pat kicked his foot against the chair. f. Pat kicked Bob the football. g. The horse kicks. h. Pat kicked his way out of the operating room. In sentence 4d, kick acts as a binary relation with arguments that fill agent and patient roles. In sentence 4f, it acts as a ternary relation with arguments that fill agent, recipient, and patient roles. In sentence 4c, it also acts as ternary relation, but now the arguments fill agent, theme, and goal. Goldberg points out that the standard approach gets tangled up in a circularity: on the one hand, the verb is claimed to have a specific n-​argument sense because it occurs with n complements, and, on the other hand, it is thought to occur with n arguments because it has an n-​argument sense. She then argues that a construction-​based explanation provides a more parsimonious explanation of these sentences (and others like them) by linking them to skeletal constructions. For instance, the ternary relationships exhibited here are directly associated with the ditransitive construction. The verb itself is then associated with a basic sense (or at least a small number of basic senses) which becomes integrated with the meaning of the construction. In other words, a construction-​based approach enables us to explain the interaction between verb meaning and constructions in a way that can be extended to other verbs.

Language Is a Neuroenhancement  105 Idioms and constructions highlight two important things about the symbolic nature of language. The first is that some aspects of semantics are integrated with grammatical structure (Kemmerer, 2019). The second is that the conceptual information encoded by words and constructions extends beyond descriptive propositional content. Rather than adopt the truth conditional semantics of analytic philosophy of language and mathematical logic, cognitive linguistics have looked to empirical research on concepts. As I discussed in previous chapters, this research suggests that concepts do not simply encode necessary and sufficient conditions for category membership. Matlock and Winter (2015) propose that cognitive linguists view meaning as perspectival, dynamic, encyclopedic, and usage-​ based. They view meaning as perspectival because the scenes depicted by linguistic utterances are often construed in a way that reflects a particular viewpoint or perspective (Langacker, 1987a; Talmy, 1983). Speakers of English, for instance, may describe a distribution of stars as a constellation, a cluster of stars, specks of light in the sky, etc.—​each of which reflects different ways of construing the same collection of physical objects (Langacker, 1987b). Meaning is viewed as dynamic because it is built up during semantic processing in an incremental and context-​sensitive fashion. Meaning is viewed as encyclopedic because it includes general knowledge tied to situations (more on this later). Finally, meaning is viewed as usage-​based because it is often shaped by our past experiences with the relevant words and constructions. Matlock and Winter note that these characterizations should not be taken to be mutually exclusive. Each of these characteristics fits well with the approach developed in this book, but let’s focus on the third for the moment. Frame semantics (Fillmore, 1977, 1982) emphasizes the encyclopedic nature of meaning. Fillmore summarizes (2012, p. 712): The idea behind frame semantics is that speakers are aware of possibly quite complex situation types, packages of connected expectations, that go by various names—​frames, schemas, scenarios, scripts, cultural narratives, memes—​and the words in our language understood with such frames as their presupposed background.

A parade case for the notion of a frame is the observation that the meaning of the words such as buy, cost, pay, and sell seem interconnected. Fillmore proposes that they are all tied to a commercial event frame. Another example

106  Abstract Concepts and the Embodied Mind of how stored background information can influence meaning is the contrast between how readers are likely to respond to the sentences I spent three hours on land this morning and I spent three hours on the ground this afternoon (Fillmore, 1977). The former is likely to be interpreted in terms of a background scene involving travel by boat and the latter is likely to be interpreted in terms of a background scene involving air travel. Background knowledge can be relevant to the meaning of an individual term. The concept of a vegetarian relies in part on the recognition that many people in our culture have diets that include meat (Fillmore, 1985). The encyclopedic nature of meaning opens the door to a cognitive approach to language. After all, it leaves room for concepts to encompass both sensorimotor and linguistic information. But a commitment to cognitive linguistics does not settle the issue. We have already seen that Barsalou and colleagues emphasize the importance of situations in our concepts while downplaying the importance of language—​relegating it to the role of a shortcut. They are not alone. It is quite common to view constructions as intermediaries for deeper semantic processing carried out by means of experientially based simulations. To give another example, embodied construction grammar (Bergen & Chang, 2005) connects constructions to embodied schemas, which are general cognitive structures that rely on grounded mechanisms. Others, though, argue that constructions influence conceptual processing. For example, Kemmerer (2006, 2019) proposes that speakers of different languages are likely to construe objects and events in different ways when relying on linguistically encoded information. At a minimum, this amounts to a strong commitment to the influence of language on thinking for speaking (Slobin, 1996). I propose that the symbolic properties of language support our capacity to encode concepts, particularly ones with disembodied content. Several caveats are warranted. For one, just as an appeal to experiential grounding is not a direct appeal to conscious imagery (because simulation can be unconscious and automatic), an appeal to a cognitive role for language is not a direct appeal to inner speech. While inner speech likely serves important cognitive functions in a way that fits with my approach (Dove et al., 2020), this approach also holds that linguistic information contributes to unconscious conceptual processing. Second, I propose that language enhances rather than replaces experientially grounded cognition. Language extends our cognitive abilities by enabling us to leverage a compositional symbol system. Third, I propose that linguistic representations are components of

Language Is a Neuroenhancement  107 distributed, multimodal concepts. Despite these caveats and the complexities that they introduce, the issue at hand is a straightforward one: Is some conceptual content encoded in representations tied to the language system, or is this system merely a vehicle for eliciting nonlinguistic simulations of experience?

6.2  Thinking Without Words For various practical reasons, most research on grounded cognition has involved language-​based tasks. Linguistic stimuli are convenient and handy prompts for experiments. Words, after all, demarcate shared concepts. While the ideas at the heart of grounded cognition clearly suggest the possibility of thought without language, little effort has been made by researchers to disentangle the contribution of language to the conceptual system in the context of embodied cognition. Despite this neglect, there are good reasons to speculate that thinking can occur without words (Pinker, 1994). For one, there is the question of how infants and young children are able to acquire concepts that are encoded in a particular natural language. The capacity to learn word meanings seems to require the prior ability to formulate hypotheses about what the words might mean (Fodor, 1981). Perhaps the most famous neuropsychology case study relating to independence of language and thought involves a Canadian monk known as “Brother John” (Lecours & Joanette, 1980). Brother John was an educated man who served on the editorial staff of the periodical put out by his order. He was epileptic and during his seizures experienced what Lecours and Joanette (1980, pp. 4–​5) refer as “spells” of paroxysmal aphasia where he would remain conscious but lose all productive and receptive language skills, including inner speech. These spells could be either relatively short (1–​5 minutes) or relatively long (1–​11 hours). During these spells, Brother John was able to function remarkably well. For example, he was able to recognize objects, use tools, navigate familiar and novel environments, and even manage social interactions. Famously, one of Brother John’s long spells occurred as he was traveling by train from Italy to a small town in Switzerland. As he arrived, he began to experience a long spell of dysphasia. Despite having never been there before, he managed to get his bags, disembark from the train, find a hotel, check in by means of handing over his passport, and order a meal by pointing to the menu. Unfortunately, he ordered fish—​something he did not

108  Abstract Concepts and the Embodied Mind like to eat. There are many reasons to be cautious with respect to drawing firm conclusions from this remarkable case study. Given the inherent limitations of this sort of anecdotal report, we should treat it as suggestive at best. Naturally occurring linguistic isolates provide further support for the potential independence of thought from language. Dramatic instances of children raised in isolation from other people—​such as the cases of Genie (Curtiss, 1977) and Victor of Aveyron (Lane, 1976)—​capture the imagination, but far too much remains unresolved concerning the details of their upbringing and the nature of their deficits. Fortunately, less dramatic cases are available. Donald (1993) points to documented instances of deaf people from pre-​literate societies. Despite their lack of exposure to formal writing systems or fully realized sign language, there seems to be little reason to suspect that such individuals were incapable of some forms of sophisticated cognition. Indeed, adults without language may not be that rare in the real world. Schaller (2012) relays her experience teaching sign language to Ildefonso, a profoundly deaf 27-​year-​old Mexican immigrant. When they met, Ildefonso was not only unable to sign but appeared to be completely unaware of how sign language functions. Despite this, he worked as a farmhand for much of his life. Echoing Brother John’s story, Ildefonso recounts how, as a boy, he was sent on plane trip by his father to pick apples with a group of men in upstate New York. When he got there, the men were arrested, but he was not. Somehow, he was able to find and join another group of workers (Schaller, 2012, pp. 200–​201). Two caveats are warranted with respect to Ildefonso’s case. The first is that, although Schaller describes Ildefonso as obviously intelligent, she did not carry out careful studies of his cognitive abilities. The second is that it is not clear that Ildefonso was completely languageless. Several studies have uncovered the systematic qualities of the gestural communications of deaf children with their nonsigning hearing parents (Goldin-​Meadow & Feldman, 1977; Goldin-​Meadow & Mylander, 1998; Goldin-​Meadow, Mylander, & Butcher, 1995). It thus remains possible that Ildefonso acquired some language or language-​like representational abilities at some point in his life. Having made these caveats, though, it seems fair to say, that prior to acquiring American Sign Language (ASL), Ildefonso’s cognitive abilities manifestly outpaced his linguistic ones. Careful studies of neurological patients—​and corresponding brain imaging research with neurotypical participants—​provide further support for the claim that some cognitive skills are independent of language. The

Language Is a Neuroenhancement  109 apparent independence of some aspects of mathematical reasoning from language provides an intriguing example. A study of three profoundly aphasic patients found that they were able to solve mathematical problems that involved recursion and structure-​dependent operations (Varley et al., 2005; see also Rossor, Warrington, & Cipolotti, 1995). This study fits with the brain imaging evidence implicating the existence of a neurological dissociation between language and mathematical processing (Monti, Parsons, & Osherson, 2012). Even so, the preservation of these skills in such patients remains somewhat surprising because some research indicates a correlation and possible neuroanatomical overlap between linguistic and mathematical processing on certain tasks (Baldo & Dronkers, 2007; although see Fedorenko & Varley, 2016, for a critical review of this evidence). Admittedly, the relationship between aphasia and calculation difficulties, often referred to as acalculia, can be somewhat difficult to ascertain in clinical contexts. Comorbidity is common but not universal (Fedorenko & Varley, 2016; Rosselli & Ardila, 1989). Some researchers distinguish primary forms of acalculia that involve a fundamental deficit in calculation abilities from secondary forms that result from other deficits, including but not limited to deficits associate with linguistic processing (Ardila & Rosselli, 2002). However, it is often difficult to distinguish primary from secondary forms (Ardila & Rosselli, 1990, 2002). Significantly, there is a recognized type of acalculia that occurs with right hemisphere damage (Grana, Hofer, & Semenza, 2006). While this type has traditionally been categorized as secondary form of acalculia (one that is caused by impaired spatial processing abilities), recent evidence suggests that right hemisphere damage can be associated with impairments of core calculation abilities that are independent of spatial processing (Benavides-​Varela et al., 2017). This leads to a distributed view of the calculation system in which right hemisphere structures play a primary rather than a secondary role (Semenza & Benavides-​Varela, 2017). The example of mathematical calculation is compelling because it provides a clear instance of an undeniably higher-​level cognitive skill that appears to rely at least in part on neural mechanisms that are not directly tied to the language system. This example is also useful precisely because there are many reasons to think that mathematical reasoning is enhanced—​both in a synchronic and a diachronic sense—​by natural language and other external symbol systems. To be clear, the available evidence shows that sophisticated mathematical cognition can occur in the absence of language, but no one would argue that all mathematical cognition is independent of language.

110  Abstract Concepts and the Embodied Mind

6.3  Word and Thought One conclusion that we might draw from the previous discussion is that it is difficult to disentangle language and thought. There are two leading explanations for this difficulty. The first holds the line with respect to the independence of thought and language and maintains that the neural mechanisms responsible for language processing are largely separate from and independent of those involved in conceptualization and other cognitive processes. It explains away the apparent entanglement of language and thought by an appeal to our reliance on language for communication. We have a hard time deciding where language ends and concepts begin because we rely on language to express conceptual content. The second explanation contends that language and thought are difficult to disentangle precisely because they are entangled. In other words, the difficulty is a consequence of the significant diachronic and synchronic roles that language plays in our cognitive lives. I suggest that we need to shift from the former explanation to the latter one.

6.3.1  Early Opposition Borghi (2020) offers two possible sources for the antipathy among early advocates of grounded cognition toward cognitive views of language. The first is a shared opposition to anything associated with amodal theories of cognition. Many, if not most, amodal theories assume that cognition is handled by a language-​like representational format (Fodor, 1975). Given this assumption, it is not surprising that a research program aimed at rejecting amodal symbols would object to the idea that linguistic representations play a role in our conceptual processing. The second is the widespread perception that cognitive views of language fall afoul of the symbol grounding problem and can thus be rejected a priori. There are good reasons to question the reasoning behind each of these sources of antipathy. Let’s begin with the association with amodal cognition. Early approaches to grounded cognition tended to treat words, phrases, and sentences as merely the means by which situated simulations of nonlinguistic experience are engaged (Barsalou, 2016b; Zwaan, 2016). Although the separation of the language system from the conceptual system may well have been inspired by a rejection of amodal cognition, this separation recapitulates the

Language Is a Neuroenhancement  111 traditional commitment to modularity (Fodor, 1983). This modularity, however, fits poorly with the fundamental tenets of grounded cognition. After all, language itself is something we experience in an embodied and grounded way (Borghi & Binkofski, 2014; Dove, 2014, 2018). If the reuse of experiential resources is important to conceptual processing, then why wouldn’t this include reuse of representations associated with linguistic experience? Certainly, we gather a great deal of what we know about the world through the activities of talking, listening, writing, reading, etc. It seems reasonable to suppose that part of the representations that are automatically engaged in conceptual tasks are experiential representations associated with the processing of language. In the end, a commitment to grounded cognition does not require rejecting a cognitive view of language. Now, let’s turn to the symbol grounding problem. Although many argue that it provides a compelling reason to adopt a fully grounded approach to cognition (Barsalou, 1999, 2008, 2016a; Glenberg, 1997; Zwaan & Madden, 2005), I showed in Chapter 3 that it leaves room for intermediate accounts that contain both embodied and disembodied representations. For this reason, it cannot be used to exclude cognitive views of language out of hand.

6.3.2  Reimagining the Role of Language The shift toward a cognitive conception of language has been anticipated by earlier theories that are compatible with the basic tenets of grounded cognition. A good place to start is with Vygotsky (2012), who proposed that internalized language can serve as scaffold for learning. On his view, inner speech can help the child organize, plan, and remember actions (Alderson-​Day & Fernyhough, 2015). Clark (2006) offers a grounded version of the scaffold thesis that emphasizes the degree to which language is a physical transformation of our “cognitive niche.” The act of labeling helps learners become attuned to perceptual commonalities and overcome the inherent complexity and noisiness of perceptual inputs (Clark, 1998; Lupyan & Clark, 2015). More broadly, language creates a novel set of perceptual objects and targets for action—​which enables us to model the world by means of the manipulation of a shared symbol system. Clark (2008, p. 47) summarizes: The computational value of a public system of essentially context-​free, arbitrary symbols, lies . . . in the way such a system can push, pull, tweak,

112  Abstract Concepts and the Embodied Mind cajole, and eventually cooperate with various non-​arbitrary, modality-​rich, context-​sensitive forms of biologically basic encoding.

On his view, language serves as a cognitive scaffold because it is an effective and readily available external support for our thinking. Becoming a speaker of a public language creates a flexible and multipurpose form of extended cognition (see also Rowlands, 2003). The language-​specific character of this scaffold can be seen in the cross-​linguistic diversity in the meanings of words and constructions (Kemmerer, 2019). While the scaffold proposal emphasizes the importance of language as an external symbol system (albeit one that can have inner influences due to the cognitive, and likely metacognitive, effects of inner speech), another historical precursor to the recent cognitive views emphasizes the internal role that language might play. The dual code theory (Paivio, 1971, 1986, 1991) holds that our concepts are encoded in two formats: one that is sensorimotor-​ based and another that is language-​based. The former employs imagistic nonverbal representations called imagens, and the latter employs verbal representations called logogens. The dual code theory proposes that most concepts employ both formats, but abstract concepts tend to depend more on linguistic representations: that is, concepts with low imageability are associated primarily with verbal representations while highly imageable concepts are associated with both verbal representations and perceptual ones. Overall, the focus in dual code theory on the dynamic relationship between experience and mental representations seems to be in keeping with the basic tenets of grounded cognition. However, an important aspect of the dual code theory—​namely, its view of language as an independent symbol system—​is incompatible with early variants. More recently, several researchers have proposed that the language system, or at least our experience of language, plays a significant role in our grounded conceptual system. Examples of specific theories that embrace the influence of language include (in alphabetical order) the embodied conceptual combination or ECCo theory (Lynott & Connell, 2010), the language and situated simulation or LASS theory (Barsalou et al., 2008), the language and associations in thinking or LASSO theory (Tillas, 2015), the symbol interdependency or SI theory (Louwerse & Jeuniaux, 2010), and word as social tool or WAT theory (Borghi & Cimatti, 2009). Although all these theories agree that language is an important tool for thinking, there is significant disagreement among them concerning what kind of tool it is and how it works. They differ

Language Is a Neuroenhancement  113 with respect to the degree to which they acknowledge the external and internal influences of language on cognition. These theories form a cline with respect to the relative importance that they assign to language. A minimalist proposal holds that language is merely a shortcut that saves time and cognitive effort. According to the LASS theory, for instance, language underwrites a superficial form of semantic processing that enables us to avoid the deeper processing associated with situated simulations. Language, in other words, provides an occasional and somewhat reliable substitute for grounded cognition. A less constrained view emphasizes the degree to which linguistic information stored within concepts can serve as an important cognitive shortcut in certain contexts and with certain tasks (Connell & Lynnott, 2013). This view proposes that such a shortcut enables us to leverage the associative information contained in language in a way that is computationally efficient (Connell, 2019). A stronger view of the contribution of language to our concepts emerges from the recognition that the structure of language itself is likely to reflect important aspects of the world. Some theories propose that conceptual processing relies in part on distributional information derived from spoken and written language. The core idea is that the meaning of a word is constrained by the company it keeps (Firth, 1957). On this sort of view, language represents a rich symbolic medium that enables an alternative means of accessing meaning (Louwerse & Jeuniaux, 2008). In keeping with this idea, several researchers have created models of semantic memory that incorporate both distributional and experiential information (Andrews, Vigliocco, & Vinson, 2009; Davis & Yee, 2021; Günther, Rinaldi, & Marelli, 2019; Johns & Jones, 2012). Hybrid theories that view cognition in terms of experiential and symbolic components can themselves be differentiated with respect to the contribution of language. Zwaan (2016) outlines a pluralistic account in which—​using his terminology—​ abstract and grounded symbols constrain each other. Louwerse and Jeuniaux (2008), on the other hand, emphasize the greater importance of linguistic information. All the cognitive views of language that we have considered so far treat linguistic symbols as fundamentally disembodied. It is possible, though, to offer an account of the role that language plays in cognition that is more in keeping with grounded cognition. The WAT theory, for example, offers a grounded account of both experiential and linguistic processing. This theory emphasizes the importance of language as both an external and an internal cognitive tool. According to this theory (Borghi & Binkofski, 2014, p. 19),

114  Abstract Concepts and the Embodied Mind Words can be seen as tools because, similar to physical tools, they allow us to act in the world, together with and in relation to other individuals; they are social also since they are acquired and used in a social context.

The WAT theory thus adopts a full-​throated version of the scaffold thesis. It also goes further and argues that abstract and concrete concepts differ in their neural representation: specifically, abstract concepts recruit more brain areas associated with linguistic and social cognition (Borghi et al., 2019). The WAT theory has four fundamental tenets (Borghi et al., 2019). The first is that abstract and concrete concepts differ in how they are acquired, with abstract concepts relying more on language and social interaction and concrete concepts relying more on sensorimotor experience. This diachronic difference is then reflected in a synchronic difference in how these two types of concepts are processed. The second is that the neural representation of abstract concepts involves interoceptive, linguistic, and social brain networks more than concrete concepts. The third is that, because of the greater involvement of linguistic networks, the mouth motor system should be more active during the processing of abstract concepts. The fourth is that linguistic variation should have a greater impact on abstract concepts than it does on concrete concepts. In addition to these tenets, the WAT theory has emphasized the degree to which language can serve as a vehicle for inner grounding and flexibly enhance metacognition (Borghi et al., 2019). Early versions of grounded cognition tended to focus on concrete concepts and deliberately rejected cognitive views of language. This rejection created a rift between grounded cognition and related approaches that envisioned an important cognitive role for language in embodied cognition (e.g., Clark, 2006, 2008; Paivio, 1986). More recently, though, the mainstream of grounded cognition has come to embrace cognitive views of language. The remaining controversy concerns the nature and extent of language’s influence.

6.3.3  The LENS Theory This book proposes a flexible, multilevel, and multimodal account of concepts. The heterogeneous class of abstract concepts creates several significant theoretical challenges for such an approach. In this chapter, I defend the claim that language plays an important role in overcoming the problem

Language Is a Neuroenhancement  115 of disembodiment. Language helps because it is more than a scaffold; it is also a neuroenhancement. I refer to this account as the language is an embodied neuroenhancement and scaffold or LENS theory (Dove, 2019). The LENS theory holds that external scaffolding (Buckner, 2018) is not enough to explain the contribution of language to our concepts. Acquiring language involves the development of sophisticated skills that involve the coordination of multiple modalities and the ability to manipulate an independent combinatorial symbol system. The LENS theory incorporates Clark’s suggestion that language is an external support for the embodied mind. Language is just one of several externally sourced symbol technologies that we are able to acquire (Clark, 2008). Our understanding of mathematics, for instance, often depends on such technologies. To give a prosaic example, learning how to perform long division on paper often involves the grounded manipulation of and interaction with physical symbols (Landy, Allen, & Zednik, 2014; Landy & Goldstone, 2007b). Although language begins as an external symbol system that we learn to manipulate in an embodied and grounded way, it is also special because there are few cognitive domains that are not impacted by language to some degree or another. The LENS theory departs from Clark’s account in emphasizing the internal role played by the neurologically realized language system in our concepts. Theories of embodied cognition come in two main varieties: those that emphasize the influence of the body on the mind and those that emphasize the importance of body–​world couplings. Viewing language as an external symbol technology that becomes an integrated part of the neurologically realized conceptual system bridges these conceptions. The core idea is that the acquisition of language transforms the cognitive lives of children by means of offering a new symbolic medium for thinking that has many of the favorable properties identified by supporters of amodal cognition (Dove, 2011, 2014, 2018).3 Language involves the manipulation of our physical and social niche, but it also provides children with new representational abilities that enhance their cognitive reach (Dove, 2014, 2018). 3 The LENS theory shares some fundamental commitments with the WAT theory. Both theories view the contribution of language in terms of a flexible, multilevel, and multimodal approach to grounded cognition. Both emphasize that language can serve as an external and internal cognitive tool. Where they differ is in their details. While the WAT theory focuses on the developmental connection of language to social action, the LENS theory focuses on the internal synchronic role of the language system. For a detailed side-​by-​side comparison of these theories co-​written by their respective creators, see Dove et al. (2020).

116  Abstract Concepts and the Embodied Mind This dual benefit is a consequence of the skills needed to be a fluent speaker of a natural language. For the sake of explanatory convenience, let’s focus on oral communication (everything I say will apply mutatis mutandis to visual forms of linguistic communication, such as signing or even writing). To be successful, a child needs to learn to recognize—​and ultimately produce by means of precise bodily actions—​the speech sounds of a particular language. In addition to acquiring phonological competence, the child needs to acquire morphological competence; that is, they must learn how morphemes combine to form words. They also need to acquire syntactic competence to be able to combine words (and morphemes) to form constructions, phrases, and sentences. They need to acquire semantic competence to understand the mostly language-​specific meanings of these morphemes, words, and constructions. Often, this requires acquiring new concepts rather than mapping previously acquired ones onto linguistic forms (Bowerman & Levinson, 2001). Beyond this, they need to develop the pragmatic discourse-​related skills needed to comprehend and produce meaningful conversations. In sum, becoming a competent speaker of a natural language involves a suite of skills. The question then is not whether natural language competence itself fits with the idea of embodied cognition, but rather to what degree such grounded and extended skills influence our concepts and thoughts. A caveat is warranted. Viewing language as a neuroenhancement is inherently constructivist. The suggestion is, after all, that the human conceptual system is significantly altered and transformed through the dynamic and sustained interaction with others through the medium of a natural language. Such constructivism predicts that there should be cross-​linguistic diversity in how the world and our experience of it are conceptualized (Kemmerer, 2019). It also likely requires the rejection of strongly nativist conceptions of the acquisition process that posit a rich, genetically determined language faculty.4 Nevertheless, we should not infer that it excludes the influence of evolutionary factors. For one, there are approaches to evolution that emphasize the relevance of developmental and environmental influences, including evo-​devo theories (West-​Eberhard, 2005) and what has been called the extended evolutionary synthesis (Laland et al., 2015). For another, it has been suggested that the human brain and language might have co-​evolved 4 I do not have the space to get into a protracted discussion with respect to the status of specific claims concerning our innate endowment for language (a good place to start is Cowie, 1999, 2017). It is worth noting, though, that nativist claims have become progressively weaker and more general over time.

Language Is a Neuroenhancement  117 in such a way that the former may have shaped an appropriate sociocultural niche, and the latter may have shaped an appropriate neural niche (Deacon, 1997). Third, alternative developmental explanations have been offered for the features of language acquisition that are often taken as evidence for a genetic endowment for language (Christiansen & Chater, 2016; Culicover, 1999; Dove, 2012). Now that I have provided the basic outlines of the LENS theory, I need to present evidence for it. Viewing language as a neuroenhancement leads to several specific predictions: 1. Words, acting as labels, should enrich our concepts in ways that help us generalize away from experiential particulars. 2. Some disembodied conceptual content should be encoded in the associations of words with other words. 3. The structural properties of language should help us draw inferences that go beyond our immediate experience. 4. Discourse-​ level knowledge linked to conversations and narratives should contribute to the flexibility of concepts—​ particularly abstract ones. 5. Given predictions 1–​4, the language system should be more engaged with abstract concepts than it is with concrete concepts. In the sections that follow, I outline the evidence supporting each of these claims. I conclude that there is a strong empirical case for the claim that language acts as a neuroenhancement.

6.4  Labels In general, words and idioms are semantically arbitrary. Putting aside onomatopoeia and rough phonological associations with semantic complexity—​ and even these are often a matter of degree rather than clear examples of semantic transparency—​most words are (somewhat) arbitrarily connected to their conceptual content. Different languages will associate different words with similar concepts. Moreover, the same word can shift meaning over time. This shift can be dramatic, as in the case of the English word hussy, which originates from a contraction of the Middle English word for housewife, or it can be subtle, as in the case of the English word peruse, which at one time

118  Abstract Concepts and the Embodied Mind meant to read carefully but now means to browse or skim.5 This semantic arbitrariness fits well with purely communicative conceptions of language. After all, on such conceptions, words label preexisting concepts that are handled by fully independent neuromechanisms (e.g., Fodor, 1975; Gleitman & Papafragou, 2005; Pinker, 1994). The LENS theory, though, builds on a tradition that emphasizes the influence of language on cognition (e.g., Clark, 1998; Quine, 1960; Vygotsky, 2012). Here, I argue that the semantic arbitrariness of words underwrites their capacity to augment and extend our cognitive powers. Neuropsychological case studies provide an initial motivation for a cognitive view of language. It has been known for some time that patients with aphasia often experience impairments in nonlinguistic cognitive tasks (Goldstein, 1948; Noppeney & Wallesch, 2000). Aphasics often have trouble with categorization tasks that require the identification of a specific attribute shared by different items rather than a global comparison between them (Cohen, Kelter, & Woll, 1980; Semenza, Bisiacchi, & Romani, 1992). The patient identified as LEW provides a compelling case study (Davidoff & Roberson, 2004; Roberson, Davidoff, & Braisby, 1999). LEW experiences great difficulty in naming pictures of actions or objects. Despite this anomia, LEW can respond appropriately to spoken definitions (Druks & Shallice, 2000). This suggests that LEW’s semantic knowledge is preserved to some degree. What is important for our purposes is that LEW struggles with taxonomic classification tasks that involve a single dimension (such as shape or color) but not with thematic classification tasks that involve broader comparisons (Davidoff & Roberson, 2004; Roberson et al., 1999). Following up on this research, Lupyan and Mirman (2013) compared the performance of 12 aphasic participants with 12 age and education matched controls on a categorization task. Presented with sets of 20 pictures, participants were asked to select all that fit a given criterion. Lupyan and Mirman hypothesized that the aphasic participants would experience greater difficulty when the criterion was, in their words, “low-​dimensional” (e.g., things that are green) than when the criterion was “high-​dimensional” (e.g., farm animals). This prediction was borne out: not only were the participants with aphasia selectively impaired on the low-​dimensional task, but the degree of their impairment also was correlated with the degree of their anomia 5 The historical evolution of these terms can be found in their respective Oxford English dictionary entries.

Language Is a Neuroenhancement  119 (Lupyan & Mirman, 2013). In a related experiment with neurologically intact participants, a similar selective impairment for taxonomic categorization and not thematic categorization was induced by means of a verbal interference task (Lupyan, 2009). All of this suggests that there is a causal link between language and at least some cognitive tasks (Boutennet & Lupyan, 2015; Lupyan, 2012b). What mechanisms are responsible for this? The LENS theory predicts that labels should act both as a scaffold for cognition and an internal neuroenhancement. Let’s begin with the scaffold idea first. Labels create a novel set of perceptual objects and targets for action (Clark, 2006). As such, they may help learners become attuned to perceptual commonalities and overcome the inherent complexity and noisiness of perceptual inputs (Clark, 1998; Lupyan & Clark, 2015). In 9-​month-​old children, categorization is facilitated by the simultaneous presentation of words but not tones (Balaban & Waxman, 1997). Distinct labels (but not distinct tones, sounds, or emotional expressions) have been found to help this cohort with an object individuation task (Xu, 2002). By the time they are a year old, infants have learned to expect nouns to be linked to categories of objects (Waxman & Markow, 1995). As we get older, labels may be able to assist categorization and concept formation in more sophisticated ways. Yuan and colleagues (2011) suggest that while preschool children learn individual words, they also learn-​to-​learn through linguistic tasks. Such learning occurs on multiple levels and carries over to nonlinguistic cognitive domains. There is also evidence that labels continue to influence learning into adulthood. Labeled categories are easier to learn than unlabeled categories even when the relevant experience is held constant and the labels are redundant (Lupyan, Rakison, & McClelland, 2007). Our current focus is on the role played by labels. All too often, discussions of conceptual grounding simply ignore the contributions of linguistic representations. It seems likely that this is due to an implicit or unstated commitment to the functional independence of the conceptual system. No matter how you slice it, though, this lacuna is problematic. Any successful account of the neuromechanisms responsible for semantic memory needs to provide an explanation of how the conceptual system and the language system interact. Pulvermüller (2013, 2017) proposes that distributed linguistic representations serve as anchors in the formation of action perception circuits (APCs). Learning a language, on this account, leads to the formation of these distributed circuits by means of both Hebbian and anti-​Hebbian learning

120  Abstract Concepts and the Embodied Mind mechanisms. In other words, linguistic forms serve as a means of stabilizing and organizing grounded representations. This proposal raises a question: Does this anchoring alter or transform our concepts? Some accounts of the role of language in thought would seem to answer this question negatively. For instance, it has been proposed that linguistic forms merely serve as symbolic placeholders for multimodal simulations (Zwaan, 2016). On this approach, labels merely underwrite our capacity to use language as a heuristic shortcut that can be deployed when conditions do not require complex task performance. There are at least three problems facing such a deflationary approach. First, as symbolic placeholders, linguistic representations likely facilitate generalization. For instance, behavioral data suggest that verbal cues (such as the spoken word dog) activate more general representations than do nonverbal cues (such as the sound of a dog barking; Edmiston & Lupyan, 2015; Lupyan & Bergen, 2015). In keeping with this, the label-​feedback hypothesis (LPH) proposes that labels actively modulate both perceptual and conceptual processes (Lupyan, 2012a). Lupyan and Bergen (2015) propose that language acts as a control system that “programs the mind” by enabling the active manipulation of sensorimotor representations. Second, because words are public and shareable symbols, they enable us to leverage the social character of language. Several philosophers of language have emphasized the degree to which the linguistic function of natural kind terms such as water or gold depend in part on their ability to track referents by means of socially determined causal links (Burge, 1973; Kripke, 1980). Although this literature has focused largely on referents that we directly experience, abstract concepts would seem to be particularly good candidates for this sort of external support. Given that we cannot directly experience atoms, genes, or tectonic plates, it seems likely that we would rely more heavily on socially determined causal links to refer to these things. The presence of a shareable label would facilitate the formation of such links. Third, the world’s roughly 6,500 languages (Hammarström, 2016) exhibit a great deal of cross-​linguistic variation in the meanings of morphemes, words, and constructions (Bohnemeyer, 2021; Evans, 2011; Majid, 2015; Malt & Majid, 2013; Moore et al., 2015). Kemmerer (2022) argues that the combination of this cross-​linguistic semantic diversity with the fundamental assumptions of grounded cognition entails a form of linguistic relativity. The morphemes, words, and constructions of a particular language “provide learners with invitations to build brand new semantic structures—​a process

Language Is a Neuroenhancement  121 that is often challenging because the sought-​after meanings are often elusive, peculiar, and far from universal” (p. 16). Theories of grounded cognition hold that these structures are encoded, at least in part, in the reuse of experiential representations. This knowledge is likely to be engaged on some occasions by the relevant kinds of entities, events, and states. In other words, language-​specific meanings are likely to influence certain nonlinguistic cognitive tasks. This amounts to a form of linguistic relativity. Verbal labels may enhance and transform our conceptual abilities. They may serve as anchors for grounded concepts. In addition, they may serve as online modulators of categorization processes. These functional roles may provide an explanation for the cross-​linguistic diversity of lexical concepts.

6.5  Word Associations The statistical patterns of words and larger chunks of language are a rich source of information about the world and its contents (Davis & Yee, 2021; Günther, Rinaldi, & Marelli, 2019). In this section, I explore the idea that the conceptual system leverages this aspect of linguistic experience. Several lines of research suggest that abstract concepts lean heavily on this sort of information. Distributional models treat concepts in terms of knowledge of statistical patterns derived from spoken and written language. Several computational models that extract statistical regularities from large corpuses have been developed, including the latent semantic analysis model (Landauer & Dumais, 1997; Landauer, Foltz, & Laham, 1998), the hyperspace analog to language model (Lund & Burgess, 1996), and the latent Dirichlet allocation model (Blei, Ng, & Jordan, 2003). The core idea behind these models is that the meaning of a word is in part constrained by the company it keeps (Firth, 1957). They assume that semantic relatedness can be constrained by information gleaned from aggregating the linguistic contexts in which a given word appears. These models have been shown to perform remarkably well on lexical access and lexical similarity tasks (Louwerse, 2011, 2018). Working from the observation that linguistic and nonlinguistic experiences appear to be independent yet complementary sources of information about the world, several researchers propose that we should adopt a hybrid approach that combines grounded simulations and distributional knowledge (Andrews, Frank, & Vigliocco, 2014; Davis & Yee, 2021; Günther

122  Abstract Concepts and the Embodied Mind et al., 2019; Louwerse & Jeuniaux, 2010; Riordan & Jones, 2010). Working within a statistical learning framework, for instance, Andrews, Vigliocco, and Vinson (2009) suggest that language-​based distributional data are likely to be more helpful with abstract concepts than nonlinguistic experiential data. Hypothesizing that the most effective semantic representations would involve the statistical combination of both types of data, they developed a model that combined them. This model’s performance on several cognitive tasks correlated with previously gathered behavioral evidence better than models that exclusively relied on data of a single type (for further evidence of the advantages of hybrid models, see Bruni, Tran, & Baroni, 2014, and Steyvers, 2010). The success of this integrated model and others like it suggests that the ability to take advantage of the distributional information contained within natural language would be a useful enhancement to an experientially based conceptual system (Davis & Yee, 2021; Günther et al., 2019). The acquisition of color concepts by congenitally blind people provides a particularly strong real-​world case of the potential importance of distributional information (Bi, 2021; Lupyan et al., 2020). Such individuals have been shown to have a remarkable understanding of color space and the color of objects (Dimitriva-​Radojichikj, 2015; Lenci et al., 2013; Shepard & Cooper, 1992). In a recent study examining the contributions of sensory experience to concepts by comparing the semantic knowledge of verbs associated with vision gained by congenitally blind individuals with that of sighted individuals, Bedny and colleagues (2019 p. 114) conclude, the best explanation of our results is that sighted and blind individuals share both (i) relevant knowledge of the meanings of “visual verbs,” and (ii) common pragmatics, that lead them to interpret the request for similarity judgments in terms of the relevant respects for this domain (e.g., modality, temporal duration, etc.).

The fact that congenitally and early blind individuals can acquire a sophisticated understanding of color concepts in the absence of direct sensorimotor experience fits poorly with industrial-​strength variants of grounded cognition. It resonates, however, with those accounts that emphasize the importance of linguistic experience and, in particular, the richness of language as a source of information about the world. In keeping with this, researchers have shown that it is possible to recover a significant amount of information about

Language Is a Neuroenhancement  123 color from the distributional structure of color language (Kim, Elli, & Bedny, 2019; Lewis, Zettersten, & Lupyan, 2019). A recent study finds that a region of the left dorsal anterior temporal lobe (ATL) supports the knowledge of object colors in congenitally or early blind participants and sighted controls (Wang et al., 2020). The picture of concepts that emerges from this section bears a striking resemblance to a recent proposal motivated by a commitment to associationism. Tillas (2015) argues that language serves “as grist to the mill of cognition” and defends a position that he refers to as LASSO. The LASSO hypothesis identifies three important types of associations: (1) associations between a word and a concept that is grounded in sensorimotor systems, (2) associations between grounded concepts, and (3) associations between words. The LENS theory is a broad account of the role of language in embodied cognition and thus leaves room for nonassociative links. Nevertheless, it is clearly committed to the importance of each of these associative sources of information. There is a key point of disagreement between the LENS and LASSO theories, though: while the LASSO theory maintains the strict separation the language and conceptual systems, the LENS theory offers an integrated account.

6.6  Syntax A natural language is a structured symbolic system consisting of stored lexical items and combinatorial rules or principles (Jackendoff, 2007). Some cognitive scientists have proposed that these structural properties reflect those of an underlying amodal representational system referred to as a language of thought (Fodor, 1975). Supporters of grounded cognition, however, deny the existence of an amodal mentelese. This raises the question of how the structural properties of language relate to those associated with nonlinguistic cognition. Linguistic symbols are syntactically re-​combinable. This fact underwrites some of the core properties of the language system. For example, the ability to recursively combine symbols in accordance with an implicit understanding of rules or principles explains the inherent productivity of language (Pinker, 1994). Re-​combinability also explains the systematicity of language (Fodor, 1975; Fodor & Pylyshyn, 1988; Pinker, 1994). The idea behind systematicity is that our ability to produce and comprehend a sentence such as Elmo sings

124  Abstract Concepts and the Embodied Mind to Big Bird is intimately connected to our ability to produce and comprehend structurally analogous sentences such as Big Bird sings to Elmo. Finally, syntactic re-​combinability seems to contribute to the stimulus independence of language (Chomsky, 1966). We need to be careful here, though, because it is far from clear that thinking without words must involve noncombinatorial representational systems. To the contrary, animal cognition is likely be productive, systematic, and stimulus-​ independent. Certainly, supporters of perceptual symbols have argued that these properties can be captured by nonlinguistic embodied representations grounded in action, emotion, and perception systems (Barsalou, 1999; Barsalou & Prinz, 1997). Given this, any argument in support of the importance of language as a novel representational format should not depend on the mere fact that language is a combinatorial system but should rather focus on the specific combinatorial properties exhibited by language. In a theoretical exploration of animal and human cognition, Camp (2009) proposes that stimulus-​independence and re-​combinability should be taken to be degree properties. She proposes that natural language enhances the combinatorial properties of nonlinguistic cognition in at least four ways. First, natural language is likely to lead us to consider more thoughts because it enables one to hear the thoughts of others. Second, the semantic arbitrariness of linguistic symbols makes it easier to reproduce the same (or, at least, similar) thought in different situations. Third, the transparent syntactic structure of natural language highlights the potential re-​combinability of thoughts and thus encourages us to entertain thoughts that we might not have considered otherwise. Finally, natural language provides a sufficiently rich expressive medium to allow one to represent truth values and inferential relations among thoughts. Camp contends that these enhancements mean that a creature with language is likely to enjoy a general cognitive advantage over a creature that does not have language. Camp’s ideas can be applied to grounded cognition. The suggestion then becomes that the stimulus independence and re-​combinability of language differs in degree from nonlinguistic grounded representational systems. Other researchers have recently offered similar proposals. For instance, Tillas (2017, p. 107) proposes that language facilitates our capacity to form endogenous thoughts, which are, in his words, “thoughts that we activate in a top-​down manner or in the absence of the appropriate stimuli.” Clearly, this mirrors Camp’s suggestion that language may offer a greater degree

Language Is a Neuroenhancement  125 of stimulus independence. Similarly, Lynott and Connell (2010) propose that conceptual combination arises from the interaction between the linguistic and simulation systems. This fits with Camp’s suggestion that the re-​ combinability of language may encourage novel thoughts. Syntactic properties may also support inferential reasoning. To drive home this point, Weiskopf (2010) asks us to consider Chomsky’s famous sentence, Colorless green ideas sleep furiously. While it is true that linguists have differed over the degree to which this sentence is semantically or syntactically deviant (Harris, 1993), there is little question that it is difficult to perceptually simulate. Nevertheless, from its linguistic structure alone, we can infer that if this sentence is true then the relevant ideas are colorless and green, and they sleep furiously. In a related vein, Shallice and Cooper (2013) argue that the syntactic properties of recursion and argument role filling support certain computational abilities that are beyond the reach of nonlinguistic representational systems. Following the general presuppositions of the research literature, I have been treating concepts as a unified phenomenon. However, some have suggested that we should distinguish between different types of semantic competence. Marconi (1997) proposes that we should distinguish between referential and inferential semantic competence. Referential competence mediates our capacity to connect words to objects and events in the world. Inferential competence, on the other, hand mediates our capacity to connect words to other words as part of inferential reasoning. It depends crucially on abilities that enable us to verbally express ourselves and understand what others are saying. Calzavarini (2017, p. 163) explains: Such “intralinguistic” abilities are semantic because, in order to exercise them, a speaker must possess an internalized network specifying semantic connections between a given word (e.g., cat) and other words of a natural language (e.g., animal, meow).

Reviewing the extant neuropsychological and neuroscience data, Calzavarini (2017) finds evidence that our referential and inferential abilities are supported by (at least partially) distinct neuroanatomical regions. While the LENS theory is not committed to the viability of the referential–​inferential distinction, one of the core desiderata of any theory of concepts is to explain how concepts support inferences. Given this, the suite of abilities associated with “inferential semantic competence” are within the theoretical purview of

126  Abstract Concepts and the Embodied Mind the LENS theory. Indeed, it proposes that the language system, in virtue of its symbolic features, supports such processes.

6.7  Conversations Linguistic communication relies on more than just the syntactic combination of words in sentences. Much of it depends on our capacity to engage in conversations within specific contexts. Barsalou (2009, 2012, 2016b) ties conceptual grounding to the simulation of situated actions. Conversations would certainly seem to be an important subclass of these. They are not only an efficient means of gathering information about the world, but also the medium of many of our social interactions. Given their central importance to human experience, any theory that ties conceptual representation to experiential systems should at least entertain the possibility that simulations of situated linguistic actions contribute to the representation of semantic knowledge. Conversations are a kind of cooperative joint activity (Pickering & Garrod, 2021). Even when there are important disagreements between participants or divergent motivations for interlocuters, successful contributions to a conversation are constrained by their degree of relevance to shared communicative goals (Grice, 1989; Sperber & Wilson, 1986). Pickering and Garrod (2021) explore the core elements of cooperative joint activity through an example: they consider the interactive dynamics of two people assembling a piece of flat-​pack furniture. This task encourages participants to coordinate their actions in such a way that they help each other. One person might hold two boards together while the other person turns the screws joining them. This example enables Pickering and Garrod to introduce the notion of a shared workspace that encompasses both the relevant objects and the behaviors of the individuals. Using their terminology, dialogues are also joint activities and should also be understood in terms of a shared workspace. An important consequence of the interactive nature of dialogues is that they engender conceptual alignment in addition to behavioral alignment. This conceptual alignment is likely one of the reasons that conversations can play such a central role in the acquisition of conceptual knowledge. In particular, the combination of a shared workspace and an active process of alignment may well enable individuals to overcome well-​known problems

Language Is a Neuroenhancement  127 concerning the fixing of semantic reference (Quine, 1960). There is good reason to think that dialogues are an important contributor to our capacity to acquire rich semantic knowledge—​with the caveat that the existence of nonverbal individuals suggests that rich semantic knowledge in the absence of linguistic communication is possible. The observation that social interaction tied to language use is central to human learning is certainly not a new one. What is important in the context of conceptual grounding is answering the question of whether or not this interaction leads to the formation of concepts that are tied to linguistic actions and the representations associated with them. An obvious reason to think that conversation-​ based knowledge contributes to the representation of abstract concepts is that several abstract concepts refer directly to aspects of language use that are directly tied to conversations. Consider the concepts assert, cajole, deny, lie, and promise. Each of these refers to a specific type of speech act (Searle, 1969). Each of these concepts involves semantic knowledge pertaining to how we communicate through language. Given this, it seems likely that they are grounded in situations tied to linguistic communication. Concepts are collections of knowledge that generate and support inferences and predictions. From a grounded perspective, the representation of this knowledge is spread across multiple action, emotion, and perception systems; it is likely to encode information associated with exteroceptive-​ motor and interoceptive-​ motor dependencies—​ how our actions are likely to influence objects and our inner experiences, respectively. The LENS theory predicts that pragmatic knowledge associated with producing and comprehending utterances within the context of a conversation—​or, if you prefer, a dialogue—​plays a particularly important role in many abstract concepts. Consider an abstract concept such as democracy. The idea is that learning about democracies involves learning about how to talk about them. Verbal utterances are an important part of the shared workspace in conversations relating to this topic, and our semantic knowledge of this category partly depends on our ability to generate speech about category exemplars. In keeping with a grounded view, democracy should rely on the neural reuse of experiential circuits within the brain’s language system. The LENS theory claims that language contributes to the flexibility of our concepts in two different but equally important ways. First, the symbolic nature of linguistic expressions facilitates our ability to capture concepts that

128  Abstract Concepts and the Embodied Mind encode objects, relations, and events that occur in diverse situations and are more distributed across time and space (Davis, Altmann & Yee, 2020; Wilson-​Mendenhall et al., 2013). As Yee (2019) aptly puts it, words provide a source of invariance for concepts with experientially diverse content. The grounding of linguistic expressions in sensorimotor systems enhances our ability to draw connections between various situations in ways that we might not be able to do otherwise. Second, knowledge tied to our ability to produce and understand conversations plays an important role in our capacity to fit a concept to a particular context (Chow et al., 2014). Pragmatic and discourse-​ related knowledge indigenous to the language system helps us tune our concepts to current circumstances. We can see these two roles at work in an example borrowed from Goldstone and Kersten (2003). Imagine hearing the term buffalo paper for the first time. How would you interpret it? Several possibilities might come to mind. It could refer to paper in the shape of a buffalo, a particularly coarse type of paper, gift paper covered in buffaloes, or perhaps an essay on buffalo. You might even think of a more fanciful alternative such as sticky paper used to catch buffaloes (in a manner analogous to fly paper). Or, being familiar with buffalo plaid, you might think of paper covered in that pattern. Our capacity to generate these options is presumably tied to several factors. The symbolic nature of English enables us to join the words buffalo and paper together. Significantly, however, the meaning of the combined expression is not immediately transparent. Instead, we are forced to generate a series of possibilities. While these possibilities are clearly linked to the concepts buffalo and paper, they are also somewhat open-​ended. Generating them involves a consideration of why one might put these words together—​to what communicative purpose this linguistic act might serve. In other words, understanding novel concepts may depend in part on a familiarity with and understanding of their potential role in discourse and how they might relate to linguistic structures. Traditionally, theories of concepts have posited fixed conceptual cores (Machery, 2015). From this perspective, the task of deriving a contextually appropriate interpretation of a word used in context involves separate mechanisms that enable the selection among stored alternatives. As we have seen, some contemporary grounded accounts of concepts, including the LENS theory, adopt a contextualist view of concepts. From this perspective, the realization of an individual concept varies with task and context. This

Language Is a Neuroenhancement  129 requires a greater interaction between mechanisms associated with semantic processing and those associated with semantic control. Given that abstract concepts go beyond our direct experience and are thus especially flexible, conversation-​based knowledge is likely to be especially relevant for them. Conversations provide a means by which we can dynamically coordinate the content of abstract concepts. The LENS theory proposes that discourse-​related knowledge shapes the context-​and task-​specific realization of our concepts. Conversations provide a means of examining and refining our conceptual knowledge. Rehearsing conversations with others and self-​directed inner speech may help us fine-​tune and adjust our concepts (Clark, 2006; Kompa, 2019). These processes are relevant to all concepts, but they are more important for abstract ones because their content is more complex, heterogeneous, and temporally extended. There is a widespread recognition that inner speech serves an important metacognitive role in the development and maintenance of semantic knowledge (Alderson-​Day & Fernyhough, 2015). Because abstract concepts are less rooted in direct experience and more flexible with respect to context, we are more likely to depend on others to help us understand their content (Borghi et al., 2019). The metacognitive role of inner speech may help us secure the meanings of abstract concepts and alleviate the uncertainty that would tend to accompany their use. In particular, grounded mechanisms associated with producing and comprehending speech may play a predictive role within the realization of abstract concepts. All in all, there are good reasons to think that knowledge associated with our capacity to participate in conversations supports the flexibility of our abstract concepts by serving as a source of inner grounding (Borghi et al., 2019). Despite this, there has not been a great deal of direct investigation into the role of conversational knowledge. One reason for this is that most investigations of have examined semantic processing using experimental paradigms that ignore the dynamic interactive nature of conversations. In other words, they have adopted a kind of methodological solipsism (Fodor, 1991). Any meaningful examination of the influence of conversational mechanisms in our concepts must involve interactive paradigms (Dove et al., 2020). If the LENS and WAT theories are on the right track with respect to the importance of conversational knowledge, then future studies will need to examine the dynamic contribution of pragmatic and discourse-​related factors to the realization of abstract concepts.

130  Abstract Concepts and the Embodied Mind

6.8  Abstract Concepts in the Brain As I have emphasized throughout this book, our ability to acquire and use abstract concepts raises important empirical and theoretical questions for any approach to concepts that appeals to sensorimotor grounding. Because abstract concepts generally refer to objects and events that we do not directly experience, it is unclear how they can be captured by representations grounded in affective and sensorimotor systems (Chatterjee, 2010; Dove, 2009; Mahon, 2015). Despite the pressing nature of the problems posed by abstract concepts, the evidence for grounded cognition has tended to involve concepts that refer to objects and events that are intuitively connected to our sensorimotor experience (Kemmerer, 2015c). The LENS theory provides an account of how language augments a conceptual system that is grounded in action, emotion, and perception systems: the ability to manipulate linguistic symbols simultaneously transforms our cognitive niche and provides an effective means of processing that information. It agrees with the WAT theory that language serves as both a social and a cognitive tool (Borghi et al., 2019). The LENS theory provides an explanation of how the neural reuse of the language system enables us to encode and process aspects of our conceptual knowledge. Language is a neurocognitive enhancement and, as such, is likely to be involved in concepts of all sorts of stripes. However, we are more likely to rely on it in the processing of abstract concepts than in the processing of concrete concepts. In this section, I consider evidence suggesting that abstract concepts are represented in the brain differently than are other concepts. A significant portion of this evidence specifically implicates the language system. The idea that we process abstract concepts differently than other concepts is supported by a robust and diverse body of evidence that goes back decades. An important early indication of a possible functional asymmetry was the discovery of concreteness effects, in which concrete words exhibit several processing advantages over abstract words (Paivio, 1986; Wattennmaker & Shoben, 1987). In keeping with our discussion of the multiple measures employed to characterize the abstract–​nonabstract distinction (or spectrum), similar effects have been found with other measures. Consider, for instance, the measure of body–​object interaction (BOI), a measure meant to capture the perceived ease with which a human body can physically interact with category exemplars. Experiments reveal that words with higher BOI ratings are

Language Is a Neuroenhancement  131 processed more efficiently than words with lower BOI ratings (Siakaluk et al., 2008; Yap et al., 2012).6 Neuropsychological case studies provide further reason to think that the language system is particularly important for abstract concepts. As mentioned in Chapter 4, left hemisphere damage has been correlated with the degree of impairment for the processing of abstract words in studies of patients who present with aphasia (Goodglass, Hyde, & Blumstein, 1969), deep dyslexia (Franklin, Howard, & Patterson, 1995; Shallice & Warrington, 1975), and deep dysphasia (Katz & Goodglass, 1990; Martin & Saffran, 1992). A subset of patients with semantic dementia (SD), which primarily affects the anterior and inferior portions of both temporal lobes, exhibits reverse concreteness effects (Bonner et al., 2009; Reilly & Peelle, 2008; Yi, Moore, & Grossman, 2007). Reverse concreteness effects have also been found in patients with herpes simplex encephalitis, which disrupts the ATLs (Sirigu, Duhamel, & Poncet, 1991; Warrington & Shallice, 1984). A recent study of seven patients who had undergone a selective unilateral anterior temporal resection (four involving the right hemisphere and three involving the left hemisphere) compared their performance on semantic processing tasks to healthy controls and a group of patients with a more general semantic impairment associated with a selective amygdalo-​hippocampectomy (Loiselle et al., 2012). While both groups of patients exhibited impaired semantic processing when compared to the controls, only the patients with selective anterior lobe resection were particularly impaired with concrete words in comparison to abstract words. Brain imaging research suggests that there is a neurological asymmetry between the areas activated by abstract and concrete concepts. Participants in one study were visually presented with three nouns in the form of a triangle and asked to decide which of the two bottom nouns was most semantically similar to the top noun (Sabsevitz et al., 2005). Abstract nouns elicited greater activation in the left superior temporal and left inferior frontal cortex, while concrete nouns elicited greater activation in a bilateral network of association areas. In another experiment, participants read simple sentences 6 Interestingly, these sort of concreteness effects are not found with all the proposed measures. Controlling for age of acquisition, context availability, familiarity, imageability, and other variables, Kousta and colleagues (2011) found that abstract words have a reaction time advantage (rather than a disadvantage) over concrete words. They suggest that this reverse concreteness effect may be because, as a group, abstract concepts rely more on emotion.

132  Abstract Concepts and the Embodied Mind that contained pairs of abstract, concrete, or mixed (abstract-​concrete and concrete-​abstract) words (Sakreida et al., 2013). The fully abstract pairs engaged the left middle temporal gyrus while the fully concrete pairs engaged a fronto-​parietal network. Consistent with these findings, several studies find that abstract words elicit greater activation than concrete words in middle/​ superior regions of the left temporal lobe (Binder et al., 2005; Giesbrecht, Gamblin, & Swaab, 2004; Noppeney & Price, 2004; Sabsevitz et al., 2005) and regions of the left inferior frontal gyrus (LFG; Binder et al., 2005; Giesbrecht et al., 2004; Goldberg, Perfetti, & Schneider, 2006; Noppeney & Price, 2004; Sabsevitz et al., 2005). Although there is some variability in the imaging data (Binder, 2007), meta-​analyses find that these areas are the most likely to show increased activation with abstract concepts (Binder et al., 2009; Del Maschio et al., 2021; Wang et al., 2010, 2018). The neuroimaging findings are bolstered by an experiment carried out with healthy participants. Accuracy on a lexical decision task decreased with abstract concepts when repetitive transcranial magnetic stimulation (rTMS) was applied over the left inferior frontal gyrus and the left superior temporal gyrus, but accuracy decreased with concrete concepts when rTMS was applied over the right superior temporal gyrus (Papagno et al., 2009). The generalization that emerges from the neuroscientific evidence is that, when we compare the activity associated with abstract words to that associated with concrete words in various semantic tasks, the major brain areas that most reliably exhibit greater activation are the left middle/​superior ATL and the left inferior frontal gyrus. The left middle/​superior has been linked to high-​level speech perception and sentence comprehension, and the left inferior frontal gyrus (which includes Broca’s area) has been linked to several types of language processing, including auditory-​verbal short-​term memory (Kemmerer, 2015c).

6.9  Language and the Embodied Mind Zwaan and Madden (2005) use a pair of analogies to highlight the difference between traditional computational views of cognition and embodied ones: they liken the computational mind to a bricklayer who assembles structures out of well-​defined mental units and the embodied mind to a beachcomber who builds structures out of whatever has washed up on shore. While a beachcomber may shape and modify what he finds, much

Language Is a Neuroenhancement  133 of the original character remains. Grounded representations similarly preserve aspects of their experiential origins. Without making too much of the analogy itself, it is worth pointing out that one of the things that washes ashore is a collection of bricks (i.e., a natural language). A supporter of embodied cognition thus faces a choice: either maintain that the language system is separate from our conceptual system or provide some explanation of how the two are integrated. In this chapter, I have argued that language augments the embodied mind. The symbolic character of language offers several cognitive benefits. One of these is the general absence of a direct connection between linguistic representations and their referents. This arbitrariness (which may come in degrees) may help them anchor disembodied knowledge. An additional benefit arises from the fact that linguistic symbols are syntactically re-​combinable in a way that is independent of the combinatorial properties of nonlinguistic grounded cognition. This independent structural flexibility makes it easier to generate new thoughts and encode unexpected connections between thoughts (Lynott & Connell, 2010). The acquisition of a natural language alters a child’s cognitive purview by offering a new medium through which to capture experience. Experience with language leads to the development of a distributed neural system able to manipulate linguistic symbols in a compositional and productive fashion. The neurologically realized language system amounts to a distributed action–​perception control system that likely relies on hierarchically organized network hubs. Linguistic forms themselves are grounded because they involve actions, sights, and sounds, but they are free to capture content in a manner that is not tied to their grounding. The language system makes an important contribution to our capacity to acquire and employ concepts, particularly abstract ones.

7 Heterogeneity Researchers have employed several measures to trace the contours of the abstract–​ concrete distinction. These measures include body–​ object interaction (Siakaluk et al., 2008), concreteness (Marschark & Paivio, 1977), context-​availability (Schwanenflugel & Shoben, 1983), emotional valence (Kousta et al., 2011), imageability (Paivio, 1986), interoceptive strength (Connell, Lynott, & Banks, 2018), semantic richness (Recchia & Jones, 2012), and strength of perceptual experience (Connell & Lynott, 2012). The result of this research is a mixed picture: while these measures correlate up to a point, they are not equivalent (Kousta et al., 2011). While it could be that we have just not come up with either the correct understanding of abstract concepts or the most accurate measure, it is also possible that abstract concepts are heterogeneous (Barsalou, Dutriaux, & Scheepers, 2018; Borghi et al., 2018; Dove, 2016). This would not be that surprising. After all, researchers have found striking neuroanatomical differences between various types of concrete concepts (Forde & Humphreys, 2005). To give an example, a lot of effort has been spent assessing the hypothesis that concepts referring to artifacts and living things are handled by different brain regions (Warrington & Shallice, 1984). While abstract concepts have generally been treated as a unitary whole, there is no compelling a priori reason to think that they should not exhibit the same sort of variation as concrete concepts. Indeed, given their distal connection to experience and the fact that they can range from socially oriented and emotion-​laden concepts to philosophical and scientific concepts, there seem to be prima facie reasons to expect such variation with abstract concepts (Dove et al., 2020). In keeping with this, several recent studies have explored the rich organization of diverse types of abstract concepts (Catricalà et al., 2014; Desai, Reilly, & van Dam, 2018; Fingerhut & Prinz, 2018; Fischer & Shaki 2018; Ghio, Vaghi, & Tettamanti, 2013; Harpaintner, Trumpp, & Kiefer, 2018; Mellem et al. 2016; Roversi, Borghi & Tummolini, 2013). In this chapter, I argue that our ability to process abstract concepts of various stripes is dependent on the multidimensional nature of our Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0007

Heterogeneity  135 conceptual system. Different abstract concepts rely on different sorts of grounded representations. More specifically, they often differ in the degree to which they evoke emotional, interoceptive, and linguistic systems. Because of this variability, abstract concepts exhibit synchronic flexibility.

7.1  Affective Embodiment Some researchers have sought to find evidence suggesting that abstract concepts—​or at least certain classes of them—​tend to be grounded in different experiential systems rather than concrete concepts. This effort is part of divide-​and-​conquer defense of grounded cognition. The strategy is straightforward: examine a particular class of abstract concepts and then demonstrate that they engage experiential systems different from those of most concrete concepts. The cumulative success of this strategy would then provide inductive support for the claim that most, if not all, abstract concepts are grounded. We can see this strategy at work in recent work examining the role played by the emotions in abstract concepts.

7.1.1  The Case for the Emotions Strikingly, many of the experiments cited in support of conceptual grounding fail to explicitly manipulate emotionality. This lacuna is surprising because many theories hold that concepts are partially grounded in affective systems (e.g., Barsalou, 2008; Fischer & Zwaan, 2008; Gallese & Lakoff, 2005). Efforts to address this oversight have produced evidence implicating the emotions in abstract concepts. Much of the recent interest in the possible importance of the emotions to abstract concepts is due to the discovery of emotionality effects (which are analogous in many ways to concreteness effects). Kousta et al. (2011) found that abstract words have a statistical reaction time advantage over concrete words in lexical processing tasks (reversing the usual concreteness effect). Clearly, this advantage cannot be explained in terms of the traditional explanations of facilitation effects—​that is, by either the dual code theory or the context availability theory. They theorize that faster reaction time might be due to the overall tendency for abstract concepts to have a greater emotional content. This proposal would be in keeping with the finding

136  Abstract Concepts and the Embodied Mind that emotional valence, regardless of whether it is negative or positive, facilitates lexical processing (Kousta, Vinson, & Vigliocco, 2009; Newcombe et al., 2012) The affective embodiment account (AEA) holds that emotion systems play a dominant role in the grounded representation of abstract concepts. Support for this view is provided by a regression analysis of a diverse set of 1,446 English words, which finds that abstract words have a general tendency to have more affective associations than concrete words (Vigliocco et al., 2014). Abstract words also tend to engage the rostral anterior cingulate cortex (rACC)—​an area associated with emotion processing—​to a greater extent than do concrete words (Vigliocco et al., 2014). The suggestion on offer is that much of the distinct behavior of abstract concepts can be explained by their emotionality. The observed behavioral and neurological differences between abstract and concrete concepts can largely be explained in terms of the fact that abstract concepts tend to be more grounded in emotion systems.

7.1.2  A More Complicated Picture The issue of scope immediately comes to mind. Even if it is true that emotions play a causal role in the processing of some abstract concepts, it is not immediately clear that they are central to all of them. Even from an intuitive perspective, there are reasons to be concerned about the explanatory reach of the AEA. Although abstract concepts as a group may have more emotional content than concrete concepts, many abstract concepts do not appear to be grounded emotional experience (e.g., mathematical and scientific concepts, not to mention the closed-​class morphemes discussed in Kemmerer, 2019). Worse, any involvement of the emotions in these concepts seems likely to be highly variable and inconsistent. It’s hard to imagine that the square root of negative 1 or the ideal gas law evoke similar emotional responses in all students who encounter them. A recent functional magnetic resonance imaging (fMRI) study employing high resolutions found that different subregions of the anterior temporal lobe (ATL) responded to emotional valence and abstractness (Wang, Wang, & Bi, 2019). Neuroimaging also appears to back up the claim that certain types of concepts are less reliant on the emotions. A recent meta-​analysis of brain imaging studies finds that the cortical activation patterns elicited

Heterogeneity  137 by numerical and emotional concepts differ from those elicited by other abstract concepts (Desai et al., 2018). Given this, it seems reasonable to question the degree to which the distinction between abstract and concrete concepts can be reformulated in terms of the contribution of the emotions (Dove, 2014; Shallice & Cooper, 2013). Some of the specific details of the studies cited in support of the AEA also raise questions for the theory. For example, these studies generally match concrete and abstract words on imageability. Emotion words, though, tend to be rated as both abstract and imageable (Altarriba & Bauer, 2004). Matching stimuli on imageability may thus generate an unrepresentative set of highly imageable abstract words (Skipper & Olson, 2014). These studies also hold age of acquisition constant, which could be problematic because abstract concepts are often acquired later in development than concrete concepts (Barca, Mazzuca, & Borghi, 2017; Borghi et al., 2018). One variable not controlled for in these studies is hedonic valence. Indeed, the abstract words used in them were rated as higher in valence than the concrete words. This raises the question of whether valence is a confound of concreteness. Skipper and Olson (2014) controlled for these variables and made two important findings: (i) the rACC responded to emotional valence, and (ii) it responded more to concrete concepts than abstract ones (reversing the effect observed by Vigliocco et al., 2014). The story is also complicated by the apparent relevance of action systems to the processing of abstract emotion words. Abstract emotion words elicit increased activity in the same areas activated somatotopically by face words (Dreyer & Pulvermüller, 2018; Moseley et al., 2012). Given that facial movements are important for emotion expression and detection, we have reason to think that they might play a role in the simulation of emotion concepts. There is evidence that the motor system is causally relevant to the semantic processing of these concepts. For example, one case study found that a patient with a focal brain lesion in the supplementary motor cortex was selectively impaired in abstract-​emotional word processing (Dreyer et al., 2015). People with an autism spectrum condition (ASC) often demonstrate alexithymia—​a difficulty with identifying and describing emotions in oneself and in others (Tager-​Flusberg, 1992). An event-​related fMRI experiment involving high-​functioning individuals with an ASC found selectively reduced activation in motor areas for emotion words when compared with matched control words (Moseley et al., 2015). Although this evidence is merely correlational, the observation

138  Abstract Concepts and the Embodied Mind that individuals who are members of a population that generally struggles with emotion words exhibit hypoactivation of the motor system while processing them is suggestive. A different alternative explanation for the data used to support the AEA is also available. Lenci, Lebani, and Passaro (2018) point out that the affective content of abstract concepts could be a byproduct of co-​occurrence statistics. In support of this hypothesis, they show that abstract words in Italian (even ones that do not have high emotive value) tend to co-​occur with contexts of higher emotive values. This evidence raises the possibility that linguistic associations may also play an important role in abstract concepts—​including those with affective content. The possible links between abstract concepts and both the motor system and emotion-​laden verbal contexts provide some initial support for a multimodal approach involving distributed neural circuits. Emotions remain a somewhat neglected factor in abstract concepts, but they are unlikely to be a panacea for the broader challenges that abstract concepts pose to embodied cognition. Two conclusions can be drawn from our critical evaluation of the AEA. First, there are compelling reasons to think that the emotions are causally relevant to how we encode and process abstract concepts. Second, there are compelling reasons to doubt that the emotions play the dominant role in abstract concepts.

7.2  Dimensions of Variation Rather than support a universally quantified, single-​factor explanation of the differences between abstract and concrete concepts, the evidence just reviewed points to the multimodal nature of abstract concepts. This raises an important question: Do different types of abstract concepts rely on different forms of grounding? Here, I survey several bodies of evidence suggesting that this is the case.

7.2.1  Semantic Richness Several studies have found that words with more associated semantic information are processed more efficiently in word recognition tasks than are words with less associated semantic information (for a review, see Pexman, 2012). Such semantic richness effects have been found in relation to several

Heterogeneity  139 dimensions (e.g., Buchanan, Westbury, & Burgess, 2001; Duñabeitia, Avilés, & Carreiras, 2008; Grondin, Lupker, & McRae, 2009; Schwanenflugel & Shoben, 1983), and this has led some researchers to offer multidimensional accounts of conceptual structure (e.g., Pexman et al., 2008; Yap et al., 2012). Recently, this approach has been extended to abstract concepts. For instance, Zdrazilova and Pexman (2013) examined the effects of six semantic-​ richness variables (context availability, sensory experience rating, emotional valence, emotional arousal, semantic neighborhood, and number of associates) on a lexical decision task and a semantic categorization task. They found that the effects varied with the task: context availability facilitated lexical decision, and both sensory experience and positive valence facilitated semantic categorization. They conclude that abstract meaning might be in part grounded in “situations, emotions, and sensory experience” (p. 1316). In another study, Moffat and colleagues (2015) compared the effects of emotional experience and context availability on a semantic categorization task. Emotional experience facilitated the processing of abstract words but inhibited the processing of concrete words, whereas context availability facilitated both. One of the important features of the semantic-​richness research is that the effects associated with different dimensions are often task-​dependent (Hargreaves & Pexman, 2012; Moffat et al., 2015; Muraki, Sidhu, & Pexman, 2020; Siakaluk, Knol, & Pexman, 2014; Zdrazilova & Pexman, 2013). Early in the exploration of concreteness effects, there was an intense debate about whether they could be explained exclusively by either context availability or dual coding. Following this debate, evidence emerged that both might be relevant (e.g., Holcomb et al., 1999; Jessen et al., 2000). Echoing this shift toward a pluralistic explanation of concreteness effects, the careful examination of semantic richness has shown it to be a multifaceted phenomenon. In the end, this research provides further evidence that abstract concepts are grounded in multiple systems because sensorimotor experience, emotionality, and context availability all seem to contribute to semantic richness effects within specific contexts.

7.2.2  The Situated Conceptualization Framework One approach to concepts that shares many of the core commitments with the elastic approach advocated in this book proposes that abstract concepts

140  Abstract Concepts and the Embodied Mind involve situated simulations (Barsalou, 1999, 2009).1 According to this approach, conceptual simulations do not occur in isolation but instead include background settings, events, and even introspective responses. For example, instances of the concept bicycle would involve the simulation of our experience with bicycles in relevant real-​world situations. The situated conceptualization framework builds on a number of basic assumptions (Barsalou et al., 2018). The first is that the core function of the conceptual system is to facilitate situated action by means of processing situations. A situation is defined as (p. 2), “a setting where agents encounter other agents, objects and events, which activate goals, affect and bodily states, leading to actions and outcomes.” Situated conceptualizations enable us to categorize these elements, make predictions, and ultimately carry out appropriate actions. Second, situated conceptualizations become encoded in long-​term memory. Third, encountering new situations often activates relevant stored situated conceptualizations, which in turn enable the construction of multimodal simulations. Fourth, stored situated conceptualizations underwrite our capacity to generate simulations of imagined counterfactual simulations. Early support for this framework was provided by a study in which participants generated typical properties for three abstract concepts (truth, freedom, invention), three concrete concepts (bird, car, sofa) and three intermediate concepts (cooking, farming, carpeting; Barsalou & Wiemer-​Hastings, 2005). This study had two core findings: (1) participants generated situational properties with both concrete and abstract concepts, and (2) participants tended to generate more event and introspective properties with abstract concepts. The authors proposed that abstract and concrete concepts are generally associated with different aspects of situations: abstract concepts tend to focus on social aspects while concrete concepts tend to focus on physical entities and actions. In a more fully realized experiment employing similar methodology, participants tended to produce fewer entity properties, more introspective properties, and more relational properties with abstract than with concrete concepts (Wiemer-​Hastings & Xu, 2005).

1 Both approaches are committed to an account of the conceptual system that is flexible, multimodal, and multilevel. They differ in at least three major respects: in contrast to the situated conceptualization framework, the approach developed in this book views grounded language as an important part of our conceptual system rather than a superficial shortcut, leaves room for amodal representations to play a limited role, and is not committed to an explicit theory of situations.

Heterogeneity  141 While this approach is promising, there is a sense in which it merely pushes the problem up a level. It seems likely that there are important differences in the types of situations associated with abstract and concrete concepts (Davis, Altmann, & Yee, 2020). There is no obvious a priori reason to assume that the relevant concepts tied to these disparate situations can be captured solely by means of grounded representations. This raises the question of whether an appeal to situated simulations reflects a unified theoretical solution to the problems posed by abstract concepts or simply a relabeling of them. There are reasons to think that other systems may be involved. One study examined the conceptualization of actions at different levels of abstractness (Spunt, Kemmerer, & Adolphs, 2016). The engagement of a set of left hemisphere regions increased as the abstractness of the descriptions increased. These regions overlap with functional networks that have been associated with mental state reasoning, the default mode network, episodic memory, and abstract concept processing. These regions are far removed from sensorimotor regions. Recently, Barsalou, Dutriaux, and Scheepers (2018) have argued that we need to abandon the distinction between concrete and abstract concepts because it is no longer useful. Part of their reasoning depends on the recognition that concrete concepts can vary greatly in how they are grounded (Kiefer & Barsalou, 2013). They propose that the abstract concepts are likely to vary in a similar way. An expanded property generation study provides some support for this hypothesis (Harpaintner et al., 2018). Participants listed properties for 296 abstract concepts, and hierarchical cluster analyses revealed three main subgroups: a cluster dominated by sensorimotor features (e.g., observation and fitness); a second dominated by internal states, emotions, and social relations (e.g., nightmare and argument); and a third dominated by verbal associations (e.g., theory and dignity). Figure 7.1 portrays the results of the hierarchical cluster analyses. Barsalou, Dutriuax, and Scheepers go on to suggest that the concrete–​ abstract distinction should be replaced by two different distinctions that would apply to all types of concepts: the distinction between external and internal situational elements and the distinction between situational elements and situational integrations. This proposal fits with other attempts to disentangle a general problem of abstraction or generalization from the bare distinction between concrete and abstract concepts (Dove, 2016; Myachykov & Fischer, 2019). The recognition that some concepts rely more heavily on our capacity to integrate situations fits well with the evidence cited earlier in support of the need for hierarchical representation.

(a)

Cluster 1

Cluster 2

Cluster 3

Cluster 1

(b)

Cluster 2

Cluster 3

Averaged relative frequencies

1.00

0.75

0.50

0.25

0.00 SC

IS/E

SM

VA

SC

IS/E SM Category

VA

SC

IS/E

SM

VA

Figure 7.1  Results of hierarchical cluster analyses of feature types, derived from property listings of 296 abstract concepts (Harpaintner, Trumpp, & Kiefer, 2018). (a) Dendrogram visualizing the k =​3 cluster solution. The different clusters are marked by different colors. (b) Box plots depicting generated features per cluster. SM, sensorimotor feature; IS/​E, internal state/​emotion; SC, social constellation; VA, verbal association.

Heterogeneity  143

7.2.3  Multidimensional Approaches A new research program has emerged in grounded cognition that builds on the mixed results of the divide-​and-​conquer strategy. This program begins with the recognition that abstract concepts are not a unitary phenomenon. Troche, Crutch, and Reilly (2014; see also Crutch et al., 2013) provide an early demonstration of the promise of a multidimensional approach to abstract concepts. Rather than rely on an intuitive notion of abstractness, they investigated how the meanings of 400 concrete and abstract English nouns are distributed in a multidimensional space using hierarchical cluster analysis. Participants rated the nouns along 12 dimensions. Factor reduction yielded three latent factors that the authors characterize as affective association, perceptual salience, and magnitude. When the original words were plotted for these three factors, abstract and concrete words were associated with unique but somewhat overlapping topographies within this space. A more recent study provides a fuller picture of how multidimensional research might proceed. Villani et al. (2019) set out to assess how abstract concepts are represented in the minds of Italian speakers by using ratings involving several dimensions (see also Ghio et al., 2013). This study builds on two previous studies involving Italian: one (Barca, Burani, & Arduino, 2002) that provides norms for 625 Italian concrete nouns along several psycholinguistic (e.g., written frequency, spoken frequency, syllable length, and letter length) and semantic variables (e.g., familiarity, concreteness, imageability, and age of acquisition) and another (Della Rosa et al., 2010) that provides norms for 417 Italian words on seven dimensions, including mode of acquisition, concreteness, imageability, familiarity, age of acquisition, context availability, and abstractness. This study is unusual because it focuses exclusively on abstract concepts and because it examines these concepts along several novel dimensions suggested by the explicitly multidimensional word as social tool (WAT) theory. The researchers presented 425 abstract Italian words to 304 participants and asked them to rate them on a 7-​point scale along 15 dimensions. In addition to the typical dimensions of abstractness, concreteness, imageability, and context availability, they included dimensions related to sensorimotor experience (perceptual strength in five modalities and body–​object interaction), linguistic experience (age of acquisition and mode of acquisition), inner experience (emotionality, interoception, and metacognition), social experience (social valence and social metacognition), and the contributions of hand and mouth effectors due

144  Abstract Concepts and the Embodied Mind to previous research indicating that the former are more important for abstract concepts and the latter are more important for concrete concepts (for a review of this evidence, see Borghi et al., 2019). A principal component analysis (PCA) of the data identified a three-​ component solution. The authors dubbed these the concreteness/​abstractness, the sensorimotor, and the inner grounding and social components. With respect to the concreteness/​abstractness component, concreteness was explained primarily by the imageability, body–​object interaction, perceptual strength, and hand effector dimensions, and abstractness was explained primarily by the social metacognition, mode of acquisition, and age of acquisition dimensions. This pattern is consistent with the hypothesis that the more abstract a concept is the more likely we are to depend on others for an understanding of its content (Borghi et al., 2018). The sensorimotor component involved most of the relevant perceptual modalities, including the hand effector dimensions. Finally, the inner grounding and social component was linked to the emotions, interoception, metacognition, and the social dimensions. A cluster analysis then indicated four major kinds of abstract concepts. The first cluster consists of physical, spatiotemporal, and quantitative concepts (73 words). This cluster includes physical concepts (e.g., acceleration, color, and matter), concepts for bodily sensations (e.g., cold, shiver, and vertigo), concepts for quantities (e.g., dose, meter, and price), mathematical concepts (e.g., addition, number, and sum), temporal concepts (e.g., beginning, day, and season), and spatial concepts (e.g., area, destination, and space). It was characterized by relatively high ratings on the concreteness/​abstractness component. The second cluster are self and sociality concepts (81 words). This cluster includes positive trait concepts (e.g., ability, charm, and enthusiasm), and concepts related to social norms (e.g., conflict, kindness, and protest), social institutions (e.g., civilization, family, and job), and somewhat neutral social situations (e.g., crowd and party). It was characterized by higher ratings on the inner grounding and social component. The third cluster are philosophical and spiritual concepts (125 words). This cluster includes concepts for imaginary entities (e.g., fate, ghost, and magic spell), religious concepts (e.g., devotion, faith, and salvation), concepts for intellectual disciplines (e.g., history, linguistics, and philosophy), concepts related to decision-​making (e.g., analysis, logic, and reason), and, somewhat counterintuitively, negative trait concepts (e.g.,

Heterogeneity  145 dishonesty, greed, and rudeness). It was characterized by low scores on the sensorimotor and the inner grounding and social components. The fourth cluster are emotional and inner state concepts (146 words). This cluster includes emotion concepts (e.g., anger, hate, and wonder), mental state concepts (e.g., boredom, exasperation, and satisfaction), and concepts for highly emotional social situations (e.g., deceit, revenge, and scandal). It is characterized by high scores across the three identified components. This study suggests that abstract concepts are not defined relative to a single dimension. Although its findings do not match up perfectly with the findings from property generation studies (Barsalou & Wiemer-​Hastings, 2005; Harpainter et al., 2018; Wiemer-​Hastings & Xu, 2005)—​perhaps due to the large number of rating dimensions employed or to general differences in the experimental paradigms—​there is a clear sense in which all of these studies point to the importance of the emotions, inner grounding, social experience, and language to abstract concepts. In addition, they point to the existence of distinct subtypes of abstract concepts. The question then becomes whether there are studies that offer more direct evidence of the causal relevance of these different factors. Villani et al. (2021) employed a behavioral interference pattern to investigate this question. Participants were asked to rate the difficulty of three types of concrete concepts (natural objects, tools. and food concepts) and abstract concepts (philosophical and spiritual concepts; physical, spatio-​temporal, and quantity concepts; and emotional, mental state, and social concepts).2 They were randomly assigned to five conditions: a condition in which they were asked to squeeze a ball following the rhythm of a metronome, one in which they rhythmically chewed gum, one in which they rhythmically pronounced the syllable “ba” (a standard technique for articulatory suppression), one in which they monitored their heart beating, and a control condition in which they carried out the task without any interference. The results support two conclusions that are relevant for my purposes. First, when abstract and concrete concepts were compared, the difference in the difficulty ratings of the abstract concepts was greatest in the heart beating and gum chewing conditions. Exploratory analyses suggest that 2 The authors based the groupings of abstract concepts on their earlier study (Villani et al., 2019). For the purposes of this new study, they decided to collapse the emotional and inner state cluster together with the self and sociality cluster. They justify this action by appealing to difficulties in differentiating these subgroups.

146  Abstract Concepts and the Embodied Mind abstract concepts were judged more difficult in the heart beating condition than all other conditions and that concrete concepts were judged less difficult in the gum chewing condition than all other conditions. This pattern fits with the hypothesis that abstract concepts depend more on interoception and language than do concrete concepts. An important caveat to this conclusion is that articulatory suppression failed to affect abstract concepts more than concrete ones. Second, when the different kinds of abstract concepts were compared, some interesting differences emerged. The ball squeezing conditions increased the difficulty judgments with the physical, spatiotemporal, and quantity concepts to a larger extent than with the heart beating and gum chewing conditions. Moreover, the heart beating condition interfered more with the emotional, mental state, and social concepts and the philosophical and spiritual concepts than it did with the physical, spatiotemporal, and quantity concepts. These differences are consistent with the hypotheses that physical, spatiotemporal, and quantity concepts are relatively more concrete—​or at least grounded in manual action—​than the other two types and that emotional, mental state, and social concepts and philosophical and spiritual concepts are more dependent on interoception than physical, spatiotemporal, and quantity concepts. Desai, Reilly, and van Dam (2018) investigated the neural basis of four types of abstract concepts through neuroimaging meta-​analyses. They examined numerical concepts, emotion concepts, concepts associated with morality, and theory of mind concepts. Numerical concepts exhibited significant clusters of activation bilaterally in the intraparietal sulcus (IPS) and portions of the superior parietal lobule (SPL). These activations are consistent with the hypothesis that numerical cognition is grounded in spatial and grasping-​ related representations. Emotion concepts generated increased activations in areas associated with emotion processing and regulation. Concepts associated with morality and theory of mind showed a partial overlap with these emotion areas but also showed increased activation in a set of distributed areas often associated with concrete concept processing. Three main findings emerged from the analyses. First, the representation of abstract concepts is more widespread than is often assumed. Second, different types of concepts are represented in different ways. Third, several areas—​including inferior parietal, posterior cingulate, and medial prefrontal cortex—​were engaged by all the abstract concept types. These areas have been linked to the processing of concrete concepts. All this fits with a broadly componential view of concepts in which abstract concepts depend partially on grounded representations

Heterogeneity  147 associated with various affective, mental, and social experiences (Binder et al., 2016; Wang et al., 2018).

7.3  Multiple Mechanisms For several decades, researchers have explored the possibility that different types of concrete concepts might be handled by neuroanatomically distinct brain areas. Within neuropsychology, clinical studies of patients with various types of brain damage have examined the degree to which classes of concepts are particularly impaired and which are not. This research has inspired and guided brain imaging studies with neurotypical participants. As outlined earlier, a robust body of behavioral research points to the heterogeneity of abstract concepts. Building on this evidence and the core ideas of grounded cognition, a number of research labs have begun to look into the possible contributions of multiple brain mechanisms to the neural representation of different types of abstract concepts. The heterogeneity of abstract concepts fits naturally with a grounded approach to cognition for at least three reasons. First, grounded theories hold that conceptual knowledge is encoded in a distributed fashion across different affective and sensorimotor modalities. This allows for the possibility that different types of abstract concepts might be handled by different cross-​modal patterns; that is, some modalities might contribute more to the representation of certain abstract concept kinds than they might to others. Second, grounding connects concept representation to our experiences of category exemplars and the situations in which we encounter them. It seems likely that, for example, affective experiences are more likely to be relevant for certain abstract concepts than others. Third, conceptual grounding opens up a possible connection between how a concept is synchronically represented and how it is acquired. This suggests that differences in conceptual representation might emerge from different modes of acquisition. For example, certain abstract concepts might depend more heavily on language for their acquisition than others. Given these theoretical considerations and the evidence of heterogeneity derived from property-​listing and rating studies, it is not surprising that a research program examining the neural underpinnings of varieties of abstract concepts has emerged. While this research program is at an early stage of development, the initial results are promising.

148  Abstract Concepts and the Embodied Mind Researchers have begun to recognize that previous studies of abstract concepts are primarily designed to uncover the differences between abstract and concrete concepts (Dreyer & Pulvermüller, 2018). The search for contrasting activation patterns between these broad categories can lead investigators to miss overlapping ones. Studies that are not focused on these differences might well find interesting similarities between the activation patterns associated with both types. For instance, several neuroimaging studies implicate diverse experiential systems in abstract concepts. For instance, both concrete and abstract words have been found to elicit distributed activation in sensorimotor systems (Pexman et al., 2007). Pexman and colleagues also found that abstract words are associated with more widespread cortical activation, including temporal, parietal, and frontal regions previously associated with semantic processing. This pattern could be explained by the reliance of abstract concepts on association areas and fits with the distributed models of concepts (Binder et al., 2016; Wang et al., 2018). One recent study (Harpaintner et al., 2020) consciously builds on previous property listing data. This study compared the fMRI responses to matched words for abstract concepts empirically associated with motor features (e.g., fitness) and visual features (e.g., beauty) in a lexical decision task. It also employed two localizer tasks associated with motor activity and visual perception; fMRI responses were recorded when participants moved their hands and looked at pictures. Motor abstract concepts elicited greater activations in bilateral fronto-​parietal areas and subcortical areas, and these activations partially overlapped with those found in the hand movements. Visual abstract concepts, on the other hand, elicited increased activations in occipital and temporal areas, which partially overlapped with the activity associated with looking at the pictures.3 Several studies have uncovered intriguing links between specific types of abstract concepts and specific experiential systems. We have already discussed studies linking abstract emotion concepts with increased activity in the motor cortex (Dreyer et al., 2015; Dreyer & Pulvermüller, 2018; Moseley et al., 2013). Abstract physical concepts (energy, momentum, etc.) can be differentiated to some degree based on the degree to which they modulate activity in specific action and perception areas (Mason & Just, 2016). 3 A caveat is warranted here because the findings with the visual abstract concepts required using a more lenient statistical threshold (Harpainter et al., 2020).

Heterogeneity  149 Physical concepts related to periodicity (e.g., frequency) have been linked with increased activation in postcentral and parietal regions associated with performing rhythmic movements (Chen, Zattore, & Penhune, 2006). Small-​number concepts (1–​9) have been found to elicit activity in motor regions contralateral to the hand that participants would use for counting (Tschentscher et al., 2012). If we adopt a grounded view of abstract concepts, then we should expect them to involve distributed neural patterns that are connected to their experiential content. The evidence just outlined suggests that this is indeed the case. Abstract concepts can be differentiated in part by the distributed activation that they elicit in grounded systems much the way concrete concepts can be. Admittedly, the particular grounded systems tend to include those associated with emotions, inner experience, and language. Nevertheless, the point remains that abstract concepts appear to also rely on distributed experiential representations. As I emphasized in my earlier discussion of the evidence supporting grounded cognition, theoretical advances have been held back by an a priori commitment to strong embodiment. When we move to a more inclusive conception of grounding that leaves room for distributed and hierarchical representations, the issues become more nuanced. In keeping with this, some recent studies provide a picture of the richness of the neural underpinnings of abstract concepts. For example, one study comparing the fMRI responses elicited by the contextual processing of the individual abstract concepts convince and arithmetic (Wilson-​Mendenhall et al., 2013) found that the processing of the former generated increased activation in brain regions associated with mentalizing and social cognition and that the processing of the latter generated increased activation in areas associated with numerical cognition. A recent meta-​analysis of fMRI studies (Desai et al., 2018) sought to find the similarities and differences between the neural correlates of different types of abstract concepts. This study examined four broad types of abstract concepts—​numerical, emotion, moral, and mental concepts—​and generated three main findings. The first was that the activations elicited by abstract concepts were widespread, including not only affective and sensorimotor regions but also higher-​level heteromodal regions. The second was that the abstract concept types elicited specific activations that reflected the experiential qualities of their content. Number concepts elicited increased bilateral activation of the IPS, which is in keeping with the idea that they

150  Abstract Concepts and the Embodied Mind are associated with spatial and motor processing. Emotion concepts engaged areas associated with emotion processing and regulation, such as the left amygdala and right orbitofrontal complex (OFC). Moral and mental concepts elicited activation that overlapped somewhat with the emotions but also included higher-​level areas associated with the semantic processing of concrete concepts, including the angular gyrus (AG), the precuneus, and the ATL. The third finding was that all of the concept types evoked activations in common brain areas—​in particular, regions of the inferior parietal, posterior cingulate, and medial prefrontal cortex. These findings suggest that concepts are encoded by means of a complex interaction of experiential representations, association areas, and semantic hubs. Desai, Reilly, and van Dam (2018) propose that an event-​or situation-​based approach provides a compelling account of both the richness of abstract concepts and their commonalities. They write (p. 12) that, “processes that are common across these abstract domains appear to be related to event and relation processing, episodic recall, imagery and emotion processing, and are also triggered by concrete concepts.” This approach offers a multilevel view conceptual grounding that connects our concepts to experiential mechanisms without relying on a reductive view of embodiment that restricts our concepts to modality-​specific areas. A separate study applied multivariate pattern analysis to fMRI data to characterize the neural activation patterns associated with 28 individual abstract concepts (Vargas & Just, 2020). There were two important take-​aways from this study. First, the neural activations elicited by the 28 concepts were consistent enough that concepts could be reliably classified within and across participants. Second, a group-​level factor analysis identified three semantic dimensions central to the processing of the abstract concepts. These dimensions corresponded to the degree to which a concept was verbally represented, was experienced as internal to the individual, and was associated with social experience. The verbal representation dimension provided an interesting gradient of abstractness. Not only was it the case that abstract concepts evoked both less activation in perceptual areas and more activation in regions associated with language processing, but there seems to be a continuum of abstractness defined in terms of the relative importance of language systems compared to perceptual ones. Collectively, these brain-​imaging studies support and extend the findings from the behavioral studies. They provide evidence that the heterogeneity of abstract concepts can be explained by the contribution of multiple brain

Heterogeneity  151 mechanisms associated with aspects of our experience. In other words, they contribute to a multimodal and multilevel view of grounding that offers a unified explanation of the heterogeneity of abstract and concrete concepts.

7.4  Grounding Reconsidered As mentioned in earlier chapters, some criticize weak versions of conceptual grounding as less theoretically interesting than strong versions (e.g., Shapiro, 2019). The complaint is often made that weak grounding does not represent a scientific revolution or paradigm shift (in the sense attributed to Kuhn, 1970). We are now able to identify the significant problems with this line of attack. The first is something that it pains me as a philosopher to admit: there is no meaningful correlation between what is theoretically interesting and what is true. This is not merely a problem for philosophers, though. Cognitive scientists are also subject to this theoretical bias. Indeed, one reason to be cautious about 4E (Embodied, Embedded, Enactive, and Extended) cognition in general is that it involves theories that would be exciting if true (Prinz, 2006). Second, and more importantly, this skepticism fails to address a central motivation for weak grounding: the existence of robust bodies of evidence implicating heteromodal brain areas in our concepts. While it is true that this evidence threatens industrial-​strength versions of grounded cognition, we need to keep in mind that other evidence threatens amodal approaches. Whether or not weak grounding represents a radical enough break from traditional views of cognition to count as a scientific revolution at some level of philosophical remove, it represents a significant break from the amodal orthodoxy. Third, it is not clear that weak grounding—​despite the milquetoast character of this terminology—​does not represent the very sort of dramatic theoretical change that would reasonably constitute a paradigm shift. Importantly, the very idea of a scientific revolution is tied to the sociology and history of science. As such, revolutions are likely to be easier to recognize in the past than they are in the present. When we are trying to identify them in a forward-​looking fashion, we need to be careful. After all, it is difficult to judge which contemporary theoretical innovations are going to have the greatest influence on future science. At a minimum, though, such

152  Abstract Concepts and the Embodied Mind revolutions require the abandonment of significant theoretical assumptions and the embracing of new core tenets. I submit that there are good reasons to think that this bar is met by weakly grounded theories. Against the background of the polemical debate between supporters of traditional amodal theories and strongly grounded ones, weak grounding appears to be little more than an act of fence-​sitting. This appearance, however, misses the radical nature of this class of theories. As I point out in Chapter 2, strongly grounded theories often fail to challenge the background assumptions of traditional approaches to concepts. For one, they generally accept the notion that concepts are relatively invariant (at least in adults).4 In addition, they do not allow for the sort of hierarchical representations that appear to be so important for explaining our capacity to generalize. Finally, they lean heavily on the notion of modal specificity. As I have shown, this conception is a holdover from an overly simplistic and outdated understanding of cortical organization. I defend an elastic conception of grounding that adopts a flexible, multimodal, and multilevel approach to concepts. Each of these design features challenges core assumptions of the traditional approaches (amodal or strongly grounded). The commitment to flexibility abandons invariantism. The commitment to multimodality amounts to more than a recognition that concepts are encoded in distributed representations. The elastic account developed in this book abandons the requirement that concepts rely only on representations associated with a single modality and explicitly leaves room for representations associated with multimodal experience, such as those involved in taste and the perception of space. It also does a better job of accommodating the contribution of the emotions and also allows for grounded representations associated with the processing of language to contribute to our concepts without requiring the adoption of a full-​throated dual code framework. Finally, the commitment to multilevel representation directly abandons the goal shared by both traditional amodal and strongly grounded theories to locate concepts at a single representational level. The current view appeals to higher-​order representations in a way that is consistent with recent developments occurring within the deep learning and predictive coding research programs.

4 This is a generalization, and there certainly have been exceptions. For instance, Barsalou’s perceptual symbol system theory (1999) acknowledged the potential flexibility of our concepts.

Heterogeneity  153 In conclusion, it is possible to explain the heterogeneity of abstract concepts in a way that shares much with grounded explanations of the heterogeneity of concrete concepts. This perspective highlights the particular importance of affective, language, and social systems to abstract concepts. Behavioral and brain-​based evidence suggests that abstract concepts are processed by means of the combined activation of higher-​level semantic and experiential regions.

8 Growth and Development In this book, I have examined the ways in which established views of grounded cognition struggle to capture the somewhat messy class of abstract concepts. I have argued that meeting these challenges requires rethinking the very nature of grounding—​both in terms of its underlying neural realization and its connection to elements of our physical and social niches. The elastic view of grounding that emerges from this rethinking is dynamic and constructive: concepts are something we do rather than something we have, they are shaped by experience but remain adaptive with respect to context and task, and they can be transformed by cultural artifacts such as language. The elastic account makes clear predictions about the growth and development of our conceptual system during our lifespans. First, the grounding of concepts should change throughout the lifespan. This prediction is a natural consequence of the multimodal nature of grounding, the cumulative effects of experience, and the lack of invariance. Second, different types of concepts should be learned through different types of experiences. Some may rely on external sensorimotor processes associated with direct interaction with category exemplars; others may rely on internal sources of grounding such as emotion and interoception. Third, concepts should vary in terms of when they are acquired. External symbolic and social resources should contribute to the acquisition of abstract concepts. These resources are likely to be more important for the acquisition of abstract concepts that do not rely heavily on other internal sources of grounding. All of this is consistent with a commitment to the flexibility of the conceptual system. Flexibility occurs at two time scales: the synchronic and the diachronic. Synchronic flexibility concerns the way in which our concepts can be influenced by immediate context and other discourse factors. Diachronic flexibility concerns the way in which our concepts change throughout the course of development. Up until now, I have primarily focused on synchronic flexibility. In this chapter, focus on diachronic flexibility and review the evidence supporting the predictions just outlined.

Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0008

Growth and Development  155

8.1  The Acquisition of Abstract Words The ability to formulate thoughts about morality, the origins of the universe, and the nature of existence seems uniquely human. And yet many developmental researchers have focused primarily on the acquisition of concrete words. This may in part be because they already have their hands full: the acquisition of concrete words is hard enough to explain. Figuring out how children learn the meaning of a word such as dog from observing dogs doing dog-​like things in the child’s immediate physical vicinity is itself a daunting theoretical task because dogs are rarely isolated from the complex sensory environments in which they occur (Medina et al., 2011; Quine, 1960). This problem only becomes more acute when we focus on words that refer to events rather than objects (Maguire & Dove, 2008; Snedeker & Gleitman, 2004). Nevertheless, the acquisition of abstract words remains a pressing theoretical issue that researchers working within the framework of grounded cognition are beginning to investigate. One way to examine the development of abstract concepts is to compare when children tend to acquire abstract and concrete word meanings. Research on subjective ratings of age of acquisition (AoA; Kuperman, Stadthagen-​Gonzalez, & Brysbaert, 2012) suggests that abstract words tend to be acquired later than concrete words. Ponari, Norbury, and Vigliocco (2018) find that abstract words represent approximately 10% of the estimated vocabulary of children aged 4 and 40% of the total vocabulary of children aged 12. A sharp increase in the number of abstract words occurs in children aged 8–​9. Despite this general delay, abstract words are also a part of children’s earliest vocabularies. Reggin, Muraki, and Pexman (2021) examined parental report data about the vocabularies of children younger than 3 years (Frank et al., 2017) and found that 96 of the 600 words in the database were abstract. These words included a group of closed-​class words made up of conjunctions (e.g., and, or), determiners (e.g., the, all), exclamations (e.g., yes, no, hi, bye), prepositions (e.g., for, to, by, with), and pronouns (e.g. they, it) and a group of open-​class words made up of adjectives (e.g., yucky, careful, pretty), adverbs (e.g., away, now, better), and verbs (e.g., be, like, think). This evidence shows that while abstract word meanings may be generally harder to acquire than concrete word meanings, they are not so difficult that abstract concepts cannot be part of our earliest vocabularies. Presumably, the frequency of these abstract words and their centrality to fundamental communicative exchanges may contribute to their early emergence.

156  Abstract Concepts and the Embodied Mind According to the elastic grounding approach, some features of our conceptual system should help us acquire abstract content. In particular, the hierarchical system of conceptual representation should enable us to generalize from immediate experience and encode multimodal information. This system should be active for most concepts. It may also underwrite a capacity to represent individual concepts with differing degrees of abstraction in different situations (Barsalou, Dutriaux, & Scheepers, 2018). In the previous chapters, I identified several sources of grounding that appear to be more important for abstract concepts than concrete ones, including affective, interoceptive, social, and linguistic experience. Here, I am going to review some of evidence concerning the development of children’s understanding of word meaning that implicates these types of experience. This evidence suggests that each of these types of experience makes important contributions to the acquisition of abstract concepts (or at least some subset of them). Significantly, their impact appears to vary with age; that is, some appear to have greater influence early in development and others appear to have greater influence at later stages.

8.1.1  Affective Grounding One major account of abstract word learning holds that affective information is a particularly important component abstract word meanings. This leads to the prediction that abstract words rated with higher valence ratings (the degree to which they elicit positive or negative feelings; Warriner, Kuperman, & Brysbaert, 2013) should be learned before those with lower valence ratings. This prediction appears to hold for subjectively reported AoA: namely, positive and negative abstract words tend to be acquired before more neutral abstract words, and positive abstract words tend to be acquired before negative ones (Ponari, Norbury, & Vigliocco, 2018). Ponari and colleagues found that abstract positive words exhibit a processing advantage in a lexical decision task (comparing the responses to positive, negative, and neutral words) for 8-​and 9-​year-​olds. Importantly, no such advantage was found for positive concrete words, and no valence effects were found for children aged 6–​7. Children aged 11–​12 exhibited a processing advantage with abstract positive words in comparison to abstract negative words but exhibited no processing advantage with valenced abstract words (positive or negative) in comparison to neutral abstract words. The researchers suggest that this latter absence of

Growth and Development  157 an effect may be due to this cohort’s increased knowledge of neutral abstract words. A different lexical decision study comparing children aged 5, 6, and 7 found a processing advantage for positive abstract words over neutral abstract words with the children aged 6 and 7 (Lund, Sidhu, & Pexman, 2019). The presence of valence effects in these relatively young children may be explained by the fact that this study used less abstract words than did the earlier study (Reggin, Muraki, & Pexman, 2021). Early valence effects have also been found in a recognition memory task (Kim, Sidhu, & Pexman, 2020). Children aged 7–​8 remembered abstract negative words more accurately than abstract neutral words. As with the experiments reviewed earlier, there was no valence effect with concrete words. In a vocabulary learning experiment (Ponari, Norbury, & Vigliocco, 2020), children aged 7–​9 provided more accurate definitions of valenced words, while children aged 9–​10 showed no such advantage. The extant evidence is admittedly inchoate. Valence is just one dimension of emotional experience. It would be interesting to know the degree to which other dimensions, such as arousal, are relevant. Nevertheless, this circumscribed body of evidence supports some initial conclusions: valence information seems to play a significant role in the acquisition of abstract vocabulary up until the age of 9, a period of rapid growth of abstract word knowledge. After this, its influence appears to wane (Vigliocco, Ponari, & Norbury, 2018). Part of the reason for this loss of influence may be the expansion of neutral abstract word knowledge. In sum, the extant evidence suggests that affective information makes important contributions during a specific stage of development.

8.1.2  Linguistic Grounding A second major account of the development of abstract word knowledge maintains that abstract word meanings are learned through language. There are a couple of versions of this idea. For instance, one proposal focuses on the contributions of distributional information in word-​to-​word associations (Andrews, Vigliocco, & Vinson, 2009; Bi, 2021; Bruni, Tran, & Baroni, 2014; Landauer & Dumais, 1997; Lupyan et al., 2020). Another focuses on the way in which children may be able leverage their knowledge of linguistic constructions to learn word meanings (Gleitman, 1990; Landau & Gleitman, 1985). The language is an embodied neuroenhancement and scaffold (LENS)

158  Abstract Concepts and the Embodied Mind account is an inclusive one and accommodates both of these influences in addition to other discourse-​related, pragmatic, and dialogic factors associated with language use (Borghi et al., 2019; Dove et al., 2020). One method used to investigate word learning involves gathering subjective ratings of a word’s mode of acquisition (MoA). Adults are asked to rate whether words are acquired through perceptual experience or through language (Wauters et al., 2003). MoA correlates with AoA but does not match up with it exactly. Indeed, MoA has been found to predict reaction times on a lexical decision task even when AoA is held constant (Kousta et al., 2007). One study asked participants to rate a set of abstract and concrete Italian nouns on seven dimensions: abstractness, age of acquisition, concreteness, context availability, familiarity, imageability, and mode of acquisition (Della Rosa et al., 2010). Results suggest that concrete words are mainly acquired through experience and abstract words are mainly acquired through language. Further research using MoA suggests that children rely on perceptual experience in grades 1–​3 and on language in grades 4–​6 (Borghi et al., 2019). A central question is why children tend to acquire some words before others. Hills and colleagues (2010) found that a word’s contextual diversity (the number of distinct word types that a word co-​occurs with in the linguistic environment) predicted when that word was acquired. In particular, they found that two aspects of linguistic context appeared to be relevant: namely, proximity and semantic association. First, a word is more likely to be acquired if it is in close proximity to a number of other words in the child’s linguistic input. Second, a word is more likely to be acquired if it is semantically related to other words that the child already knows. The authors suggest that the role of association fits well with the proposal that abstract words are learned through language. Some of the very same studies used to motivate the relevance of valence information can be used to motivate the relevance of language. For example, Lund, Sidhu, and Pexman (2019) found that children with more advanced language skills responded more quickly in an auditory lexical decision task to neutral abstract words than did children with less advanced language skills. Ponari, Norbury, and Vigliocco (2020) found that children acquired knowledge of neutral abstract words despite displaying no valence effect. They suggest that this might be explained by a reliance on language rather than emotion as a bootstrapping mechanism. Of course, other experiential factors cannot be ruled out.

Growth and Development  159 One intriguing line of research examines the role of the mouth motor system in abstract concepts (for a comprehensive review, see Borghi et al., 2019). The idea behind this research is that the functional role played by the language system in abstract concepts should lead to the engagement of the mouth motor system. This hypothesis fits with the finding that abstract concepts tend to be rated as more associated with the mouth than are concrete words (excluding words for food; Granito, Scorolli, & Borghi, 2015; Borghi & Zarcone, 2016). It is supported by evidence of facilitation effects with adults and interference effects with adults and children. Let’s consider facilitation first. In a task requiring participants to decide whether definitions fit with abstract or concrete words, responses with the mouth (as opposed to the hand) were faster with abstract words than concrete ones (Borghi & Zarcone, 2016). In a follow-​up experiment, response times in a word recognition task were faster for abstract words in the mouth effector condition (speaking into a microphone) than in the hand effector condition (pressing a button), while response times with concrete words exhibited the reverse pattern (Mazzuca et al., 2018). A bit of caution is warranted because the researchers found no facilitation effect in an initial lexical decision task. Now, let’s consider interference. In an articulatory suppression condition during a concreteness judgment task, adult participants were significantly slower to respond to abstract concepts than they were to concrete concepts (Fini et al., 2021). With children, a novel sort of interference effect has been found. Reasoning that extensive use of a pacifier during early childhood might disrupt the relevant activity of the mouth motor system needed to process the linguistic and social experiences associated with abstract concepts, Barca, Mazzuca, and Borghi (2020; see also 2017) gave children in grade 3 a semantic categorization task involving abstract, concrete, and emotion words. Children who had used a pacifier extensively in early childhood responded more slowly than did children who had not. The effect was more pronounced with abstract words than with concrete and emotion words. As was the case in our earlier discussion of the emotions, the extant evidence of learning through language remains somewhat underdeveloped. Nevertheless, it is substantial enough to provide initial support for the conclusion that language makes important contributions to the acquisition of abstract concepts—​particularly at later stages of childhood. Clearly, questions remain. Perhaps the most significant of these is the question of whether language plays a more central role with some abstract concepts than

160  Abstract Concepts and the Embodied Mind it does with others. So far, the data seem somewhat equivocal. Some studies seem to suggest that language provides a general leg up for all concepts. For example, aspects of the linguistic input, such as proximity and association, may help children acquire concepts across the board. On the other hand, other studies seem to suggest that language is particularly important later in development for the acquisition of emotionally neutral concepts. Notably, these hypotheses are not mutually exclusive: language could be both an influential cognitive enhancement from the earliest stages of word learning and especially important later in development for the acquisition of certain types of abstract concepts.

8.1.3  A Multimodal Approach Taken as a whole, the evidence reviewed so far supports the proposition that both emotion and language can serve as bootstrapping mechanisms for the acquisition of abstract word knowledge. It supports the conjecture that affective grounding is more important early in development and that language becomes more important later when children acquire neutral abstract words. This suggests that a multimodal approach is needed. One reason to adopt a multimodal perspective is that other types of grounding appear to be important for abstract concepts. Interoceptive strength ratings are higher for abstract words than concrete words (Connell, Lynott, & Banks, 2018; Lynott et al., 2019). A study examining the words and gestures used to explain word meanings found that participants frequently referenced introspective states and people when communicating abstract meanings and objects and entities when communicating concrete meanings (Zdrazilova, Sidhu, & Pexman, 2018). Reggin, Muraki, and Pexman (2021) carry out several analyses on an updated (Brysbaert & Biemiller, 2017) set of AoA data collected from the testing of children for their vocabulary knowledge across elementary, middle, and high school grades (Dale & O’Rourke, 1981). Given that these data have been gathered through objective means (i.e., explicit testing), it seems likely to be more reliable than the subjective methods often used to get AoA estimates. The researchers found that interoception, mouth action, and word valence facilitated the acquisition of abstract words more than concrete words. They also found that contextual diversity facilitated the acquisition of words whether they were abstract or concrete.

Growth and Development  161 There has been an increasing interest in and experimental research on the influence of iconicity in the acquisition and use of words (Nielsen & Dingemanse, 2021). Iconicity is typically construed in terms of sensorimotor associations between linguistic forms and features of their referents. Dingemanse and colleagues (2015, p. 604) define it as “the resemblance-​ based mapping between aspects of form and meaning.” Onomatopoeia is the most obvious and familiar form of iconicity. When spoken, the words meow and moo sound like the actions that they identify (although the phonological character of these words is likely to be language-​dependent to some degree). The iconic links between words and referents can be less direct, however. Consider the English words crunch, murmur, and slurp. They each contain a clear sound-​related resemblance, but they also have specific semantic content that extends beyond this resemblance. The iconic connection can be even more distant. For instance, there is a cross-​linguistic tendency for words referring to small objects to contain high front vowels such as /​i/​as in the English words tiny and wee (Blasi et al., 2016). Iconicity is not limited to spoken language. In British sign language (BSL), the sign book involves depicting the leaves of a book (Murgiano, Motamedi, Vigliocco, 2021; Thompson et al., 2012). There is evidence that adults are sensitive to iconicity. In a lexical decision task, iconic English words were processed more quickly than non-​ iconic words (Sidhu, Vigliocco, & Pexman, 2020). A lexical decision study involving patients with aphasia found a similar facilitation effect (Meteyard et al., 2015). Iconic signs in BSL were produced faster than non-​iconic ones (Vinson et al., 2015). A word of caution is warranted though because these effects are relatively small and limited in scope. A robust and growing body of evidence suggests that iconicity plays a role in word acquisition, particularly early in development (Nielson & Dingemanse, 2021; Perniss & Vigliocco, 2014). Studies have found that young children are able to leverage iconic relationships to learn new words (Asano et al., 2015; Imai et al., 2008). This includes being able to take advantage of sound symbolism patterns borrowed from foreign languages (Kantartzis, Imai, & Kita, 2011). The words learned earliest in the process of language acquisition tend to be the most iconic. One study obtained iconicity ratings for early words in English and Spanish (Perry, Perlman, & Lupyan, 2015). Children tended to acquire words rated high in iconicity, such as bubbles and splash, before words that were not, such as person and bench. Another study found that early-​learned signs in BSL tended to be the

162  Abstract Concepts and the Embodied Mind most iconic (Thompson et al., 2012). Perry and colleagues (2018) found that young English-​speaking children use iconic words more frequently than non-​iconic ones. They also found that adults exhibited the reverse pattern except when speaking to children: in child-​directed speech, adults revert to using iconic words more frequently. There is general agreement among researchers that iconicity contributes to word learning. Current evidence suggests that iconic words are prominent in early vocabularies and that they are easier to learn than non-​iconic words. What remains controversial is the depth and scope of its influence (Nielsen & Dingemanse, 2021). On the one hand, some have proposed that iconicity plays a central role in word learning throughout development (Imai & Kita, 2014; Imai et al., 2008; Perniss & Vigliocco, 2014). Imai and Kita (2014) offer one of the clearest expressions of this with their sound symbolism bootstrapping hypothesis. They suggest that iconicity helps children learn that words can represent objects and events, helps them associates words with their referents, and helps them pick out referents in complex environments. On the other hand, others propose that iconicity may contribute to early word learning but fades dramatically as a significant resource over time (Massaro & Perlman, 2017; Nielsen & Dingemanse, 2021). This constrained view is supported by the fact that languages vary with respect to the degree of iconicity they employ (Imai & Kita, 2014; Nielsen & Dingemanse, 2021). Further research is needed to decide which of these approaches—​or some intermediate one—​is correct. Rather than leave the topic with such an anodyne conclusion, though, I wish to lay my cards on the table and offer a speculative account of how iconicity might be relevant to word learning. The LENS approach explicitly appeals to the pragmatic, discourse, and dialogic aspects of language. This inclusiveness is not a calculated hedge but rather a principled position that falls out of a theoretical commitment to the importance of task and context in both the acquisition and use of language. More broadly, it fits with the commitments of the elastic grounding approach. Against this background, iconicity might serve as one grounded developmental resource among many. Perceptual salience, manipulability, and emotionality may also facilitate word learning. In addition, other linguistic factors, such as word length, frequency, and associations, may play a role. Other aspects of verbal communication, including the presence of accompanying iconic gestures or acts of pointing and eye gaze, are also likely to be important. Murgiano, Motamedi, and Vigliocco (2021) contrast two overarching views of language: the systemic and the situated. They contend that a great

Growth and Development  163 deal of psycholinguistic research has focused on the systematic population-​ level properties of language. This perspective enables researchers to examine the structural components of a language (or dialect) at both synchronic and diachronic time scales. They suggest that this perspective ignores the fact that this systematic structure emerges from acts of communication that occur within specific physical environments and social contexts. The situated approach to language acknowledges the systematicity of language, but it focuses on the degree to which linguistic communication often relies on social interactions between cooperative interlocutors who can leverage available grounded resources to convey meaning (Bohn & Köymen, 2018; Tomasello, 2008). It also allows for a nondeflationary view of the contribution of iconicity to word learning and use because it views iconicity as one nonarbitrary communicative resource among many available for speakers throughout the lifespan. Murgiano and colleagues offer the following assessment (2021, p. 10): The wide range of evidence concerning iconic and indexical communicative cues suggests that language as it is used in online, face-​to-​face interaction between both adults and children cannot be confined to a system of context-​independent, linguistic components, nor can it be assigned only a grounding role, useful only in the early stages of language learning. Rather, language being a dynamic, multimodal system situated in given communicative and physical context, adaptively uses multiple arbitrary and non-​arbitrary cues (arbitrary words and onomatopoeia available in the linguistic system, as well as iconic gesture, deixis and prosody) to contribute to context-​dependent meaning making.

By these lights, iconicity is a consistently useful aid to comprehension that may be particularly valuable during early word learning. It serves as one of many available situated and nonarbitrary cues available to children in their communicative interactions with others and provides support for both comprehension and learning. Our focus is on the acquisition of abstract words. Two findings involving iconicity seem particularly relevant. First, words high in iconicity are acquired before words that are low in iconicity. Indeed, iconicity is a statistically reliable predictor of AoA even when semantic associations are controlled for (Perry et al., 2015, 2018). Second, as children develop, they begin to produce less iconic words more frequently than iconic ones (Perry et al., 2018). One

164  Abstract Concepts and the Embodied Mind possible contributing factor to these developmental patterns is that “iconicity is inimical to abstraction because iconic forms are too connected to specific contexts and sensory depictions” (Lupyan & Winter, 2018, p. 1). In other words, the non-​arbitrariness of iconic words may interfere with their capacity to represent abstract concepts. The picture that emerges from our brief survey of research on word learning is that children appear to rely on multiple grounded cues to learn words. These include concreteness, emotions, iconicity, interoception, and language. While these cues are available throughout the course of development, each appears to play a greater role at a certain stage of development. In keeping with their heterogeneity, abstract words associated with high valence tend to be acquired before neutral abstract words. All in all, research on word learning provides compelling support for a multimodal view of concepts.

8.2  Relational Categories The evidence just outlined implicates both inner experiential systems and language in the emergence of abstract concepts. One question that remains unanswered is the extent to which language provides a leg up with respect to the acquisition of specific abstract concepts. Some evidence suggests that linguistic bootstrapping is helpful from the very beginning of word learning (e.g., Hills et al., 2010). Other evidence suggests that language plays a greater role later in development and is most relevant to the acquisition of neutral abstract words (e.g., Lund et al., 2019). In this section and the next, I consider evidence implicating language in the acquisition of specific types of abstract concepts. I focus on those that depict relational categories and those that depict mental states, respectively. As we have seen in previous chapters, the differences between abstract and concrete concepts appear to be complex and multidimensional. One reason for this is that abstract concepts are heterogeneous. While a great deal of attention has been paid to the relative importance of broad semantic domains (such as the aesthetic, moral, social, and scientific domains), the ways in which words connect to the world can also be important (Gentner, 1983, 2006). The distinction between concepts that refer to object categories and those that refer to relational categories is a good example. While the members of object categories such as pug share many intrinsic features, the

Growth and Development  165 members of a relational category such as pet do not. Indeed, they can vary widely in their intrinsic features. Cats, fish, monkeys, snakes, animatronic toys, and even rocks can be pets. Consider another example, the category of a barrier. Fences, hedges, mountains, rivers, pillows, foreign languages, and lack of money are examples of possible barriers (Gentner, Anggoro, & Klibanoff, 2011). The literature on the acquisition and use of relational concepts extends well beyond the issue of conceptual representation. Surveying this literature would require a significant digression. Rather than attempt to provide a global summary, I propose to focus on the issue of the degree to which language helps children learn relational concepts. I start off with a couple of related experiments that directly implicate the contribution of language to their acquisition. Gentner, Anggoro, and Klibanoff (2011) investigated relational category learning in children aged 3, 4, 5, and 6. Each child was given instances of relational concepts by means of cards depicting objects that might serve as its arguments (Figure 8.1). For example, the relation cutter for was illustrated by the combination of a watermelon drawing (the entity that is cut) and a knife drawing (the entity that does the cutting). Half of the children were given the exemplars without a specific label (the NoLabel condition: e.g., “The knife goes with the watermelon”), and the other half were given the exemplars with a novel label used in a relational construction (the RelLang Initial Phase:

Entity

Operator Test Phase:

Entity

Taxonomic

Thematic

Relational

Alternatives

Figure 8.1  Sample set of images depicting a cutter for relation from Experiment 1a (Gentner, Anggoro & Klibanoff, 2011).

166  Abstract Concepts and the Embodied Mind condition: e.g., “The knife is the dax for the watermelon”). Following this, the children were asked to complete another example. They were presented with a drawing of a piece of paper and asked to choose the item from a set of three alternatives that has the same relation as the knife did to the watermelon. These items offered a taxonomic match, a thematic match, and a relational match, respectively. This experiment was designed to see whether hearing relational language would help children acquire the relevant concept from an elicited comparison. The results (Figure 8.2) demonstrated an age-​sensitive effect: the 3-​year-​olds did not learn the concept under either condition, the 4-​to-​5-​ year-​olds learned the concept only in the RelLang condition, and 6-​year-​olds acquired the concept equally in both conditions. It appears that relational language helps the 4-​to-​5-​year-​olds, while 6-​year-​olds are able to learn the concept from the comparison alone. In a follow-​up experiment, the researchers sought to find out whether progressive alignment, with or without relational language, can help 3-​year-​olds learn concepts associated with relational categories. To test this, they had the children interact with a puppet, “Sammy.” For each relational category, the

1.0

Proportion of Relational Choices

0.9

**

0.8

**

0.7 *

0.6 0.5 0.4

Chance

0.3 0.2 0.1 0.0 3 Years

4–5 Years No Label

6 Years

Rel Language

Figure 8.2  Results from Experiment 1a (Gentner, Anggoro & Klibanoff, 2011). Six-​year-​olds in both conditions and 5-​year-​olds in the RelLang condition made relation responses above chance (.33) rates.

Growth and Development  167 children and Sammy were shown four pairs of cards depicting an entity and an operator. The first half of the initial phase involved “close” pairs in which the objects and entities associated with a relational were highly similar (e.g., a watermelon paired with a knife and an orange paired with a different knife as examples of the relation cutter for). The second half of the initial phase involved “far” pairs that were less similar (e.g., a pine tree paired with an ax and a log with a saw). In the RelLang condition, the experimenter says to the children at the beginning, “In this game, we are going to teach Sammy the word dax. We are going to show him what dax means.” The test phase was then identical to the experiment outlined earlier. In the condition where the children were provided with progressive alignment without a relational label, they responded by selecting the alternatives at the level of chance, but in the condition where the children were provided with both progressive alignment and a label, they chose the relational alternative at a level that was higher than chance. This suggests that children of this age can learn a relational category from progressive alignment when this progressive alignment is combined with appropriate labeling. A central take-​away from these experiments is that exposure to relational language can support the formation of relational concepts. This conclusion fits with the results of other studies. Christie and Gentner (2014) tested children aged 2 and 3 using a relational match-​to-​sample task that has proved difficult for nonhuman primates, including most monkey species (Thompson & Oden, 2000). Chimpanzees with symbolic training (but not those without this training) can accomplish this task (Thompson, Oden, & Boysen, 1997). Inspired by the apparent importance of symbolic training, Christie and Gentner used this established task to investigate the degree to which linguistic experience might aid children with relational learning. After finding that children aged 2 and 3 fail this task, the researchers gave groups of children from each cohort pre-​test symbolic training involving the application of the words same and different. Following this training, children aged 3 (but not children aged 2) were able to accomplish the relational match-​to-​sample task. In a different experiment, they provided no pre-​test training but instead introduced a novel label, truffet, that they applied to the standard used in the task. Strikingly, children from both age groups were then able to accomplish the task. These results fit with other findings that support the conclusion that labels help children learn relational concepts by inviting them to compare things that may lack obvious or salient similarities (Gentner & Namy, 2004; Namy & Gentner, 2002).

168  Abstract Concepts and the Embodied Mind Gentner (2010) suggests that two design features make human cognition special: our capacity for relational cognition and our facility with symbolic systems that enhance this capacity. She maintains that language can serve as a springboard to make the sort of structural comparisons and to recognize the ways in which different situations may align with each other. This influence fits with the LENS view that language acts as a cognitive neuroenhancement that helps us recognize and learn concepts. Research on the acquisition of relational concepts thus provides a clear picture of how children might leverage their acquired symbolic competence to extend their conceptual repertoire. To be clear, this research does not show that language is the only means by which relational concepts are learned. To the contrary, it is based on the presumption that analogical reasoning is independent from language. Gentner offers a mutual bootstrapping theory in which language and analogical reasoning form a positive feedback system. Language facilitates the development of our relational cognition and is in turn facilitated by relational cognition.

8.3  Theory of Mind Let’s now consider another case study: research on the emergence of theory of mind in young children. Much of the research on this topic has been pursued without any commitment to the research program of embodied or grounded cognition. In other words, it represents an independent investigation of the acquisition of a developmentally significant body of abstract knowledge. Given this context—​namely, a research effort that is not focused on abstract concepts as a difficult challenge to a particular theoretical approach—​the discovery of a robust body of evidence pointing to the importance of language is revealing and gives added weight to the LENS theory. Much of early word learning is focused on concrete words. The task of learning the meanings of these words may well involve word-​to-​world mappings (Gleitman et al., 2005). These mappings could arise from the observation of statistical contingencies between words and perceptually salient objects or events in the physical environment. Communicative gestures such as pointing and eye gaze may help children pick out the appropriate referents and track these contingencies. Gleitman and colleagues (2005) argue that certain abstract word meanings, such as those associated with psychological states, require structure-​to-​world mappings to be acquired.

Growth and Development  169 Our ability to infer, imagine, and reason about the mental states of others is widely recognized as a core feature of human social cognition (Leslie, 2001). Consequently, researchers have expended a great deal of effort trying to uncover the development course of the skills involved in understanding others. A great deal of research attention has been placed on the development of the ability to attribute false beliefs to others under various conditions (Wimmer & Perner, 1983; although see Bloom & German, 2000, for a critical assessment of this practice). What has emerged from this research is a robust body of evidence suggesting that language can play an important causal role in the development of theory of mind. There are three leading types of theoretical explanations for the apparent influence of language on the development of theory of mind (de Villiers & de Villiers, 2014). The first holds that learning to apply mental state terms provides the child with an invaluable leg up with respect to the acquisition of theory of mind skills. In other words, the labels themselves play an important role here. This explanation is supported by the observation that frequent use of mental terms by young children with parents, siblings, and friends is correlated with later degree of success on false-​belief tasks (Brown, Donelan-​ McCall, & Dunn, 1996; Dunn et al., 1991; Ruffman, Slade, & Crowe, 2002). The second type of theoretical explanation holds that conversations provide essential input for learning about the mental states of others (Harris, 2005; Peterson & Siegal, 1995). Learning to talk explicitly about mental states is important, but so is learning the pragmatic dynamics of everyday discourse. De Villiers and de Villiers (2014, p. 314) point to the following possible exchange as an example of how even mundane conversations might provide a window into the beliefs of others.

I am going to have chocolate spread on my toast. That’s Marmite!

While there is no explicit discussion of beliefs, the ability to understand this conversation requires sensitivity to the possibility of ignorance and false belief. This view is supported in part by a series of studies involving deaf participants (Courtin & Melot, 1998; Peterson, 2009). Severe delays in the acquisition of theory of mind skills have been observed in neurocognitively unimpaired deaf children of hearing parents who are not fluent signers (Courtin & Melot, 1998; Peterson, 2009; Peterson & Siegal, 1995; Schick et al.,

170  Abstract Concepts and the Embodied Mind 2007). Because these children generally acquire sign language outside of the home, they tend to be relatively late signers. In contrast, children who grow up in a household with a fluent signer tend to be early signers. A series of studies compared these groups with respect to their performance on theory of mind measures and found that the children of nonsigning hearing parents lag behind children with at least one signing deaf parent (Peterson & Siegal, 1999; Woolfe, Want, & Siegal, 2002). This finding fits well with evidence that the interactions between deaf mothers and deaf children are similar in content, extent, and frequency with those between hearing mothers and children, while the interactions between hearing mothers and deaf children fall short on each of these dimensions (Meadow et al., 1981). An account that connects theory of mind understanding to access to relevant conversations provides a compelling explanation for the lag in the acquisition of theory of mind skills in late signers when compared to early signers. This idea is bolstered by research involving hearing children that demonstrates the effectiveness of conversation-​based interventions on false-​belief tasks (Appleton & Reddy, 1996). The third type of explanation links the emergence of theory of mind abilities to the child’s mastery of certain linguistic constructions (Astington & Jenkins, 1999; de Villiers, 2007). Here, the idea is that aspects of linguistic competence play a causal role in the acquisition of theory of mind understanding. Researchers have focused on the hypothesis that the ability to formulate complement clauses might contribute to the emergence of theory of mind. Building on their previous example, de Villiers and de Villiers (2014, p. 314) offer the following pair of sentences: Joanna thought that it was chocolate spread. But it was really Marmite.

The idea is that an increased familiarity with and understanding of certain constructions may facilitate the child’s ability to think about the mental states of others. In other words, complement clauses may provide a useful means of representing the contents of other people’s minds. A diverse body of evidence supports this proposal. For one, the ability to handle complement syntax is a robust predictor of performance on false-​ belief tasks (de Villiers & Peyers, 2002; de Villiers & de Villiers, 2012; Low, 2010). Strikingly, this relationship holds for several populations of children who experience language delays—​including deaf children of hearing

Growth and Development  171 parents (de Villiers & de Villiers, 2012; Pyers & Senghas, 2009; Schick et al., 2007), children with specific language impairment (de Villiers, Burns, & Pearson, 2003; Farrant, Mayberry, & Fletcher, 2012; Miller, 2004), and high-​ functioning individuals with an autism spectrum disorder (ASD; Tager-​Flusberg & Joseph, 2005). Hale and Tager-​Flusberg (2003) compared the effectiveness of three interventions on preschoolers: the first involved training children on false-​belief tasks, the second involved training children on sentential complements, and the third involved training children on relative clauses (which are similar in syntactic complexity to sentential complements). Children trained on false-​belief tasks showed improvement on theory of mind tasks but did not show a corresponding improvement on language tasks. Children trained on sentential complements, on the other hand, showed improvement on both types. Finally, children trained on relative clauses showed no appreciable improvement in theory of mind understanding. This evidence supports the notion that the acquisition of sentential complements contributes to the emergence of theory of mind in preschoolers. The influence of language on false-​belief tasks does not appear to be limited to early development. For instance, verbal shadowing has been shown to disrupt performance on a nonverbal false-​belief task in neurotypical adults (Newton & de Villiers, 2007). In sum, researchers have offered three distinct explanations for the apparent influence that language has on the development of theory of mind. Each of these enjoys some significant empirical support. We should keep three things in mind when assessing them. First, none of these explanations excludes the others. It is entirely possible that they all hold and that language plays each of these roles to a greater or lesser degree. Another possibility is that we need a more nuanced account that incorporates elements of each. For instance, Berio (2021) argues that children learning false-​belief reasoning leverage culturally embedded schemata that are facilitated by language acquisition. Second, none of these explanations requires that language is necessary for theory of mind. While language may help, it need not be the only developmental pathway. Indeed, one of the most successful interventions yet to be investigated involves training children to understand cartoon-​like thought bubbles (Wellman, Hollander, & Schult, 1996; Wellman & Peterson, 2013). Notably, this intervention involves the use of external physical symbols. Third, questions remain about the importance of language to adult comprehension of theory of mind. For example, a neuropsychological case

172  Abstract Concepts and the Embodied Mind study of a patient, P. H., who shows a significant language impairment that includes the sort of constructions identified as developmentally important, found that he had intact false-​belief reasoning abilities (Apperly et al., 2006). Each of these proposed influences of language on the acquisition of psychological concepts fits well with the LENS theory. For instance, the LENS theory predicts that the ability to affix a label to categorize objects or events transforms our capacity to conceptualize and think about them. While this is true of concepts in general, this cognitive leverage seems especially important with respect to abstract concepts. While it may be difficult to provide a clear definition of abstractness, there is a manifest sense in which the mental states of others are not part of our direct experience. The fact that mental state labels appear to facilitate the emergence of theory of mind can thus be seen as a specific instance of a general phenomenon. The LENS theory also predicts that the informational features of conversations provide support for the acquisition of psychological concepts such as those involved in theory of mind. Finally, the LENS theory emphasizes the degree to which the representational properties of language can influence our concepts. This would underwrite the sort of syntactic bootstrapping that appears to be associated with the acquisition of theory of mind.

8.4  Developmental Language Disorder Much of the research discussed in the previous two sections points to the causal relevance of language for the acquisition of at least some abstract concepts. One means of evaluating this hypothesis is to study special populations in which language is selectively impaired or preserved to see what effects these circumstances might have on the acquisition of concepts in general and abstract concepts in specific. To date, there has not been a tremendous amount of research on this front. However, some recently gathered evidence is suggestive. Vigliocco, Ponari, and Norbury (2018) review some of their published (Vigliocco, Ponari, & Norbury, 2017) and unpublished data concerning the acquisition of abstract word knowledge in atypical development. These studies compare abstract word knowledge in typically developing (TD) children and children with developmental language disorder (DLD).1 1 Vigliocco, Ponari, and Norbury (2018) also refer to unpublished data involving children with an ASD. The researchers found that, once they had controlled for language impairment, the ASD

Growth and Development  173 Children with DLD experience relative delays in the acquisition of vocabulary (McGregor et al., 2013) and aspects of morphosyntactic competence (Rice, 2013) that cannot be explained by the existence of other sensory, social, or neurodevelopmental deficits. Vigliocco and colleagues point out that previous research suggests that children with DLD are less likely to take advantage of statistical learning (Evans, Saffran, & Robe-​Torres, 2009) and syntactic bootstrapping (Shulman & Guberman, 2007). They propose that, if language is a primary means of acquiring abstract word knowledge, then children with DLD should experience particular difficulties with the acquisition of abstract word meanings. Using an auditory lexical decision task to test for implicit semantic knowledge and a definition task to test for explicit knowledge, they found no selective disadvantage for abstract words in comparison to TD children. Instead, they found a general disadvantage for all types of words. These results fit well with the proposal that language provides a general benefit to word learning, but they do not fit well with the proposal that language is particularly important for the acquisition of abstract concepts. However, a lot more research on children experiencing language delays needs to be done before any firm conclusions can be drawn.

8.5  Multiple Cues Despite the long-​standing body of evidence showing a cognitive difference between abstract and concrete concepts, researchers have only recently begun to focus specifically on their acquisition in development. Nevertheless, a substantial and growing body of research looking into the emergence of abstract words in the vocabularies of children provides compelling support for a multimodal approach to concepts. From the beginning, factors associated with linguistic context, such as proximity and association, appear to contribute to the acquisition of concepts. Affective information appears to be particularly helpful during an early period of rapid growth of abstract word learning, which tends to occur from 8 to 9 years of age. After that period, children begin to acquire neutral abstract words. During this period, abstract word learning appears to rely heavily on language. Nonarbitrary children performed similarly to the TD and DLD children. Given the preliminary nature of the results, the well-​known variability of deficits experienced by children with an ASD, and the possibility of multiple confounds, I have chosen not to include this finding in the discussion.

174  Abstract Concepts and the Embodied Mind aspects of language use, such as iconicity and indexicality, make important contributions to early word learning but appear to fade in terms of their importance to word learning (but not their importance to online face-​to-​face speech) as abstract words become more prevalent. Although the focus on the development of abstract concepts as a group is largely a recent phenomenon, developmental researchers have examined the acquisition of specific types of abstract concepts over a longer period of time and in greater detail. Research on both relational and psychological concepts implicates language as an important causal influence in their development. Supporters of grounded cognition are sometimes accused of conflating the circumstances in which a concept is learned and the mechanisms by which it is represented (e.g., Shapiro, 2019). While it is certainly the case that these need to be distinguished, the very notion of grounding presupposes an intimate connection between learning experiences and representation. To be clear, the elastic approach offers an account of the neurological mechanisms involved in our concepts. It holds that these mechanisms depend on neural reuse and are flexible, multimodal, and hierarchical. They encompass cortical regions associated with emotions and language in addition to those associated with both internal and external sensorimotor experiences. In this chapter, I have provided evidence that diverse grounded cues shape the acquisition of abstract concepts.

9 Metaphor Responding to the problems posed by abstract concepts for grounded cognition, I have defended a theory of concepts that assigns a central role to distributed neural reuse, recognizes the importance of external symbolic technologies such as language, and emphasizes the degree to which concepts are a developmental phenomenon. The purpose of this chapter is to examine the role that metaphors might play in capturing abstract concepts. I argue that our ability to use metaphors provides further evidence of the elasticity of our conceptual system. There are obvious reasons to look for a connection between abstract concepts and metaphor. We often turn to metaphors when faced with complex or elusive topics. Metaphors can, for example, help demystify difficult scientific concepts. They are especially useful at communicating ideas to those who lack the appropriate mathematical or theoretical background. Famously, the expanding universe can be thought of as a loaf of raisin bread baking in an oven. This metaphor enables physicists to express the idea that everything is moving away from everything else. Metaphors may also help us develop new scientific hypotheses and models for phenomena that we do not completely understand. The Rutherford-​Bohr solar system model of the atom is another well-​known example. Metaphors also seem handy when we need to give voice to deeply held feelings. They often play a central role in testaments of love. Consider the country standard, “You are my sunshine.”1 In sum, metaphors enable us to give shape and structure to what might otherwise be amorphous and difficult to perceive. They also help us to organize and package our thoughts. At times, we might not even aware of this more aspect. For instance, I did not make a conscious decision to use only cosmological metaphors in this paragraph.

1 Although this song is often perceived to be sweet, even saccharin, it is actually about the desperation and pain associated with rejection. Listen to the words carefully, and you will be surprised by their aggressive and threatening tone. The common misperception of the emotional tenor of this song is likely due to both the charming nature of the melody and the solar metaphor. Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0009

176  Abstract Concepts and the Embodied Mind Given the apparently functional connection between metaphor and abstract concepts, it is reasonable to conjecture that there might be an underlying cognitive connection between them. Indeed, linguists and cognitive scientists have gathered a diverse body of evidence suggesting that metaphorical thinking plays important roles in our cognitive lives. Some have even proposed that conceptual metaphors are responsible for our capacity for abstract concepts (Gibbs, 2006; Lakoff & Johnson, 1980). In this chapter, I examine the evidence offered in support of this view as well as the criticisms that it has faced. Ultimately, I conclude that what we know about metaphor fits well with the sort of flexible, multimodal, and multilevel account that I have been defending in this book. A word of caution, though. I do not claim that grounding provides a full explanation of metaphor across the board. There is an ongoing debate among researchers about whether metaphor should be primarily conceived of as a grounded phenomenon or a discourse phenomenon. I defend the position that metaphor cannot be understood fully from either perspective alone (Gibbs, 2017; Hampe, 2017; Littlemore, 2019; Müller, 2017). In other words, I agree with those who claim that metaphor is a complex phenomenon that involves language, thought, and communication (e.g., Steen, 2008). While aspects of metaphor likely depend on embodiment, metaphors are also shaped and influenced by culture and discourse factors. Any ecologically valid account of metaphor needs to recognize and explain the complex factors that are responsible for the functionality of metaphors.

9.1  Conceptual Metaphor Theory Metaphors have promise as a solution to the problems posed by abstract concepts because they relate abstract content to more concrete experiential phenomena. For example, we often talk about difficulty in terms of weight and importance in terms of size. More broadly, we often use experience-​based metaphors to talk about concepts with referents that are not immediately tied to our experiences. From an embodied or grounded perspective, these metaphors may be more than convenient linguistic devices; they may reflect an underlying form of metaphorical thinking. Santiago and colleagues (2012, p. 1051) provide a succinct summary of this idea.

Metaphor  177 Abstract concepts without a clear reference in the perceptual world constitute an essential part of the human conceptual system. Think of morality, time, magnitude, social status, intimacy, to name a few well-​studied examples. Lacking direct correspondences with perceptuo-​ motor experiences, how can people represent and reason about them? A prominent answer to this question is the Metaphor View . . . which proposes that our conceptual system recycles concrete concepts to help understanding the abstract. Representations of location, motion, size, color, brightness, weight, smell, temperature and other perceptually based dimensions of experience are used to understand abstract concepts as if, at least in part, the latter were examples of such concrete experiences.

According to what the authors refer to as the Metaphor View, abstract concepts are thought of in terms of mental metaphors tied to perception and physical action (see also Jamrozik et al., 2016). This view is commonly associated with conceptual metaphor theory (CMT). Working primarily from language data (often intuition-​based or anecdotal but sometimes culled from corpuses of natural speech or written communication), cognitive linguists have found that metaphors are not a rare feature of language use but rather a common and central feature of it (Lakoff & Johnson, 1980, 1999). For example, English speakers regularly talk about theories as if they were buildings (Kövecses, 2010, p. 6): Is that the foundation for your theory? The theory needs more support. We need to construct a strong argument for that. We need to buttress the theory with solid arguments.

Sentences such as these suggest that we sometimes understand one conceptual domain in terms of another. Cognitive linguists have proposed that this linguistic behavior is explained by an underlying conceptual metaphor. Correspondences between the target domain and the source domain allow for a directional mapping between these domains that enables us to think about the former in terms of the latter. As a consequence, this metaphor enables us to use our knowledge of buildings to structure and organize our thoughts about theories. At times, linguistic form may determine the assignment of what constitutes the source and target domains of the metaphor. Consider the different

178  Abstract Concepts and the Embodied Mind metaphorical interpretations that would arise in response to utterances of the following sentences under typical discourse conditions (Glucksberg & Keysar, 1990): This butcher is a surgeon. This surgeon is a butcher.

The first sentence would generally be interpreted as commending the skill of the butcher. It treats butchering as the target domain and surgery as the source domain and maps the skills of the surgeon onto those of the butcher. The second sentence reverses this relationship. It highlights the lack of skill of the surgeon by mapping the presumed relative lack of skill of the butcher (given the less fine-​grained demands of their job) onto the surgeon. Importantly, CMT treats these conceptual metaphors as a cognitive phenomenon and not a linguistic one. To make this clear, researchers have adopted the convention employed by concept researchers (and used throughout this book) of representing concepts by means of fully uppercased words. For instance, the preceding sentences are evidence for an underlying conceptual metaphor of theories are buildings. Similar linguistic evidence has been used to infer the existence of conceptual metaphors such as an argument is war, love is a journey, and ideas are food (Lakoff & Johnson, 1980). As useful as this convention is, we need to be careful to recognize that the use of verbal labels may give the false impression that conceptual domain individuation is a simple matter. Because the labels are familiar, they tend to reify the purported conceptual content and make it seem more specific, tractable, and immediate than it may be in reality. This can become apparent when researchers critically examine specific metaphors. Semino (2008), for instance, argues that an analysis of a large corpus of metaphor use in written British English shows that, rather than the oft-​cited conceptual metaphor of argument is war, the relevant conceptual metaphor at work is better characterized as antagonistic communication is physical contact. Further questions arise because researchers occasionally use synonymous terms to describe domains in a way that gives the impression of difference where there might not be any. For example, the following conceptual metaphors have been identified: familiarity is proximity, social distance is physical distance, and intimacy is closeness. To what degree should we consider the proposed conceptual domains characterized

Metaphor  179 by the terms of proximity, physical distance, and closeness to be distinct? Do these represent separate conceptual domains, or are they to be explained by a single, more general conceptual domain? Perhaps careful corpus-​based research or controlled lab experiments could resolve some of these issues. Nevertheless, it is important to keep in mind that using specific words to identify the respective conceptual metaphors may give the impression that these issues are more settled than they are in fact. These challenges reflect an underlying epistemic limitation of the standard research program of CMT. This limitation arises because of the attempt to infer truths about how we think from observations about how we talk (Murphy, 1996, 1997). The indirectness of this inference is problematic because it can often be difficult to rule out alternative explanations for the observed patterns of linguistic behavior. After all, there could be other language-​related reasons for the prevalence of metaphor. For instance, some have suggested that conventional metaphors are learned idiomatic associations (Keysar & Bly, 1999; Keysar et al., 2000). Recognizing these epistemic limitations, some researchers have sought more direct evidence supporting the existence of conceptual metaphors. In particular, they have examined the degree to which metaphor might be a grounded phenomenon. Perhaps the clearest expression of this is Bergen’s (2012) metaphorical simulation hypothesis. The core idea is straightforward: conceptual metaphors should involve neural simulation in the experiential areas associated with the source domain. The metaphorical simulation hypothesis predicts that we should find evidence of causally relevant activity in experiential areas associated with the source domain of the metaphor (e.g., Desai, 2021). We now have a possible explanation of how some abstract concepts might be handled by a grounded semantic system: namely, they are structured by neurological links to experiential domains. Our ability to think about abstract contents may thus be explained in part by our capacity to think about them metaphorically in terms of concepts that are more directly tied to our grounded experience. Does this solution work? The first thing to note is that questions of scope arise immediately. There are good reasons to suspect that conceptual metaphors are not up to the task of capturing the full content of abstract concepts. Conceptual metaphors require a mapping between two conceptual domains. They thus depend on correspondences between a source and target domain. For such a mapping to get off the ground, it would seem that the target domain must already be structured to some degree (Murphy, 1996,

180  Abstract Concepts and the Embodied Mind 1997). It just does not seem possible to map the structure of one fully developed concept onto a completely unstructured one (Jamrozik et al., 2016). Metaphoric mappings also exhibit a degree of flexibility that seems incompatible with the thesis that metaphors fully structure abstract concepts. It is an obvious point—​but an important one nonetheless—​that the same source domain can be used with different target domains. After all, speakers of English use the journey metaphor to talk about more than romantic relationships. Not only do we employ the life is a journey metaphor (Kövecses, 2010), but novel journey metaphors are possible. In a world still reeling from Covid-​ 19, it is not hard to imagine how we might use a pandemic is a journey to structure what we say and think about the situation in which we find ourselves. Our ability to generate this metaphor seems to require a preexisting understanding of pandemics in addition to a ready-​made journey schema. The one-​ to-​ many relationship between target domains and source domains is a common feature of conceptual metaphors. Consider the following list of metaphors associated with the concept of happiness (Kövecses, 2010, p. 111): Happiness is a fluid container: She was bursting with joy. Happiness is heat/​fire: Fires of joy were kindled by the birth of her son. Happiness is a natural force: I was overwhelmed by joy. Happiness is a physical force: He was hit by happiness. Happiness is a social superior: They live a life ruled by happiness. Happiness is an opponent: She was seized by joy. Happiness is a captive animal: All joy broke loose as the kids opened their presents. Happiness is insanity: The crowd went crazy with joy. Happiness is a force dislocating the self: He was beside himself with joy. Happiness is a disease: Her good mood was contagious.

This list is revealing in a couple of ways. First, the items themselves are clearly similar to one another. In particular, they all exhibit a general force-​dynamic pattern (Leonard & Talmy, 1988). Second, each is an instance of a general metaphor that can be applied to other emotions (Kövecses, 2010). For example, we could use the same general metaphors to talk about anger instead of happiness. In other words, not only is it the case that emotions such as anger and happiness share individual metaphors (such as the heat/​fire

Metaphor  181 metaphor), but it is also the case that they share groups of metaphors. Of course, it may well be the case that some specific metaphors are more suitable for anger than happiness and vice versa. The question then becomes how to explain this suitability. The existence of an underlying nonmetaphorical understanding of anger or happiness seems required. While CMT may be able to capture important generalizations about language use and possibly the underlying cognitive processes, it does not appear to be up to the task of providing a full account of the content of individual abstract concepts. The fact that individual source domains can be mapped onto multiple target domains and individual target domains can be captured in terms of multiple source domains throws into question strong versions of the Metaphor View.2 Fortunately, the theoretical framework developed in this book does not require us to hypothesize that the content of individual abstract concepts is entirely captured by conceptual metaphors; our focus is instead on the degree to which the reactivation of the relevant experiential representations is causally relevant to conceptual processing. Faced with the inherent limitations of using linguistic data to draw inferences about cognitive processes, an active and robust research program examining embodied metaphors has emerged. Researchers have gathered a diverse body of evidence implicating affective and sensorimotor representations in the semantic processing of metaphors. This evidence in many ways echoes the evidence gathered in support of grounded cognition using concrete concepts. Let’s begin with some of the behavioral studies. Various effects have been found that implicate conceptual metaphors associated with verticality. For example, participants respond more quickly to sentences referring to larger or smaller amounts (Coffee contains a lot of caffeine vs. Tea contains little caffeine) when the sentences referring to the larger amounts appear on the top of screen and when the sentences referring to smaller amounts appear at the bottom of the screen (Langston, 2002). When asked to generate a random number, people who are looking up tend to generate larger numbers and people who are looking down tend to generate smaller numbers (Winter &

2 These considerations should not be seen as providing a knockdown argument against CMT in general. CMT is an ambitious theoretical proposal that seeks to explain linguistic and cognitive behavior at several time scales (Lakoff, 1987; Lakoff & Johnson, 1999). It seeks to capture not only widespread patterns of language use but also universal features of language acquisition, language change, and aspects of the evolution of human cognition. Given the breadth of these ambitions, supporters of CMT may be able to take my criticisms of the Metaphor View in stride.

182  Abstract Concepts and the Embodied Mind Matlock, 2013). Studies such as these are taken as evidence for the existence of a more is up conceptual metaphor. Quantity, though, is not the only thing that appears to be metaphorically associated with verticality. After all, we often associate positive things with a higher vertical position (the top, the peak, the pinnacle, etc.) and negative things with a lower one (the bottom, the lowest, the nadir, etc.). In keeping with this, evaluations of positive words are faster when they appear higher on the screen and evaluations of negative words are faster when they appear lower on the screen (Meier & Robinson, 2004). Emotionally positive words are also recalled better when they are presented at the top of the screen and negative words when they are presented at the bottom (Crawford et al., 2007). Appealing to evidence such as this, researchers posit a positive is up conceptual metaphor. Researchers have also explored the relationship between verticality and our representation of social power. Participants are faster at identifying a target presented in the upper portion of a computer screen after judging that a word such as king identifies a powerful person and in the lower portion after judging that a word such as servant identifies a powerless person (Zanolie et al., 2012). It is important to acknowledge that there have been some issues regarding the reproducibility of this research. A recent failure of reproduction has gained some notoriety. In an influential article, Williams and Bargh (2008) describe the findings of two experiments that seem to demonstrate a link between physical warmth and feelings of interpersonal warmth: Experiment 1 found that participants who briefly held a cup of warm coffee (as opposed to a cup of cold coffee) judged a target person as being more generous and caring, and Experiment 2 found that participants who held a warm (rather than a cold) therapeutic pad were more likely to exhibit prosocial behavior. These results were challenged by subsequent research. A follow-​up study to Experiment 1 found that the effect could be modulated by social context (Citron & Goldberg, 2014b), and three highly powered attempts to replicate Experiment 2 failed to find the effect at all (Lynott et al., 2014). At a minimum, these studies point to a need to scrutinize initial findings. As was the case with concrete concepts (see Chapter 3), the existence of grounded metaphors is supported by research in cognitive neuroscience (Desai, 2021). Here, too, we can group these studies into those that implicate perception representations and those that implicate action representations. A good example of the former used rapid event-​related functional magnetic resonance imaging (fMRI) to compare how participants processed sentences

Metaphor  183 containing textual metaphors and literal sentences matched with respect to meaning (e.g. Sam had a rough day vs. Sam had a bad day). This study found that the metaphors activated texture-​selective nonprimary somatosensory cortex (Lacey, Stilla, & Sathian, 2012). Another study that compared sentences containing taste metaphors and literal sentences found that the taste words elicited increased activation in primary and secondary gustatory cortex (Citron & Goldberg, 2014a). Studies comparing literal motion verbs with those expressing metaphorical motion (Chen, Widick, & Chatterjee, 2008) or fictive motion (which occurs when a motion verb is used with a subject that cannot move: e.g., The path comes into the garden; Wallentin, Lund et al., 2005; Wallentin, Ostergaard et al., 2005) find activation in motion-​ sensitive areas that are adjacent to primary visual areas. The evidence involving action systems is somewhat inconsistent. For instance, while it is the case that a study has found that arm-​related and leg-​ related action words in idiomatic and literal sentences elicited somatotopic activation in the motor strip (Boulenger, Hauk, & Pulvermüller, 2009), other studies indicate that the effector-​specific activations associated with the semantic processing of action words does not appear if the words are used in metaphors (Aziz-​Zadeh et al., 2006) or idioms (Raposo et al., 2009). The activation of sensorimotor areas during metaphor processing may occur at a later point in time than it does with the processing of literal semantic contents. In an event-​related potential (ERP) study (Bardolph & Coulson, 2014), participants made upward or downward movements with marbles as they read words that had literal (ascend, descend) or metaphorical (inspire, defeat) vertical associations. Congruency effects were found with both types of words when the associations matched the direction of the movements, but their time signatures were different: with the literal movement words the effects emerged at 200–​300 ms after word onset but after 500 ms with the metaphoric movement words. The delay with the metaphors suggests that the relevant sensorimotor simulations are not automatically engaged in the same way that they are with the literal action words. This undermines the suggestion that metaphorical simulation acts in the same way that simulation is supposed to work in concrete concepts. Furthermore, it raises the question of whether such metaphorical simulation is causally relevant to the conceptual processing or merely an aftereffect. Exhibiting a common cross-​linguistic pattern that fits well with the metaphor view, English uses the same prepositions to express spatial and temporal relationships (e.g., at the corner and at noon). Kemmerer (2005)

184  Abstract Concepts and the Embodied Mind conducted a series of experiments with four patients with left perisylvian lesions. He found evidence of a double dissociation between the spatial and temporal meanings of the same prepositions: two of the participants exhibited an impairment for the spatial meanings but a preserved ability to process the temporal meanings, while the other two exhibited the reverse pattern. This double dissociation suggests that the spatial and temporal meanings are processed independently. Whatever historical role a time is space conceptual metaphor might play in explaining the existence of these parallel meanings, it does not seem to be necessary for online semantic processing. The behavioral and neurocognitive evidence suggests that embodied or grounded metaphors may contribute to the representation of abstract concepts. Significant questions remain, however, concerning their reach. Here, too, we can distinguish between a stronger and a weaker view of embodied or grounded metaphor (Jamrozik et al., 2016). The stronger view holds that metaphor comprehension relies entirely on automatically engaged simulation grounded in unimodal sensorimotor areas, and the weaker view leaves room for other influences, including language, culture, and proximal discourse factors. Taken as a whole, the evidence favors the weaker view.

9.2  Language, Discourse, and Culture A primary criticism of the CMT approach to metaphor (as a linguistic phenomenon) is that it leaves out too much of what actually makes metaphors tick. For instance, it fails to capture how the effectiveness of metaphors can be shaped by sociocultural norms and factors associated with specific discourses. Gibbs (2017) points out that researchers use several distinct methods to investigate metaphor that involve different levels of analysis—​ including evolutionary, historical, sociocultural, discourse, linguistic, cognitive, psychological, embodied, and neural levels. He suggests that it is unlikely that an account tied to any one of these levels of analysis will capture the generalizations associated with all the others. Metaphors are simply too complex, diverse, and multifactorial to be explained by a single-​level theory. Casasanto (2014) argues that metaphors are often shaped by cultural experience in a way that is not directly tied to the body. He uses the metaphor conservative is right/​liberal is left as an example. There is no obvious embodied grounding for this association. Because people differ in which

Metaphor  185 metaphorical direction they lean, the metaphor is disconnected from the directional linguistic metaphor in English good is right/​left is bad. Even when we focus on this directional metaphor, we find that it’s apparent association with handedness is at the population level and not necessarily tied to individual bodily experience. After all, many people are left-​handed and evidence suggests that while right-​handers tend to associate positive ideas and emotions with the right side of space and negative ideas and emotions with the left, left-​handers exhibit the reverse pattern (Casasanto, 2009). This body-​specific metaphor can itself be transformed by experience: Casasanto and Chryikou (2011) had right-​handed participants carry out a motor fluency task for 12 minutes with a cumbersome glove on either their dominant or nondominant hand. Those participants who used the glove on their left hand (preserving the normal dominance of their right hand) still showed a preference for stimuli presented on the right, but those who used it on their right hand switched their spatial bias. Specific motor experience can change implicit associations. Examples showing the influence of cultural artifacts such as language can be found in studies examining the way that passage of time is conceptualized in terms of a mental line. Research indicates that speakers of Spanish tend to conceptualize time from left to right (Flumini & Santiago, 2013; Santiago et al., 2007). Hypothesizing that this might be due in part to the orientation of their writing system, Ouellet and colleagues (2010) compared the response times of speakers of Spanish with that of speakers of Hebrew (which is read from right to left) in a temporal judgment task. Speakers of Spanish responded more quickly with their left hand to words associated with the past and more quickly with their right hand to words associated with the future, while speakers of Hebrew exhibited the opposite pattern. While the elastic grounding approach predicts that experience with different natural languages should result in different grounded metaphors, it also predicts that grounded metaphors themselves should be flexible and experience dependent. This prediction is supported the fact that providing participants with a brief exposure to mirror-​reversed orthography can reverse the orientation of the congruency effects on temporal judgments associated with a particular language (Casasanto & Bottini, 2014). It is also supported by another recent experiment which found that, after introducing a novel metaphor connecting time and weight (the past is heavy and the present is light), congruency effects emerged in weight judgments of books that appeared new or old (Slepian & Ambady, 2014).

186  Abstract Concepts and the Embodied Mind Shen and Porat (2017) argue that there is a clear sense in which linguistic and conceptual metaphors behave differently. Recent evidence suggests that conceptual metaphors can exhibit bidirectional influence. Consider the conceptual metaphor affection is warmth. Much of the initial evidence for this conceptual metaphor involves linguistic utterances that seem to link the abstract target domain with the concrete one. However, some recent evidence suggests that the cognitive influence goes in both directions. For example, participants who recalled experiencing social exclusion gave lower estimates of room temperature than those who recalled experiencing social inclusion (Zhong & Leonardelli, 2008). Such bidirectional effects are not uncommon and include other associations between abstract and concrete domains (e.g., Schneider et al., 2011; Zhong & Liljenquist, 2006). The problem is that linguistic metaphors are commonly unidirectional. There is no conventional linguistic metaphor warmth is affection. Even if such a metaphor were to exist—​say, in a poetic context—​speakers and hearers would need to use an abstract source domain to structure their understanding of a concrete target domain. Shen and Porat provide a twofold solution to the problem of directionality: they suggest that (i) verbal metaphors depend on prelinguistic, bidirectional conceptual associations and that (ii) language creates the directional source-​target relationship. There is also evidence that conceptual metaphors can fail to emerge in the expected fashion with some cognitive tasks. Casasanto (2008) scrutinizes the similarity is closeness conceptual metaphor in a behavioral experiment. Participants performed a similarity judgment on stimuli presented on a computer screen. The influence of spatial proximity of these judgments depended on the nature of the similarity judgment: stimuli paired closer together were judged to be more similar than stimuli paired farther apart when the task involved conceptual similarity but were judged to be less similar when the judgment involved perceptual similarity. This experiment undermines the notion that we can draw a direct link between how we talk about similarity and how we think about it. Responding to this argument and the evidence offered in support of it, Grady and Ascoli write (2017, p. 30; emphasis in the original), “It is possible that there is a conceptual link between proximity and similarity (including perceptual similarity) but that link this link is ‘trumped’ in the context of a particular judgment task, and quite possibly other contexts.” They go on to suggest that CMT is compatible with such flexibility and is not committed to the idea that the conceptual metaphor is at work

Metaphor  187 in all circumstances. In other words, immediate contextual factors and task demands may supplant the influence of the underlying conceptual metaphor. Steen (2017) argues that metaphors work differently when they are used deliberately for effect. He claims that deliberate metaphors succeed as a means of communication by acting as a metaphor shared by language users within a discourse. This special use requires that the interlocutors attend to the source domain as a referential domain in its own right. In other words, deliberate metaphors succeed because they are consciously recognized and treated as metaphors. They are processed differently than nondeliberate metaphors. Following an introduction such as Shall I compare thee to a summer’s day?, the audience is aware that the speaker is intentionally offering a metaphoric comparison. With the utterance of a sentence such as This theory needs support, interlocutors may not consciously attend to the fact that the theories are buildings metaphor is being employed. The role of attention in metaphors highlights the flexible nature of their use. Jensen (2017) proposes that we should think of metaphors in terms of whole-​body interactions rather than in terms of static mappings. We should question how people do metaphor. By these lights, metaphor is neither strictly social nor strictly cognitive. He uses the term “metaphoricity” to highlight the dependency of metaphors on the presence and interaction with other people.3 The term is meant to capture the fact that metaphors are not the outcome of a cognitive cross-​domain mapping or a linguistic convention but a process of creating a double meaning on the spot for the purposes of communication. Metaphors are the result of an “inter-​personal ecology” (p. 257). Cameron and colleagues (2009, p. 63) explain the core insight behind a dynamical approach: “a metaphor is no longer a static, fixed mapping, but a temporary stability emerging from the activity of interacting systems of socially situated language use and cognitive activity.” From this perspective, the distinction between deliberate and nondeliberate metaphor is a matter of degree and just one of the scalar dimensions by which we might characterize metaphors in use.

3 One might think that written metaphors serve as a counterexample to this approach. Jensen (2017) suggests that, as a specialized activity that focuses on linearly presented chunks of language, writing offers distinct opportunities for metaphor. As such we should expect to find differences in metaphor use in writing from that offered through the immediate bodily dynamics of face-​to-​face conversation.

188  Abstract Concepts and the Embodied Mind

9.3  The Living and the Dead (or Sleeping) Any theory of metaphor must deal with the fact that metaphors can vary greatly in terms of the degree to which they are conventionalized. Some metaphors are so conventionalized that many speakers may not even realize that they are metaphors. These are often referred to as dead, frozen, or sleeping metaphors. One of the initial insights behind CMT was that many metaphors are hiding in plain sight. A lot of our everyday language is metaphoric—​such as when we talk about supporting a theory, spending time, and seeing the truth. Such expressions seem to depend on linguistic convention and are part of the semantic knowledge required to be a fluent speaker of English. Other metaphors are creative and novel. Their status as metaphors is inescapable. Consider Roald Dahl’s (1988, p. 6) delightfully gross metaphor in the following passage: The parents looked upon Matilda as nothing more than a scab. A scab is something you have to put up with until the time comes when you can pick it off and flick it away. Mr. and Mrs. Wormwood looked forward enormously to the time when they could pick their little daughter off and flick her away, preferably into the next county or even further than that.

While the comparison of a child to a scab unusual, the metaphor is easily understood. It highlights the disdain that Matilda’s mother and father have for her and the comfort that they take from the temporariness of their parental obligations. Philosophers of language have been trying to figure out how to handle metaphor within the context of their semantic theories (Hills, 2017). The distinction between conventional and novel metaphors has proved to be a significant theoretical challenge. There is general agreement that a single theory of metaphor is desirable (although some suggest that no theory is possible: e.g., Davidson, 1978). For a long time, two main theoretical camps dominated: those who want to treat metaphors as a semantic phenomenon, accommodating metaphor within what is said by an utterance (e.g., Stern, 2000) and those who want to treat it as a pragmatic phenomenon, treating metaphor as something implied by what is said (e.g., Camp, 2006; Grice, 1989). Recently, a new approach has emerged that treats metaphor as just one of the contextually sensitive means by which we understand what is said beyond the literal content of the words (e.g., Wearing, 2006). While it

Metaphor  189 would take us too far afield to evaluate the philosophical arguments for these approaches, it is worth noting that conventional metaphors would seem to fit better with the semantic approach and that novel metaphors would seem to fit better with the pragmatic and contextualist approaches. Our survey of the cognitive science literature suggests that metaphor may be a complicated phenomenon that is not well-​suited to a single-​level account. In an attempt to deal with this apparent complexity, Casasanto (2014) proposes that we distinguish between mental metaphors associated with cross-​domain influences on cognition and linguistic metaphors that are lexicalized. This account of metaphor is meant to resolve the tension between the apparent contribution of grounded mechanisms and the culturally specific metaphors that become enshrined in language. Steen (2008, 2017) offers a hierarchical approach that distinguishes metaphor in thought, language, and discourse. Such multilevel and multidimensional accounts of metaphor raise the possibility that conventional and novel metaphors may differ in their relative dependence on grounding and language. Some recent evidence from cognitive neuroscience suggests that there are important differences in how conventional and novel metaphors are processed (for a detailed discussion of this evidence and its implications, see Desai, 2021). Saygin and colleagues (2010) carried out an fMRI experiment in which they compared motion sentences (The wild horse crossed the barren field), fictive motion sentences (The hiking trail crossed the barren field), and static sentences. Motion-​sensitive visual areas were activated significantly more with the motion sentences than with the other two sentence types, but they were also more active with the fictive sentences than they were with the static ones. A skeptic of grounded metaphor might take the findings of decreased activation with fictive and metaphoric contexts as supporting the claim that sensorimotor activation during conceptual tasks is generally epiphenomenal. There is even some evidence that can seem to support this conjecture. Building on the finding that sentences describing upward or downward motion can interfere with the shape perception of a visual object presented in the relevant region (Richardson et al., 2003), Bergen and colleagues (2007) carried out a number of experiments comparing the responses to subject nouns or main verbs in sentences triggered by visual imagery. They found interference effects with literal sentences referring to the relevant portion of the visual field but not with metaphorical or abstract sentences. We need to be careful though. After all, the categorical dismissal of metaphorical grounding fits poorly with

190  Abstract Concepts and the Embodied Mind other findings that implicate grounding in metaphor. In addition, a study comparing patients with Parkinson’s disease to age-​matched controls found that the Parkinson’s patients exhibited a deficit with both idiomatic and metaphorical action sentences (Fernandino et al., 2013a). Metaphorical grounding appears to be somewhat flexible (Desai, 2021). For example, Desai and colleagues (2013) compared the neural responses to three analogous types of sentences: those expressing “literal” action (The daughter grasped the flowers), those expressing “metaphoric” action (The public grasped the idea), and those expressing “abstract” action (The public understood the idea). Two of their findings are relevant to our present concerns. First, in the “literal” and “metaphoric” conditions, they found that the degree of activity in primary motor and visual motion areas was inversely correlated with the familiarity ratings of the sentences. This fits with the proposal that metaphors undergo a gradual process of conventionalization as they gain familiarity (Bowdle & Gentner, 2005). Second, in the “metaphor” and “abstract” conditions, they found increased activation in left temporal regions. This fits with the notion that their content is stored at least in part by means of linguistic associations.

9.4  Development The preceding sections have surveyed evidence implicating both sensorimotor simulations and the language system in metaphor comprehension. This evidence even suggests that their respective influences may wax and wane depending on the sort of metaphor involved: in general, novel metaphors seem to rely more heavily on grounded simulations, while conventionalized metaphors seem to rely more heavily on language. When we turn our attention to the development of children’s ability to comprehend and produce metaphors, we find further evidence of this division of labor.

9.4.1  Neurotypical Development Tracking the developmental course of metaphor has proved a challenging endeavor. Early studies found that children do not become proficient with metaphors until relatively late in development, with a gradual transition

Metaphor  191 from a predominantly literal understanding of speech to one that includes metaphors beginning at the ages of 8 and 9 (Asch & Nerlove, 1960; Vosniadou, 1987; Winner, 1979; Winner, Rosentiel, & Gardner, 1976). Echoing a familiar pattern in psycholinguistic research, researchers have used methodological innovations such as multiple-​choice tasks or tasks involving toys to provide evidence of earlier metaphor comprehension (Melogno, Pinto, & Levy, 2012). Using these techniques, research has found that metaphor comprehension might begin at 4–​5 years of age (Cicone, Gardner, & Winner, 1981; Vosniadou et al., 1984; Waggoner & Palermo, 1989). Other studies show that children as young as 5 can understand “sensorial metaphors” that involve comparisons within a single sensory domain, such as the sun is an orange and the river is a snake (Gentner, 1988; Siltanen, 1986). Rundblad and Annaz (2010a) found that metaphor comprehension gradually improved throughout childhood and into adulthood. Not all metaphors are created equal. Some are particularly common, and others appear to be more dependent on context, community, and cultural factors. Within CMT, it has become standard to distinguish primary and complex conceptual metaphors (Grady, 1997). Primary metaphors arise from nearly ubiquitous correlations in our experiences in which abstract categories co-​occur with more directly embodied categories. Perhaps the easiest way to understand this term is to consider examples. Take the conceptual metaphor of more is up. The idea is that this metaphor emerges through an experiential connection between an increase in quantity and the perception of height. Piles grow and containers fill up. Other primary metaphors such as affection is warmth, difficulty is weight, importance is size, and time is a motion along a path also seem to originate, at least in part, from our physical experiences. These experiences give us access to correlations that we are able to leverage into mappings between the relevant conceptual domains (Lakoff & Johnson, 1980). Complex conceptual metaphors such as theories are buildings involve more than these universal experiences. Some recent evidence also suggests that language-​specific metaphors may build on preexisting nonlinguistic embodied mappings. Whereas speakers of Dutch tend to talk of musical pitch in terms of height (the way that we do in English), speakers of Farsi tend to talk of it in terms of thickness (Shayan, Ozturk, & Sicoli, 2011). These different linguistic metaphors appear to influence how Dutch and Farsi speakers reproduce recently heard musical pitches in the presence of irrelevant spatial information involving either height

192  Abstract Concepts and the Embodied Mind or thickness (Dolscheid et al., 2013). A follow-​up study found that prelinguistic infants are sensitive to both the pitch-​height and the pitch-​thickness mappings (Dolscheid et al. 2014). The acquisition of metaphor in typically developing (TD) individuals exhibits several important features. First, comprehension and production abilities appear to surface early during the preschool years. Second, the capacity to produce and understand metaphors does not appear to be an all-​ or-​nothing phenomenon but is instead the result of a gradual developmental process in which the number and complexity of the metaphors that children are able to process increases over time (Melogno, Pinto, & Levi, 2012). Third, some metaphors—​particularly those associated with fundamental sensorimotor experiences—​seem easier to acquire than others. Fourth, even these metaphors can be shaped by culture (Winter & Matlock, 2017). For example, Almohamadi (2017; discussed in Littlemore, 2019) examined metaphor comprehension in Arabic-​speaking children aged 3, 4, 5, and 6 and found that their understanding of several primary metaphors improved significantly over this time period. Fifth, metaphor begins as a largely cognitive phenomenon and only gradually becomes a linguistic one. Although children begin to make sense of metaphor during their preschool years, they do not become fully competent consumers and producers of metaphor until they acquire better language skills, greater encyclopedic knowledge, and more sophisticated pragmatic competencies associated with the interpretation of an interlocutor’s communicative intent.

9.4.2  Sensory Differences Grounding links conceptual representation to experiential systems. In this way, it shares a commitment to the importance of experience to our concepts with empiricism. Hume (2000) recognized that this commitment created a challenge for empiricism: the question of sensory concepts in individuals who experience sensory deficits. Hume bit the bullet and argued that certain concepts were inaccessible to such individuals. He writes (p. 15): If it happen, from a defect of the organ, that a man is not susceptible to any species of sensation, we always find, that he is as little susceptible of the correspondent ideas. A blind man can form no notion of colours; a deaf man of sounds.

Metaphor  193 We have already reviewed evidence demonstrating a rich knowledge of visual concepts in congenitally blind participants, so Hume’s presumptive assessment was incorrect. An elastic approach can explain this richness by appealing to the multimodal nature of experience and the influence of language. Nevertheless, it remains plausible that certain metaphors should be understood and used differently by people who experience the world differently. A few studies suggest that differences in the use of primary metaphors can be found in children and adults with visual impairments (Littlemore, 2019). Iossifova and Marmolejo-​Ramos (2012) sought to examine the use of primary metaphors relating to vision by TD children, children with visual-​ motor impairments, and blind children between the ages of 3 and 8. The participants were asked to point at spatial locations (in front of you, behind you, and down) and temporal locations (yesterday, today, and tomorrow). They found that visually impaired children were less able to relate space to time than TD children. A study involving early-​blind adults and sighted controls looked for evidence of a mental time line (Rinaldi et al., 2018). Participants were asked to classify words as referring to the future or the past by moving their hands forward (away from their body) and backward (toward their body). Sighted participants responded more quickly when the semantic content was metaphorically congruent with the response direction (words for the future with forward movements and words for the past with backward movements). The early-​blind participants did not exhibit this facilitation effect. The authors conjecture that the absence of the sagittal mental time line in the early-​blind participants might be explained by the fact that they experience space differently (Schinazi, Thrash, & Chebat, 2016). These studies suggest that the absence of certain types of experiences may decrease the likelihood of certain metaphors. However, much more research needs to be done before any firm conclusions can be drawn. One reason for caution is that languages encode space-​time mappings in multiple ways. For example, in Aymara, a language spoken in western Bolivia, speakers use the same word to indicate front and past and a different word for back and future (Núñez & Sweetser, 2006). This contrasts with the more common tendency to construe future events in terms of a spatial metaphor involving what lies in front of us. Further evidence of the flexibility of these conceptual metaphors is provided by the fact that Mandarin speakers appear to use both future-​in-​front/​past-​at-​back and past-​in-​front/​future-​at-​back mappings (Gu, Zheng, & Swerts, 2019).

194  Abstract Concepts and the Embodied Mind Evidence from sign language provides a more complex picture of how sensory differences may influence metaphors understanding and use. Given that sign languages are full-​blown natural languages, it should come as no surprise that metaphors are common. Indeed, research on sign languages shows that metaphors play a key role in the creation of signs, particularly those for abstract concepts (Taub, 2001; Wilcox, 2000). Meir and Cohen (2018) provide a clear example: in Israeli Sign Language (ISL) the sign for eat involves repeatedly moving a handshape that mimics holding a small object with the fingers coming together toward the mouth. The sign for learn involves the same movement and handshape directed toward the temple. The iconic sign is thus mapped onto the abstract domain of mental activities. There is reason to think that metaphor works somewhat differently in sign languages than it does in spoken languages. Meir (2010) notes that metaphors in sign languages often involve a double-​mapping: an iconic mapping between linguistic form and some concrete domain and a metaphorical mapping between that concrete domain and an abstract one. Meir proposes that a double-​mapping constraint (DMC) operates in sign languages. According to the DMC, metaphors must in general fit with the iconic mapping associated with the signs themselves and the source domain. Metaphors that violate this constraint should be rare and disfavored. The DMC is meant to explain why it can be difficult to translate some metaphors from spoken languages into a sign language. Meir (2010) highlights this by the example of the metaphor associated with sentences such as Time flies. The sign fly in ISL involves flapping one’s hands in a way that evokes the use of wings. Meir argues that the iconicity of this sign fails to highlight the speed of flight and thus does not support the shift in meaning needed for the appropriate metaphorical reading of the sentence. In sum, research is beginning to show that sensory differences—​either with respect to bodily experiences or the medium of communication—​can influence the understanding and use of metaphors. This research not only fits with the proposal that bodily experience shapes metaphor but also with the proposal that language shapes metaphors. The rich iconic medium of sign language provides opportunities for metaphor that are not always available in spoken languages. In turn, the absence of iconicity in spoken languages may free up opportunities for metaphorical extension that are more difficult with more iconic equivalents in sign languages.

Metaphor  195

9.4.3  Autism Spectrum Disorders and Asperger’s Syndrome Individuals with an autism spectrum disorder (ASD)4 who have good language skills pose a different kind of test case for metaphor. As is true of ASD in general, this circumscribed subgroup is highly heterogeneous with respect to several cognitive dimensions, including such things as learning abilities, rigidity of thought, scope of interests, and social engagement (Casanova & Casanova, 2019). Given this variability and inherent complexity, we should be cautious when drawing broader conclusions about metaphor from research involving this population. Littlemore (2019) identifies an apparent paradox in the research on metaphor in ASD. On the one hand, a robust body of literature suggests that high-​functioning individuals with an ASD have difficulties understanding metaphor. On the other hand, some studies indicate that they are able to generate novel readings of metaphors and produce them in their speech. I am going to consider each in turn, starting with the evidence of a deficit. Children with an ASD have been found to experience difficulties with figurative language in general (Dennis, Lazenby, & Lockyer, 2001; Mackay & Shaw, 2004; Vulchanova, et al., 2015). Several studies find that people with an ASD experience particular difficulties with metaphor (Happé, 1993; Kasirer & Mashal, 2014; Melogno, Pinto, & Levi, 2012; Rundblad & Annaz, 2010b). For instance, high-​functioning individuals with an ASD who were given the Test of Language Competence (Wiig & Secord, 1992) performed poorly on subtests evaluating inferences and metaphor processing (Minshew, Golstein, & Siegel, 1997). Another study found that children with an ASD exhibit impaired metaphor comprehension in relation to measures of mental age (Runblad & Annaz, 2010b). Research specifically focused on individuals with AS suggests that they experience difficulties with metaphor comprehension (Faust, 2012). In an ERP study, researchers used the amplitude of particular waveform component, the N400 (a negative peak in the waveform that occurs at 400 ms that

4 There has been a recent effort to include Asperger’s syndrome (AS) within ASD. This effort is driven in large part by difficulties associated with differentiating between individuals with an ASD who have good language skills and those with AS. I refer to both here, not because I wish to reject this effort, but because some of the relevant research focuses specifically on people who are explicitly identified as having AS. It is interesting to note that Asperger himself had patients with more severe impairments and was committed to the idea of an autism spectrum (Silberman, 2015).

196  Abstract Concepts and the Embodied Mind has been associated with semantic integration; Key, Dove, & Maguire, 2005), to track the degree of effort applied to semantic integration (Gold, Faust, & Goldstein, 2010). Literal and novel metaphors elicited larger N400s in the AS participants than in the controls. The investigators propose that differences in linguistic processing explain the difficulties in metaphor comprehension experienced by AS individuals. Using a divided visual field paradigm, Gold and Faust (2010) found that control participants showed a right hemisphere advantage in reaction times, but AS participants did not. Building on previous research implicating a special role for the right hemisphere in novel metaphor comprehension, they propose that right hemisphere dysfunction may provide an explanation for the difficulties that AS individuals have with novel metaphors. Now, let’s consider the evidence for a lack of an impairment. Much of this evidence involves novel metaphors, which may depend more on grounded processing. Highly verbal individuals with an ASD are often able to offer creative interpretations of novel metaphors that build on phonological and semantic associations (Melogno, D’Ardia et al., 2012). Several studies have found that participants with an ASD were able to comprehend novel metaphors as well as age-​matched controls. Children with an ASD (aged 12–​15) had more difficulties with comprehending conventional metaphors than TD children (e.g., sharp tongue) but showed no difference with novel metaphors (e.g., joy bits). Adults with AS (aged 22–​68) have been found to process novel metaphors similarly to age-​matched TD adults (Hermann et al., 2013). Kasirer and Mashal (2014) examined the verbal creativity of adults with an ASD in comparison to age-​matched TD adults. They used a multiple-​choice task involving conventional and novel metaphors to measure comprehension and a sentence completion task to measure verbal creativity. The participants with an ASD performed similarly to the TD adults in their comprehension but outperformed them with respect to generating novel metaphors. In other words, the participants with an ASD exhibited remarkable verbal creativity with respect to novel metaphors. What emerges from this brief survey of the available evidence is that individuals with an ASD often process metaphors differently than do TD individuals. While a large portion of this evidence suggests many experience difficulties with metaphor, a smaller portion suggests that some individuals have heretofore underappreciated verbal skills that enable them to process novel metaphors in creative and unusual ways. That these skills have been overlooked is not entirely surprising. After all, levels of intelligence of children with an ASD have regularly been underestimated

Metaphor  197 (Dawson et al., 2007). The selective difficulty for conventional metaphors may be due in part to weaknesses in vocabulary and general knowledge (Kasirer & Mashal, 2014). It is worth emphasizing, however, that individuals with ASD vary widely in terms of their cognitive skills. The fact that highly verbal individuals with an ASD have been found to have the ability to understand and use (novel) metaphors successfully does not mean that others will not struggle with them. Indeed, there is an unfortunate tendency in the research literature to study highly verbal ASD individuals (for all sorts of pragmatic reasons) and draw inferences from them about autism in general. The variability of abilities, behavioral traits, and interests exhibited by individuals with an ASD exposes the obvious flaw in this methodology. To put it simply, the class of people identified as having an ASD is just too heterogeneous to make such broad inferences. The fact that one subgroup of individuals with an ASD have certain abilities—​particularly a subgroup that fails to exhibit one of autism’s central diagnostic features (language difficulties)—​is insufficient grounds for generalizing about the disorder. Indeed, difficulties with metaphor have been linked to several aspects of ASD, including impaired executive function (Landa & Goldberg, 2005), right hemisphere dysfunction (Gold & Faust, 2010), impaired theory of mind (Happé, 1993; Mackay & Shaw, 2004), and a lack of semantic knowledge (Norbury, 2005). Rather than treat the research on highly verbal individuals as conclusive for ASD, we should view it as provisional. What is important for my purposes is that it provides further evidence that both grounding and language play distinct roles in metaphor processing.

9.5  Metaphors Are Elastic Some have looked to metaphors as a panacea for the problems posed by abstract concepts for grounded cognition. The Metaphor View holds that our ability to encode abstract concepts depends on our capacity to recycle representations from concrete domains. This compelling proposal encounters theoretical and empirical challenges, though. On the theoretical front, it seems unlikely that abstract conceptual domains are shaped entirely by concrete source domains. For such cross-​domain mappings to get off the ground in the first place, the relevant abstract domains must have some preexisting structure. This point is only strengthened by the facts that a single concrete domain can be used as a source for multiple distinct

198  Abstract Concepts and the Embodied Mind abstract domains and that a single abstract domain can be characterized in terms of multiple distinct concrete domains. This one-​to-​many and many-​ to-​one flexibility belies any proposal holding that the source domain fully structures the target domain. On the empirical front, the evidence for conceptual metaphors is often indirect—​involving evidence from linguistic productions—​and limited in scope. The strongest evidence for grounding comes from a subset of abstract concepts that seem to enjoy a special connection to primary metaphors. When we step back from this bold conjecture, examine the available research for evidence of causal influence, and consider the phenomenon of metaphor in all its complicated glory, a different picture emerges. In keeping with the flexible, multimodal, and multilevel account developed in the previous chapters, metaphors depend on both grounding and language. The degree to which metaphors rely on these influences can vary. Novel metaphors seem to rely more on grounding, and conventionalized metaphors seem to rely more on language. The influence of grounding and language can also shift in development. Initial cross-​domain mappings may help young children recognize and understand new types of information, but these mappings can gradually become lexicalized—​inheriting directionality and ultimately becoming shared cultural artifacts. Even highly conventionalized “sleeping” metaphors can become “waking” metaphors through the dynamic interaction of people in the context of a particular discourse (Müller, 2017). Metaphor is not an alternative to the sort of conceptual system that I have been proposing but rather an outcome of it. This perspective enables us to go beyond the observation that some metaphors engage action, emotion, and perception systems. It allows us to acknowledge that metaphors are shaped by language, evolve over the course of development, and remain flexible in relation to task and context. Metaphors are the complicated and multifaceted consequence of an inherently flexible grounded cognitive system interacting with language, culture, and discourse. To put it simply, metaphors are elastic.

10 The Elastic Mind This book began with a theoretical puzzle: despite the substantial body of evidence suggesting that concrete concepts are grounded in experiential mechanisms, there are reasons to doubt that abstract concepts work that way. The rapidly growing body of research on abstract concepts within cognitive science, developmental psychology, neuropsychology, and psycholinguistics only complicates and extends this puzzle. Throughout this book, I have argued that the theoretical challenges posed by abstract concepts force us to revise our conception of grounded cognition. Abstract concepts are the product of an elastic mind. The reuse of experiential mechanisms remains central, but this reuse is distributed, flexible, and multilevel. Conceptual content is not only encoded by representations indigenous to sensorimotor areas, but also by representations from areas involved with emotion, interoception, and social cognition. This updated view also points to our reliance on external features of our cognitive niche, such as language and culture. The chapter unfolds in three stages. First, I show how the arguments from the previous chapters come together to support an elastic view of concepts. Next, I explore what this view tells us about abstract concepts themselves. Finally, I identify the ways in which the elasticity of our conceptual system transforms the research program of grounded cognition. This will point to new avenues of research and illustrate the empirical promise of the elastic approach.

10.1  Causal Relevance The rethinking of grounded cognition started off with a critical reappraisal of the evidence for and against neural reuse. Both supporters and critics of neural simulation acknowledge that sensorimotor systems are engaged by conceptual processing. The core disagreement concerns the causal relevance of this activity. Supporters of grounded cognition view this activity as necessary for the relevant cognitive processes, and critics contend that it is Abstract Concepts and the Embodied Mind. Guy Dove, Oxford University Press. © Oxford University Press 2022. DOI: 10.1093/​oso/​9780190061975.003.0010

200  Abstract Concepts and the Embodied Mind epiphenomenal. There is an unfortunate tendency to evaluate the evidence in all-​or-​nothing terms: if one is theoretically inclined to support grounding, then one is likely to generalize broadly from positive evidence implicating sensorimotor mechanisms in semantic memory; if one is theoretically opposed to grounding, then one is likely to put great emphasis on the fact that modality-​specific activation is not always found in neuroimaging studies (e.g., Bedny et al., 2008; Postle et al., 2008; Raposo et al., 2009), neuropsychological case studies (e.g., Arévalo, Baldo, & Dronkers, 2012; Kemmerer et al., 2012), and transcranial magnetic stimulation (TMS) studies (e.g., Papeo et al., 2009; Papeo, Pascual-​Leone, & Caramazza, 2013).1 Ultimately, the polemical nature of this debate interferes with an honest appraisal of the experimental literature. The truth is that compelling evidence for the causal relevance of experiential reactivation has been found, but this evidence is circumscribed and incomplete (Ostarek & Bottini, 2021). One reason for the predominance of such dichotomous thinking is the assumption that concepts must be realized in an invariant fashion. Early versions of grounding treated simulation as an automatic, inflexible, and necessary feature of our concepts. Against this background, evidence of missing activation or even varying patterns of activation in different contexts counted against the causal relevance of neural reactivation to conceptual processing. The elastic mind approach maintains that this assumption is mistaken. If we look at motor planning (which is itself an important model for grounded cognition), we find the simultaneous presence of reliability and flexibility. Our movements are regular and consistent, but they are also adapted to the task at hand and the physical context. In other words, they exhibit what Grafton (2020) refers to as “physical intelligence.” Requiring complete invariance would lead to a terrible account of the activity of the motor system. I suggest that the same is true for concepts. Given that a great deal of research implicates action and perception systems in conceptual processing, it is reasonable to conclude that the neural systems we use to experience objects, actions, and events in the world can 1 Amodal theorists often suppose that, because of the universally quantified nature of strong embodiment theses, they only need to find evidence of some exceptions. This strategy is regularly employed without circumspection. After all, what is good for the goose is good for the gander. Critics of the classical approach can similarly focus on universally quantified versions of amodal theories—​ which, to be frank, are most of them. By these lights, the abundant evidence of causal relevance of grounding in particular instances suggests that strong amodal approaches are similarly doomed. Supporters of amodal representations have started to recognize this problem and have begun to identify a limited causal role for grounding (e.g., Kompa, 2019; Machery, 2015).

The Elastic Mind  201 be reused to internally simulate those objects, actions, and events at later points in time. The real question is the scope of this influence. How central are such simulations to our concepts, particularly those that we would pretheoretically identify as abstract?

10.2  Theoretical Challenges The next step in my reappraisal of grounding was to identify several important theoretical challenges posed by abstract concepts. Grounded cognition provides a compelling account of how our brains can leverage experientially derived information to form concepts, but it struggles with accounting for concepts that go beyond experience. Abstract concepts are challenging because they refer to objects, properties, and events that are less manipulable and perceivable; rely on culturally derived information; and tend to be more sensitive to context. Ultimately, they pose three distinct theoretical problems for grounded cognition: the problems of generalization, disembodiment, and flexibility. Recent research on abstract concepts suggests that they are not a monolith but instead a heterogeneous class. Some have argued that this heterogeneity inoculates grounded cognition from criticism. In contrast, I have argued that abstract concepts pose several problems because they are heterogeneous. These challenges fit with different properties that different theorists have associated with abstractness (i.e., generality, intangibility, and context availability).

10.3  Hierarchical Representations I then proceeded to tackle each of these problems in succession. This effort began with a critical appraisal of the existing orthodoxy concerning grounded cognition. I questioned the common presumption that grounding must involve modality-​ specific sensorimotor representations. Critics of grounding often see this idea as their principal target and tend to view any move to weak embodiment as little more than an ad hoc retreat from this thesis. In contrast, I showed that there are good reasons to doubt the modal-​ specificity constraint. For one, the designation of a sensory modality is a rough-​and-​ready anatomical characterization anchored to an anachronistic

202  Abstract Concepts and the Embodied Mind view of sensation that is incompatible with the current emphasis on the importance of multimodal signaling. Many areas of the cortex that have traditionally been thought to be unimodal have been revealed to be multimodal. Furthermore, some of the areas thought to provide evidence for conceptual grounding—​such as those involved in gustation and spatial processing—​are demonstrably multimodal. Reorienting our conception of grounded cognition to one that fully embraces multimodality leaves room for researchers to explore the degree to which higher-​level perceptual and motor representations encode our concepts. Researchers are now free to investigate the degree to which higher-​ level representations help bind together sensorimotor features associated with our concepts (Kuhnke, Kiefer, & Hartwigsen, 2021). In keeping with this, I demonstrated that an emerging body of evidence from cognitive neuroscience and neuropsychology implicates higher-​level representations in semantic memory. Additional evidence points to semantic “hubs” that interact with experiential “spokes.” Hub-​and-​spoke theories provide another way in which hierarchical representations might help with the problem of generalization.

10.4  The Influence of Language Next, I turned to the problem of disembodiment. Here, I argued that language contributes to our capacity to acquire and employ concepts that go beyond our direct experience. Language plays an important role in the acquisition and use of many, if not most, of our concepts, but it tends to be more important for abstract concepts—​particularly ones with emotionally neutral content. Grounded cognition contends that our concepts are constructed out of the elements of our experiences. Since a great deal of our experiences involve or are mediated by language, language itself is a potential source of grounding. The language is an embodied neuroenhancement and scaffold (LENS) theory proposes that language enhances cognition in two important ways: (i) it transforms our external cognitive niche by offering external access to a shareable physical symbol system, and (ii) it transforms our conceptual system by giving us internal access to a compositional and productive set of grounded symbols that are for the most part disembodied with respect to their semantic content (leaving room for iconicity and other sources of non-​arbitrariness).

The Elastic Mind  203 Linguistic forms are grounded because they involve actions, sights, and sounds, but they are also free to capture content in a manner that is not specifically tied to their grounding. The LENS theory supplies specific predictions concerning the contributions of language to our concepts. First, the presence of linguistic labels should help us collect and access multimodal information associated with category exemplars. Second, the associations of words with other words should help us encode semantic content. Third, syntactic properties should constrain and support our concepts. Fourth, knowledge related to the generation and understanding of conversations should help us interpret abstract concepts. Substantial bodies of evidence support each of these hypotheses. Although I have emphasized the degree to which language represents a solution to the problem of disembodiment, I should also point out that it likely helps our brains overcome the problems of generalization and flexibility. For instance, having access to linguistic symbols is likely to increase our capacity for generalization because labels can serve as a neurocognitive means of connecting disparate experiences of category exemplars. Conversation-​based knowledge is also likely to help us deploy concepts in a flexible manner.

10.5  Synchronic and Diachronic Flexibility Next, I focused on the problem of flexibility. Flexibility can be defined either synchronically or diachronically. Synchronically, abstract concepts vary more in how they are realized than do concrete concepts. Diachronically, the representation of abstract concepts changes over time, and different sorts of abstract concepts are acquired at different stages of development. Importantly, flexibility is a matter of degree. All concepts are flexible; it is just that abstract concepts are more flexible. This fits with evidence that abstract concepts tend to rate higher in terms of their semantic diversity (Hoffman, Lambon Ralph, & Rogers, 2013) and situational systematicity (Davis, Altmann, & Yee, 2020). A prediction of the elastic approach is that abstract concepts themselves should vary in the sorts of information they encode. This generalization is supported by a growing number of behavioral and imaging studies that implicate the flexible contribution of emotion, language, and social cognition systems to abstract concepts. Although the abstract–​concrete distinction has played an important role in adult research on semantic memory for several decades, the acquisition

204  Abstract Concepts and the Embodied Mind of abstract concepts has only recently become a prominent research topic. Fortunately, there has been an explosion of research on the acquisition of abstract word meanings. What emerges from this research is that sources of grounding play different roles at different stages of development. Early in the acquisition process, nonarbitrary aspects of language use, such as iconicity and indexicality, make important contributions to word learning. These aspects appear to fade in terms of their importance to word learning (but not their importance to online face-​to-​face speech) as abstract words become more prevalent. Linguistic factors such as proximity and association appear to be important throughout development. Affective information appears to be particularly helpful during a period of rapid growth of abstract word learning that occurs later in development (around 8–​9 years old). After that period, children begin to acquire neutral abstract words. This process appears to rely heavily on language.

10.6  Metaphor: A Case Study Metaphors enable us to think about abstract domains in terms of more concrete ones. This has led some to speculate that metaphors might be an important component of abstract concepts. A controversy exists, though, concerning whether metaphor is primarily an embodied/​grounded phenomenon or a linguistic/​discourse phenomenon. The elastic approach provides a synthetic solution to this debate. Metaphorical thinking incorporates both elements: that is, it depends on both grounded simulations and language. This approach fits with three generalizations from the neurocognitive and neuropsychological research on metaphor. The first is that the influence of grounded simulations is limited in scope. Much of the positive evidence involves primary metaphors that typically emerge early in development and are shared by multiple distinct concepts. The second is that grounded simulations are sensitive to context. Several studies indicate that they are more likely to occur with novel metaphors than with conventional ones. The third is that conventional metaphors tend to rely more on linguistic information than do novel metaphors. There is a tendency to view metaphor as a unified solution to the theoretical challenges posed by abstract concepts. This idea fits poorly with the extant evidence, which suggests that metaphors are flexible, variable, and shaped by discourse. The elastic approach enables us to reconsider metaphor

The Elastic Mind  205 in a way that recognizes its complexity. Metaphor represents a further stretching of our grounded resources that facilitates our capacity to think in ways that go beyond our experience.

10.7  The Elasticity of Abstract Concepts Throughout this book, I have considered the ways in which the challenges posed by abstract concepts should cause us to rethink the nature of grounded cognition. I have argued that a diverse body of evidence suggests that our concepts are handled by a distributed, flexible, and hierarchical neurological system. In this section, I am going to turn the tables a bit and ask what the existence of such a conceptual system reveals about abstract concepts themselves. In other words, I am going to examine the ways in which the elasticity of our conceptual system should cause us to rethink the nature of abstract concepts. To accomplish this, I am going to adapt an example developed by Wilson (2006). He examines the seemingly straightforward abstract concept of hardness. He points out that philosophers have traditionally disagreed about whether it picks out a physical property, a perceptual property, or some combination of both, but they have rarely questioned the assumption that it has a fixed and determinate content. He argues that, when one looks carefully at how this concept is applied within the material sciences, it becomes apparent that no widely accepted definition of hardness exists. An indication of this is that there are several distinct ways of measuring hardness. Intriguingly, each of these is grounded in the different sorts of actions that we might carry out on objects: when faced with the task of testing a material for hardness—​say a piece of wood, metal, or plastic—​we are likely to punch at it, rap on it, or squeeze it. Wilson points out that material scientists have operationalized these actions and developed calibrated machines that provide distinct objective measurements (Figure 10.1). Two considerations are important for Wilson’s case that there is no single underlying concept. The first is that these measures often yield conflicting results. A particular material may rate as relatively hard on one measure but not on another. The second is that different measures appear to work better with different materials: what is useful for testing the hardness of metals is not as useful for testing the hardness of plastics. Wilson goes on to quote

206  Abstract Concepts and the Embodied Mind (a)

Punching

Brinell indentor

(b)

Squeezing

Durometer

(c)

Rapping

Scleroscope

Figure 10.1  Different ways of measuring hardness (Wilson, 2006). Each form of action is associated with a distinct apparatus designed to provide a calibrated assessment of hardness.

the Metals Handbook published by the American Society for Metals (Wilson, 2006, p. 337): The definition of hardness varies depending upon the experience or background of the person conducting the test or interpreting the data. To the metallurgist, hardness is the resistance to indentation; to the design engineer, a measure of flow stress; to the lubrication engineer, the resistance to wear; to the mineralogist, the resistance to scratching; and to the machinist, the resistance to cutting.

The Elastic Mind  207 This variability suggests that it is unlikely that there is some underlying microstructural property that explains the various phenomena associated with hardness. Defenders of the orthodox conception of invariant concepts would likely posit that hardness is polysemous. They might point to previous instances in the history of science in which this sort of semantic disagreement was eventually resolved and suggest that we just haven’t hit on the correct definition of hardness yet. In support of this idea, they might highlight the fact that each of the listed definitions involves an appeal to resistance. A weakness of this gambit is that it remains an open question whether resistance is itself applied in a consistent manner. Alternatively, a traditionalist might raise the possibility that hardness is not a natural kind and may ultimately need to be eliminated as a general term in the material sciences. Notice, though, that this proscriptive suggestion in no way undermines the usefulness of our current conceptions of hardness. Wilson takes a different route. He suggests that scientific concepts rarely have precise contents that can be expressed by means of necessary and sufficient conditions but, instead, are provisional cognitive tools adapted to specific tasks against the background of shifting practical and theoretical concerns. This conditionality is the rule rather than the exception. Many of the rich details of Wilson’s innovative account are beyond the purview of this book. What is important for my purposes is that, even within the constrained, deliberate, and quantified context of applied science, abstract concepts can exhibit a sui generis flexibility that is grounded in experience. The different conceptions of hardness in the material sciences emerge through a process of abstracting away from grounded interactions with objects. Their usefulness, though, relies in large part in their being formalized, operationalized, and measured by calibrated means. Rather than think about scientific concepts in terms of an idealized theory of the world, it is useful to examine them in the context of how they are applied in practice. When we do, we find that fixing the content of abstract concepts takes effort and often occurs in a piecemeal fashion. Abstract concepts are built up over time through dynamic interaction with the world; they emerge through a gradual process of research and development. A distributed, flexible, and hierarchical conceptual system has the resources to stretch grounded representations to handle abstract contents in a way that can be adapted to circumscribed contexts. This process is often

208  Abstract Concepts and the Embodied Mind iterative and may ultimately lead to increasing levels of abstraction. It may involve stretching concepts tied to direct sensorimotor experience (as it appears to be with the various conceptions of hardness), but it may also involve stretching concepts associated with complex affective, linguistic, or social experiences. The heterogeneity of abstract concepts is a predictable feature of the elastic mind. Given the effort required to stretch grounded resources to handle disembodied contents, we should expect abstract concepts to vary in terms of the grounded mechanisms they employ.

10.8  The Future Is Now Research programs change over time in response to new evidence and theories. Early advocates of grounded cognition focused almost exclusively on concrete concepts and the role that the redeployment of sensorimotor representations might play in them. As researchers expanded their experimental purview to include different varieties of concepts, they have begun to speculate that the emotions and other inner resources might play an important role, that higher-​level areas might need to be part of the story, and that simulations might vary with task and context. In these pages, I have defended explicit versions of these hypotheses and argued that the human conceptual system is fundamentally elastic. Our concepts rely on the neural reuse of experiential systems, but they do so in a way that matches up poorly with the theoretical presuppositions of strong embodiment. If this is correct, then we need to rethink how we investigate semantic memory. Elastic concepts require different research methods and experimental paradigms. For instance, if concepts are fundamentally multimodal, then we need to explore multiple possible sources of grounding. It is simply not enough to combine intuitive sensorimotor associations with positive evidence. We saw a clear demonstration of this with the affective embodiment account (AEA) approach, which attempted to draw broad generalizations from circumscribed evidence implicating the emotions in abstract concepts. This approach failed to be comprehensive in at least two ways. First, there are good reasons to think that many abstract concepts do not rely exclusively on emotional content (Desai, Reilly, & van Dam, 2018). Second, even when we limit our focus to those abstract concepts that are likely involve the emotions, evidence suggests that other sensorimotor systems are also involved in their processing (Dreyer & Pulvermüller, 2018; Moseley et al., 2012). These

The Elastic Mind  209 conceptual oversights even led researchers to make problematic assumptions in designing experiments (Skipper & Olson, 2014). Strong embodiment offers a facile parsimony, but it is unsustainable. Not only does it fit poorly with our current understanding of how action and perception systems work, but too many experiments implicate multimodal areas in conceptual tasks. Fortunately, abandoning this orthodox notion of how grounding works enables researchers to explore the role that hierarchical organization might play in specific concepts. Recent research on conceptual hierarchies provides a compelling proof of concept for an updated research program that leaves room for higher-​level experiential representations. Evidence from neuropsychological case studies of patients with forms of aphasia also point to the need for further research examining the role of hierarchical organization in semantic memory. The flexibility of concepts requires perhaps the most significant changes in methodology and experimental design. Much of the currently available research on concepts relies on categorization and linguistic processing tasks that involve the application of individual concepts in isolation. Our ability to generalize from this research rests heavily on the presumption of invariance. Flexible concepts, however, require experiments that manipulate task and context. Researchers need to explore the ways in which social factors and situations influence the activation of experiential systems. This should involve not only novel manipulations of task but also the embrace of interactive paradigms involving multiple participants and dynamic contexts. Our ability to acquire and use abstract knowledge is unrivaled in the natural world and requires explanation. Given that abstract knowledge lies at the heart of what makes us human, answering the question of how this knowledge is encoded within our concepts is central to our ability to understand ourselves. A successful theory of grounded cognition will need to embrace multimodal representations, hierarchical architecture, linguistic and cultural scaffolding, and semantic flexibility. In other words, it will need to recognize and explain the elasticity of our thoughts.

References Adams, F., & Aizawa, K. (2008). The bounds of cognition. Boston, MA: Blackwell. Adorni, R., & Proverbio, A. M. (2012). The neural manifestation of the word concreteness effect: An electrical neuroimaging study. Neuropsychologia, 50, 880–​891. Aggujaro, S., Crepaldi, D., Pistarini, C., Taricco, M., & Luzzatti, C. (2006). Neuroanatomical correlates of impaired retrieval of verbs and nouns: Interaction of grammatical class, imageability and actionality. Journal of Neurolinguistics, 19, 174–​194. Aizawa, K. (2017). Cognition and behavior. Synthese, 194(11), 4269–​4288. Akinina, Y., Dragoy, O. V., Ivanova, M. V., Iskra, E. V., Soloukhina, O. A., Petryshevsky, A. G., Fedina, O. N., Turken, A., Shklovsky, V. M., & Dronkers, N. F. (2019). Grey and white matter substrates of action naming. Neuropsychologia, 131, 249–​265. Alderson-​Day, B., & Fernyhough, C. (2015). Inner speech: Development, cognitive functions, phenomenology, and neurobiology. Psychological Bulletin, 141, 931–​965. Allen, C., & Bekoff, M. (1997). Species of mind: The philosophy and biology of cognitive ethology. Cambridge, MA: MIT Press. Almohammadi, A. (2017). The development of metaphor understanding in Arabic-​ speaking children. PhD dissertation, School of Commuinication, Education, and Public Policy, King’s College. Altarriba, J., & Bauer, L. M. (2004). The distinctiveness of emotion concepts: A comparison between emotion, abstract, and concrete words. American Journal of Psychology, 117(3), 389–​410. Amedi, A., Hofstetter, S., Maidenbaum, S., & Heimler, B. (2017). Task selectivity as a comprehensive principle for brain organization. Trends in Cognitive Science, 21, 307–​310. Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard University Press. Anderson, M. L. (2003). Embodied cognition: A field guide. Artificial Intelligence, 149(1), 91–​130. Anderson, M. L. (2010). Neural reuse: A fundamental organizational principle of the brain. Behavioral and Brain Sciences, 33, 245–​313. Anderson, M. L. (2014). After phrenology: Neural reuse and the interactive brain. Cambridge, MA: MIT Press. Andrews, M., Frank, S., & Vigliocco, G. (2014). Reconciling embodied and distributional accounts of meaning in language. Topics in Cognitive Science, 6, 359–​370. Andrews, M., Vigliocco, G., & Vinson, D. (2009). Integrating experiential and distributional data to learn semantic representations. Psychological Review, 116, 463–​498. Apperly, I. A., Samson, D., Carroll, N., Hussain, S., & Humphreys, G. (2006). Intact first-​ and second-​order false belief reasoning in a patient with severely impaired grammar. Social Neuroscience, 1(3–​4), 334–​348. Appleton, M., & Reddy, V. (1996). Teaching three year-​olds to pass false belief tests: A conversational approach. Social Development, 5(3), 275–​291. Ardila, A., & Rosselli, M. (1990). Acalculias. Behavioral Neurology, 3, 39–​48.

212 References Ardila, A., & Rosselli, M. (2002). Acalculia and dyscalculia. Neuropsychology Review, 12, 179–​231. Arévalo, A., Baldo, J. V., & Dronkers, N. F. (2012). What do brain lesions tell us about theories of embodied semantics and the human mirror neuron system? Cortex, 48, 242–​254. Arévalo, A., Perani, D., Cappa, S. F., Butler, A., Bates, E., & Dronkers, N. (2007). Action and object processing in aphasia: From nouns and verbs to the effect of manipulability. Brain and Language, 100, 79–​94. Aristotle. (1995). Aristotle: Selections (T. Irwin & G. Fine, Trans.). Indianapolis, IN: Hackett. Asano, M., Imai, M., Kita, S., Kitajo, K., Okada, H., & Thierry, G. (2015). Sound symbolism scaffolds language development in preverbal infants. Cortex, 63, 196–​205. Asch, S., & Nerlove, H. (1960). The development of double-​function terms in children: An exploratory investigation. In B. Kaplan & S. Wapner (Eds.), Perspectives in psychological theory: Essays in honor of Heinz Werner (pp. 47–​60). New York: International Universities Press. Astington, J. W., & Jenkins, J. M. (1999). A longitudinal study of the relation between language and theory-​ of-​ mind development. Developmental Psychology, 35(5), 1311–​1320. Auer, E. T., Bernstein, L. E., Sungkarat, W., & Singh, M. (2007). Vibrotactile activation of the auditory cortices in deaf versus hearing adults. Neuroreport, 18(7),645–​648. Aziz-​Zadeh, L., Wilson, S. M., Rizzolatti, G., & Iacoboni, M. (2006). Congruent embodied representations for visually presented actions and linguistic phrases describing actions. Current Biology, 16, 1818–​1823. Bak, T. H., & Hodges, J. R. (2004). The effects of motor neurone disease on language: Further evidence. Brain and Language, 89, 354–​361. Bak, T. H., O’Donovan, D. G., Xuereb, J. H., Boniface, S., & Hodges, J. R. (2001). Selective impairment of verb processing associate with pathological changes in Brodmann areas 44 and 45 in the motor neurone-​disease-​dementia-​aphasia syndrome. Brain, 124(Pt 1), 103–​120. Balaban, M. T., & Waxman, S. R. (1997). Do words facilitate object categorization in 9-​ month-​old infants? Journal of Experimental Child Psychology, 64(1), 3–​26. Baldo, J. V., & Dronkers, N. F. (2007). Neural correlates of arithmetic and language comprehension: A common substrate? Neuropsychologia, 45, 229–​235. Barca, L., Borghi, A. M., Binkofski, F., & Tummolini, L. (Eds.). (2018). Varieties of abstract concepts: Development, use and representation in the brain theme issue. Philosophical Transactions of the Royal Society B, 373(1752), 1–​7. Barca, L., Burani, C., & Arduino, L. S. (2002). Word naming times and psycholinguistic norms for Italian nouns. Behavior Research Methods, 34(3), 424–​434. Barca, L., Mazzuca, C., & Borghi, A. M. (2017). Pacifier overuse and conceptual relations of abstract and emotional concepts. Frontiers in Psychology, 8, Article 2014, 1–​19. Barca, L., Mazzuca, C., & Borghi, A. M. (2020). Overusing the pacifier during infancy sets a footprint on abstract words processing. Journal of Child Language, 47(5), 1084–​1099. Bardolph, M., & Coulson, S. (2014). How vertical hand movements impact brain activity elicited by literally and metaphorically related words: An ERP study of embodied metaphor. Frontiers in Human Neuroscience, 8, Article 1031, 1–​10. Barr, R. A., & Caplan, L. J. (1987). Category representations and their implications for category structure. Memory and Cognition, 15(5), 397–​418.

References  213 Barrós-​Loscertales, A., González, J., Pulvermüller, F., Ventura-​Campos, N., Bustamante, J. C., Costumero, V., Parcet, M. A., & Ávila, C. (2012). Reading salt activates gustatory brain regions: fMRI evidence for semantic grounding in a novel sensory modality. Cerebral Cortex, 22(11), 2554–​2563. Barsalou, L. W. (1983). Ad hoc categories. Memory & Cognition, 11, 211–​227. Barsalou, L. W. (1987). The instability of graded structure: Implications for the nature of concepts. In U. Neisser (Ed.), Concepts and conceptual development: Ecological and intellectual factors in categorization (pp. 101–​ 140). Cambridge: Cambridge University Press. Barsalou, L. W. (1991). Deriving categories to achieve goals. In G. H. Bower (Ed.), Psychology of learning and motivation (pp. 1–​64). New York: Academic Press. Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22, 577–​609. Barsalou, L. W. (2008). Grounded cognition. Annual Review of Psychology, 59, 617–​645. Barsalou, L. W. (2009). Simulation, situated conceptualization, and prediction. Philosophical Transactions of the Royal Society B, 363, 1281–​1289. Barsalou, L. W. (2010). Grounded cognition: Past, present, and future. Topics in Cognitive Science, 2, 716–​724. doi:10.1111/​j.1756-​8765. 2010.01115.x Barsalou, L. W. (2012). The human conceptual system. In M. Spivey, K. McRae, & M. Joanisse (Eds.), The Cambridge handbook of psycholinguistics (pp. 239–​ 258). New York: Cambridge University Press. Barsalou, L. W. (2015). Can cognition be reduced to action? Processes that mediate stimuli and responses make human action possible. In A. K. Engel, K. J. Friston, & D. Kragic (Eds.), The pragmatic turn: Toward action-​oriented views of cognitive science (pp. 81–​96). Cambridge, MA: MIT Press Barsalou, L. W. (2016a). On staying grounded and avoiding quixotic dead ends. Psychonomic Bulletin & Review, 23(4), 1122–​1142. Barsalou, L. W. (2016b). Situated conceptualization: Theory and applications. In Y. Coello & M. H. Fischer (Eds.), Foundations of embodied cognition, vol. 1: Perceptual and emotional embodiment (pp. 11–​37). East Sussex, UK: Psychology Press. Barsalou, L. W. (2017). Cognitively plausible theories of concept composition. In Y. Winter & J. A. Hampton (Eds.), Compositionality and concepts in linguistics and psychology (pp. 9–​30). London: Springer. Barsalou, L. W. (2020). Challenges and opportunities for grounding cognition. Journal of Cognition, 3(1), 31. Barsalou, L. W., Breazeal, C., & Smith, L. B. (2007). Cognition as coordinated non-​ cognition. Cognitive processing, 8(2), 79–​91. Barsalou, L. W., Dutriaux, L., & Scheepers, C. (2018). Moving beyond the distinction between concrete and abstract concepts. Philosophical Transactions of the Royal Society B, 373, 20170144, 1–​11. Barsalou, L. W., & Prinz, J. J. (1997). Mundane creativity in perceptual symbol systems. In S. M. Smith, J. Vaid, & T. B. Ward (Eds.), Conceptual structures and processes: Emergence, discovery, and change (pp. 267–​307). Washington, DC: American Psychological Association. Barsalou, L. W., Santos, A., Simmons, W. K., & Wilson, C. D. (2008). Language and simulation in conceptual processing. In M. De Vega, A. M. Glenberg, & A. C. Graesser (Eds.), Symbols and embodiment: Debates on meaning and cognition (pp. 245–​284). New York: Oxford University Press.

214 References Barsalou, L. W., & Wiemer-​Hastings, K. (2005). Situating abstract concepts. In D. Pecher & R. Zwaan (Eds.), Grounding cognition: The role of perception and action in memory, language, and thought (pp. 129–​163). Cambridge: Cambridge University Press. Barth, H., Kanwisher, N., & Spelke, E. S. (2003). The construction of large number representations in adults. Cognition, 86, 201–​221. Barth, H., LaMont, K., Lipton, J., Dehaene, S., Kanwisher, N., & Spelke, E. S. (2006). Non-​ symbolic arithmetic in adults and young children. Cognition, 98, 199–​222. Bartlett, F. (1932). Remembering: A study in experimental and social psychology. Cambridge: Cambridge University Press. Bauer, A. J., & Just, M. A. (2017). A brain-​based account of basic-​level concepts. Neuroimage, 161, 196–​205. Bechtel, W. (2008). Mental mechanisms: Philosophical perspectives on cognitive neuroscience. London: Routledge Bedny, M., Caramazza, A., Grossman, E., Pascual-​Leone, A., & Saxe, R. (2008). Concepts are more than percepts: The case of action verbs. Journal of Neuroscience, 28, 11347–​11353. Bedny, M., Caramazza, A., Pascual-​ Leone, A., & Saxe, R. (2012). Typical neural representations of action verbs develop without vision. Cereberal Cortex, 22(2), 286–​293. Bedny, M., Koster-​Hale, J., Elli, G., Yazzolino, L., & Saxe, R. (2019). There’s more to sparkle than meets the eye: Knowledge of vision and light verbs among congenitally blind and sighted individuals. Cognition, 189, 105–​115. Beilock, S. L., Lyons, I. M., Mattarella-​Micke, A., Nusbaum, H. C., & Small, S. L. (2008). Sports experience changes the neural processing of action language. Proceedings of the National Academy of Sciences, USA, 105, 13269–​13273. Benavides-​Varela, S., Piva, D., Burgio, F., Passarini, L., Rolma, G., Meneghello, F., & Semenza, C. (2017). Re-​assessing acalculia: Distinguishing spatial and purely arithmetical deficits in right-​hemisphere damaged patients. Cortex, 88, 151–​164. https://​ doi.org/​10.1016/​j.cortex.2016.12.014 Bergen, B. K. (2012). Louder than words: The new science of how the mind makes meaning. New York: Basic Books. Bergen, B. K., & Chang, N. C. (2005). Embodied construction grammar in simulation-​ based language understanding. In M. Fried & J.-​O. Östman (Eds.), Construction grammars: Cognitive grounding and theoretical extensions (pp. 147–​190). Amsterdam: John Benjamins. Bergen, B. K., Lindsay, S., Matlock, T., & Narayanan, S. (2007). Spatial and linguistic aspects of visual imagery in sentence comprehension. Cognitive Science, 31, 733–​764. Berio, L. (2021). Culturally embedded schemata for false belief reasoning. Synthese, 199, 285–​314. Bi, Y. (2021). Dual coding of knowledge in the human brain. Trends in Cognitive Sciences, 25(10), 883–​895. Binder, J. R. (2007). Effects of word imageability on semantic access: Neuroimaging studies. In J. Hart & M. A. Kraut (Eds.), Neural basis of semantic memory (pp. 149–​181). Cambridge: Cambridge University Press. Binder, J. R. (2016). In defense of abstract conceptual representations. Psychonomic Bulletin & Review, 23(4), 1096–​1108. Binder, J. R., Conant, L. L., Humphries, C. J., Fernandino, L., Simons, S. B., Aguilar, M., & Desai, R. H. (2016). Toward a brain-​based componential semantic representation. Cognitive Neuropsychology, 33(3–​4), 130–​174.

References  215 Binder, J. R., & Desai, R. H. (2011). The neurobiology of semantic memory. Trends in Cognitive Sciences, 15, 527–​536. Binder, J. R., Desai, R. H., Graves, W. W., & Conant, L. L. (2009). Where is the semantic system? A critical review and meta-​analysis of 120 functional neuroimaging studies. Cerebral Cortex, 19, 2767–​2796. Binder, J. R., Westbury, C., McKiernan, K., Possing, E., & Medler, D. (2005). Distinct brain systems for processing concrete and abstract concepts. Journal of Cognitive Neuroscience, 17, 905–​917. Binney, R. J., Embleton, K. V., Jeffries, E., Parker, G. J. M., & Lambon Ralph, M. A. (2010). The ventral and inferolateral aspects of the anterior temporal temporal lobe are crucial in semantic memory: Evidence from a novel direct comparison of distortion-​corrected fMRI, rTMS, and semantic dementia. Cerebral Cortex, 20, 2728–​2738. Binney, R. J., Parker, G. J. M., & Lambon Ralph, M. A. (2012). Convergent connectivity and graded specialization in the rostral human temporal lobe as revealed by diffusion-​ weighted imaging probabilistic tractography. Journal of Cognitive Neuroscience, 24, 1998–​2014. Bird, H., Franklin, S., & Howard, D. (2001). Age of acquisition and imageability ratings for a large set of words, including verbs and function words. Behavior Research Methods, Instruments, and Computers, 33, 73–​79. Blasi, D. E., Wichman, S., Hammarström, H., Stadler, P. F., & Christiansen, M. H. (2016). Sound-​meaning association biases evidenced across thousands of languages. PNAS, 113(39), 10818–​10823. Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. Journal of Machine Learning Research, 3, 993–​1022. Bloom, P., & German, T. P. (2000). Two reasons to abandon the false belief task as a test of theory of mind. Cognition, 77, B25–​B31. Blouw, P., Solodkin, E., Thagard, P., & Eliasmith, C. (2016). Concepts as semantic pointers: A framework and computational model. Cognitive Science, 40(5), 1128–​1162. Bocanegra, Y., García, A. M., Lopera, F., Pineda, D., Baena, A., Ospina, P., Alzate, D., Buriticá, O., Moreno, L., Ibáñez, A., & Cuetos, F. (2017). Unspeakable motion: Selective action-​verb impairments in Parkinson’s disease patients without mild cognitive impairment. Brain and Language, 168, 37–​46. Bocanegra, Y., García, A. M., Pineda, D., Buritica, O., Villegas, A., Lopera, F., Gómez, D., Gómez-​Aria, C., Cardona, J. F., Trujillo, N., & Ibáñez, A. (2015). Syntax, action verbs, action semantics, and object semantics in Parkinson’s disease: Dissociability, progression, and executive influences. Cortex, 69, 237–​254. Bohn, M., & Köymen, B. (2018). Common ground and development. Child Development Perspectives, 12(2), 104–​108. Bohnemeyer, J. (2021). Ten lectures on field semantics and semantic typology. Leiden: Brill. Bonini, L., Maranesi, M., Livi, A., Fogassi, L., & Rizzolatti, G. (2014). Space-​dependent representation of objects and other’s actions in monkey ventral premotor grasping neurons. Journal of Neuroscience, 34(11), 4108–​4119. Bonner, M. F., & Grossman, M. (2012). Gray matter density of auditory association cortex relates to knowledge of sound concepts in primary progressive aphasia. Journal of Neuroscience, 32, 7986–​7991. Bonner, M. F., Vesely, L., Price, C., Anderson, C., Richmond, L., Farag, C., Avants, B., & Grossman, M. (2009). Reversal of the concreteness effect in semantic dementia. Cognitive Neuropsychology, 26, 568–​579.

216 References Boone, W., & Piccinini, G. (2016). The cognitive neuroscience revolution. Synthese, 193(5), 1509–​1534. Borghesani, V., Buiatti, M., Eger, E., & Piazza, M. (2018). Conceptual and perceptual dimensions of word meaning are recovered rapidly and in parallel during reading. Journal of Cognitive Neuroscience, 29, 1–​14. Borghesani, V., & Piazza, M. (2017). The neuro-​cognitive representations of symbols: The case of concrete words. Neuropsychologia, 105, 4–​17. Borghi, A. M. (2020). A future of words: Language and the challenge of abstract concepts. Journal of Psychology, 3(1), 1–​18. Borghi, A. M., Barca, L., Binkofski, F. Castelfranchi, C., Pezzulo, G., & Tummolini, L. (2019). Words as social tools: Language, sociality and inner grounding in abstract concepts. Physics of Life Reviews, 29, 120–​153. Borghi, A., & Binkofski, F. (2014). Words as social tools: An embodied view on abstract concepts. New York: Springer. Borghi, A. M., Binkofski, F., Castelfranchi, C., Cimatti, F., Scorolli, C., & Tummolini, L. (2017). The challenge of abstract concepts. Psychological Bulletin, 143(3), 263–​292. Borghi, A. M., & Cimatti, F. (2009). Words as tools and the problem of abstract words meanings. In N. Taatgen & H. van Rijn (Eds.), Proceedings of the 31st Annual Conference of the Cognitive Science Society (pp. 2304–​2309). Amsterdam: Cognitive Science Society. Borghi, A. M., & Zarcone, E. (2016). Grounding abstractness: Abstract concepts and the activation of the mouth. Frontiers in Psychology, 7, Article 1498, 1–​16. Boulenger, V., Hauk, O., & Pulvermüller, F. (2009). Grasping ideas with the motor system: Semantic somatotopy in idiom comprehension. Cerebral Cortex, 19(8), 1905–​1914. Boulenger, V., Shtyrov, Y., & Pulvermüller, F. (2012) . When do you grasp the idea? MEG evidence for instantaneous idiom understanding. Neuroimage, 59(4), 3502–​3513. Boutonnet, B., & Lupyan, G. (2015) . Words jump-​start vision: A label advantage in object recognition. Journal of Neuroscience, 35(24), 9329–​9335. Bowdle, B. F., & Gentner, D. (2005). The career of metaphor. Psychological Review, 112, 193–​216. Bowerman, M., & Levinson, S. C. (Eds.). (2001). Langauge acquisition and conceptual development. Cambridge: Cambridge University Press. Bozeat, S., Lambon Ralph, M. A., Patterson, K., Garrard, P., & Hodges, J. R. (2000). Non-​ verbal semantic impairment in semantic dementia. Neuropsychologia, 38, 1207–​1215. Brown, J. R., Donelan-​McCall, N., & Dunn, J. (1996). Why talk about mental states? The significance of children’s conversations with friends, siblings, and mothers. Child Development, 67, 836–​849. Bruni, E., Tran, G. B., & Baroni, M. (2011). Distributional semantics from texts and images. In S. Pado & Y. Peirsman (Eds.), Proceedings of the EMNLP 2011 geometrical models for natural language semantics GEMS 2011 workshop (pp. 22–​32). Stroudsburg, PA: Association for Computational Linguistics. Bruni, E., Tran, N. K., & Baroni, M. (2014). Multimodal distributional semantics. Journal of Artificial Intelligence Research, 49, 1–​47. Brunyé, T. T., Walters, E. K., Ditman, T. Gagnon, S. A., Mahoney, C. R., & Taylor, H. A. (2012). The fabric of thought: Priming tactile properties during reading influences direct tactile perception. Cognitive Science, 36(8), 1449–​1467. Brysbaert, M., & Biemiller, A. (2017). Test-​based age-​of-​acquisition norms for 44 thousand English word meanings. Behavior Research Methods, 49(4), 1520–​1523.

References  217 Buccino, G., Riggio, L., Melli, G., Binkofski, M. F., Galese, V., & Rizzolati, G. (2005). Listening to action-​related sentences modulates the activity of the motor system: A combined TMS and behavioral study. Cognitive Brain Research, 24(3), 355–​363. Buchanan, L., Westbury, C., & Burgess, C. (2001). Characterizing semantic space: Neighborhood effects in word recognition. Psychonomic Bulletin & Review, 8, 531–​544. Buckner, C. (2018). Scaffolding intuitive rationality. In L. De Bruin, S. Gallagher, & A. Newen (Eds.), The Oxford handbook of 4E cognition (pp. 821–​840). Oxford: Oxford University Press. Burge, T. (1973). Reference and proper names. Journal of Philosophy, 70(14), 425–​439. Caggiano, V., Fleischer, F., Pomper, J. K., Giese, M. A., & Their, P. (2016). Mirror neurons in monkey premotor area F5 show tuning for critical features of visual causality perception. Current Biology, 26, 3077–​3082. Caggiano, V., Fogassi, L., Rizzolatti, G., Casile, A., Giese, M. A., & Thier, P. (2012). Mirror neurons encode the subjective value of an observed action. Proceedings of the National Academy of Sciences, 109, 11848–​11853. Caggiano, V., Fogassi, L., Rizzolatti, G., Pomper, J. K., Thier, P., Giese, M. A., & Casile, A. (2011). A view-​based encoding of mirror neurons of area f5 in macaque premotor cortex. Current Biology, 21(2), 144–​148. Caggiano, V., Fogassi, L., Rizzolatti, G., Thier, P., & Casile, A. (2009). Mirror neurons differentially encode peripersonal and extrapersonal space of monkeys. Science, 324, 403–​406. Calzavarini, F. (2017). Inferential and referential lexical semantic competence: A critical review of the supporting evidence. Neurolinguistics, 44, 163–​189. Calzavarini, F. (2021). The conceptual format debate and the challenge from global supramodality. British Journal for the Philosophy of Science. 1–​31. doi:10.1086/​717564 Cameron, L., Maslen, R., Todd, Z., Maule, J., Stratton, P., & Stanley, N. (2009). The discourse dynamics approach to metaphor and metaphor-​ led discourse analysis. Metaphor and Symbol, 24(2), 63–​89. Camp, E. (2006). Contextualism, metaphor, and what is said. Mind & Language, 21(3), 280–​309. Camp, E. (2009). Putting thoughts to work: Concepts, systematicity, and stimulus-​independence. Philosophy and Phenomenological Research, 78, 275–​311. Cangelosi, A., & Riga, T. (2006). An embodied model for sensorimotor grounding and grounding transfer: Experiments with epigenetic robots. Cognitive Science, 30, 673–​689. Caramazza, A., Anzellotti, S., Strnad, L., & Lingnau, A. (2014). Embodied cognition and mirror neurons: A critical assessment. Annual Review of Neuroscience, 37, 1–​15. Cardona, J. F., Kargieman, L., Sinay, V., Gershanik, O., Gelormini, C., Amoruso, L., Roca, M., Pineda, D., Trujillo, N., Michon, M., Garcia, A., Szenkman, D., Bekinschtein, T., Manes, F., & Ibanez, A. (2014). How embodied is action language? Neurological evidence from motor diseases. Cognition, 131, 311–​322. Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: MIT Press. Carey, S. (1991). Knowledge acquisition: Enrichment or conceptual change? In S. Carey & R. Gelman (Eds.), The epigenesis of mind: Essays on biology and cognition (pp. 257–​ 291). Hillsdale, NJ: Lawrence Erlbaum. Carston, R. (2002). Linguistic meaning, communicated meaning and cognitive pragmatics. Mind & Language, 17, 127–​148. Casanova, E. L., & Casanova, M. F. (2019). Defining Autism: A Guide to Brain, Biology, and Behavior. Philadelphia: Jessica Kingsley Publishers.

218 References Casasanto, D. (2008). Similarity and proximity: When does close in space mean close in mind? Memory and Cognition, 36, 1047–​1056. Casasanto, D. (2009). Embodiment of abstract concepts: Good and bad in right-​and left-​ handers. Journal of Experiment Psychology: General, 138(3), 351–​367. Casasanto, D. (2014). Experiential origins of mental metaphors: Language, culture, and the body. In M. Landau, B. P. Meier, & M. D. Robinson (Eds.), The power of metaphor: Examining its influence on social life (pp. 249–​268). American Psychological Association. Casasanto, D., & Bottini, R. (2014). Mirror reading can reverse the flow of time. Journal of Experimental Psychology: General, 143(2), 473–​479. Casasanto, D., & Bottini, R. (2014). Mirror reading can reverse the flow of time. Journal of Experimental Psychology: General, 143(2), 473–​479. Casasanto, D., & Chrysikou, E. G. (2011). When left is right: Motor fluency shapes abstract concepts. Psychological Science, 22(4), 419–​422. Casasanto, D., & Lupyan, G. (2015). All concepts are ad hoc concepts. In S. Laurence & E. Margolis (Eds.), The conceptual mind: New directions in the study of concepts (pp. 543–​ 566). Cambridge, MA: MIT Press. Caspers, S., Zilles, K., Laird, A. R., & Eickhoff, S. B. (2010). ALE meta-​analysis of action observation and imitation in the human brain. Neuroimage, 50(3), 1148–​1167. Catani, M. (2017). A little man of some importance. Brain, 140(11), 3055–​3061. Catricalà, E., Della Rosa, P. A., Plebani, V., Vigliocco, G., & Cappa, S. F. (2014). Abstract and concrete categories? Evidences from neurodegenerative diseases. Neuropsychologia, 64, 271–​281. Cecchetti, L., Ricciardi, E., Handjaras, G., Kupers, R., Ptito, M., & Pietrini, P. (2016). Congenital blindness affects diencephalic but not mesencephalic structures in the human brain. Brain Structure and Function, 221, 1465–​1480. Chao, L. L., Haxby, J. V., & Martin, A. (1999). Attribute–​based neural substrates in temporal cortex for perceiving and knowing about objects. Nature Neuroscience, 2, 913–​919. Chatterjee, A. (2010). Disembodying cognition. Language and Cognition, 2, 79–​116. Chemero, A. (2011). Radical embodied cognitive science. Cambridge, MA: MIT Press. Chen, E., Widick, P., & Chatterjee, A. (2008). Functional-​anatomical organization of predicate metaphor processing. Brain and Language, 107(3), 194–​202. Chen, J. L., Zatorre, R. J., Penhune, V. B. (2006). Interactions between auditory and dorsal premotor cortex to musical rhythms. Neuroimage, 32, 1771–​1781. Chiò, A., Pagani, M., Agosta, F., Calvo, A., Cistaro, A., & Filippi, M. (2014). Neuroimaging in amyotrophic lateral sclerosis: Insights into structural and functional changes. Lancet Neurology, 13(12), 1228–​1240. Chirimuuta, M. (2020). Naturalism and the philosophy of colour ontology and perception. Philosophy Compass, e12694, 1–​9. Chomsky, N. (1966). Cartesian linguistics: A chapter in the history of rationalist thought. New York: Harper and Row. Chow, H. M., Mar, R. A., Xu, Y., Liu, S., Wagage, S., & Braun, A. R. (2014) . Embodied comprehension of stories: Interactions between language regions and modality-​specific neural systems. Journal of Cognitive Neuroscience, 26(2), 279–​295. Christiansen, M. H., & Chater, N. (2016). The now-​or-​never bottleneck: A fundamental constraint on language. Behavioral and Brain Sciences, 39, 1–​72. Christie, S., & Gentner, D. (2014). Language helps children succeed on a classic analogy task. Cognitive Science, 38, 383–​397.

References  219 Cicone, M., Gardner, H., & Winner, E. (1981). Understanding the psychology in psychological metaphors. Journal of Child Language, 8(1), 213–​216. Cisek, P., & Kalaska, J. F. (2004). Neural correlates of mental rehearsal in dorsal premotor cortex. Nature, 431, 993–​996. Citron, F. M., & Goldberg, A. E. (2014a). Metaphorical sentences are more emotionally engaging than their literal counterparts. Journal of Cognitive Neuroscience, 26(11), 2585–​2595. Citron, F. M., & Goldberg, A. E. (2014b). Social context modulates the effect of physical warmth on perceived interpersonal kindness: A study of embodied metaphors. Language and Cognition, 6(1), 1–​11. Clark, A. (1986). Being There: Putting Brain, Body, and World Together Again. Cambridge, MA: MIT Press. Clark, A. (1998). Magic words: How language augments human computation. In J. Boucher & P. Carruthers (Eds.), Language and thought: Interdisciplinary themes (pp. 162–​183). Cambridge: Cambridge University Press. Clark, A. (2003). Natural-​born cyborgs. New York: Oxford University Press. Clark, A. (2006). Language, embodiment, and the cognitive niche. Trends in Cognitive Sciences, 10(8), 370–​374. doi:10.1016/​j.tics.2006.06.012 Clark, A. (2008). Supersizing the mind: Embodiment, action, and cognitive extension. Oxford: Oxford University Press. Clark, A., & Toribo-​Mateas, J. (1994). Doing without representing. Synthese, 101(3), 401–​431. Cohen, J. (2009). The red and the real: An essay on color ontology. Oxford: Oxford University Press. Cohen, R., Kelter, S., & Woll, G. (1980). Analytical competence and language impairment in Aphasia. Brain and Language, 10(2), 331–​347. Coltheart, M., Marshall, J. C., & Patterson, K. (Eds.). (1980). Deep dyslexia. London: Routledge & Kegan Paul. Conca, F., & Tettamanti, M. (2018). Conceptual semantics as grounded in personal experience. Phenomenology and Mind, 14, 98–​116. Connell, L. (2019). What have labels ever done for us? The linguistic shortcut in conceptual processing. Language, Cognition and Neuroscience, 34, 130–​1318. Connell, L., & Lynott, D. (2012). Strength of perceptual experience predicts word processing performance better than concreteness or imageability. Cognition, 125, 452–​465. Connell, L., & Lynott, D. (2013). Flexible and fast: Linguistic shortcut affects both shallow and deep conceptual processing. Psychonomic Bulletin & Review, 20(3), 542–​550. Connell, L., & Lynott, D. (2014a). I see/​hear what you mean: Semantic activation in visual word recognition depends on perceptual attention. Journal of Experimental Psychology: General, 143, 527–​533. Connell, L., & Lynott, D. (2014b). Principles of representation: Why you can’t represent the same concept twice. Topics in Cognitive Science, 6, 390–​406. Connell, L., Lynott, & Banks, B. (2018). Interoception: The forgotten modality in perceptual grounding of abstract and concrete concepts. Philosophical Transactions of the Royal Society B, 373, 20170143, 1–​9. Connell, L., Lynott, D., & Dreyer, F. (2012). A functional role for modality-​specific perceptual systems in conceptual representations. PLOS ONE, 7(3), Article e33321, 1–​7. Cordes, S., Gelman, R., & Gallistel, C. R. (2002). Variability signatures distinguish verbal from nonverbal counting for both large and small numbers. Psychological Bulletin and Review, 8, 698–​707.

220 References Courtin, C., & Melot, A. (1998). Development of theories of mind in deaf children. In M. D. Clark & M. Marschark (Eds.), Psychological perspectives on deafness (pp. 79–​102). Malwah, NJ: Erlbaum. Cousins, K. A. Q., Ash, S., & Grossman, M. (2018). Production of verbs related to body movement in amyotrophic lateral sclerosis (ALS) and Parkinson’s Disease (PD). Cortex, 100, 127–​139. Coutanche, M. N., & Thompson-​Schill, S. L. (2015). Creating concepts from converging features in human cortex. Cerebral Cortex, 25(9), 2584–​2593. Cowie, F. (1999). What’s within? Nativism reconsidered. Oxford: Oxford University Press. Cowie, F. (2017). Innateness and language. Stanford Encyclopedia of Philosophy, E. N. Zalta (Ed.). https://​plato.stanford.edu/​archives/​fall2017/​entries/​innateness-​language/​ Crawford, L. E., Margolies, S. M., Drake, J. T., & Murphy, M. E. (2007). Affect biases memory location: Evidence for the spatial representation of affect. Cognition and Emotion, 20(8), 1153–​1169. Croft, W., & Cruse, D. A. (2004). Cognitive linguistics. Cambridge: Cambridge University Press. Crutch, S. J., & Jackson, E. C. (2011). Contrasting graded effects of semantic similarity and association across the concreteness spectrum. Quarterly Journal of Experimental Psychology, 64, 1388–​1408. Crutch, S. J., Troche, J., Reilly, J., & Ridgway, G. R. (2013). Abstract conceptual feature ratings: The role of emotion, magnitude, and other cognitive domains in the organization of abstract conceptual knowledge. Frontiers in human neuroscience, 7, Article 186, 1–​14. Crutch, S. J., & Warrington, E. K. (2005). Abstract and concrete concepts have structurally different representational frameworks. Brain, 128, 615–​627. Crutch, S. J., & Warrington, E. K. (2010). The differential dependence of abstract and concrete words upon associative and similarity-​based information: Complementary semantic interference and facilitation effects. Cognitive Neuropsychology, 27, 46–​71. Cuccio, V., & Gallese, V. (2018). A Peircean account of concepts: Grounding abstract in phylogeny through a comparative neuroscientific perspective. Philosophical Transactions of the Royal Society B, 373, 20170128, 1–​10. Culicover, P. W. (1999). Syntactic nuts. New York: Oxford University Press. Curtiss, S. (1977). Genie: A pscyholinguistic study of a modern-​ day wild child. New York: Academic Press. Dahl, R. (1988). Matilda. New York: Viking. Dale, E., & O’Rourke, J. (1981). The living word vocabulary, the words we know: A national vocabulary inventory. Dayton, MN: Dome, Inc. Damasio, A. R. (1989). The brain binds entities and events by multiregional activation from convergence zones. Neural Computation, 1(1), 123–​132. Davidoff, J., & Roberson, D. (2004). Preserved thematic and impaired taxonomic categorization: A case study. Language and Cognitive Processes, 19(1), 137–​174. Davidson, D. (1978). What metaphors mean. Critical Inquiry, 5(1), 31–​47. Davis, C. P., Altmann, G. T. M., & Yee, E. (2020). Situational systematicity: A role for schema in understanding the differences between abstract and concrete concepts. Cognitive Neuropsychology. Davis, C. P., & Yee, E. (2018). Features, labels, space, and time: Factors supporting taxonomic relationships in the anterior temporal lobe and thematic relationships in the angular gyrus. Language, Cognition and Neuroscience, 34(10), 1347–​1357.

References  221 Davis, C. P., & Yee, E. (2021). Building semantic memory from embodied and distributional language experience. WIREs Cognitive Science, 12(5), 1–​29. Dawson, M., Soulières, I., Gernsbacher, M. A., & Mottron, L. (2007). The level and nature of autistic intelligence. Psychological Science, 18(8), 657–​662. de Villiers, J. G. (2007). The interface of language and theory of mind. Lingua, 117(11), 1858–​1878. de Villiers, J. G., & Pyers, J. E. (2002). Complements to cognition: A longitudinal study of the relationship between complex syntax and false-​belief-​understanding. Cognitive Development, 17, 1037–​1060. de Villiers, P. A., Burns, F., & Pearson, B. (2003). The Role of language in the theory of mind development of language-​ impaired children: Complementing theories. In B. Beachley, A. Brown, & F. Conlin (Eds.), Proceedings of the 27th Annual Boston University Conference on Language Development (pp. 232–​242). Somerville, MA: Cascadilla Press. de Villiers, P. A., & de Villiers, J. G. (2012). Deception dissociates from false belief reasoning in deaf children: Implications for the implicit versus explicit theory of mind distinction. British Journal of Developmental Psychology, 30(1), 188–​209. de Villiers, P. A., & de Villiers, J. G. (2014). The role of language in theory of mind development. Topics in Language Disorders, 34(4), 313–​328. De Wilde, E., Vanoverberghe, V., Storms, G., & De Boeck, P. (2003). The instantiation principle re-​evaluated. Memory, 11(6), 533–​548. Deacon, T. (1997). The symbolic species: The co-​evolution of language and the brain. New York: W. W. Norton. Deen, B., & McCarthy, G. (2010). Reading about the actions of others: Biological motion imagery and action congruency influence brain activity. Neuropsychologia, 48, 1607–​1615. Dehaene, S., Piazza, M., Pinel, P., & Cohen, L. (2003). Three circuits for number processing. Cognitive Neuropsychology, 20, 487–​506. Del Maschio, N., Fedeli, D., Garafalo, G., & Buccino, G. (2022). Evidence for the concreteness of abstract language: A meta-​analysis of neuroimaging studies. Brain Sciences, 12(1), 32, 1–​18. Della Rosa, P. A., Catricalà, E., Vigliocco, G., & Cappa, S. F. (2010). Beyond the abstract-​ concrete dichotomy: Mode of acquisition, concreteness, imageability, familiarity, age of acquisition, context availability and abstractness norms for a set of 417 Italian words. Behavior Research Methods, 42(4), 1042–​1048. Dennis, M., Lazenby, A. L., & Lockyer, L. (2001). Inferential language in high-​ functionchildren with autism. Journal of Autism and Developmental Disorders, 31(1), 47–​54. Desai, R., Herter, T., Riccardi, N., Rorden, C., & Fridriksson, J. (2015). Concepts within reach: Action performance predicts action language processing in stroke. Neuropsychologia, 71, 217–​224. Desai, R., Reilly, M., & van Dam, W. (2018). The multifaceted abstract brain. Philosophical Transactions of the Royal Society B: Biological Sciences, 373, 20170122, 1–​19. Desai, R. H. (2021). Are metaphors embodied? The neural evidence. Psychological Research, 1–​16. 10.1007/​s00426–​021–​01604–​4. Advance online publication. Desai, R. H., Binder, J. R., Conant, L. L., Mano, Q. R., & Seidenberg, M. S. (2011) . The neural career of sensory-​motor metaphors. Journal of Cognitive Neuroscience, 23(9), 2376–​2386.

222 References Desai, R. H., Conant, L. L., Binder, J. R., Park, H., & Seidenberg, M. (2013). A piece of the action: Modulation of sensory-​motor regions by idioms and metaphors. NeuroImage, 83, 862–​869. Devlin, J. T., Rushworth, M. F. S., & Matthews, P. M. (2005). Category-​related activation for written words in the posterior fusiform is task specific. Neuropsychologia, 43, 69–​74. Di Paolo, E. (2009). Extended life. Topoi, 28(1), 9–​21. Di Pellegrino, G., Fadiga, L., Gallese, V., & Rizzolatti, G. (1992). Understanding motor events: A neurophysiological study. Experimental Brain Research, 91, 176–​180. Dimitrova-​Radojichikj, D. (2015). Concepts of colors in children with congenital blindness. Journal of Special Education and Rehabilitation, 16(1–​2), 7–​16. Dingemanse, M., Blasi, D. E., Lupyan, G., Christiansen, M. H., & Monaghan, P. (2015). Arbitrariness, iconicity, and systematicity in language. Trends in Cognitive Sciences, 19(10), 603–​615. Dolscheid, S., Hunnius, S., Casasanto, D., & Majid, A. (2014). Prelinguistic infants are sensitive to space-​pitch associations found across cultures. Psychological Science, 25(6), 1256–​1261. Dolscheid, S., Shayan, S., Majid, A., & Casasanto, D. (2013). The thickness of musical pitch: Psychophysical evidence for linguistic relativity. Psychological Science, 24(5), 613–​621. Donald, M. (1993). Précis of origins of the modern mind: Three stages in the evolution of culture and cognition. Behavioral and Brain Sciences, 16, 739–​791. Dove, G. (2009). Beyond perceptual symbols: A call for representational pluralism. Cognition, 110, 412–​431. Dove, G. (2011). On the need for embodied and dis-​embodied cognition. Frontiers in Cognition, 1, Article 242, 1–​13. Dove, G. (2012). Grammar as a developmental phenomenon. Biology and Philosophy, 27(5), 615–​637. Dove, G. (2014). Thinking in words: Language as an embodied medium of thought. Topics in Cognitive Science, 6, 371–​389. Dove, G. (2016). Three symbol ungrounding problems: Abstract concepts and the future of embodied cognition. Psychonomic Bulletin & Review, 23(4), 1109–​1121. Dove, G. (2018). Language as a disruptive technology: Abstract concepts, embodiment, and the flexible mind. Philosophical Transactions of the Royal Society B, 373, 20170135, 1–​9. Dove, G. (2019). More than a scaffold: Language is a neuroenhancement. Cognitive Neuropsychology, 1–​24. Dove, G. (2021). The challenges of abstract concepts. In M. D. Robinson & L. E. Thomas (Eds.), Handbook of embodied psychology: Thinking, feeling, and acting (pp. 171–​195). Cham, Switzerland: Springer. Dove, G., Barca, L., Tummolini, L., & Borghi, A. M. (2020). Words have a weight: Language as a source of inner grounding and flexibility in abstract concepts. Psychological Research, 1–​17. 10.1007/​s00426–​020–​01438–​6. Advance online publication. Downing, P. E., & Peelen, M. V. (2011). The role of occipitotemporal body-​selective regions in person perception. Cognitive Neuroscience, 2, 186–​203. Dretske, F. (1981). Knowledge and the flow of information. Cambridge, MA: MIT Press. Dreyer, F. R., Frey, D., Arana, S., von Saldern, S., Picht, T., Vajkoczy, P., & Pulvermuller, F. (2015). Is the motor system necessary for processing action and abstract emotion words? Evidence from focal brain lesions. Frontiers in Psychology, 6, Article 1661, 1–​17.

References  223 Dreyer, F. R., & Pulvermuller, F. (2018). Abstract semantics in the motor system? An event-​related fMRI study on passive reading of semantic word categories carrying abstract emotional and mental meaning. Cortex, 100, 52–​70. Druks, J., & Shallice, T. (2000). Selective preservation of naming from description and the restricted preverbal message. Brain and Language, 72(2), 100–​128. Duñabeitia, J. A., Avilés, A., & Carreiras, M. (2008). NoA’s ark: Influence of the number of associates in visual word recognition. Psychonomic Bulletin & Review, 15, 1072–​1077. Dunn, J., Brown, J., Slomkowski, C., Tesla, C., & Youngblade, L. (1991). Young children’s understanding of other people’s feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 1352–​1366. Edmiston, P., & Lupyan, G. (2015). What makes words special? Words as unmotivated cues. Cognition, 143, 93–​100. Edwards, K. (2011). Higher-​ level concepts and their heterogeneous implementations: A polemical review of Edouard Machery’s doing without concepts. Philosophical Psychology, 24(1), 119–​133, doi:10.1080/​09515089.2010.544520 Eliasmith, C. (2013). How to build a brain: An architecture for neurobiological cognition. Oxford: Oxford University Press. Evangeliou, M. N., Raos, V., Galletti, C., & Savaki, H. E. (2009). Functional imaging of the parietal cortex during action execution and observation. Cerebral Cortex, 19(3), 624–​639. Evans, G. A. L., Lambon Ralph, M. A., & Woollams, A. M. (2012). What’s in a word? A parametric study of semantic influences on visual word recognition. Psychonomic Bulletin & Review, 19, 325–​331. Evans, N. (2011). Semantic typology. In J. J. Song (Ed.), The Oxford handbook of linguistic typology (pp. 504–​533). New York: Oxford University Press. Fargier, R. (2016). From modal brain structures to word meaning. In Y. Coello & M. H. Fischer (Eds.), Foundations of embodied cognition: Conceptual and interactive embodiment (pp. 11–​28). New York: Routledge. Farrant, B. M., Mayberry, M. T., & Fletcher, J. (2012). Language, cognitive flexibility, and explicit false belief understanding: Longitudinal analysis in typical development and specific language impairment. Child Development, 83(1), 223–​235. Faust, M. (2012). Thinking outside the left box: The role of the right hemispheres in novel metaphor comprehension. In M. Faust (Ed.), Advances in the neural substrates of language: Toward a synthesis of basic science and clinical research (pp. 425–​448). Malden, MA: Wiley-​Blackwell. Fedorenko, E., & Varley, R. (2016). Language and thought are not the same thing: Evidence from neuroimaging and neurological patients. Annals of the New York Academy of Sciences, 1369(1), 132–​153. Fernandino, L., Binder, J. R., Desai, R. H., Pendl, S. L., Humphries, C. J., Gross, W. L., Conant, L. L., & Seidenberg, M. S. (2015). Concept representation reflects multimodal abstraction: A framework for embodied semantics. Cerebral Cortex, 26(5), 2018–​2034. doi:10.1093/​cercor/​bhv020. Fernandino, L., Conant, L. L., Binder, J. R., Blindauer, K., Hiner, B., Spangler, K., & Desai, R. H. (2013b). Where is the action? Action sentence processing in Parkinson’s disease. Neuropsychologia, 51, 1510–​1517. Fernandino, L., Conant, L. L., Binder, J. R., Blindauer, K., Hiner, B., Spangler, K., & Desai, R. H. (2013a). Parkinson’s disease disrupts both automatic and controlled processing of action verbs. Brain and Language, 127, 65–​74.

224 References Fernandino, L., Humphries, C. J., Conant, L. L., Seidenberg, M. S., & Binder, J. R. (2016). Heteromodal cortical areas encode sensory-​motor features of word meaning. Journal of Neuroscience, 36(38), 9763–​9769. Fernandino, L., Humphries, C. J., Seidenberg, M. S., Gross, W. S., Conant, L., & Binder, J. R. (2015). Predicting brain activation patterns associated with individual lexicalconcepts based on five sensory-​motor attributes. Neuropsychologia, 76, 17–​26. Fernandino, L., & Iacoboni, M. (2010). Are cortical maps based on body parts or coordinated actions? Implications for embodied semantics. Brain and Language, 112, 44–​53. Ferrari, P. F., Gallese, V., Rizzolatti, G., & Fogassi, L. (2003). Mirror neurons responding to the observation of ingestive and communicative actions in the monkey ventral premotor cortex. Europen Journal of Neuroscience, 17, 1703–​1714. Ferrari, P. F., Rozzi, S., & Fogassi, L. (2005). Mirror neurons responding to observation of actions made with tools in the monkey ventral motor cortex. Journal of Cognitive Neuroscience, 17, 212–​226. Fillmore, C. J. (1977). Scenes-​and-​frames semantics. In A. Zampolli (Ed.), Linguistic structures processing (pp. 55–​81). Amsterdam: North Holland. Fillmore, C. J. (1982). Frame semantics. In Linguistic Society of Korea (Ed.), Linguistics in the morning calm: Selected papers from SICOL-​1991 (pp. 111–​137). Seoul: Hanshin. Fillmore, C. J. (1985). Frames and the semantics of understanding. Quaderni di Semantica, 6, 222–​254. Fillmore, C. J. (2012). Encounters with language. Computational Lingusitics, 38(4), 701–​718. Fillmore, C. J., & Kay, P. (1993). Construction grammar. Stanford, CA: CSLI Publications. Fillmore, C. J., Kay, P., & O’Connor, M. C. (1988). Regularity and idiomaticity in grammatical constructions: The case of let alone. Language, 64, 501–​538. Fingerhut, J., & Prinz, J. J. (2018). Grounding evaluative concepts. Philosophical Transactions of the Royal Society B, 373, 20170142, 1–​7. Fini, C., Zannino, G. D., Orsoni, M., Carlesimo, G. A., Benassi, M., & Borghi, A. M. (2021). Articulatory suppression delays processing of abstract words: The role of inner speech. Quarterly Journal of Experimental Psychology, 17470218211053623. Advance online publication. Firth, J. R. (1957). Papers in linguistics 1934–​1951. London: Oxford University Press. Fischer, M. H., & Shaki, S. (2018). Number concepts: Abstract and embodied. Philosophical Transactions of the Royal Society B, 373, 20170125, 1–​8. Fischer, M. H., & Zwaan, R. A. (2008). Embodied language: A review of the motor system in language comprehension. Quarterly Journal of Experimental Psychology, 6, 825–​850. Flombaum, J. (2002). The evolution of guesstimation. Cambridge, MA: Harvard University Press. Flumini, A., & Santiago, J. (2013). Time also flies from left to right if it is needed! In M. Knauff, M. Pauen, N. Sebanz, & I. Wachmuz (Eds.), Proceedings of the 36th annual conference of the cognitive science society (pp. 2315–​2320). Austin, TX: Cognitive Science Society. Fodor, J. A. (1975). The language of thought. Cambridge, MA: MIT Press. Fodor, J. A. (1981). The present state of the innateness controversy. In representations (pp. 257–​316). Cambridge, MA: MIT Press. Fodor, J. A. (1983). The modularity of mind. Cambridge, MA: MIT Press. Fodor, J. A., & Pylyshyn, Z. W. (1988). Connectionism and cognitive architecture: A critical analysis. Cognition, 28, 3–​71.

References  225 Fogassi, L., Ferrari, P. F., Gesierich, B., Rozzi, S., Chersim, F., Rizzolatti, G. (2005). Parietal lobe: From action organization to intention understanding. Science, 308, 662–​667. Forde, M. E., & Humphreys, G. W. (Eds.). (2005). Category specificity in the brain and mind. London: Taylor & Francis. Foxe, J. J., Wylie, G. R., Martinez, A., Schroeder, C. E., Javitt, D. C., Guilfoyle, D., Ritter, W., & Murray, M. M. (2002). Auditory-​somatosensory multisensory processing in auditory association cortex: An fMRI study. Journal of Neurophysiology, 88(1), 540–​543. Frank, M. C., Braginsky, M., Yurovsky, D., & Marchman, V. A. (2017). Wordbank: An open repository for developmental vocabulary data. Journal of Child Language, 44, 677–​694. Franklin, S., Howard, D., & Patterson, K. (1995). Abstract word anomia. Cognitive Neuropsychology, 12(5), 549–​566. Fulkerson, M. (2014). Rethinking the senses and their interactions: The case for sensory pluralism. Frontiers in Psychology, 5, Article 1426, 1–​14. Gainotti, G. (2006) . Anatomical, functional, and cognitive determinants of semantic memory disorders. Neuroscience and Biobehavioral Reviews, 30, 577–​594. Gallese, V. (2005). Embodied simulation: From neurons to phenomenal experience. Phenomenology and the Cognitive Sciences, 4, 23–​48. Gallese, V. (2009). Mirror neurons, embodied simulation, and the neural basis of social identification. Psychoanalytic Dialogues, 19, 519–​536. Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Action recognition in the premotor cortex. Brain, 119(Pt.2), 593–​609. Gallese, V., Fogassi, L., Fadiga, L., & Rizzolatti, G. (2002). Action representation and the inferior parietal lobe. In B. Hommel & W. Prinz (Eds.), Attention and performance XIX: Common mechanisms in perception and action (pp. 247–​266). Oxfork: Oxford University Press. Gallese, V., & Lakoff, G. (2005). The brain’s concepts: The role of the sensory–​motor system in reason and language. Cognitive Neuropsychology, 22, 455–​479. Gallistel, C. R. (1990). The organization of learning. Cambridge, MA: MIT Press. Garagnani, M., & Pulvermüller, F. (2016). Conceptual grounding of language in action and perception: A neurocomputational model of the emergence of category specificity and semantic hubs. European Journal of Neuroscience, 43(6), 721–​737. García, A. M., & Ibáñez, A. (2014). Words in motion: Motor-​language coupling in Parkinson’s disease. Translational Neuroscience, 5(2), 152–​159. García, A. M., Bocanegra, Y., Herrera, E., Moreno, L., Carmona, J., Baena, A., Lopera, F., Pineda, D., Melloni, M., Legaz, A., Muñoz, E., Sedeño, L., Baez, S., & Ibáñez, A. (2018). Parkinson’s disease compromises the appraisal of action meanings evoked by naturalistic texts. Cortex, 100, 111–​126. Garrard, P., & Carroll, E. (2006). Lost in semantic space: A multi-​modal, non-​verbal assessment of feature knowledge in semantic dementia. Brain, 129, 1152–​1163. doi:10.1093/​brain/​awl069 Gauthier, I., Anderson, A. W., Tarr, M. J., Skudlarski, P., & Gore, J. C. (1997). Levels of categorization in visual recognition studied using functional magnetic resonance imaging. Current Biology, 7(9), 645–​651. Gentner, D. (1983). Structure-​mapping: A theoretical framework for analogy. Cognitive Science, 7(2), 155–​170. Gentner, D. (1988). Metaphor as structure mapping: The relational shift. Child Development, 59(1), 47–​59.

226 References Gentner, D. (2006). Why verbs are hard to learn. In R. M. Golinkoff & K. Hirsh-​Pasek (Eds.), Action meets word: How children learn verbs (pp. 544–​564). Oxford: Oxford University Press. Gentner, D. (2010). Bootstrapping children’s learning: Analogical processes and symbol systems. Cognitive Science, 34, 752–​775. Gentner, D., Angggoro, F. K., & Klibanoff, R. S. (2011). Structure mapping and relational language support children’s learning of relational categories. Child Development, 82(4), 1173–​1188. Gentner, D., & Goldin-​Meadow, S. (Eds.). (2003). Language in mind: Advances in the study of language and thought. Cambridge, MA: MIT Press. Gentner, D., & Namy, L. L. (2006). Analogical processes in language learning. Current Directions in Psychological Science, 15, 297–​301. Gerfo, E. L., Oliveri, M., Torriero, S., Salerno, S., Koch, G., & Caltagirone, C. (2008). The influence of rTMS over prefrontal and motor areas in a morphological task: Grammatical vs. semantic effects. Neuropsychologia, 46, 764–​770. Ghazanfar, A. A., & Schroeder, C. E. (2006). Is neocortex essentially multisensory? Trends in Cognitive Science, 10(6), 278–​285. Ghio, M., Vaghi, M. M. S., & Tettamanti, M. (2013). Fine-​grained semantic categorization across the abstract and concrete domains. PLOS ONE, 8(6), e67090. Gibbs, R. W. (2006). Embodiment and cognitive science. New York: Cambridge University Press. Gibbs, R. W. (2017). The embodied and discourse views of metaphor: Why these are not so different and how they can be brought closer together. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 319–​334). Cambridge: Cambridge University Press. Gibson, J. J. (1986). The ecological approach to visual perception. Hillsdale, NJ: Lawrence Erlbaum. Giesbrecht, B., Gamblin, C., & Swaab, T. (2004). Separable effects of semantic priming and imageability on word processing in human cortex. Cerebral Cortex, 14(5), 521–​529. Gilbert, C. D., & Li, W. (2013). Top-​down influences on visual processing. Nature Reviews Neuroscience, 14, 350–​363. Gilboa, A., & Marlatte, H. (2017). Neurobiology of schemas and schema-​mediated memory. Trends in Cognitive Science, 21(8), 618–​631. Gleichgerrcht, E., Fridriksson, J., Rorden, C., Nesland, T., Desai, R., & Bonilha, L. (2016). Separate neural systems support representations for actions and objects during narrative speech in post stroke aphasia. NeuroImage: Clinical, 10, 140–​145. Gleitman, L. (1990). The structural sources of verb meanings. Language Acquisition: A Journal of Developmental Linguistics, 1(1), 3–​55. Gleitman, L., & Papafragou, A. (2005). Language and thought. In K. Holyoak & B. Morrison (Eds.), Cambridge handbook of thinking and reasoning (pp. 633–​ 661). Cambridge: Cambridge University Press. Gleitman, L. R., Cassidy, K., Nappa, R., Papafragou, A., & Trueswell, J. C. (2005). Hard words. Language Learning and Development, 1(1), 23–​64. Glenberg, A. M. (1997). What memory is for. Behavioral and Brain Sciences, 20(1), 1–​19. Glenberg, A. M., & Gallese, V. (2012). Action-​based language: A theory of language acquisition, comprehension, and production. Cortex, 48, 905–​922. Glenberg, A. M., & Kaschak, M. P. (2002). Grounding language in action. Psychonomic Bulletin & Review, 9, 558–​565.

References  227 Glenberg, A. M., & Robertson, D. A. (2000). Symbol grounding and meaning: A comparison of high-​dimensional and embodied theories of meaning. Journal of Memory and Langauge, 43, 379–​401. Glenberg, A. M., Sato, M., Cattaneo, L., Riggio, L., Palumbo, D., & Buccino, G. (2008). Processing abstract language modulates motor system activity. Quarterly Journal of Experimental Psychology, 61, 905–​919. Glock, H-​J. (2010). Concepts, Abilities, and Propositions. Grazer Philosophische Studien, 81(1), 115–​134. Glucksberg, S., & Keysar, B. (1990). Understanding metaphorical comparisons: Beyond similarity. Psychological Review, 97(1), 3–​18. Gold, R., & Faust, M. (2010). Right hemisphere dysfunction and metaphor comprehension in young adults with Asperger syndrome. Journal of Autism and Developmental Disorders, 40(7), 800–​811. Gold, R., Faust, M., & Goldstein, A. (2010). Semantic integration during metaphor comprehension in Asperger syndrome. Brain and Language, 113(3), 124–​134. Goldberg, A. (1995). A construction grammar approach to argument structure. Chicago: University of Chicago Press. Goldberg, R. F., Perfetti, C. A., & Schneider, W. (2006). Distinct and common cortical activations for multimodal semantic categories. Cognitive, Affective, & Behavioral Neuroscience, 6(3), 214–​222. Goldinger, S. D., Papesh, M. H., Barnhart, A. S., Hansen, W. A., & Hout, M. C. (2016). The poverty of embodied cognition. Psychonomic Bulletin and Review, 23(4), 959–​978. Goldin-​Meadow, S., & Feldman, H. (1977). The development of language-​like communication without a language model. Science, 197, 401–​403. Goldin-​Meadow, S., & Mylander, C. (1998). Spontaneous sign systems created by deaf children in two cultures. Nature, 391, 279–​281. Goldin-​Meadow, S., Mylander, C., & Butcher, C. (1995). The resilience of combinatorial structure at the word level: Morphology in self-​styled gesture systems. Cognition, 56, 195–​262. Goldstein, K. (1948). Language and language disturbances: Aphasic symptom complexes and their significance for medicine and theory of language. New York: Grune & Stratton. Goldstone, R. L., Feng, Y., & Rogosky, B. (2005). Connecting concepts to the world and each other. In D. Pecher & R. Zwaan (Eds.), Grounding cognition: The role of perception and action in memory, language, and thinking (pp. 292–​314). Cambridge: Cambridge University Press. Goldstone, R. L., & Hendrickson, A. T. (2009). Categorical perception. WIREs Cognitive Science, 1(1), 69–​78. Goldstone, R. L., & Kersten, A. (2003). Concepts and categorization. In A. F. Healy & R. W. Proctor (Eds.), Comprehensive handbook of psychology, vol. 4: Experimental psychology (pp. 599–​621). Hoboken, NJ: Wiley. Gonzalez, J., Barros-​ Loscertales, A., Pulvermuller, F., Meseguer, V., Sanjuan, A., Belloch, V., & Avila, C. (2006). Reading cinnamon activates olifactory brain regions. Neuroimage, 32, 906–​912. González-​Redondo, R., García-​García, D., Clavero, P., Gasca-​Salas, G., García-​Eulate, R., Zubieta, J., Arbizu, J., Obeso, J., & Rodríguez-​Oroz, M. (2014). Grey matter hypometabolism and atrophy in Parkinson’s disease with cognitive impairment: A two step process, Brain, 137(8), 2356–​2367.

228 References Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. Cambridge, MA: MIT Press. Goodglass, H., Hyde, M. R., & Blumstein, S. (1969). Frequency, picturability and availability of nouns in aphasia. Cortex, 5, 104–​119. Grady, J. (1997). Foundations of meaning: Primary metaphors and primary scenes. PhD dissertation, Department of Linguistics, University of California at Berkeley. Grady, J., & Ascoli, G. (2017). Sources and targets in primary metaphor theory: Looking back and thinking ahead. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 27–​45). Cambridge: Cambridge University Press. Grafton, S. (2020). Physical intelligence: The science of how the body and the mind guide each other through life. New York: Pantheon. Grafton, S., & Hamilton, A. (2007). Evidence for a distributed hierarchy of action representation in the brain. Human Movement Science, 26(4), 590–​616. Grana, A., Hofer, R., & Semenza, C. (2006). Aclalculia from a right hemisphere lesion: Dealing with where in multiplication procedures. Neuropsychologia, 44(14), 2972–​2986. Granito, C., Scorolli, C., & Boghi, A. M. (2015). Naming a Lego world: The role of language in the acquisition of abstract concepts. PLOS ONE, 10(1), e0114615, 1–​21. Graziano, M. S. A. (2008). The intelligent movement machine: An ethological perspective on the primate motor system. Oxford: Oxford University Press. Graziano, M. S. A. (2018). The space between us: A story of neuroscience, evolution, and human nature. Oxford: Oxford University Press. Graziano, M. S. A., & Aflalo, T. N. (2007). Mapping the behavioral repertoire onto the cortex. Neuron, 56(2), 239–​251. Graziano, M. S. A., & Gross, C. G. (1994). The representation of extrapersonal space: A possible role for bimodal, visual-​tactile neurons. In M. S. Gazzaniga (Ed.), The cognitive neurosciences (pp. 1021–​1034). Cambridge, MA: MIT Press. Grice, P. (1989). Studies in the ways of words. Cambridge, MA: Harvard University Press. Grondin, R., Lupker, S. J., & McRae, K. (2009). Shared features dominate semantic richness effects for concrete concepts. Journal of Memory and Language, 60, 1–​19. Gu, Y., Zheng, Y., & Swerts, M. (2019). Which is in front of Chinese people, past or future? The effect of language and culture on temporal gestures and spatial conceptions of time. Cognitive Science, 43, e12804, 1–​32. Günther, F., Rinaldi, L., & Marelli, M. (2019). Vector-​space models of semantic representation from a cognitive perspective: A discussion of common misconceptions. Perspectives on Psychological Science, 14(6), 1006–​1033. Hale, C. M., & Tager-​Flusberg, H. (2003). The influence of language on theory of mind: A training study. Developmental Science, 6(3), 346–​359. Halsband, U., Schmitt, J., Weyers, M., Binkofski, F., Grützner, G., & Freund, H. J. (2001). Recognition and imitation of pantomimed motor acts after unilateral parietal and premotor lesions: A perspective on apraxia. Neuropsychologia, 39, 200–​216. Hammarström, H. (2016). Linguistic diversity and language evolution. Journal of Language Evolution, 1, 19–​29. Hampe, B. (2017). Embodiment and discourse: Dimensions and dynamics of contemporary metaphor theory. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 3–​24). Cambridge: Cambridge University Press. Hampton, J. A. (1979). Polymorphous concepts in semantic memory. Journal of Verbal Learning and Verbal Behavior, 18, 441–​461.

References  229 Happé, F. (1993). Communicative competence and theory of mind in autism: A test of relevance theory. Cognition, 48, 101–​119. Hargreaves, I. S., & Pexman, P. M. (2012). Does richness lose its luster? Effects of extensive practice on semantic richness in visual word recognition. Frontiers in Human Neuroscience, 6, Article 234, 1–​11. Harnad, S. (1990). The symbol grounding problem. Physica D, 42, 335–​346. Harpaintner, M, Sim, E-​J., Trumpp, N. M., Ulrich, M., & Kiefer, M. (2020). The grounding of abstract concepts in the motor and visual system: An fMRI study. Cortex, 124, 1–​22. Harpaintner, M., Trumpp, N. M., & Kiefer, M. (2018). The semantic content of abstract concepts: A property listing study of 296 abstract words. Frontiers in Psychology, 9, Article 1748, 1–​16. Harris, P. L. (2005). Conversation, pretense, and theory of mind. In J. W. Astington & J. A. Baird (Eds.), Why language matters for theory of mind (pp. 70–​83). New York: Oxford University Press. Harris, R. A. (1993). The linguistic wars. New York: Oxford University Press. Hauk, O. (2016). What does it mean? A review of the neuroscientific evidence for embodied lexical semantics. In G. Hickok & S. Small (Eds.), Neurobiology of language (pp. 777–​788). New York: Elsevier. Hauk, O., Johnsrude, I., & Pulvermüller, F. (2004). Somatopic representation of action words in human motor and premotor cortex. Neuron, 41, 301–​307. Hauser, M. D., & Spelke, E. S. (2004). Evolutionary and developmental foundations of human knowledge: A case study of mathematics. In M. Gazzaniga (Ed.), The cognitive neurosciences (3rd ed., pp. 853–​864). Cambridge, MA: MIT Press. Heit, E, & Barsalou, L. W. (1996). The instantiation principle in natural categories. Memory, 4(4), 413–​451. Henrich, J. (2020). The weirdest people in the world: How the west became psychologically peculiar and particularly prosperous. New York: Farrar, Straus and Giroux. Hermann, I., Haser, V., van Elst, L. T., Ebert, D., Müller-​Feldmeth, D., Riedel, A., & Konieczny, L. (2013). Automatic metaphor processing in adults with Asperger syndrome: A metaphor interference effect task. European Archives of Psychiatry and Clinical Neuroscience, 263(Suppl 2), S177–​S187. Herrera, E., & Cuetos, F. (2012). Action naming in Parkinson’s disease patients on/​off dopamine. Neuroscience Letters, 513, 219–​222. Herrera, E., Rodriguez-​Fereiro, J., & Cuetos, F. (2012). The effect of motion content in action naming by Parkinson’s disease patients. Cortex, 48, 900–​904. Hesslow, G. (2012). The current status of the simulation theory of cognition. Brain Research, 1428, 71–​79. Hickok, G. (2014). The myth of mirror neurons: The real neuroscience of communication and cognition. New York: W. W. Norton & Company. Hills, D. (2017). Metaphor. Stanford Encyclopedia of Philosophy. https://​plato.stanf​ord. edu/​archi​ves/​fall2​017/​entr​ies/​metap​hor/​. Hills, T. T., Maouene, J., Riordan, B., & Smith, L. B. (2010). The associative structure of language: Contextual diversity in early word learning. Journal of Memory and Language, 63(3), 259–​273. Hodges, J., Bozeat, S., Lambon Ralph, M. A., Patterson, K., & Spatt, J. (2000). The role of conceptual knowledge in object use evidence from semantic dementia. Brain, 123, 1913–​1925.

230 References Hodges, J. R., Graham, N., & Patterson, K. (1995). Charting the progression in semantic dementia: Implications for the organisation of semantic memory. Memory, 3, 463–​495. Hoffman, P. (2016). The meaning of life and other abstract words: Insights from neuropsychology. Journal of Neuropsychology, 10, 317–​343. Hoffman, P., Binney, R. J., & Lambon Ralph, M. A. (2015). Differing contributions of inferior prefrontal and anterior temporal cortex to concrete and abstract conceptual knowledge. Cortex, 63, 250–​266. Hoffman, P., & Lambon Ralph, M. A. (2011). Reverse concreteness are not a typical feature of semantic dementia: Evidence for the hub-​and-​spoke model of conceptual representation. Cerebral Cortex, 21, 2103–​2112. Hoffman, P., & Lambon Ralph, M. A. (2018) . From percept to concept in the ventral temporal lobes: Graded hemispheric specialization based on stimulus and task. Cortex, 101, 107–​118. Hoffman, P., Lambon Ralph, M. A., & Rogers, T. T. (2013). Semantic diversity: A measure of semantic ambiguity based on variability in the contextual usage of words. Behavior Research Methods, 45, 718–​730. Hoffman, P., & Woollams, A. M. (2015). Opposing effects of semantic diversity in lexical and semantic relatedness decisions. Journal of Experimental Psychology: Human Perception and Performance, 41, 385–​402. Holcomb, P. J., Kounios, J., Anderson, J. E., & West, W. C. (1999). Dual-​coding, context-​ availability, and concreteness effects in sentence comprehension: An electrophysiological investigation. Journal of Experimental Psychology: Learning, Memory and Cognition, 25, 721–​742. Huang, H. W., Lee, C. L., & Federmeier, K. D. (2010). Imagine that! ERPs provide evidence for distinct hemispheric contributions to the processing of concrete and abstract concepts. NeuroImage, 49(1),1116–​1123. Hume, D. (2000). An enquiry concerning human understanding, T. Beauchamp (Ed.). Oxford: Oxford University Press. Hung, J., Wang, X., Wang, X., & Bi, Y. (2020). Functional subdivisions in the anterior lobes: A large scale meta-​analytic investigation. Neuroscience and Biobehavioral Reviews, 115, 134–​145. Hurley, S. (2008). The shared circuits model (SCM): How control, mirroring, simulation can enable imitation, deliberation, and mindreading. Behavioral and Brain Sciences, 31, 1–​58. Hutto, D. D., & Myin, E. (2013). Radicalizing enactivism: Basic minds without content. Cambridge, MA: MIT Press. Ibáñez, A., Cardona, J. F., Dos Santos, Y. V., Blenkmann, A., Aravena, P., Roca, M., Hurtado, E., Nerguizian, M., Amoruso, L., Gómez-​Arévalo, G., Chade, A., Dubrovsky, A., Gershanik, O., Kochen, S., Glenberg, A., Manes, F., & Bekinschtein, T. (2013). Motor-​language coupling: Direct evidence from early Parkinson’s disease and intracranial cortical recordings. Cortex, 49, 968–​984. Imai, M., & Kita, S. (2014). The sound symbolism bootstrapping hypothesis for language acquisition and language evolution. Philosophical Transactions of the Royal Society B, 369, 20130298, 1–​13. Imai, M., Kita, S., Nagumo, M., & Okada, H. (2008). Sound symbolism facilitates early verb learning. Cognition, 109, 54–​65. Iossifova, R., & Marmolejo-​Ramos, F. (2012). Spatial and temporal deixis. The role of age and vision in the ontogeny of a child’s spatial and temporal cognition. Journal of Speech and Language Pathology, 2(2), 75–​98.

References  231 Iurilli, G., Ghezzi, D., Olcese, U., Lassi, G., Nazzaro, C., Tonini, R., Tucci, V., Benfenati, F., & Medini, P. (2012). Sound-​driven synaptic inhibition in primary visual cortex. Neuron, 73(4–​2), 814–​828. Jackendoff, R. (2007). Linguistics in Cognitive Science: The state of the art. Linguistic Review, 24, 347–​401. Jamrozik, A., McQuire, M., Cardillo, E. R., & Chatterjee, A. (2016). Metaphor: Bridging embodiment to abstraction. Psychological Bulletin & Review, 23, 1080–​1089. Jankovic, J. (2008). Parkinson’s disease: Clinical features and diagnosis. Journal of Neurology, Neurosurgery, Psychiatry, 79(4), 368–​376. Jensen, T. W. (2017). Doing metaphor: An ecological perspective on metaphoricity. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 257–​276). Cambridge: Cambridge University Press. Jessen, F., Heun, R., Erb, M., Granath, D., Klose, U., Papassotiropoulos, A., & Grodd, W. (2000). The concreteness effect: Evidence for dual coding and context availability. Brain and Language, 74, 103–​112. Johns, B. T., & Jones, M. N. (2012). Perceptual inference from global lexical similarity. Topics in Cognitive Science, 4(1), 103–​120. Johnson-​Frey, S. H. (2004). The neural bases of complex tool use in humans. Trends in Cognitive Sciences, 8(2), 71–​78. Kable, J. W., Kan, I. P., Wilson, A., Thompson-​Schill, S. L., Chatterjee, A. (2005). Conceptual representations of action in the lateral temporal cortex. Journal of Cognitive Neuroscience, 17, 1855–​1870. Kable, J. W., Lease-​Spellmeyer, J., & Chatterjee, A. (2002). Neural substrates of action event knowledge. Journal of Cognitive Neuroscience, 14, 795–​805. Kalénine, S., Buxbaum, L. J., & Coslett, H. B. (2010). Critical brain regions for action recognition: Lesion symptom mapping in left hemisphere stroke. Brain, 133, 3269–​3280. Kalmijn, A. J. (1982). Electric and magnetic field detection in elasmobranch fishes. Science, 218(4575), 916–​918. Kantartzis, K., Imai, M., & Kita, S. (2011). Japanese sound-​symbolism facilitates word learning in English-​speaking children. Cognitive Science, 35(3), 575–​586. Kaschak, M. P., Madden, C. J., Therriault, D. J., Yaxley, R. H., Aveyard, M., Blanchard, A. A., & Zwaan, R. A. (2005). Perception of motion affects language processing. Cognition, 94(3), B79–​89. Kaschak, M. P., Zwaan, R. A., Aveyard, M., & Yaxley, R. H. (2006). Perception of auditory motion affects language processing. Cognitive Science, 30(4), 733–​744. Kasirer, A., & Mashal, N. (2014). Verbal creativity in autism: Comprehension and generation of metaphoric language in high-​functioning autism spectrum disorder and typical development. Frontiers in Human Neuroscience, 8, Article 615, 1–​8. Katz, R. B., & Goodglass, H. (1990). Deep dysphasia: Analysis of a rare form of repetition disorder. Brain and Language, 39(1), 153–​185. Keeley, B. (2000). Shocking lessons from electric fish: The theory and practice of multiple realization. Philosophy of Science, 67(3), 444–​465. Keeley, B. (2002). Making sense of the senses: Individuating modalities in humans and other animals. Journal of Philosophy, 99(1), 5–​28. Keeley, B. (2009). The role of neurobiology in differentiating the senses. In J. Bickle (Ed.), The Oxford handbook of philosophy and neuroscience (pp. 226–​250). Oxford: Oxford University Press.

232 References Keeley, B. (2015). Nonhuman animal senses. In M. Matthen (Ed.), The Oxford handbook of philosophy of perception (pp. 853–​870). Oxford: Oxford University Press. Kemmerer, D. (2005). The spatial and temporal meanings of English prepositions can be independently impaired. Neuropsychologia, 43(5), 797–​806. Kemmerer, D. (2006). The semantics of space: Integrating linguistic typology and cognitive neuroscience. Neuropsychologia, 44(9), 1607–​1621. Kemmerer, D. (2010). How words capture visual experience: The perspective from cognitive neuroscience. In B. Malt & P. Wolff (Eds.), Words and the mind: How words capture human experience (pp. 289–​329). New York: Oxford University Press. Kemmerer, D. (2015a). Are the motor features of verb meanings represented in the precentral motor cortices? Yes, but within the context of a flexible, multilevel architecture for conceptual knowledge. Psychonomic Bulletin and Review, 22, 10680–​1075. Kemmerer, D. (2015b). Are we ever aware of our concepts? A critical question to the Global Neuronal Workspace, Integrated Information, and Attended Intermediate-​ Level Representation theories of consciousness. Neuroscience of Consciousness, 1, 1–​10. Kemmerer, D. (2015c). Cognitive neuroscience of language. New York: Psychology Press. Kemmerer, D. (2017). Categories of object concepts across languages and brains: The relevance of nominal classification systems to cognitive neuroscience. Language, Cognition, and Neuroscience, 32, 401–​424. Kemmerer, D. (2019). Concepts in the brain: The view from cross-​linguistic diversity. New York: Oxford University Press. Kemmerer, D. (2021). What modulates the Mirror Neuron System during action observation? Multiple factors involving the action, the actor, the observer, the relationship between actor and observer, and the context. Progress in Neurobiology, 205, 102128. Kemmerer, D. (2022a). Cognitive Neuroscience of Language, 2nd Edition. New York: Routledge. Kemmerer, D. (2022b). Grounded cognition entails linguistic relativity. Manuscript submitted for publication.Kemmerer, D., & Gonzalez-​Castillo, J. (2010). The two level theory of verb meaning: An approach to integrating the semantics of action with the mirror system. Brain and Language, 112(1), 54–​76. Kemmerer, D., Rudrauf, D., Manzel, K., & Tranel, D. (2012). Behavioral patterns and lesion sites associated with impaired processing of lexical and conceptual knowledge of actions. Cortex, 48, 826–​848. Key, A. P. F, Dove, G. O., & Maguire, M. J. (2005). Linking brainwaves to the brain: An ERP primer. Developmental Neuropsychology, 27, 183–​215. Keysar, B., & Bly, B. M. (1999). Swimming against the current: Do idioms reflect conceptual structure? Journal of Pragmatics, 12(2), 1559–​1578. Keysar, B., Shen, Y., Glucksberg, S., & Horton, W. S. (2000). Conventional language: How metaphorical is it? Journal of Memory and Language, 43(4), 576–​593. Kiefer, M., & Barsalou, L. (2013). Grounding the human conceptual system in perception, action, and internal states. In M. Beisert, A. Herwig, & W. Prinz (Eds.), Action science: Foundations of an emerging discipline (pp. 381–​407). Cambridge, MA: MIT Press. Kiefer, M., & Pulvermüller, F. (2012). Conceptual representations in mind and brain: Theoretical developments, current evidence, and future directions. Cortex, 48, 805–​825. Kiefer, M., Sim, E.-​J., Herrnberger, B., & Hoenig, K. (2008). The sound of concepts: Four markers for a link between auditory and conceptual brain systems. Journal of Neuroscience, 28, 12224–​12230.

References  233 Kilner, J. M., & Lemon, R. N. (2013). What we currently know about mirror neurons. Current Biology, 23, R1057–​R1062. Kim, J. M., Sidhu, D. M., & Pexman, P. M. (2020). Effects of emotional valence and concreteness on children’s recognition memory. Frontiers in Psychology, 11, 615041, 1–​9. Kim, J. S., Elli, G. V., & Bedny, M. (2019). Knowledge of animal appearance among sighted and blind adults. Proceedings of the National Academy of Sciences, 116(23), 11213–​11222. Kohler, E., Keysers, C., Umiltà, M. A., Fogassi, L., Gallese, V., & Rizzolatti, G. (2002). Hearing sounds, understanding actions: Action representation in mirror neurons. Science, 297(5582), 846–​848. Kompa, N. A. (2019). Language and embodiment: Or the cognitive benefits of abstract representations. Mind and Language, 36, 27–​47. Kontra, C., Goldin-​Meadow, S., & Beilock, S. L. (2012). Embodied Learning Across the Life Span. Topics in Cognitive Science, 4, 731–​739. Kosslyn, S. M., Alpert, N. M., & Thompson, W. L. (1995). Identifying objects at different levels of hierarchy: A positron emission tomography study. Human Brain Mapping, 3(2), 107–​132. Kousta, S.-​T., Della Rosa, P. A., Cappa, S. F., & Vigliocco, G. (2007). The effect of mode of acquisition on lexical representation and processing. Paper presented at the 13th annual conference on architectures and mechanisms for language processing, Turku, Finland. Kousta, S.-​T., Vigliocco, G., Vinson, D. P., Andrews, M., & Del Campo, E. (2011). The representation of abstract words: Why emotion matters. Journal of Experimental Psychology: General, 140, 14–​34. Kousta, S.-​T., Vinson, D. P., & Vigliocco, G. (2009). Emotion words, regardless of polarity, have processing advantage over neutral words. Cognition, 112, 473–​481. Kövecses, Z. (2010). Metaphor: A practical introduction. Oxford: Oxford University Press. Kripke, S. (1980). Naming and necessity. Cambridge, MA: Harvard University Press. Kuhn, T. S. (1970). The structure of scientific revolutions (2nd ed., enlarged). Chicago: University of Chicago. Kuhnke, P., Kiefer, M., Hartwigsen, G. (2021). Task-​dependent recruitment of modality-​ specific and multimodal regions during conceptual processing. Cerebral Cortex, 31(7), 3475–​3493. Kuperman, V., Stadthagen-​Gonzalez, H., & Brysbaert, M. (2012). Age-​of-​acquisition ratings for 30,000 English words. Behavior Research Methods, 44(4), 978–​990. Kurata, K., & Tanji, J. (1986). Premotor cortex neurons in macaques: Activity before distal and proximal forelimb movements. Journal of Neuroscience, 6(2), 403–​411. Lacey, S., Stilla, R., & Sathian, K. (2012). Metaphorically feeling: Comprehending textural metaphors activates somatosensory cortex. Brain and Language, 120(3), 416–​421. Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press. Lakoff, G., & Johnson, M. (1980). Metaphors we live by. Chicago: University of Chicago Press. Lakoff, G., & Johnson, M. (1999). Philosophy in the flesh: The embodied mind and its challenge to western thought. New York: Basic Books. Lakoff, G., & Núñez, R. E. (2000). Where mathematics comes from: How the embodied mind brings mathematics into being. New York: Basic Books. Laland, K. N., Uller, T., Feldman, M., Sterelny, K., Muller, G., Moczek, A., Jablonka, E., & Odling-​Smee, J. (2015). The extended evolutionary synthesis: Its structure, assumptions and predictions. Proceedings of the Royal Society B, 282: 20151019.

234 References Lambon Ralph, M. A., Jefferies, E., Patterson, K., & Rogers, T. (2017). The neural and computational bases of semantic cognition. Nature Reviews Neuroscience, 18, 42–​55. Lambon Ralph, M. A., McClelland, J. L., Patterson, K., Galton, C. J., & Hodges, J. R. (2001). No right to speak? The relationship between object naming and semantic impairment: Neuropsychological evidence and a computational model. Journal of Cognitive Neuroscience, 13, 341–​356. Lambon Ralph, M. A., Sage, K., Jones, R. W., & Mayberry, E. J. (2010). Coherent concepts are computed in the anterior temporal lobes. Proceedings of the National Academy of Sciences, 107, 2717–​2722. Landa, R. J., Goldberg, M. C. (2005). Language, social, and executive functions in high functioning autism: A continuum of performance. Journal of Autism and Developmental Disorders, 35, 557–​573. Landau, B., & Gleitman, L. R. (1985). Language and experience: Evidence from the blind child. Cambridge, MA: Harvard University Press. Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato’s problem: The latent semantic analysis theory of acquisition, induction and representation of knowledge. Psychological Review, 104, 211–​240. Landauer, T. K., Foltz, P. W., & Laham, D. (1998). Introduction to latent semantic analysis. Discourse Processes, 25, 259–​284. Landy, D., Allen, C., & Zednik, C. (2014). A perceptual account of symbolic reasoning. Frontiers in Psychology, 5, 275. doi:10.3389/​fpsyg.2014.00275 Landy, D., & Goldstone, R. L. (2007a). Formal notations are diagrams: Evidence from a production task. Memory and Cognition 35(8), 2033–​2040. doi:10.3758/​BF03192935 Landy, D., & Goldstone, R. L. (2007b). How abstract is symbolic thought? Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(4), 720–​733. Lane, H. (1976). The wild boy of Aveyron. Cambridge, MA: Harvard University Press. Langacker, R. (1987a). Foundations of cognitive grammar: Theoretical prerequisites. Stanford, CA: Stanford University Press. Langacker, R. W. (1987b). Nouns and verbs. Language, 63(1), 53–​94. Langston, W. (2002). Violating orientational metaphors slows reading. Discourse Processes, 34, 381–​310. Lebois, L. A., Wilson-​Mendenhall, C. D., & Barsalou, L. W. (2015a). Are automatic conceptual cores the gold standard of semantic processing? The context-​dependence of spatial meaning in grounded congruency effects. Cognitive Science, 39(8), 1764–​1801. Lebois, L. A. M., Wilson-​Mendenhall, C. D., & Barsalou, L. W. (2015b). Putting everything in context. Cognitive Science, 39, 1987–​1995. Lecours, A. R., & Joanette, Y. (1980). Linguistic and other psychological aspects of paroxysmal aphasia. Brain and Language, 10(1), 1–​23. LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–​444. Lehky, S. R., & Tanaka, K. (2016). Neural representation for object recognition in inferotemporal cortex. Current Opinion in Neurobiology, 37, 23–​35. Lehmann, D., Pascual-​Marqui, R. D., Strik, W. K., & Koenig, T. (2010). Core networks for visual-​concrete and abstract thought content: A brain electric microstate analysis. NeuroImage, 49(1), 1073–​1079. doi:10.1016/​j.neuroimage.2009.07.054. Lenci, A., Baroni, M., Cazzolli, G., & Marotta, G. (2013). BLIND: A set of semantic feature norms from the congenitally blind. Behavior Research Methods, 45(4), 1218–​1233. Lenci, A., Lebani, G. E., & Passaro, L. C. (2018). The emotions of abstract words: A distributional semantic analysis. Topics in Cognitive Science, 10(3), 550–​572.

References  235 Leslie, A. M. (2001). Theory of mind. In The International Encyclopedia of the Social & Behavioral Sciences. 15652–​15656. Lewis, M., Zettersten, M., & Lupyan, G. (2019). Distributional semantics as a source of visual knowledge. PNAS Proceedings of the National Academy of Sciences of the United States of America, 116(39), 19237–​19238. Lingnau, A., & Downing, P. E. (2015). The lateral occipitotemporal cortex in action. Trends in Cognitive Sciences, 19, 268–​277. Lipton, J. S., & Spelke, E. S. (2003). Origins of number sense: Large number discrimination in human infants. Psychological Science, 15, 396–​401. Littlemore, J. (2019). Metaphors in the mind: Sources of variation in embodied metaphor. Cambridge: Cambridge University Press. Loftus, E. F. (1979). The malleability of human memory: Information introduced after we view an incident can transform memory. American Scientist, 67, 312–​320. Löhr, G. (2017). Abstract concepts, compositionality, and the contextualism-​invariantism debate. Philosophical Psychology, 30(6), 689–​710. Löhr, G. (2019). Embodied cognition and abstract concepts: Do concept empiricists leave anything out? Philosophical Psychology, 32(2), 161–​185. Loiselle, M., Rouleau, I., Nguyen, D. K., Dubeau, F., & Joubert, S. (2012). Comprehension of concrete and abstract words in patients with selective anterior temporal lobe resection and in patients with selective amygdalo-​hippocampectomy. Neuropsychologia, 50, 630–​639. Louwerse, M. (2011). Symbol interdependency in symbolic and embodied cognition. Topics in Cognitive Science, 3, 273–​302. Louwerse, M. (2018). Knowing the meaning of a word by the linguistic and perceptual company it keeps. Topics in Cognitive Science, 10(3), 573–​589. Louwerse, M. , & Jeuniaux, P. (2008). Language comprehension is both embodied and symbolic. In M. De Vega, A. M. Glenberg, & A. C. Graesser (Eds.), Symbols, embodiment, and meaning (pp. 309–​326). Oxford: Oxford University Press. Louwerse, M., & Jeuniaux, P. (2010). The linguistic and embodied nature of conceptual processing. Cognition, 114(1), 96–​104. Low, J. (2010). Preschoolers’ implicit and explicit false belief understanding: Relations with complex syntactical mastery. Child Development, 81(2), 597–​615. Lund, K., & Burgess, C. (1996). Producing high-​dimensional semantic spaces from lexical co-​occurrence. Behavior and Research Methods, Instruments, & Computers, 28, 203–​208. Lund, T. C., Sidhu, D. M., & Pexman, P. M. (2019). Sensitivity to emotion information in children’s lexical processing. Cognition, 190, 61–​71. Lupyan, G. (2009). Extracommunicative functions of language: Verbal interference causes selective categorization impairments. Psychonomic Bulletin & Review, 16(4), 711–​718. Lupyan, G. (2012a). Linguistically modulated perception and cognition: The label-​feedback hypothesis. Frontiers in Psychology, 3, Article 54, 1–​13. Lupyan, G. (2012b). What do words do? Toward a theory of language-​augmented thought. In B. H. Ross (Ed.), The psychology of learning and motivation, vol. 57 (pp. 255–​297). Amsterdam: Elsevier. Lupyan, G., & Bergen, B. (2015). How language programs the mind. Topics in Cognitive Science, 8(2), 408–​442. doi:10.1111/​tops.12155 Lupyan, G., & Clark, A. (2015). Words and the world: Predictive coding and the language-​ perception-​cognition interface. Current Directions in Psychological Science, 24(4), 279–​ 284. doi:10.1177/​0963721415570732

236 References Lupyan, G., & Mirman, D. (2013). Linking language and categorization: Evidence from aphasia. Cortex, 49(5), 1187–​1194. Lupyan, G., Rahman, R. A., Boroditsky, L., & Clark, A. (2020). Effects of language on visual perception. Trends in Cognitive Sciences, 24(11), 930–​944. Lupyan, G., Rakison, D. H., & McClelland, J. L. (2007). Language is not just for talking: Labels facilitate learning of novel categories. Psychological Science, 18(12), 1077–​1083. Lupyan, G., & Winter, B. (2018). Language is moe abstract than you think, or, why aren’t languages more iconic? Philosophical Transactions of the Royal Society of London. Series B, 373, 20170137, 1–​11. Lynott, D., & Connell, L. (2009). Modality exclusivity norms for 423 object properties. Behavior Research Methods, 41, 558–​564. Lynott, D., & Connell, L. (2010). Embodied conceptual combination. Frontiers in Psychology, 1, Article 212, 1–​14. Lynott, D., & Connell, L. (2013). Modality exclusivity norms for 400 nouns: The relationship between perceptual experience and surface word form. Behavior Research Methods, 45(2), 516–​526. Lynott, D., Connell, L., Brysbaert, M., Brand, J., & Carney, J. (2019) . The Lancaster sensorimotor norms: Multidimensional measures of perceptual and action strength for 40,000 English words. Behavior Research Methods, 1–​21. Lynott, D., Corker, K. S., Wortman, J., Connell, L., Donnellan, M. B., Lucas, R. E., & O’Brien, K. (2014). Replication of experiencing physical warmth promotes interpersonal warmth by Williams and Bargh (2008). Social Psychology, 45(3), 216–​222. Machery, E. (2007). Concept empiricism: A methodological critique. Cognition, 104, 19–​46. Machery, E. (2009). Doing without concepts. New York: Oxford University Press. Machery, E. (2010). Précis of doing without concepts. Behavioral and Brain Sciences, 33, 195–​206. Machery, E. (2015). By default. In S. Laurence & E. Margolis (Eds.), The conceptual mind: New directions in the study of concepts (pp. 567–​588). Cambridge, MA: MIT Press. Machery, E. (2016). The amodal brain and the offloading hypothesis. Psychonomic Bulletin & Review, 23, 1090–​1095. MacKay, G., & Shaw, A. (2004). A comparative study of figurative language in children with autistic spectrum disorders. Child Language Teaching and Therapy, 20, 13–​32. Macoir, J. (2009). Is a plum a memory problem? Longitudinal study of the reversal of the concreteness effect in a patient with semantic dementia. Neuropsychologia, 47, 518–​535. Maguire, M. J., & Dove, G. O. (2008). Speaking of events: Event word learning and event representation. In T. F. Shipley & J. M. Zacks (Eds.), Understanding events: From perception to action (pp. 193–​218). Oxford: Oxford University Press. Mahon, B. Z. (2015). What is embodied about cognition? Language, Cognition and Neuroscience, 30(4), 420–​429. Mahon, B. Z., & Caramazza, A. (2005). The orchestration of sensory-​motor systems: Clues from neuropsychology. Cognitive Neuropsychology, 22, 480–​494. Mahon, B. Z., & Caramazza, A. (2008). A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content. Journal of Physiology Paris, 102, 59–​70.

References  237 Majid, A. (2015). Comparing lexicons cross-​linguistically. In J. R. Taylor (Ed.), The Oxford handbook of the word (pp. 364–​379). New York: Oxford University Press. Mak, E., Su, L., Williams, G., Firbank, M., Lawson, R., Yarnall, A., Duncan, G., Owen, A., Khoo, T., Brooks, D., Rowe, J., Barker, R., Burn, D., & O’Brien, J. (2015). Baseline and longitudinal grey matter changes in newly diagnosed Parkinson’s disease: ICICLE-​PD study. Brain, 138(10), 2974–​2986. Maley, C. J. (2017). Toward analog neural computation. Minds and Machines, 1–​15. doi:10.1007/​s11023–​017–​9442–​5 Malt, B. C., & Majid, A. (2013). How thought is mapped into words. WIREs Cognitive Science, 4, 583–​597. Marconi, D. (1997). Lexical competence. Cambridge, MA: MIT Press. Margolis, E., & Laurence, S. (1999). Concepts: Core readings. Cambridge, MA: MIT Press. Margolis, E., & Laurence, S. (2015). The conceptual mind: New directions in the study of concepts. Cambridge, MA: MIT Press. Marinkovic, K., Dhond, R. P., Dale, A. M., Glessner, M., Carr, V., & Halgren, E. (2003). Spatiotemporal dynamics of modality-​specific and supramodal word processing. Neuron, 38(3), 487–​497. Markman, A. B., & Dietrich, E. (2000). In defense of representation. Cognitive Psychology, 40(2), 138–​171. doi:10.1006/​cogp.1999.0727 Marr, D. (1982). Vision: A computational investigation into the human representation and processing of visual information. Cambridge, MA: MIT Press. Marschark, M., & Paivio, A. (1977). Integrative processing of concrete and abstract sentences. Journal of Verbal Learning and Verbal Behavior, 16, 217–​231. Martin, A. (2007). The representation of object categories in the brain. Annual Review of Psychology, 58, 25–​45. Martin, N., & Saffran, E. M. (1992). A computational account of deep dysphasia: Evidence from a single case study. Brain and Language, 43, 240–​274. Mason, R. A., & Just, M. A. (2016). Neural representations of physics concepts. Psychological Science, 27, 904–​913. Massaro, D. W., & Perlman, M. (2017). Quantifying iconicity’s contribution during language acquisition: Implications for vocabulary learning. Frontiers in Communication, 2, Article 4, 1–​14. Matlock, T., & Winter, B. (2015). Experimental semantics. In B. Heine & H. Narrog (Eds.), The Oxford handbook of linguistic analysis (pp. 1–​25). Oxford: Oxford University Press. Matthen, M. (2005). Seeing, doing and knowing: A philosophical theory of sense perception. Oxford: Oxford University Press. Mazzuca, C., Lugli, L., Nicoletti, R., & Borghi, A. M. (2018). Abstract, emotional, and concrete concepts and the activation of mouth hand effectors. PeerJ: Life and Environment, 6, e5987, 1–​26. McCaffrey, J. (2015) . Reconceiving conceptual vehicles: Lessons from semantic dementia. Philosophical Psychology, 28, 337–​354. McCaffrey, J., & Machery, E. (2012). Philosophical issues about concepts. Wiley Interdisciplinary Reviews: Cognitive Science, 3, 265–​279. doi:10.1002/​wcs.1166 McCarthy, J. J. (1982). Prosodic structure and explitive infixation. Language, 58(3), 574–​590. McGregor, K. K., Oleson, J., Bahnsen, A., & Duff, D. (2013). Children with developmental language impairment have vocabulary deficits characterized by limited breadth and depth. International Journal of Language & Communication Disorders, 48(3), 307–​319.

238 References McRae, K., Nedjadrasul, D., Pau, R., Lo, B. P.‐H., & King, L. (2018). Abstract concepts and pictures of real‐world situations activate one another. Topics in Cognitive Science, 10(3), 518–​532. Meadow, K. P., Greenberg, M. T., Erting, C., & Carmichael, H. (1981). Interactions of deaf mothers and deaf preschool children: Comparisons with three other groups of deaf and hearing dyads. American Annals of the Deaf, 156(4), 454–​468. Meck, W. H., Church, R. M. (1983). A mode control model of counting and timing processes. Journal of Experimental Psychology: Animal Behavioral Processes, 9, 320–​334. Medin, D. L., & Schaffer, M. M. (1978). Context theory of classification learning. Psychological Review, 85, 207–​238. Medin, D. L., & Smith, E. E. (1981). Strategies and classification learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 7, 241–​253. Medina, T. N., Snedeker, J., Trueswell, J. C., & Gleitman, L. R. (2011). How words can and cannot be learned by observation. PNAS, 108(22), 9014–​9019. Meier, B. P., & Robinson, M. D. (2004). Why the sunny side is up: Associations between affect and vertical position. Psychological Science, 15(4), 243–​247. Meir, I. (2010). Iconicity and metaphor: Constraints on metaphorical use of iconic forms. Language, 86, 865–​896. Meir, I., & Cohen, A. (2018). Metaphor in sign languages. Frontiers in Psychology, 9, Article 1025, 1–​13. Mellem, M. S., Jasmin, K. M., Peng, C., & Martin, A. (2016). Sentence processing in anterior superior temporal cortex shows a social-​emotional bias. Neuropsychologia, 89, 217–​224. Melloni, M., Sedeño, L., Hesse, E., García-​Cordero, I., Mikulan, E., Plastino, A., Marcotti, A., López, J. D., Bustamante, C., Lopera, F., Pineda, D., García, A. M., Manes, F., Trujillo, N., & Ibáñez, A. (2015). Cortical dynamics and subcortical signatures of motor-​language coupling in Parkinson’s disease. Scientific Reports, 5, 11899. Melogno, S., D’Ardia, C., Pinto, M. A., & Levi, G. (2012). Explaining metaphors in high-​ functioning autism spectrum disorder children: A brief report. Research in Autism Spectrum Disorders, 6(2), 683–​689. Melogno, S., Pinto, M. A., & Levi, G. (2012). Metaphor and metonymy in ASD children: A critical review from a developmental perspective. Research in Autism Spectrum Disorders, 6(4), 1289–​1296. Menary, R. (2007). Cognitive integration: Mind and cognition unbounded. Basingstoke, UK: Palgrave Macmillan. Menary, R. (2010). The extended mind. Cambridge, MA: MIT Press. Meredith, M. (2001). Human veronasal organ function: A critical review of best and worst cases. Chemical Senses, 26(4), 433–​445. Mesulam, M. (1998). From sensation to cognition. Brain, 121, 1013–​1052. Meteyard, L., Bahrami, B., & Vigliocco, G. (2007). Motion detection and motion verbs –​ Language affects low level visual perception. Psychological Science, 18(11), 1007–​1013. Meteyard, L., Cuadrado, S. R., Bahrami, B., & Vigliocco. (2012). Coming of age: A review of embodiment and the neuroscience of semantics. Cortex, 48, 788–​804. Meteyard, L., Stoppard, E., Snudden, D., Cappa, S. F., & Vigliocco, G. (2015). When semantics aids phonology: A processing advantage for iconic word forms in aphasia. Neuropsychologia, 76, 264–​275. Meteyard, L., Zokaei, N., Bahrami, B., & Vigliocco, G. (2008). Visual motion interferes with lexical decision on motion verbs. Current Biology, 18(17), R732–​R733.

References  239 Meyer, K., & Damasio, A. (2009). Convergence and divergence in a neural architecture for recognition and memory. Trends in Neurosciences, 32(7), 376–​382. Michel, C. (2020). Concept contextualism through the lens of Predictive Processing. Philosophical Psychology, 33(4), 624–​647. Miller, C. A. (2004). False belief and sentence complement performance in children with specific language impairment. International Journal of Language & Communication Disorders, 39(2), 191–​213. Milner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford: Oxford University Press. Minshew, N. J., Goldstein, G., & Siegel, D. (1997). Neuropsychologic functioning in autism: Profile of a complex information processing disorder. Journal of the International Neuropsychological Society, 3, 303–​316. Moffat, M., Siakaluk, P. D., Sidhu, D. M., & Pexman, P. M. (2015). Situated conceptualization and semantic processing: Effects of emotional experience and context availability in semantic categorization and naming tasks. Psychonomic Bulletin & Review, 22, 408–​419. Monti, M., Parsons, L., & Osherson, D. (2012). Thought beyond language: Neural dissociation of algebra and natural language. Psychological Science, 23, 914–​922. Moore, R., Donelson, K., Eggleston, A., & Bohnemeyer, J. (2015). Semantic typology: New approaches to crosslinguistic variation in language and cognition. Linguistics Vanguard, 1, 189–​200. Moseley, R., Carota, F., Hauk, O., Mohr, B., & Pulvermuller, F. (2012). A role for the motor system in binding abstract emotional meaning. Cerebral Cortex, 22, 1634–​1647. Moseley, R. L., Mohr, B., Lombardo, M. V., Baron-​Cohen, S., Hauk, O., & Pulvermüller, F. (2013) . Brain and behavioral correlates of action semantic deficits in autism. Frontiers in Human Neuroscience, 7, Article 725, 1–​10. Moseley, R. L., Shtyrov, Y., Mohr, B., Lombardo, M. V., Baron-​Cohen, S., & Pulvermüller, F. (2015). Lost for emotion words: What motor and limbic brain activity reveals about autism and semantic theory. Neuroimage, 104, 413–​422. Müller, C. (2017). Waking metaphors: Embodied cognition in multimodal discourse. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 297–​316). Cambridge: Cambridge University Press. Muraki, E. J., Sidhu, D. M., & Pexman, P. M. (2020). Heterogeneous abstract concepts: Is ponder different from dissolve?. Psychological Research, 1–​17. Advance online publication. doi:10.1007/​s00426-​020-​01398-​x Murgiano, M., Motamedi, Y., & Vigliocco, G. (2021). Situating language in the real-​world: The role of multimodal iconicity and indexicality. Journal of Cognition, 4(1), 38, 1–​18. Murphy, G., & Medin, D. L. (1985). The role of theories in conceptual coherence. Psychological Review, 92(3), 289–​316. Murphy, G. L. (1996). On metaphoric representation. Cognition, 60, 173–​204. Murphy, G. L. (1997). Reasons to doubt the present evidence for metaphoric representation. Cognition, 62, 99–​108. Murphy, G. L. (2002). The big book of concepts. Cambridge, MA: MIT Press. Murphy, G. L., & Lassaline, M. E. (1997). Hierarchical structure in concepts and the basic level of categorization. In K. Lamberts & D. R. Shanks (Eds.), Studies in cognition. Knowledge, concepts and categories (pp. 93–​131). MIT Press. Murphy, G. L., & Wisniewski, E. J. (1989). Categorizing objects in isolation and in scenes: What a superordinate is good for. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15(4), 572–​586.

240 References Murray, M. M., Lewkowicz, D. J., Amedi, A., & Wallace, M. T. (2016). Multisensory processes: A balancing act across the lifespan. Trends in Neurosciences, 39(8), 567–​579. Myachykov, A., & Fischer, M. H. (2019). A hierarchical view of abstractness: Grounded, embodied, and situated aspects. Comment on words as social tools: Language, sociality, and inner grounding in abstract concepts by Anna M. Borghi et al. Physics of Life Reviews, 29, 161–​163. Namy, L. L., & Gentner, D. (2002). Making a silk purse out of two sow’s ears: Young children’s use of comparison in category learning. Journal of Experimental Psychology: General, 131(1), 5. Negri, G. A. L., Rumiati, R. I., Zadini, A., Ukmar, M., Mahon, B. Z., & Caramazza, A. (2007). What is the role of motor simulation in action and object recognition? Evidence from apraxia. Cognitive Neuropsychology, 24(8), 795–​816. Newcombe, P. I., Campbell, C., Siakaluk, P. D., & Pexman, P. M. (2012). Effects of emotional and sensorimotor knowledge in semantic processing of concrete and abstract nouns. Frontiers in Human Neuroscience, 6, Article 275, 1–​15. Newman, A., De Bruin, L., & Gallagher, S. (Eds.). (2018). The Oxford handbook of 4E cognition. Oxford: Oxford University Press. Newton, A. M., & de Villiers, J. G. (2007). Thinking while talking: Adults fail nonverbal false-​belief reasoning. Psychological Science, 18, 574–​579. Nielsen, A. K., & Dingemanse, M. (2021). Iconicity in word learning and beyond: A critical review. Language and Speech, 64(1), 52–​72. Noe, A. (2004). Action in perception. Cambridge, MA: MIT Press. Noppeney, U., & Price, C. J. (2004). Retrieval of abstract semantics. NeuroImage, 22, 164–​170. Noppeney, U., & Wallesch, C.-​W. (2000). Language and cognition: Kurt Goldstein’s theory of semantics. Brain and Langauge, 44(3), 367–​386. Norbury, C. F. (2005). The relationship between theory of mind and metaphor: Evidence from children with language impairment and autism spectrum disorder. Developmental Psychology, 23(3), 383–​399. Nunberg, G., Sag, I. A., & Wasow, T. (1994). Idioms. In S. Everson (Ed.), Language (pp. 491–​538). Cambridge: Cambridge University Press. Núñez, R. E., & Sweetser, E. (2006). With the future behind them: Convergenct evidence from Aymara language and gesture in the crosslinguistic comparison of spatial construals of time. Cognitive Science, 30, 401–​450. Ohala, M., & Ohala, J. (1972). The problem of aspiration in Hindi. Annual Bulletin (Research Institute of Logopedics and Phoniatrics, Faculty of Medicine, University of Tokyo, Japan), 39–​46. Olofsson, J. K., Bowman, N. E., Khatibi, K., & Gottfried, J. A. (2012). A Time-​Based Account of the Perception of Odor Objects and Valences. Psychological Science, 23(10), 1224–​1232. Oosterhof, N. N., Tipper, S. P., & Downing, P. E. (2012). Viewpoint (in)dependence of action representations: An MVPA study. Journal of Cognitive Neuroscience, 24(4), 975–​989. Ostarek, M., & Bottini, R. (2021). Towards strong inference in research on embodiment: Possibilities and limitations of causal paradigms. Journal of Cognition, 4(1), 5. Ouellet, M., Santiago, J., Israeli, Z., & Gabay, S. (2010). Is the future the right time? Experimental Psychology, 57(4), 308–​314. Paivio, A. (1971). Imagery and verbal processes. New York: Holt, Rinehart & Winston.

References  241 Paivio, A. (1986). Mental representations: A dual coding approach. New York: Oxford University Press. Paivio, A. (1991). Dual coding theory: Retrospect and current status. Canadian Journal of Psychology, 45, 255–​287. Paivio, A., Yuille, J. C., & Madigan, S. A. (1968). Concreteness, imagery and meaningfulness values for 925 words. Journal of Experimental Psychology Monograph Supplement, 76(1), part 2. Papagno, A., Fogliata, E., Catricalà, C., & Miniussi. (2009). The lexical processing of abstract and concrete nouns. Brain Research, 1263, 78–​86. Papeo, L., Pascual-​Leone, A., & Caramazza, A. (2013). Disrupting the brain to validate hypotheses on the neurobiology of language. Frontiers in Human Neuroscience, 7, Article 148, 1–​8. Papeo, L., Vallesi, A., Isaja, A., & Rumiati, R. I. (2009). Effects of TMS on different stages of motor and non-​motor verb processing in primary motor cortex. PLOS ONE, 4, e4508. Parish, P. (1963). Amelia Bedelia. New York: Harper & Row. Patterson, K., Graham, N., & Hodges, J. R. (1994). The impact of semantic memory loss on phonological representations. Journal of Cognitive Neuroscience, 6, 57–​69. Patterson, K., Nestor, P. J., & Rogers, T. T. (2007). Where do you know what you know? The representation of semantic knowledge in the human brain. Nature Reviews Neuroscience, 8, 976–​987. Pecher, D. (2018). Curb your embodiment. Topics in Cognitive Science, 10(3), 501–​517. Pecher, D., Zeelenberg, R., & Barsalou, L. W. (2003). Verifying properties from different modalities for concepts produces switching costs. Psychological Science, 14, 119–​124. Penfield, W., & Boldrey, E. (1937). Somatic motor and sensory representation in the cerebral cortex of man as studied by electrical stimulation. Brain, 60, 389–​443. Penfield, W., & Rasmussen, T. (1950). The cerebral cortex of man: A clinical study of localization of function. New York: Macmillan. Perniss, P., & Vigliocco, G. (2014). The bridge of iconicity: From a world of experience to the experience of language. Philosophical Transactions of the Royal Society B, 369, 20130300, 1–​13. Perry, L. K., Perlman, M., & Lupyan, G. (2015). Iconicity in English and Spanish and its relation to lexical category and age of acquisition. PLOS ONE, 10(9), e0137147, 1–​17. Perry, L. K., Perlman, M., Winter, B., Massaro, D. W., & Lupyan, G. (2018). Iconicity in the speech of children and adults. Developmental Science, 21(3), e12572. Peterson, C. C. (2009). Development of social-​cognitive and communicative skills in children born deaf. Scandinavian Journal of Psychology, 50(5), 475–​483. Peterson, C. C., & Siegal, M. (1995). Deafness, conversation, and theory of mind. Journal of Child Psychology and Psychiatry, 36(3), 459–​474. Peterson, C. C., & Siegal, M. (1999). Representing inner worlds: Theory of mind in autistic, deaf, and normal hearing children. Psychological Science, 10, 126–​129. Petro, L. S., Paton, A. T., & Muckli, L. (2017). Contextual modulation of visual cortex by auditory signals. Philosophical Transactions of the Royal Society B, 372, Article 1714, 1–​9. Pexman, P. M. (2012). Meaning-​level influences on visual-​word recognition. In J. S. Adelman (Ed.), Visual-​word recognition: Vol. 2. Meaning and context, individuals and development (pp. 24–​43). Hove, UK: Psychology Press. Pexman, P. M. (2019). The role of embodiment in conceptual development. Language, Cognition and Neuroscience, 34(10), 1274–​1283.

242 References Pexman, P. M., Hargreaves, I. S., Edwards, J. D., Henry, L. C., & Goodyear, B. G. (2007). Neural correlates of concreteness in semantic categorization. Journal of Cognitive Neuroscience, 19, 1407–​1419. Pexman, P. M., Hargreaves, I. S., Siakaluk, P. D., Bodner, G. E., & Pope, J. (2008). There are many ways to be rich: Effects of three measures of semantic richness on visual word recognition. Psychonomic Bulletin & Review, 15, 161–​167. Pezzulo, G., & Castelfranchi, C. (2007). The symbol detachment problem. Cognitive Process, 8, 115–​131. Piazza, M., & Dehaene, S. (2004). From number neurons to mental arithmetic: The cognitive neuroscience of number sense. In M. Gazzaniga (Ed.), The cognitive neurosciences (3rd ed., pp. 865–​875). Cambridge, MA: MIT Press. Pickering, M. J., & Clark, A. (2014). Getting ahead: Forward models and their place in cognitive architecture. Trends in Cognitive Sciences, 18(9), 451–​456. Pickering, M. J., Garrod, S. (2021). Understanding dialogue: Language use and social interaction. Cambridge: Cambridge University Press. Pinker, S. (1994). The language instinct. New York: Harper Collins. Pinker, S. (1997). How the mind works. New York: W. W. Norton & Co. Plaut, D. C., & Shallice, T. (1993) . Deep dyslexia: A case study in connectionist neuropsychology. Cognitive Neuropsychology, 10, 377–​500. Pobric, G., Jefferies, E., & Lambon Ralph, M. A. (2007) . Anterior temporal lobes mediate semantic representation: Mimicking semantic dementia by using rTMS in normal participants. Proceedings of the National Academy of Sciences, 104, 20137–​20141. Pobric, G., Lambon Ralph, M. A., & Jefferies, E. (2009) . The role of the anterior temporal lobes in the comprehension of concrete and abstract words: rTMS evidence. Cortex, 45(9), 1104–​1110. Pobric, G., Jefferies, E., & Lambon Ralph, M. A. (2010). Amodal semantic representations depend on both left and right anterior temporal lobes: New rTMS evidence. Neuropsychologia, 48, 1336–​1342. Ponari, M., Norbury, C. F., & Vigliocco, G. (2018). Acquisition of abstract concepts is influenced by emotional valence. Developmental Science, 21(2), 10.1111/​desc.12549. Ponari, M., Norbury, C. F., & Vigliocco, G. (2020). The role of emotional valence in learning novel abstract concepts. Developmental Psychology, 56(10), 1855–​1865. Postle, N., McMahon, K. L., Ashton, R., Meredith, M., & de Zubicaray, G. I. (2008). Action word meaning representations in cytoarchitectonically defined primary and premotor cortices. NeuroImage, 43, 634–​644. Prinz, J. J. (2002). Furnishing the mind: Concepts and their perceptual basis. Cambridge, MA: MIT Press. Prinz, J. J. (2006). Putting the brakes on enactive perception. Psyche, 12, 1–​19. Pulvermüller, F. (2001). Brain reflections of words and their meaning. Trends in Cognitive Sciences, 5, 517–​524. Pulvermüller, F. (2013). How neurons make meaning: Brain mechanisms for embodied and abstract-​symbolic semantics. Trends in Cognitive Sciences, 17(9), 458–​470. Pulvermüller, F. (2017). Neural reuse of action perception circuits for language, concepts, and communication. Progress in Neurobiology, 160, 1–​44. Pulvermüller, F., & Garagnani, M. (2014). From sensorimotor learning to memory cells in prefrontal and temporal association cortex: A neurocomputational study of disembodiment. Cortex, 57, 1–​21. Pulvermüller, F., Hauk, O., Nikulin, V. V., & Ilmoniemi, R. J. (2005). Functional links between motor and language systems. European Journal of Neuroscience, 21(3), 793–​797.

References  243 Pyers, J. E., & Senghas, A. (2009). Language promotes false-​belief understanding: Evidence from learners of a new sign language. Psychological Science, 20(7), 805–​812. Pylyshyn, Z. W. (1984). Computation and cognition: Toward a foundation for cognitive science. Cambridge, MA: MIT Press. Quine, W. V. (1960). Word and object. Cambridge, MA: MIT Press. Ramsey, W. M. (2007). Representation reconsidered. Cambridge: Cambridge University Press. Raos, V., Evangeliou, M. N., & Savaki, H. E. (2004). Observation of action: Grasping with the mind’s hand. Neuroimage, 23, 193–​201. Raos, V., Evangeliou, M. N., & Savaki, H. E. (2007). Mental simulation of action in the service of action perception. Journal of Neuroscience, 27, 12675–​12683. Raposo, A., Moss, H. E., Stamatakis, E. A., & Tyler, L. K. (2009). Modulation of motor and premotor cortices by actions, action words and action sentences. Neuropsychologia, 47, 388–​396. Ratcliffe, M. (2012). What is touch? Australasian Journal of Philosophy, 90(3), 413–​432. Recanati, F. (2004). Literal meaning. Cambridge: Cambridge University Press. Recchia, G., & Jones, M. N. (2012). The semantic richness of abstract concepts. Frontiers in Human Neuroscience, 6, Article 315, 1–​16. Reggin, L. D., Muraki, E. J., & Pexman, P. M. (2021). Development of abstract word knowledge. Frontiers in Psychology, 12, 686478, 1–​11. Reilly, J., Harnish, S., Garcia, A., Hung, J., Rodriguez, A. D., & Crosson, B. (2014). Lesion symptom mapping in nonfluent aphasia: Can a brain be both embodied and disembodied? Cognitive Neuropsychology, 31, 287–​312. Reilly, J., & Peelle, J. E. (2008). Effects of semantic impairment on language processing in semantic dementia. Seminars in Speech and Language, 29(1), 32–​43. Reilly, J., Peelle, J. E., Garcia, A., & Crutch, S. J. (2016). Linking somatic and symbolic rerpresentation in semantic memory: The dynamic multilevel representation f ramework. Psychonomic Bulletin & Review, 23, 1002–​1014. Repetto, C., Colombo, B., Cipresso, P., & Riva, G. (2013). The effects of rTMS over the primary motor cortex: The link between action and language. Neuropsychologia, 51, 8–​13. Ricciardi, E., Bonino, D., Pellegrini, S., & Pietrini, P. (2014). Mind the blind brain to understand the sighted one! Is there a supramodal cortical functional architecture? Neuroscience and Biobehavioral Reviews, 41, 64–​77. Rice, G. E., Hoffman, P., Binney, R. J., & Lambon Ralph, M. A. (2018). Concrete versus abstract forms of social concept: An fMRI comparison of knowledge about people versus social terms. Philosophical Transactions of the Royal Society B, 373, 20170136, 1–​16. Rice, G. E., Lambon Ralph, M. A., & Hoffman, P. (2015). The roles of left versus right anterior temporal lobes in conceptual knowledge: An ALE meta-​analysis of 97 functional neuroimaging studies. Cerebral Cortex, 25(11), 4374–​4391. Rice, M. L. (2013). Language growth and genetics of developmental language disorder. International Journal of Speech-​Language Pathology, 15(3), 223–​233. Richardson, D. C., Spivey, M. J., Barsalou, L. W., & McRae, K. (2003). Spatial representations activated during real-​time comprehension of verbs. Cognitive Science, 27, 767–​780. Rinaldi, L., Vecchi, T., Fantino, M., Merabet, L. B., & Cattaneo, Z. (2018). The ego-​moving metaphor of time relies on visual experience: No representation of time along the sagittal space in the blind. Journal of Experimental Psychology: General, 147(3), 444–​450. Riordan, B., & Jones, M. N. (2010). Redundancy in perceptual and linguistic experience: Comparing feature-​based and distributional models of semantic information. Topics in Cognitive Science, 3, 303–​345.

244 References Rivlin, R., & Gravelle, K. (1984). Deciphering the senses: Expanding the world of human perception. New York: Simon & Schuster. Rizzolatti, G., Fadiga, L., Gallese, V., & Fogassi, L. (1996). Premotor cortex and the recognition of motor actions. Cognitive Brain Research, 3, 131–​141. Rizzolatti, G., & Fogassi, L. (2014). The mirror mechanism: Recent findings and perspectives. Philosophical Transactions of the Royal Society of London B, 369(1644), Article 20130420, 1–​12. Roberson, D., Davidoff, J., Braisby, N. (1999). Similarity and categorization: Neuropsychological evidence for a dissociation in in explicit categorization tasks. Cognition, 71(1), 1–​42. Roberts, A., Nguyen, P., Orange, J. B., Jog, M., Nisbet, K. A., & McRae, K. (2017). Differential impairments of upper and lower limb movements influence action verb processing in Parkinson’s disease. Cortex, 97, 49–​59. Rogers, T. T., Hocking, J., Noppeney, U., Mechelli, A., Gorno-​Tempini, M. L., Patterson, K., & Price, C. J. (2006). Anterior temporal cortex and semantic memory: Reconciling findings from neuropsychology and functional imaging. Cognitive, Affective, & Behavioral Neuroscience, 6(3), 201–​213. Rolls, E. T. (2012). Invariant visual object and face recognition: Neural and computational bases, and a model, VisNet. Frontiers in Computational Neuroscience, 6, 35. Rosch, E. (1978). Principles of categorization. In B. B. Lloyd & E. Rosch (Eds.), Cognition and categorization (pp. 28–​49). Hillsdale, NJ: Erlbaum. Rosch, E., & Mervis, C. B. (1975). Family resemblance: Studies in the internal structure of categories. Cognitive Psychology, 7, 573–​605. Rosch, E., Mervis, C. B., Gray, W., Johnson, D., & Boyes-​Braem, P. (1976). Basic objects in natural categories. Cognitive Psychology, 8, 382–​439. Rosselli, M., & Ardila, A. (1989). Calculation deficits in patients with right and left hemisphere damage. Neuropsychologia, 27(5), 607–​617. Rossor, M. N., Warrington, E. K., & Cipolotti, L. (1995). The isolation of calculation skills. Journal of Neurology, 242, 78–​81. Roversi, C., Borghi, A. M., & Tummolini, L. (2013). A marriage is an artefact and not a walk that we take together: An experimental study on the categorization of artefacts. Review of Philosophy and Psychology, 4(3), 527–​542. Rueschemeyer, S.-​A., Lindemann, O., van Rooij, D., van Dam, W., & Bekkering, H. (2010). Effects of intentional motor actions on embodied language processing. Experimental Psychology, 57(4), 260–​266. Ruffman, T., Slade, L., & Crowe, E. (2002). The relation between children’s and mothers’ mental-​state language and theory-​of-​mind understanding. Child Development, 73(3), 734–​751. Rundblad, G., & Annaz, D. (2010a). Development of metaphor and metonymy comprehension: Receptive vocabulary and conceptual knowledge. British Journal of Developmental Psychology, 28(Pt 3), 547–​563. Rundblad, G., & Annaz, D. (2010b). The atypical development of metaphor and metonymy comprehension in children with autism. Autism, 14(1), 29–​46. Rupert, R. (2010). Cognitive systems and the extended mind. New York: Oxford University Press. Rupert, R. (2011). Embodiment, consciousness, and the massively representational mind. Philosophical Topics, 39(1),99–​120.

References  245 Rust, N. C., & DiCarlo, J. J. (2010). Selectivity and tolerance (“invariance”) both increase as visual information propagates from cortical area V4 to IT. Journal of Neuroscience, 30(39), 12978–​12995. Sabsevitz, D., Medler, D., Seidenberg, M., & Binder, J. (2005). Modulation of the semantic system by word imageability. Neuroimage, 27(1), 188–​200. Sadato, N., Pascual-​Leone, A., Grafman, J., Ibañez, V., Deiber, M-​P., Dold, G., & Hallett, M. (1996). Activation of the primary visual cortex by Braille reading in blind subjects. Nature, 380(6574), 526–​528. Sakreida, K., Scorolli, C., Menz, M. M., Heim, S., Borghi, A. M., & Binkofski, F. (2013). Are abstract action words embodied? An fMRI investigation at the interface between language and motor cognition. Frontiers in Human Neuroscience, 7(125), 1–​13. Salmazo-​Silva, H., Parente, M., Rocha, M., Baradel, R., Cravo, A., Sato, J. R., Godinho, F., & Carthery-​Goulart, M. T. (2017). Lexical retrieval and semantic memory in Parkinson’s disease: The question of noun and verb dissociation. Brain and Language, 165, 10–​20. Santiago, J., Lupáñez, J., Pérez, E., & Funes, M. J. (2007). Time (also) flies from left to right. Psychonomic Bulletin & Review, 14(3), 512–​516. Santiago, J., Ouellet, M., Román, A. and Valenzuela, J. (2012). Attentional factors in conceptual congruency. Cognitive Science, 36, 1051–​1077. Sathian, K., & Stilla, R. (2010). Cross-​modal plasticity of tactile perception in blindness. Restorative Neurology and Neuroscience, 28(2), 271–​281. Saygin, A. P., McCullough, S., Alac, M., & Emmorey, K. (2010). Modulation of BOLD response in motion-​sensitive lateral temporal cortex by real and fictive motion sentences. Journal of Cognitive Neuroscience, 22, 2480–​2490. Schaller, S. (2012). A man without words (2nd ed.). Berkeley, CA: University of California Press. Schick, B., de Villiers, P., de Villiers, J., & Hoffmeister, R. (2007). Language and theory of mind: A study of deaf children. Child Development, 78(2), 376–​396. Schinazi, V. R., Thrash, T., & Chebat, D. R. (2016). Spatial navigation by congenitally blind individuals. WIREs: Cognitive Science, 7(1), 37–​58. Schneider, I. K., Rutjens, B. T., Jostmann, N. B., & Lakens, D. (2011). Weighty matters: Importance literally feels heavy. Social Psychological and Personality Science, 2(5), 474–​478. Schwanenflugel, P. J., Harnishfeger, K. K., & Stowe, R. W. (1988). Context availability and lexical decisions for abstract and concrete words. Journal of Memory and Language, 27, 499–​520. Schwanenflugel, P. J., & Shoben, E. J. (1983). Differential context effects in the comprehension of abstract and concrete verbal materials. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 82–​102. Scorolli, C., Binkofski, F., Buccino, G., Nicoletti, R., Riggio, L., & Borghi, A. (2011). Abstract and concrete sentences, embodiment and languages. Frontiers in Psychology, 2, 227. Scorolli, C., & Borghi, A. M. (2007). Sentence comprehension and action: Effector specific modulation of the motor system. Brain Research, 1130, 119–​124. Scorolli, C., Borghi, A. M., & Glenberg, A. (2009). Language-​induced motor activity in bi-​manual object lifting. Experimental Brain Research, 193, 43–​43. Scorolli, C., Jacquet, P. O., Binkofski, F., Nicoletti, R., Tessari, A., & Borghi, A. (2012). Abstract and concrete phrases processing differently modulates cortico-​spinal excitability. Brain Research, 1488, 60–​71.

246 References Searle, J. (1969). Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press. Seghier, M. L. (2013). The angular gyrus: Multiple functions and multiple subdivisions. Neuroscientist, 19(1), 43–​61. Semenza, C., & Benavides-​Varela, S. (2017). Reassessing lateralization in calculation. Philosophical Transactions of the Royal Society B, 373, 20170044. http://​dx.doi.org/​ 10.1098/​rstb.2017.0044 Semenza, C., Bisiacchi, P. S., & Romani, L. (1992). Naming disorders and semantic representations. Journal of Psycholinguistic Research, 21(5), 349–​364. Semino, E. (2008). A corpus-​based study of metaphors for speech activity in British English. In S. Gries & A. Stefanowitsch (Eds.), Corpus-​based approaches to metaphor and metonymy (pp. 36–​62). New York: De Gruyter Mouton. Shallice, T., & Cooper, R. (2013). Is there a semantic system for abstract words? Frontiers in Human Neuroscience, 7, Article 175, 1–​10. Shallice, T., & Warrington, E. K. (1975). Word recognition in a phonemic dyslexic patient. Quarterly Journal of Experimental Psychology, 27, 187–​199. Shams, L., Kamitani, Y., & Shimojo, S. (2000). What you see is what you hear. Nature, 408, 788. Shapiro, L. (2019). Embodied cognition (2nd ed.). New York: Routledge. Shayan, S., Ozturk, O., & Sicoli, M. (2011). The thickness of pitch: Cross-​modal metaphors in Farsi, Turkish and Zapotec. Senses & Society, 6, 19–​29. Shea, N. (2018). Metacognition and abstract concepts. Philosophical Transactions of the Royal Society B, 373, 20170133, 1–​7. Shen, Y., & Porat, R. (2017). Metaphorical directionality: The role of language. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 62–​81). Cambridge: Cambridge University Press. Shepard, R. N., & Cooper, L. A. (1992). Representation of colors in the blind, color-​blind, and normally sighted. Psychological Science, 3(2), 97–​104. Shields, C. (2020). Aristotle. The Stanford Encyclopedia of Philosophy. https://​plato.stanf​ ord.edu/​archi​ves/​fall2​020/​entr​ies/​aristo​tle/​. Shimojo, S., & Shams, L. (2001). Sensory modalities are not separate modalities: plasticity and interactions. Current Opinion in Neurobiology, 11, 505–​509. Shulman, C., & Guberman, A. (2007). Acquisition of verb meaning through syntactic cues: A comparison of children with autism, children with specific language impairment (SLI) and children with typical language development (TLD). Journal of Child Language, 34(2), 411–​423. Siakaluk, P. D., Knol, N., & Pexman, P. M. (2014). Effects of emotional experience for abstract words in the Stroop task. Cognitive Science, 38, 1698–​1717. Siakaluk, P. D., Pexman, P. M., Sears, C. R., Wilson, K., Locheed, K., & Owen, W. J. (2008). The benefits of sensorimotor knowledge: Body-​object interaction facilitates semantic processing. Cognitive Science, 32, 591–​605. Sidhu, D. M., Heard, A., & Pexman, P. M. (2016). Is more always better for verbs? Semantic richness effects and verb meaning. Frontiers in Psychology, 7, Article 798, 1–​14. Sidhu, D. M., Kwan, R., Pexman, P. M., & Siakaluk, P. D. (2014). Effects of relative embodiment in lexical and semantic processing of verbs. Acta Psychologica, 149, 32–​39. Sidhu, D. M., Vigliocco, G., & Pexman, P. M. (2020). Effects of iconicity in lexical decision. Language and Cognition, 12(1), 164–​181. Sifneos, P. E. (1996) . Alexithymia: Past and present. American Journal of Psychiatry, 153(7 Suppl), 137–​142.

References  247 Silberman, S. (2015). Neurotribes: The Legacy of Autism and the Future of Neurodiversity. New York: Avery. Siltanen, S. A. (1986). “Butterflies are rainbows?”: A developmental investigation of metaphor comprehension. Communication Eductation, 1, 1–​12. Simmons, W. K., & Barsalou, L. W. (2003). The similarity-​in-​topography principle: Reconciling theories of conceptual deficits. Cognitive Neuropsychology, 20, 451–​486. Simmons, W. K., Ramjee, V., Beauchamp, M. S., McRae, K., Martin, A., & Barsalou, L. (2007). A common neural substrate for perceiving and knowing about color. Neuropsychologia, 45(12), 2802–​2810. Sirigu, A., Duhamel, J. R., Poncet, M. (1991). The role of sensorimotor experience in object recognition: A case of multimodal agnosia. Brain, 114, 2555–​2573. Skipper, L. M., & Olson, I. R. (2014). Semantic memory: Distinct neural representations for abstractness and valence. Brain and Language, 130, 1–​10. Slepian, M. L., & Ambady, N. (2014). Simulating sensorimotor metaphors: Novel metaphors influence sensory judgments. Cognition, 130(3), 309–​314. Slobin, D. I. (1996). From thought and language to thinking for speaking. In J. Gumperz & S. Levinson (Eds.), Rethinking linguistic relativity (pp. 70–​96). Cambridge: Cambridge University Press. Smith, E. E., & Medin, D. L. (1999). The exemplar view. In S. Laurence & E. Margolis (Eds.), Concepts: Core readings (pp. 207–​222). Cambridge, MA: MIT Press. Snedeker, J., & Gleitman, L. R. (2004). Why it is hard to label our concepts. In D. G. Hall & S. R. Waxman (Eds.), Weaving a lexicon (pp. 257–​293). Cambridge, MA: MIT Press. Speed, L., & Majid, A. (2020). Grounding language in the neglected senses of touch, taste, and smell. Cognitive Neuropsychology, 37(5–​6), 363–​392. Spence, C., & Driver, J. (1998). Auditory and audiovisual inhibition of return. Perception & Psychophysics, 60, 125–​139. Spence, C., Nicholls, M. E. R., & Driver, J. (2000). The cost of expecting events in the wrong sensory modality. Perception & Psychophysics, 63, 330–​336. Sperber, D., & Wilson, D. (1986). Relevance: Communication and cognition. Oxford, UK: Blackwell. Spitsyna, G., Warren, J. E., Scott, S. K., Turkheimer, F. E., & Wise, R. J. (2006). Converging language streams in the human temporal lobe. Journal of Neuroscience, 26(28), 7328–​7336. Sporns, O., Honey, C. J., & Kotter, R. (2007) . Identification and classification of hubs in brain networks. PLOS ONE, 2, e1049. Spunt, R. P., Kemmerer, D., & Adolphs, R. (2016). The neural basis of conceptualizing the same action at different levels of abstraction. Social Cognitive and Affective Neuroscience, 11(7), 1141–​1151. doi:10.1093/​scan/​nsv084 Stanfield, R. A., & Zwaan, R. A. (2001). The effect of implied orientation derived from verbal context on picture recognition. Psychological Science, 12(2), 153–​156. Stapleton, M. (2013). Steps to a properly embodied cognitive science. Cognitive Systems Research, 22, 1–​11. Steen, C. J. (2017). Attention to metaphor: Where embodied cognition and social interaction can meet, but may not often do so. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 279–​296). Cambridge, UK: Cambridge University Press. Steen, G. J. (2008). The paradox of metaphor: Why we need a three-​dimensional model of metaphor. Metaphor and Symbol, 23(4), 213–​241. Sterelny, K. (2010) . Minds: Extended or scaffolded? Phenomenology and the Cognitive Sciences, 9, 465–​481.

248 References Stern, J. (2000). Metaphor in context. Cambridge, MA: MIT Press. Steyvers, M. (2010). Combining feature norms and text data with topic models. Acta Psychologica, 133(3), 234–​243. Stotz, K. (2010). Human nature and cognitive-​ developmental niche construction. Phenomenology and the Cognitive Sciences, 9(4), 483–​501. Stramandinoli, F., Marocco, D., & Cangelosi, A. (2012). The grounding of higher order concepts in action and language: A cognitive robotics model. Neural Networks, 32, 165–​173. Striem-​ Amit, E., Cohen, L., Dehaene, S., & Amedi, A. (2012). Reading with sounds: Sensory substitution selectively activates the visual word form area in the blind. Neuron, 76, 640–​652. Striem-​Amit, E., Wang, X., Bi, Y., & Caramazza, A. (2018). Neural representation of visual concepts in people born blind. Nature Communications, 9, 5250. Stringaris, A., Medford, N., Giampietro, V., Brammer, M., & David, A. (2007). Deriving meaning: Distinct neural mechanisms for metaphoric, literal, and non-​meaningful sentences. Brain and Language, 100(2), 150–​162. Strnad L., Peelen M. V., Bedny M., & Caramazza A. (2013). Multivoxel pattern analysis reveals auditory motion information in MT+​of both congenitally blind and sighted individuals. PLOS ONE, 8(4), e63198. Sutton, J. (2010). Exograms and interdisciplinarity: History, the extended mind, and the civilizing process. In R. Menary (Ed.), The extended mind (pp. 189–​226). Cambridge, MA: MIT Press. Tager-​Flusberg, H. (1992). Autistic children’s talk about psychological states: Deficits in the early acquisition of theory of mind. Child Development, 63(1), 161–​172. Tager-​Flusberg, H., & Joseph, R. (2005). How language facilitates the acquisition of false belief in children with autism. In J. W. Astington & J. Baird (Eds.), Why language matters for theory of mind (pp. 298–​318). New York: Oxford University Press. Taikh, A., Hargreaves, I. S., Yap, M., & Pexman, P. M. (2015). Semantic classification of pictures and words. Quarterly Journal of Experimental Psychology, 68, 1502–​1518. Talmy, L. (1983). How languages structure space. In L. P. Acredolo & H. L. Pick (Eds.), Spatial orientation: Theory, research, and application (pp. 225–​ 282). New York: Plenum Press. Tanaka, K. (2003). Columns for complex visual object features in the inferotemporal cortex: Clustering of cells with similar but slightly different stimulus selectivities. Cerebral Cortex, 13, 90–​99. Taub, S. F. (2001). Language from the body: Iconicity and metaphor in american sign language. Cambridge: Cambridge University Press. Taylor, L. J., Evans, C., Greer, J., Senior, C., Coventry, K. R., & Ietswaart, M. (2017). Dissociation between semantic representations for motion and action verbs: Evidence from patients with left hemisphere lesions. Frontiers in Human Neuroscience, 11, Article 35. Tettamanti, M., Buccino, G., Saccuman, M., Gallese, V., Danna, M., Scifo, P., Fazio, F., Rizzolatti, G., Cappa, S. F., & Perani, D. (2005). Listening to action-​related sentences activates fronto-​ parietal motor circuits. Journal of Cognitive Neuroscience, 17, 273–​281. Thill, S., Pado, S., & Ziemke, T. (2014). On the importance of rich grounding of concepts: Perspectives from embodied cognitive science and computational linguistics. Topics in Cognitive Science, 6, 545–​558.

References  249 Thompson, E. (2007). Mind in life: Biology, phenomenology, and the sciences of the mind. Cambridge, MA: Harvard University Press. Thompson, E., & Stapleton, M. (2009). Making sense of sense-​making: Reflections on enactive and extended mind. Topoi, 28(1), 23–​30. Thompson, R. K. R., & Oden, D. L. (2000). Categorical perception and conceptual judgments by nonhuman primates: The paleological monkey and the analogical ape. Cognitive Science, 24, 363–​396. Thompson, R. K. R., Oden, D. L., & Boysen, S. T. (1997). Language-​naive chimpanzees (pan troglodytes) judge relations between relations in a conceptual matching-​to-​ sample task. Journal of Experimental Psychology: Animal Behavior Processes, 23, 31–​43. Thompson, R. L., Vinson, D. P., Woll, B., & Vigliocco, G. (2012). The road to language learning is iconic: Evidence from British sign language. Psychological Science, 23(12), 1443–​1448. Thompson-​Schill, S. L., Aguirre, G. K., D’Esposito, M., & Farah, M. J. (1999). A neural basis for category and modality specificity in semantic knowledge. Neuropsychologia, 37, 671–​676. Tillas, A. (2015). Langauge as grist to the mill of cognition. Cognitive Processes, 16(3), 219–​243. Tillas, A. (2017). On the origins of endogenous thoughts. Cognitive Processing, 18(2), 107–​117. Tkach, D., Reimer, J., & Hatsopoulos, N. G. (2007). Congruent activity during action and action observation in the motor cortex. Journal of Neuroscience, 27, 13241–​13250. Toglia, M. P., & Battig, W. F. (1978). Handbook of semantic word norms. New York: Erlbaum. Tomasello, M. (2008). The origins of human communication. Cambridge, MA: MIT Press. Tranel, D., Kemmerer, D., Adolphs, R., Damasio, H., & Damasio, A. N. (2003). Neuralcorrelates of conceptual knowledge of actions. Cognitive Neuropsychology, 20, 409–​432. Tranel, D., Manzel, K., Asp, E., & Kemmerer, D. (2008). Naming static and dynamic actions: Neuropsychological evidence. Journal of Physiology, Paris, 102, 80–​94. Troche, J., Crutch, S., & Reilly, J. (2014). Clustering, hierarchical organization, and the topography of abstract and concrete nouns. Frontiers in Psychology, 5(360), 1–​10. Trumpp, N. M., Kliese, D., Hoenig, K., Haarmeier, T., & Kiefer, M. (2013). Losing the sound of concepts: Damage to auditory association cortex impairs the processing of sound-​related concepts. Cortex, 49, 474–​486. Tschentscher, N., Hauk, O., Fischer, M. H., & Pulvermüller, F. (2012). You can count on the motor cortex: fMRI reveals embodied number processing. NeuroImage, 59, 3139–​3148. Tversky, B. (2019). Mind in motion: How action shapes thought. New York: Basic Books. Tversky, B., & Hemenway, K. (1984). Objects, parts, and categories. Journal of Experimental Psychology: General, 113, 169–​193. Umiltà, M. A., Escola, L., Intskirveli, I., Grammont, F., Rochat, M., Caruana, F., Jezzini, A., Gallese, V., & Rizzolati, G. (2008). When pliers become fingers in the monkey motor system. PNAS, 105(6), 2209–​2213. Umiltà, M. A., Kohler, E., Gallese, V., Fogassi, L., Fadiga, L., Keysers, C., & Rizzolatti, G. (2001). I know what you are doing: A neuropsychological case study. Neuron, 32(1), 155–​165. Urgesi, C., Candidi, M., & Avenanti, A. (2014). Neuroanatomical substrates of action perception and understanding: An anatomical likelihood estimation meta-​analysis

250 References of lesion-​symptom mapping studies in brain-​injured patients. Frontiers in Human Neuroscience, 8, Article 344, 1–​17. van Ackeren, M. J., Schneider, T. R., Musch, K., & Rueschemeyer, S.-​A. (2014). Oscillatory neuronal activity reflects lexical-​semantic feature integration within and across sensory modalities in distributed cortical networks. Journal of Neuroscience, 34(43), 14318–​4323. van Dantzig, S., Pecher, D., Zeelenberg, R., & Barsalou, L. W. (2008). Perceptual processing affects conceptual processing. Cognitive Science, 32(3), 579–​590. van den Heuvel, M. P., & Sporns, O. (2013). Network hubs in the human brain. Trends in Cognitive Science, 17, 683–​696. van Kemenade, B. M., Seymour, K., Wacker, E., Spitzer, B., Blankenburg, F., & Sterzer, P. (2014). Tactile and visual motion direction processing in hMT+​/​V5. NeuroImage, 84, 420–​427. Vandenberghe, R., Nobre, A. C., & Price, C. J. (2002) . The response of left temporal cortex to sentences. Journal of Cognitive Neuroscience, 14, 550–​560. Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. Cambridge, MA: MIT Press. Vargas, R., & Just, M. A. (2020). Neural representations of abstract concepts: Identifying underlying neurosemantic dimensions. Cerebral Cortex, 30(4), 2157–​2166. Varley, R. A., Klessinger, N. J., Romanowski, C. A., & Siegal, M. (2005). Agrammatic but numerate. PNAS, 102(9), 3519–​3524. Vigliocco, G., Kousta, S. -​T., Della Rosa, P. A., Vinson, D. P., Tettamanti, M., Devlin, J. T., & Cappa, S. F. (2014). The neural representation of abstract words: The role of emotion. Cerebral Cortex, 24, 1767–​1777. Vigliocco, G., Ponari, M., & Norbury, C. (2017). The role of linguistic information in learning abstract words: Evidence from children with specific language impairment (SLI). Proceedings of the 39th Annual Meeting of the Cognitive Science Society, 3441–​ 3446. Vigliocco, G., Ponari, M., & Norbury, C. F. (2018). Learning and processing abstract words and concepts: Insights from typical and atypical development. Topics in Cognitive Science, 10(3), 533–​549. Vigliocco, G., Vinson, D. P., Druks, J., Barber, H., & Cappa, S. F. (2011) . Nouns and verbs in the brain: A review of behavioural, electrophys-​iological, neurophysiological and imaging studies. Neuroscience & Biobehavioral Reviews, 35, 407–​426. Vigliocco, G., Vinson, D. P., Lewis, W., & Garrett, M. F. (2004). Representing meanings of object and action words: The featural and unitary semantic space hypothesis. Cognitive Psychology, 48, 422–​488. Villani, C., Lugli, L., Liuzza, M. T., & Borghi, A. M. (2019). Varieties of abstract concepts and their multiple dimensions. Language and Cognition, 11(3), 403–​430. Vinson, D., Thompson, R. L., Skinner, R., & Vigliocco, G. (2015). A faster path between meaning and form? Iconicity facilitates sign recognition and production in British Sign Language. Journal of Memory and Langauge, 82, 56–​85. Visser, M., Jefferies, E., Embleton, K. V., & Lambon Ralph, M. A. (2012). Both the middle temporal gyrus and the ventral temporal area are crucial for multimodal semantic processing: Distortion-​corrected fMRI evidence for a double gradient of information convergence in the temporal lobes. Journal of Cognitive Neuroscience, 24, 1766–​1778. Visser, M., Jefferies, E., & Lambon Ralph, M. (2010). Semantic processing in the anterior temporal lobes: A meta-​analysis of the functional neuroimaging literature. Journal of Cognitive Neuroscience, 22, 1083–​1094.

References  251 Vosniadou, S. (1987). Children and metaphors. Child Development, 58(3), 870–​885. Vosniadou, S., Ortony, A., Reynolds, R. E., & Wilson, P. T. (1984). Sources of Difficulty in the Young Child’s Understanding of Metaphorical Language. Child Development, 55(4), 1588–​1606. Vukovic, N., Feurra, M., Shpektor, A., Myachykov, A., & Shtyrov, Y. (2017). Primary motor cortex functionally contributes to language comprehension: An online rTMS study. Neuropsychologia, 96, 222–​229. Vulchanova, M., Saldaña, D., Chahboun, S., & Vulchanov, V. (2015). Figurative language processing in atypical populations: The ASD perspective. Frontiers in Human Neuroscience, 9, Article 24, 1–​11. Vygotsky, L. (2012). Thought and language revised and expanded. Cambridge, MA: MIT Press. Waggoner, J. E., & Palermo, D. S. (1989). Betty is a bouncing bubble: Children’s comprehension of emotion-​descriptive metaphors. Developmental Psychology, 25(1), 152–​163. Wall, M. B., Lingnau, A., Ashida, H., & Smith, A. T. (2008). Selective visual responses to expansion and rotation in the human MT complex revealed by functional magnetic resonance imaging adaptation. European Journal of Neuroscience, 27(10), 2747–​2757. Wallentin, M., Lund, T. E., Ostergaard, S., Ostergaard, L., & Roepstorff, A. (2005). Motion verb sentences activate left posterior middle temporal cortex despite static context. Neuroreport, 16(6), 649–​652. Wallentin, M., Nielson, A. H., Vuust, P., Dohn, A., Roepstorff, A., & Lund, T. E. (2011). BOLD response to motion verbs in left posterior middle temporal gyrus during story comprehension. Brain and Language, 119, 221–​225. Wallentin, M., Ostergaard, S., Lund, T. E., Ostergaard, L., & Roepstorff, A. (2005). Concrete spatial language: See what I mean? Brain and Language, 92(3), 221–​233. Wang, J, Baucom, L. B., & Shinkareva, S. V. (2013). Decoding abstract and concrete concept representations based on single-​trial fMRI data. Human Brain Mapping, 34(5), 1133–​1147. Wang, J., Conder, J. A., Blitzer, D. N., & Shinkareva, S. V. (2010). Neural representation of abstract and concrete concepts: A meta-​analysis of imaging studies. Human Brain Mapping, 31(10), 1459–​1468. Wang, X., Men, W., Gao, J., Caramazza, A., & Bi, Y. (2020). Two forms of knowledge representation in the human brain. Neuron, 107(2), 383–​393.e5. Wang, X., Wang, B., & Bi, Y. (2019). Close yet independent: Dissociation of social from valence and abstract semantic dimensions in the left anterior temporal lobe. Human Brain Mapping, 40(16), 4759–​4776. Wang, X., Wu, W., Ling, Z., Xu, Y., Fang, Y., Wang, X., Binder, J. R., Men, W., Gao, J. H., & Bi, Y. (2018). Organizational principles of abstract words in the human brain. Cerebral Cortex, 28(12), 4305–​4318. Warriner, A. B., Kuperman, V., & Brysbaert, M. (2013). Norms of valence, arousal, and dominance for 13,915 English lemmas. Behavior Research Methods, 45(4), 1191–​1207. Warrington, E. K. (1975). The selective impairment of semantic memory. Quarterly Journal of Experimental Psychology, 27, 635–​637. Warrington, E. K., & Shallice, T. (1984). Category specific semantic impairments. Brain, 107, 829–​854. Watson, C. E., & Chatterjee, A. (2011). The functional neuroanatomy of actions. Neurology, 76, 1428–​1434.

252 References Watson, C. E., Cardillo, E. R., Ianni, G. R., & Chatterjee, A. (2013). Action concepts in the brain: An activation likelihood estimation meta-​analysis. Journal of Cognitive Neuroscience, 25, 1191–​1205. Wattenmaker, W., & Shoben, E. (1987). Context and the recallability of concrete and abstract sentences. Journal of Experimental Psychology: Learning, Memory and Cognition, 13(1), 140–​150. Wauters, L., Tellings, A., Van Bon, W., & Van Haaften, A. (2003). Mode of acquisition of word meanings: The viability of a theoretical construct. Applied Psycholinguistics, 24(3), 385–​406. doi:10.1017/​S0142716403000201 Waxman, S. R., & Markow, D. B. (1995). Words as invitations to form categories: Evidence from 12-​to 13-​month-​old infants. Cognitive Psychology, 29(3), 257–​302. Wearing, C. (2006). Metaphor and what is said. Mind & Language, 21(3), 310–​332. Weiskopf, D. (2007). Concept empiricism and the vehicles of thought. Journal of Consciousness Studies, 14, 156–​183. Weiskopf, D. (2010). Embodied cognition and linguistic comprehension. Studies in History and Philosophy of Science, 41, 294–​304. Wellman, H. M. (1990). The child’s theory of mind. Cambridge, MA: MIT Press. Wellman, H. M., Hollander, M., & Schult, C. A. (1996). Young children’s understanding of thought bubbles and thoughts. Child Development, 67(3), 768–​788. Wellman, H. M., & Peterson, C. C. (2013). Deafness, thought bubbles, and theory of mind development. Developmental Psychology, 49(12), 2357–​2367. Wellsby, M., Siakaluk, P. D., Owen, W. J., & Pexman, P. M. (2011). Embodied semantic processing: The body–​object interaction effect in a non-​manual task. Language and Cognition, 3, 1–​14. West-​ Eberhard, M. J. (2005). Developmental plasticity and the origin of species differences. PNAS, 102(Suppl 1), 6543–​6549. https://​doi.org/​10.1073/​pnas.0501844102 Whalen, J., Gallistel, C. R., & Gelman, R. (1999). Non-​verbal counting in humans: The psychophysics of number representation. Psychological Science, 10, 130–​137. Wheatley, T., Weisberg, J., Beauchamp, M. S., & Martin, A. (2005). Automatic priming of semantically related words reduces activity in the fusiform gyrus. Journal of Cognitive Neuroscience, 17, 1871–​1885. Wiemer-​Hastings, K., & Xu, X. (2005). Content differences for abstract and concrete concepts. Cognitive Science, 29(5), 719–​736. Wiig, E. H., & Secord, W. (1992). Test of language competence (Expanded edition). New York: The Psychological Corporation. Wilcox, P. (2000). Metaphor in American Sign Language. Washington, DC: Gallaudet University Press. Willems, R. M., & Casasanto, D. (2011). Flexibility in embodied understanding. Frontiers in Psychology, 2, Article 116, 1–​11. Willems, R. M., Hagoort, P., & Casasanto, D. (2010). Body-​specific representation of action words: Neural evidence from right-​and left-​handers. Psychological Science, 21, 67–​74. Williams, L. E., & Bargh, J. A. (2008). Experiencing physical warmth promotes interpersonal warmth. Science, 322, 606–​607. Wilson, A. D., & Golonka, S. (2013). Embodied cognition is not what you think it is. Frontiers in Psychology, 4, Article 58, 1–​13. Wilson, M. (2002). Six views of embodied cognition. Psychonomic Bulletin and Review, 9, 625–​636.

References  253 Wilson, M. (2006). Wandering significance: An essay on conceptual behavior. Oxford: Oxford University Press. Wilson-​Mendenhall, C. D., Simmons, W. K., Martin, A., & Barsalou, L. W. (2013). Contextual processing of abstract concepts reveals neural representations of non-​linguistic semantic content. Journal of Cognitive Neuroscience, 25, 920–​935. Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining Function of wrong beliefs in young children’s understanding of deception. Cognition, 13, 103–​128. Winkielman, P., Coulson, S., & Niedenthal, P. (2018). Dynamic grounding of emotion concepts. Philosophical Transactions of the Royal Society B, 373: 20170127, 1–​9. Winner, E. (1979). New names for old things: The emergence of metaphoric language. Journal of Child Language, 6(3), 469–​491. Winner, E., Rosenstiel, A. K., & Gardner, H. (1976). The development of metaphoric understanding. Developmental Psychology, 12(4), 289–​297. Winter, B., & Matlock, T. (2013). More is up . . . and right: Random number generation along two axes. In M. Knauff, M. Pauen, N. Sebanz, & I. Wachsmuth (Eds.), Proceedings of the 35th annual conference of the cognitive science society (pp. 3789–​3974). Austin, TX: Cognitive Science Society. Winter, B., & Matlock, T. (2017). Primary metaphors are both cultural and embodied. In B. Hampe (Ed.), Metaphor: Embodied cognition and discourse (pp. 99–​ 116). Cambridge: Cambridge University Press. Woolfe, T., Want, S. C., & Siegal, M. (2002). Signposts to development: Theory of mind in deaf children. Child Development, 73(3), 768–​778. Wurm, M. F., Ariani, G., Greenlee, M. W., & Lingnau, A. (2016). Decoding concrete and abstract representations during explicit and implicit conceptual processing. Cerebral Cortex, 26(8), 3390–​3401. Wurm, M. F., & Caramazza, A. (2019). Distinct roles of temporal and frontoparietal cortex in representing actions across vision and language. Nature Communications, 10, Article 289, 1–​10. Wurm, M. F., Caramazza, A., & Lingnau, A. (2017). Action categories in lateral occipitotemporal cortex are organized along sociality and transitivity. Journal of Neuroscience, 37(3), 562–​575. Wurm, M. F., & Lingnau, A. (2015). Decoding actions at different levels of abstraction. Journal of Neuroscience, 35(20), 7727–​7735. Xu, F. (2002). The role of language in acquiring object kind concepts in infancy. Cognition, 85(3), 223–​250. Xu, F. (2003). Numerosity discrimination in infants: Evidence for two systems of representations. Cognition, 89, B15–​B25. Xu, F., & Arriaga, R. I. (2007). Number discrimination in 10-​month-​old infants. British Journal of Developmental Psychology, 25, 103–​108. Xu, F., Spelke, E. S., & Goddard, S. (2005). Number sense in human infants. Developmental Science, 8, 88–​101. Yap, M. J., Pexman, P. M., Wellsby, M., Hargreaves, I. S., & Huff, M. (2012). An abundance of riches: Cross-​task comparison of semantic richness effects in visual word recognition. Frontiers in Human Neuroscience, 6(72), 1–​10. Yee, E. (2019). Abstraction and concepts: When, how, where, what and why? Language, Cognition and Neuroscience, 34(10), 1257–​1265.

254 References Yee, E., Ahmed, S., & Thompson-​Schill, S. L. (2012). Colorless green ideas can prime furiously. Psychological Science, 23(4), 364–​369. Yee, E., & Thompson-​Schill, S. L. (2016). Putting concepts into context. Psychonomic Bulletin & Review, 23(4), 1015–​1027. Yi, H. A., Moore, P., & Grossman, M. (2007). Reversal of the concreteness effect for verbs in semantic dementia. Neuropsychology, 21(1), 9–​19. York, C., Olm, C., Boller, A., McCluskey, L., Elman, J., Seltzer, E., Chahine, L., Woo, J., Rascovsky, K., McMillan C., & Grossman, M. (2014). Action verb comprehension in amyotrophic lateral sclerosis and Parkinson’s disease. Journal of Neurology, 261, 1073–​1079. Yuan, S., Perfors, A., Tenenbaum, J. B., & Xu, F. (2011). Learning individual words and learning about words simultaneously. Proceedings of the Annual Conference of the Cognitive Science Society, 33, 3280–​3285. Zanolie, K., van Dantzig, S., Boot, I., Wijnen, J., Schubert, T. W., Giessner, S. R., & Pecher, D. (2012). Mighty metaphors: Behavioral and ERP evidence that power shifts attention on a vertical dimension. Brain and Cognition, 78(1), 50–​58. Zdrazilova, L., & Pexman, P. M. (2013). Grasping the invisible: Semantic processing of abstract words. Psychonomic Bulletin & Review, 20, 1312–​1318. Zhong, C-​B., & Leonardelli, G. J. (2008). Cold and lonely: Does social exclusion literally feel cold? Psychological Science, 19(9), 838–​842. Zhong, C-​B., Liljenquist, K. (2006). Washing away your sins: Threatened morality and physical cleansing. Science, 313, 1451–​1452. Zihl, J., & von Cramon, D. Y. (1983). Selective disturbance of movement vision after bilateral brain damage. Brain, 106(Pt 2), 313–​340. Zihl, J., von Cramon, D., Mai, N., & Schmid, C. (1991). Disturbance of movement vision after bilateral posterior brain damage: Further evidence and follow up observations. Brain, 114(Pt 5), 2235–​2252. Zipser, K., Lamme, V. A. F., & Schiller, P. H. (1996). Contextual modulation in primary visual cortex. Journal of Neuroscience, 16(22), 7376–​7389. Zwaan, R. A. (2014). Embodiment and language comprehension: Reframing the discussion. Trends in Cognitive Sciences, 18, 229–​234. Zwaan, R. A. (2016). Situation models, mental simulations, and abstract concepts in discourse comprehension. Psychonomic Bulletin & Review, 23, 1028–​1034. Zwaan, R. A., & Madden, C. J. (2005). Embodied sentence comprehension. In D. Pecher & R. Zwaan (Eds.), Grounding cognition: The role of perception and action in memory, language, and thought (pp. 224–​245). New York: Cambridge University Press.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–​53) may, on occasion, appear on only one of those pages. 4E cognition, 18–​22, 151 behaviorism, 22 semantic memory, 19–​20 symbol grounding problem, 18–​22, 46 abstract concepts. See also specific topics brain, 130–​32 concrete concepts and, 164–​65 distributed neural pattern, 147–​51 elasticity, 205–​8, 206f experiential systems, 148–​49 grounded cognition and, 51 heterogeneity, 5–​6, 134, 147, 164–​65 (see also heterogeneity) mouth motor system, 159 organization, 134 reconsidered, 62–​63 semantic richness, 139 theoretical challenge, 51 abstract words, acquisition, 155–​64 affective grounding, 156–​57 earliest vocabularies, 155 linguistic grounding, 157–​60 meanings, 155 mode, 158 multimodal approach, 160–​64 structure-​to-​world mappings, 168 abstract–​concrete distinction, 4–​5 abstraction, 4–​5 vs. abstractness, 54 degree, 52–​53 abstractness vs. abstraction, 54 empirical measures, 5 acalculia, 109 acquisition, word abstract words, 155–​64 (see also abstract words, acquisition) concrete words, 155

action perception circuits, 119–​20 action systems abstract emotion word processing, 137–​38 concepts, 31–​36 concepts, processing, 200–​1 higher-​level representations, 81–​88, 82f–​84f metaphor, 183 activation modality-​specific, 199–​200 sensorimotor, metaphor, 183 spreading, 40 actotopy, 82–​83, 132–​33 ad hoc categories, 13 concepts, 13–​14 Adorni, R., 56 affect areas, multimodal reengagement, 23 grounding, abstract word acquisition, 156–​57 processing, metaphor, 181 representations, 12 affective embodiment, 135–​38 emotions, case for, 135–​36 more complicated picture, 136–​38 affective embodiment account, 136–​ 38, 208–​9 age of acquisition, abstract words vs. concrete words, 155 positive/​negative vs. neutral, 156–​57 alexithymia, 137–​38 Almohamadi, A., 192 Altmann, G. T. M., 61–​62 amodal representations/​view anterior temporal lobes, 88–​90 anterolateral lobes, 92 arguments for, 41–​44

256 Index concepts, 2–​3, 38–​39, 49, 88 definition, 25 disembodiment, 54 embodiment, 89, 90 information integration, multiple sources, 53–​54 language, 110–​11 language of thought, 123 Machery on, 39–​40 mirror neurons, 87–​88 modality-​specific information, 17 modified/​hybrid approach, 86–​87 modular approach, 45 semantic pointers, 93–​94 sensorimotor features, 96 supporters, resources, 49 system, 11 amyotrophic lateral sclerosis, 34–​35 Andrews, M., 121–​22 Anggoro, F. K., 165f, 165–​66 Annaz, D., 190–​91 anterior cingulate cortex, rostral, 136, 137 anterior temporal lobes, 37–​38, 88–​90, 122–​23, 136–​37 aphasia abstract word processing, 56 calculation and, 109 categorization, 118–​19 iconicity, 161 left hemisphere, 131 paroxysmal, 107–​8 primary progressive, logopenic variant, 31 semantic, 61 semantic memory, 209 Aristotle, 58, 59 Ascoli, G., 186–​87 Asperger’s syndrome, metaphors, 195–​97 autism spectrum disorders, 195–​97 metaphors, 137–​38 automaticity, 11 Barca, L., 159 Bargh, J. A., 182 Barr, R. A., 5 Barsalou, L. W., 13–​15, 17–​18, 21, 26, 46–​ 47, 126, 141 Bauer, A. J., 94–​95

Bedny, M., 122 behaviorism, 4E approaches and, 22 Bengio, Y., 74–​75 Bergen, B. K., 120, 179, 189–​90 Berio, L., 171–​72 Berkeley, G., 3, 51 Binkofski, F., 54 Blouw, P., 93 body in mind, 25–​49 action systems, 31–​36 perception systems, 26–​31 simulation and its discontents, 36–​44 alternative hypotheses, 38–​41 amodal representations, 41–​44 dissociation, 36–​38 symbol grounding problem, 44–​47, 50 weak vs. strong embodiment, 47–​49 body-​mind connections, 19 body-​world couplings, 19 body–​object interaction, 5, 56, 130–​31, 134, 143–​44 body, on mind, 19 Bonner, M. F., 31 bootstrapping, linguistic, 164 emotion, 158, 160 language, 160 mutual bootstrapping theory, 168 syntactic, 172–​73 Borghi, A. M., 31, 54, 110, 159 brain, conceptual, 10–​24 dogmas concept research, 10–​12 questioning, 12–​15 elasticity and 4E cognition, 18–​22 grounding, 22–​24 mind, something we do with, 15–​18 Brother John, 107–​8 Buccino, G., 32 Buckner, C., 75 byproduct views, 38–​39 Calzavarini, F., 125–​26 Cameron, L., 187 Camp, E., 124–​25 Caplan, L. J., 5 Caramazza, A., 54, 86–​87 Casasanto, D., 13–​15, 184–​85, 186, 189 Chemero, A., 21

Index  257 Christie, S., 167 Chrysikou, E. G., 184–​85 Clark, A., 111–​12 classical sandwich view, 11, 12 co-​occurrence statistics, 138 cognition. See also specific topics 4E (see 4E cognition) embedded, 18–​19 embodied (see embodied cognition) embodied vs. grounded, 22–​23 enactive, 18–​19 extended, 18–​19 grounded abstract concepts, 51 definition, 22–​23 structure and dynamics, rethinking, 4–​5 metacognition, social, 5, 143–​44 neural reuse in, flexible, 19 symbolic, language as, 23–​24 cognitive robotics, 52–​53 Cohen, A., 194 competence/​performance distinction, 22 computationalism, 20, 22, 44–​45 concept empiricism, 3 concept research. See also specific topics early, 10 exemplar theories, 10–​11 prototype theories, 10–​11 theory-​theories, 10–​11 concept(s). See also specific topics abstract–​discrete distinction, 4–​5 abstract/​abstraction, 4–​6 acquisition, 1–​2 ad hoc, 13–​14 amodal view, 2–​3 definitions, 13–​14, 15, 25 grounded/​grounding (see grounded concepts) hierarchies, 52–​53 higher order, 52–​53 nonlinguistic, 2 as representation production, 16 as scientific category, 13 as skills vs. mental objects, 16 stability, as illusion, 14 as static objects, 12 uses, 1, 13

concepts abstraction, 6–​9 object vs. relational categories, 164–​65 conceptual brain. See brain, conceptual conceptual metaphor theory, 176–​84 criticisms, 184 flexibility, 186–​87 conceptual representation, stimulus and context on, 24 conceptualization, defined, 18 concrete concepts abstract concepts and, 164–​65 heterogeneity (see heterogeneity) concrete words, acquisition, 155, 168 meanings, 155 mode, 158 concreteness/​concreteness effects, 5, 32–​33, 56, 130–​31, 134. See also imageability anterior temporal lobes, 90 context availability, 60, 139 definition, 55 dual coding, 139 emotionality effects, 135–​36 interference, 159, 164 mode of acquisition, 158 multidimensional research, 143–​44 reverse, 131 semantic dementia, 131 valence, 137 word learning, 164 Connell, L., 14–​15, 16–​17, 27, 56 constructions, 103–​6 cross-​linguistic, 112 cross-​linguistic variation, 120–​21 emergence theory of mind, 170, 171–​72 LENS theory, 116, 157–​58, 170, 171–​72 word meaning learning, 157–​58 context availability, 60, 62, 135–​36, 139, 143–​44, 158, 201 on conceptual representation, 24 context-​dependence, 13–​14 context-​dependent embodies simulation (CODES), 14–​15 context-​independence, 12, 13–​14 contextual diversity, 158, 160 contextualism, 13, 14, 59–​60

258 Index conversations, 126–​29 theory of mind, development, 169–​70 Cooper, R., 125 Coulson, S., 14–​15 Croft, W., 101–​2 cross-​modality modal representations, 37–​38, 44, 47, 76, 88 plasticity, 70 processes, 71, 72 Cruse, D. A., 101–​2 Crutch, S. J., 5, 143 Cuccio, V., 46–​47 cues, multiple, growth and development, 173–​74 Dahl, R., 188 Davis, C. P., 61–​62 de Villiers, J. G., 169, 170–​71 de Villiers, P. A., 169, 170–​71 dead metaphors, 188–​90 deaf children, theory of mind acquisition, 169–​70 deep learning networks/​methods, 74–​75 deflationary, 39, 42–​43 Dehaene, S., 42–​43 dementia, semantic, 48 amodal processing, 37–​38, 88–​89 concreteness/​concreteness effects, 131 processing deficits, 61 concrete concepts +​abstract conceptual knowledge, 56–​57 dependence, context-​, 13–​14 Desai, R., 146–​47, 149–​50, 190 development. See also growth and development; specific types neurotypical, 190–​92 developmental language disorder, 172–​73 diachronic flexibility, 154, 203–​4. See also growth and development dichotomous thinking, 199–​200 Dingemanse, M., 161 disembodiment, 54–​58, 97 amodal representations, 54 language and, 202–​3 LENS theory, 114–​15, 202–​3 dissociation, 36–​38 dogmas, concept research, 10–​12 questioning, 12–​15

Donald, M., 108 double-​mapping constraint, 194 dual code theory, 112, 135–​36, 139, 152 Dutriaux, L., 141 dynamic multilevel reactivation framework, 91–​92 dynamical systems theory, 21 ECCo theory, 112–​13 elastic grounding approach, 156, 162, 185 elastic mind. See elasticity elasticity, 154, 199–​209. See also flexibility 4E cognition, 18–​22 abstract concepts, 205–​8, 206f causal relevance, 199–​201 flexibility, synchronic and diachronic, 203–​4 future is now, 208–​9 hierarchical representations, 201–​2 language, influence, 202–​3 metaphor, case study, 204–​5 metaphors, 197–​98 theoretical challenges, 201 embedded cognition, 18–​19 embodied abstraction theory, 91, 92 embodied cognition, 2–​4 4E cognition, 18–​19 vs. classical sandwich view, 12 contextualism, 13, 14 definition, 22–​23 empiricism vs., 3 fundamentals, 12 vs. grounded cognition, 22–​23 invariantism, 13 task-​dependence, 14–​15 embodied conceptual combination theory, 112–​13 embodied construction grammar, 106 embodiment affective, 135–​38 emotions, case for, 135–​36 more complicated picture, 136–​38 amodal representations, 89, 90 problem, 54–​58 strong, 39, 96, 209 vs. neural reuse, 208 a priori commitment to, 90, 149 weak embodiment vs., 47–​49

Index  259 weak, 3–​4, 96, 201–​2 weak vs. strong, 47–​49 emergence theory of mind, 170, 171–​72 emotionality, 135–​36, 139, 143–​44, 162 emotions, 135–​36. See also affect empiricism, 192 neo-​, 3–​4, 39 new concept (concept), 3 enactive cognition, 18–​19 epigenetic robotics, 52–​53 exemplar theories, 10–​11 experiential systems, 36 neural reuse, 36, 45–​46, 127, 200–​1, 208 explanatory pluralism, 21 extended cognition, 18–​19 extended evolutionary synthesis, 116–​17 exteroception, 6–​7, 45–​46 extrastriate body area, 78 false-​belief tasks, language on, 170–​72 Faust, M., 195–​96 Fernadino, L., 35 Fillmore, C. J., 105–​6 flexibility, 154. See also elasticity degree, 203 diachronic, 154, 203–​4 (see also growth and development) problem, 58–​62 synchronic, 154, 203–​4 frame semantics, 105–​6 frozen metaphors, 188–​90 Fulkerson, M., 68 Gallese, V., 46–​47 Garrod, S., 126 generalization, 64 grounding, 76–​77 neural mechanisms, 64–​96 (see also hierarchies and hubs) problem, 52–​54 Gentner, D., 165f, 165–​66, 167, 168 Ghazanfar, A. A., 70 Gibbs, R. W., 184 Gleitman, L. R., 168 Gold, R., 195–​96 Goldberg, A., 104 Goldstone, R. L., 39, 128 Golonka, S., 14–​15

Grady, J., 186–​87 Grafton, S., 200 Gravelle, K., 65–​66 Graziano, M. S. A., 76–​77, 82–​83 Grossman, M., 31 grounded cognition definition, 22–​23 structure and dynamics, rethinking, 4–​5 grounded concepts, 2–​3, 4 grounding, 22–​24. See also specific topics elastic grounding approach, 156, 162, 185 reconsidered, 151–​53 symbol grounding problem, 44–​47, 50, 64–​65, 110, 111 symbol ungrounding problem, 50 theories, 3–​4 weak, 47, 71, 77, 151–​52 growth and development, 154–​74 abstract words, acquisition, 155–​64 affective grounding, 156–​57 linguistic grounding, 157–​60 multimodal approach, 160–​64 cues, multiple, 173–​74 developmental language disorder, 172–​73 iconicity, 161–​64, 173–​74, 194, 202–​4 relational categories, 164–​68, 165f, 166f theory of mind, 168–​72 Hale, C. M., 170–​71 hardness, 205–​7, 206f Harnad, S., 44–​46 Hauk, O., 32 heterogeneity, 134–​53 abstract concepts, 134 affective embodiment, 135–​38 emotions, 135–​36 more complicated picture, 136–​38 grounding reconsidered, 151–​53 multiple mechanisms, 147–​51 variation, dimensions, 138–​47 multidimensional approaches, 143–​47 semantic richness, 138–​39 situated conceptualization framework, 139–​41, 142f

260 Index heteromodality areas/​hubs, 7, 23, 70–​71, 91–​92, 95–​96, 149–​50, 151 convergence zones, 77 representations, 6–​7, 76, 96 heuristic offloading, 39 Hickok, G., 65, 72 hierarchical organization/​ hierarchies, 72–​77 black box, opening, 72–​76, 74f conceptual, 52 elasticity, 201–​2 generalization and grounding, 76–​77 hierarchies and hubs, 64–​96 hierarchical organization, 72–​77 black box, opening, 72–​76, 74f generalization and grounding, 76–​77 higher-​level representations, 77–​88 action systems, 81–​88, 82f–​84f perception systems, 77–​80, 79f hubs and spokes, 88–​94 anterior temporal lobes, 88–​90 multiple hubs, 91–​92 semantic pointers, 93–​94, 94f multimodal and multilevel representations, 94–​96 unimodality, beyond, 64–​72 higher order concepts, 52–​53 higher-​level representations, 77–​88 action systems, 81–​88, 82f–​84f perception systems, 77–​80, 79f Hinton, G., 74–​75 Hoffman, P., 90 homunculus, motor, 81, 82f horizontal intraparietal sulcus, 42–​43 hubs and spokes, 88–​94, 202 anterior temporal lobes, 88–​90 multiple hubs, 91–​92 semantic pointers, 93–​94, 94f human middle temporal cortex (hMT+), 78 Hume, D., 3, 51, 192–​93 Hurley, S., 11 hyperspace analog to language model, 121

onomatopoeia, 161, 163 sign language, 194 word learning, 203–​4 idioms, 101–​3, 105, 117–​18, 179, 183, 189–​90 imageability, 5, 55, 134. See also concreteness/​concreteness effects affective embodiment account studies, 137 concreteness, 144 definition, 55 maximum perceptual strength rating vs, 56 mode of acquisition, 158 multidimensional approaches, 143–​44 semantic diversity vs., 61 verbal representations and, 112 imagens, 112 Imai, M., 162 independence, context-​, 12, 13–​14 indexicality, 173–​74, 203–​4 inferential competence, 125 inner speech, 106–​8, 111, 112, 129 inter-​personal ecology, 187 interoception, 6–​7, 8–​9, 154 abstract concepts, 145–​46 abstract words, acquisition, 160 conceptual content encoding, 199 definition, 65–​66 sensations, 45–​46 WAT theory, 143–​44 word learning, 164 intraparietal sulcus, 146–​47 horizontal segment of, 42–​43 invariantism, 11 embodied cognition, 13 Iossifova, R., 193

iconicity, 161–​64, 173–​74 definition, 161 LENS theory and, 202–​3

Kaschak, M. P., 27 Kasirer, A., 196 Keeley, B., 66

Jensen, T. W., 187 Jeuniaux, P., 113 Joanette, Y., 107–​8 John, Brother, 107–​8 Johnsrude, I., 32 Just, J. A., 94–​95

Index  261 Kemmerer, D., 78, 79–​80, 106, 120–​ 21, 183–​84 Kersten, A., 128 Kiefer, M., 29–​30 Kita, S., 162 Klibanoff, R. S., 165f, 165–​66 Kompa, N. A., 48–​49 Kousta, S.-​T., 135–​36 label, 111, 117–​21 label-​feedback hypothesis, 120 Lambon Ralph, M. A., 14–​15 Landy, D., 39 language. See also metaphor; specific topics amodal representations, 110–​11 concept acquisition, 2 developmental disorder, 172–​73 discourse, culture and, 184–​87 disembodiment, 202–​3 elasticity, 202–​3 false-​belief tasks, 170–​72 influence, 202–​3 inner speech, 106–​8, 111, 112, 129 linguistic bootstrapping, 160 meaning, 98–​107 modularity, 110–​11 neural reuse, 130 relational categories, 164–​68, 165f, 166f semantic memory, 97–​98 sensorimotor representations, 120 sign, 194 situated view, 162–​63 as symbolic cognition, 23–​24 systemic view, 162–​63 theory of mind, development, 168–​72 of thought, 123 amodal representations, 123 in word and thought, reimagining, 111–​14 language and associations in thinking theory, 112–​13, 123 language and situated simulation theory, 112–​13 language as neuroenhancement, 97–​133 abstract concepts, brain, 130–​32 conversations, 126–​29 embodied mind, 132–​33 fundamentals, 97–​98 labels, 117–​21

symbolic medium, 98–​107 syntax, 123–​26 thinking without words, 107–​9 word and thought, 110–​17 LENS theory, 114–​17 (see also LENS theory) opposition, early, 110–​11 role of language, reimagining, 111–​14 word associations, 121–​23 language is an embodied neuroenhancement and scaffold (LENS). See LENS theory LASS theory, 112–​13 LASSO theory, 112–​13, 123 latent Dirichlet allocation model, 121 latent semantic analysis model, 121 lateral occipitotemporal cortex, 75, 78–​80, 79f, 88 Lebani, G. E., 138 Lecours, A. R., 107–​8 LeCun, Y., 74–​75 Lenci, A., 138 LENS theory, 114–​17, 157–​58 abstract concepts, brain, 130–​32 conversations, 126–​29 disembodiment, 114–​15, 202–​3 growth and development, 162 labels, 117–​21 language, 168, 172 syntax, 123–​26 word associations, 121–​23 lexical decision, 27 lexicon, mental, 101 Lingnau, A., 80 linguistic grounding, abstract word acquisition, 157–​60 linguistic metaphor, 189 Littlemore, J., 195 Locke, J., 3, 51 logogens, 112 logopenic variant, primary progressive aphasia, 31 long-​term memory, 11–​12, 13, 14, 16, 101, 140 Louwerse, M., 113 Lund, T. C., 158 Lupyan, G., 13–​15, 118–​19, 120 Lynott, D., 14–​15, 16–​17, 27, 56

262 Index Machery, E., 11, 13, 38–​40, 41 Madden, C. J., 132–​33 Mahon, B. Z., 54 Majid, A., 30 Maley, C. J., 20 Marconi, D., 125 Marmolejo-​Ramos, F., 193 Marshal, N., 196 mathematical reasoning, 108–​9 Matlock, T., 105 maximum perceptual strength rating, 56 Mazzuca, C., 159 meaning language, 98–​107 word acquisition, 155 word constructions, 157–​58 Meir, I., 194 memory long-​term, 11–​12, 13, 14, 16, 101, 140 recognition, valence effects, 157 semantic (see semantic memory) mental lexicon, 101 mental metaphor, 189 metacognition social, 5, 143–​44 WAT theory, 114 metaphor, 175–​98 abstract concept connection, 175–​76 action systems, 183 attention, 187 case study, elasticity, 204–​5 conceptual metaphor theory, 176–​84 conceptual, verticality and, 181–​82 development, 190–​97 autism spectrum disorders/​ Asperger’s, 195–​97 neurotypical, 190–​92 sensory differences, 192–​94 for effect, deliberate use, 187 elastic, 197–​98 experience-​based, 176 inter-​personal ecology, 187 language-​specific, 191–​92 language, discourse, and culture, 184–​87 language, thought, and communication, 176 linguistic, 189

mappings, 180 mental, 189 metaphorical simulation hypothesis, 179 novel vs. conventional, 198 primary, 191, 193 sensorimotor system activation, 183 sleeping, 188–​90 target/​source domain, one-​to-​many relationships, 180–​81 thought, language, and discourse, 189 Metaphor View, 177, 197–​98 metaphorical simulation hypothesis, 179 metaphoricity, 187 Meteyard, L., 53–​54 mind. See also theory of mind body in, 25–​49 (see also body in mind) body-​mind connections, 19 elastic (see elasticity) embodied (language as), 132–​33 other-​species-​of-​mind problem, 66 something we do with, 15–​18 Mirman, D., 118–​19 mirror neurons, 84–​88 modality-​specific activation, 199–​200 modality-​specific processing, 27 modality-​specific representation, 65 anterior temporal lobes, 88–​89 cortical organization, 69–​71 grounded simulations, 68 multilevel conceptual grounding, 150 neuroimaging studies, 200 semantic memory, 53 sensorimotor system, dynamic body interactions with environment, 50 (see also specific modalities; specific topics) conceptual tasks, 12 information sources, 17 neo-​empiricism, 39 weak vs. strong embodiment, 48 sensorimotor systems, 70–​71, 86–​87, 96, 201–​2 symbol ungrounding problem, 50 weak grounding, 77 modality-​specific systems, 2–​3 modality-​switching cost, perception systems, 26

Index  263 mode of acquisition, word, 158 modularity, language, 110–​11 Moffat, M., 139 morphemes, 5–​6, 99–​100, 101–​2, 116, 120–​21, 136–​37 Motamedi, Y., 162–​63 motion agnosia, 75 motor cortex, 81–​85, 82f–​84f motor homunculus, 81, 82f motor systems. See action systems mouth motor system, abstract concepts, 159 multidimensional approaches, 143–​47 multilevel representations, 94–​96 multimodality abstract word acquisition, 160–​64 representations, 94–​96 sensorimotor and affective cortex reengagement, 23 multiple cues, growth and development, 173–​74 multiple hubs, 91–​92 multisensory processing, 69–​70, 71–​ 72, 75–​76 Muraki, E. J., 155, 160 Murgiano, M., 162–​63 mutual bootstrapping theory, 168 neo-​empiricism, 3–​4, 39 Nestor, P. J., 53 network hubs, 77 neural representation, 19–​22. See also specific topics neural reuse, 75–​76, 174, 175 causal relevance, 199–​201 conceptual processing, 49 evidence, review, 199–​200 experiential systems, 36, 45–​46, 127, 200–​1, 208 flexible, in cognition, 19 language system, 130 neural simulation, 17, 53, 75–​76, 86, 87–​ 88, 179, 199–​200 neurotypical development, 190–​92 new concept empiricism, 3 Niedenthal, P., 14–​15 Norbury, C. F., 155, 156–​57, 158, 172–​73

number sense/​approximation, amodal representations, 41–​43 Nunberg, G., 102 occipitotemporal cortex, lateral, 75, 78–​80, 79f, 88 offloading hypothesis, 38–​41 olfaction, 68 Olson, I. R., 137 onomatopoeia, 117–​18, 161, 163 other-​species-​of-​mind problem, 66 Oullet, M., 185 Pado, S., 52 Papagno, C., 57 Parkinson’s disease, 34–​35 Passaro, L. C., 138 Patterson, K., 53 perception systems, 26–​31, 77–​80, 79f conceptual processing, 200–​1 higher-​level representations, 77–​ 80, 79f interference and facilitation, 27 modality-​switching cost, 26 task and context, 27 perceptual strength, 56, 143–​44 Perry, L. K., 161–​62 Pexman, P. M., 139, 148, 155, 158, 160 phonemes, 99–​100 Piazza, M., 42–​43 Pickering, M. J., 126 Pinker, S., 59 plasticity, cross-​modal, 70 Plato, 58 pluralism, explanatory, 21 Ponari, M., 155, 156–​57, 158, 172–​73 Porat, R., 186 posterior fusiform gyrus, 27–​28, 29 posterior middle temporal gyrus, 78–​80 posterior superior temporal sulcus, 78 premotor cortex, 84–​85, 87–​88 primary metaphors, 191, 193 primary progressive aphasia, logopenic variant, 31 Prinz, J. J., 46–​47 problems, three. See three problems proprioception, 65–​66 prototype theories, 10–​11

264 Index Proverbio, A. M., 56 Pulvermüller, F., 32, 119–​20 recognition memory, valence effects, 157 referential competence, 125 Reggin, L. D., 155, 160 Reilly, J., 92, 143 Reilly, M., 146–​47, 149–​50 relational categories, language, 164–​68, 165f, 166f representation. See also specific types of cognitive neuroscientists vs. philosophers, 21–​22 neural, 19–​22 (see also specific topics) reuse experiential mechanisms, 199 neural (see neural reuse) reverse concreteness effects, 131 Rivlin, R., 65–​66 robotics, cognitive/​epigenetic, 52–​53 Rogers, T. T., 53 rostral anterior cingulate cortex, 136, 137 Rundblad, G., 190–​91 Sag, I. A., 102 Santiago, J., 176–​77 Saygin, A. P., 189–​90 Schaller, S., 108 Scheepers, C., 141 Schroeder, C. E., 70 Scorolli, C., 31 semantic aphasia, 61 semantic control system, 14–​15 semantic dementia, 48 amodal processing, 37–​38, 88–​89 concreteness/​concreteness effects, 131 processing deficits, 61 concrete concepts +​abstract conceptual knowledge, 56–​57 semantic diversity, 61, 62, 120–​21, 203 semantic memory 4E cognition and, 19–​20 abstract-​concrete distinction, 203–​4 categories encoded, 26 distributional +​experiential information, 113 hierarchical organization, 23, 209 higher level representations, 202

language, 97–​98 mental lexicon, 101 mirror neurons, 87 motor involvement, 81, 83f multilevel models, 53 multiple hubs, 91 neuromechanisms, 49, 119 semantic dementia, 37–​38 sensorimotor mechanisms, 199–​200 touch, taste, and smell, 30 visual areas, higher level, 79–​80 semantic pointers, 93–​94, 94f semantic richness, heterogeneity, 138–​39 semantic tasks, perception systems in, 26 Semino, E., 178 senses differences, metaphor use, 192–​94 differentiating/​individuating, 65–​69 experience, 3 modality definition, 65–​66 theoretical approaches, 66–​68 pluralism, 68–​69 vestibular, 65–​66 sensorimotor cortex multimodal reengagement, 23 textual metaphors activating, 30 sensorimotor processing metaphor, 181 semantic memory, 199–​200 sensorimotor representations, 7, 12, 50 basic-​level concepts, 95–​96 conceptual processing, 36 grounded view, 12, 45–​46, 201–​2, 208 language manipulating, 120 metaphors, semantic processing, 181 modality-​specific, 201–​2 offloading hypothesis, 38–​40 sensorimotor system, modality-​specific representation, 70–​71, 86–​87, 96, 201–​2 dynamic body interactions with environment, 50 conceptual tasks, 12 information sources, 17 neo-​empiricism, 39 weak vs. strong embodiment, 48 Shallice, T., 125

Index  265 Shapiro, L., 46 Shea, N., 5 Shen, Y., 186 SI theory, 112–​13 Sidhu, D. M., 158 sign language, 194 Simmons, W. K., 28 simulation, 36–​44 alternative hypotheses, 38–​41 amodal representations, 41–​44 definition and example, 64–​65 dissociation, 36–​38 neural, 17, 53, 75–​76, 86, 87–​88, 179, 199–​200 situated conceptualization framework, 139–​41, 142f situated simulations, 139–​40 situated view, language, 162–​63 situation, defined, 140 situational systematicity, 61–​62, 203 Skipper, L. M., 137 sleeping metaphors, 188–​90 sound symbolism bootstrapping hypothesis, 162 spatial processing, 69 spatial representation, 69 speech. See also language; word inner, 106–​8, 111, 112, 129 speed, 11 Speed, L., 30 spreading activation, 40 Stanfield, R. A., 26–​27 Stapleton, M., 46–​47 Steen, G. J., 187, 189 stimulus. See also specific types and topics on conceptual representation, 24 strong embodiment, 39, 96, 209 vs. neural reuse, 208 a priori commitment to, 90, 149 weak embodiment vs., 47–​49 superior parietal lobule, 146–​47 supramodality areas/​hubs, 23, 91 representations, 7–​8 symbol grounding problem, 44–​47, 50, 64–​65, 110, 111 symbol interdependency theory, 112–​13 symbol ungrounding problem, 50

symbolic medium, language as, 98–​107 synchronic flexibility, 154, 203–​4 synesthesia, 66–​67 syntactic bootstrapping, 172–​73 syntax, 123–​26 complex, on false-​belief tasks, 170–​71 systemic view, language, 162–​63 Tager-​Flusberg, H., 170–​71 task-​dependence, 14–​15 temporal lobes, anterior, 88–​90 theory of mind, development conversations, 169–​70 deaf children, 169–​70 language, 168–​72 linguistic constructions, 170–​71 mental state terms, 169 theory of mind, emergence, 170, 171–​72 theory-​theories, 10–​11 Thill, S., 52 thinking/​thought. See also specific topics as action, 15–​16 language of, 123 without words, 107–​9 word and, 110–​17 (see also word and thought) Thompson-​Schill, S. L., 37 three problems, 50–​63 abstract concepts reconsidered, 62–​63 embodiment, 54–​58 flexibility, 58–​62 generalization, 52–​54 Tillas, A., 123, 124–​25 topicalization, 103 transformational abstraction, 75 Umiltà, M. A., 83–​84 unimodality, beyond, 64–​72 valence abstract word acquisition, 157 recognition memory, child, 157 van Dam, W., 146–​47, 149–​50 verticality, conceptual metaphors, 181–​82 vestibular senses, 65–​66 Vigliocco, G., 121–​22, 155, 156–​57, 158, 162–​63, 172–​73 Villani, C., 143–​44, 145

266 Index Vinson, D., 121–​22 visual system/​perception cross-​modal/​heteromodal representations, 76–​77 hierarchical organization, 73–​75, 74f Vygotsky, L., 111 Warrington, E. K., 5 Wasow, T., 102 WAT theory, 112–​14, 129, 130 multidimensional approaches, 143–​44 weak embodiment, 3–​4, 96, 201–​2 vs. strong embodiment, 47–​49 weak grounding, 47, 71, 77, 151–​52 Weiskopf, D., 125 Wiemer-​Hastings, K., 5 Williams, L. E., 182 Wilson, A. D., 14–​15 Wilson, M., 205–​7, 206f Winkielman, P., 14–​15 Winter, B., 105

word and thought, 110–​17 LENS theory, 114–​17 opposition, early, 110–​11 role of language, reimagining, 111–​14 word as social tool theory. See WAT theory word(s). See also abstract words; concrete words associations, 121–​23 learning, concreteness, 164 thinking without, 107–​9 Wurm, M. F., 80 Xu, X., 5 Yee, E., 37, 52, 61–​62, 127–​28 Yuan, S., 119 Zdrazilova, L., 139 Ziemke, T., 52 Zwaan, R. A., 26–​27, 113, 132–​33