To be or not to be a word : new reflections on the definition of word 9781443862073, 144386207X

Citation preview

To be or not to be a Word

To be or not to be a Word: New Reflections on the Definition of Word

Edited by

Iraide Ibarretxe-Antuñano and José-Luis Mendívil-Giró

To be or not to be a Word: New Reflections on the Definition of Word, Edited by Iraide Ibarretxe-Antuñano and José-Luis Mendívil-Giró This book first published 2014 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2014 by Iraide Ibarretxe-Antuñano, José-Luis Mendívil-Giró and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-6207-X, ISBN (13): 978-1-4438-6207-3

TABLE OF CONTENTS

Introduction ................................................................................................. 1 Words, Words, Words… Iraide Ibarretxe-Antuñano and José-Luis Mendívil-Giró Chapter One ................................................................................................. 6 The Word in Phonology Gorka Elordieta Chapter Two .............................................................................................. 66 Word-Defining Properties from a Lexical Approach to Morphology Elena Felíu-Arquiola Chapter Three ............................................................................................ 93 On a Grammatically Relevant Definition of Word and why it belongs to Syntax Antonio Fábregas Chapter Four ............................................................................................ 131 From Agglutination to Polysynthesis: Towards a Biological Characterization of the Spoken Word Juan-Carlos Moreno-Cabrera Chapter Five ............................................................................................ 164 Words as Constructions: Some Reflections in the Light of Constructionist Approaches Francisco Gonzálvez-García Chapter Six .............................................................................................. 189 Word in Cognitive Grammar Ricardo Maldonado and Boris Fridman-Mintz Contributors ............................................................................................. 244 Index ........................................................................................................ 247

INTRODUCTION WORDS, WORDS, WORDS… IRAIDE IBARRETXE-ANTUÑANO AND JOSÉ-LUIS MENDÍVIL-GIRÓ

“Words, words, words”… This is how Hamlet replies to Polonius’ question: “What do you read, my lord?” in Shakespeare’s famous, and much quoted, play Hamlet (Act II, sc. ii). Although apparently simple, this phrase raises a much deeper and complex issue since… What is really a “word”? This is precisely the “question” addressed in this book: the nature and definition of what a “word” is in linguistics. As happens with Hamlet’s answer, it is not easy to define what a “word” is. Indeed, the term subsumes a wide range of phenomena explored from an even wider array of perspectives (see Cruse et al. (2002), Dixon and Aikhenvald (2002), DiSciullo and Williams (1987), among others). And, yet, “words” are the most accessible linguistic units from the speaker’s introspection viewpoint; however, they are, at the same time, an incredibly elusive reality for the linguist. “Words” have been the main linguistic units in the Western grammatical tradition based on Greco-Latin grammar. Linguistic structuralism, in an attempt to avoid using this problematic concept, put forward the notion of morpheme as the central unit for scientific analysis. Unfortunately, structuralist scholars did not meet their purpose. Nowadays, the term “word” has not disappeared from the research agenda in Linguistics. Rather, it is still one of the key concepts in research dealing with the human language where issues such as the definition, theoretical status, limits, characteristics, and psycholinguistic reality of the notion of word are still controversial and, at the same time, open for debate. All these issues will be discussed at length in the following pages. This book is the result of a scientific meeting organized by the research group

2

Introduction

Sylex (Syntax and Lexis) at the University of Zaragoza, Spain.1 It offers an up-to-date overview of the latest discussions on the nature of “word” in modern linguistics. Indeed, the main contribution of this book lies in that it gathers together under a single collective volume different views of what a “word” is from a wide range of diverse methodological and theoretical linguistic frameworks such as phonological theory (Elordieta), linguistic typology (Moreno-Cabrera), lexical generative morphology (FelíuArquiola), generative syntax (Fábregas), cognitive grammar (Maldonado and Fridman-Mintz), and construction grammar (Gonzálvez-García). Despite their different backgrounds, all these papers are geared towards the same goal: to offer a detailed account of what a “word” is in their respective fields. Elordieta’s chapter provides an overview of the most relevant aspects of the concept of word in generative phonology (most commonly referred to as the phonological or prosodic word): the demarcation of the word in the speech chain, the nature of the relations between morphology and phonology (i.e. morphophonology, Lexical Phonology), conditions on phonological word formation, and the relationship between morphosyntactic words and phonological words. According to Elordieta, the lack of isomorphism or correspondence between the morphosyntactic and phonological sides of the grammatical word reveals itself in the prosodic aspects of clitics, affixes and compound words. The next two chapters provide a formal approach to words from two contrasting perspectives of the notion of word which represent a long and fruitful debate on the internal architecture of the Human Faculty of Language within the Generative tradition (in the sense that the study of what a word is equals studying how grammars are organized). Thus, FelíuArquiola’s contribution takes a lexical approach to morphology, while Fábregas presents a syntactic (“constructivist”) approach to the notion of word. Felíu-Arquiola’s chapter starts addressing a central question: “Is it legitimate to adopt an approach in which morphological processes are not considered the result of syntactic operations?” After claiming that this is indeed the case, she argues that languages seem to show lexical integrity effects, although neither in the same degree nor in relation to the same kind of phenomena. Therefore, she proposes that from a lexical approach 1 This book falls within the scope of the Spanish Government funded research projects TAC (HUM2007-64200) and MovEs (FFI2010-14903). We would like to thank all contributors for their expertise and patience as well as Rosario Caballero, Paula Cifuentes-Pérez, Francisco Gonzálvez-García, and Javier Valenzuela for their help with the manuscript.

Words, Words, Words…

3

to morphology words can be considered as items that tend to show lexical integrity properties and that tend to behave as closed semantic domains, in contrast to phrases and sentences. Since words and phrases seem to show different morphologic, syntactic and semantic behaviors, it should be possible to consider that they are produced by different types of operations. As a consequence, morphological processes–processes that give rise to complex words–can be considered different from syntactic operations, since their outputs tend to behave in a different way. Fábregas’ chapter departs from two questions: Is there any grammatically relevant definition of word? If such a definition exists and we can talk about words as distinct units in the grammatical knowledge of speakers, how are they defined? In other words, is the definition different enough from what we know about syntax to allow us saying that morphology is a distinct component? The author proposes an affirmative answer to the first question, which seems to place him within a lexicalist perspective; however, he gives a negative answer to the second question, which leads him to propose a purely syntactic definition of word. According to Fábregas, this strategy allows him to account for the so-called Lexical Integrity effects without renouncing to the more theoretically ambitious assumption that there is only one generative component in language. More specifically, the author makes a syntactic definition of word using phasetheoretical arguments and considering the deactivation of the syntactic derivation of chunks of structure in relation to interface conditions. Should this be the case, his model would predict that all languages have words, to the extent that in all languages syntax is assumed to exist and syntax is a system that deactivates active features in order to allow structures to be processed by the interfaces. The book continues with Moreno-Cabrera’s chapter, which somehow serves as a link between formal and functionalist-cognitivist approaches. Moreno-Cabrera’s chapter proposes a critical overview of a robust (but not very spread) tradition, i.e. a purely naturalistic approach to the nature of linguistic units in general and words in particular. In his contribution, Moreno-Cabrera focuses on the spoken word from the perspective of speech and its impact on the morphological typology of languages. He argues that classical morphological typology is almost exclusively based on the written word and, therefore, is of little use for establishing a word typology relevant to spoken languages. He concludes that some of the word types envisaged in classical morphological typology do not define language types and may, instead, be interpreted as manifestations of universal properties of human speech. He then shows that the spoken word can be conceived of as a linguistic realization of “chunks” in short-term

4

Introduction

memory tasks. Later he relates this approach with Dahl’s (2004) notion of “packets” and with De Reuse’s (2009) proposal concerning the polysynthetic word. He concludes that the quantitative limits of words as speech chunks coincide with the “magical number seven, plus or minus two” originally proposed by Miller (1956). The two chapters that close this book come from a cognitive linguistic tradition. Gonzálvez-García focuses on word and construction grammar (see Hoffmann and Trousdale 2013 for a review), whereas Madonado and Fridman-Mintz draw on Langacker’s (1987, 1991, 2008) Cognitive Grammar to define what a word is. Cognitive Linguistics is not a homogeneous framework, but an umbrella term subsuming different approaches that share basic theoretical and epistemological tenets. Therefore, words in construction grammar approaches are considered “constructions”, i.e. form-meaning pairs (or “conventional associations of a given form and its corresponding semantics and/or discoursal function”) and words in Cognitive Grammar are regarded as minimal symbolic units (composed of a semantic pole and a phonological pole). Regardless of internal differences, both sub-models coincide in considering morphology and syntax not as separate and/or (in)dependable units of analysis but as members of a grammar-lexicon (semantics/pragmatics) continuum. Apart from their specific views on what a “word” is, each paper provides a succinct description of their corresponding frameworks and focuses on certain word-related crucial phenomena. Gonzálvez-García provides an overview of the main constructionist approaches that exist nowadays with a special focus on the characteristics they have in common. Later he describes how Golberg’s Construction Grammar accounts for compounding, coercion and the saber ‘know’ + secondary predicate constructions in Spanish. In turn, and after briefly introducing some of the basic mechanisms in Langacker’s model (especially, base/profile, compositional path, and scope), Maldonado and Fridman-Mintz propose a cognitive account of what a syllable is from a cognitive phonetics perspective. They use this notion of syllable to define what a word is and to discuss several related issues such as cohesion, separability, independence, and informativity. All in all, this collection of papers offers different perspectives that will contribute to provide some answers to the myriad of questions that a simple phrase such as “to be or not to be a Word” arises.

Words, Words, Words…

5

References Cruse, D. A., F. Hundsnurscher, M. Job and P. R. Lutzeier (eds.) 2002, Lexikologie/Lexicology: Ein Internationales Handbuch zur Natur and Struktur von Wörtern und Wortschätzen/An international handbook on the nature and structure of words and vocabularies, Berlin: Mouton de Gruyter Dahl, Ö. 2004, The Growth and Maintenance of Linguistic Complexity, Amsterdam: John Benjamins De Reuse, W. J. 2009, “Polysynthesis as a typological feature. An attempt at a characterization from Eskimo and Athabascan perspectives”, M.A. Mahieu and N. Tersis (eds.) Variations on Polysynthesis. The Eskaleut Languages, Amsterdam: John Benjamins, 19-34 Dixon, R. M. W. and A. Y. Aikhenvald (eds.) 2002, Word: A CrossLinguistic Typology, Cambridge: Cambridge University Press DiSciullo A. M. and E. Williams 1987, On the Definition of Word, Cambridge (Mass.): MIT Press Hoffmann, H. and G. Trousdale (eds.) 2013, The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press Langacker, R. 1987, Foundations of Cognitive Grammar, Vol. 1. Theoretical Prerequisites, Stanford: Stanford University Press —. 1991, Foundations of Cognitive Grammar, Vol. 2: Descriptive Application, Stanford: Stanford University Press —. 2008, Cognitive Grammar, Oxford: Oxford University Press Miller, G. A. 1956, “The magical number seven, plus or minus two: Some limits on our capacity for processing information”, Psychological Review 63, 81-97

CHAPTER ONE THE WORD IN PHONOLOGY GORKA ELORDIETA

1. Introduction Included in a volume that contains chapters addressing aspects of the word from different linguistic disciplines, this chapter provides a short overview of the evidence provided by the sound component of grammar (phonetics and phonology) for a domain or grammatical entity referred to as a “Word”, that is, the Phonological Word. The chapter is conceived as a quick mini-guide of the most relevant issues pertaining to the word in phonology, without stopping at any depth in any of them. It is mainly intended for a reader with basic knowledge of phonology who would like to have access to an overview of some of the main issues on the word in phonology, especially in the context of a volume containing papers from different modules of grammar and which may attract the attention of a generalist reader or a reader not specialized in phonology. The chapter may prove particularly useful for anyone interested in Spanish linguistics, *

First and foremost, I would like to thank the editors of this volume, Iraide Ibarretxe-Antuñano and José-Luis Mendívil-Giró, for their interest in having me as a contributor to this volume, and above all, for their infinite patience while awaiting my submission (which was delayed due to several circumstances). Parts of the contents of this article were presented as an invited talk at the Workshop “On the nature of the word”, at the University of Zaragoza on October 7-8, 2010, organized by Sylex, the research group on syntax and the lexicon of that university and to which the editors of this volume belong. I am grateful to the members of that group for their invitation and to the participants in that event for their comments and feedback. This work forms part of the research activities of the Group of Theoretical Linguistics (HiTT, in its Basque acronym) funded by the Basque Government (reference code IT769-13), of the Research and Formation Unit on Theoretical and Diachronic Linguistics (UFI11/14) of the University of the Basque Country (UPV/EHU), and the funding from these institutions is gratefully acknowledged.

The Word in Phonology

7

as the paper contains many references to this language, in the form of examples illustrating specific issues under discussion and specially in the form of analyses of my own on the prosody of clitics, affixes, and compound words in Spanish (to my knowledge, there is no other analysis of the prosodic structure of affixes, compound words, and clitics in Spanish).2 There are many aspects related to the phonological word that will not be dealt with in this paper, because the breadth of the topic of “the word in phonology” exceeds the limits of space for the papers in this volume. A careful, detailed coverage of the topic certainly deserves a book rather than an article of 60-some pages. This constraint on paper size also prevents me from illustrating every issue with examples. But with the snapshots presented in the paper I hope to offer at least an initial service to readers interested in getting to know about phonological studies of the word, so that they can proceed later to the specialized literature on each topic.3 The reminder of the chapter is organized as follows: section 2 presents some aspects of the phonetic and phonological cues for the demarcation of words. Section 3 offers a brief overview of the main framework devised for studying word-internal phonology, namely Lexical Phonology. Section 4 focuses on the conditions on the formation of prosodic words (such as minimum-size requirements), and the correspondences between phonological words and syntactic terminal nodes, as part of the object of study of Prosodic Phonology. In this section, particular attention is paid to the prosodic structures of function words adjacent to lexical words, affixed words, and compound words. Finally, section 5 addresses the relevance for the domain of the word of recent proposals on the syntax-phonology interface that advocate for a direct mapping from morphosyntax to phonology, in the sense that morphological and syntactic phases (or rather the Spell-Out domains of phases) are mapped onto PF as domains where phonological phenomena of various sorts may apply.

2. Phonetic and phonological demarcation of words In most common everyday speech situations, phonological words are not separated or preceded by pauses. That is, speakers of different languages do not indicate word boundaries by pauses, which would be the most obvious means of identifying words. Thus, the question arises 2

This slight bias arises from the fact that this chapter is a written development of an oral presentation at the University of Zaragoza, Spain, for an audience of researchers in General and Spanish linguistics. 3 Hall (1999a) is also an excellent review article on the phonological word.

8

Chapter One

whether there are any phonetic or phonological cues that serve to demarcate the edges of words in the speech chain. Several studies have reported the existence of different types of cues for word boundaries present in the world’s languages. At least in higher levels of the prosodic hierarchy (utterance, intonational phrase, intermediate phrase, phonological phrase), the left edges or beginnings of words have been found to be pronounced with a higher degree of articulatory strength, in the form of greater duration of initial consonants and/or initial syllables, VOT or aspiration in initial consonants, glottalization of word-initial vowels, or bigger resistance to assimilation and lenition processes. In contrast, word-medial consonants and vowels are more prone to weakening effects, such as lenition, reduction, and deletion. Diachronically, strengthening processes may result in favoring obstruents in word-initial positions. Inversely, weakening processes affecting word-medial segments may lead to the occurrence of sonorants word-medially (cf. Martinet 1955, Hock 1991, 1992, Vennemann 1993). On the other hand, the ends or right edges of words can also be strengthened. At higher levels of the prosodic hierarchy, word-final syllables can be signaled by lengthening, accompanied by other suprasegmental cues such as the presence of boundary tones. These phenomena are clearer the higher the prosodic constituent the word starts or ends. The following references of phonetic analysis can be consulted: Quené (1992), Fougeron and Keating (1997), Byrd and Saltzman (1998), Fougeron (1999), Byrd (2000), Cho and Keating (2001), Cho (2002, 2004, 2006), Keating et al. (2003), Cho, McQueen and Cox (2007), inter alia. For a more phonological discussion of word-initial versus word-internal positions, see Beckman (1998), Kirchner (1998, 2001), Smith (2005) and Barnes (2005), among others. Stress at the left and right edges of words is very common among the world’s languages. Of the 306 languages surveyed by Hyman (1977), 114 (37.3%) had initial stress, and 97 (31.7%) had final stress. Initial and final stress systems are thus found in 211 languages, 69% of the languages consulted. These languages were both quantity-sensitive and quantityinsensitive. Gordon (2002) offers a survey of 262 quantity-insensitive languages, divided into languages with fixed single stress, languages with dual stress (primary and secondary), and languages with binary and ternary stress. Of the 262 languages, 85 languages have initial stress and 75 have final stress (61.08% of the languages).4 Finally, Goedemans and van der 4

These figures are not found in any single table in Gordon’s (2002) article. Rather, they are the result of my own calculations on the location of primary stress, taking

The Word in Phonology

9

Hulst (2011) claim that of the 282 languages with fixed stress they analyzed, 92 have initial stress and 51 have final stress. Thus, 143 of the 282 languages have initial or final stress (50.70%). But the percentages of stress systems marking word edges as stress locations increase substantially once penultimate stress is considered. This pattern is very common among fixed stress systems. Penultimate stress is only one syllable away from the word’s right edge, and Gordon (2001) hypothesizes that penultimate stress may derive from final stress due to phonetic reasons against having final stress in phrase-final position. In phrase-final position, a word with final stress would have to realize both the tonal configuration for stress and the one for a boundary tone. If the tonal specifications for stress and the boundary tone were of the opposite type, such as a rising tone or a high tone for stress and a falling tone or a low tone for the boundary tone (as in declaratives), the crowding that originates could motivate retraction of stress to the penultimate syllable. This retraction in phrase-final position could then be reanalyzed as penultimate stress at the word level. Another aspect not favoring the last syllable of a word in phrase-final position as a bearer of stress is that at higher level prosodic constituents the last syllable or vowel is weakened, as manifested in many languages by vowel devoicing, reduction, neutralization or deletion (cf. Hock 1999, Barnes 2005). A fact in support of these phonetic explanations is the asymmetry between the right and left edges. The equivalent of penultimate stress on the left edge of the word is peninitial stress (i.e. stress on the postinitial syllable), and the number of languages with this pattern is much lower than the number of languages with penultimate stress. If we add languages with penultimate stress to the number of languages with initial and final stress, the percentage of languages with “edge” stress is a clear majority. In Hyman’s (1977) corpus, there are 77 languages with penultimate stress, which added to initial and final stress amount to 94.11% of the total languages considered. Gordon (2002) reports 76 languages with stress on the penultimate syllable, which together with initial and final stress languages gives a total percentage of 90.07% of the languages surveyed. Finally, in Goedemans and van der Hulst’s (2011) classification, there are 110 languages with penultimate stress, and thus 253 of the languages studied would have initial, final or penultimate stress, i.e. 89.71% of the total. According to these surveys, then, between 89% and 94% of the languages would have stress at a word’s edge or on the penultimate syllable, only one syllable from the word’s edge. These figures would Gordon’s figures for single and dual stress patterns and binary and ternary stress patterns across the article.

10

Chapter One

reflect quantitatively what Trubetzkoy (1939) called the demarcative function of stress. Word phonotactics may also be another cue for word edges. For instance, in English monomorphemic words obstruent clusters are permitted at word-end codas, but not word-internally. In Dutch, word-final syllables are longer than word-internal syllables. In several languages, certain consonants, consonant clusters and vowels are banned from wordinitial or word-final positions, or alternatively, only certain consonants may be allowed in word-initial position or word-final codas. Dixon and Aikhenvald (2002: 14-15) mention several cases (cf. the references therein for details). In Bare, aspirated consonants can only occur word-initially, and nasalized vowels can only occur word-finally; in Zoque, there are onsets with three and four consonants, but only in word-initial position; in Estonian, one can find syllables beginning with a vowel only in wordinitial position; in Basque, stops and affricates can be codas only wordfinally (cf. Hualde 2003: 34). The presence of such segments or sequences of segments at the left or right edge of a word in these languages is an indicator of word boundaries. Phonotactic restrictions may also hold of certain vowel or consonant sequences occurring across word edges. In Fijian, vowel sequences may be pronounced as diphthongs word-medially, but when they straddle two words they have to be pronounced in different syllables (cf. Dixon and Aikhenvald 2002: 15); in many languages certain sequences of consonants are only permitted across word boundaries, not internal to the word (e.g. in Basque, sequences of stop or affricate+consonant are only permitted across words, not within a word, with the exception of stop+liquid onset clusters; cf. Hualde 2003: 35). The presence of such vowel or consonant sequences serves as an indicator of word boundaries. In some cases, phonotactic restrictions may be negatively focalized, in the sense that specific consonants may be banned from appearing at the left or right edge of a word. In Yingkarta, laterals are not permitted wordinitially, and words cannot begin with rhotics in several languages, such as Tariana (cf. Dixon and Aikhenvald 2002: 15) and Basque (cf. Hualde 2003: 37, although the restriction is relaxed in present day). In German, a word cannot end in a short lax vowel (Hall 1999b). For phonotactic constraints as diagnoses of the word in Dutch, see Booij (1999). In general, the occurrences of such consonants or vowels indicate that there are no word-boundaries aligned with those consonants, i.e. that no word is starting or ending there.5 5

See Flack (2009) for an analytical treatment of phonotactic restrictions at word edges in Optimality Theoretic terms.

The Word in Phonology

11

Also, historical processes of word-initial strengthening and wordmedial lenition or weakening may end up shaping the phonotactics of a language so that obstruents are favored in word-initial position and sonorants are favored word-medially. One of the main observations in Prosodic Phonology is that certain phonological phenomena may be bound to the limits of the phonological word. For instance, Nespor and Vogel (1986, ch. 4) give examples of assimilation processes of different kind in several languages applying within words but not between words or even between a word and a clitic (cf. also Peperkamp 1997, Hall and Kleinhenz 1999, Dixon and Aikhenvald 2002: 17-18, Itô and Mester 2009, and Vigário 2010 for additional phenomena and recent discussion). Closely related would be non-stress or non-tonal suprasegmental processes, such as nasalization or vowel harmony, which spread over consonants or vowels within a phonological word, but not across words, as in Turkish, Hungarian, Finnish, Sundanese, Terena, etc. Altogether, prosodic domain-initial and domain-final strengthening, edge-located stress, prosodic word-bounded phonological rules and wordedge phonotactics serve a delimitative function (Trubetzkoy 1939), which aids word recognition and lexical acquisition or bootstrapping (cf. Christophe and Dupoux 1996). Indeed, in recent decades there has been increasing attention directed towards word segmentation in children and adults, and the results indicate that listeners make use of a number of phonotactic, allophonic, coarticulation and prosodic cues in word segmentation. Some references could be Mehler et al. (1981), Otake et al. (1983), Hirsh-Pasek et al. (1987), Cutler and Norris (1988), Friederici and Wessels (1993), Saffran, Aslin and Newport (1996), Saffran, Newport and Aslin (1996), McQueen (1998), Jusczyk, Houston and Newsome (1999), Jusczyk, Hohne and Bauman (1999), Smith and Hawkins (2000), Johnson and Jusczyk (2001), Bagou, Fougeron and Frauenfelder (2002), or Fernandes, Ventura and Kolinsky (2007). For a recent overview with many more references, see chapter 1 of De la Cruz (2012).

3. Phonology internal to the Word: Morphophonology and Lexical Phonology The surface form of a word may reveal the occurrence of phonological phenomena of various sorts affecting the features of the segments contained in the underlying form, such as assimilation, dissimilation, epenthesis, and deletion. These phenomena may occur in non-derived environments, with respect to an underlying form, or in derived

Chapter One

12

environments, when two or more morphemes combine to form a complex word. As for the latter, it has been a longstanding observation that morphemes in the same phonological context (the same “structural description”, in SPE-terminology) may display different behaviors with respect to a given phonological phenomenon. In Peninsular Spanish, for instance, there is a process of interdentalization of /t/ and /k/, by which these voiceless stops are spirantized to /ș/ (/s/ in Latin American Spanish and some southern varieties of Peninsular Spanish) when they occur in root-final position and are followed by a suffix beginning with /i/. But the process does not apply with all /i/-initial suffixes. With the superlative and the diminutive suffixes, for example, no interdentalization applies. The examples in (1a) (center column) illustrate cases of application of the process for the root-final consonants in the lefthand column, and those in (1b) (right-hand column) illustrate cases of nonapplication of the process: (1) místi/k/-o ‘mystic’

a. misti/ș/-ismo b. misti/k/-ísimo mystic-ism mystic-SUPERL

clási/k/-a ‘classic’

clasi/ș/-ismo classic-ism

clasi/k/-ísimo classic-SUPERL

opa/k/-o ‘opaque’

opa/ș/-idad opac-ity

opa/k/-ísimo opaque-SUPERL

decen/t/-e ‘decent’

decen/ș/-ia decent-cy

decen/t/-ísimo, decen/t/-illo decent-superl., decent-DIM

poten/t/-e ‘powerful’

poten/ș/-ia powerful-ness

poten/t/-ísimo, poten/t/-illo powerful-superl., powerful-DIM

Such differences in the application of phonological rules have been interpreted by Lexical Phonology as evidence that morphemes are organized in different levels or strata in the lexicon, and that phonological rules/processes are set to operate in different levels or strata. Namely, if a morpheme undergoes or causes a phonological process in a certain phonological context, and if another morpheme with the same phonological context does not undergo or cause the same process, lexical phonologists would assume that the two morphemes belong to different levels or strata (cf. Kiparsky 1982, 1985, Mohanan 1982, 1986, Halle and Mohanan 1985, Hargus and Kaisse 1993, Hanson and Inkelas 2009, among others). Phonological rules are specified to apply at level/stratum 1, 2, 3...n, the same way that morphemes are attached at levels or strata 1, 2, 3…n, and

The Word in Phonology

13

thus phonological rules operating at level 1, for instance, will only apply to morphemes attached at that level; phonological rules operating at level 2 will only apply to morphemes attached at level 2, and so on. There are two competing views on the interaction between morphology and phonology: the interactionist position claims that phonology and morphology go in tandem in each level, cyclically (cf. Mascaró 1976, Pesetsky 1979, Kiparsky 1982, 1985, Booij and Lieber 1993, among others), and the noninteractionist view claims that phonology only operates after all morphology has taken place (cf. Chomsky and Halle 1968, Halle and Mohanan 1985, Halle and Vergnaud 1987, Halle, Harris and Vergnaud 1991, among others). In an interactionist model, morphological bracketing indicating the morphological composition of an earlier level may not be available to the next level (i.e. bracket erasure), and in the noninteractionist model all morphological bracketing needs to be available when the phonology applies, since phonological rules are specified to apply at different levels. According to the interactionist view of Lexical Phonology, the derivation would proceed as follows: first, there would be the underived roots, on which certain structure-building operations apply (such as syllable structure or stress assignment); then, the resulting form is sent to Level 1 morphology and phonology, so that morphemes belonging to that level are added and the phonological operations specified at that level apply on the new structures created; then, the resulting form goes to Level 2 morphology and phonology, where morphemes of that level are added and the phonological rules specified for that level apply; and so on. The basic schema can be illustrated as in (2):6

6

It is still a matter of debate in Generative Linguistics whether word-formation takes place fully in the lexicon or fully in syntax, or whether morphology is split between derivational/compounding morphology and inflectional morphology, in the sense that the former takes place in the lexicon and the latter takes place in syntax or after syntax. For references, see Halle (1973), Aronoff (1976), DiSciullo and Williams (1987), Baker (1988), Hale and Keyser (1993), Halle and Marantz (1993), Noyer (1997), Embick and Noyer (2007), among others, and chapters 2 and 3 in this volume.

Chapter One

14 (2) MORPHOLOGY

PHONOLOGY

Underived roots

Underived roots

Level 1

Level 1

Level 2

Level 2

(…)

(…)

Level n

Level n

According to this view, most lexical phonological rules are cyclic (cf. the Strict Cycle Condition of Mascaró 1976), although there have been proposals that lexical rules can be noncyclic as well, especially those applying in the last morphological level of the word (i.e. the so called word-level, cf. Booij and Rubach 1987, Borowsky 1993). In Lexical Phonology, the application and underapplication/nonapplication of interdentalization in Spanish would be explained by assigning this phonological rule to the level or stratum in morphology where affixes such as -ismo, -idad and -ia are attached to the roots. If these morphemes were attached at, say, Level/Stratum 1, interdentalization would apply to these suffixes because it is specified to apply at this level, and only at this level. The underapplication or non-application of the process to the superlative and diminutive suffixes would be accounted for if these suffixes belonged to Level/Stratum 2 in morphology. The phonological rule of interdentalization would not apply to any morpheme in this level (cf. García-Bellido 1987 as well).7

7

The caveat of so-called “bracketing paradoxes” is in order here. In English, the word unhappier has the morphological bracketing [[un-happy]-er], as it can be paraphrased as ‘more unhappy’, rather than ‘not happier’, but the comparative suffix -er attaches to bisyllabic stems only, so it should attach to happy and not to unhappy. Another example is ungrammaticality, where un- must be assumed to attach first to grammatical, as the meaning is ‘the property of not being grammatical’, but the lexical morphology and phonology of English suggest that ity is a Level 1 affix and un- is a Level 2 affix. There have been alternative approaches to bracketing paradoxes, including Williams (1981), Selkirk (1982), Pesetsky (1985), Spencer (1988), Sproat (1988) and Beard (1991), among others. Limits of space prevent me from discussing these here, and the reader is referred to the sources.

The Word in Phonology

15

Lexical Phonology received contemporary resistance as well, as there were attempts to reinterpret affix classes in representational terms, by positing affixation to either roots or words (instead of Class I and Class II affixes, cf. Selkirk 1984) or by advocating the presence of prosodic constituents below the word level (cf. Inkelas 1990; see next section for a review of Prosodic Phonology at the word level). On the other hand, the existence of levels/strata and cyclicity imply the notion of derivation from an underlying representation to a surface representation going through cycles, hence the possibility of having intermediate levels of representation. The advent of Optimality Theory in the 90s brought the rejection of derivations and cycles. However, recent versions of OT acknowledge the impossibility of abandoning cycles and strata from phonology altogether, and some serial or derivational relationship between strata is assumed. For instance, in Bermúdez-Otero’s (2011, 2012, 2013) Stratal OT framework three strata are distinguished: stem-level, word-level and phrase-level, giving rise to at least three cycles. Finally, there have been recent proposals to rethink levels and strata based on the Phase Theory of syntactic derivations (Chomsky 2001). Marantz (2007), Embick and Marantz (2008) and Embick (2010) claim that the category-assigning heads n, v and a are phases that are spelled out, and hence any new addition of a categorial head constitutes a phase. Thus, there can be inner and outer phases, resembling the cycles or levels of Lexical Phonology, where allomorphy operates (see also Pak 2008: ch. 6). Marvin (2002) and DiSciullo (2004, 2005) call these “morphological phases”, and in this line of thinking, Samuels (2009, 2011, 2012) named this model “Phonological Derivation by Phase”. This is certainly an interesting and promising line of research that deserves to be explored further, as it relates cyclicity and morphophonology to the general functioning of grammar, in terms of phases and cyclic spell-out. The reader is referred to the mentioned sources and to chapter 3 (this volume) for discussion. Let us now turn our attention to aspects of phonological words having to do with their formation. First, I will briefly review conditions on phonological or prosodic word formation, such as minimum-size constraints. Then I will deal with the mapping from morphosyntax to phonology, concretely with the delimitation of prosodic words. Both aspects of phonological/prosodic word formation are areas of study of the subpart of phonology known as Prosodic Phonology.

16

Chapter One

4. Prosodic Word formation Prosodic Phonology is the part of phonology having to do with suprasegmental phenomena and phonological constituency above the segment, from the Mora to the Utterance. Prosodic Phonology is also the subarea of phonology that deals with the formation or derivation of prosodic constituents from morphosyntactic constituency following certain principles or conditions, such as alignment with the edges of morphosyntactic constituents (roots, prefixes, words, phrases, clauses, sentences). These prosodic constituents are the Phonological or Prosodic Word, the Phonological Phrase, the Intonational Phrase, and the Utterance (Nespor and Vogel 1986; Selkirk 1986; Itô and Mester 2009; Vigário 2010, among others). I will address the issue of the alignment of morphosyntactic and prosodic constituents in subsection 4.2, focusing on the Prosodic Word. But first let us briefly review some well-formedness conditions imposed on prosodic words.

4.1. Conditions on Prosodic Words First of all, there is an apparently universal condition on word stress, namely that a prosodic word can only have a maximum of one primary stress. Notice that this condition is not equivalent to stating that a prosodic word can only have one syllable that is stressed, since there are many languages in which words can have one or more syllables bearing secondary stress, apart from the syllable with primary stress. The observation is that the maximum number of primary stresses a word can have is one. Notice that this constraint should not also be taken as a minimum requirement on the number of stressed syllables in a prosodic word, i.e. as a condition stating that prosodic words must have at least one (primarily) stressed syllable, as there are languages where words are lexically unaccented (e.g. (Tokyo) Japanese and (Northern Bizkaian) Basque). Another widespread (although not universal) condition that prosodic words need to fulfill is minimum size. In many languages, a prosodic word needs to be minimally bisyllabic or bimoraic, that is, it needs to contain at least two syllables or two moras (cf. McCarthy and Prince 1986, 1990, 1993, Nespor and Vogel 1986, among others). Whether the restriction is imposed on syllables or moras is language-dependent (cf. Kenstowicz 1994: 640-646, Hayes 1995: 87-88, Downing 2006). Such restrictions are usually imposed on content or lexical words (nouns, adjectives, verbs, adverbs) but not on function words (e.g. prepositions, determiners,

The Word in Phonology

pronouns). The minimum schematically as in (3):

size

condition

17

could

be

represented

(3) Minimal word a. Bisyllabic (V.V)Ȧ b. Bimoraic (V)Ȧ = (C)VV , (C)VC P

P

In Spanish, for instance, lexical words must be minimally of the form (C)VX, where X can be C or V in a diphthong: pan ‘bread’, sol ‘sun’, luz ‘light’, ser ‘be’, ley ‘law’, pie ‘foot’, etc. CV lexical words are normally disallowed, with the exception of certain borrowings like té ‘tea’, and certain conjugated forms of verbs with the shape CVC in the infinitive, e.g. ser : sé ‘be, imperative’, ir : va ‘go, 3rd sg. pres.’, dar : dé ‘give, 1st/3rd sg. pres. subj’.8 On the other hand, function words can be monosyllabic with the shape (C)V: a ‘to’, de ‘of’, mi ‘my’, la ‘the, fem.’, que ‘that’, etc. McCarthy and Prince (1986) argued that the minimum size constraints for words arise from the fact that words contain feet, and normally feet do obey minimal size constraints, in the sense that they have to be minimally bimoraic or bisyllabic. Thus, word minimality, stated as a constraint calling for a minimally binary structure, stems from foot minimality conditions, demanding such minimally binary structure. For different languages, see Fitzpatrick-Cole (1990), Cabré (1993), Kenstowicz (1994), Leben and Ahoua (1997), Hall (1999b), Downing (1999) and Parker (1999), among many others. In fact, Hayes (1995) claims that word minimality constraints should really be understood as language-specific constraints on degenerate feet, composed of only one mora or one syllable. Namely, if a language shows minimal size restrictions on words it is because it does not allow degenerate feet; and conversely, if a language permits monosyllabic CV words (i.e. monomoraic and monosyllabic), it must be because the language does not ban degenerate feet. However, more recent work (cf. Garrett 1999, Gordon 1999, Downing 2006) shows that word minimality does not arise from foot size restrictions: there are languages which permit degenerate feet but impose minimum size 8 The verb haber ‘have’ has CV forms in the present indicative (he ‘1st sg.’, ha ‘3rd sg.’). However, this verb is not used nowadays as a lexical verb but as an auxiliary verb used to convey perfective aspect. On the other hand, the musical notes do, re, mi, fa, la, si also have the shape CV, but these can hardly be considered lexical words, in the sense that they do not denote objects, concepts or events.

Chapter One

18

restrictions on phonological words (e.g. Cahuilla); languages demanding minimally disyllabic words have moraic feet rather than syllabic feet (e.g. Carib, Hixkaryana); monosyllables with the shape CVC are accepted as minimal words in some languages where only CVV syllables count as heavy (e.g. Malayalam, Lardil); unbounded stress languages, which offer no evidence for foot structure, may also show word minimality conditions. The reader is referred to Garrett (1999) and Downing (2006: ch. 3) for alternative accounts of word minimality. Finally, it is important to remember that many languages allow words to be composed of a single CV syllable, that is, that word minimality is by no means a universal or dominant constraint on the world’s languages.

4.2. Correspondence of morphosyntactic terminal nodes and prosodic words 4.2.1. The level of the Prosodic Word in Prosodic Phonology Lexical items (nouns, adjectives, verbs, adverbs) are terminal nodes in syntax. Under the assumption of Generative Grammar that there is a mapping between syntactic structure and prosodic structure, the null theory should be to presume that the most transparent mapping between the two structures would be one in which each non-empty syntactic terminal node corresponded to a phonological or prosodic word. Hence: (4)

X0 Ÿ Ȧ

This straightforward relationship is illustrated in the example from Spanish in (5), where each of the syntactic heads or terminal nodes constitutes a prosodic word: (5)

C0 | que that

D0 | la the

N0 | mujer woman

V0 | volvió returned

P0 | a to

N0 | Bilbao Bilbao

Ÿ

(que)Ȧ (la)Ȧ (mujer)Ȧ (volvió)Ȧ (a)Ȧ (Bilbao)Ȧ

However, things are not so simple in the correspondence between syntactic heads and prosodic words. There are mismatches which blur such a direct mapping, and I will present some of them in what follows. Crosslinguistically, it is very common for function words (determiners, pronouns, complementizers, prepositions, conjunctions) not to carry stress.

The Word in Phonology

19

These syntactic heads do not constitute independent prosodic words, and lean on adjacent lexical heads to form prosodic words. Such function words are called “clitics”, and the prosodic leaning or attachment is called “cliticization”. In the case of (5) above, then, the mapping into prosodic words would be as in (6), where the complementizer que, the determiner la, and the preposition a cliticize onto the lexical words to their right:9 (6) C0 | que that

D0 | la the

N0 | mujer woman

V0 | volvió returned

P0 | a to

N0 | Bilbao Bilbao

Ÿ

(que la mujer)Ȧ (volvió)Ȧ (a Bilbao)Ȧ

In the eighties there arose two models of the interface between syntactic structure and prosodic structure: the Relation-Based Approach (RBA) (Nespor and Vogel 1982, 1986, Hayes 1984/1989) and the EndBased Approach (EBA) (Selkirk 1980, 1986, 1996, Chen 1987, Selkirk and Tateishi 1988, 1991, Selkirk and Shen 1990, inter alia). I reproduce here parts of the review of these theories in Elordieta (2008). The structure in (7) would be a schematic prosodic tree, according to the RBA. Notice that at each level there may be more than one constituent, symbolized by the parentheses. (The following abbreviations are used: U = Utterance; IP = Intonational Phrase; PPh = Phonological Phrase; CG = Clitic Group; PWd = Prosodic Word; Ft = Foot; Syl = Syllable).

9

A question worth considering is whether the cliticization of que and la onto the head noun mujer takes place in one swoop for both function words or iteratively, i.e. la attaches to mujer, creating a prosodic word, and que attaches to this prosodic word. For a discussion on European Portuguese, cf. Vigário (2003, chapters 5-6).

Chapter One

20 (7)

U IP

PPh

CG

Ft

Syl

(IP)

(PPh)

(CG) …

…

PPh

(PPh)

CG

(CG)

...

PWd (PWd)

…

PWd (PWd)

(Ft)

Ft

(Ft)

(Syl) ...

...

Syl

...

…

...

(Syl) ...

In this model, functional elements such as determiners, auxiliaries, possessive pronouns, and conjunctions are included in phonological words, clitic groups or phonological phrases with adjacent heads. That is, these elements may be either independent phonological words, which combine with other words to form phonological phrases, they may be clitics which attach to phonological words to create clitic groups, or they may be affixes which combine with a stem to form a phonological word. This is the typology of phonological patterns of so-called function words. For the sake of discussion, let us present the algorithms for the creation of phonological words and clitic groups that Nespor and Vogel propose: (8) Ȧ domain (Ȧ = phonological word) A.The domain of Ȧ is Q (Q = terminal element of a syntactic tree) or B.I. The domain of Ȧ consists of a. a stem; b. any element identified by specific phonological and/or morphological criteria c. any element marked with the diacritic [+W]. II. Any unattached elements within Q form part of the adjacent Ȧ closest to the stem; if no such Ȧ exists, they form a Ȧ on their own.

The Word in Phonology

21

Phonological words may thus be equal to or smaller than the terminal element in a syntactic tree (i.e. Q), as expressed by A and B in (8), respectively. Possibility A refers to phonological words which are composed of the stem and all affixes, or by the two members of a compound together (e.g. Greek, Latin; cf. Nespor and Vogel 1986: 110116, Nespor and Ralli 1996). Possibility BIa is exemplified by those cases in which only a stem plus affixes counts as a phonological word, that is, with each member of a compound word forming its own phonological word (e.g. Sanskrit, Turkish, Italian; cf. Nespor and Vogel 1986: 117-122, Nespor and Ralli 1996). 10 Both possibilities can coexist in the same language, although one option is always the least favored one (Nespor and Ralli 1996, Peperkamp 1997). Part Ib of possibility B refers to the distinctions that some languages show between prefixes and suffixes in terms of phonological word formation. In Hungarian and Italian prefixes are specified to form independent phonological words, as opposed to suffixes, which combine with the stem to form one phonological word (cf. Nespor and Vogel 1986: 122-134). Thus, there are affixes which form phonological words on their own by virtue of satisfying minimal prosodic size requirements such as bisyllabicity (e.g. Yidin; cf. Nespor and Vogel 1986: 134-136). Part Ic refers to affixes which are idiosyncratically specified to form independent words, as in Dutch (cf. Nespor and Vogel 1986: 136-140), hence the diacritic [+W]. Part II of possibility B responds to the necessity of obeying the Strict Layer Hypothesis (cf. Selkirk 1981, 1984, 1986, Nespor and Vogel 1986), which says that every level of prosodic constituency must exhaustively parse the entire segmental string. That is, every segment must be part of a phonological word, a phonological phrase, an intonational phrase, and an utterance. By this condition, elements that do not qualify as stems, such as conjunctions, complementizers and clitics, also form a Ȧ, either by attaching to a Ȧ within Q or by themselves. The End-Based Approach (EBA) has been the model that has been followed by most researchers in the area of the syntax-phonology interface. Selkirk’s (1986) original proposal is that the edges of phrases and heads are mapped onto prosodic structure as edges of prosodic constituents. This simple algorithm is part of Universal Grammar, and it is a directional parameter, so that in some languages prosodic edges are 10

Reiss (2003) offers a reanalysis of vowel harmony in Hungarian that renders superfluous the need to assume that each member of a compound constitutes an independent prosodic word, as traditionally argued in the literature. It could be that other cases could be reanalyzed the same way.

Chapter One

22

inserted at the left edges of XPs or X0s, and in other languages they are inserted at the right edges of these constituents. The parameter is represented in (9): (9) a. Left alignment: b. Right alignment:

[X0 X0]

Ÿ Ÿ

Ȧ[

]Ȧ

A crucial axiom in the EBA is the Principle of Invisibility of Functional Categories (Selkirk 1984), which states that functional categories (determiners, unstressed pronouns, complementizers, prepositions, conjunctions) do not project edges of prosodic constituents, unlike lexical categories (nouns, verbs, adjectives, adverbs). The following example from Shanghai Chinese (cf. Selkirk and Shen 1990) illustrates this principle. Prosodic word boundaries are inserted at the left edge of lexical X0s, as proven by the fact that pronouns, auxiliaries, prepositions, quantifiers and classifiers do not project prosodic edges and are hence included in one prosodic word with the lexical heads on their left. Syntactic structure is represented in square brackets on the left, above the words, and prosodic structure is represented on the right, in round brackets: (10) a. V [Pro Part.] Ÿ (V Pro Part.) taN ‘noN leq hit you part. ‘(s)he has hit you’ b. V [N Part.] Ÿ (V) (N Part.) taN ‘mo leq hit horse part ‘(s)he has hit the horse’ (11) a. V [P [N]SN ]SP Ÿ (V P) (N) ‘z ‘laq ‘zawNhe live in Shanghai ‘live in Shanghai’ b. V [Q Classif. N ]NP Ÿ (V Q Class.) (N) taw ?iq pe ‘zo pour a cup tea ‘pour a cup of tea’

Recently, Selkirk (2011) has argued that prosodic domains are not created simply by a parametric choice of left or right edges of X0s or XPs, but by a default mapping mechanism that matches X0s or XPs (including clauses) with prosodic words and phonological phrases, wrapping both left and right edges of these constituents (hence generalizing Truckenbrodt’s

The Word in Phonology

23

1995, 1999 Wrap-XP constraint). This is the Match Theory of the syntaxphonology interface, which would be based on principles like: (a) Match Clause (a clause in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it Ț, in phonological representation); (b) Match Phrase (a phrase in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it ij, in phonological representation); (c) Match Word (a word in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it Ȧ, in phonological representation). Phonological markedness constraints (such as prosodic weight or binarity considerations, rhythm) interplay with these general algorithms, and can impose deviations from the most transparent match between syntactic and prosodic constituency. Selkirk (2011) shows the advantages of Match Theory over the EBA, hence opening a promising new line of investigation that should be bound to produce interesting results. However, work within this framework has only focused on the match between syntactic and phonological phrases (Match Phrase), not on the match between syntactic and prosodic words, although research is expected to be carried out on this end as well. For the purposes of this chapter, then, I will proceed with a review of a sample of the most representative research on the prosody of words within the EBA. Selkirk (2011) suggests that many of the assumptions of most recent work within the EBA, such as recursivity of prosodic structure or the prosodic dependency of functions words (to be discussed below), are also to be adopted by Match Theory for the word level. 4.2.2. The prosody of function words (clitics) The EBA, applied in a simple manner, would not distinguish between affixes and unstressed functional syntactic heads, as both would be integrated in prosodic words together with adjacent lexical heads. Selkirk (1995) provides a representation of the different prosodic relationships between lexical heads and function words. She assumes that recursivity is allowed in prosodic structure (i.e. that a prosodic constituent of type p may dominate another node of type p), and a principle formerly known as Exhaustivity (i.e. that a prosodic constituent p must exhaustively dominate prosodic constituents of the immediately inferior level, p-1). According to Selkirk (1995), function words may: (a) integrate almost as affixes with an adjacent lexical head, in a prosodic word; (b) adjoin to the prosodic word that an adjacent lexical word constitutes (violating recursivity); (c) attach directly to the prosodic constituent higher than the Prosodic Word, attach

24

Chapter One

directly to the Phonological Phrase, the prosodic constituent higher than the Prosodic Word (violating Exhaustivity);11 or (d) constitute independent prosodic words, if focalized or pronounced in isolation. Thus, possibility (a) would represent a parallelism or similarity between function words and affixes. Possibility (b) would correspond to a cliticization of function words to the level of the Prosodic Word. Possibility (c) would amount to a cliticization to the level of the Phonological Phrase. Finally, possibility (d) would constitute a case of non-affixation and non-cliticization of function words. Selkirk (1995) coins the following terms for the function words integrated with lexical words in these four ways: (a) internal clitics, (b) affixal clitics, (c) free clitics, and (d) prosodic words, respectively. The four types of prosodizations of function words are represented schematically in (12a-d), where the function word precedes the lexical word. The function words in contexts (12a-c) have received the name “proclitics” in the literature. A function word that is focalized or pronounced in isolation will be stressed (will receive a pitch accent), and will hence be projected as a prosodic word (12d). The symbol Ȧ stands for a prosodic word, ij stands for a phonological phrase, and V stands for the syllables that the function words and lexical words are composed of (for the sake of simplicity, I assign one syllable to function words, as they are most commonly monosyllabic, and one or more syllables to lexical words, as they are most commonly polysyllabic). “Fnc” and “‘Lex” stand for function word and lexical word, respectively. (12’a-d) presents bracketed representations of the different prosodic structures in (12a-d). The schemes in (12) and (12’) are adapted from Selkirk (1995), with the omission of the Foot level.

11

This possibility is rejected by Itô and Mester (2009) in favor of adjunction to the Prosodic Word. Other scholars have proposed prosodic analyses with attachment of function words to the Phonological Phrase (cf. Hall 1999b or Kabak and Schiering 2006, for German).

The Word in Phonology

25

(12) a. internal clitics b. affixal clitics c. free clitics d. prosodic words ij

ij

Ȧ

Ȧ

V

V(V)

Fnc

Lex

V Fnc

ij V Ȧ

V(V)

Fnc

ij

Ȧ

Ȧ

Ȧ

V(V)

V

V(V)

Lex

Fnc

Lex

Lex (12’) a. ((Fnc Lex)Ȧ)ij b. ((Fnc (Lex)Ȧ)Ȧ)ij c. (Fnc (Lex)Ȧ)ij d.((Fnc)Ȧ(Lex)Ȧ)ij

For the purposes of this chapter, I will concentrate on structures (12ac)/(12’a-c) and will leave aside (12d)/(12’d), which does not constitute an interesting case of prosodic dependency on the lexical word. The prosodic structures in (12a-c)/(12’a-c) are illustrated by different dialects of SerboCroatian (cf. Selkirk 1995, citing Zec 1993). In Serbo-Croatian, lexically unaccented words receive a high pitch accent on its first mora when preceded by a lexical word in the same phrase. However, when a function word precedes the unaccented word, three possibilities arise, depending on the dialect. They are exemplified in (13a-c), for the sequence formed by the preposition u ‘to’ and the unaccented words graad ‘city’ and glaavu ‘head’. In Eastern Herzegovina (13a), the accent is inserted on the function word (the preposition u). In Šrem/Maþva, the accent may be inserted on the function word, or it may fall on the initial mora of the lexical word, after which it spreads onto the function word. Finally, in Belgrade the accent falls on the initial mora of the lexical word, without spreading onto the function word. (13) a. Eastern Herzegovina: ú grad ‘to the city’ b. Šrem, Maþva: ú glaavu/ú gláavu ‘into (the) head’ c. Belgrade: u gráad ‘to the city’

26

Chapter One

The facts are accounted for under the assumption that accents are associated to the first mora of a Ȧ, and that function words are internal clitics in Eastern Herzegovina ((14a)), affixed clitics in Šrem/Maþva ((14b)), and free clitics in Belgrade ((14c)).12 (14) a. Eastern Herzegovina: ((ú graad)Ȧ)ij b. Šrem, Maþva: ((ú (graavu)Ȧ)Ȧ)ij / ((ú (gráavu)Ȧ)Ȧ)ij c. Belgrade: (u (gráad)Ȧ)ij

Function words may also integrate with lexical words that precede them, with similar prosodic structures to those observed for function words integrated with lexical words following them. Stressless function words that are attached in different forms to lexical words are called enclitics. Like proclitics, they can be internal clitics, affixed clitics, and free clitics, apart from independent prosodic words, of course. The structures are represented in (15) and (15’):

12

The spreading of pitch accents to the preceding moras in a prosodic word appears to be a general phenomenon in Serbo-Croatian. Thus, the second pattern of the Šrem/Maþva dialects must be explained by assuming that an accent is assigned to the first mora of the most embedded Ȧ and that the accent spreads onto the preceding mora in its own Ȧ, this being the recursive, outer Ȧ (cf. Selkirk 1995: 462). In Ito and Mester’s (2009) view, according to which attachment of function words to ij is rejected and only adjunction to Ȧ is allowed, it is not trivial to know how one would account for the difference between the patterns in (13b) and (13c). If both had the same prosodic structure of Ȧ-adjunction, namely (14b), the difference would have to be explained by assuming that in Šrem/Maþva it is either the top Ȧ -level (Ȧ-max, in their model) or both the top and bottom levels that are the domain of accent assignment, whereas in Belgrade it is only the bottom level that is the domain of accent assignment. The implications of this interpretation escape my objectives in this chapter.

The Word in Phonology

27

(15) a. internal clitics b. affixal clitics c. free clitics d. prosodic words ij

ij

Ȧ

Ȧ

V(V) Lex

V Fnc

Ȧ V(V)

ij

Ȧ V

Fnc

Lex

ij V

Ȧ

Ȧ

V(V)

V(V)

V

Fnc

Lex

Fnc

Lex (15’) a. ((Lex Fnc)Ȧ)ij b. (((Lex)Ȧ Fnc)Ȧ)ij c. ((Lex)Ȧ Fnc)ij d.((Lex)Ȧ(Fnc)Ȧ)ij

Like for proclitics, I discuss structures (15a-c)/(15’a-c), leaving aside (15d)/(15’d). Possibilities (15a-c)/(15’a-c) are illustrated by postverbal (direct and indirect) object pronouns in the Lucanian, Neapolitan and Standard dialects of Italian (Peperkamp 1997). In Lucanian, postverbal pronouns form a prosodic word with the verb they attach to, as evidenced by the main stress assigned to the penultimate syllable of the whole string formed by the verb+pronouns (cf. (16a)). In Neapolitan, stress is assigned to the penultimate syllable of the verb and to the penultimate syllable of the string formed by the postverbal pronouns (cf. (16b)). In Standard Italian, stress is assigned to the penultimate syllable of the verb (cf. (16c)): (16) a. Lucanian b. Neapolitan c. Standard Italian

man:atΩ-mí-llΩ send-me-it cónta-tí-llo tell-you-it pórta-me-lo bring-me-it

‘send it to me’ ‘tell it to you’ ‘bring it to me’

Bearing in mind that the most basic stress pattern in Italian assigns stress to the penultimate syllable of a Ȧ, the facts can be explained straightforwardly. 13 The string formed by a verb and the pronouns following it would form a single prosodic word in Lucanian, i.e. 13

The analysis provided here is a slight adaptation of Peperkamp’s (1997), with the only difference that it follows more closely Selkirk’s (1995) general proposal for the prosodization of function words.

28

Chapter One

postverbal pronouns are internal clitics in this dialect (cf. (17a)), hence the whole string is one domain for stress assignment. In Neapolitan, the pronouns are adjoined to the prosodic word projected by the verb, i.e. postverbal pronouns are affixed clitics (cf. (17b)), hence the verb gets one stress and the whole (recursive) Ȧ gets another stress. In Standard Italian, it could be that the postverbal pronouns are attached directly to the phonological phrase, i.e. they are free clitics (cf. (17c)). That is why only the verb receives stress; the clitics are not in a Ȧ and thus do not need to be stressed:14 (17) a. Lucanian b. Neapolitan c. Standard Italian

(man:atԥ míllԥ)Ȧ ((cónta)Ȧ tíllo)Ȧ ((pórta)Ȧ melo)ij

For more cases in other languages, the reader could refer to Booij (1996), Peperkamp (1997), Hall (1999b), and Vigário (2003), among others. Let us then have a quick look at the behavior of proclitics and enclitics in Spanish. Let us start with proclitics, such as determiners, preverbal object pronouns, or possessive pronouns: (18) a. la casa the house b. lo vimos it we-saw c. mis amigos my friends

Since the determiners, preverbal object pronouns, and possessive pronouns in (18) are unstressed, they do not constitute independent prosodic words. Stress is assigned to the final, penultimate or antepenultimate syllable

14 Similar considerations to those discussed in footnote 12 would hold here as well. If ij-attachment is avoided, the patterns in (17b) and (17c) would need to be distinguished by assuming that stress assignment in Neapolitan (17b) applies at both Ȧ-min and Ȧ-max levels, whereas in Standard Italian it applies at the lower, Ȧ-min level only (Ȧ-max would correspond to the higher level, once the clitics are adjoined). Or, alternatively, it would have to be assumed that the postverbal clitics in Neapolitan form one Ȧ which is joined with the Ȧ of the verb to form a compound Ȧ. Ito and Mester (2009) actually argue that polysyllabic function words in English and German are Ȧs that combine with lexical Ȧs to form recursive (i.e. compound) Ȧs.

The Word in Phonology

29

of the lexical word (usually the penultimate). 15 Although the stress facts would not inform us whether Spanish proclitics are internal, affixed or free clitics, there is some reason to believe that they are not internal clitics. According to Harris (1983), there is a phonological rule operating in Spanish that converts root-initial flaps into trills in word-initial position. The working of this rule can be seen in the alternations in (19), where the root-initial flaps are trills word-initially (see also Peperkamp 1997: 96-97): (19) a. Root: /ᢪect-/ b. Root: /ᢪeg-/ c. Root: /ᢪupt-/

di-[ᢪ]ect-o direct (noun) di-[ᢪ]ig-ir direct (verb) e-[ᢪ]upc-ión eruption

[r]ect-o straight [r]eg-ir rule (verb) [r]upt-ura rupture

Thus, underlying flaps turn into trills at the left edge of a prosodic word. The rule would be formalized as in (20), adapting it slightly from Harris (1983): (20) /ᢪ/ > [r] / Ȧ(__

Rhotic-initial lexical words following proclitics in Spanish are pronounced with trills rather than flaps (cf. (21)). The appearance of trills would suggest that there is a Ȧ-boundary at the left edge of the lexical words, and hence that proclitics in Spanish are either affixed clitics or free clitics, but not internal clitics (cf. (22)). Similar structures would have to be posited for the sequences in (18): (21) a. la [r]ecta the straight ‘the straight line / the straight one’ b. los [r]igió them (s)he-ruled ‘(s)he ruled them’ c. tu [r]uptura your rupture 15

See Hualde (2006, 2009) for more data on the stresslessness of function words in Spanish. Hualde (2006) argues that function words are not inherently stressless and that there is stress deletion when function words appear unstressed. They only appear stressed in surface when they occur in phrase-final position. Hualde (2009) makes this rule of stress deletion extensive to all word-level structures, including derivation and compounding.

Chapter One

30 (22) a. ((la (recta)Ȧ)Ȧ)ij or b. ((los (rigió)Ȧ)Ȧ)ij or c. ((tu (ruptura)Ȧ)Ȧ)ij or

(la (recta)Ȧ)ij (los (rigió)Ȧ)ij (tu (ruptura)Ȧ)ij

Another piece of evidence comes from the application of e-epenthesis word-initially. It is a well-known fact of Spanish that an epenthetic /e/ is inserted at the left edge of stems starting with the sequence formed by /s/ and another consonant, when the root occurs at the left edge of a word. Hence, compare the pairs of words in (23a,c,e), with e-epenthesis, with those in (23b,d,f), where the sequence SC in word-internal position is pronounced with s as the coda of a syllable and C as the onset of the following syllable: (23) a. eslavo ‘Slavic’ c. esfera ‘sphere’ e. estático ‘static’

b. yugoslavo ‘Yugoslavian’ d. estratosfera ‘stratosphere’ f. aerostático ‘aerostatic’

When a proclitic precedes a stem starting with sC, there is eepenthesis, thus showing that there is a Ȧ-boundary between the proclitic and the stem. The proclitic ends in a vowel, so as to permit in principle the syllabification of s as the coda of the proclitic and the following C as the onset of the following syllable: (24) a. la eslava ‘the Slavic (one), fem.’ b. tu esfera ‘your sphere’ c. la estática ‘the static (one), fem.’

As for enclitics, it seems clearer that they are free clitics, as suggested by the fact that stress assignment is calculated on the basis of the domain of the lexical head, leaving out the enclitics. Enclitics are found with imperatives, as in (24a-b), or nonfinite forms such as the infinitive or the gerund, as in (25c):

The Word in Phonology

31

(25) a. vénde-lo sell-it b. cómpra-se-los buy-to.them-them ‘buy them from them’ c. leyéndo-te-la reading-you-it ‘reading it to you’

These stress facts can be explained if we assume that only the lexical head projects a prosodic word, leaving the clitics out. Thus, the postverbal personal pronouns would be attached directly to the level of the Phonological Phrase, ij, as in Standard Italian and unlike European Portuguese, where they are internal clitics (Vigário 2003).16 (26) ((vénde)Ȧlo)ij, ((cómpra)Ȧselos)ij, ((leyéndo)Ȧtela)ij

The available evidence in Spanish, then, suggests that in this language proclitics at least allow for a possible analysis in which they would be adjoined to the Ȧ of the lexical word, creating a recursive Ȧ, and that enclitics would be adjoined to ij (the other possible analysis for proclitics is that they are adjoined to ij as well). Clitics associate with lexical heads to their left or to their right. The direction of association is usually coherent with the syntactic relationship between the clitics and the lexical heads. That is, clitics will usually attach to the lexical head with which they form a syntactic phrase. There is a substantial amount of literature on the topic, which cannot be reviewed here for limits of space. The reader is referred to the work mentioned in the whole of this section. In some cases, there might be a preference of association which is independent of syntax. In fact, Klavans (1982) claims that all clitics are specified for direction of attachment. For instance, Klavans (1985) claims that clitics in Nganhacara attach to the lexical head preceding them in a separate syntactic phrase, not with the lexical head following them and with which they share a syntactic phrase. In this language, then, there would only be enclitics. It would be important to know how frequent are the divergences between the syntactic associations and the prosodic associations of clitics and lexical heads. In principle, such 16

The same observation expressed in note 14 for Standard Italian would apply for these cases from Spanish. Namely, if these constructions involved Ȧ-adjunction only, one would have to assume that stress assignment holds at the Ȧ-min level only, that is, the bottom Ȧ-level.

32

Chapter One

mismatches would not seem advantageous for the processing of a sentence. It is important to point out that the model advocated by Selkirk (1995) blurs a categorical distinction between affixes and clitics at the prosodic level. In this model, there would be no differences in prosodic word formation between internal clitics ((12a), (15a)) and affixes, as both would be integrated closely and directly with adjacent lexical heads in a prosodic word. There are no prosodic word boundaries (i.e. Ȧ-boundaries) between function words and lexical words in the case of internal clitics, and hence the close degree of prosodic cohesion between function words and lexical words in these cases would resemble that of affixes and roots or stems. The difference between affixes and internal clitics would thus be a morphosyntactic one, not a prosodic one. There is substantial literature on the morphosyntactic differences between affixes and clitics, but a review or discussion of these properties exceeds the goals of this chapter. The interested reader is referred to Zwicky (1977, 1985), Klavans (1982, 1985), Zwicky and Pullum (1983), Nevis (1988), Spencer (1991), Anderson (1992, 2005), Halpern (1995), Halpern and Zwicky (1996), Everett (1996), van der Leeuw (1997), Gerlach and Grijzenhout (2001), Spencer and Luís (2012), among others, and to references mentioned therein. The basic distinction between affixes and clitics is that affixes have a closer degree of morphological cohesion with roots and stems they attach to, and the evidence for this conclusion comes from facts such as the high degree of morphological and lexical selection exhibited by affixes as compared to clitics, the existence of morphophonological and semantic idiosyncrasies in stem+affix combinations but not between a clitic and its host, or that clitics are always attached at the edges of the combination of a stem and an affix (i.e. a suffix cannot appear following an enclitic, and a prefix cannot precede a proclitic). Some researchers have advocated the existence of a prosodic constituent between the level of the Prosodic Word and the Phonological Phrase: the Clitic Group. This constituent would contain a prosodic word and one or more clitics (cf. Nespor and Vogel 1986, Hayes 1984/1989, Cohn 1989, Vogel 1990, Vogel and Spinu 2007). However, the necessity of this constituent has been rejected by several authors (Zec 1988, 1993, Inkelas 1990, Selkirk 1995, Booij 1996, Vigário 2003, among others), based on the observation that there are no phonological phenomena or phonological patterns that only apply at the level of the Clitic Group, without also applying at the level of the Prosodic Word. Hence, the prosodic attachment of clitics can be accounted for by making reference to the level of the Prosodic Word and the Phonological Phrase, by assuming

The Word in Phonology

33

different patterns of attachment of clitics to lexical words, as expressed in (12)-(12’), (15)-(15’), (17), (22), and (26). 4.2.3. The prosody of affixes Prefixes and suffixes can also display different degrees of prosodic cohesion with the root or stem they attach to. The different types of integration between roots/stems and prefixes, for instance, could be represented in (27)/(27’), with identical structures to those of proclitics (cf. (12)/(12’)). The possible prosodic structures of suffixes are represented in (28)/(28’), with a parallelism with enclitics.17 (27) a. internal pref. b. Ȧ-adjoined pref c. ij-attached pref d. prosodic words ij ij ij ij

Ȧ V

V(V)

Pref Stem

V

Ȧ V Pref

Ȧ V(V)

Ȧ

Ȧ

Ȧ

V(V)

V

V(V)

Stem

Pref

Pref

Stem

Stem (27’) a. ((Pref Stem)Ȧ)ij c. (Pref (Stem)Ȧ)ij

17

b. ((Pref (Stem)Ȧ)Ȧ)ij d. ((Pref)Ȧ(Stem)Ȧ)ij

I will not draw a complete parallelism with proclitics and enclitics, however, I will refrain from using the terms “internal prefix/suffix”, “affixal prefix/suffix”, and “free prefix/suffix”. These terms have not been used in the literature, and the term “affixal prefix/suffix” sounds completely redundant.

Chapter One

34

(28) a. internal suff b. Ȧ-adjoined suff c. ij-attached suff d. prosodic words ij ij ij ij

Ȧ V(V)

V

Ȧ V

Stem Suff

Ȧ V(V)

V

V(V)

Suff

Stem

Ȧ

Suff

Ȧ

Ȧ

V(V)

V

Stem

Suff

Stem (28’) a. ((Stem Suff)Ȧ)ij c. ((Stem)Ȧ Suff)ij

b. (((Stem)Ȧ Suff)Ȧ)ij d. ((Stem)Ȧ(Suff)Ȧ)ij

4.2.3.1. Prefixes An interesting observation holding in different languages is that there may be two types of prefixes, those that are integrated prosodically with the base they attach morphologically to, and those that have a more independent status from their morphological base. The former would be analyzed as having the prosodic structure in (27a)/(27’a), and the latter would have one of the structures in (27b-d)/(27’b-d). In European Portuguese, for instance, the words in (29) with the prefixes in- and desattached in a synchronically opaque, non-compositional way allow for the first vowel of the stem to reduce, following the process of European Portuguese that reduces unstressed vowels. However, in the words in (30) it is not possible to find reduced vowels stem-initially (Vigário 2003: 170173, 228-229). It is relevant to point out that the prefixes in (30) are stressed:18 (29) a. pos-pôr (underlying /‫ܧ‬/ in initial syllable pronounced as [u]) ‘to postpone’ b. pro-mover (underlying /‫ܧ‬/ in initial syllable pronounced as [u]) ‘to promote’ c. pre-ver (underlying /‫ܭ‬/ in initial syllable pronounced as [ܺ]) ‘to foresee’

18 For ease of reading, I introduce morpheme boundaries in all the examples involving affixed and compound words (see section 4.2.4), which should not be taken as literal orthographic transcriptions.

The Word in Phonology

35

(30) a. pré-accentual (maintenance of underlying /‫ܭ‬/ in initial syllable) ‘pre-stressed’ b. pró-comunista (maintenance of underlying /‫ܧ‬/ in initial syllable) ‘procommunist’

Polysyllabic prefixes (usually disyllabic) are also non-cohering. They are stressed and do not present vowel reduction or other processes typical of weak, unstressed syllables. These are prefixes like poli-, super-, extra-, mega-, or inter-. Vigário (2003: 229-232) argues that polysyllabic prefixes and stressed monosyllabic prefixes like those in (30) should be analysed as independent Ȧs: (pré)Ȧ (accentual)Ȧ; (pró)Ȧ (comunista)Ȧ; (inter)Ȧ (cultural)Ȧ; (mega)Ȧ (concerto)Ȧ. Thus, unstressed monosyllabic prefixes as in (29) would present the prosodic structure in (27a)/(27’a), and stressed monosyllabic prefixes and polysyllabic prefixes would be represented with a similar structure to (27d)/(27’d). Vigário (2003: 218-232) presents evidence that the Ȧs formed by the prefixes and the stems in the latter cases are not immediately dominated by a ij, but by a prosodic constituent Ȧmax (which in Vigário 2010 is renamed as “Prosodic Word Group”, PWG). That is, although there are two Ȧs, the two Ȧs behave as if they are a single Ȧ. For instance, the second Ȧ is the one bearing main prosodic prominence, regardless of its position in a ij or regardless of which of the two Ȧs is under narrow focus. Unlike non-final Ȧs in a ij, the prefix cannot have the most prominent accent in the ij under narrow focus; rather, it falls on the rightmost stressed syllable, i.e. the second Ȧ. Another piece of evidence in favor of a Ȧmax or PWG constituent is that it is the domain of non-back vowel deletion (cf. Vigário 2003: 225). For limits of space, I will not get into details and illustrating examples, and will hence refer the reader to Vigário (2003, 2010). Similarly, in Italian there are cohering and non-cohering prefixes. As Peperkamp (1997: 70-72) shows, monosyllabic prefixes are unstressed, except ex-, which can be considered more like the first member of a compound (see Peperkamp 1997: 73, 83): de-ambulare ‘to walk around’, pre-avviso ‘pre-announcement’, ri-flettere ‘to think’, ri-guardare ‘to concern’, ri-piegare ‘to make do with’, pre-visioni ‘forecast’. Unlike European Portuguese, when these prefixes are used in a synchronically transparent and compositional way, they are still unstressed. This can be seen with semantically transparent versions of some of the words above: ri-flettere ‘to bend again’, ri-guardare ‘to look again’, ri-piegare ‘to fold again’, pre-visioni ‘preview’. Polysyllabic prefixes, on the other hand, are lexically stressed: extra-coniugale ‘extramarital’, super-veloce ‘superfast’, arci-vescovo ‘archbishop’, inter-disciplinare ‘interdisciplinary’. That these

36

Chapter One

prefixes bear stress is shown by the possibility of having lengthened nonfinal open syllables, which is a correlate of stress in Italian, e.g. s[úӃ]pervel[óӃ]ce, and by the possibility of presenting a lax mid vowel [‫]ܭ‬, as in [ѓ]xtra-coniugale, [ѓ]xtra-parlamentare ‘extraparliamentary’. Peperkamp (1997) argues that monosyllabic prefixes in Italian are incorporated with the stem in a single Ȧ, whereas disyllabic prefixes form independent Ȧs. That is, monosyllabic prefixes would have the prosodic structure in (27a)/(27’a), and disyllabic prefixes would have the prosodic structure in (27d)/(27’d). In Spanish, like in European Portuguese, there is evidence that some prefixes at least are non-cohering, i.e. they are not integrated in the same Ȧ with the stem. To start with, consonant-final prefixes may give rise to heterosyllabic consonant sequences that are not found morphemeinternally or between a root/stem and a derivative or inflectional suffix, e.g. ab-negado ‘unselfish/sacrificed’, con-llevar ‘to imply’, ad-judicar ‘to award’, sub-yugar ‘to subdue’ (cf. Pensado 1999: 4452). On the other hand, the syllabification of consonant-final prefixes when followed by vowel- or liquid-initial stems constitutes evidence that the prefixes are not cohering. Consider the following words with the prefix sub-, for instance. A syllabification in which the final consonant of this prefix forms the onset of a vowel-initial stem is accepted by native speakers, but it is more marked than one in which the final b constitutes the coda of the syllable of the prefix (cf. (31)). This pattern can be observed as well in cases with a liquid-initial stem; although b+liquid onset clusters are perfectly well formed in Spanish, b can be syllabified as the coda. In fact, for some words the alternative with the complex onset seems even more marked or unacceptable (cf. (32b-c)):19 19 Pensado (1999: 4452) claims that sub- is always syllabified with b in the coda, but modern pronunciation of some words containing this prefix reveal more flexibility. Whether the prefix-final b is syllabified as an onset of the vowel- or liquid-initial stem seems to be lexically determined, most likely subject to the degree of semantic compositionality or transparency for native speakers. Hence, for words which were already prefixed in Latin, whose original compositional meaning is blurred in current Spanish and are of common use it is more acceptable to have complex onsets, and in fact it may be the only acceptable syllabification: su.bas.tar ‘to auction’ (< Lat. sub-hastare), su.bal.ter.no ‘subordinate’ (< Lat. subalternus), su.bor.di.nar ‘to subordinate’ (< Lat. sub-ordinare), su.bur.bio ‘suburb’ (< Lat. sub-úrbium), su.ble.var ‘to revolt’ (< Lat. sub-levare), su.bli.me ‘sublime’ (< Lat. sub-limis). Words that already had the prefix sub- in Latin but are not of common use, like sub.ro.gar (< Lat. sub-rogare) (cf. (32b)) show a clear tendency towards having b as a coda, not an onset. An interesting case is that of suburbano, which is syllabified as sub.ur.ba.no with the compositional meaning referred to the

The Word in Phonology (31) a. sub-acuático ‘sub-aquatic’ b. sub-índice ‘sub-index’ c. sub-oficial ‘sub-officer’ (32) a. sub-rayar ‘to underline’ b. sub-rogar ‘to substitute’ c. sub-lingual ‘sub-lingual’

sub.a.cuá.ti.co

?su.ba.cuá.ti.co

sub.ín.di.ce

?su.bín.di.ce

sub.o.fi.cial

?su.bo.fi.cial

sub.rayar

?su.bra.yar

sub.ro.gar

*su.bro.gar

sub.lin.gual

*su.blin.gual

37

A second piece of evidence of non-cohering prefixes in Spanish comes from /s/- aspiration and /n/-velarization. In southern Spanish and several dialects of Latin American Spanish /s/ is aspirated in coda position, e.g. la[h] cosa[h] ‘the things’ (cf. la[s] cosa[s] in Standard Peninsular Spanish and non-aspirating dialects); de[h]preciar ‘to despise’ (cf. de[s]-preciar in Standard Peninsular Spanish and non-aspirating dialects). Interestingly, Harris (1993: 183-184) cites one example from these varieties containing the prefix des- before a vowel-initial stem in which /s/ is aspirated even though this consonant appears to be an onset of the following syllable (cf. (33a)). Notice the contrast with the monomorphemic word in (33b), with a non-aspirated /s/. (33) a. des-echo ‘waste’ b. deseo ‘desire’

de.[h]e.cho de.se.o

In aspirating varieties, /n/ is velarized in coda position, word-internally or word-finally: lento [ɇleƾto], camión [kaಬmjoƾ]. According to Harris (1993: 182), similar to the examples with des-, there is velarization of /n/ adjective ‘urban’, but as su.bur.ba.no with the non-compositional meaning of ‘metropolitan bus’, where the adjective is used as a noun. Non-technical neologisms like subrayar ‘to underline’ and subestimar present more balanced preferences for either syllabification. Certainly, the syllabification tendencies of consonant-final prefixes in Spanish deserve to be explored in more detail, with judgments from as many speakers as possible, from different dialectal areas. But the distinction between transparent or compositional meaning vs. opaque or noncompositional meaning is an important one, as the behavior of prefixes with respect to vowel epenthesis will show further below (see note 17).

38

Chapter One

of the prefix en- even if this consonant appears to be the onset of the following vowel-initial syllable of the base. The example Harris (1993) gives is the one in (34a), to be contrasted with a monomorphemic word like the one in (34b): (34) a. en-amora ‘(s)he enamors’ b. enano ‘dwarf’

e.[ƾ]a.mo.ra e.[n]a.no

Although the speakers from aspirating and velarizing dialects I have consulted (from the Canary Islands and the Dominican Republic) do not agree with the data in (33a) and (34a) (i.e. they neither aspirate nor velarize prefix-final consonants), the facts reported by Harris (1993) would provide an interesting point in favor of the relative independence of prefixes from stems. Peperkamp (1997: 100-104) argues that prefixes are syllabified independently of the base at the lexical level or stratum, and that resyllabification applies at a postlexical stratum. A third piece of evidence of the existence of non-cohering prefixes in Spanish comes from the process of e-epenthesis at the beginning of roots or stems beginning with the cluster sC, as shown above in section 4.2.2 (cf. the pairs in (23): eslavo ~ yugoslavo; esfera ~ estratosfera; estático ~ aerostático). Peperkamp (1997: 89) provides examples in which Spanish prefixes are followed by stems that apparently have undergone eepenthesis from Latin to Spanish, as they were originally sC-initial in Latin. What these examples suggest is that the stem-initial sC clusters are treated as word-initial sequences:20 (35) a. in-estable ‘unstable’ b. sub-especie ‘subspecies’ c. super-estrella ‘superstar’

The final phenomenon showing that Spanish has non-cohering prefixes is Ѷ-strengthening in word-initial position, that is, the process that turns /‫ݐ‬/ 20

Here too, the degree of semantic compositionality or transparency in the combination of prefix and base is relevant in order to account for cases with no epenthesis of e, like in-spirar, in-scribir or in-struir. Peperkamp (1999: 89-94) discusses some examples like these, and argues that in these cases the prefixes are cohering, i.e. integrated in the same Ȧ with the stem.

The Word in Phonology

39

into /r/ in word-initial position. Several examples were provided in (19), in section 4.2.2: di[ᢪ]ecto ~ [r]ecto; di[ᢪ]igir ~ [r]egir; e[ᢪ]upción ~ [r]uptura. Prefixed words with base-initial trills suggest that the prefixes are noncohering: (36) a. semi-[r]ecto ‘angle of 45º’ b. co-[r]egir ‘to correct’ c. pre-[r]uptura ‘pre-rupture’

(from Peperkamp 1997: 97)

Like monosyllabic prefixes, polysyllabic prefixes like super-, mega-, hiper-, multi-, poli-, micro-, macro-, extra-, semi-, anti-, and archi- in Spanish are also non-cohering, as evidenced by (35c) and (36a). Of the four possible prosodic structures for prefixes in (27), the one that would represent more appropriately the prosodization of Spanish noncohering prefixes would be (27b)/(27’b), that is, a structure in which prefixes adjoin to the Ȧ of the base, forming a recursive Ȧ. Both monosyllabic and polysyllabic prefixes are stressless, which suggests that they cannot be independent Ȧs, hence (27d)/(27’d) cannot be the right prosodic structure for prefixes in Spanish. (27a)/(27’a) cannot be the correct structure, either, since prefixes do not show a close degree of cohesion with the base. Finally, (27c)/(27’c) has to be discarded as well, as there is evidence that prefixes form part of a Ȧ and are not directly incorporated to ij: in Spanish, emphatic stress found in political and public discourse falls on the initial syllable of a Ȧ, apart from regular stress (e.g. cónstitución ‘constitution’, cándidatúra ‘candidacy’, prófesionál ‘professional’), and prefixes may surface with emphatic stress (e.g. réformulación ‘reformulation’, dés-propósito ‘absurdity’, ím-posibilidád ‘impossibility’). Hence, the prosodic structure that would represent the prosodization of (productive, transparent, compositional) prefixes in Spanish would be: (37) ((Pref (Stem)Ȧ)Ȧ)ij

Peperkamp (1997: 104) argues that the prefix is outside the prosodic domain which contains the stem, that is, a Ȧ, and that syllabification applies in that Ȧ, excluding the prefix, even if the Ȧ starts with a vowel. That is why aspiration and velarization may apply to /s/ and /n/ in some Spanish dialects. Later, at a postlexical stratum, resyllabification applies

40

Chapter One

and makes the final consonant of the prefix appear as the onset of the following syllable. 4.2.3.2 Suffixes It has been a long standing cross-linguistic observation in prosodic morphology and phonology that suffixes are usually closely integrated with the stem, in the sense that it is much less frequent to find heterosyllabic consonant clusters at the border between stem and affix that are not permitted morpheme-internally, or stem- or root-final consonants that are syllabified as codas in the presence of a vowel-initial suffix. This is certainly true for Romance languages. In Germanic languages there are non-cohering suffixes. In English, for instance, -hood, -ness, -less, -ful, -ly, -like or -ship, among others, can appear in words like child.hood, kind.ness, spot.less, cup.ful, abrupt.ly, dumb.like, or friend.ship, where the heterosyllabic clusters are disallowed morpheme-internally. They are also stress-neutral, using the terminology of classical lexical phonology, as they do not give rise to stress shift after having added one more syllable (e.g. páin-páinful-páinfulness, chíld-chíldless-chíldlessness). These are socalled Class-II suffixes (Siegel 1974, Allen 1979, Selkirk 1982, Kiparsky 1982, Halle and Mohanan 1985, Giegerich 1999, among others). On the other hand, interestingly, cohering suffixes are all vowel-initial, so there are no issues with consonant clusters (-ity, -ion, -al, -ous, -ic, -ate, -ive, among others). They are also stress-shifting (dráma-dramátic, grámmargrammátical-grammaticálity). In other words, Class-I suffixes show a closer degree of cohesion with the stem than Class-II suffixes, as if they clearly formed one Ȧ. Thus, Class-I suffixes could have the prosodic structure in (28a)/(28’a), whereas Class-II suffixes could be represented as in (28b)/(28’b) or (28c)/(28’c). The fact that these suffixes have selectional restrictions regarding the category of the base to which they attach and that they are category-changing suggests their containment in a (recursive) Ȧ with the base they attach to (that is, the structure in (28b)-(28’b)), rather than their attachment to a ij without a relation with the base, as in (28c)(28’c). In any case, neither Class-II suffixes nor Class-I prefixes bear lexical stress, which suggests that these suffixes are not independent Ȧs, and hence (28d)/(28’d) would be excluded. As for Spanish, it seems clear that suffixes are cohering. There are several facts suggesting this. First, their phonological shape is quite telling: most derivational and inflectional suffixes start with a vowel if the root or stem ends in a consonant and with a consonant if the root or stem ends in a vowel. Cases with V+C or C+C sequences are a minority, and some of these cases are resolved by the insertion of an epenthetic infix or

The Word in Phonology

41

by vowel deletion (cf. Pensado 1999: 4459). This fact prevents the appearance of heterosyllabic consonant clusters which are not found morpheme-internally. Second, unlike enclitics (cf. (25), in section 4.2.2), suffixes form part of a single domain of stress assignment with the stem (except the adverb-forming -mente, on which see below), as illustrated by the following examples from nominal and verbal morphology, respectively, where the morphological breakdown is provided in round brackets: (38) a. ní.ño (niñ-o) ‘child (masc.)’ b. ni.ñé.ro (niñ-er-o) ‘babysitter (masc.)’ c. ni.ñe.rí.a (niñ-er-í-a) ‘childish thing’ (39) a. cán.tas (cant-a) ‘(s)he sings’ b. can.tá.ba (cant-a-ba) ‘(s)he sang’ c. can.ta.rí.a (cant-a-rí-a) ‘(s)he would sing’

A third piece of evidence would come from the lack of aspiration and velarization of root- or stem-final /s/ and /n/, respectively, when followed by vowel-initial suffixes in those dialects that Harris (1993) reports to have aspiration and velarization at the boundary of a prefix and a stem (cf. (33a)-(34a) above). In (40), found in Harris (1993: 182), the root-final /s/ is aspirated when appearing in the coda (cf. (40a)) but not when appearing in the onset, after the addition of the vowel-initial plural suffix -es (cf. (40b)). In (40), the root-final /n/ is velarized in a coda (cf. (41a)), but not in an onset, once the plural suffix is added. (40) a. tos ‘cough’ b. tos-es ‘coughs’ (41) a. fin ‘end’ b. fin-es ‘ends’

to[h] to. [s]es fi[ƾ] fi. [n]es

42

Chapter One

The lack of aspiration of the /s/ and the lack of velarization of the /n/ in onset position differ from the aspiration and velarization of the final /s/ and /n/ of the prefixes des- and en- in onset position in (33a) and (34a) above. That is, the morphological boundary between the root and a suffix is not of the same type from a prosodic point of view as the one between a prefix and a root or stem. The asymmetry between prefixes and suffixes in Spanish regarding their degree of cohesion with the stem or base they attach to can be explained if prefixes adjoin to the Ȧ containing the base, and suffixes form one Ȧ with the base, i.e. they are internal to it rather than adjoined to it. That is, suffixes would have the prosodic structure in (42a), whereas prefixes would have the prosodic structure in (42b). Peperkamp (1997) and Vigário (2003) assume the same prosodic structure for suffixes in Italian and European Portuguese, respectively. In fact, it is a common observation across languages that suffixes are more cohesive with the stem than prefixes. No language has been reported to have prefixes more cohering than suffixes: (42) a. ((Stem Suff)Ȧ)ij b. ((Pref (Stem)Ȧ)Ȧ)ij

It should be pointed out, however, that there are suffixes which are not integrated in the same Ȧ with the root or stem. Class II suffixes of English would be one example. In Spanish, the disyllabic suffix -mente is another case. This suffix, which attaches to adjectives to form manner adverbs (equivalent to English -ly), is the only suffix which bears its own stress independent of the stem. The stem bears primary stress, and it bears secondary stress. Actually, mente-adverbs are the only words with secondary stress in Spanish: (43) a. rápida-ménte ‘quick-ly’ b. cuidadósa-ménte ‘careful-ly’

rápidamènte cuidadósamènte

In Dutch, there are disyllabic affixes which are independent Ȧs, judging from their behavior with respect to deletion under coordination. In this language, a word can be deleted when it is repeated in coordinated structures. These words can be fully independent lexical categories, like nouns (cf. (44a)), but also stems of prefixed forms (cf. (44b)) and suffixes (cf. (44c,d)). The examples are from Gussenhoven and Jacobs (2011: 256):

The Word in Phonology

43

(44) a. grote maten en kleine maten ‘big sizes and small sizes’ b. in voer en uit voer ‘import and export’ c. zwanger schap en moeder schap ‘pregnancy and maternity’ d. paars achtig en groen achtig ‘purple-like and green-like’

The last suffix, -achtig, should be compared to its monosyllabic equivalent -ig, which cannot be deleted under identity in coordination. This suffix is monosyllabic, contains a lax vowel, and is unstressed: (45) a. paars ig en groen ig *paars ig en groen ig ‘purple-like and green-like’

Assuming with Booij (1996) that only independent Ȧs are deleted under identity in coordination, the difference between the two related Dutch affixes would be that -achtig is a Ȧ, whereas -ig is not, as it forms a single Ȧ with the stem. Thus, non-cohering suffixes would have the representation in (28d), in which suffixes are independent Ȧs: (46) ((Stem)Ȧ(Suff)Ȧ)ij

4.2.4. The prosody of compound words Compound words constitute an interesting scenario for the interface between morphosyntax and prosody. There is evidence that the two members forming the compound (roots or words) may have any of the three prosodic structures in (47), parallel to clitics and affixes. The two members may be prosodically integrated in the same Ȧ (cf. (47a)), form two independent Ȧs (cf. (47c)), or the first member may be adjoined to the Ȧ that the second member constitutes by itself. Unlike clitics and affixes, I have not found any cases of compounds in which the first member is directly adjoined to a ij. Further work is needed to confirm whether this is a real gap among the options, and if so, a principled explanation will have to be posited.

Chapter One

44 (47) a. ij

b. ij

Ȧ

Ȧ

V(V) V(V) Rt/Wd

Rt/Wd

V(V) Rt/Wd

c. ij Ȧ

Ȧ

Ȧ

V(V)

V(V)

V(V)

Rt/Wd Rt/Wd

Rt/Wd (47’) a. ((Rt/Wd – Rt/Wd)Ȧ)ij c. ((Rt/Wd)Ȧ(Rt/Wd)Ȧ)ij

b. ((Rt/Wd (Rt/Wd)Ȧ)Ȧ)ij

Italian has compounds illustrating the three prosodic structures above. Root+root compounds with old Greek or Latin roots have only one primary stress, and no evidence of juncture between the members of the compound. Thus, the following compounds have only one primary stress and present /s/-voicing on the initial /s/ of the second member of the compound, which is only possible within a Ȧ, not across Ȧs (Peperkamp 1997: 134): (48) a. cromo-soma ‘chromosome’ b. filo-sofo ‘philosopher’

cromo-[z]óma filó-[z]ofo

On the other hand, there are compounds with one secondary stress on the first member and one primary stress on the second member (cf. (49a,b)) and compounds with one primary stress in each member (cf. (49c)). The latter are formed by two nouns, in which the second one modifies the first one. Unlike regular morphological compounds, the first member in these complex words can be inflected. Some scholars call them juxtapositions or syntactic compounds. The presence of open lax vowels constitutes evidence of stress, as these vowels can only appear in stressed syllables in Italian (Peperkamp 1997: 124):

The Word in Phonology

45

(49) a. terra-cotta t[‫ܭ‬Ғ]rra-c[‫ܧ‬ғ]tta ‘terracotta (lit. earth cooked)’ b. porta-ombrelli p[‫ܧ‬Ғ]rta-ombr[‫ܭ‬ғ]lli ‘umbrella-stand (lit. hold umbrellas)’ c. parola modello par[‫ܧ‬ғ]la mod[‫ܭ‬ғ]llo ‘model word (lit. word model)’

/s/-voicing does not occur on the second member of these compounds: (50) a. campo-santo ‘cemetery (lit. field holy)’ b. scatola sorpresa ‘surprise box (lit. box surprise)’

campo-[s]anto scatola [s]orpresa

These two types of compounds have different prosodic structures. The ones illustrated in (48) have the prosodic structure (47a)/(47’a) (i.e. the two members of the compound form a single Ȧ), and the ones illustrated in (49)-(50) have the prosodic structure (47c)/(47’c) (i.e. each member of the compound constitutes an independent Ȧ). Interestingly, there seems to be variation on the latter type of compounds among native Italian speakers. Some speakers pronounce these compounds without open lax vowels in the first member of the compound (thus presumably without stress), and /s/-voicing does not occur on the second member: (51) a. copri-letto ‘cover-bedspread (lit. cover bed)’ b. reggi-seno ‘bra (lit. hold breast)’

c[o]pri-letto r[e]ggi-[s]eno

Peperkamp’s (1997: 125-129) analysis of this variation is that compounds with open lax vowels are formed by two independent Ȧs, but compounds without open lax vowels have a prosodic structure in which the first member is not an independent Ȧ and is adjoined to the Ȧ that the second member does constitute. The structure would hence be the one in (47b)/(47’b). The absence of open vowels in the first member of the compound would be explained by the fact that stress is only assigned to Ȧs, and these members are not independent Ȧs. The absence of /s/-voicing would be explained by the fact that there is still a Ȧ-boundary at the beginning of the second word. In Spanish, there are compound words with one stress and compound words with two stresses. Penultimate stress is the most common accentuation pattern among the former type of compounds, except for

46

Chapter One

compounds formed of Greek or Latin old roots, which present antepenultimate stress (cf. (52a)): (52) a. bió-grafo ‘biographer’ b. saca-córchos ‘corkscrew (lit. extracts-corks)’ c. cara-dúra ‘cheeky (lit. face hard)’

Compounds with two stresses, one on each member of the complex word, are similar constructions to Italian syntactic compounds or juxtapositions (cf. (49c)):21 (53) a. hómbre rána ‘scubadiver (lit. man-frog)’ b. cóche cáma ‘sleeping car (lit. car-bed)’ c. fútbol sála ‘indoor football/soccer (lit. football/soccer room’)

Like in Italian, the latter type of compounds can be easily represented with the prosodic structure in (47c)/(47’c), that is, each member of the compound constituting a Ȧ. As for compounds with one stress, I would venture that they are of two types. Compounds formed by old Greek or Latin roots (cf. (52a)) have antepenultimate stress rather commonly, and hence their main stress can fall on the first member of the compound (besides bió-grafo in (52a), cf. bió-logo ‘biologist’, filó-sofo ‘philosopher’, polí-glota ‘polyglot’, francó-filo ‘Francophile’, etc). It seems reasonable to posit a prosodic structure like (47a)/(47’a) for these compound words, that is, one in which the two roots form a single Ȧ, and where the whole set of syllables is a single window for stress assignment. The fact that the members of these compounds are roots rather than words (i.e. they are bound roots) also contributes to the non-transparency or opacity of the internal division, and hence to the absence of Ȧ-juncture. One-stress compounds which are not composed of Greek or Latin bound roots (cf. (52b)) but of independently occurring words have a different prosodic structure. It would seem more appropriate to presume that they have the structure in (47b)/(47’b), where the first member does not constitute a Ȧ by itself and adjoins to the Ȧ formed by the second 21

See Hualde (2006) for additional data and analysis of compounds with one and two stresses as word-level and phrase-level compounds, respectively.

The Word in Phonology

47

member, thus giving rise to a recursive Ȧ. The evidence is only indirect. In sections 4.2.2 and 4.2.3.1 I referred to the process discussed by Harris (1983) that strengthens an underlying word-initial /ᢪ/ to /r/ in Spanish (cf. the examples in (19) di[ᢪ]ecto ~ [r]ecto; di[ᢪ]igir ~ [r]egir; e[ᢪ]upción ~ [r]uptura). I used this process as evidence for diagnosing the presence of a Ȧ-boundary aligned with the left edge of lexical words and stems or bases following proclitics and non-cohering prefixes (cf. (36)). Although there are no real compounds in which the second member is one of the words with the alternation /ᢪ/~/r/, for made-up but perfectly plausible and transparent one-stress compounds like the ones in (54), native speakers agree that the second member is pronounced with a trill. These pronunciations would inform of the presence of a Ȧ-boundary between the two members of the compound: (54) a. euro-[r]uptura ‘euro-rupture (i.e. breaking of (part of) the European Union or the euro)’ b. espalda-[r]ecta ‘straight back (i.e. someone whose back is straight or stiff)’

Thus, as in Italian, Spanish compound words would present the three prosodic structures in (47)/(47’). Many more examples of compounds displaying the prosodic structures in (47a)/(47’) in different languages can be found in the literature, and there is no space here to present more. The reader can consult Nespor and Vogel (1986: 109-144), Nespor and Ralli (1996) or Vigário (2003, 2010), among others.22 To end this section, I will bring forth Vigário’s (2003, 2010) advocacy for an alternative treatment of recursive Ȧ-structure. As stated in section 4.2.3.1, Vigário (2003, 2010) advocates for the existence of a prosodic constituent intermediate between the Ȧ and the ij, namely the Compound Prosodic Word (represented as Ȧmax in Vigário 2003), or more recently, the Prosodic Word Group (PWG in Vigário 2010). Thus, for this author, the prosodic structure in (47d)/(47’d), where the Ȧs corresponding to each of the words in the compound are dominated by ij is incorrect. Rather, those Ȧs should be dominated by a PWG:

22

An interesting point that deserves to be mentioned is the absence (to the best of my knowledge) of the mirror image of (47b)/(47’b), in which the second member of the compound is adjoined to a Ȧ formed by the first member. Closer scrutiny will tell whether this proves to be a real gap.

Chapter One

48 (55)

PWG

Ȧ Ȧ | | V(V) V(V) | | Rt/Wd Rt/Wd

The PWG is extendable to the higher level of the recursive Ȧ-structures considered by Selkirk (1995) for clitics/function words and affixes in (12), (15), (27), and (28), and is reminiscent of the Maximal Prosodic Word or Ȧ-max of Itô and Mester (2009). The evidence adduced by Vigário (2003, 2010) comes from European Portuguese and other languages. The first piece of evidence from European Portuguese is the level or degree of stress. The first member of a compound word has a weaker stress than its corresponding independent, non-compound Ȧ. The second type of evidence from European Portuguese comes from the (non)application of vowel deletion across Ȧs. In this language, Ȧ-final vowels are deleted when certain segmental and prosodic conditions on the following word are met (cf. Vigário 2003: 73-114). Within compounds, vowel-deletion applies in a more constrained way. For instance, non-back mid vowels and /a/ at the end of the first member do not get deleted when the second member receives main stress in the compound. In non-compound sequences of Ȧs contained within a ij, however, stress plays no role for non-back vowel deletion (Vigário 2003: 238, Vigário 2010). The third type of phenomena of European Portuguese adduced as evidence in favor of the PWG comes from narrow (contrastive) focus stress. Within a ij, narrow focus stress may be carried by any of the Ȧs that are narrowly focalized. If a compound word is narrowly focalized, however, narrow focus stress invariably falls on the second member (i.e. rightmost Ȧ). Besides these data, additional evidence from other languages supports the difference between non-compound sequences of Ȧs and compound words, as presented and argued by Vigário (2010). In English, for instance, the stress pattern is different. Two independent words present a weak-strong pattern within a ij, whereas two members of a compound word present a strong-weak pattern: blàck bóard vs. bláckboard, grèen hóuse vs. gréenhouse, tèachers’ fríend vs. téachers’ union, etc. Dutch also presents similar contrasts (with secondary stress on the second member of

The Word in Phonology

49

compounds),23 and in this language processes such as progressive fricative devoicing and regressive obstruent voicing apply obligatorily across Ȧs within compounds but optionally across Ȧs within ijs. In Baule, processes such as tonal upstep and word-initial and word-final vowel deletion apply across members of compounds and compound-like structures but not across Ȧs in phrases. In Hungarian, the stress pattern of compound words differs from that of non-compound sequences of words in that the stress of the second member is lost, and each member of a compound also constitutes an independent domain for vowel harmony and palatalization. In Warlpiri, the stress of the second member of a compound is also weaker than in a non-compound sequence, and like in Hungarian, vowel harmony does not occur across members of a compound. In French, each member of a compound forms a domain for glide formation and closed syllable adjustment, like an independent Ȧ; however, a compound is also different from a sequence of words in a phrase, as evidenced by the obligatory presence of a schwa in the penultimate syllable of a foot, unlike in sequences of words. In Korean and Japanese, two words within a compound constitute a single Accentual Phrase, whereas they are independent Accentual Phrases when appearing in a phrase. Vigário (2010) also provides arguments against positing completely recursive Ȧ structures for compound words, that is, a structure like (56): (56)

Ȧ | V(V) | Rt/Wd

Ȧ

Ȧ | V(V) | Rt/Wd

One of the arguments provided by Vigário (2010) is that in recursive structures the differences between the higher and lower levels are of degree or strength, as the prosodic category is the same (cf. Frota 2000, Ladd 2008). Thus, if the structure of compounds were the one in (56), we should expect to find gradient differences between non-compound and compound structures. However, the differences are categorical and nongradient. That is, the differences are of type rather than of degree. In 23

In Dutch and English there are compounds without the strong-weak pattern, but presumably these are syntactic compounds, composed of two PWGs (cf. Vigário 2010).

50

Chapter One

Turkish, for instance, stress at the Ȧ level is located at the right edge, but in compounds stress falls on the leftmost member. A similar difference is found in English and Dutch, as mentioned above (at least in morphological compounds, not so much in syntactic compounds), as well as in Swedish. Vigário’s (2010) claim is that the PWG is not a constituent that is only relevant for compound words. Rather, this author argues that it is a regular prosodic constituent and that all Ȧs are dominated by the PWG in the default case, obeying the Exhaustivity condition of prosodic structure. That is, the PWG would be a constituent intermediate between the Ȧ and the ij: (57) ij | PWG | Ȧ

Apart from the evidence presented above suggesting that the PWG is different from ij and Ȧ, Vigário (2010) adduces the fact that the PWG is subject to size constraints, like other prosodic categories (i.e. minimality conditions for Ȧs, ijs and Ls -Intonational Phrases-, cf. Nespor and Vogel 1986, Ghini 1993, Selkirk 2000, Frota 2000, D’Imperio et al. 2005, Frota et al. 2007). In Baule, for instance, a Ȧ does not project its own PWG if monosyllabic, and is integrated with the preceding Ȧ or Ȧs in the same PWG. In European Portuguese, words that are formed by the initial letters of organizations are pronounced in such a way that each initial letter is a Ȧ. In cases of four initial letters, the four Ȧs are grouped in two PWGs. Another argument in favor of the PWG as a prosodic constituent is its non-isomorphism with morphosyntax. In default terms, the PWG would correspond to an X0 in syntax (in parallel to the correspondence between ij and XP), but there are cases in which two X0s may form one PWG, as in Baule, or cases in which a single X0 is composed of two PWGs, like in European Portuguese initial letter pronunciations or combinations of prefix+stem or stem+suffix, as in Italian, European Portuguese, Spanish and Dutch, as shown in sections 4.2.3.1 and 4.2.3.2 above. In my view, the proposal of the PWG needs further clarification. If the PWG and the Ȧ are assumed to choose different edges for stress assignment, as argued by Vigário (2010) for English, Dutch, Swedish and Turkish, and the PWG and the Ȧ coexist in the prosodic structure of any morphosyntactic word, then we should expect to see both types of stress being assigned (i.e. stresses at both edges). However, the types of stress are in complementary distribution: the stress pattern of Ȧs is observed for

The Word in Phonology

51

single, non-compound words, and the stress pattern of PWGs is observed for compound words and similar structures. It would be desirable to have a better understanding of the relationship between the PWG and the Ȧ, and more specifically, of the distribution and degree of presence of the PWG in the prosodic hierarchy.

5. Recent proposals without prosodic constituency This article should not end without making reference to recent movements in the area of the interfaces of phonology with morphology and syntax. As already mentioned in section 3, several authors have claimed that the traditional levels and strata of Lexical Phonology should be rethought in terms of phases, namely as morphological phases (cf. Marvin 2002, DiSciullo 2004, 2005). Marantz (2007), Embick and Marantz (2008) and Embick (2010) argue that the heads n, v and a create phases that are spelled out cyclically, in the spirit of Multiple Spell-Out of the phase-based theory of linguistic derivations (cf. Uriagereka 1999, Chomsky 2001). More accurately, it is spell-out domains of phases that are spelled out, namely the complements of the phasal heads. Since derivational morphemes are specified for category, they create phases. Some phonologists have adopted this claim and have advocated for the abandonment of morphological levels or strata where phonological rules are specified to apply, and have claimed that it is morphological phases that create domains where phonological processes operate. In this regard, see Pak (2008: chapter 6) and especially Samuels (2009, 2011, 2012), who calls the model “Phonological Derivation by Phase”. The attractiveness of this view is that it derives the concepts of levels or strata from general concepts of generative minimalist grammar, such as the Phase Theory of syntactic derivations (Chomsky 2001), applied word-internally. It relates cyclicity and morphophonology to the general functioning of grammar, in terms of phases and cyclic spell-out. Thus, the theory that the above mentioned scholars defend has the conceptual advantage that it is more general, more parsimonious, and hence more optimal (see also Scheer 2012 for discussion). A direct implication of an application of phase theory and cyclic spellout to phonology is that prosodic constituency is deemed superfluous, both at the word level and above. Thus, Pak’s (2008) and Samuels’s (2009, 2011, 2012) claim is that the Prosodic Hierarchy should be rejected altogether, and that prosodic constituents as such do not exist (see also Scheer 2011, 2012 for criticism). Rather, it is the spell-out domains of phases both at the word level and at the phrasal and clausal levels where

52

Chapter One

phonological rules are set to operate. Phonological rules are lexically specified to apply cyclically, at each morpheme phase, or non-cyclically (i.e. postlexically). The reader is referred to the mentioned sources for more detailed discussion, which I cannot undertake here for limits of space. I will only say here that the position advocated by these researchers deserves to be followed closely for the advantages I have mentioned above. And it remains to be seen whether the promises set by this alternative theory can indeed account for all the phenomena analyzed in terms of prosodic constituency, thus showing that prosodic phonology is entirely superfluous and that no domains such as Ȧ, ij or Ț (the latter being an abbreviation for the Intonational Phrase) are needed in phonology. Much work has to be done in the future in this regard. Just to point out an intriguing topic of investigation, one wonders how a theory based on morphological and syntactic configurations can account for the distinctions in prosodic structure for the same syntactic elements in dialects of the same language. For instance, as shown in section 4.2.2, function words are related prosodically to the following word in different ways in three dialects of Serbian: internal clitics in Eastern Herzegovina, affixed clitics in Šrem/Maþva, and free clitics in Belgrade (cf. (13)-(15)). And postverbal clitics are prosodized differently in three dialects of Italian: internal clitics in Lucanian, affixed clitics in Neapolitan, and free clitics in Standard Italian (cf. (16)-(17)). The differences between the examples in the dialects of Serbian and Italian are apparently minimal. If the syntactic structure is the same for the three dialects in each language, and hence if the spell-out domains of phases are the same across dialects, it would be interesting to see how a phasal spell out theory would account for the differences in phonological behavior. Elordieta (1997, 1999) also shows the difficulties that prosodic constituency as proposed in prosodic phonology has in accounting for phonological processes occurring in some languages between morphemes and words that are syntactically closely related, to the exclusion of morphemes and words that do not have the same degree of syntactic cohesion.24 In Lekeitio Basque vowel assimilation, French liaison and Irish initial consonant mutation occur between morphemes or syntactic heads between which a relation of feature checking or feature validation holds (as 24

This is just a very brief summary presented here of the view defended in Elordieta (1997, 1999). The interested reader can consult the original works or the longer reviews in Elordieta (2007, 2008), from which the summary presented here is taken directly.

The Word in Phonology

53

understood in the minimalist framework of grammatical derivations). These heads would constitute feature chains (e.g. {C, T}, {T, v}, {T, D}, {v, D} and {D, N}), and the idea would be that the cohesion of feature chains is represented or made visible in other components of grammar, namely that this syntactic cohesion is reflected in the components of grammar where heads and their features are spelled out. The morphemes realizing the heads in feature chains form phonological constituents, and as such, certain phonological processes may be specified to apply in them.25 In Elordieta (1997, 1999) it was proposed that these phonological constituents could not be identical to phonological or prosodic words, since a lexical verb and an inflected auxiliary may each bear their own stress. Hence, it was suggested that feature chains were not directly mapped to phonological structure but to an intermediate structure, the level of Morphological Structure (MS), argued for in the theory of Distributed Morphology. The claim is that feature chains are realized or represented at the level of MS as morphosyntactic units, MS-words, if the heads are spelled out linearly adjacent, either as a result of incorporation or by being in a spec-head or in a head-complement configuration. From MS, feature chains would be mapped into the phonological component proper as constituents or domains where phonological processes may apply. That is, in addition to domains formed at PF by prosodic properties of morphemes, PF also contains constituents which are mapped from this intermediate component between syntax and PF. This does not mean that all languages manifest (all) feature chains overtly in terms of phonological operations, since languages differ in the density of rules of external sandhi that they have. The opposite scenario would be that a language generalizes the phonological relationship between heads related by feature chains to any linear sequence of a functional head and a lexical head. One advantage of Elordieta’s (1997, 1999) proposal would be that it offers a principled explanation for the Principle of the Categorial Invisibility of Function Words (PCI), that is, for the stipulation that functional categories are included in the same prosodic constituent with lexical categories (cf. Selkirk 1984 and subsequent work in prosodic phonology). Another positive aspect is that the theory recurs to a notion that already exists independently in the grammar, such as feature checking relationships. Samuels (2009: 318-327) subsumes Elordieta’s proposal into the more general theory of cyclic spell-out, under the argument that feature checking and validation are processes that are part of phase mergers, and 25

Epstein and Seely (2002) argue that each and every syntactic object formed by two heads whose features enter in a feature valuation operation is mapped to the interface levels PF and LF.

54

Chapter One

that the domains created by feature checking coincide with or are part of spell-out domains of phases. Part of her claim would also be that the theory of Phonological Derivation by Phase is more general, as it would also account for phonological phenomena occurring over stretches of material that are more extensive than two adjacent heads, such as phrasal and clausal levels. Thus, the same as there are morphological phases, there would be spell-out domains of syntactic phases at the phrase and clause levels that are interpreted as domains of operations of phonological phenomena. Further empirical and theoretical work is necessary in order to confirm whether Phonological Derivation by Phase is the general model of the syntax-phonology interface that its proponents advocate, and whether the hierarchical prosodic constituency as devised in Prosodic Phonology should be rejected and abandoned altogether. This would certainly be a significant development in linguistic theory, due to the implications it would have for our understanding of how formal grammar works at the interface of syntax and phonology.

6. Conclusions In this chapter, I made an attempt to provide a general summary of the main aspects of the phonology of words. The first section summarized some of the phonetic and phonological cues for word demarcation, such as the articulatory strengthening of consonants and vowels at word edges, the tendency of stress to align with syllables at or near the word edge in fixed stress systems, the phonotactic distributions of allophones and phonemes with respect to word edges, and the application of phonological processes within the limits of prosodic words (Ȧs). The second section succinctly presented some notions of morphophonology, that is, phonology as triggered by the addition of morphemes. The third section (the longest one) was devoted to prosodic phonology, with a review of issues such as word-minimality conditions and the correspondences between morphosyntactic words (syntactic heads or X0s) and prosodic words (Ȧs). Within the latter issue, the attention was on the different prosodic structures that cliticized function words, affixed words and compound words adopt in different languages, with special attention to Spanish. Reviewing such a vast topic in the format of a regular-size paper is a necessarily complicated task, as the issues are just are just glimpsed at and not examined at any depth. Moreover, certain points have not even been dealt with, such as the acquisition of words in L1 (and L2) phonology, or the psycholinguistic aspects of the processing of words. But it is my hope that at least the aspects presented constitute a useful starting guide or

The Word in Phonology

55

reference for the non-expert in the topic of the word in phonology, with a substantial amount of bibliographical references on the different subtopics.

References Allen, M. 1979, Morphological investigations, Ph Diss., University of Connecticut Anderson, S. 2005, Aspects of the theory of clitics, Oxford: Oxford University Press Aronoff, M. 1976, Word formation in generative grammar, Cambridge: Cambridge University Press Bagou, O., C. Fougeron and U. Frauenfelder 2002, “Contribution of prosody to the segmentation and storage of ‘words’ in the acquisition of a new mini-language”, Proceedings from the 5th Speech Production Workshop: Models and Data, Germany: Kloster Seeon Baker, M. 1988, Incorporation: A theory of grammatical function changing, Chicago: The University of Chicago Press Barnes, J. 2005, Strength and weakness at the interface: Positional neutralization in phonetics and phonology, Berlin and New York: Mouton de Gruyter Beard, R. 1991, “Decompositional composition: The semantics of scope ambiguities and ‘bracketing paradoxes’”, Natural Language and Linguistic Theory 9, 195-229 Bermúdez-Otero, R. 2011, “Cyclicity”, M. van Oostendorp, C. Ewen, E. Hume and K. Rice (eds.) The Blackwell companion to phonology, vol. 4, Malden, MA: Wiley-Blackwell, 2019-2048 —. 2012, “The architecture of grammar and the division of labour in exponence”, J. Trommer (ed.) The morphology and phonology of exponence, Oxford: Oxford University Press, 8-83 —. 2013, “The stem-level syndrome”, Paper presented to the UPenn Linguistics Department, Speaker Series, Philadelphia, 11 April 2013 Booij, G. 1996, “Cliticization as prosodic integration”, The Linguistic Review 13, 219-242 —. 1999, “The role of the prosodic word in phonotactic generalizations”, T. A. Hall and U. Kleinhenz (eds.), Studies on the phonological word, Amsterdam: John Benjamins, 45-72 Booij, G. and R. Lieber 1993, “On the simultaneity of morphological and prosodic structure”, S. Hargus and E. Kaisse (eds.) Studies in Lexical Phonology, San Diego: Academic Press, 23-44 Booij, G. and J. Rubach 1987, “Postcyclic versus postlexical rules in Lexical Phonology”, Linguistic Inquiry 18, 1-44

56

Chapter One

Borowsky, T. 1993, “On the word level”, S. Hargus and E. Kaisse (eds.) Studies in Lexical Phonology, San Diego: Academic Press, 199-234 Byrd, D. 2000, “Articulatory vowel lengthening and coordination at phrasal junctures”, Phonetica 57, 3-16 Byrd, D. and E. Saltzman 1998, “Intragestural dynamics of multiple prosodic boundaries”, Journal of Phonetics 26, 173-20 Cabré, T. 1993, Estructura grammatical i lexicó: el mot mínim català, PhD diss., Universitat Autònoma de Barcelona Chen, M. 1987, “The syntax of Xiamen tone sandhi”, Phonology Yearbook 4, 109-149 Cho, T. 2002, The effects of prosody on articulation in English, New York: Routledge —. 2004, “Prosodically-conditioned strengthening and vowel-to-vowel coarticulation in English”, Journal of Phonetics 32, 141-176 —. 2006, “Manifestation of prosodic structure in articulation: Evidence from lip kinematics in English”, L. Goldstein, D. H. Whalen and C. T. Best (eds.) Laboratory Phonology 8, Berlin/New York: Mouton de Gruyter, 519-548 Cho, T. and Keating, P. 2001, “Articulatory strengthening at the onset of prosodic domains in Korean”, Journal of Phonetics 28, 155-190 Cho, T., J. McQueen and E. Cox 2007, “Prosodically driven phonetic detail in speech processing: The case of domain-initial strengthening in English”, Journal of Phonetics 35, 210-243 Chomsky, N. 2001, “Derivation by phase”, M. Kenstowicz (ed.) Ken Hale: A life in language, Cambridge (MA): MIT Press, 1-54 Chomsky, N. and M. Halle 1968, The sound pattern of English, New York: Harper and Row Christophe, A., and E. Dupoux 1996, “Bootstrapping lexical acquisition: The role of prosodic structure”, The Linguistic Review 13, 383-412 Cohn, A. 1989, “Stress in Indonesian and bracketing paradoxes”, Natural Language and Linguistic Theory 7, 167-216 Cutler, A. and D. Norris 1988, “The role of strong syllables in segmentation for lexical access”, Journal of Experimental Psychology: Human Perception and Performance 14, 113-121 Dixon, R. W. and A. Y. Aikhenvald 2002, Word: A cross-linguistic typology, Cambridge: Cambridge University Press De la Cruz, I. 2012, Chunking the input: On the role of frequency and prosody in the segmentation strategies of adult bilinguals, PhD diss., University of the Basque Country D’Imperio, M., G. Elordieta, S. Frota, P. Prieto and M. Vigário 2005, “The phonetics and phonology of intonational phrasing in Romance”, P.

The Word in Phonology

57

Prieto, J. Mascaró and M.-J. Solé (eds.) Segmental and prosodic issues in Romance phonology, Amsterdam and Boston: John Benjamins, 131153 DiSciullo, A.-M. 2004, “Morphological phases”, J.-Y. Yoon (ed.) Generative Grammar in a Broader Perspective: The 4th GLOW in Asia, Seoul: The Korean Generative Grammar Circle and Cognitive Science, 113-137 DiSciullo, A.-M. 2005, Asymmetry in morphology, Cambridge (MA): MIT Press DiSciullo, A.-M. and E. Williams 1987, On the definition of Word, Cambridge (MA), MIT Press Downing, L. 1999, “Prosodic stem  prosodic word in Bantu”, T. A. Hall and U. Kleinhenz (eds.) Studies on the phonological word, Amsterdam: John Benjamins, 73-98 Downing, L. 2006, Canonical forms in prosodic morphology, Oxford: Oxford University Press Elordieta, G. 1997, Morphosyntactic feature chains and phonological domains, PhD diss., University of Southern California —. 1999, “Phonological cohesion as a reflex of morphosyntactic feature chains”, Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, Stanford (CA): CSLI Publications, 175-189 —. 2007, “Segmental phonology and syntactic structure”, G. Ramchand and Ch. Reiss (eds.) The Oxford Handbook of Linguistic Interfaces, Oxford: Oxford University Press, 125-178 —. 2008, “An overview of theories of the syntax-phonology interface”, ASJU - International Journal of Basque Linguistics and Philology 42, 209-286 Embick, D. 2010, Localism versus globalism in morphology and phonology, Cambridge (MA): MIT Press Embick, D. and A. Marantz 2008, “Architecture and blocking”, Linguistic Inquiry 39, 1-53 Embick, D. and R. Noyer 2007, “Distributed Morphology and the syntax/morphology interface”, G. Ramchand and Ch. Reiss (eds.) The Oxford Handbook of Linguistic Interfaces, Oxford: Oxford University Press, 289-324 Epstein, S. D. and T. D. Seely 2002, “Rule applications as cycles in a level-free syntax”, S. D. Epstein and T. D. Seely (eds.) Derivation and explanation in the Minimalist Program, Cambridge: Blackwell, 65-89 Everett, D. 1996, Why there are no clitics: An alternative perspective on pronominal allomorphy, Dallas: SIL

58

Chapter One

Fernandes, T., P. Ventura and R. Kolinsky 2007, “Statistical information and coarticulation as cues to word boundaries: A matter of signal quality”, Perception & Psychophysics 69, 856-864 Fitzpatrick-Cole, J. 1990, “The minimal word in Bengali”, A. Halpern (ed.) Proceedings of the Ninth West Coast Conference on Formal Linguistics, Stanford: Stanford Linguistics Association, 157-170 Flack, K. 2009, “Constraints on onsets and codas of words and phrases”, Phonology 26, 269-302 Fougeron, C. 1999, “Prosodically conditioned articulatory variation: A review”, UCLA Working Papers in Phonetics 97, 1-73 Fougeron, C. and P. Keating 1997, “Articulatory strengthening at edges of prosodic domains”, Journal of the Acoustical Society of America 101, 3728-3740 Friederici, A. M. and J. M. I. Wessels 1993, “Phonotactic knowledge of word boundaries and its use in infant speech perception”, Perception & Psychophysics 54, 287-295 Frota, S. 2000, Prosody and focus in European Portuguese, London: Garland Frota, S., M. D’Imperio, G. Elordieta, S. Frota, P. Prieto and M. Vigário 2007, “The phonetics and phonology of intonational phrasing in Romance”, P. Prieto, J. Mascaró and M.-J. Solé (eds.) Segmental and prosodic issues in Romance phonology, Amsterdam and Boston: John Benjamins, 131-153 García-Bellido, P. 1987, “La ordenación de las reglas y la Fonología Léxica”, Estudios de Lingüística Universidad de Alicante 4, 107-126 Garrett, E. 1999, “Minimal words aren’t minimal feet”, UCLA Working Papers in Linguistics 1, 68-105 Gerlach, B. and J. Grijzenhout 2001, Clitics in phonology, morphology and syntax, Amsterdam: John Benjamins Ghini, M. 1993. Phonological Phrase formation in Italian, MA thesis, University of Toronto. Giegerich, M. 1999, Lexical strata in English: morphological causes, phonological effects, Cambridge: Cambridge University Press Goedemans, R. and H. van der Hulst 2011, “Fixed stress locations”, M. S. Dryer and M. Haspelmath (eds.) The world atlas of language structures online, ch. 14 [http://wals.info/chapter/14. Accessed on 2013-11-13], München, Max Planck Digital Library Gordon, M. 1999, Syllable weight: phonetics, phonology and typology, PhD diss., UCLA —. 2001, “The tonal basis of weight criteria in final position”, Proceedings of the Chicago Linguistics Society 36, 141-156

The Word in Phonology

59

—. 2002, “A factorial typology of quantity insensitive stress”, Natural Language and Linguistic Theory 20, 491-552 Gussenhoven, C. and H. Jacobs 2011, Understanding phonology [3rd ed.], London: Hodder-Arnold Hale, K. and S. J. Keyser 1993, “On argument structure and the lexical expression of syntactic relations”, K. Hale and S. J. Keyser (eds.) The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, Cambridge (MA): MIT Press, 53-109 Hall, T. A. 1999a, “The phonological word: a review”, T.A. Hall and U. Kleinhenz (eds.) Studies on the phonological word, Amsterdam: John Benjamins, 1-22 —. 1999b, “Phonotactics and the prosodic structure of German function words”, T. A. Hall and U. Kleinhenz (eds.) Studies on the phonological word, Amsterdam: John Benjamins, 99-131 Hall, T. A. and U. Kleinhenz 1999, Studies on the phonological word, Amsterdam: John Benjamins Halle, M. 1973, “Prolegomena to a theory of word formation”, Linguistic Inquiry 4, 3-16 Halle, M., J. Harris and J-R. Vergnaud 1991, “A re-examination of the stress erasure convention and Spanish stress”, Linguistic Inquiry 22, 141-159 Halle, M. and A. Marantz 1993, “Distributed Morphology and the pieces of inflection”, K. Hale and S. J. Keyser (eds.) The View from Building 20: Essays in linguistics in honor of Sylvain Bromberger, Cambridge (MA): MIT Press, 111-176 Halle, M. and K. P. Mohanan 1985, “Segmental phonology of modern English”, Linguistic Inquiry 16, 57-116 Halle, M. and J.-R. Vergnaud 1987, An essay on stress, Cambridge (MA): MIT Press Halpern, A. 1995, On the placement and morphology of clitics, Stanford: CSLI Publications Halpern, A. and A. Zwicky 1996, Approaching second: Second position clitics and related phenomena, Stanford: CSLI Publications Hanson, K. and S. Inkelas 2009, The nature of the word: Studies in honor of Paul Kiparsky, Cambridge (MA): MIT Press Hargus, S. and E. Kaisse 1993, Phonetics and Phonology 4: Studies in Lexical Phonology, San Diego: Academic Press Harris, J. 1983, Syllable structure and stress in Spanish, Cambridge (MA): MIT Press

60

Chapter One

Hayes, B. 1989, “The prosodic hierarchy in meter”, P. Kiparsky and G. Youmans (eds.) Rhythm and meter, San Diego: Academic Press, 201260 —. 1995, Metrical stress theory: Principles and case studies, Chicago: University of Chicago Press Hirsh-Pasek K., D. G. Kemler Nelson, P. W. Jusczyk, K. Wright Cassidy, B. Druss and L. Kennedy 1987, “Clauses are perceptual units for young infants”, Cognition 26, 269-286 Hock, H. H. 1991, Principles of historical linguistics [2nd edition], Berlin and New York: Mouton de Gruyter —. 1992, “Initial strengthening”, W. U. Dressler, H. C. Luschützky, O. E. Pfeiffer and J. R. Rennison (eds.) Phonologica 1988, Cambridge: Cambridge University Press, 101-110 —. 1999, “Finality, prosody, and change”, O. Fujimura, B. Joseph and B. Palek (eds.) Proceedings of LP’98: Item and order in language and speech, Prague: Charles University, 15-30 Hualde, J. I. 2003, “Segmental phonology”, J. I. Hualde and J. Ortiz de Urbina (eds.) A grammar of Basque, Berlin and New York: Mouton de Gruyter, 15-65 —. 2006, “Stress removal and stress addition in Spanish”, Journal of Portuguese Linguistics 5, 59-89 —. 2009, “Unstressed words in Spanish”, Language Sciences 31, 199-212 Hyman, L. 1977, “On the nature of linguistic stress”, L. Hyman (ed.) Studies in stress and accent (Southern California Occasional Papers in Linguistics), Department of Linguistics, University of Southern California, 37-82 Inkelas, S. 1990, Prosodic constituency in the lexicon, New York: Garland Itô, J. and A. Mester 2009, “The extended prosodic word”, J. Grijzenhout and B. Kabak (eds.) Phonological domains: Universals and deviations, Berlin and New York: Mouton de Gruyter, 135-194 Johnson, E. K. and P. W. Jusczyk 2001, “Word segmentation by 8-montholds: When speech cues count more than statistics”, Journal of Memory and Language 44, 548-567 Jusczyk, P. W., E. A. Hohne and A. Bauman 1999, “Infants’ sensitivity to allophonic cues for word segmentation”, Perception & Psychophysics 61, 1465-1476 Jusczyk, P. W., D. M. Houston and M. Newsome 1999, “The beginnings of word segmentation in English-learning infants”, Cognitive Psychology 39, 159-207

The Word in Phonology

61

Kabak, B. and Schiering, R. 2006, “The phonology and morphology of function word contractions in German”, Journal of Comparative Germanic Linguistics 9, 53-99 Keating, P., T. Cho, C. Fougeron and C. Hsu 2003, “Domain-initial articulatory strengthening in four languages”, J. Local, R. Ogden and R. Temple (eds.) Phonetic interpretation (Papers in Laboratory Phonology 6), Cambridge: Cambridge University Press, 143-161 Kenstowicz, M. 1994, Phonology in generative grammar, London: Blackwell Publishers Kiparsky, P. 1982, “Lexical morphology and phonology”, The Linguistic Society of Korea (ed.) Linguistics in the morning calm, Seoul: Hanshin, 3-91 —. 1985, “Some consequences of Lexical Phonology”, Phonology Yearbook 2, 85-138 Kirchner, R. 1998, An effort-based approach to consonant lenition, PhD diss., UCLA —. 2001, “Phonological contrast and articulatory effort”, L. Lombardi (ed.) Segmental phonology in Optimality Theory, Cambridge: Cambridge University Press, 79-117 Klavans, J. 1982, Some problems in a theory of clitics, Bloomington: Indiana University Linguistics Club —. 1985, “The independence of syntax and phonology in cliticization”, Language 61, 95-120 Kleinhenz, U. 1996, “The prosody of German clitics”, ZAS Working Papers in Linguistics 6, 81-95 Ladd, D. R. 2008, Intonational phonology [2nd edition], Cambridge: Cambridge University Press Leben, W. and F. Ahoua 1997, “Prosodic domains in Baule”, Phonology 14, 113-132 Marantz, A. 2008, “Phases and words”, S.-H. Choe (ed.) Phases in the theory of grammar, Seoul: Dong In, 191-222 Martinet, A. 1955, Économie des changements phonétiques: traité de phonologie diachronique, Bern: Francke Marvin, T. 2002, Topics in the stress and syntax of words, PhD diss., MIT Mascaró, J. 1976, Catalan phonology and the phonological cycle, PhD diss., MIT McCarthy, J. and A. Prince 1986, Prosodic morphology, Report no. RuCCS-TR-32, New Brunswick, NJ: Rutgers University Center for Cognitive Science, http://ruccs.rutgers.edu/tech_rpt/pm86all. pdf

62

Chapter One

McCarthy, J. and A. Prince 1990, “Foot and word in prosodic morphology: The Arabic broken plural”, Natural Language and Linguistic Theory 8, 209-283 McCarthy, J. and A. Prince 1993, Prosodic morphology: Constraint interaction and satisfaction, Report no. RuCCS-TR-3, New Brunswick, NJ: Rutgers University Center for Cognitive Science. [Rutgers Optimality Archive #482-1201]. McQueen, J. M. 1998, “Segmentation of continuous speech using phonotactics”, Journal of Memory and Language 39, 21-46 Mehler, J., J. Y. Dommergues, U. Frauenfelder and J. Seguí 1981, “The syllable’s role in speech segmentation”, Journal of Verbal Learning and Verbal Behavior 20, 298-305 Mohanan, K.P. 1982, Lexical Phonology, PhD diss., MIT —. 1986, The theory of Lexical Phonology, Dordrecht: Reidel Nespor, M. and A. Ralli 1996, “Morphology-phonology interface: Phonological domains in Greek compounds”, The Linguistic Review 13, 357-382 Nespor, M. and I. Vogel 1982, “Prosodic domains of external sandhi rules. The structure of phonological representations”, H. van der Hulst and N. Smith (eds.) The structure of phonological representations, Part I, Dordrecht: Foris, 225-255 Nespor, M. and I. Vogel 1986, Prosodic Phonology, Dordrecht: Foris Nevis, J. A. 1988, Finnish particle clitics and general clitic theory, New York and London: Garland Noyer, R. 1997, Features, positions and affixes in autonomous morphological structure, New York and London: Garland Otake, T., G. Hatano, A. Cutler and J. Mehler 1993, “Mora or syllable? Speech segmentation in Japanese”, Journal of Memory and Language 32, 258-278 Pak, M. 2008, The postsyntactic derivation and its phonological reflexes, PhD diss., University of Pennsylvania Parker, S. 1999, “On the behavior of definite articles in Chamicuro”, Language 75: 552-562 Pensado, C. 1999, “Morfología y fonología. Fenómenos morfofonológicos”, I. Bosque and V. Demonte (dirs.) Gramática descriptiva de la lengua española (vol. 3), Madrid: Espasa Calpe, 4423-4504 Peperkamp, S. 1997, Prosodic words, HIL Dissertations 34, Holland Academic Graphics Pesetsky, D. 1979, “Russian morphology and lexical theory”, Ms., MIT —. 1985, “Morphology and logical form”, Linguistic Inquiry 16, 193-246

The Word in Phonology

63

Quené, H. 1992, “Durational cues for word segmentation in Dutch”, Journal of Phonetics 20, 331-350 Reiss, Ch. 2003, “Deriving the feature-filling/feature-changing contrast: An application to Hungarian vowel harmony”, Linguistic Inquiry 34, 199-224 Saffran, J. R., R. N. Aslin and E. L. Newport 1996, “Statistical learning by 8-month-old infants”, Science 274, 1926-1928 Saffran, J. R., E. L. Newport and R. N. Aslin 1996, “Word segmentation: The role of distributional cues”, Journal of Memory and Language 35, 606-621 Samuels, B. 2009, The structure of phonological theory, PhD diss., Harvard University —. 2011, Phonological architecture: A biolinguistic perspective, Oxford: Oxford University Press —. 2012, “Consequences of phases for morphophonology”, Á. Gallego (ed.) Phases: Developing the framework, Berlin and New York: Mouton de Gruyter, 251-282 Scheer, T. 2011, A guide to morphosyntax-phonology interface theories. How extra-phonological information is treated in phonology since Trubetzkoy's Grenzsignale, Berlin and New York: Mouton de Gruyter —. 2012, “Chunk definition in phonology: Prosodic constituency vs. phase structure”, M. Bloch-Trojnar and A. Bloch-Rozmej (eds.) Modules and Interfaces, Lublin: Wydawnictwo KUL, 221-253 Selkirk, E. 1980, “The role of prosodic categories in English word stress”, Linguistic Inquiry 11, 563-605 —. 1981, “On the nature of phonological representation”, J. Anderson, J. Laver and T. Meyers (eds.) The cognitive representation of speech, Amsterdam: North Holland, 379-388 —. 1982, The syntax of words, Cambridge (MA): MIT Press —. 1984, Phonology and syntax: The relation between sound and structure, Cambridge (MA): MIT Press —. 1986, “On derived domains in sentence phonology”, Phonology Yearbook 3, 371-405 —. 1995, “The prosodic structure of function words”, J. Beckman, L. Walsh-Dickey and S. Urbanczyk (eds.) Papers in Optimality Theory (University of Massachusetts Occasional papers in Linguistics 18), Amherst: GLSA, 439-469 —. 2000, “The interaction of constraints on prosodic phrasing”, M. Horne (ed.) Prosody: Theory and experiments, Dordrecht: Kluwer, 231-261

64

Chapter One

—. 2011, “The syntax-phonology interface”, J. A. Goldsmith, J. Riggle and A. Yu (eds.) The handbook of phonological theory [2nd edition], Oxford: Wiley Blackwell, 435-484 Selkirk, E. and T. Shen 1990, “Prosodic domains in Shanghai Chinese”, S. Inkelas and D. Zec (eds.) The phonology-syntax connection, Chicago: The University of Chicago Press, 313-337 Selkirk, E. and K. Tateishi 1988, “Constraints on minor phrase formation in Japanese”, M. G. Larson and D. Brentari (eds.) Proceedings of the 24th Annual Meeting of the Chicago Linguistics Society, Chicago: Chicago Linguistics Society, 316-336 Selkirk, E. and K. Tateishi 1991, “Syntax and downstep in Japanese”, C. Georgopoulos and R. Ishihara (eds.) Essays in honour of S.-Y. Kuroda, Dordrecht: Kluwer, 519-543 Siegel, D. 1974, Topics in English morphology, PhD diss., MIT Smith, J. 2005, Phonological augmentation in prominent positions, London and New York: Routledge Smith, R. and S. Hawkins 2000, “Allophonic influences on word-spotting experiments”, A. Cutler, J. M. McQueen and R. Zondervan (eds.) Proceedings from the Workshop on Spoken Word Access Processes (SWAP), Nijmegen: Max-Planck-Institute for Psycholinguistics, 139142 Spencer, A. 1988, “Bracketing Paradoxes and the English Lexicon”, Language 64: 663-682 —. 1991, Morphological Theory: An introduction to word structure in generative grammar, Oxford: Blackwell Spencer, A. and A.R. Luís 2012, Clitics: An Introduction, Cambridge: Cambridge University Press Sproat, R. 1988, “Bracketing paradoxes, cliticization and other topics: The mapping between syntactic and phonological structure”, M. Everaert, A. Evers, R. Huybregts and M. Trommelen (eds.) Morphology and Modularity, Dordrecht: Foris, 339-360 Trubetzkoy, N. 1939, Grundzüge der Phonologie. Travaux du cercle linguistique de Prague 7 [English translation Principles of phonology, by C. Baltaxe, Berkeley : University of California Press] Truckenbrodt, H. 1995, Phonological phrases: Their relation to syntax, focus and prominence, PhD diss., MIT —. 1999, “On the relation between syntactic phrases and phonological phrases”, Linguistic Inquiry 30, 219-255 Uriagereka, J. 1999, “Multiple Spell-Out”, S. Epstein and N. Hornstein (eds.) Working minimalism, Cambridge (MA): MIT Press, 251-282 van der Leeuw, F. 1997, Clitics: Prosodic studies, Amsterdam: LOT

The Word in Phonology

65

Vennemann, T. 1993, “Language change as language improvement”, Ch. Jones (ed.) Historical Linguistics: Problems and Perspectives, London and New York: Longman, 319-344 Vigário, M. 2003, The prosodic word in European Portuguese, Berlin and New York: Mouton de Gruyter —. 2010, “Prosodic structure between the Prosodic Word and the Phonological Phrase”, The Linguistic Review 27, 485-530 Williams, E. 1981, “On the notions ‘Lexically related’ and ‘Head of a word’”, Linguistic Inquiry 12, 245-274 Zec, D. 1988, Sonority constraints on prosodic structure, PhD diss., Stanford University —. 1993, “Rule domains and phonological change”, S. Hargus and E. Kaisse (eds.) Studies in Lexical Phonology, San Diego: Academic Press, 365-405 Zwicky, A. 1977, On clitics, Bloomington: Indiana University Linguistics Club —. 1985, “Clitics and particles”, Language 61, 283-305 Zwicky, A. and G. Pullum 1983, “Cliticization vs. inflection: English ‘NT”, Language 59: 502-513

CHAPTER TWO WORD-DEFINING PROPERTIES FROM A LEXICAL APPROACH TO MORPHOLOGY* ELENA FELÍU-ARQUIOLA

1. Introduction: the place of morphology and the definition of word One of the main questions that have stimulated the development of morphological theory during the last three decades is related to determining the place of morphology in the general architecture of grammar. On the one hand, morphology has been claimed to be an independent component or module with its own units and principles. This is the case of both scholars who defend the existence of syntactic word-internal structure (DiSciullo and Williams 1987, Ackema and Neeleman 2004, 2007, Williams 2007, among others) and those who consider that derived words are “amorphous”, i.e. lack internal structure (Anderson 1992, for example). On the other hand, it has been upheld that morphological processes can be accounted for using the theoretical mechanisms of other components, mainly syntax. A paradigmatic example is represented by Lieber’s (1992) *

An earlier version of this chapter was presented at the meeting “Jornadas Sylex 2010” (University of Zaragoza, October 2010). I would like to thank the organizers for giving me the opportunity to participate, as well as the audience for their useful comments and suggestions. Remaining errors are my own responsibility. This work has been partially supported by the research project “Lexicon, syntax and morphological variation: deverbal nominalizations” (ref. FFI2008-00603), from the Spanish Ministry of Science and Innovation, as well as by the International Excellence Campus Project in Natural and Cultural Heritage, headed by the University of Jaén.

Word-Defining Properties from a Lexical Approach to Morphology

67

Deconstructing Morphology, where we find an attempt to reduce word formation to operations belonging to phrasal syntax.1 Finally, Distributed Morphology (Halle and Marantz 1993, Marantz 1997) combines a syntactic approach to morphology with the existence of a post-syntactic morphological component whose nature can be described as interpretative (Fábregas 2006).2 In fact, the debate about “what is the place of morphology?” is implicitly related to one of the main topics in this volume: what is a word? (cf. Spencer 1991: 453). The main question here is whether words have morphological, syntactic or semantic properties that make them different from phrases and sentences. The main claim in this chapter is that one of the points in which these two questions meet is the notion of lexical integrity. Lexical approaches to morphology consider that morphological processes are different (to a greater or lesser extent) from syntactic operations. This implies recognizing the existence of some properties, most of them related to the notion of lexical integrity, which would characterize words in opposition to phrases. This relationship between the definition of word and the place of morphology through the notion of lexical integrity will be the core of this paper, which is organized as follows: in section 2 I summarize a number of word-defining properties from a lexical approach to morphology. At the same time, I discuss the notion of lexical integrity from an empirical as well as theoretical point of view. In section 3 I present recent revisions of the Lexical Integrity Hypothesis and I contrast cases which seem to follow this principle with other ones which do not. The final objective is to check if this notion can still be considered the foundation of morphological definitions of word and, therefore, of lexical approaches to morphology (section 4).

1

The proposal that complex words are built in the syntax was previously defended in the work of Baker (1988). 2 Similarly, Mendívil-Giró (2010) proposes a syntactic approach to morphology, in the sense that he does not recognize any difference between word syntax and phrasal syntax. However, the main discrepancy regarding Distributed Morphology lies in his consideration of the word as the central feature of human language, instead of as an epiphenomenon resulting from the interaction of syntax with the possibility of certain morpheme combinations to assume a distributional reality (Julien 2007).

Chapter Two

68

2. Two dimensions of lexical integrity The notion of lexical integrity can be understood in two ways: from an empirical as well as from a theoretical perspective. In the former case, the term “lexical integrity” encompasses a number of widely debated phenomena whereby words differ from phrases in having (or lacking) certain properties (as discussed in Bresnan and Mchombo 1995, Spencer 2005, Williams 2007, Booij 2009, Haspelmath 2011). Among the characteristics that would define words negatively in contrast with phrases and sentences, we can mention the fact that word-internal constituents lack the possibility of extraction, syntactic modification, coordination, ellipsis or anaphoric reference, as illustrated below in relation to Spanish.3 In the first place, it has been pointed out that word-internal constituents cannot be extracted by syntactic operations like relativization, topicalization or interrogative formation, as shown in (1b) regarding the second constituent of a compound: (1)

a. María ha comprado un lavaplatos ‘Mary has bought a dishwasher’ b. *¿[Qué]i ha comprado María un [lava [t]i]? ‘[What]i has Mary bought a [ti washer]?’

Secondly, word-internal constituents cannot receive independent syntactic modification. This is illustrated in (2). The example in (2a) shows that, contrary to what happens with an independent adjective, a root inside a derived noun cannot be modified by an adverb. In turn, (2b) exemplifies the ungrammaticality of an adjective modifying the second constituent of a compound: (2) a. b.

muy ebrio / ‘very drunk’ / *un lava[platos sucios] ‘a [dirty dish]washer’

*[muy ebri]edad ‘[very drunken]ness’

The third test has to do with the coordination of sublexical elements. As shown in (3), neither affixes nor members of a compound seem to undergo coordination in Spanish:4

3

Lexical integrity in Spanish has been recently addressed by Bosque (2012a, 2012b) and Fábregas (2011, 2012). See also Fábregas (this volume). 4 We will come back to this point in section 3.1.1.

Word-Defining Properties from a Lexical Approach to Morphology (3) a. b.

69

*Tienes que des- y rehacer el trabajo ‘You must undo and redo the work’ *Van a construir una casa [cuna y cuartel] lit. ‘They are going to build a [cradle and quarter] house’ ‘A nursery and a residential barrack are going to be built’

In the fourth place, constituents of complex words cannot undergo ellipsis in coordinated structures. This is illustrated in (4a) for prefixed verbs and in (4b) for compound nouns: (4) a. b.

*María congeló el pescado y Juan lo des[congeló] ‘Mary froze the fish and John de[frosted] it’ *Compré un [[lava][platos]], y María varios [[lava][frutas]] (adapted from Varela 1990: 37) lit. ‘I bought a dishwasher, and Mary some fruit[washers]’ ‘I bought a dishwasher, and Mary some finger bowls’

The last well-known phenomenon has to do with anaphoric reference. Since Postal (1969), complex words have been claimed to be “anaphoric islands” in two ways. On the one hand, it seems that a word-internal constituent cannot be the antecedent of a pronoun (outbound anaphora). This is the case of derived words (5a) and compounds (5b) in Spanish:5 (5) a. b.

*Los nuevos [[darwin]iista]s no están totalmente de acuerdo con éli (Varela, 1990: 37) ‘New darwinites do not completely agree with him (= Darwin)’ *He comprado un [[lava][platos]i] pero no losi lava bien (Varela 1990: 37) ‘I have bought a dishwasher but it doesn’t wash them (= dishes) well’

On the other hand, it does not seem possible to have anaphoric elements inside a word (inbound anaphora), as the famous English example in (6a) shows. A similar case is illustrated in (6b) for Spanish: (6) a. b.

McCarthyite vs. *him-ite (Postal 1969: 213-214) *Perón fue recibido con entusiasmo por los él-istas (él = Perón) (Piera 2009: 41) ‘Perón was received with enthusiasm by him-ites (him = Perón)’

Let us now turn to the theoretical dimension of lexical integrity. On the basis of the data presented in (1-6), some scholars have proposed a 5

See Fábregas (2012) for a discourse-pragmatic account of this phenomenon in Spanish based on the notion of shared reference.

70

Chapter Two

theoretical principle in order to restrict the relationship between morphology and syntax. In other words, this principle would prevent syntax from having access to the inside of complex words, and has received different names, e.g. the Lexical Integrity Hypothesis (LIH), Principle of Lexical Integrity or Lexical Hypothesis (cf. Jackendoff 1972, Siegel 1974, Lapointe 1980 DiSciullo and Williams 1987, Anderson 1992, Williams 2007, among others). The LIH has played a crucial role in morphological theory, and has provided the foundations to a number of approaches that defend–to a greater or lesser extent–the independence of morphology from syntax. This independence is understood in two main ways. The most radical one, prototypically represented by Anderson’s (1992) “a-morphous morphology”, rejects the existence of syntactic structure inside complex words.6 The second view sees complex words as having internal syntactic structure, although word syntax and phrasal syntax would be different and independent from each other (Ackema and Neeleman 2007, Williams 2007). Now, how do the empirical and the theoretical dimensions of lexical integrity relate? Lexical approaches to morphology upgrade an empirical observation (not without problems, as we will see in the next section) to the category of a theoretical principle turning it into one of their basic assumptions, the LIH. In contrast, syntactic approaches to morphology analyze data showing lexical integrity effects with mechanisms taken from phrasal syntax. Their aim is to demonstrate that there is no need to propose an independent morphological component or module. This is the case of the attempts to explain the impossibility of word-internal anaphoric reference summarized in Fábregas (2011). The notion of lexical integrity was central to linguistic discussion during the 70s and 80s, when lexicalism was at its height. Nowadays, after the development of syntactic approaches to morphology in the late 80s and 90s, discussing lexical integrity may be considered old-fashioned. However, in recent years, it has become a revisited topic in morphological literature, as illustrated by the work of Lieber and Scalise (2006), Williams (2007), Booij (2009) or Haspelmath (2011), among others. This revitalization probably has to do with the fact that lexical integrity is on the basis of the definition of word from a morphosyntactic point of view, and this is a timeless matter in contemporary linguistics. In the next section I discuss some of these revisions. When possible, Spanish data related to word-formation will be used in the discussion. 6

More accurately, inside inflected and derived words, since Anderson (1992) recognizes some kind of syntactic structure in the case of compounds.

Word-Defining Properties from a Lexical Approach to Morphology

71

3. Recent revisions of the Lexical Integrity Hypothesis As Lieber and Scalise (2006) point out, the LIH must be revised for two reasons. Firstly, linguistic theory has greatly evolved since its formulation more than thirty years ago. In other words, the context in which the LIH appeared has little to do with today’s theoretical scene. In the second place, the LIH was enunciated from the observation of a small amount of data. The study of new phenomena shows that it was an excessively rigid theoretical principle, since it overlimited the interaction between morphology and syntax. Among recent revisions of the LIH, I will focus on two proposals: on the one hand, the splitting of the LIH into two subprinciples (section 3.1); on the other hand, the need to allow the relationship between morphology and semantics into the discussion (section 3.2).

3.1. Splitting up the Lexical Integrity Hypothesis into two subprinciples This proposal can be found in the work of Booij (2005, 2009), who takes the following formulation of the LIH as his starting point: “the syntax neither manipulates nor has access to the internal structure of words” (Anderson 1992: 84).7 Booij (2009) points out that this wording rules out two different types of morphology-syntax interaction: on the one hand, it bans the syntactic manipulation of parts of word-internal structure, a prohibition that he considers to be right; on the other hand, it excludes the syntactic access to word-internal structure, something with which he does not agree. 3.1.1. Syntactic manipulation of parts of word-internal structure By “manipulation”, Booij (2009) refers to two types of phenomena that have already been mentioned in this paper: the extraction of word-internal constituents by means of syntactic movement, as in (1b), and the splitting of word-internal constituents, including the syntactic modification of parts of a word (see examples in (2)) as well as the coordination of sublexical elements, as shown in (3). In Booij’s (2009) views, this kind of manipulation is rightly ruled out by the LIH. However, as he also mentions, it is not difficult to find cases which seem to be counterexamples to the generalization that word-internal 7

Principle of Lexical Integrity in Anderson’s terms.

72

Chapter Two

constituents cannot be involved in coordination structures, as shown in (7) and (8) for Spanish and German:8 (7) Prefixes: a. periodo pre- y postbélico ‘pre-war and post-war period’ b. sistemas de auto- y cogestión ‘self-management and co-management systems’ (8) Bases: a. lenta y tranquilamente ‘slowly and calmly’ b. Freund- oder Feindschaft (Booij 1985: 152) ‘friendship or enmity’

Data of this kind have often been considered apparent counterexamples (cf. Booij 1985, 2009, Bresnan and Mchombo 1995), since they can receive a prosodic analysis. In short, as defended in Booij (1985) for Dutch and German compounds, this process would involve the coordination of complete words from a morphological point of view, together with prosodic deletion: “[...] a prosodic word is deleted under identity with another phonological word in the same phonological phrase” (Booij 2009: 89). Since the LIH is formulated in terms of the relationship between morphology and syntax, these examples would not contravene it. However, as Haspelmath (2011) points out, it has also been claimed that prosody is not enough in order to explain examples like those in (7) and (8), since semantic as well as morphosyntactic factors also seem to be involved. This would mean for Haspelmath (2011) that (un)coordinability is not a reliable criterion for determining if an element must be considered a word or part of a word. One of the scholars who have recently paid attention to the cases under study is Chaves (2008). He thinks that we are not dealing with the coordination of sublexical elements, but with word-part ellipsis. In this phenomenon, stems and certain affixes can be peripherally unrealized under certain conditions, in a similar way to what happens in phrasal ellipsis. Word-part ellipsis can be of two kinds, either right-periphery ellipsis (9) or left-periphery ellipsis (10), 9 as the following English examples show: 8

Similar cases have been discussed in the literature in relation to Catalan, Dutch, English, Hungarian, Italian and Portuguese, among other languages. 9 This phenomenon has also been called coordination reduction (Booij 1985); backward and forward deletion (Wilder 1997); backward and forward coordination reduction (Kenesei 2007), among other names.

Word-Defining Properties from a Lexical Approach to Morphology (9) a. b. (10) a. b.

73

Our therapists are trained in pre- and post-natal care Most students in my school are either under- or overweight Elementary Mercury is used in gold-mining and -refining Most anti-wrinkle and -aging creams have vitamin C and retinol (examples taken from Chaves 2008: 261-262).

Chaves (2008) defends that word-part ellipsis can be accounted for in terms of a deletion operation, sensitive to phonological, morphological, semantic and syntactic properties. He develops a unified account of wordpart ellipsis and phrasal ellipsis in the framework of Head-driven Phrase Structure Grammar, in a formalization in which all the constraints are expressed in a parallel and declarative fashion. And he explicitly addresses the issue of lexical integrity: his realizational account of word-part ellipsis, based in lexical derivational constructions, would be compatible with the idea that syntax cannot manipulate word-internal structure. In other words, although Chaves (2008) recognizes the interaction of different factors in word-part ellipsis, 10 as emphasized by Haspelmath (2011), his analysis does not involve a syntactic approach to wordformation. Instead, it is based on linearization constraints specified by lexical word formation rules. So, word-part ellipsis would not be inconsistent with the idea that words have certain properties which make them different from phrases and sentences. In order to illustrate the interaction of different factors in word-part ellipsis pointed out by Smith (2000) and Chaves (2008), I will examine the Spanish data in greater detail. In particular, I will focus on examples like those in (3) and (7), repeated here as (11) and (12): (11) *Tienes que des- y rehacer el trabajo ‘You must undo and redo the work’ (12) a. periodo pre- y postbélico ‘pre-war and post-war period’ b. sistemas de auto- y cogestión ‘self-management and co-management systems’

As far as Spanish prefixes are concerned,11 we see that some of them can appear in right-periphery ellipsis constructions, as illustrated in (12), 10

This interaction had been previously pointed out in Smith (2000). Spanish examples have been considered cases of ellipsis in coordination (Bosque 1987: 95-98). Against this analysis, Rainer (1993: 36) mentions the fact that the elided element does not need to be an inmediate constituent, as in exo-y [[endo[centric]]idad] ‘exo- and endocentricity’. However, an example like exo- y endocentricidad could also be interpreted as an argument in favor of an ellipsis 11

74

Chapter Two

whereas other prefixes cannot, as shown in (11). What are the factors involved in this phenomenon? In the first place, sense identity conditions have been claimed to govern ellipsis in general, and word-part ellipsis in particular (Chaves 2008: 275). As for right-periphery ellipsis, it seems to require a weak kind of semantic identity. From my point of view, the Spanish prefixes involved in it must belong to the same semantic class, as also pointed out by Varela and Martín García (1999: 5007). They can be either two temporal prefixes, as in (12a), or two prefixes with scope on the arguments of the base (Felíu-Arquiola 2003a, 2003b), as in (12b). We could think of other examples with prefixes belonging to other semantic classes, like those expressing support or rejection (13a), spatial relations (13b) or gradative relations (13c): (13) a. manifestantes pro- y antigubernamentales ‘pro-governmental and anti-governmental demonstrators’ b. mecanismos supra- e infraestatales ‘supra-state and infra-state mechanisms’ c. situaciones de sobre- e infraprotección ‘overprotection and underprotection situations’

In addition to being members of the same semantic class, the prefixes involved in right-periphery ellipsis in Spanish often show semantic opposition. However, sense identity conditions are not enough in order to account for examples like (11). Although des- and re- belong to the same semantic class, that of aspectual prefixes, they do not usually give rise to right-periphery ellipsis.12 The same is true of negative prefixes like a- and in-, which do not usually appear in word-part ellipsis, as (14) shows: (14) *actitud a- e inmoral ‘amoral and inmoral attitude’

analysis governed by phonological restrictions (the prosodic independence of the prefix) and not by structural constraints. 12 We have only attested a couple of examples from written texts, like re- y desencuentros ‘reunions and disagreements’ or des- y relocalizaciones ‘delocalizations and relocalizations’, which would require a strong contrastive stress on both prefixes. Curiously, the result is better when the base is a noun rather than a verb (*des- y rehacer el trabajo ‘undo and redo the work’). Perhaps the category of the elided element should also be taken into account, as pointed out in Varela and Martín García (1999: 5007). Unfortunately, this task exceeds the limits of this chapter.

Word-Defining Properties from a Lexical Approach to Morphology

75

What prefixes des-, re-, a- and in- seem to have in common is their impossibility to receive phonological word status (Booij 1985) or to be prosodically independent (Chaves 2008), a necessary condition in rightperiphery ellipsis. Chaves (2008: 304) describes it as follows: “rightperiphery ellipsis deletes prosodically independent items, under morph identity”, wording which combines prosodic conditions with morphosemantic ones. If we compare prefixes des-, re-, a- and in- with those which allow right-periphery ellipsis, we will notice that the former do not come from Latin or Greek prepositions, nor coincide with prepositions in Spanish. However, prefixes like ante-, anti-, post-, pre-, pro-, sobre- come from Latin or Greek prepositions and, in addition, in some cases they have the shape of Spanish prepositions (ante, sobre). Some of them, like for example anti-, have been claimed to show a behavior in between a preposition and an affix (Serrano Dolader 2002). As Varela and Martín García (1999: 5006) point out, the prefixes that take part in coordinate structures in Spanish are those which are less fused to their bases, a property that is usually represented graphically with a hyphen (manifestación pro-gubernamental ‘pro-governmental demonstration’, movimiento anti-aborto ‘anti-abortion movement’). It is even possible to find these prefixes written as an independent word (manifestación pro gubernamental ‘pro-governmental demonstration’). In addition to appearing in word-part ellipsis, they can have semantic effect on phrasal bases (manifestación pro-[legalización de las drogas] ‘pro- druglegalization demonstration’), as will be discussed in section 3.2. Due to their special behaviour, close to that of free elements, they have been described as separable, autonomous or un-bound (RAE 2009: 669). In this sense, it must be mentioned that the (im)possibility of certain affixes to give rise to right-periphery ellipsis has lead some authors to divide affixes into two subclasses (Kenesei 2007): affixoides, which would allow right-periphery ellipsis (or backward coordination reduction, in Kenesei’s 2007 terms), and affixes, which would not participate in these structures, a division that is reminiscent of that of Siegel (1974) between Class I and Class II affixes. In short, as far as lexical integrity is concerned, word-part ellipsis constructions need not be understood as radical counterexamples, since this phenomenon seems to be highly restricted.13 In addition, it does not affect every kind of sublexical element. For example, among Spanish suffixes, only -mente takes part in word-part ellipsis, precisely an element 13

A different interpretation of word-part ellipsis constructions in Spanish in relation to the LIH can be found in Bosque (2012a, 2012b).

76

Chapter Two

coming from an independent Latin noun. Rather than a counterexample against lexical integrity, word-part ellipsis seems to highlight the fact that derivative affixes are not homogeneous regarding their possibility of being prosodically independent. This difference in the degree of prosodic independence of sublexical items varies from one language to another (see also chapter 1 in this volume). Whereas Spanish suffixes show a very low degree of prosodic independence, German ones are characterized by having this property in a high degree, as the examples in (15) show: (15) a. Freund- oder Feindschaft (Booij 1985: 152) ‘friendship or enmity’ b. trink- und eȕbar (Chaves 2008: 262) ‘drinkable and edible’ c. farb- und ausdrucklos (Smith 2000: 62) ‘colorless and expressionless’

A second fact highlighted by word-part ellipsis has to do with the need of considering derivational affixes as items with some kind of meaning. It has been shown that ellipsis of sublexical elements seems to involve sense-identity conditions. So, the fact that this phenomenon is sensitive to the semantics of sublexical items would go against an “amorphous” view of derived words (Chaves 2008: 275). 3.1.2. Access of syntax to word-internal structure Let us now turn to what Booij (2009) considers to be the second subprinciple of the LIH: the banning of syntax having access to wordinternal structure. In his opinion, it is not infrequent to find cases in which word-internal structure must be accessible to syntax. An example of this fact comes from Georgian (16), as Harris (2006a) points out: (16) sam tit-moþ’r-il-i three.OBL finger-cut.off-PP-NOM ‘(a man) with three fingers cut off’

(k’aci) (Harris 2006a: 206) man.NOM

In the preceding example, sam would act as an external modifier of a sublexical element (tit). This would be a clear counterexample to one of the word-defining properties mentioned in section 2 of this chapter: the fact that word-internal constituents cannot receive independent syntactic modification (see examples in (2)). Many other phenomena pointed out in the literature can also be considered counterexamples to the idea that syntax cannot have access to

Word-Defining Properties from a Lexical Approach to Morphology

77

word-internal structure. I will mention just a few examples. Firstly, it is well-known that Germanic languages allow phrases to appear as the first element of a compound, in what has been called “phrasal compounds”, as mentioned by Lieber (1992) and, more recently, by Lieber and Scalise (2006). The English examples in (17) illustrate this kind of structure: (17) a. a pipe and slipper husband b. a slept all day look (examples taken from Lieber 1992: 11)

The second piece of evidence comes also from Georgian. As Harris (2006b) shows, in this language personal demonstrative, reflexive and interrogative pronouns can be the first element of a compound or the base of a suffix in a derivational process. In (18) we present an example of the latter case: (18) amis-iani ‘her-ite’

(Harris 2006b: 118)

In (18) we can see a demonstrative acting as a third person pronoun, in genitive form (amis), combined with a derivational suffix similar to English -ite. It is a case of inbound anaphora, that had been claimed not to be possible since Postal (1969), as discussed in section 2 (see examples in (6)). However, Harris (2006b) shows that word-formation processes in which different kinds of fully referential pronouns are involved are routine in Georgian. Finally, among other possible examples of syntax having access to word-internal structure, I will mention the case of construction-dependent morphology. As Booij (2005, 2009) shows, some syntactic constructions require the presence of a specific word-form. He mentions some cases in Dutch; for example, the inflection of Dutch numerals (Booij 2005), which in a number of particular constructions show an affix -en. Historically, it comes from a case ending, but it has been reanalyzed as a plural suffix, since it coincides formally with one of the two Dutch plural endings (-en and -s). The interesting fact is that a construction like collective adverbial (see (19)) requires the presence of suffix -en, even in numerals like zeven ‘seven’, whose plural form is zevens for prosodic reasons: (19) met zijn zeven-en (Booij 2005: 174) with his seven-en ‘the seven of us / you / them’

78

Chapter Two

Booij (2005) proposes to treat the collective adverbial expression as a constructional schema with two slots lexically fixed. In addition, in the slot for the numeral, the presence of suffix -en must also be specified, as shown in (20): (20) [[met]P [z’n [[x]Numeral -en]N]NP]PP ‘the x of us / you / them’

In this way, it is possible to capture the fact that this syntactic construction requires the presence of a specific word-form of the numeral. Data of this kind seem to support Booij’s (2009) claim that it is necessary to distinguish between syntactic manipulation of sublexical elements and access of syntax to word-internal structure. Whereas the former would be an infrequent phenomenon, the latter seems to be more frequent, it is attested in genetically unrelated languages and it involves a wide range of phenomena (for example, Germanic phrasal compounds or Georgian inbound anaphora). Any revision of the LIH would have to reflect this difference between manipulation and access. This means that the LIH would have to be worded in a way that it would rule out the first one of these two types of morphology-syntax interaction, but not the latter.

3.2. Morphology-semantics interactions Recent revisions of the LIH coincide in pointing to phenomena in which the interaction between morphology and semantics plays a crucial role (for example, the scope of derivational prefixes in Indo-European languages). However, there is no agreement on how the facts related to the semantic scope of derivational prefixes must be understood. Depending on the kind of data analyzed, semantic scope in prefixation is seen as a counterexample to the LIH or, on the contrary, as a piece of evidence for the LIH. For some scholars, like Lieber and Scalise (2006) and Booij (2009), certain cases of phrasal scope of derivational prefixes can be considered counterexamples to the LIH. In (21) we illustrate the fact, already pointed out in Rainer and Varela (1992), that some prefixes in Spanish seem to have semantic scope over a whole NP, although from a phonological point of view prefixation takes place on the head of the NP: (21) a. manifestación pro-legalización de las drogas ‘pro- drug-legalization demonstration’ b. la postguerra fría (Piera and Varela 1999: 4375) ‘the post-Cold War period’

Word-Defining Properties from a Lexical Approach to Morphology

79

The examples in (21) show the same type of structure. Although each one of the prefixes involved adjoins to the head of an NP (prolegalización, postguerra), from a semantic point of view, the prefix affects the whole NP, as illustrated in (22): (22) a. manifestación pro-[legalización de las drogas]NP b. la post-[guerra fría]NP

These examples are in fact instances of the well-known phenomenon of bracketing paradoxes, which involves a mismatch between the formal and the semantic representation of a linguistic expression. In Spanish, prefixes are frequently involved in bracketing paradoxes, even inside word limits. This is the case with relational adjectives containing prefixes, like tridimensional ‘three-dimensional’ (Varela and Martín García 1999: 43734375), that should receive two different segmentations. From a semantic point of view, the prefix tri- seems to affect the noun dimensión (23a), whereas from a morphological point of view it must be represented adjoined to the derived adjective dimensional (23b), since tridimensión is not an attested noun: (23) a. [[tri[dimension]]al] b. [tri[[dimension]al]]

A slightly different phenomenon is illustrated in (24): (24) un exfutbolista del Barcelona (Gràcia and Lieber 2002: 1307-1308) ‘a former Barcelona footballer’

As Gràcia and Lieber (2002: 1307-1308) point out, an example like (24) can be understood in two ways, given the fact that the base noun jugador expresses a function or a relation. On the one hand, ex- can affect semantically the whole NP (ex- [futbolista del Barcelona]). In this case, un exfutbolista del Barcelona would refer to someone that played in the Barcelona football team some time ago, but not anymore. However, he could still be playing in a different team. On the other hand, prefix ex- can have scope just on the noun futbolista. In this second case, un exfutbolista del Barcelona would refer to someone that is no longer a footballer, although when he was one, he played in the Barcelona team. Another phenomenon involving prefix ex- is shown in (25): (25) mi excoche ‘my ex-car’

80

Chapter Two

As can be seen, although the prefix adjoins to the noun coche ‘car’, it affects semantically the relationship expressed by the possessive mi ‘my’. Therefore, mi excoche ‘my ex-car’ refers to an entity that is still a car, but which is no longer mine, as pointed out in Gràcia and Lieber (2002: 1309). Lieber and Scalise (2006) as well as Booij (2009) believe that data of this kind would show the accessibility of word-internal structure of complex words not only to syntax, but also to semantics. Therefore, they propose that any reformulation of the LIH would have to allow the access of rules responsible for the semantic interpretation of phrases to the internal structure of complex words. From my point of view, however, cases like those illustrated in (21-25) need not be understood as counterexamples to the LIH, although they indeed show the complex interaction between morphology and semantics. In fact, this dissociation between the formal locus of prefixation (the formal derivational base) and the element on which the prefix has scope seems to be very common in Spanish and in other Indo-European languages. In a VP like rehacer el trabajo ‘redo the work’, the prefix readjoins to the verbal base hacer, although it has semantic scope over the verb and its internal argument (re- [hacer el trabajo]). Precisely the fact that prefixes can have scope on their bases’ arguments (but not on adjuncts) is mentioned in Williams (2007) as an argument for the LIH. Williams (2007) pays attention to the fact that the semantic scope of derivational prefixes seems to be more limited than that of syntactic modifiers. He illustrates this difference with the well-known contrast between the prefix re- and the adverb again in English, shown in (26): (26) a. John again washed the dishes on Tuesday (ambiguous) b. John re-washed the dishes on Tuesday (unambiguous)

Sentence (26a) is ambiguous because again can have scope on the PP on Tuesday or not. However, in sentence (26b), prefix re- cannot include on Tuesday in its scope. As Williams (2007: 355) states it, “(a) means that a dish washing precedes the event announced in (a), but not necessarily a Tuesday dish washing. (b) on the other hand is ambiguous, in that on Tuesday can be a part of the presupposition associated with again”. The reason for this difference would lie in the fact that prefixes can only have scope on the arguments of the base to which they adjoin, but not on adjuncts. Derivational prefixes are elements of the word system– morphology in Williams’ terms–, whose scope cannot go beyond lexical boundaries. Arguments of a predicate are part of the information that is present in a lexical entry. Therefore, prefixes can have semantic scope on

Word-Defining Properties from a Lexical Approach to Morphology

81

them, but not on adjuncts, which are not codified in the lexical entry of the base. In my opinion, in Spanish it is possible to find other cases in which derivational prefixes seem to have a more limited semantic scope than syntactic expressions. The first example I would like to mention is that of quantitative prefixes (mono-, uni-, di-, bi-, multi-, etc.), which are common to many other Indo-European languages given their Greek and Latin origin. Let us compare the semantic scope of quantitative prefixes with that of syntactic quantifiers, taking the examples in (27) as a point of departure: (27) a. Dos aviones sobrevolaron un poblado indígena ‘Two planes flew over an indigenous village’ b. El bimotor sobrevoló un poblado indígena ‘The twin-engined plane flew over an indigenous village’

As is well-known, in a sentence like (27a), dos quantifies the variable aviones. But, at the same time, it can have scope on other quantifiers in the same sentence. Therefore, (27a) is ambiguous, and its interpretation depends on scope relations. If un has scope over dos (un poblado indígena > dos aviones), then we obtain the following interpretation: ‘two planes flew over an indigenous village (the same village)’. On the contrary, if dos has scope over un (dos aviones > un poblado indígena), then the interpretation obtained is the following: ‘two planes flew each over an indigenous village’. However, quantitative prefixes do not give rise to scope effects, since they can only affect semantically the variable that they directly quantify, but they cannot extend their scope outside the prefixed word. Thus, in (27b) prefix bi- quantifies the noun base motor, but it cannot establish scope relations with other quantifiers that appear in the same sentence (Felíu-Arquiola 2003c). Summing up, it seems that, as far as Spanish quantitative prefixes are concerned, there is a clear difference between the semantic scope of these sublexical elements and that of syntactic quantifiers, being the former strictly limited to the word level. I would like to mention another example of the fact that the semantic scope of derivational prefixes in Spanish seems to be more restricted than that of syntactic expressions. This is the contrast between prefix co- and syntactic modifiers like the adjective juntos ‘together’, the adverb conjuntamente ‘jointly’ and the PP entre los dos ‘between the two’, illustrated in (28):

82

Chapter Two

(28) a. María y Juan dirigen una película juntos ‘Mary and John direct a film together’ b. María y Juan dirigen una película conjuntamente ‘Mary and John direct a film jointly’ c. María y Juan dirigen una película entre los dos ‘Mary and John direct a film between the two’ (29) María y Juan codirigen una película ‘Mary and John co-direct a film’

There is a contrast between the number of readings available for the examples in (28) and in (29). On the one hand, the syntactic adjuncts of sentences (28a, 28b and 28c) cannot have scope on the coordinated subject. Therefore, the only possible interpretation of these sentences is the group reading, whereas a distributive reading of the plurality of the subject is excluded.14 On the other hand, the sentence in (29), where we find a verb prefixed with co-, has three possible interpretations. Two of them could be characterized as group readings, whereas the third one corresponds to a distributive reading, as shown in (30) (Felíu-Arquiola 2003b): (30) i. Group readings: a. ‘Mary and John direct a film one with the other’ (explicit symmetry) b. ‘Mary and John direct a film together with at least some other person’ (non-distributive implicit symmetry). ii. Distributive reading: ‘Mary and John direct each one a film with at least some other person’ (distributive implicit symmetry).

Verbs formed with prefix co- in Spanish predicate an event of a plurality, whose members relate to each other symmetrically. This plurality can be satisfied explicitly in different ways: with a coordinated 14

Some members of the audience at the meeting “Jornadas Sylex 2010” (University of Zaragoza, October 2010) mentioned the possibility of having a distributive reading of the example in (28b). For them, the sentence María y Juan dirigen una película conjuntamente could also be understood in the following way: ‘Mary and John direct each one a film with at least some other person’, in a distributive implicit symmetry reading like the one described (30ii). Therefore, the adverb conjuntamente would behave similarly to prefix co-. However, Spanish native speakers seem to differ in our grammaticality judgements, since for us, as well as for many people we have asked, this distributive interpretation is ruled out in the case of conjuntamente. Likewise, a non-distributive implicit symmetry interpretation of this sentence (like the one described in (30b)) is also impossible for us. Thus, María y Juan dirigen una película conjuntamente cannot mean ‘Mary and John direct a film together with at least some other person’.

Word-Defining Properties from a Lexical Approach to Morphology

83

subject (31a), with a morphological plural (31b) or with a con-PP (31c). This is called “explicit symmetry”: (31) a. María y Juan codirigen una película ‘Mary and John co-direct a film (one with the other)’ b. Mis amigos codirigen una película ‘My friends codirect a film (one with the other or others)’ c. María codirige una película con Juan ‘Mary co-directs a film with John’

However, sentences in (31a) and (31b) can also receive a different kind of group interpretation. In this second case, the plurality required by the verb codirigir is not satisfied by the coordinated subject or by the morphological plural subject. Instead, an unexpressed con-PP is interpreted, as shown in (32), in a “non distributive implicit symmetry” reading: (32) a. María y Juan codirigen una película (con al menos otra persona más) ‘Mary and John direct a film together (with at least some other person)’ b. Mis amigos codirigen una película (con al menos otra persona más) ‘My friends codirect a film (with at least some other person)’

This non-distributive implicit symmetry reading can also be obtained when the subject is singular, as in (33): (33) María codirige una película ‘Mary directs a film with at least some other person’

Finally, a sentence like (29) can also receive a distributive reading, as mentioned before. In this interpretation, which we call “distributive implicit symmetry”, there are as many predicates codirigir una película as members in the plurality, as can be seen in (34): (34) a. María y Juan codirigen una película ‘Mary and John direct each one a film with at least some other person’ b. Mis tres amigos codirigen una película ‘My three friends direct each one a film with at least some other person’

What is relevant for the present discussion is the fact that the presence of prefix co- does not exclude a distributive reading of the subject, contrary to what happens with syntactic adjuncts like juntos, conjuntamente or entre los dos. Once we have a distributive reading of the subject, prefix co- forces us to understand the external argument of the

84

Chapter Two

distributed predicate as a plurality, although only one member of that plurality is explicitly expressed. As Williams (2007) claims in order to explain the contrast between re- and again in English, it seems that the scope of prefix co- is fixed from the lexicon and is limited to the external argument of the base predicate. Summing up, as far as the semantic scope of derivational prefixes in languages like Spanish is concerned, different phenomena can be pointed out. On the one hand, the scope of some Spanish derivational prefixes seems to be variable: sometimes the prefix can have semantic scope just on its formal base (ex[futbolista] del Barcelona ‘Barcelona ex-footballer’), whereas in other cases a whole phrase can be semantically affected by the prefix (ex[futbolista del Barcelona] ‘a former Barcelona footballer’), although it is formally adjoined to the head. However, this kind of mismatch between the formal representation and the semantic representation of a morphologically complex expression is reminiscent of cases of bracketing paradoxes, that can also be identified inside word limits (tridimensional ‘three-dimensional). In addition, many of the prefixes with phrasal scope coincide with those that give rise to word-part ellipsis. This coincidence shows that derivational prefixes in Spanish are not homogeneous regarding their phonological independence and their semantic scope. On the other hand, the semantic scope of other prefixes like English reor Spanish co- seems to be lexically fixed, in such a way that it cannot go beyond the word limits, or maybe beyond the limits of the lexical entry, since they can affect arguments of a predicate, but not adjuncts. In this aspect, the semantic scope of these prefixes contrasts sharply with that of syntactic adjuncts. All these phenomena show that the semantic scope of derivational prefixes cannot be understood simply as a counterexample to the LIH, and that the interaction between word formation and semantics requires more research.

4. Conclusions 4.1. Conclusions on the notion of lexical integrity and on the notion of word Some conclusions can be drawn from previous sections regarding lexical integrity in its two dimensions–empirical and theoretical–as well as regarding the notion of word.

Word-Defining Properties from a Lexical Approach to Morphology

85

Let us begin with the empirical dimension of lexical integrity. In spite of the brevity of the overview in section 3, it has been shown with data from different languages that lexical integrity effects can actually be found, although not generally. As Booij (2009) claims, the access of syntax to the internal structure of complex words (phrasal bases, phrases inside compounds, lexical coreference, construction-dependent morphology) seems to be more frequent than syntactic manipulation of word-internal constituents (for instance, syntactic extraction of sublexical elements). In addition, phenomena related to lexical integrity vary from one language to another. For example, inbound anaphora is not possible in English or Spanish, but it can be found in Georgian (Harris 2006b), as mentioned in section 3.1.2. The empirical dimension of lexical integrity relates to one of the questions taken as a point of departure in this chapter: “what is a word?”. It has often been claimed that it is impossible to find morphosyntactic criteria–including those related to lexical integrity–which are crosslinguistically valid in order to define the notion of “word” (see Haspelmath 2011, for a recent proposal in this sense). However, it has also been claimed that counterexamples or exceptions to lexical integrity coexist with the fact that word-defining criteria related to word-internal cohesion and order “fulfill in an extraordinarily high number of unities that we tend to call words” (Piera 2009: 41, my translation). Summing up, there seems to be a tendency for languages to show lexical integrity effects, but neither in the same degree nor in relation to the same kind of phenomena. Therefore, from a lexical approach to morphology, words can be considered as items that tend to show lexical integrity properties and that tend to behave as closed semantic domains, in contrast to phrases and sentences. Indeed, morphological processes, i.e. processes that give rise to complex words, can be considered different (to a greater or lesser extent) from syntactic operations, since their outputs (words vs. phrases) tend to behave in a different way. As for the theoretical dimension of lexical integrity, it is related to the other question that was mentioned at the beginning of this chapter: “what is the place of morphology?”. If crosslinguistic data show that lexical integrity effects are not general, but may vary from language to language, it is difficult to maintain a theoretical principle like the LIH in its most radical formulations. Different proposals have been recently put forward in order to adapt the LIH so that it can account for problematic data. On the one hand, Lieber and Scalise (2006) reformulate the LIH in such a way that it may allow some interaction between morphology and

86

Chapter Two

syntax. They propose that morphology and syntax are blind to each other, except for the set of configurations presented in (35), in which some intermodular relation would be allowed: (35) a. [[XP] Y]Y b. [[XP] [Y]]Y c. [[XP] Y]X d. [[XP] [Y]]X (Lieber and Scalise 2006: 21)

/ / / /

[Y [XP]]Y [[Y] [XP]]Y [Y [XP]]X [[Y] [XP]]X

Configurations in (35) would account for cases like Germanic phrasal compounds or prefixes in coordinated structures, which would correspond to (35a). In addition, the authors state the “Limited Access Principle”, as shown in (36) (36) “The Limited Access Principle: Morphological Merge can select on a language specific basis to merge with a phrasal/sentential unit. There is no Syntactic Merge below the word level” (Lieber and Scalise 2006: 21, italics added).

This formulation distinguishes between Syntactic Merge and Morphological Merge. The former cannot apply below word level, whereas the latter can apply to syntactic objects (phrases or sentences) in some cases, which are determined on a language specific basis. In this way, Lieber and Scalise (2006) capture the fact that languages vary regarding the type of lexical integrity effects they show. This proposal draws a line between morphology and syntax, since morphological processes (Morphological Merge) and syntactic operations (Syntactic Merge) are considered to be different, although some interaction is allowed. However, in other proposals the boundaries between morphology and syntax are more blurred, as is the case of Booij’s (2010) Construction Morphology. In this approach, morphological patterns are represented as schemas, just like syntactic constructions. Syntactic and morphological constructions may be mutually dependent, a relationship that must be represented for each particular case. However, in Booij’s (2009) view, this does not imply that words are equal to phrases. He considers that the notion of word is defined basically by the property of cohesiveness,15 which results from the no manipulation part of the LIH. 15 Dixon and Aikhenvald (2002: 19) also consider cohesiveness to be one of the main characteristics of the grammatical word, together with fixed order and conventionalized coherence and meaning.

Word-Defining Properties from a Lexical Approach to Morphology

87

The absence of limits between morphology and syntax is also defended in Haspelmath (2011: 20), who believes that a way of abandoning the dichotomy morphology-syntax is to merge both of them into a realizational or constructional morphosyntax. In this sense, Haspelmath partially coincides with Booij (2009) in the adoption of a constructional approach. A representational approach is also adopted by Anderson (2005: 161165) in order to account for endoclitics. In his analysis, developed in the framework of Optimality Theory, lexical integrity is treated as a family of constraints that are usually high-ranked in the hierarchy. This highrankedness would explain the fact that in many languages words tend to show lexical integrity effects related to non-interruptibility, fixed order of constituents, and so on. However, as usual in this framework, constraints responsible for lexical integrity can be violated if other conditions require it. That would be what happens in the case of endoclitics in a language like Udi (Harris 2000), a frequently mentioned counterexample to lexical integrity. In this case, the positional constraints on person marker clitics are ranked higher than lexical integrity constraints. Therefore, these clitics can appear in word-internal positions. Harris (2006a: 227-228) herself also suggests a constraint-based approach in order to account for the fact that some languages like Georgian allow inbound anaphora and syntactic modification of wordinternal constituents. She considers that properties like the inseparability of word-internal elements, sublexical anaphora and the absence of syntactic modification of parts of words are not irrelevant in characterizing the notion of word. However, she claims that these properties are not absolute linguistic principles, but tendencies. In her opinion, this fact can be accounted for in an Optimality Theory framework as high-ranked constraints that may be violated in some languages. Harris (2006a: 227228) suggests that a typological difference between languages might exist regarding the degree of accessibility of syntax to non-nuclear elements in compounds and derived words. These morphological constituents would show a very low degree of accessibility in English, but they would be highly accessible in Georgian. However, she thinks that many more crosslinguistic studies on wordhood are needed in order to check this hypothesis and to really understand the notion of word. As can be seen in the previous discussion, there are two characteristics that almost all the proposals have in common. On the one hand, they include, in some way or another, language specific factors related to lexical integrity. This is an attempt to account for the fact that lexical integrity has to be understood more as a tendency rather than as an

88

Chapter Two

absolute universal. On the other hand, the majority of these proposals adopt representational approaches to linguistic expressions, which seem to allow intermodular relations more easily than derivational ones. For example, representational approaches allowing the parallel expression of different types of constraints (phonological, morphological, semantic) could offer satisfying tools for accounting for bracketing paradoxes involving the semantic effect of Spanish prefixes like those seen in section 3.2. Put differently, it seems easier to explain mismatches between different levels of linguistic description using representational approaches than using derivational ones.

4.2. Conclusions on lexical approaches to morphology I would like to end this contribution in the same way I started, addressing a question: Is it legitimate to adopt an approach in which morphological processes are not considered the result of syntactic operations? From my point of view, the answer to this question is “yes”, at least for two reasons. Firstly, for an empirical one: in spite of not being universal, there are many syntactic as well as semantic phenomena that do not apply equally at the word level as at the phrasal and sentential level (syntactic extraction, independent syntactic modification, anaphoric reference, semantic scope of derivational prefixes, etc.). Therefore, it is justified to consider that morphological processes can be–to a greater or lesser extent– different from syntactic operations, since their outputs do not behave identically. This idea can be found in Fábregas (2011: 16), when he states that “morphology would still be distinct from syntax because the resulting structures constitute different domains”, even if operations used for word formation were “taken from the same set of (perhaps universal) relationships and units that syntax operates with”. My reasoning flows in a different way: since words and phrases tend to show different morphologic, syntactic and semantic behaviors, it is possible to consider that they may be produced by different types of operations. Finally, the second reason that I would like to mention in order to support an affirmative answer to the question raised at the beginning of this section is of a methodological nature. As Spencer (2005: 94) puts it, “suppose there really are differences between morphology and syntax […] If we assume a ‘syntax-all-the-way-down’ approach we will never hypothesize possible independent morphological principles and therefore we will never find them”.

Word-Defining Properties from a Lexical Approach to Morphology

89

References Ackema, P. and A. Neeleman 2004, Beyond Morphology: Interface Conditions on Word-Formation, Oxford: Oxford University Press Ackema, P. and A. Neeleman 2007, “Morphology  Syntax”, G. Ramchand and C. Reiss (eds.) The Oxford Handbook of Linguistic Interfaces, Oxford: Oxford University Press, 324-352 Anderson, S. R. 1992, A-morphous Morphology, Cambridge: Cambridge University Press —. 2005, Aspects of the Theory of Clitics, Oxford: Oxford University Press Baker, M. 1988, Incorporation, Chicago: Chicago University Press. Booij, G. 1985, “Conjunction Reduction in Complex Words: A Case for Prosodic Phonology”, H. van der Hulst and N. Smith (eds.) Advances in Non-linear Phonology, Dordrecht: Foris, 143-160 —. 2005, “Construction-dependent Morphology”, Lingue e Linguaggio 4, 163-178 —. 2009, “Lexical Integrity as a Formal Universal: A Constructionist View”, S. Scalise et al. (eds.) Universals of Language Today, Amsterdam: Springer, 83-100 —. 2010, Construction Morphology, Oxford: Oxford University Press Bosque, I. 1987, “Constricciones morfológicas sobre la coordinación”, Lingüística Española Actual 9, 83-100 —. 2012a, “On the Lexical Integrity Hypothesis and its (In)accurate Predictions”, Iberia 4, 140-173 —. 2012b, “Tres variantes de la integridad léxica”, A. Fábregas, E. FelíuArquiola, J. Martín and J. Pazó (eds.) Los límites de la morfología. Estudios ofrecidos a Soledad Varela Ortega, Madrid: UAM, 115-132 Bresnan, J. and S. A. Mchombo 1995, “The Lexical Integrity Principle: Evidence from Bantu”, Natural Language and Linguistic Theory 13, 181-254 Chaves, R. P. 2008, “Linearization-based Word-part Ellipsis”, Linguistics and Philosophy 31, 261-307 Dixon, R. M. W. and A. Y. Aikhenvald 2002, “Word: A Typological Framework”, R. M. W. Dixon and A. Y. Aikhenvald (eds.) Word: A Cross-Linguistic Typology, Cambridge: Cambridge University Press, 1-41 DiSciullo, A. M. and E. Williams 1987, On the Definition of Word, Cambridge (Mass.): MIT Press

90

Chapter Two

Fábregas, A. 2006, “Historia de un problema: la morfología desde 1985 hasta hoy”, E. Felíu-Arquiola (ed.) La morfología a debate, Jaén: Universidad de Jaén, 63-92 —. 2011, “On Why Syntactic Phases cannot Account for Lexical Integrity Effects”, Lingue e Linguaggio 10, 3-28 —. 2012, “Islas y penínsulas anafóricas: gramática y pragmática”, Estudios filológicos 50, 23-37 Felíu-Arquiola, E. 2003a, “Morphology, Argument Structure, and Lexical Semantics: The Case of Spanish auto- and co- Prefixation to Verbal Bases”, Linguistics 41, 495-513 —. 2003b, Morfología derivativa y semántica léxica: la prefijación de auto-, co- e inter-, Madrid: Ediciones de la Universidad Autónoma de Madrid —. 2003c, “Los prefijos cuantificadores en español”, F. Sánchez Miret (ed.) Actas del XXIII Congreso Internacional de Lingüística y Filología Románica, vol II/1, Tübingen: Max Niemeyer, 317-330 Gràcia, Ll. and R. Lieber 2002, “Sobre el prefijo ex-”, M. D. Muñoz Núñez et al. (eds.) IV Congreso de Lingüística General, vol. III, Cádiz: Universidad de Cádiz, 1307-1318 Halle, M. and A. Marantz 1993, “Distributed Morphology and the Pieces of Inflection”, K. Hale and S. J. Keyser (eds.) The View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger, Cambridge (Mass.): MIT Press, 111-176 Harris, A. 2000, “Where in the Word is the Udi Clitic?”, Language 76, 593-616 —. 2006a “In other Words: External Modifiers in Georgian”, Morphology 16, 205-229 —. 2006b, “Revisiting Anaphoric Islands”, Language 82, 114-130 Haspelmath, M. 2011, “The Indeterminacy of Word Segmentation and the Nature of Morphology and Syntax”, Folia Linguistica 45, 31-80 Jackendoff, R. 1972, Semantic Interpretation in Generative Grammar, Cambridge (Mass.): MIT Press Julien, M. 2007, “On the Relation between Morphology and Syntax”, G. Ramchand and C. Reiss (eds.) The Oxford Handbook of Linguistic Interfaces, Oxford: Oxford University Press, 209-238. Kenesei, I. 2007, “Semiwords and Affixoids: The Territory between Word and Affix”, Acta Linguistica Hungarica 54, 263-293 Lapointe, S. 1980, A Theory of Grammatical Agreement, PhD. diss., Amherst: University of Massachusetts Lieber, R. 1992, Deconstructing Morphology, Chicago: University of Chicago Press

Word-Defining Properties from a Lexical Approach to Morphology

91

Lieber, R. and S. Scalise 2006, “The Lexical Integrity Hypotesis in a New Theoretical Universe”, Lingue e Linguaggio 6, 7-32 Marantz, A. 1997, “No Escape from Syntax: Don’t Try Morphological Analysis in the Privacy of your own Lexicon”, A. Dimitridades, L. Siegel et al. (eds.) University of Pennsylvania Working Papers in Linguistics, 4, 201-225 Mendívil-Giró, J. L. 2010, “Por debajo de la palabra, silencio. La sintaxis como interfaz y la naturaleza del léxico”, J. F. Val Álvaro and M. C. Horno Chéliz (eds.) La gramática del sentido: Léxico y Sintaxis en la encrucijada, Zaragoza: Prensas Universitarias de Zaragoza, 181-224. Piera, C. 2009, “Una idea de la palabra”, E. de Miguel (ed.) Panorama de la lexicología, Barcelona: Ariel, 25-49 Piera, C. and S. Varela 1999, “Relaciones entre morfología y sintaxis”, I. Bosque and V. Demonte (dirs.) Gramática Descriptiva de la Lengua Española, vol. 3, Madrid: RAE-Espasa Calpe, 4367-4422 Postal, P. M. 1969, “Anaphoric Islands”, CLS 5, 205-239 Rainer, F. 1993, “Setenta años (1921-1990) de investigación en la formación de palabras del español moderno: bibliografía crítica selectiva”, S. Varela (ed.) La formación de palabras, Madrid: Taurus, 30-70 Rainer, F. and S. Varela 1992, “Compounding in Spanish”, Rivista di Linguistica 4, 117-142 Real Academia Española 2009, Nueva gramática de la lengua española, Madrid: Espasa Serrano-Dolader, D. 2002, “Hacia una concepción no-discreta de algunas formaciones con anti- en español”, Revista Española de Lingüística 32, 387-411 Siegel, M. 1974, Topics in English Morphology, PhD diss., Cambridge (Mass.): MIT Smith, G. 2000, “Word-remnants and Coordination”, R. Thieroff, M. Tamrat, N. Fuhrhop and O. Teuber (eds.) Deutsche Grammatik in Theorie und Praxis, Tübingen: Niemeyer, 57-68 Spencer, A. 1991, Morphological Theory, Oxford: Blackwell —. 2005, “Word-formation and Syntax”, P. Štekauer and R. Lieber (eds.) Handbook of Word-Formation, Amsterdam: Springer, 73-97 Varela, S. 1990, Fundamentos de morfología, Madrid: Síntesis Varela, S. and J. Martín García 1999, “La prefijación”, I. Bosque and V. Demonte (dirs.) Gramática descriptiva de la lengua española, vol. 3, Madrid: Espasa, 4993-5040

92

Chapter Two

Wilder, C. 1997, “Some Properties of Ellipsis in Coordination”, A. Alexiadou and T. A. Hall (eds.) Studies on Universal Grammar and Typological Variation, Amsterdam: John Benjamins, 59-107 Williams, E. 2007, “Dumping Lexicalism”, in G. Ramchand and C. Reiss (eds.) The Oxford Handbook of Linguistic Interfaces, Oxford: Oxford University Press, 353-382

CHAPTER THREE ON A GRAMMATICALLY RELEVANT DEFINITION OF WORD AND WHY IT BELONGS TO SYNTAX ANTONIO FÁBREGAS

1. Why words? The concept of word is so deeply rooted in our linguistic tradition and is such a common sense idea, intuitively used by non-specialists in language, that it becomes crucial in any discussion about it to remind ourselves of why we are still discussing this issue. The reason why we still consider the concept worth discussing is, plain and simple, that we have not understood yet what the relation between morphology and syntax is. This reflects a more basic problem: we do not know yet what the organization of the cognitive linguistic capacity of speakers is. In other words, we do not know whether this capacity has to be divided into separate components–as the generative tradition coming from Chomsky (1957) and Fodor (1968) argues–or those divisions are arbitrary–as proposed within cognitive and constructional approaches to language (Lakoff 1962, Goldberg 1995). Even those that would prefer to divide it into components, like the author of this chapter, do not agree on how many components should be identified and, perhaps more crucially, how one interfaces with the other. To put it mildly, these questions have not received an answer that satisfies all researchers in the field. “Word” is, as a concept, still relevant because it is associated to all of these questions. Leaving aside its everyday, non-specialized use by common speakers, this concept implicitly states that morphology and syntax are two separate components of the grammar. It implies that words are different from phrases, and as such, they must receive a different name. Words are morphological objects and phrases are syntactic objects. Words are generated in the lexicon, which is the engine that a speaker uses to

94

Chapter Three

implement his or her morphological generative capacity, and phrases are generated in syntax. All these notions are automatically implied when a researcher uses the term “word” (Aronoff 1976, Selkirk 1982, Scalise 1984, Varela 1990, Anderson 1992, among many others). Ultimately, therefore, what we are doing when we discuss what a word looks like is nothing but determining whether morphological objects are different from syntactic objects and, if so, how. Understanding this is just a way–albeit a necessary one–to understand how morphology is different from syntax, if that is actually true. What is even more, understanding how a word is defined and how it behaves is necessary because we hope that this understanding will be a window into how morphology and syntax interact, or how they cannot interact, perhaps because they are one and the same and interaction is vacuous, perhaps because they use languages that are not mutually intelligible and interaction is practically zero. The study of what a word is, therefore, is the study of how grammars are organized. But these questions are useful only to frame more specific discussions, because they are so general that one runs into the danger of addressing them from a purely theoretical perspective that forgets data. Data, to be interesting, generally require some level of detail. From far away, cows and cars might look alike; it is only when one focuses enough and comes closer that relevant differences emerge. In order to focus a bit closer onto the problem, I pose two questions: (a) Is there any grammatically relevant definition of “word”? (b) In case that definition exists and we can talk about “words” as distinct units in the grammatical knowledge of speakers, how are they defined? Is the definition different enough from what we know from syntax to allow us to say that morphology is a distinct component?

At this stage it should be borne in mind that there are two main (families of) theories about the relation between syntax and the lexicon. The first one, Lexicalism (Halle 1973, Aronoff 1976, Lieber 1980, Scalise 1984, Varela 1990, Anderson 1992; see also chapter 2 of this volume), proposes that morphology is an independent component which precedes syntax, feeding it with words which are already labeled with a number of properties that syntactic structures must satisfy. This theory expects the answer to questions (a) and (b) to be affirmative in both cases: words must exist in some way and they must be different enough from syntax. The second big family of theories has been labeled Constructionism (Lees 1960, Baker 1988, Halle and Marantz 1993, Borer 2005, Starke 2009) and argues that there is no morphological component previous to

Grammatically Relevant Definition of Word and why it belongs to Syntax 95

syntax. Syntax thus determines solely all properties of any combination of units. Depending on authors and theories, there might still be a component which deserves to be called “morphology”, but this component comes after syntax and is generally devoted to determining how structures are spelled out in phonology (Embick and Noyer 2001). A theory belonging to this family clearly expects the answer to (b) to be negative: all differences, if they are real, come from independent principles that do not justify considering morphology different from syntax. Life would be easier for these theories if the answer to question (a) were also negative. However, it is true that something called “word” could exist in their world provided that what makes a word special can be explained by phonological reasons (as Embick and Noyer (2001) suggest, idiosyncratic properties of the phonological pieces used to spell out some heads that make them more cohesive than normal phrases), other syntactic reasons (as in Arad 2003, or as in the proposal I will put forward in this chapter) or due to principles related to still unexplored domains like parsing. It must also be noted that constructionist theories are typically generativist, and as such, they take the position that morphology has to be reduced to syntax. An equally economic solution would be to say that syntax has to be reduced to morphology or, rather, to the lexicon. In theories that are more centered on the semiotic character of languages such as Cognitive Linguistics and Construction Grammar, this is precisely what is done: there is no distinction between the two components because what looks like a syntactic structure is actually a (quite abstract) pre-assembled construction, stored in the lexicon, with open slots where other (less abstract) constructions can be inserted (see chapter 5 in this volume). As in constructionist models, the difference between one kind of object and the other kind depends on an orthogonal property–how (under)specified the slots in the construction are–not on an ontological property. As Williams (2007) says, from a theoretical perspective any system that has one generative component is preferred over any system with two or more generative components, because any phenomenon that can be accounted for with one engine can also be accounted for with two, three or fourteen engines; you only need to divide them arbitrarily and to claim that they are not in use at your own convenience. For this reason, any argument in favor of lexicalism is empirical: arguably, a single engine cannot explain some patterns of data, because the data themselves seem to come from different sources, as they do not behave alike, and the differences cannot be reduced to different structures or operations inside one single engine. This division of the generative capacity is precisely what was done in the seventies following a particular interpretation of Chomsky (1970):

Chapter Three

96

several phenomena where identified where words seemed to behave in a different way from phrases. This set of data, to which I will come back later, gave rise to the Lexical Integrity Hypothesis (Lapointe 1978) at the end of the decade. This hypothesis proposes that (a) words exist and (b) that they are different enough from phrases–specifically, because of a number of things that they cannot do–to sustain the hypothesis that morphology and syntax are different. It is also worth mentioning that any system where there are no separate components is in principle–at least theoretically–more economical than any system where components are differentiated. Although at this point we might start to discuss what one exactly means by economy, this explains to some extent that the (strongest) arguments used in generativism against the non-modular view of grammar argued for in cognitive linguistics are also empirical. For instance, one of the reasons why generativists think that syntax and world knowledge have to be kept apart is that some things are very easily thought but very difficult to express in a linguistic utterance. By way of illustration, it is not difficult to imagine a situation in which I am having a coffee with someone else in a cafe and I see our image in the mirror. However, it is difficult to express this event in a single sentence in Spanish: (1)

??Nos veo en us.ACC see.1SG in ‘I see us in the mirror’

el the

espejo mirror

Conversely, some things are easier to say than to imagine, as the famous Chomsky (1965) example reproduced in (2). (2) Colourless green ideas sleep furiously

Data like these argue in favor of differentiating between two components, simply because there is no direct correlation between what one sees in the world and what one is allowed to say, given his or her grammar. This last consideration takes us to the last point which should be mentioned in this introduction. It is very easy to make arguments in favor or against some analysis and only consider analyses that belong to one particular family; one that argues, from constructionism, that two modules are worse than one sometimes forgets that there are other linguistic theories that make the same point in order to deny the whole system that one is presupposing. I will do my best to avoid this shortcoming, and whenever possible and relevant I will try to relate my points, framed inside

Grammatically Relevant Definition of Word and why it belongs to Syntax 97

a generative model, to other theories–like Cognitive Grammar or Construction Grammar–that are not in my closest family but seek to answer the same questions that I am presenting here (see chapters 5 and 6 in this volume). This chapter is divided into two separate parts in keeping with the two questions that need to be addressed here. The first one, developed in sections 2 and 3, shows that there is a grammatically relevant definition of word, contrary to what has been claimed in some constructionist literature (for instance, Arad 2003). The second one discusses whether the notion of word presented in section 3 can be defined in a syntactic way; if that were the case, then the existence of words would not mean that morphology is independent from syntax. In sections 4 and 5, I will argue that this is the case; the notion of word is directly related to the presence of a syntactic constituent in the internal structure. This constituent becomes immediately deactivated by merge with the head. This means that the existence of words follows from syntactic principles. This makes predictions with respect to how languages would behave: if and when a structure in a language has elements which are rendered inactive by first merge, a word will be defined. It should be emphasized that the argument predicts that all languages have words to the extent that in all languages syntax is assumed to exist and syntax is a system that deactivates active features in order to allow structures to be processed by the interfaces.

2. What words are not When finding a definition of word that is grammatically relevant, it is important to remember that labeling is not classifying. The main difference is that, when one has organized objects from a domain into relevant classes, it is possible to discover other properties of the objects that fall into each one of these categories. For this reason, we expect that if we found a relevant definition of word, objects that are classified as such will share some significant properties with each other, and those properties will help us to shed some light into the nature of grammar. The same happens when animals are classified into species: those falling into the same species share some relevant properties in their physiology and behavior. If, on the other hand, what we did is just to give labels to objects, then we cannot expect those objects to share some properties with each other, in much the same way as we cannot expect all people named Stuart or Hassan to behave in the same manner or look alike. Sciences try to classify, not just to name.

Chapter Three

98

So let us begin with a couple of definitions of word that do not give us a grammatically relevant classification. The oldest definition, and perhaps the most intuitive for alphabetized speakers, is the orthographic definition. This roughly states that a word is anything separated by two blanks in the written form. This definition has some obvious shortcomings. First of all, it does not afford us a classification, because the objects falling into each class do not share properties beyond, axiomatically, that of being written with two blanks. Consider (3). In both Spanish expressions, we have a nominal base and a prefix with a certain aspectual meaning. Following this criterion, however, in (3a) the prefix and the base only form one word, while in (3b) the prefix forms two separate words. (3) a. b.

expresidente ex-president ‘former president’ ex presidente de Francia ex president of France ‘former French president’

The way (3) is written comes from a particular set of rules that a speaker has to learn in a specialized way. People like me, born before 1999, had to learn this rule way after our language acquisition years. This takes us to the second shortcoming in this definition: it is based on an artificial set of rules that change over time; rules that might not be learnt by all speakers regardless of their oral competence and that might be missing in some languages with no written tradition. The official orthography of Spanish in 2010 states that prefixes have to be written as single words with their bases (as in (3a)) except when the base has the orthographic shape of a phrase (as in (3b)). This rule has changed from previous rules, which basically always preferred the prefix written as part of the base, and may well change again in the future. The problem is that, even if this criterion gave us some natural class (which it actually does not), that natural class could not be explained directly by this criterion, because it is artificially stated as part of the cultural tradition of a language. Therefore, we should look for whatever motivated the orthographic criterion in order to determine how those natural classes were conformed in the first place. Another problem is that relying too much on this artificial form of representing natural languages might force us to reach funny typological conclusions which, when the orthography is factored out from the analysis, are simply not true (see chapter 4, this volume).

Grammatically Relevant Definition of Word and why it belongs to Syntax 99

Another definition of word, which is more integrated in linguistic studies, is the phonological definition of word. Something is considered a word when it complies with several properties like carrying primary word stress, being pronounced without natural boundaries, etc. (see chapter 1, this volume). This definition is certainly useful to determine certain interesting issues about the phonology of natural languages, but its shortcoming is that it lumps together things that do not form a natural class with respect to the rest of properties. Consider (4a), taken from Marantz (2001), and its counterpart in (4b). (4) a. b.

d’yawanna? do you want to?

(4a) would be a word according to this definition, while (4b) would not be a word. The problem is that both segments reproduce the same structure and that, if one ignores the differences following from their phonology, the two behave in the same way. Of course, we know that we can insert a negation between the verb and to in (4b), but this is not possible in (4a). This presumably follows from the phonological properties, namely that there is no phonological space in (4a) because the boundaries between the two elements have already disappeared. However, in both cases we have the same number of projections: in a conservative (and generative) estimate, at least one verb, a couple of complementizers and a tense node. Both structures seem to be able to license the requirements for the pronominal subject; both structures have do insertion; both structures can introduce an infinitive; both structures have a volitional meaning, etc. Another well-known attempt at defining words has been to use some morphological properties. A word would be any unit which is morphologically marked in some specific way; this morphological mark has to be overt. For instance, it has been proposed that desinences and theme vowels act as these kinds of markers, closing a structure and defining it as a word. This criterion, however, suffers from an important problem if applied in a systematic way: it makes very similar things to be categorized into different classes. For instance, the objects in (5) would be words, because they have theme vowels and desinences, while those in (6) would not be, despite the very obvious similarities that we would be unable to capture. (5)

a. buen-o good-desinence b. parec-e-mos seem-THV-1PL

Chapter Three

100 (6)

a. buen good b. so-mos be-1PL

It could also be objected that the morphological criterion is not valid because, by definition, it does not give us any universal classification of objects. Clearly, in a language like English, which lacks desinences and theme vowels, the criterion would have to be reformulated in such a way that it can be applied to English morphology. However, this critique might not be fair, since nothing is said about words being relevant for all languages in the world, which may well be a parameter. In the same way as some languages use topichood to license pro drop (like Japanese) and other languages use some form of agreement (like Spanish), the existence of units called “words” could be a further difference. Finally, another definition of word that probably will not take us too far is the one that bases wordhood on expressing a single semantic concept. This has been used as a criterion, especially when differentiating compounds from phrases. Consider the classic German example in (7). This example has been claimed to be a compound because, instead of expressing two separate concepts–‘big’ and ‘father’–it expresses a single concept which is not denoted by any of its parts in isolation. (7) Gross-vater big-father ‘grandfather’

This criterion, which reduces wordhood to non-compositionality, might be right. The main problem nevertheless is that, at this point, we cannot really tell since, in order to determine whether something is a word or not, we will have to determine first what counts as a simple concept and what does not. Since Generative Semantics, we are aware that what is decomposable into primitives for one researcher might not be decomposable into primitives for another one. Is bachelor really composed of not, married and adult? Is kill really composed of cause, not, to be and alive? Are we positive that adult should not be decomposed into more basic concepts? What about not which in standard analysis contains an existential quantifier? Where do we stop? Without an exhaustive and testable theory about the ontology of semantic primitives, this criterion will not take us too far. Moreover, this criterion forces us to accept that perfectly compositional units such as that in (8) cannot be considered words. And yet they do not

Grammatically Relevant Definition of Word and why it belongs to Syntax 101

seem to be very different from units with a non-compositional meaning when the other criteria discussed in this section are applied. (8) corre-dor run-er ‘runner’

3. What a word could be: the Lexical Integrity Hypothesis As already mentioned, the criterion that has proven to be more useful to define a word and the one that most researchers have adopted has to do with the syntactic properties of units, where “syntactic” is perhaps a cover term that refers to the properties that result from combining distinct units into structures. Some structures, called phrases, can take part in a variety of syntactic processes, while other structures, called words, never take part in the same processes. These contrasts are exemplified in the pairs in (9), (10) and (11). These illustrate the three major phenomena that have motivated the so-called Lexical Integrity Hypothesis. (9) Extractions a. ¿Quéi has dicho que [limpia ti]Juan? what have.2SG said that cleans t Juan? ‘What did you say that Juan cleans?’ b. *¿Quéi has dicho que Juan es [limpia-ti]? what have.2sg said that Juan is [clean-t]? *‘What did you say that Juan is a t-polisher?’ (10) Pronominal coreference a. He comprado [unos platosi], pero no me gustan proi. have.1sg bought some plates, but not me.dat like pro ‘I bought some platesi, but I don’t like themi’ b. *He comprado un [lava-platosi], pero no losi lava bien Have.1sg bought a [wash-dishes], but not them washes fine *‘I have bought a dish-washer, but it doesn’t clean them properly’ (11) Ellipsis a. Juan lavai platos, así que yo ya no tengo que [hacer-lo]i Juan washes dishes, so that I anymore not have to do-it. ‘Juan washes the dishes, so I don’t have to do it anymore’ b. *Esto es un [lavai-platos], así que yo ya no tengo que [hacer-lo]i this is a wash-dishes, so that I anymore not have to do-it. *‘This is a dish washer, so I don’t have to do it anymore’

Other phenomena (absence of modification or expansion of a constituent internal to a word, absence of coordination, absence of internal inflection, etc.) were identified as criteria that would differentiate between

102

Chapter Three

words and phrases, but later these generalizations were shown to be just tendencies. A number of counterexamples have also been attested (Lieber and Scalise 2006). Instead, the three phenomena presented in (9)-(11) seem empirically robust. The phenomenon in (9) shows that some structures are islands for extractions, while the phenomenon in (10) shows that the same kind of structures are also anaphoric islands (Postal 1969); that is, they do not allow a pronoun to make reference to one of their internal constituents. Some cases have been argued to show that not all structures looking like words are anaphoric islands (Lieber 1992) such as the example in (12). (12) Reagani-ites no longer believe in himi

However, it can be shown–following a widespread tradition in the last 20 years (Montermini 2006 and references therein)–that in these cases the referent of the pronoun is a discoursively salient participant, not the base of the word. We will summarize some of these tests to show that the phenomena are robust. First, as expected from the theory where the antecedent of the pronoun is not the base of the word, we do not have Ccondition infractions as shown in (13). (13) Hei no longer believes in Reagani-ites

The pronoun in (13) can be interpreted as ‘Reagan’; this would not be possible if the consequent of the pronoun is the proper name embedded in the word since this is a referential expression that cannot be bound by a preceding pronoun. Secondly, only strong pronouns that are able to contrastively shift topic can take part in this construction. Clitics, which in Spanish cannot be contrastive, are out in this construal, as shown in the contrasts in (14). (14) a. *Los berlusconii-anos no loi creen the Berlusconi-ans not him believe ‘Berlusconians don’t trust him’ b. Los berlusconii-anos no creen en éli the Berlusconi-ans not believe in him ‘Berlusconians don’t believe in him’ c. ??Los berlusconii-anos dicen que proi es inocente the Berlusconi-ans say that he is inocent ‘Berlusconians say that he is inocent’ d. Los berlusconii-anos dicen que éli es inocente the Berlusconi-ans say that he is inocent ‘Berlusconians say that he is inocent’

Grammatically Relevant Definition of Word and why it belongs to Syntax 103

The difference between (14a) and (14b) is that in the first case the pronoun intended to refer to ‘Berlusconi’ is weak, while in the second case it is strong. The contrast in (14c) and (14d) is that in the second case we have an overt subject, which in Spanish is typically interpreted as a contrastive topic, while in (14c) the empty subject cannot shift topic. Thus, we can see that the apparent counterexamples are not cases in which the pronoun accesses a constituent internal to the structure. Now we have a relevant contrast that we can use to define words as opposed to phrases: the internal constituents of the former do not allow displacement outside and they are not eligible for pronominal coreference.

3.1. Is the contrast enough? Three kinds of cases Remember that it is not enough to show that some unit can be defined as a word; it is also necessary to show that it is different enough from phrases to justify that we have a classification (as opposed to some naming decision). Indeed, the strategy adopted by some constructionist researchers to argue that the data in (9)-(11) cannot prove that words are special domains has been to analyze them as phrases and to show that they share the properties of phrases. Indeed, not all syntactic phrases allow for extractions of constituents. Syntactic islands exist, and in order to show that cases in (9) are enough to argue for the existence of words, it is necessary to argue that they cannot be analyzed as syntactic islands. This means that, before we proceed any further, it is necessary to refine the data that I have just presented. In the case of absence of movement, two questions immediately arise: (a) Are affixes unable to undergo movement from inside a word due to their phonological properties? If morphemes are phonologically weak, they, just like clitics, will have to be pronounced in fixed positions (attached to a base, in this case) to compensate for their phonological defectivity. Equally, moving away the base without the affixes would be blocked because in this case too the affixes will surface separated from their base. (b) Are cases in which an element cannot be extracted from a word any different from syntactic phrases that do not allow extraction of one of their constituents? If a morpheme cannot be extracted from a word, but the same relation between two elements in the syntax also prevents the elements from moving, then there is no real contrast.

With respect to the first question, it seems to be true that some of the cases where the extraction of a morpheme does not seem to be possible should be removed from the set of data that argue for words as opposed to

104

Chapter Three

phrases. This is so because the absence of extraction is only apparent and presumably due to phonological conditions. If morphemes are in principle extractable from a word and they only fail to be so because of their phonological properties, the immediate prediction is that these units can undergo movement, but do not show it in the phonological shape of the utterance. In other words, these units will undergo covert movement (May 1985). One such case might have been identified in the case of prefixes. Consider the pattern of data in (15) taken from Fábregas (2005). (15) a. Comunicación interdepartamental en la universidad communication interdepartamental in the university b. *Comunicación interdepartamental entre profesores communication interdepartamental between teachers

In (15a), the adjective interdepartamental is interpreted as an argument of the deverbal noun (‘communication between departments’), but in (15b) it must be interpreted as an adjunct, as the arguments of the noun are expressed by the PP entre profesores. The explanation is simple if we consider that in both cases the prefix inter-, which has a reciprocal meaning, needs to take scope over at least two participants; this is only possible if it takes scope at LF over the deverbal noun as in (16). (16) [XP [inter]i X [NP comunicación [YP ti departamental]]]

From an adjunct position, however, the constituent that the prefix has to escape at LF forms an island. Thus, (15b) is ungrammatical because the prefix cannot license its semantic conditions at LF as it cannot take scope over the noun that introduces the participants. This case shows that movement from some words is possible, provided that it is not reflected in the phonological representation. This strongly suggests an explanation similar to that of Embick and Noyer’s (2001). These authors suggest that the main property of affixes (as opposed to non-affixes) is that they have special phonological properties. In other words, the difference between the free form in (17a) and the prefix in (17b) would be phonological; the latter cannot reorder in the linear chain because of some phonological defective status. (17) a. entre between b. interinter

Grammatically Relevant Definition of Word and why it belongs to Syntax 105

If this phonological side is included in the explanation, many of the extractions that are impossible will have to be reconsidered to determine whether what is impossible for them is the new linear order adopted or whether the ungrammaticality is due to syntactic reasons. This, of course, does not explain why none of the two elements inside a compound can be moved away from inside the word. This is so because, trivially, the phonological shape of a compound constituent is generally identical to the one it adopts as an independent word (pace the relative weight of the stress assigned to the root). If the root itself is phonologically licensed as a single word, nothing suggests that it will not be licensed inside the compound. (18) perro policía dog police ‘police dog’

This takes us to the second issue. In some cases, extraction from inside a word is not different from syntax, because the relation between the two compound constituents, when manifested in a clearly syntactic structure, also disallows movement. In a compound formed by two nouns (19a) in which the second is an attribute that restricts the class denoted by the first, extraction is impossible (19b) because the same appositional structure in the syntax (19c) prevents extraction (19d). There is no difference between the syntactic structure and the alleged morphological one; both could be the same and this does not argue for a grammatically relevant definition of word. (19) a. Luis es un hombre anuncio Luis is a man advertisement ‘Luis is an ad-man’ b. *[anuncio]i, Juan es un hombre ti c. Vimos la ópera Fidelio we.saw the opera Fidelio ‘We saw Fidelio, the opera’ d. *[Fidelio]i, vimos la ópera ti

But there is yet a third class of cases. This is when the absence of extraction is not explained by the phonological defectivity of any of the morphemes and the same structure, when built in the syntax, allows for movement. This was the case in (9). Even if it is generally possible to extract a bare NP from the object position of a verb, the same relation

106

Chapter Three

between a verb and a noun does not allow for extraction when the whole is a word. To recapitulate, I have differentiated between three cases where a word-internal component seems not to extract: (i)

Cases where the constituent shows movement effects at LF but do not reflect them in the phonology. In these cases, it is plausible that movement is possible, but the phonological properties of the affix do not allow it to be reflected (ii) Cases where the constituent cannot be extracted, but where there is independent evidence that, even if the construction were syntactic, movement would be impossible (iii) Cases where movement is impossible and the syntactic equivalent allows for movement. This third case is the one that properly constitutes a problem for a syntactic take on words

3.2. Are words syntactic phases? One option to analyze the cases in (iii) in a way that they do not argue for the autonomy of words would be to say that words introduce syntactic domains by default, while phrases do not. In modern Minimalism, syntactic domains are phases (Chomsky 2001); that is, sets of structure which are complete and must thus be transferred to the phonological and semantic interfaces. More specifically, the proposal is that at particular points in the tree, a part of the structure created by merge has already satisfied all its formal properties. At that point, syntax does not have any further operations to perform in that chunk of structure, and therefore, it transfers it so that phonology and semantics can interpret it. As that part of the structure has abandoned syntax, any subsequent operation will not be able to manipulate the transferred structure, simply because it will not find it in the computational system. Marantz (2001), and more clearly, Arad (2003: 748 ff.) propose to extend this notion of phase to the word domain. Their reasoning is as follows: a root does not have a defined grammatical category (20a). It will not be able to be transferred to the interfaces on its own, because the lack of a grammatical category entails that the semantic component will not be able to assign an interpretation to it. However, in a further step, the root combines with a functional head and at this point it receives a grammatical category (20b). The root has thus satisfied its formal requisite of getting a grammatical category and the semantics can interpret it (at that point, the phonology will also be able to know which allomorph will be required to

Grammatically Relevant Definition of Word and why it belongs to Syntax 107

spell out the root in that context; for example, in a pair such as divide ~ divis-). It is then when it gets transferred. (20) a. [¥] b. [xP [¥]]

In Arad (2003), this has consequences for the further steps of the derivation. The material contained inside a phase is transferred to the phonology and to the semantics. From here it follows that such material will be processed by these components as a unit, and thus, that it will be possible to assign special semantics or phonology to the chunk as a whole. The interest of this step is that it is used to explain the semantic and phonological idiosyncrasies found inside the word domain, as pointed out at least since Chomsky (1970) and Halle (1973). Compare the two sentences in (21), taken by Arad (2003: 756) from Kiparsky (1982, 1997). (21) a. He hammered the nail with a rock b. *Screw the fixture to the wall with nails!

In (21a), the verb to hammer can describe an action performed by using a rock, instead of a hammer; in (21b), on the other hand, the verb to screw cannot describe an action which involves nails; the action must be performed by using screws. Arad proposes that this difference can be explained if the set formed by the root and its categorizer forms a phase. Her proposal is that interpretation is assigned to a root in the context of the first categorizer, and once this has been done, this interpretation cannot be changed by the further steps of the derivation (Arad 2003: 747). Let us see an illustration of what this proposal can account for. In (21a), the verb is constructed over a root ¥HAMMER, and in that way the verb gets its meaning defined idiosyncratically, which in this case means that the action does not need to be performed with the object that we call ‘a hammer’ (22a). In (22b), by contrast, the verb is formed over the noun screw, not over the root ¥SCREW. The verbal head combines with an already complex object, formed by a zero nominalizer and the root ¥SCREW, as represented in (22b). By the time the verb is constructed, its base, defined as a phase because it contains a root and a categorizer, has a specific meaning that the subsequent steps in the derivation cannot change. Consequently, the verb to screw must refer to an action which involves the objects we call ‘screws’. (22) a. [v [¥HAMMER]] b. [v [n [¥SCREW]]]

108

Chapter Three

To summarize the argument, Arad’s proposal is that the root and the first categorizer form a phase, and argues that the question of whether a word has an idiosyncratic semantics (or phonology) depends on whether the word itself is a phase—in which case idiosyncrasies are possible, as in (21a/22a)—or whether it contains a phase, in which case the meaning of the word must be constructed on top of whatever meaning was assigned in the previous step of the derivation, forcing the verb to denote an action which involves the particular object interpreted in the base (21b/22b). Imagine that a word is a phase. Within that proposal the absence of extraction of word constituents can be explained just as the failure to extract the interrogative from a conditional sentence. (23) reproduces the structure of life saver following Harley’s (2009) analysis in the Distributed Morphology framework. Notice that the word contains two phases in Arad’s sense, one formed by the constituent life and the first nP, which turns it into a noun, and a second one that dominates the first and contains life and the second nP. (23) [nP -ern [¥P SAVE [nP øn [¥P LIFE]]]]

If we assume this construction, the fact that life or an interrogative occupying its position cannot be extracted from inside the compound by movement can be explained by the notion of phase, in the sense of Arad. The constituent life is contained inside the phase defined by the categorizer -ø, which in turn is contained inside the higher phase defined by -er. This means that once the compound is constructed, the non-head will not be able to be extracted simply because it is part of a chunk of structure which has abandoned the computational system and has gone to the interfaces to get an interpretation. Any constituent occupying the position of life or, for that matter, any constituent below the head responsible for categorizing the compound would be equally unextractable because of the same reasons. On these grounds, the reason that explains that the interrogative what cannot be extracted from inside the compound in (9b) is the same that explains that an interrogative cannot be extracted from a relative or conditional subordinate clause: they are constituents embedded in a phase, and they are transferred out of the computational system before movement can take place.

Grammatically Relevant Definition of Word and why it belongs to Syntax 109

3.3. Against treating words like phases Despite what I have shown in the previous paragraphs, the proposal that words are phases runs into theoretical difficulties and, moreover, fails to account for the data. The notion of phase is understood in the Minimalist program as a way to minimize the information load that the computational system has to cope with. Intuitively, the idea is very simple: if there is a chunk of structure which has satisfied all its formal requisites, syntax can remove it from the space where the derivation is performed because that chunk does not need anything else. “Removing it” implies transferring it to further levels, where semantics and phonology are performed. It is crucial that the chunk of structure which has all its formal properties satisfied is transferred as soon as possible. Otherwise, the whole intuition that underlies the proposal of phases would be contradicted, as there will be at least one step in which an operation is performed on a structure which contains an unnecessary work load. As a consequence, every phase is defined by a specific head which is assumed to have all the formal characteristics required to satisfy the formal conditions of the chunk of structure embedded under it. Consider, for example, the head little v, assumed in standard minimalism to be a head that defines a phase (24). (24) [vP [DP1] v0 [VP V0 [DP2]]]

Little v is assumed to carry a set of formal features which, by a form of agreement, can check the case needed by DP2 in (24). This corresponds to the direct object of a transitive verb. Before little v is introduced in the syntactic tree, DP2 has, still, a formal property in need of satisfaction, its case. As soon as little v is introduced in the tree, it automatically uses its formal features to satisfy the case of the direct object and the structure is defined as a phase. But which chunk of structure is transferred to the interfaces, thus making its constituents unavailable for movement? It is standardly accepted that the chunk is the complement of v0, but not v0 itself or its specifier. In other words, the part that abandons syntax is VP, not vP. How do we know this? Because if vP was transferred, the DP1 contained in vP would also be eliminated from the syntax. DP1 corresponds to the agent of a transitive verb, and by assumption, it must check its case for the same reasons that DP2 had to check it. However, it cannot check it with v0 because this head has already checked the case of DP2. In other words, DP1 will have to wait for further steps of the syntactic derivation in order to get its formal properties licensed, and because of that, it follows that it cannot be part of the chunk of structure transferred to

110

Chapter Three

the interfaces. Of course, if DP1, the specifier of vP, cannot be transferred, it also follows that v0, the head of vP, cannot be transferred either, because it is not possible to transfer a head and leave the phrase that it heads behind. As shown, the standard minimalist assumptions necessarily lead us to the conclusion that the complement of a phase, but not its head or its specifier, is transferred. At this point, it is necessary to question whether Arad’s system is using exactly this technical notion of phase. The answer seems to be negative. Arad’s claim that the set formed by the root and its categorizer is a phase is technically incompatible with the way in which she assigns the semantic interpretation to the root, if by phase the author means what is standardly assumed to be a syntactic phase. Why? Given the structure in (25), the complement of the head that defines the phase is expected to be transferred to semantics where it gets an interpretation. That complement is the root. (25) [v [¥HAMMER]]

However, Arad (2003: 747) states very clearly that roots are assigned an interpretation in the environment of the first category-assigning head. This is crucial to explain the difference in meaning between a hammer and to hammer. The root does not have a stable semantics; that is, the root does not have any semantics of its own (see also Acquaviva 2009). It is rather the structures in (26a) and in (26b) that get a semantics assigned. (26) a. [n [¥HAMMER]] = ‘tool consisting of a solid head set crosswise on a handle’ b. [v [¥HAMMER]] = ‘to beat something using a blunt instrument’

However, the proposal that (25) constitutes a phase entails that the constituent transferred to the semantics will only be the root. This is difficult to reconcile with the fact that the semantics does not assign any interpretation directly to the root, but rather to the complex structures in (26). If only the root were transferred, the semantics would not be able to know if the root appears in a verbal or a nominal context, as a phasedefining head such as little v is not transferred as part of the phase. This, of course, does not mean that Arad’s analysis is not a possible explanation of the empirical contrasts that she discusses, but rather that the domain that she is describing does not behave like a standard syntactic phase such as the one discussed before. From this perspective, it seems that, even if Arad’s proposal were right and it could be extended to explain the impossibility of extracting constituents from inside words, this explanation would not qualify as supporting evidence for the claim that Lexical Integrity effects can be

Grammatically Relevant Definition of Word and why it belongs to Syntax 111

explained by the syntactic notion of phase. This is due, quite simply, to the fact that the phase used in the analysis does not display the expected behavior for a syntactic phase. Assuming that her analysis is right, it provides evidence for a different kind of domain. Moreover, being in a different phase does not block pronominal coreference in general as (27) shows; thus, even if there was a technical way of redefining phases to allow the timing just described—which there is not doubt it could be possible—the problem would still remain that a phase behaves differently from a word in allowing pronominal coreference (compare (27) with (28)). (27) If Maryi comes, John will marry heri (28) *John went to the cari washer so that iti is shinny

3.4. Interim conclusion: There are “words” Let us take stock before continuing. We have seen that there is a grammatically relevant way of defining something as a word based on its syntactic behavior or rather, absence of behavior with respect to some typical syntactic operations. We have seen that some apparent counterexamples are not real counterexamples, so this way of defining a word is reasonably robust and testable. Furthermore, it has been argued that one attempt to interpret this kind of structure called “word” as equal to the structure that we call “phrase” is not correct. This is so for two reasons. One is that the assumptions that one has to make about the timing of transference of the first structure are not identical to those necessary for the second structure. Another is that the way in which both structures interact with respect to pronominal coreference is different. Therefore, there seem to be reasonable grounds to claim that there is a classification which includes two objects: words and phrases.

4. “Words” are defined by deactivation of features: Spanish as a test language In this section I will show that, despite the real empirical differences between words and phrases and the impossibility of using standard phases to explain these differences, the grammatically relevant notion of word can be defined by syntactic principles and, thus, does not support the division between morphology and syntax. I will argue that what behaves like a “word” with respect to extractions, coreference and ellipsis is actually a syntactic constituent containing only one set of active phi features. The

112

Chapter Three

internal members of this structure either lack active features or have deactivated them as a result of merging to compose the structure. What we call a word is a syntactic constituent whose members do not have any remaining active features; what we call a phrase is a syntactic constituent whose members have active features. To give initial support to this idea, let us start with some surprising empirical contrasts. One of the phenomena that were invoked to motivate a distinction between phrases and words–and was shown later to be empirically wrong–is whether one of the constituents inside the word can be expanded into a bigger structure. The data show that this is possible, but not with any kind of phrase. Consider the following contrasts: (29) a. un pinta [uñas de plástico] a paint- [nails of plastic] ‘something to paint plastic nails’ b. *un pinta [uñas postizas] a paint [nails fake] ‘something to pain fake nails’ (30) a. un abre [botellas de porcelana] an open [bottles of porcelaine] ‘something to open porcelaine bottles’ b. *un abre [botellas delicadas] an open [bottles delicate] ‘something to open delicate bottles’ (31) a. un salva [manteles de tela] a protect [tablecloths of fabric] ‘something to protect fabric tablecloths’ b. *un salva [manteles caseros] a protect tablecloths home-made ‘something to protect home-made tablecloths’

The previous examples involve the so-called V-N compounds in Spanish. Other structures provide similar contrasts like the so-called N-N compounds. (32) a. amarillo [tirando a limón] yellow [going to lemon] ‘lemonish yellow’ b. *amarillo [limón pequeño] yellow [lemon small] ‘small lemon yellow’

Grammatically Relevant Definition of Word and why it belongs to Syntax 113 (33) a. perro [policía de aduanas] dog [police of customs] ‘customs police dog’ b. *perro [policía militar] dog police military ‘military police dog’ (34) a. hombre [anuncio de sex-shop] man [advertisement of sex-shop] ‘sex-shop ad-man’ b. *hombre [anuncio ilegal] man [advertisement illegal] ‘illegal ad-man’

In much the same fashion as V-N compounds, N-N compounds also behave like anaphoric islands as shown in example (35). (35) *Mi jersey es amarillo [limón]i, pero no creas que me gusta proi. My sweater is yellow lemon, but not think.2sg that me.dat likes pro. ‘My sweater is lemoni yellow, but don’t think that I like iti’

It should be noted that a necessary condition to allow for modification is that the compound has a compositional meaning. This is not surprising when one considers that the expansion of the internal noun into a bigger phrase requires that the noun is interpreted in the compound in a compositional way. This point can be illustrated with a compound like (36), where the whole does not have a compositional meaning and behaves like an idiom. Any modification of the non-head noun is impossible because that modification will give us a specific kind of frog and this concept is not part of the denotation of the compound. Thus, in these cases modification is not possible because it destroys the idiomatic meaning of the compound, but there is no reason to think that it is blocked by structural reasons. (36) hombre rana man frog ‘frogman’

The fact that a compound has a non-compositional meaning does not mean that the compound has to be treated as a non-syntactic object; it only means that it has to be stored in some list, but complex syntactic structures can have the same meaning properties when they are idiomatic. What hombre rana needs to be idiomatic is the same as a syntactic structure like the one in (37) needs to be idiomatic, and certainly the presence of

114

Chapter Three

pronouns and functional elements in this second structure prevents anyone from saying that it is a word in lexicalist terms. (37) dár-se-las de give-SE-them of ‘to pretend to be’

It is not possible to provide examples with internal modification in the case of N-A compounds, the third productive class of compounds in Spanish involving nouns, but this is due to independent reasons. More specifically (cf. Fábregas 2004), these compounds have a prosodic requisite that imposes that the first member of the compound, the noun, has to be bisyllabic, and this automatically precludes the noun from being modified by any kind of prepositional modifier. (38) shows that when the noun forms more than two syllables with the linking element (LE), it has to be truncated. (38) a. *naric-i-largo nose-LE-long ‘long-nosed’ b. nar-i-largo nose-LE-long ‘long-nosed’

The examples discussed here raise two immediate conclusions. The first one is that, despite the fact that the internal members of the compound cannot establish coreferential relations with pronouns, the data do not support the proposal that considers that these compounds are morphological objects. The reason is that they contain whole phrases as part of their internal structure. Two solutions have been proposed from a lexicalist perspective to this problem and in order to be able to claim that the compounds are still morphological objects. The first one is to argue that what looks like an internal phrase inside the compound is a lexical object because it has to be stored in the lexicon. It is true that this object is not lexical in the sense that it disallows functional material or in the sense that it must have a demotivated meaning. But even if these factors could be argued to be involved in any word internal modification, the point is that being listed in the lexicon does not automatically imply being a morphological object (DiSciullo and Williams 1987, Williams 2007). An object listed in the lexicon, which DiSciullo and Williams call “listeme”, is simply any kind of meaning-form pair that cannot be derived through the application of compositional and regular rules and needs to be learnt

Grammatically Relevant Definition of Word and why it belongs to Syntax 115

by speakers by heart. The lexicon contains some objects that might be classified as morphological (like the morpheme -s marking the plural), but it will also contain phonological objects (like the intonation associated to interrogatives) and, certainly, also syntactic objects, like the idiom presented in (37). Even if sequences like policía de aduanas ‘customs police’ need to be listed in the lexicon, this does not prove that they are syntactic. The second solution from a lexicalist perspective to solve this problem has been to weaken the sequential relationship between morphology and syntax and propose that, instead, both components act in parallel allowing objects built in the syntax to be taken as input for morphological objects (Ackema and Neeleman 2004). Notice, however, that this complicates the design of grammar from a theoretical perspective, and perhaps more crucially, that in the absence of independent grammatical evidence that the domain defined by a word is not analyzable in syntactic terms, this proposal needs to make that design still more complicated on the basis of a distinction that might not be motivated. The conclusion, if we look at the data without too many theoretical preconceptions, is that with respect to the possibility of having expansions of internal constituents, there do not seem to be any real differences between phrases and compounds, and therefore that both objects seem to be quite similar. It should be emphasized at this stage that both Construction Grammar and Constructionism would be able to explain this: there is no real structural difference between words and phrases. Lexicalism, on the other hand, has trouble explaining this pattern of data and, as shown above, needs to give up the projectionist idea that morphology precedes syntax and, therefore, conditions it. Let us now discuss the second conclusion that can be derived from the contrasts presented before. There is nothing wrong in having a modifier of a noun internal to a compound provided that this modifier is not an adjective. When the modifier is a prepositional phrase, the result is grammatical. The question is what the difference between adjectives and prepositional phrases in Spanish actually is. The immediate answer is that adjectives need to agree in number and gender with the constituent they modify, but there is no direct evidence that the prepositional phrase needs to have any kind of agreement in this language.

Chapter Three

116

4.1. Freezing structure through deactivation My proposal, which gets initial support from the contrasts just presented, is that what we call “words” are structures whose internal constituents do not have any active features left once the head is projected. The theoretical background for this proposal is the so-called Activity Condition (Chomsky 2001). The idea is that a syntactic constituent can only take part in syntactic operations if it still has some property to license syntactically. When a constituent has licensed all its syntactic properties, economy requirements dictate that it will not participate in any further syntactic operation, because it does not have any more properties to satisfy. If properties are codified in the form of features, what this proposal claims is that once an item does not contain any more features to check, it becomes deactivated and will be invisible for further syntactic operations. Imagine that there are two constituents, a and b, which merge together. These items are characterized by a number of features; some are interpretable ([iF] or just [F]), some are not ([uF]). If merge is a syntactic operation and having uninterpretable features is a precondition for participating in syntactic operations, it is necessary to assume that all items that take part in a merge operation have at least one uninterpretable feature checked by merge itself. Additionally, the items combined in the tree might have other uninterpretable features. Following Chomsky’s (2004) suggestion, I assume that the projecting head is the one that selects the other element. If selection is basically a problem solved at the semantic interface (Borer 2003), this does not need to be codified in terms of features: the interface will interpret the configuration in (39) accordingly, interpreting whatever element is in the complement of the other as the selected one. (39)

a [uF]

a

b [uF]

Differences arise when the other uninterpretable features that each constituent has inside the word are considered. Assuming for the sake of explicitness that, in the relevant cases, the projecting head will have at least another uninterpretable feature so that it remains active for at least another merge operation, there are two options:

Grammatically Relevant Definition of Word and why it belongs to Syntax 117

(i) After merge, b does not have any further uninterpretable features. In this situation, merge with a has deactivated b and by the time a projects, b is completely deactivated (40) After merge, b still has uninterpretable features. In this situation, merge with a has not deactivated b and by the time a projects, b is still syntactically active (41)

(ii)

(40)

a [uF] [uF]

a

(41)

b [uF]

a [uF] [uF]

a

b [uF] [uF]

I contend that the constituent in (40) is what we call a “word”; the constituent in (41) is what we call a “phrase”; there are no real differences between them, except for the fact that both constituents are active after merge in the second case, but only one remains active in the first case. From the above it follows that the internal constituents in a grammatical word will not be able to take part in further syntactic operations: as they have been deactivated, further steps of the syntactic derivation will ignore them. This explains that in (40), b will not be able to be displaced, to be coreferential with a pronoun, and to take part in ellipsis operations. To all effects, b has become frozen, will have to be interpreted as part of a, and the syntax has done with it all that it could be done. As b will not take part in more syntactic operations, it will be interpreted as part of a. It will only be able to take part in operations to the extent that it is part of a syntactic constituent headed by a, and therefore a and b will be, to all effects, interpreted as a single unit by further syntactic operations. This explains several empirical facts. First of all, the whole constituent a will be able to displace, and it will carry with it b, because b’s position inside the constituent headed by a has been defined irreversibly. Thus, the “word” will be able to move as a whole, but a or b will not be able to move without the other element. (42) Lavaplatosi, Juan sí compra ti wash-dishes, Juan yes buys ‘Dish washers, John does buy’

118

Chapter Three

Secondly, ellipsis will be possible provided that the target is the whole constituent headed by a, but it will never be possible if b is, in the absence of a, the attempted target. Consequently, ellipsis can be controlled by a whole word, but not by a part of it. (42) El [lavaplatos]i de Juan y el øi de María the wash-dishes of Juan and the ø of Mary ‘John’s dish-washer and Mary’s’

Thirdly, pronominal coreference will be possible provided that the antecedent is the whole structure headed by a, but b, having been frozen as an inactive part of that constituent, will not be able to be the antecedent. Consequently, the whole word can be the antecedent of a pronoun, but b in isolation will not. (43) Me he comprado un [lavaplatos]i, pero no me gusta proi me.dat have.1SG bought a wash-dishes, but not me likes pro ‘I bought a dish washer, but I don’t like it’

In contrast, in (41), b remains active after the constituent has been formed. Naturally, all uninterpretable features will be eventually erased, but the point is that this constituent will remain active after the constituent headed by a has been defined. This explains that a and b will not be frozen together and will be interpreted as two separate constituents for further steps of the derivation. Therefore, b in this case will be able to (a) move without a, (b) control ellipsis without a and (c) be the antecedent of a pronoun without a. Consider from this perspective how the other contrasts can be explained. It has been shown that an adjective cannot modify a noun inside what we call a grammatical word, but a prepositional phrase can. The reason is that the structure will have in that case conflicting aspirations. The adjective will require agreement, a syntactic operation that consequently requires the noun it modifies to be syntactically active, but in order to build a grammatical “word” the same noun has to become syntactically deactivated. Let us show this step by step. For the sake of explicitness, following Cinque (2010) it is assumed that the adjective is introduced as a specifier of a functional category on the noun’s functional hierarchy. This gives as a result the configuration in (44).

Grammatically Relevant Definition of Word and why it belongs to Syntax 119 (44)

FP

AP

F F

NP

The adjective requires number and gender agreement. This implies that the noun in this configuration has to be introduced with functional projections that give both a number and a gender value. However, these projections are not inside the NP. Let us dwell on why this is so. First of all, despite the fact that the noun contains a desinence, there is evidence that the desinence that comes with the noun does not carry gender (Harris 1991), as there is no direct correspondence between that marker and the gender value (45). (45) a. la man-o izquierd-a the.FEM hand left-FEM ‘the left hand’ b. el perr-o negr-o the.MASC dog black-MASC ‘the black dog’

The general mismatches between desinence and gender argue in favor of splitting what has been traditionally interpreted as “gender” into two projections, with agreement being controlled by a high functional head, belonging to the determiner domain, which contributes gender as a classification of nouns in order to restrict reference (Fábregas and Pérez 2010). Thus, gender agreement with the adjective requires the presence of a high GenderP. Consider now number. Even though nouns inside V-N compounds can appear in the plural (46a), there is evidence that this -s marker cannot be interpreted as number. This evidence comes from the behavior of mass nouns inside V-N compounds, which can appear in the plural form without the noun losing its mass noun nature, or it getting a taxonomic reading (46b).

120

Chapter Three

(46) a. limpia-[ventana-s] wash-window-s ‘window cleaner’ b. para-[brisa-s] stop-wind-s ‘wind shield’

This suggests, again, that the alleged plural marker -s is not a number projection, but a much lower constituent which does not provide the semantic interpretation associated to plurals (see also Borer 2005 for the case of English). A similar reasoning to the one I made for the case of gender and desinences is possible here. The projection responsible for number agreement is not the one that introduces the -s. Consequently, for an adjective to agree in gender and number with an NP, it is necessary that the noun should be projected up to a high functional level, including at least number projections and a gender projection associated to gender. This gives us, minimally, the structure in (47), which must be present for an adjective to agree. (47) [GenderP [NumberP [FP [AP] F0 [NP]]]]

The problem is that with GenderP and NumberP we have already moved past the domain of noun phrases to the domain of determiner phrases. Determiner phrases, however, have other kinds of syntactic requisites and needs that make it necessarily the case that they remain syntactically active after they merge with a V; minimally, they need to receive case (Longobardi 1994). Thus, when the structure in (47) is embedded under the V to form what we call a V-N compound (48), it will still remain syntactically active, because the V itself will not check its case. (48) [VP V [GenderP…]]

And at this point, in (48), there is only one possibility left: to merge a verbal functional projection that checks case in the GenderP. For the sake of explicitness, assume that this projection is vP (Chomsky 1995). (49) [vP [VP V [GenderP…]]]

The result is that then the structure will not project as a grammatical word, because GenderP is still syntactically active after merge with V.

Grammatically Relevant Definition of Word and why it belongs to Syntax 121

Let us now turn the other option. Imagine that GenderP and NumberP are not introduced, so that the whole remains syntactically inactive after merge with V–not needing to check case. This implies the structure in (50); as verbal functional projections do not have to be introduced over (50), the whole could become a noun or any other category by embedding it under the relevant lexical head. (50) [VP V [FP [AP] F0 [NP]]]

The problem is that in this case AP connects with a syntactically inactive NP and does not find the projections that can check its agreement in gender and number; thus the structure will be ungrammatical, as the adjective will not license its syntactic properties. So in the presence of an adjective, either the whole is interpreted as a “phrase” or the derivation will be ungrammatical because it fails to agree. Consider now what happens with prepositional modifiers. In Spanish, a language where the morphophonology is not shy to reflect agreement operations between nominal constituents and different categories, prepositions systematically lack morphemes to mark agreement with the noun they modify. Lacking evidence to the contrary, it is safe to assume that a preposition–unlike an adjective—does not require agreement with the noun it modifies. Consequently, the presence of a PP modifier will not require expansion of the nominal constituent into a higher order functional phrase containing number and gender. Thus, the structure in (51) can combine directly with V (52) and still, it could become syntactically deactivated after this merge operation, because the preposition does not need any kind of agreement with the head noun. (51) [NP N0 [PP]] (52) [V [NP N0 [PP]]]

Consequently, the contrasts introduced at the beginning of this section are explained as follows. In Spanish, where there is no evidence that prepositions agree with the modified noun, but there is evidence that adjectives need to agree, a prepositional phrase can modify a noun and the whole can be part of a “word”, but an adjectival phrase cannot. I argue that this is due to the fact that adjectival agreement requires an active constituent, but in order to form a “word”, the constituent needs to be syntactically inactive.

122

Chapter Three

4.2. Further evidence In this section I will quickly review some additional pieces of evidence to further support my proposal. 4.2.1. Making a “word” by removing agreement First of all, consider compounds that combine two adjectives (A-A compounds). In these compounds it has been noted that only one of them can contain agreement; the other one is frozen. (53) chicas [[sord-o] mud-as] girls deaf mute-FEM.PL. ‘deaf-mute girls’

This is explained in this proposal. To form a “word”, it is not possible for both adjectives to be syntactically active after they merge with each other. The need to agree with the extended projection of an NP keeps the adjective active, as shown in the last section. Thus, agreement in both adjectives means that they cannot form a word. If only one adjective carries agreement, the other can be deactivated syntactically after merge, giving a whole which acts as it is expected of a grammatical “word”. 4.2.2. Making a “word” by removing functional markers Secondly, it has been noticed that, inside real compounds, functional words like prepositions can be absent even if the semantic relation between the two elements of the compound requires those functional words when projected as a phrase. Indeed, losing functional markers is one of the criteria to establish that a structure moves from an idiomatic phrase to a real compound, which in my terms translates into moving from a phrase to a “word” (Val Álvaro 1999). Consider the pair in (54). (54) a. tela-araña web-spider ‘spider’s web’ b. tela de araña web of spider ‘spider’s web’

The absence of the preposition is associated to a more wordy status. Consider how this is explained in my proposal: prepositions are syntactic

Grammatically Relevant Definition of Word and why it belongs to Syntax 123

heads that introduce relations between constituents (Hale and Keyser 2002), but at the same time they can check the case requirements of the noun they take as a complement. This is the case if the noun is syntactically projected as a determiner phrase and thus can remain syntactically active. However, in order to form a compound, the noun cannot be projected as a whole determiner phrase, because it needs to become syntactically inactive after merging with the other constituent. Therefore, it does not require case to be checked; consequently, the presence of the preposition is not necessary. 4.2.3. Not all morphological objects can be placed inside words More generally speaking, my proposal gives an explanation to the well-known fact that agreement markers and functional items are generally absent from inside words. To repeat much quoted data, a deverbal derived noun does not allow the verb to contain agreement markers. Example (55) is an illustration of the generalization that the item which is not the head of a whole “word” cannot carry real agreement markers (cf. cases discussed in Bauer 1978, which can be argued to be exponents of categories sometimes used in agreement processes, but not triggering or reflecting agreement in themselves). (55) a. vende-dor sell-er b. *vende-n-dor sell-3pl-er

The explanation is the following: if the non-head of a “word” contains agreement, it will be syntactically active after merging with the head of the whole constituent. Therefore, it will not form a grammatical “word”, but either it carries agreement or it is syntactically inactive. This explains the typology of units that can participate in word formation processes. Words can be the combination of roots with other heads as (56). Roots are the maximally inactive elements in the grammar; they do not carry a grammatical category, do not have meaning by themselves and do not have selectional restrictions (Acquaviva 2008). Thus, they have satisfied all their own requirements once they are merged with a functional category. (56) cas-a house-desinence

124

Chapter Three

Words can also contain stems, provided that the marker for the stem does not carry agreement. The stem is the set formed by a root and (at least) a functional head which assigns a grammatical category to it, such as cant-a ‘sing-Th(eme) V(owel)’ in (57). Defined as a lexical category, it does not require agreement. (57) cant-a-nte sing-ThV-er ‘singer’

Finally, words can contain whole phrases provided that these whole phrases are “frozen”; which means that all their formal and semantic requisites need to be satisfied inside the phrase and cannot need to be checked by an external constituent. Thus, the verb in (58) does not show agreement, even in a language like Spanish, with an external element. Once the phrase is used as part of a word, the verb remains frozen in whatever form we used, and it cannot agree with a subject, as seen in (59b), from (59a), only marginally acceptable in Spanish but still grammatical. (58) a [slept-all-day] look (59) a. un pinta ‘he dormido todo el día’ a look have slept all the day b. Tú tienes una pinta ‘{he / *has} dormido todo el día’ You have a look ‘have / have.2SG slept all day’

4.2.4. “Words” cannot be recursive Ultimately, this explains a final property: words cannot be recursive in the way in which phrases are. A phrase of the XP type can contain a phrase of the XP type inside (60), but a word cannot contain a word inside it, if by “word” in opposition to root and stem, we mean a unit which reflects not only its grammatical category but also markers that trigger or undergo agreement. (60) the brother [of the sister [of the girlfriend [of my aunt]]]

By contrast, a word cannot contain another word (meaning a stem plus inflection) inside it: (61) *chicas sord-as-mud-as girls deaf-FEM.PL.-mute-FEM.PL.

Grammatically Relevant Definition of Word and why it belongs to Syntax 125

The limit of a “word” will be the presence of agreement markers or functional elements requiring licensing by higher projections. This explains why in a language like Spanish, where agreement is generally marked by designated morphophonological markers, the rule of thumb is that each word is characterized by agreeing or being able to trigger agreement with another unit.

4.3. Summary To summarize, my proposal is the following: (i) (ii)

The difference between phrases and words is empirically real but does not follow from a distinction between the way in which both structures are built. The difference between a phrase and a word is that in the former, but not in the latter, the non-head still contains active non-interpretable features after merging with the head. Thus, there is no real distinction between these two elements in the grammar.

This proposal explains that: (i) (ii)

No elements containing agreement can be the non-head inside words. Therefore, in a language like Spanish, words can have prepositional modifiers but not adjectival ones. (iii) Therefore, in general, words cannot have internal markers of grammatical function or agreement. (iv) therefore, words can only be built by combining roots, stems or frozen phrases with other units.

5. A short note on languages without exponents of agreement Agreement has played a relevant role in this proposal. Despite the fact that agreement does not define a word, but rather absence of agreeing features (and other syntactic properties that need to be checked after first merge) in the non-head of the word, the question remains of whether languages where agreement is not instantiated in morphophonological segments are also amenable to this analysis. The question is whether languages that have been argued not to have agreement behave in a similar way with respect to the phenomena discussed here. Consider the following sentences in Japanese. The context is set so that the whole sentence is focused, to prevent the pronoun from referring to a previously uttered sequence.

126

Chapter Three

(62) (John enters the room and says: ‘You know what?...’) Taro-ga ha-burashi-o kae-te, {ha/*sore(-ra)}-ga Taro-NOM tooth-brush-ACC change-PAST tooth/that(-PL)-NOM boroboroni nat-ta crumbling become-PAST ‘Taro changed toothbrush, and they (=the teeth) crumbled’ (63) (John enters the room and says: ‘You know what?...’) Taro-ga ki-koku shi-ta toki, {kuni/*sore}-wa Taro-NOM return-country do-PAST when, country/it-TOP kanari kawatte shimatte i-ta quite change have be-past ‘When Taro returned to his country, it (=the country) had changed a lot’ (64) (John enters the room and says: ‘You know what?...’) Totsuzen ie-kara shuk-ka shi-ta suddenly house-from come.out-fire do-PAST ga {hi/*sore}-wa ni-jikan-go-ni tome-rare-ta but fire/it-top 2-hour-after-at stop-pass-past ‘Suddenly fire came out of a house, but it (=the fire) was put off 2 hours later’

In all these sentences we have a compound involving a noun: haburashi ‘tooth-brush’ (62), ki-koku ‘return-country’ (63) and shuk-ka ‘come.out-fire’. In all these cases, despite the absence of markers for agreement in this language, the noun inside the compound cannot be the antecedent of the pronoun sore. Compare these examples with (65). Here, instead of the compound kikoku, the non-“word” equivalent is used. In this case, the pronoun can corefer: (65) (John enters the room and says: ‘You know what?...’) Taro-ga kuni-ni kaet-ta toki sore-wa Taro-NOM country-to return-PAST when it-TOP kanari kawatte shimatte i-ta quite change have be-PAST ‘When Taro returned to his country, it (=the country) had changed a lot’

What changed in this case? The noun is now introduced by a postpositional marker ni ‘to’, and the verb now carries the inflection and is not combined with a verb that carries the inflection, as in (63). What we see is a pattern very similar to the one we observed in Spanish, a language where agreement is present as a syntactic process.

Grammatically Relevant Definition of Word and why it belongs to Syntax 127

This is also expected in my proposal. On the assumption that all languages have syntax, all languages must have features that make units syntactically active at least for some part of the derivation. Therefore, we also expect to see cases in which the activity of the non-head inside a syntactic constituent remains after merging with the head and other cases where the non-head becomes inactive after the first merge. Agreement might not be active in some languages, but other syntactic properties–like case checking, combination with tense, etc.—will be active and motivate the distinction. In conclusion, this proposal allows us to say that all languages have structures that can be called “words”, even if they lack agreement.

6. Final conclusions In this chapter I struggled with the definition of word. I argued that a definition of word based on its semantic, phonological or morphological properties does not give us a concept that can be used to identify units with special properties (beyond those that, in a circular fashion, we used to define the concept itself). However, I shown that an exploration of a word’s syntactic properties shows that the Lexical Integrity Hypothesis was right to focus attention on those structures that are opaque for processes such as movement, coreference or ellipsis in order to define structures as words, where “word” means “a structure whose internal structure lacks syntactic activity”. The special properties of a word as opposed to a phrase are never positive: they all involve not being able to take part in a process where (at least some) phrases can take part. The second step was to show that this syntactic definition of word does not provide us with any argument in favor of a two-engine model of language. This is so basically because the distinction between the two kinds of objects can be explained with one single engine. In other words, by proposing that, in what we call a word, the first merge between the head and the non-head deactivates the non-head. Empirical differences stem from here: being deactivated, the non-head freezes in that position and is unambiguously interpreted as belonging to the constituent headed by the other element. Consequently, the non-head will not be able to take part in syntactic processes unless these processes are determined by the head of the “word”, in which case, by transitivity, they will also involve the non-head. Let us finally address the consequences that can be derived from the ongoing. If my suggestion survives further inquiry and proves to be solid, it might mean that morphology, as a generative level of grammar, does not

128

Chapter Three

have any reality, but could still be a convenient way to refer to the analysis of syntactic structures containing elements that have been deactivated as early as possible in the syntactic derivation. In that case, this study would mean another step towards ascribing morphology entirely to syntax. However, this would not be (yet) everything needed in order to reduce the two traditional components to one: we would still have to address the question of apparently purely morphological markers (like desinences and theme vowels) and find a place for them in a system where the lexicon is just a stored list of units and the only way of combining them is syntax. Would they correspond to some syntactic function or would they have to be stored as parts of a lexical item with some mechanism that explains their distribution without reference to morphological properties? This last issue will have to be left for further research.

References Ackema, P. and A. Neeleman 2004, Beyond Morphology: Interface Conditions on Word Formation, Oxford: Oxford University Press Acquaviva, P. 2009, “Roots and Lexicality in Distributed Morphology”, A. Galani, D. Redinger and N. Yeo (eds.) York-Essex Morphology Meeting 2, York: University of York, 1-21 Anderson, S. A. 1992, Amorphous Morphology, Cambridge: Cambridge University Press Arad, M. 2003, “Locality constraints on the interpretation of roots: the case of Hebrew denominal verbs”, Natural Language and Linguistic Theory 21 (4), 737-778 Aronoff, M. 1976, Word Formation in Generative Grammar, Cambridge (Mass.): MIT Press Baker, M. 1988, Incorporation. A Theory of Grammatical Function Changing, Chicago: University of Chicago Press Bauer, L. 1978, The Grammar of Nominal Compounding, with Special Reference to Danish, English and French, Odense: Odense University Press Borer, H. 2003, “Exo-skeletal vs. endo-skeletal explanations: Syntactic projections and the lexicon”, J. Moore and M. Polinsky (eds.) The nature of explanation in linguistic theory, Chicago, CSLI, 31-67 —. 2005, Structuring Sense Vol. 1: In Name Only, Oxford: Oxford University Press Chomsky, N. 1957, Syntactic Structures, The Hague, Mouton & Co —. 1965, Aspects of the Theory of Syntax, Cambridge (Mass.): MIT Press

Grammatically Relevant Definition of Word and why it belongs to Syntax 129

—. 1970, “Remarks on nominalization”, R. A. Jacobs and P. S. Rosenbaum (eds.) Readings in English Transformational Grammar, Waltham: Ginn and Co., 232-286 —. 2001, “Derivation by Phase”, M. Kenstowicz (ed.) Ken Hale: A Life in Language, Cambridge (Mass.): MIT Press, 1-52 —. 2004, “Beyond Explanatory Adequacy”, A. Belletti (ed.) Structures and Beyond. The Cartography of Syntactic Structures, vol. 3, Oxford: Oxford University Press, 104-131 Cinque, G. 2010, The Syntax of Adjectives: A Comparative Study, Cambridge (Mass.): MIT Press DiSciullo, A.-M. and E. Williams 1987, On the Definition of Word, Cambridge (Mass.): MIT Press Embick, D. and R. Noyer 2001, “Movement Operations after Syntax”, Linguistic Inquiry 32 (4), 555-598 Fábregas, A. 2004, “Prosodic Constraints and the Difference between Root and Word Compounding”, Lingue e Linguaggio 2, 303-339 —. 2005, The Definition of the Grammatical Category in a Syntactically Oriented Morphology, PhD diss., Madrid: Instituto Universitario Ortega y Gasset / Universidad Autónoma de Madrid (Available at http://ling.auf.net/lingbuzz) Fábregas, A. and I. Pérez 2010, “Gender agreement on adverbs in Spanish”, Journal of Portuguese Linguistics 7, 25-47 Fodor, J. 1968, Psychological Explanation: An Introduction to the Philosophy of Psychology, New York: Random House Goldberg, A 1995, Constructions. A Construction Grammar Approach to Argument Structure, Chicago: University of Chicago Press Hale, K. and S. J. Keyser 2002, Prolegomenon to a Theory of Argument Structure, Cambridge (Mass.): MIT Press Halle, M. 1973, “Prolegomena to a theory of word formation”, Linguistic Inquiry 4 (1), 3-16 Halle, M. and A. Marantz 1993, “Distributed Morphology and the Pieces of Inflection”, K. Hale and S. J. Keyser (eds.) The View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger, Cambridge (Mass.): MIT Press, 111-176 Harley, H. 2009, “Compounding in Distributed Morphology”, R. Lieber and P. Stekauer (eds.) The Oxford Handbook of Compounding, Oxford: Oxford University Press, 122-149 Kiparsky, P. 1982, “Lexical morphology and phonology”, Linguistics in the Morning Calm. Selected Papers from SICOL-1981, Seoul, Hanshin Publishing Co., 3-91

130

Chapter Three

—. 1997, “Remarks on Denominal Verbs”, A. Alsina, J. Bresnan and P. Sells (eds.) Complex Predicates, Stanford (CA): CSLI Publications, 473-499 Lakoff, G. 1962, Deep Surface Grammar, Indiana: Indiana University Linguistics Club Lapointe, S. 1978, A Theory of Grammatical Agreement, PhD diss., Amherst: University of Massachusetts Lees, R. 1960, The Grammar of English Nominalizations, The Hague: Mouton de Gruyter Lieber, R. 1980, On the Organization of the Lexicon, PhD diss., University of New Hampshire. —. 1992, Deconstructing Morphology, Chicago: University of Chicago Press Lieber, R. and S. Scalise 2006, “The Lexical Integrity Hypothesis in a New Theoretical Universe”, Lingua e Linguaggio 5 (1), 7-32 Longobardi, G. 1994, “Reference to kinds and proper names: a Theory of N movement in Syntax and Logical Form”, Linguistic Inquiry 25 (3), 609-665 Marantz, A. 2001, Words, ms., MIT May, R. 1985, Logical Form: its Structure and Derivation, Cambridge (Mass.): MIT Press Montermini, F. 2006, “A New Look on Word-Internal Anaphora on the Basis of Italian Data”, Lingua e linguaggio 5, 127-148 Postal, P. 1969, “Anaphoric islands”, R. Binnick et al. (eds.) Papers from the 5th Regional meeting of the Chicago Linguistic Society, Chicago: CLS, 209-239 Scalise, S. 1984, Generative Morphology, Dordrecht: Foris Selkirk, E. 1982, The Syntax of Words, Cambridge (MA): MIT Press Starke, M. 2009, “Nanosyntax: A short primer to a new approach to language”, P. Svenonius, G. Ramchand, M. Starke and K. T. Taraldsen (eds.) Nordlyd 36, 1-6 Val Álvaro, J. F. 1999, “La composición”, I. Bosque and V. Demonte (dirs.) Gramática Descriptiva de la Lengua Española, Madrid: Espasa, 4757-4843 Varela, S. 1990, Fundamentos de Morfología, Madrid: Crítica Williams, E. 2007, “Dumping lexicalism”, G. Ramchand and C. Reiss (eds.) The Oxford Handbook of Linguistic Interfaces, Oxford: OUP, 353-382

CHAPTER FOUR FROM AGGLUTINATION TO POLYSYNTHESIS: TOWARDS A BIOLOGICAL CHARACTERIZATION OF THE SPOKEN WORD JUAN-CARLOS MORENO-CABRERA

1. Introduction In this chapter I focus on the spoken word from the perspective of speech and its impact on the morphological typology of languages. In the first section I argue that classical morphological typology is almost exclusively based on the written word and, therefore, is of little use for establishing a word typology relevant to spoken languages. In fact, it can be argued that some of the word types envisaged in classical morphological typology (notably, agglutination and polysynthesis) do not define language types but can, instead, be interpreted as manifestations of universal properties of human speech. In the second section it is argued that the spoken word can be conceived of as a linguistic realization of the chunks in the short-term memory (STM) tasks proposed by Miller (Miller 1956, Cowan 2000). In section 3, I discuss the proposal by Dahl (2004) concerning the linguistic definition of these speech chunks, “packets” in his own words. Whereas a recent proposal concerning the polysynthetic word put forward by De Reuse (2009) is presented in section 4. This author proposes a mechanism of productive non-inflectional concatenation (PNC) which not only is valid for all languages but, in my view, can be used to account for certain polysynthetic phenomena in speech, which, therefore, renders it is useless for a typological classification of languages. The final section shows that the quantitative limits of words as speech chunks coincide with the “magical number seven, plus or minus two” originally proposed by Miller (1956) and discussed in detail by Cowan (2000).

132

Chapter Four

2. Word types in classical morphological typology: written word vs. spoken word 2.1. The classical morphological classification of languages In his discussion on the possible genetic affiliation of all agglutinative languages, Tóth (2007: 1) presents the classical morphological classification of languages in the following terms: Morphological: analytic (eg. English) – isolating (eg. Chinese) – synthetic (eg.most Indo-European languages) – fusional (eg. Indo-European, Semitic) – agglutinative (eg. Uralic, Altaic) – polysynthetic (eg. Eskimo, Ainu) – oligosynthetic (eg. Nahuatl)

In this classification we find all the usual language types, including the popular isolating, agglutinative, fusional and polysynthetic types. In this chapter I will try to show that these types, except the fusional type, are of no use for classifying languages. The reason is that they are based on some misconceptions that result from the way in with languages are represented in the written medium.

2.2. The isolating language type myth The idea that languages such as Chinese and Vietnamese are morphologically isolating to a certain extent is very popular in linguistic literature and also elsewhere. In this section I will show that the concept isolating or monosyllabic type comes from a written-language oriented linguistic typology (Moreno-Cabrera 2008). In order to prove this, it is necessary to review some of the facts and fantasies about the Chinese writing system. Until the 19th century, Chinese writing has been regarded as ideographic. However, modern research in Chinese writing has demonstrated that this idea is wrong (DeFrancis 1984: 133-148, Unger 1990, 2003, Hannas 1997: 104-109). As Coulmas (2003: 57) states, “it is now widely recognized that the Chinese writing system is best described as a large syllabary with strong semantic elements that make up for the phonetic imprecision”. Accordingly, each Chinese character corresponds to a syllable; therefore, if we count the number of characters in a text we are able to know (except for a few cases) how many syllables are included in it. In addition, one of the norms of Chinese writing is that characters are to be written separately even if two or more characters correspond to one single word in speech; as

From Agglutination to Polysynthesis

133

a consequence, all characters in a text are separated by the same distance no matter whether they constitute different words or not. Moreover, wordderiving affixes are submitted to the same orthographic rules and are also separated from the root they are attached to by a blank space in spite of its word-forming function. Let me introduce here some examples of Chinese words. These examples illustrate the fact that in Chinese there are word-deriving suffixes as Packard (2000) has demonstrated beyond any reasonable doubt. One of these word-forming suffixes is the -yuán suffix meaning ‘person whose job is X’; let me introduce some examples in order to show this important point (Packard 2000: 72). (1) 哈 chuán ‘boat’ Ÿ 哈⛧ chuányuán ‘crew member’ ⷵ xué ‘study’ Ÿⷵ⛧ xuéyuán ‘student’ 㟨 jiao ‘teach’ Ÿ㟨⛧ jiaoyuán ‘teacher’ 㔴ヤ bàomù ‘announce’ Ÿ㔴ヤ⛧ bàomùyuán ‘announcer’ ≬十 bâoyù ‘child care’ Ÿ≬十⛧ bâoyùyuán ‘child-care worker’ 摖徼 câigòu ‘purchase’ Ÿ摖徼⛧ bâoyùyuán ‘purchasing agent’ ⒉兂 chunà ‘receive and pay out’ Ÿ⒉兂⛧ chunàyuán ‘cashier’ 扟┷ yùndòng ‘sport’ Ÿ扟┷⛧ yùndòngyuán ‘athlete’ 㖖⺋ zhídâo ‘advise’ Ÿ㖖⺋⛧ zhídâoyuán ‘adviser’ 䪣䴅 jánjiu ‘research’ Ÿ䪣䴅⛧ jánjiuyuán ‘research worker’1

If we examine the Chinese characters of the given words above, we will appreciate that all of them are separated from each other by blank spaces. For those with no knowledge of Chinese it will be impossible to determine how many words there are in the Chinese expressions given in (1); it will only be possible to count the number of written characters. For example, in chuányuán ‘crew member’ we see two characters and in zhídâoyuán ‘adviser’ there are three characters; but, as the pinyin roman transliteration shows, we have only one word and not two words (in the first example) or three (in the second example). Let us now consider the following Chinese sentence and its corresponding official pinyin transliteration (from Coulmas 2003: 59): 1

For typographical reasons I am using the circumflex accent for making the third tone in Mandarin Chinese.

134

Chapter Four

‘China underwent big changes during the past several years’ Fig. 4-1. Chinese sentence with official pinyin transliteration

The Roman script transliteration follows the basic rules of pinyin orthography. These rules determine which grammatical and lexical elements should be separated by blank spaces. Many of these criteria are similar to those of the Western orthographic tradition. We can see that zhǀngguó ‘China’ is analyzed as one word instead of two (zhǀng ‘Chinese’ and guó ‘country’), which is reasonable, since similar expressions such as Deutschland ‘Germany’ or Magyarország ‘Hungary’ are regarded as one word in German and Hungarian, respectively. The bisyllabic word biànhuà ‘change’ contains a lexical stem (biàn) and a grammatically used lexical element (huà, denoting transition). Hence, it is reasonable to consider biànhuà as one word. Finally, díquè ‘really’ is clearly a twosyllable word whose elements cannot be separated. This would suffice to show that Chinese is not a monosyllabic language at all. Following the conventions of pinyin orthography, we get nine words for twelve characters. The conventional rules of pinyin require a functional element like de, a universal subordinator, zhè, a determiner, and an adverb like hên to be written separately. But, as in the case of Western orthography, this does not account for actual Chinese speech. The functional particle de is always uttered together with some other word; in fact, its tone depends on that of the adjacent word with which it co-occurs. Similarly, degree adverbs as hên are always attached to the word they modify. Finally, determiners as zhè are also attached in pronunciation to the word they determine. In order to show the actual pronunciation of this sentence, we could relax some of the rules of the pinyin orthography to obtain: (2) zhǀngguó zhèjî niánde biànhuà díquè hêndà China these several years GEN change really very big ‘China underwent big changes during the past several years’

In this case, we have only six words; exactly one half of the number of characters in the Chinese writing version.

From Agglutination to Polysynthesis

135

As a result of the preceding discussion, it is easy to see that the characterization of Chinese as monosyllabic comes from the idiosyncratic properties of Chinese writing. Hannas (1997: 176) explicitly states that the idea that Chinese is monosyllabic “is based not on the language as it is spoken (and, presumably, as internalized by its speakers), but rather on the way the language was and is conventionally written”. A detailed criticism of this myth can be found in DeFrancis (1984: 177-189). There are reports saying that monosyllabic words account for just 12 percent of the contemporary Chinese lexicon; other estimates vary between 5 per cent and 30 percent.2 In a few words, “clearly, the notion that Chinese, absolutely or even relative to other languages is monosyllabic is untenable” (Hannas 1997: 176). From the above discussion it should be clear by now that it is Chinese writing what is monosyllabic and isolating, not Chinese speech. Therefore, the characterization of Chinese as predominantly monosyllabic and isolating is a serious mistake caused by the written language bias in traditional linguistics (Moreno-Cabrera 2008). The same point can be made for Vietnamese. Thus, although Vietnamese is written in a variety of the Roman alphabet, its orthographic conventions require each syllable to be written separately. As a result, a typical Vietnamese text typically consists, with only a few exceptions, in a series of syllables separated by blank spaces without any indication of what syllables constitute a single prosodic word, just as in the case of Chinese writing. It is worth noting that, in Vietnam, Chinese writing and a local variant thereof have been in use until the first decades of the 20th century (Dìn-Hoà 1996: 691). In their study of the prosodic word in Vietnamese, Schierin and Bickel state:  Although there is a strong tendency for monosyllabic grammatical words, certain word classes (place names, loan words) as well as compounds and reduplicated words may be polysyllabic in Vietnamese (Schiering and Bickel 2007: 6)

 In order to illustrate this point, the authors offer the following examples: 

2

See Hannas (1997: 176) for the relevant sources.

136

Chapter Four

Fig. 4.2. Possible word forms in Vietnamese

Schierin and Bickel (2007) also point out that some of the usual criteria for defining polysyllabic words such as non-interruptability and ordering constraints do not always apply in a strict way in Vietnamese. Nevertheless, it cannot be said that Vietnamese is a monosyllabic language.

2.3. Reichenbach’s contention: all languages are agglutinative An agglutinative language is usually defined as a language in which words are formed by joining morphemes together. The following characterization can be found in the English Wikipedia: An agglutinative language is a form of synthetic language where each affix typically represents one unit of meaning (such as “diminutive”, “past tense”, “plural”, etc.), and bound morphemes are expressed by affixes (and not by internal changes of the root of the word, or changes in stress or tone). Additionally, and most importantly, in an agglutinative language affixes do not become fused with others, and do not change form conditioned by others. Synthetic languages that are not agglutinative are called fusional languages; they sometimes combine affixes by “squeezing” them together, often changing them drastically in the process, and joining several meanings in one affix (for example, in the Spanish word comí [I ate], the suffix -í carries the meanings of indicative mood, active voice, past tense, first person singular subject and perfective aspect). 3 (http://en.wikipedia.org/wiki/Agglutinative_language)

If we go back to the Chinese examples discussed in the preceding section we will easily infer that this language is agglutinative. We have 3

This page was consulted in February 2011



From Agglutination to Polysynthesis

137

seen that Chinese grammatical words are usually monosyllabic and attach to the preceding or following word in an agglutinative way. So, this language can qualify as agglutinative following the preceding definition. What is clear is that, in general, Chinese is not fusional: its grammatical morphemes do not fuse with the lexical word they are attached to. Indeed, it can also be shown that fusional languages are also agglutinative. Take, for instance, the Spanish example introduced in the preceding quotation. There is certainly morphological fusion in the comí verbal form. But a form like this usually appears in real conversational Spanish with some agglutinative elements attached to it as in me lo comí ‘I ate it up’ normally uttered as [melokomí], that is, as one prosodic word (see chapter 1, this volume). This extremely usual form is distinctively agglutinative: a me (first personal pronoun in dative case) and a lo (third personal pronoun in accusative case) can be easily distinguished and isolated. A similar observation was made by the eminent philosopher and logician Hans Reichenbach in the first chapter of his introduction to symbolic logic: The grammatical definition of word as a group of letters separated from others by an interval is not satisfactory: what is one word in one language may be expressed by several words in another, whereas propositions are always translated into propositions. The German language is noted for its long compound words, such as Eisenbahnknotenpunkt, which means ‘railroad junction’ in English. The rule stating that the words ‘rail’ and ‘road’ are written in one word, whereas ‘junction’ is written separately, is merely conventional. The Turkish language, which uses many suffixes, sometimes expresses a whole sentence in one word. Thus the Turkish, word ‘alabÕlece÷im’ is a sentence and means ‘I shall be able to buy’. We should not forget that in all languages the division into words more or less disappears as soon as we speak, since in talking we make no intervals between most words (Reichenbach 1947: 5)

This author noted the written language bias of classical morphological typology and pointed out that in real speech the division into words as established in the written language tends to disappear. This observation is absolutely right. In the Spanish example discussed above, the written expression me lo comí ‘I ate it up’ corresponds to one single word in colloquial Spanish. As we have seen, this conclusion is also valid for Chinese and Vietnamese. All these observations lead us to suspect that, in fact, all human spoken languages are agglutinative to a greater or lesser degree. This does not mean that all languages are morphologically alike: there are indeed clear

138

Chapter Four

differences in morphology between them. The degree of morphological fusion is a useful parameter for establishing typological differences in the field of morphology (Comrie 1981: 48, Bybee 1985: 44-46, Whaley 1997: 133, Bossong 2004). What I am questioning here is the relevance of isolation and agglutination as parameters of cross-linguistic morphological variation. In section four I will also question the concept of polysynthesis in a similar way. Certainly, Reichenbach was right when he said that in speech the orthographic division into words tends to disappear. But this does not mean that speech consists of an uninterrupted flow of endless utterances. The physiological constraints of human speech must determine the maximal and minimal length of linguistic utterances. They are therefore expected to be delivered in manageable speech units both from a production and a reception point of view. This aspect will be discussed in the following section.

3. Internet packets and Miller’s Magical Number Seven, Plus or Minus Two In his book on linguistic complexity, Dahl (2004) proposed the following interesting analogy: Files sent on the Internet are delivered in “packets” of a pre-specified length. Suppose in fact similar principles hold for spoken language, that is, that utterances are delivered in units we can call “packets”, and that those represent the maximal domain of incorporation processes (what does not fit in one packet cannot be incorporated) (Dahl 2004: 257)

Before considering the linguistic consequences of this idea, let me briefly speak about Internet packets. The key idea of packet switching is that of each communication is divided into individual, equal-sized packets. These packets are then sent individually to their destination through the network, and the entire message is reassembled when all the packets arrive. On the Internet, a typical packet length is about one kilobyte, or a thousand characters. Consequently, a large message may be divided into thousands of individual packets. The packets can be fragmented into smaller units depending on the maximum amount of data that can pass through a certain layer. The fragmentation of transmitted information into packets has to do with the notion of maximum transmission unit (MTU). This unit can be defined in the following way: the MTU of a

From Agglutination to Polysynthesis

139

communications protocol of a layer is the size of the largest protocol data unit that the layer can pass onwards. The following diagram shows an example of fragmentation of transmitted data following the Internet Protocol in his fourth version (Ipv4): Original IPv4 Packet Identification: 1956 Fragment Offset = 0 Total Lenght: 1500 1480 Octets of Data

IPv4 Fragmentation MTU = 666

1st Fragment Identification: 1956 Fragment Offset = 0 Total Lenght: 660 640 Octets of Data

2nd Fragment Identification: 1956 Fragment Offset = 80 Total Lenght: 660 640 Octets of Data

3rd Fragment Identification: 1956 Fragment Offset = 160 Total Lenght: 220 200 Octets of Data Fig. 4-3. An example of fragmentation of transmitted data

As can be seen in the diagram, the maximum transfer unit equals in this case 666 bytes of data. The original Ipv4 packet has a total length of 1500 bytes including 1480 bytes of data. Therefore, this packet has to be fragmented into two units of 660 bytes and one of 220 bytes, conveying 640, 640 and 220 bytes of data, respectively. This equals the 1480 bytes of data in the original packet. The additional amount of bytes contains information in headers used to identify the fragments, the order in which they should be reassembled and additional information about the source and the destination. All this fragmentation is clearly determined by the limitations and constraints of the physical medium through which the data will circulate in order to reach its destination.

140

Chapter Four

In speech we also have a series of data that must be transmitted by means of the activity of the speech organs producing an acoustic signal that will be physically and psychically processed by the hearer. It is clear that this signal is subject to strong production and perception constraints that will severely limit the amount of information that can be sent in one single communicative action both in production and processing. In an influential paper, Miller (1956) discussed some of the limitations that short-term memory (STM) imposes on human information processing. Miller noticed that the memory span of young adults was around seven elements called chunks, regardless whether the elements were digits, letters, words, or other units. Later research revealed that span does depend on the category of chunks used (e.g. span is around seven for digits, around six for letters, and around five for words), and even on features of the chunks within a category. In a comprehensive paper on Miller’s proposal, Cowan4 (2000) gives the following definition of chunk, taken from Miller’s paper: A chunk must be defined with respect to associations between concepts in long-term memory. I will define the term chunk as a collection of concepts that have strong associations to one another and much weaker associations to other chunks concurrently in use (Cowan 2000: 89)

Cowan tries to be more specific about the definition of chunks in serial recall experiments as follows: Given the importance of binding, a chunk might be described in these procedures as a direct association between adjacent items, with together become bound to only one higher-order serial position or spatial location in the stimulus array as a group (Cowan 2000: 157)

In the experiments of serial recall, the binding of items to particular serial positions within lists comprising the material that must be held in the focus of attention seems to play an important role. This has a clear direct counterpart in modern theoretical linguistics: it is the adjacency principle. This principle requires case assigners to be adjacent to the phrase that receives Case (Cook and Newson 2007: 158). In addition, there is an important principle in current theoretical linguistics that seems to be related to the constraints discussed above. This is the locality principle, defined by Moro in the following way: 4

I thank Cedric Boeckx for directing me to this paper and providing me with a copy.

From Agglutination to Polysynthesis

141

If recursion in a language guarantees that there is no upper limit to the amount of computable structure (except for the biological limits of memory and length of life), then locality is a limit on dependences: it is a filter that eliminates some potentially available dependencies and thereby reduces the quantity of computable information in a syntactic structure (Moro 2008: 88, author’s italics)

Following Moro, this locality principle has a great biological significance not only in language production and processing but also in language acquisition: When children start to learn the use of the pronominal system, the locality principles in their syntax make it much easier for them to avoid taking into consideration all the possible pronominal co-references and converge toward the correct interpretation. The number of co-references that are a priori possible is drastically reduced (Moro 2008: 196)

Pothos and Juola (2000) have shown that Cowan’s 7 (plus or minus 2) limit for short-term memory is operative in word processing. Their analysis of eight languages (Bulgarian, Czech, Dutch, English, Estonian, French, Gaelic and German) based on words confirms a limitation of STM limit of about four chunks. These authors conclude: Linguistic dependency structure is contained primarily with a range of not more that about five word tokens, which would be consistent with the capacity of STM span being of the order of 3-5 chunks. This provides additional source of evidence for Cowan’s proposal (Pothos and Juola 2000: 139)

As a result of the preceding discussion, it seems reasonable to assume that the necessary fragmentation of spoken discourse into minimal meaningful units is due to the constraints imposed by the physical and psychical media through which linguistic signals are transmitted. Experts on the biology of language have indeed identified some of these physiologically-determined units. One of them is the breath-group posited by Lieberman (2006: 324-340). Speech production must be made compatible with respiration, one of the basic biological functions of human beings. Lieberman characterizes the breath-group as follows: A breath-group is a phonologic construct that attempts to take account of biological constraints reflecting the mark of evolution. […] The basic ‘normal’ breath-group reflects the biologic constraints imposed by breathing. The prosody of a sentencethe acoustic cues that delimit the

142

Chapter Four end of a sentenceusually follow from articulatory maneuvers that represent the state of minimal departure from quiet respiration. […] This requires anticipating the approximate length of the sentence before we utter a single word (Lieberman 2006: 329)

From the above discussion, it becomes clear that the spoken word, which seems to be midway between the phonological word (Dixon and Aikhenvald 2002: 13-18, see chapter 1 this volume), the semantic word and the psycholinguistic word (Packard 2000: 7-13), could be identified with those information chunks presenting an optimal solution to the constraints on speech production and perception. These meaningful minimal units of speech must comply at the same time with the wellformedness requirements of the two interfaces posited by current research in the Chomskyan minimalist program: the articulatory-perceptual interface (A-P) and the Conceptual-Intentional interface (C-I) (Hornstein, Nunes and Grohmann 2005: 8-9). This means that spoken words should function at the same time as units of the automatized speech articulation and perception processes and as units of the mental processes building up the semantics of a linguistic expression. Miller, Cowan and other authors refer to words as information chunks. But they do not give a detailed and precise definition of the spoken word as a chunk. In the following section, I review a proposal which, in my opinion, can be interpreted in the direction sketched in this section.

4. On the linguistic nature of the spoken word: Dahl’s packets as Miller’s chunks In his book on the growth and maintenance of linguistic complexity, Dahl (2004: 209-260) discusses at length morphological incorporation, a phenomenon that gives rise to the so-called incorporating and polysynthetic linguistic type, that will be discussed in the following section. Dahl uses the concept of incorporating pattern defining it as follows: I shall use the term incorporating pattern as a general term for a one word pattern which contains more than one lexical element, including both phenomena that have been called incorporation in earlier treatments and those that are usually called ‘compounds’ (Dahl 2004: 210)

In his discussion of incorporating patterns, Dahl offers a detailed linguistic specification of the vague intuition stated by Reichenbach in the quotation introduced in the second section of this chapter. Indeed, as

From Agglutination to Polysynthesis

143

Reichenbach said, the limits between orthographic words tend to disappear in speech, but they do so in an extremely constrained way as Dahl (2004: 252-259) shows. Incorporating patterns show constraints on productivity: many of them are unproductive or have a limited productivity. They also show a tendency towards lexicalization and tend to refer to unitary concepts denoting stereotypical, institutionalized or permanent objects and activities. In addition, incorporation patterns show a very limited syntactic structure and cannot usually be subjected to productive syntactic processes such as modification. Furthermore, the elements inside these incorporation patterns cannot be emphasized or focused. In order to identify and characterize the linguistic unit produced by the incorporating patterns and following the Internet analogy I introduced at the beginning of the second section of this chapter, Dahl introduces the packet unit. Utterances are delivered in units that we can call ‘packets’. This scholar gives us the following definition: A packet would typically be a word or a short phrase, pronounced as at most one prosodic unit (with a unitary intonational contour) (Dahl 2004: 257)

These packets represent the maximal domain of incorporation processes, and as a consequence, what does not fit into one packet cannot be incorporated (Dahl 2004: 257). These units have two salient properties (Dahl 2004: 257): (i) (ii)

They are highly integrated; in production they probably correspond to a single command at some level. There are strong constraints on their internal complexity. They are like containers with limited capacity.

When a complex linguistic expression is built, a speaker puts as much structure as possible into each packet (Dahl 2004: 257). This author also offers us a list of the main constraints on packets (Dahl 2004: 258): (i)

Lexical referential expressions (LRE) do not easily fit into packets. A LRE is a noun phrase containing at least one lexical element and having specific reference. Examples of LRE: the pen, John, Mary. (ii) An element of a packet may not be independently focused, emphasized (F). (iii) An element of a packet should not have an internal syntactic structure of its own (SE).

In order to illustrate these constraints, Dahl proposes the examples listed below. I have made two modifications to Dahl’s sentences. In order

144

Chapter Four

to highlight the fact that one packet corresponds to one speech unit, I have modified the orthographic conventions of English regarding the separation of words. All the distinct orthographic words belonging to a packet are written as one word (I also give in angular parentheses the conventional orthographic form). In addition, I provide parallel Spanish examples (that are a translation of the corresponding English examples) and I apply the same orthographic criteria as in English: (3) Shedrinkstea / bèbeté  one packet. (4) Mary drinkstea / María bèbeté  two packets (LRE,1). (5) Mary drinks thetea / María bebe elté  three packets (LRE,2). (6) Mary drinks TEA/ María bebe TÉ  three packets (F,1). (7) Mary drinks teaandcoffee / María bebe tèycafé  three packets (SE,1). (8) Ella bebeté  two packets (F,1).

The abbreviations in parentheses refer to the constraints determining the number of packets in each sentence. For example, following Dahl, in the sentence Mary drinks tea there are two packets (for this reason this sentence appears in the list as Mary drinkstea). The noun Mary is a LRE and cannot be integrated into the drinkstea packet. Moreover, the noun tea can be integrated into the drinkstea packet because it is not a lexical referential expression; once it is provided with a definite article as in the tea it becomes an LRE and therefore cannot be integrated into the drink packet, as can be seen in (5), with three packets. The pronoun she is not an LRE (it is referential but it does not have lexical contents), so it can be incorporated into the drinkstea packet, as can be seen in example (3). The words in capitals are focused and, therefore, constitute one single packet, since a focused element cannot be part of a bigger packet. In this sense, it is interesting to compare the English example (3) with the Spanish example (8). They roughly translate each other, but there is one important difference concerning the status of the personal pronoun. In Spanish, the subject pronoun does not normally appear in normal or neutral speech as can be seen in the Spanish translation of example (3). When it does, it is focused, so in sentence (8) we emphasize the fact that ella ‘she’ drinks tea and not él ‘he’ or ellos ‘they’. Since a subject pronoun in Spanish is always focused, it must constitute a single packet; therefore, sentence (8) consists of two packets and not of one like in English. If we look at the conventional written forms of the given sentences we can appreciate a drastic difference in the number of units. For example,

From Agglutination to Polysynthesis

145

sentence (7), Mary drinks tea and coffee, contains three packets (Mary, drinks, teaandcoffee), since the expression tea and coffee has a syntactic structure and, therefore, cannot appear inside a bigger packet. In spite of this, this sentence is segmented into five words following the conventions of the written language. The same applies to the corresponding Spanish example. It is interesting to point out that these packets comply with the definition of chunk seen in the preceding section. Following this definition, the elements inside this unit have strong associations to one another and much weaker associations to other chunks. Consider for example the two packet sequence drinks thetea. It is clear that the article the has a strong association to tea and a much weaker association to drinks. Moreover, the and tea are adjacent. All the packets identified by Dahl abide by these two conditions on chunks. The written form of language has had and still has a tremendous influence on linguistic science. This point can be illustrated with the famous example used by Sapir (1921) in his influential presentation of language: the farmer kills the duckling. After analyzing in detail the morphological, syntactic and semantic aspects of this example, in six dense pages Sapir concludes: In this short sentence of five words there are expressed, therefore, thirteen distinct concepts, of which three are radical and concrete, two derivational, and eight relational. (Sapir 1921: 89)

It is certainly true that the example has five orthographic words. But are there really five words in this sentence? If we use Dahl’s packet concept and apply the proposed criteria, it is clear that in speech this sentence should be divided into three packets: thefarmer, kill and theduckling. These packets are real minimal units that can be identified with spoken words. In English orthography, the farmer kills the duckling has five words, but in everyday speech it is clear that it consists of only three spoken words; these three words correspond to the three radical and concrete concepts identified by Sapir in his insightful discussion. This point has a great importance, since we can reasonably expect the number of elements inside a packet to be drastically constrained, as we assume that we are dealing with a basic unit in the linguistic speech process, both from an articulatory and a perceptual point of view. In order to determine the number of elements inside a packet, I will consider all the meaningful constituents; these include traditional lexemes and morphemes. In the packets identified in the given examples we find up to four elements. For example, shedrinkstea has four meaningful elements:

146

Chapter Four

she, drink, s, and tea. Other packets have three elements (teaandcoffee), two elements (thetea), or only one (Mary). We can speculate from these data that the maximal number of elements inside a packet can be estimated in about five components. This roughly coincides with Miller’s number. But in order to further explore the limits of the maximal number of elements in the spoken word we must consider languages of the so-called polysynthetic type, in which we can find spoken words with a spectacular accumulation of meaningful elements.

5. Polysynthesis and De Reuse’s productive non-inflectional concatenation Polysynthetic languages are said to have holophrastic words, i.e. words containing all the information conveyed by whole sentences in languages of other types.5 Let me introduce now an example from Chukchi (Skorik 1961: 102), a polysynthetic, incorporating and agglutinating language: (9) Tëmeyngëlevtpëgtërkën t-ë-meyng-ë-levt-pëgt-ë-rkën 1.SG.SUBJ-great-head-hurt-PRES.1 ‘I have a fierce headache’

This word consists of five meaningful elements. Two of them are morphemes (the first and the last elements) and the rest are lexemes. Some of these elements are joined by the neutral vowel ë with no morphological function. It can also be seen that this word qualifies quite well as a packet in Dahl’s sense. 6 There are neither lexical referential elements, nor emphasized or focused elements; moreover, there are no intonational pauses inside it.

5 Mithun (2009: 3) notices that the term polysynthesis was coined in 1816 by Duponceau to describe words in American languages containing many morphemes. The traditional concept of polysynthesis should be carefully distinguished from the parameter of polysynthesis proposed by Baker (1996), which can be properly understood only from the perspective of the generative theory in which this parameter was proposed. Its formulation is rather technical and reads, in the simplest possible terms, as follows: “Every argument of a head element must be related to a morpheme in the word containing that head” (Baker 1996: 14). 6 I refer the reader to the previous section.

From Agglutination to Polysynthesis

147

Perhaps the paradigmatic example of polysynthetic languages is provided by the widely extended Eskimo family. In an interesting and innovative paper, De Reuse (2009) examines the traditional concept of polysynthesis and proposes that this phenomenon is a by-product of a concatenation process. In De Reuse’s paper, it is called productive noninflectional concatenation (PNC). This process is different from both inflectional and derivational processes and also from syntactic concatenation. In our view, it is a specific formulation of Reichenbach’s vague observation that the limits between words tend to disappear in colloquial speech. In the following table (De Reuse 2009: 22), the main differences between PNC and the morphological and syntactic processes are summarized.

Fig. 4-4. Differences between PNC and morphological/syntactic processes

De Reuse illustrates these properties using expressions from Central Siberian Yupik (CSY), a language of the Eskimo family. These sentences come from the colloquial everyday language, since this language lacks a written tradition similar to that of the more familiar European languages. I will consider some of the CSY examples proposed by De Reuse together with some corresponding Spanish examples. These Spanish examples do not come from the written language; they are also colloquial everyday expressions. This is coherent with the point of view adopted in this chapter; we are interested in the biologically constrained units of normal everyday speech, not in the culturally-conditioned units of written language.

148

Chapter Four

As happens with syntactic processes and unlike morphological processes, PNC is a productive process. In order to exemplify this point De Reuse proposes the following CSY expression (De Reuse 2009: 23). (10) Productivity in Central Siberian Yupik Eskimo (CSY) neghyaghtughyugumayaghpetaa negh-yaghtugh-yug -uma-yagh -pete -aa eat -go.to.V -want.to.V-PAST-FRUSTR-INFRN-IND.3S.3S7(7) ‘It turns out s/he wanted to go eat it, but…’

This example contains seven meaningful elements and can be analyzed as a packet in Dahl’s terms. De Reuses points out that, in this example, only the lexeme negh and the morphemeaa are obligatory; the other elements can be optionally left out to produce at least thirty different words. In colloquial Spanish we can also find similar examples showing productivity. Consider the following colloquial Spanish expression: (11) [kenotelandáo] ke-no-te-lo-ha-n-da-o (8) CONJ.-NEG-2SIO-3SGDO-AUX-3PL-GIVE-PART8 ‘That they haven’t given it to you’

The conventional orthographic form of this sentence is and comprises six orthographic words. This expression is uttered with a single intonation contour and cannot be interrupted. If, for example, a pause is made after ke and this element receives a main accent we obtain two sentences (that can be roughly translated as ‘What? They haven’t given it to you’). Moreover, it is fully productive. As in the CSY case, the first four elements are optional; combining them we can obtain the following normal expressions in colloquial Spanish, among many others: (12) a. b. c. d. 7

[ketelandáo] ‘that they have given it to you’ [notelandáo] ‘they haven’t given it to you’ [nolandáo] ‘they haven’t given it’ [ketandáo] ‘that they have given you’, etc.

PAST = past tense, FRUSTR = frustrative, INFRN = inferential evidential, IND.3S.3S = indicative mood third person subject acting of third person object. 8 CONJ = conjunction; NEG = negation, 2SIO = second person singular indirect object, 3SGDO = third person singular direct object, AUX = auxiliary, 3PL = third person plural verbal marker, PART = participial ending.

From Agglutination to Polysynthesis

149

The CSY and Spanish expressions can also be analyzed as packets in Dahl’s sense containing seven and eight meaningful elements respectively. Recursion is characteristic of PNC. Here is an example from CSY proposed by De Reuse (2009: 24) (13) iitghesqesaghtiisqaa itegh -sqe -yaghtugh-sqe –aa (5) come.in-ask.to.v-go.to.v -ask.to.v-IND.3S.3S ‘Hei asked himj to go ask himk to come in’

In this expression, the -sqe ‘ask to V’ element is used recursively since it appears twice with a specific function in each occurrence. In colloquial Spanish we can also find examples of this type of recursion. For example, in the noun phrase we can easily find two occurrences of the Spanish definite article as in the following expression: (14) [losdelosníños] l-o-s-de-l-o-s-niñ-o-s (10) ART-DESIN-PL-PREP-ART-DESI-PL-child-DESI-PL9 ‘those of the children’

In this expression the definite article los appears twice. The scope of the first one is de los niños and niños is being determined by the second one. This is a perfectly normal and productive expression in colloquial Spanish that in written form consists of four orthographic words: . Nevertheless, it can be analyzed as one packet in Dahl’s sense containing only one LRE (losniños). It is interesting to note that it contains ten meaningful elements, since the morpheme o, appearing three times, could be substituted for the morpheme a to obtain different feminine gender versions of this expression such as: losdelasniñas ‘those (masculine) of the little girls’ or lasdelasniñas ‘those (feminine) of the little girls’. This Spanish example can therefore be seen as a much more extreme case of PNC than that of CSY proposed by De Reuse. The concatenation property has already been illustrated with the CSY and the Spanish examples analyzed so far. It is also observed in the following examples. PNC also shows cases of variable order. De Reuse (2009: 25) illustrates this point with two CSY sentences: 9

ART = article, DESIN= desinential morpheme identified in traditional grammar as a masculine morpheme, PL = plural

150

Chapter Four

(15) aananiitkaa aane -nanigh -utke -aa (4) go.out-cease.to.V-V.on.account.of-IND.3S.3S ‘He ceased going out on account of it’ (16) aanutkenanighaa aane -utke -nanigh -aa (4) go.out-V.on.account.of-cease.to.V-IND.3S.3S ‘He ceased going out on account of it’

As seen in the above examples, the -utke lexical base can appear both before and after the -nanigh base. The expressions obtained have exactly the same meaning. In Spanish we can also find this property of PNC in expressions containing clitics and in colloquial speech. For example, the following two expressions are equivalent in Spanish (RAE and Asele 2009: 2663): (17) [tesebé]

2SGOBJ REFL see-3SG ‘you are seen’ (18) [setebé]

REFL 2SGOBJ see-3SG ‘you are seen’10

The first expression is considered vulgar and is not accepted in formal Spanish, but it exists and can occasionally appear in literary works (RAE and Asele 2009: 2663); it is completely equivalent to the second standard version. In addition, we find more extreme cases of this PNC variable order in colloquial Spanish. In these cases, a rather unexpected interaction with morphology can be observed, as in the following examples (RAE and Asele 2009: 195): (19) márche-n-se get out-3PL-REFL ‘get out of here (you all)’

10

2SG = Second person singular pronoun, OBJ = object, REFL = reflexive morpheme here used in an impersonal sense, INF = infinitive morpheme, 3SGDO = third person singular direct object pronoun, 3SG = third person singular inflectional ending.

From Agglutination to Polysynthesis

151

(20) márche-se-n get out- REFL-3PL ‘get out of here (you all)’

In the first form the third person plural morpheme -n precedes the reflexive pronoun se. This is the standard form. But a second version in which the third person plural inflectional morpheme follows the reflexive can occasionally be heard in colloquial Spanish, although it is considered vulgar and incorrect. The vulgar form márchensen ‘get out of here you all’ has been also documented (RAE and Asele 2009: 195). In it, there are two third person plural morphemes illustrating the recursivity property of PNC referred to in the above discussion. The interaction with external syntax is one of the properties of PNC. In the following example (De Reuse 2009: 26), an adjectival phrase modifies an incorporated noun. (21) atan aangelghiimeng ata -n aange-lghii -meng father-POSS.2S.ABS.SG be.big-INTR.PTCP-MOD.SG11 (22) qikmilguuq qikmigh-lgu -uq dog -have.N-IND.3S ‘Your father has a big dog’

The noun qikmigh ‘dog’ has been morphological incorporated into the lgu verbal base. In spite of that, the adjectival phrase aangelgiimeng ‘big’ is modifying the incorporated noun. This shows that the PNC concatenated elements are accessible to syntactic relations. The same property can be observed in Spanish PNC. For example, in the following standard Spanish sentence, a concatenated clitic pronoun (lo) anaphorically refers to a preceding proper noun (Juan): (23) A Juan lo v-i ayer to John 3SGDO see-1SG yesterday ‘I met John yesterday’

I give the Spanish standard orthographic form for the analyzed sentence. In colloquial Spanish, we find three packets or spoken words

11

POSS.2S.ABS.SG = absolutive singular case, second-person singular possessor, INTR.PTCP = intransitive participial morpheme; it, can be a nominalizer, MOD.SG = modalis singular, a type of oblique case.

Chapter Four

152

obtained by a PNC process. Below, I offer a rough phonetic transcription of this sentence: (24) [axuán lobí ayer]

The five orthographic words in the standard form are in fact an analysis of the three spoken words of the corresponding colloquial language expression. The colloquial form [lobí] (the orthographic ) is one spoken word consisting of three meaningful elements. The lo pronoun in this form is syntactically related to the [xuán] () component of the spoken word [axuán] (). Therefore, we can see that Spanish offers an example of interaction with syntax based on PNC expressions with incorporated pronouns, in a similar way to CSY does. The final feature characterizing PNC is category change. This is exemplified by the following CSY examples (De Reuse 2009: 26): (25) Verb: yughagh- ‘to pray’ Ÿ Noun: yughaghvig- ‘church’ Ÿ Noun: yughaghvigllag- ‘big church’ Ÿ Verb: yughaghvigllange- ‘to acquire a big church’

In Spanish we can also find similar enchained category changes: (26) Noun: precio ‘price’ Ÿ Noun: justiprecio ‘evaluation’ Ÿ Verb: justipreciar ‘evaluate’ Ÿ Adverb: justipreciadamente ‘in an evaluative way’

The second noun in this chain is obtained by composing the initial noun with the adjective justo ‘fair, right’. Although this compounding process is usually studied as a morphological derivation, it can be also seen as a result of a PNC process. The following category change is clearly derivative, but in the last item we can see again a composition with the adverbial -mente ending that can also be understood as the result of a PNC process. In his paper, De Reuse (2009: 27) maintains that, in fact, PNC is an additional type of morphology. In consequence, some sort of correspondence between PNC, inflection, and derivation is to be expected when comparing different languages, as we have just seen. Concerning the traditional linguistic typology positing a distinct polysynthetic type, De Reuse states the following: Another consequence of this view is that there is no such thing as a “polysynthetic type” of languages. The languages called “polysynthetic” are languages with large or notable amounts of PNC. The difference

From Agglutination to Polysynthesis

153

between polysynthetic languages and non-polysynthetic languages is thus a quantitative difference, not a qualitative one. The stereotypical “polysynthetic type” is represented by the languages that happen to have many (a hundred or more) PNC elements, such as CSY (De Reuse 2009: 28)

In addition, this author maintains that PNC is also present in languages that are not normally assigned to the polysynthetic type as I have noted above when discussing the Spanish examples: As mentioned already, another consequence of this view is that what is traditionally called derivational is now split into derivation and productive derivation, i.e. PNC. But one should not conclude from this that only polysynthetic languages have PNC elements, and that non-polysynthetic languages have derivation only. Indeed, PNC elements are not completely absent from non-polysynthetic languages. Some of the affixes traditionally called “derivational” in Indo-European languages are productive, and among these productive ones, some have recursion as well (De Reuse 2009: 28)

The Spanish examples in this section are precisely meant to prove this point.

6. Reaching the limits of the spoken word: an interlinguistic assessment of Miller’s number 7 Plus or Minus 2 In the preceding sections, Dahl’s packet analysis and De Reuse’s PNC have been presented and explained in order to clarify and specify the linguistic characteristics of the spoken word as a speech unit strongly constrained by human biological factors. One of these constraints has to do with the optimal number of elements that one spoken word can accommodate as a production and perception unit. In this chapter I have introduced and discussed Miller’s proposal that 7 plus or minus 2 is the optimal quantity of information that can be handled efficiently by the working short-term memory. I will try now to determine if this 5/9 figure can be used to characterize the maximum number of elements included in one spoken word or packet in Dahl’s sense. The following is a selection of the examples encountered so far in this chapter:

154

Chapter Four

(27) Chukchi Tëmeyngëlevtpëgtërkën. t-ë-meyng-ë-levt-pëgt-ë-rkën (5) 1.SG.SUBJ-great-head-hurt-PRES.1 ‘I have a fierce headache’ (28) Yupik neghyaghtughyugumayaghpetaa negh-yaghtugh-yug -uma-yagh -pete –aa (7) eat -go.to.V -want.to.V-PAST-FRUSTR-INFRN-IND.3S.3S12 ‘It turns out s/he wanted to go eat it, but…’ (29) Colloquial Spanish kenoteloandáo ke-no-te-lo-ha-n-da-o (8) CONJ.-NEG-2SIO-3SGDO-AUX-3PL-GIVE-PART13 ‘That they haven’t given it to you’ (30) losdelosniños l-o-s-de-l-o-s-niñ-o-s (10) ART-DESIN-PL-PREP-ART-DESI-PL-child-DESI-PL14 ‘those of the children’

The fact that the examples in colloquial Spanish contain the spoken words with the largest quantity of meaningful elements in this chapter is surprising, since we have also seen examples from languages of a wellknown polysynthetic reputation. On the other hand, all the examples of spoken words in this chapter present a number of meaningful elements consistent with Miller’s seven plus or minus two. In order to check this point on a more comprehensive empirical base, I will now consider all the examples given in another contemporary collective work on polysynthesis (Evans and Sasse 2002). In their introduction to this book, Evans and Sasse give some notable examples of a high degree of polysynthesis in one word. I include here the most salient cases:

12

PAST = past tense, FRUSTR = frustrative, INFRN = inferential evidential, IND.3S.3S = indicative mood third person subject acting of third person object. 13 CONJ = conjunction; NEG = negation, 2SIO = second person singular indirect object, 3SGDO = third person singular direct object, AUX = auxiliary, 3PL = third person plural verbal marker, PART = participial ending. 14 ART = article, DESIN= desinential morpheme identified in traditional grammar as a masculine morpheme, PL = plural

From Agglutination to Polysynthesis

155

(31) Bininj Gun-Wok (Australian, Evans and Sasse 2002: 2) Nga-ban-marne- yawoih-dulk-djobge-ng (7) 1SGSUBJ-3PLOBJ-BEN-again-tree-cut-PP15 ‘I cut another tree for them’ (32) Cayuga (Iroquoian, Evans and Sasse 2002: 2) T-e-hen-atat-hona’t-a-yé:thw-ahs (7) DU-FUT-3PLMASC:SUBJ-REFL-potato-SJ-plant-BEN:PUNC16 ‘They will plant potatoes for each other’

Although this form has eight elements, the a vowel is not a meaningful element; it is only a meaningless vowel joining two stems (SJ). Evans and Sasse (2002: 3) also mention the following West Greenlandic example: (33) West Greenlandic (Inuit) (Fortescue 1983: 97) Aliiku-sersu-i-llammas-sua-a-nerar-ta-ssa-galuar-paal-li (12) entertainment-provide-SEMITRANS-one:good:at-COP-say:that-REP-FUTsure:but- 3PLSUBJ/3SGOBJ-but17 ‘However, they will say that he is a great entertainer, but…’

In another West Greenlandic holophrastic word eleven elements can be found: (34) West Greenlandic (Inuit) (Fortescue 2002: 264) Aju-nngit-su-liur-vigi- nnit- tuar- tu-u-nngil-aq (11) be.good-NEG-PART-make-have.as.place.of-ANTIPAS-all.the.time-PARTbe-NEG-3SIND18 ‘He is not (much of) a benefactor’

In the Australian language Bininj Gun-wok, words with seven meaningful elements can be found:

15

1SG = first person singular, SUBJ=subject, 3PL = third person plural, OBJ = object, BEN = benefactive, PP= past perfective. 16 DU = dual, FUT = future, 3PL = third person plural, MASC = masculine, SUBJ = subject, REFL = reflexive, SJ = stem joiner, BEN = benefactive, PUNC = punctual 17 SEMITRANS = semitransitive, COP = copula, REP = repetitive, FUT = future, 3PL = third person plural affix, SUBJ = subject, OBJ = object. 18 NEG = negative, PART = participle, ANTIPAS= antipassive, 3S = third person singular, IND = Indicative.

156

Chapter Four

(35) Bininj Gun-wok (Evans 2002: 19) Aban-yawoih-warrgah-marne-ganj-ginje-ng (7) 1/3PL-again-wrong-BEN-meat cook-PP19 ‘I cooked the wrong meat for them again’

In this word, certain nominal roots have been incorporated inside the verb. The rest of the examples given by Evans in this paper are shorter, occasionally one can find words with five or four meaningful elements. Baker (2002) includes some words from the Australian language Ngalakan with five meaning elements. This is one of the examples: (36) Ngalakan (Baker 2002: 80) ngurr-mu-birtdin-marninyh-nya (5) 12PS-VEG-bread-make+FUT20 ‘We’ll made some lilyseed bread’

The Georgian verb can have a complex morphological structure. The following is an extreme example: (37) Georgian (Boeder 2002: 89) uku-she-mo-g-a-brun-a (7) back-into-hither-OBJECT- VERSION- return-3SG (aorist) ‘He made you return here’

The Yenisean language Ket also has a very complex verbal morphology. The following verbal form (Werner 1995: 156) contains eight elements: (38) Ket (Drossard 2002: 238) da-nan-bet-q-i-n-dang-et (8) 3SG:F:S-bread-make-DET-EV-IMPF-1PLO:O-make ‘She is making bread’

In Apache it is not difficult to find seven or even eight element words:

19

1/3PL= third person singular subject upon third person singular lower animate object, BEN = benefactive, PP= past perfective. 20 12 = first person inclusive affix, P = plural, S = subject, VEG = vegetable noun class, FUT = future

From Agglutination to Polysynthesis

157

(39) Apache (Jung 2002: 172) yi-l-k’e-i-n-o-‘an-na… (8) 3O-P-LEX-3O-ASP-3S-tell-NARR21 ‘They told them….’

We can also find eight elements in the following word from Rembarrnga, an Australian language: (40) Rembarrnga (Nordlinger and Saulwick 2002: 188) yarran-ba-bak-rdi-rditj-mae-rri-niny (8) 3>1P-P-BEN-DUP-return-STM-RR-PP22 ‘They all would return to us’

In this brief survey, we have encountered only two Greenlandic words having more components (11 and 12 respectively) than the ten elements of the Spanish word [losdelosniños] ‘those of the children’. Nevertheless, this Spanish word can be easily expanded by adding the a ‘to’ preposition to obtain the eleven element word [alosdelosniños] ‘to those of the children’. Note that in this complex Spanish word none of the components except the lexical item niños ‘children’ has its own accent: all these elements are clitics and cannot appear alone in normal Spanish speech; they need the support of a lexical item. We can also see that alosdelosniños can be analyzed as a packet in Dahl’s sense, since it has only one lexical referential component (losniños) and only one main accent on the first syllable of niños. In general, all the words in this section can be analyzed as packets in Dahl’s sense obtained through a PNC process in De Reuse’s sense. As can be easily seen by direct inspection, none of these words contains two lexical referential expressions, a focused component or a phrase with an internal syntactic structure. All of them seem to have been produced in agreement with Dahl’s description: When a complex linguistic expression is built, a speaker puts as much structure as possible into each packet. If there is too much structure for one packet, or if any other constraint on what can be done within a packet is violated, the material has to be divided into two (Dahl 2004: 257-258)

21 3O= third person object, P= postposition, LEX = lexical prefix, ASP = aspect, NARR = narrative 22 3>1 =third person acting on first person, P = Plural, BEN = benefactive, DUP = reduplication, STM = stem-forming affix, RR = reflexive/reciprocal, PP = past perfective.

158

Chapter Four

All the examples included in the two volumes on polysynthesis considered in this chapter seem to be also in agreement with Miller’s number seven plus or minus two. The majority of the complex words in the so-called polysynthetic languages (and also in colloquial Spanish) include between five and nine components. We have seen in this chapter some examples of words containing ten, eleven and even twelve components, but these cases are by no means the norm, neither in the socalled polysynthetic languages nor in colloquial European type languages such as Spanish. Of course, the recursive rules of derivational morphology can be applied to obtain words of a virtually unlimited length. I include here an invented German example: (41) German (Heine and Kuteva 2007: 269) Autobahnrastplatztoilettenreinigungspersonalbedarfabstellschrankschlüsseld iensttelefonnummer

This monster word contains at least fourteen meaningful elements: (42) Auto-bahn-rast-platz-toiletten-reinigungs-personal-bedarf-abstell-schrankschlüssel-dienst-telefon-nummer (14) car-line-rest-place-toilet-cleaning-staff-need-storing-locker-key-servicephone-number ‘the phone number of the key service for lockers of the cleaning staff of public conveniences of rest places of highways’

Heine and Kuteva point out about this word: We have never heard a German speaker uttering this compound, and it probably never will be uttered; however, it is a grammatically correct instance of productive recursive compounding (Heine and Kuteva 2007: 269)

The adjustment of the spoken word to Miller’s seven (plus or minus two) figure is by no means an unexpected and surprising result. The limits of the spoken word must be determined by the producing and processing limitations of humans. This means that the spoken word as a producing and processing unit should be very similar in all languages, since such limitations are exactly the same for all humans. In fact, we have seen that colloquial Spanish (the same could be said concerning colloquial English, French, Russian or German) is very similar to languages of the Eskimo family, with respect to the complexity of spoken words.

From Agglutination to Polysynthesis

159

7. Conclusions In this chapter I have discussed the role of written and spoken words for establishing a sound typology of language. The conclusion is that neither the written nor the spoken word are useful to build up a classification into language types based on morphology. I have tried to show that the classical typological classification into an isolated, an agglutinative, an inflective and a polysynthetic type, lacks an empirically adequate basis. As shown in the first section of this chapter, the isolating (and monosyllabic) type is based on written language conventions completely alien to the true linguistic nature of the languages they transcribe. Chinese and Vietnamese are not isolating and monosyllabic languages. All that can be said is that the Chinese and the Vietnamese take the syllable as the basis of their respective writing systems. I have also argued that the so-called agglutinative language type is also a myth. The writing conventions of European languages have a clear analytic bias and tend to separate, in the written form of linguistic expression, constituents that are normally grouped together in speech. So the English expression the farmer kills the duckling is segmented into five elements in writing, but in normal speech only three elements can be properly justified (thefarmer, kills, theduckling). As Shockey (2003: 25, 32) shows, written expressions such as I should think or but I think in… appear in everyday informal speech as single spoken words agglutinating three and four meaningful elements, respectively. If we consider only standard written English we can easily conclude that this language is quite isolating, but if we pay attention to real spoken English, it will be evident that this language is predominantly agglutinative, like the rest of the spoken languages of the world. The second perspective is to be preferred if we follow one of the main tenets of modern linguistics: the primacy of spoken or signed language over written language. From this, it follows that the agglutination concept cannot serve as a criterion for a typological classification of languages. All languages are agglutinative. The differences between languages lie in the degree of morphological integration of the agglutinated grammatical elements; for example, in the spoken word thetoys, the plural ending s is clearly more integrated morphologically into the lexeme toy than the definite article the. This integration degree of the article can vary between languages. For example, in Classical Arabic, the definite article has a higher degree of morphological integration in the determined element, so in the noun phrase ‘al ja:ru l-hasanu ‘the good neighbor’(Fischer 1997: 213) both the

160

Chapter Four

noun (ja:ru ‘neighbor’) and the adjective (hasanu ‘good’) incorporate the definite article (al- in the first case and l- in the second case), unlike English. Some authors use this criterion to propose a gradual morphological fusion parameter (Comrie 1981: 48, Bybee 1985: 44-46, Whaley 1997: 133, Bossong 2004: 433). In the rest of the chapter I explored the biological constraints on agglutination determining a maximal meaningful unit for speech production and perception. In the second section, I considered a proposal concerning the necessity of delivering speech in certain units determined by the strong constraints imposed by the channels through with linguistic signals are transmitted. I presented Miller’s influential proposal concerning the number of information chunks that can be efficiently handled by the short-term memory of humans. In the third section, I tried to apply Miller’s findings about chunks to the analysis of speech into basic meaningful units, called spoken words here. I presented a programmatic and preliminary proposal by Dahl’s concerning these speech units, called packets by this author. It has been shown that Dahl’s proposal offers a promising point of departure in the enterprise of characterizing the main meaningful units of speech. In the fourth section I explained De Reuse’s proposal concerning the linguistic process originating spoken words. This process is characterized by this author as productive non-inflectional concatenation (PNC). In examining the different properties of this process, I argued, following De Reuse’s paper, that it must be operative in all languages and not only in those of the supposed polysynthetic type. In fact, I showed that colloquial Spanish qualifies as a polysynthetic language if we pay attention to the spoken word only. In the fifth section, I explored the limits of the spoken words as produced by the PNC process. On the basis of several examples of highly polysynthetic languages (including colloquial Spanish), I concluded that Miller’s number seven plus or minus two is adequate for characterizing the maximum number of elements that PNC driven spoken words can have. To sum up, in the present chapter I have shown that languages are very similar to each other in this respect, and that this similarity is determined by some of the biological constraints of human linguistic performance.

From Agglutination to Polysynthesis

161

References Baker, B. 2002, “How referential is agreement? The interpretation of polysynthetic dis-agreement morphology in Ngalakan”, N. Evans and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 51-85 Baker, M. C. 1996, The Polysynthesis Parameter, Oxford: Oxford University Press Boeder, W. 2002, “Syntax and morphology of polysynthesis in the Georgian verb”, N. Evans and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 87-111 Bossong, G. 2004, “Analytizität und Synkthetizität. Kasus und Adpositionen im typologischen Vergleich”, U. Hinrichs (ed.) Die europäischen Sprachen auf dem Wege zum analytischen Sprach typ, Wiesbaden: Harrassowitz, 431-452 Bybee, J. L. 1985, Morphology. A Study of the relation between meaning and form, Amsterdam: John Benjamins Comrie, B. 1981, Language Universals and Linguistic Typology. Syntax and Morphology, Oxford: B. Blackwell. Cook, V. J. and M. Newson 2007, Chomsky’s Universal Grammar. An introduction, Oxford: Blackwell Cook, C. and A. Johns 2009, “Determining the semantics of Inuktitut postbases”, M.-A. Mahieu and N. Tersis (eds.) (2009): Variations on Polysynthesis. The Eskaleut Languages, Amsterdam: John Benjamins, 149-170 Coulmas, F. 1996, The Blackwell Encyclopedia of Writing Systems, Oxford: Blackwell Cowan, N. 2000, “The magical number 4 in short-term memory: A reconsideration of mental storage capacity”, Behavioral and Brain Sciences 24, 87-185 Dahl, Ö. 2004, The Growth and Maintenance of Linguistic Complexity, Amsterdam: John Benjamins De Reuse, W. J. 2009, “Polysynthesis as a typological feature. An attempt at a characterization from Eskimo and Athabascan perspectives”, M.A. Mahieu and N. Tersis (eds.) Variations on Polysynthesis. The Eskaleut Languages, Amsterdam: John Benjamins, 19-34 DeFrancis, J. 1984, The Chinese Language. Fact and Fantasy, Honolulu: University of Hawai’i Press. Dìn-Hoà, N. 1996, “Vietnamese”, P. T. Daniels and W. Bright (eds.) The World’s writing systems, Oxford: Oxford University Press, 691-695

162

Chapter Four

Dixon, R. M. W. and A. Y. Aikhenvald 2002, “Word: a typological framework”, R. M. W. Dixon and A. Y. Aikhenvald (eds.) Word. A cross-linguistic typology, Cambridge: Cambridge University Press, 141 Drossard, W. 2002, “Ket as a polysynthetic language, with special reference to complex verbs”, N. Evans and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 223-256 Evans, N. 2002, “The true status of grammatical object affixes: evidence from Bininj Gun-wok”, Evans, N. and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 15-50 Evans, N. and H.-J. Sasse (eds.) 2002, Problems of Polysynthesis, Berlin, Akademie Verlag Evans, N. and H.-J. Sasse 2002, “Introduction: problems of polysynthesis”, N. Evans and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 1-13 Fischer, W. 1997, “Classical Arabic”, R. Hetzron (ed.) The Semitic Languages, London: Routledge, 187-219 Fortescue, M. 1983, A Comparative Manual of Affixes for the Inuit Dialects of Greenland, Canada, and Alaska, Copenhagen: Nyt Nordisk —. 2002, “The rise and fall of polysynthesis in the Eskimo-Aleut family”, N. Evans and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 257-275 Hannas, W. C. 1997, Asia’s Orthographic Dilemma, Hawai’i: University of Hawai’i Press Heine, B. and T. Kuteva 2007, The genesis of grammar. A reconstruction, Oxford: Oxford University Press Hornstein, N., J. Nunes and K. K. Grohmann 2005, Understanding Minimalism, Cambridge: Cambridge University Press Jung, D. 2002, “Clause combining in Apache”, Evans, N. and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag, 167-183 Lieberman, Ph. 2006, Toward and evolutionary biology of language, Cambridge (Mass.): The Belknap Press of Harvard University Press. Mahieu, M.-A. and N. Tersis (eds.) 2009, Variations on Polysynthesis. The Eskaleut Languages, Amsterdam: John Benjamins. Miller, G. A. 1956, “The magical number seven, plus or minus two: Some limits on our capacity for processing information”, Psychological Review 63, 81-97 Mithun, M. 2009, “Polysynthesis in the Arctic”, M.-A. Mahieu and N. Tersis (eds.) Variations on Polysynthesis. The Eskaleut Languages, Amsterdam: John Benjamins, 3-17

From Agglutination to Polysynthesis

163

Moreno-Cabrera, J. C. 2008, “The Written Language Bias in Linguistic Typology”, Cuadernos de lingüística 15, 117-137 Moro, A. 2008, The Boundaries of Babel. The Brain and the Enigma of Impossible Languages, Cambridge (Mass.): The MIT Press Nhàn, N. T. 1984, The syllabeme and patterns of word formation in Vietnamese, PhD Diss., NYU Nordlinger, R. and A. Saulwick 2002, “Infinitives in Polysynthesis: the case of Rembarrnga”, N. Evans and H.-J. Sasse (eds.) Problems of Polysynthesis, Berlin: Akademie Verlag,185-201 Packard, J. L. 2000, The morphology of Chinese. A linguistic and Cognitive Approach, Cambridge: Cambridge University Press. Pothos, E. M. and P. Juola 2000, “Linguistic structure and short term memory”, Behavioral and Brain Sciences 24(1), 138-139 RAE and Asele 2009, Nueva Gramática de la Lengua Española, Madrid: Espasa Reichenbach, H. 1947, Elements of Symbolic Logic, New York: Dover. Sapir, E. 1921, Language. An Introduction to the Study of Speech, New York: Harcourt Brace & Company Schiering, R. and B. Bickel 2007, “Does Vietnamese have prosodic words? A Mon-Khmer development and its typological significance”, Austroasiatic Workshop (Available at http://www.rene.punksinscience.org/SchieringBickelAustroasiatic-Handout.pdf) Shockey, L. 2003, Sound Patterns of Spoken English, Oxford: Blackwell Skorik, P. Y. 1961, Grammatika þukotskogo jazyka: Fonetika i morfologija imennyx þastej reþi , Leningrad: Nauka Thompson, L. C. 1963, “The Problem of the Word in Vietnamese”, Word 19, 39-52 Tóth, A. 2007, “Are all agglutinative languages related to one another?”, The Hague: Mikes International (Available at http://mek.niif.hu/05500/05523/05523.pdf) Unger, M. 1990, “The very idea: the notion of ideogram in China and Japan”, Monumenta Nipponica 45, 391-411 —. 2003, Ideogram: Chinese characters and the myth of disembodied meaning, Honolulu: University of Hawai’i Press Whaley, L. J. 1997, Introduction to Typology. The Unity and Diversity of Language, London: SAGE Werner, H. 1995, Zur typologie der Jenissej-Sprachen, Wiesbaden: Harrassowitz

CHAPTER FIVE WORDS AS CONSTRUCTIONS: SOME REFLECTIONS IN THE LIGHT OF CONSTRUCTIONIST APPROACHES* FRANCISCO GONZÁLVEZ-GARCÍA

1. Introduction: Back to constructions The origins of Construction Grammar (henceforth CxG) can be traced back to the publication of two key seminal works at the University of California at Berkeley, USA, in the late 80’s: the study of the family of existential there constructions in English by George Lakoff (1987: 462585) (e.g. There’s Harry on the porch), and the publication in Language only one year later of a leading article by Charles Fillmore, Paul Kay and Mary O’Connor (1988[2003]) on scalar inferential constructions with let alone (e.g. Max won’t eat shrimp, let alone squid). An important aspect shared by these two works is the vindication of the centrality of constructions as free-standing theoretical entities with their own inherent meaning and/or function. This stance is diametrically different from the Chomskyan conception of constructions as mere taxonomic artifacts which may prove useful in linguistic description but which are nonetheless devoid of any theoretical status and explanatory power (Chomsky 1995: 170). Thus, in stark contrast to Chomskyan linguistics, CxG concentrates on the rehabilitation of the notion of *

This research has been financed by research projects FFI2010-17610/FILO and FFI2010-19380/FILO of the Spanish Ministry of Education and Science as well as research project PGDIT-INCITE09 204 155PR of the Xunta de Galicia. A preliminary version of this chapter was presented at the Jornadas Sylex 2, “Sobre la naturaleza de la palabra”, held at the University of Zaragoza over the period 7-8 October 2010. I should like to thank the organizers of this event for having invited me to take part in it as well as the audience for most constructive suggestions and criticisms. Any remaining error or inaccuracy is my sole responsibility.

Words as Constructions

165

construction as a key unit in linguistic theory and thus central to a comprehensive description and explanation of the knowledge that language users have and the different uses they put language to. The considerable impact that the revitalization of the notion of construction has had on the present-day linguistic scenario can be clearly seen in the fact that this notion has been invoked by formalist and functionalist linguists alike. Thus, at the formalist extreme, of particular relevance are neoconstructionist proposals such as the one formulated by Hagit Borer, which aim to combine the Minimalist Program (Chomsky 1995) with the notion of construction (Borer 2003) or recent works by Booij (2010, 2013), which propose a promising synthesis of CxG(s), grammaticalization theory, the Simpler Syntax model (Culicover and Jackendoff 2005) and psycholinguistics. On the functional end, particularly outstanding is the Lexical Constructional Model developed by Ricardo Mairal Usón and Francisco José Ruiz de Mendoza (Ruiz de Mendoza and Mairal Usón 2008, Mairal Usón and Ruiz de Mendoza 2009), which draws on functionalist models such as Robert Van Valin’s Role and Reference Grammar (Van Valin 2005), on the one hand, and cognitivist/constructionist approaches such as (Cognitive) Construction Grammar (Goldberg 1995, 2006) and Lakoff’s Idealized Cognitive Models (Lakoff 1987), on the other (see www.lexicom.es for further details). Half-way between the two extremes lies the Simpler Syntax model proposed by Ray Jackendoff and Peter Culicover (Culicover and Jackendoff 2005, Jackendoff 2010), whose main strength, in our opinion, lies in demonstrating the viability of a parallel architecture that integrates in a fruitful, coherent way a number of assumptions of Chomskyan linguistics, Cognitive Linguistics and CxG(s). The very existence of constructionist approaches like these is but a clear indication of the crucial role that the notion of construction has played in bridging the gap between moderate Chosmkyan positions, on the one hand, and functionalist, cognitivist and/or constructionist approaches, on the other, within functional-cognitive space in the 21st century (see further Gonzálvez-García and Butler 2006). This chapter is structured as follows: section 2 provides an overview of GxC(s), with special focus on the different strands of constructionist approaches in the present-day linguistic scenario. Section 3 is concerned with the foundational assumptions that bind these strands together in a family of constructionist approaches. Section 4 concentrates on the definition and status of constructions in CxC(s). At a higher level of specificity, section 5 examines the implications of GxC(s) for the study of morphology, with special focus on the analysis of compounds and the phenomenon of coercion in contemporary Spanish. The final part of this

166

Chapter Five

chapter illustrates the relevance of words as constructions, in the sense proposed by Dąbrowska (2009), with a view to accounting for the dynamic syntactico-semantic building up of constructions with secondary predicates after saber ‘know’ in present-day Spanish.

2. Constructionist approaches in the 21st century Ever since the publication of the two seminal works mentioned above, there has been a proliferation of different formulations of CxG over the last twenty years. In line with Goldberg (2013), we shall assume that the different formulations of CxG can be appropriately referred to under the label of “constructionist approaches”. As noted by Goldberg (2003a: 223), a common assumption adhered to by all these versions of a constructionist approach is that the user’s knowledge of language consists of a vast network of interrelated constructions or a constructicon. A necessarily non-exhaustive listing of the main exponents of constructionist approaches is given in what follows: (i) The version of Construction Grammar proposed by Charles Fillmore and Paul Kay, originally without a formalist orientation (Fillmore, Kay and O’Connor 1988[2003]), has gradually converged with a formal, generative type of Linguistics in general and unification-based grammars in particular, under the rubric of Berkeley Construction Grammar (Fillmore and Kay 1995). Very recently, the cross-fertilization of Berkeley Construction Grammar and HeadDriven Phrase Structure Grammar (Pollard and Sag 1994) has given rise to the emergence of Sign-Based Construction Grammar (Boas and Sag 2012, Sag 2012, Michaelis 2009, 2013). (ii) A second important formulation of CxG is the one proposed by George Lakoff in 1987, whose most direct offspring is Embodied Construction Grammar, developed by Benjamin Bergen and Nancy Chang (Bergen and Chang 2005), a model which has a number of important points of convergence, from a theoretical and methodological standpoint, with Fluid Construction Grammar (Steels and De Beule 2006). (iii) The formulation of CxG proposed by Adele E. Goldberg, deeply rooted in the work of George Lakoff and Charles Fillmore (Goldberg 1995, 2003a, 2003b), has gradually taken on a decidedly cognitivist orientation, to the extent that in her latest monograph (Goldberg 2006), the author herself labels her strand of CxG “Cognitive Construction Grammar” (see also Goldberg 2009a, b, Boas 2013). This formulation of CxG is the main point of reference for the so-called collostructional approach developed by Stefan Th. Gries and Anatol Stefanowitsch, which aims to endow Cognitive Linguistics and CxG(s) with more stringent empirical validity, thus invoking quantitative corpus linguistics as well as a number of highly sophisticated and rigorous statistical techniques (Stefanowitsch and Gries 2003, Stefanowitsch 2013). The different applications

Words as Constructions

167

of CxG, in particular the Goldbergian formulation, to contrastive grammar have been one of the sustaining threads of Boas’ work and his usage-based lexical-constructional approach (Boas 2003), recently rechristened as a framesemantic constructional approach, due to the pivotal role that FrameNet plays on this model (Boas 2006, 2008).1

R. W. Langacker’s Cognitive Grammar (henceforth CG) (1987, 1991, 1999, 2003a, 2003b, 2005, 2009) qualifies as an exponent of a constructionist approach. Its theoretical and methodological assumptions in general and its full-blown defense of a usage-based approach in particular have played a crucial role in the development of CxGs within the present-day functional-cognitive scenario. Closely connected with CG is W. Croft’s Radical Construction Grammar (2001, 2003). Unlike the other formulations of CxG mentioned so far, Radical Construction Grammar has a prominent typological orientation. Last but not least, reference must be made to the work of Paul Hopper, Joan Bybee, Sandra Thompson and Barbara Fox, inter alia, which has been central to the evolution of Emergent Grammar (Hopper 1987, 1988) into Interactional Linguistics (Bybee and Hopper 2001, Hopper 2001, Thompson 2002, Ford, Fox and Thompson 2003, Bybee 2003, Bybee and McClelland 2005, Bybee and Eddington 2006). However, it should be emphasized that some of the scholars affiliated with this program, such as Barbara Fox and Sandra Thompson, have recently distanced themselves from the constructionist premises in invoking fragments instead of constructions as the basic units of analysis (Thompson 2002, Fox and Thompson 2007). The concept of fragment is thus invoked to refer to prefabricated units or pieces of naturally-occurring texts which are stored in the minds of language users and pasted in an improvised way in the interaction with other interlocutors in accordance with the communicative demands of a specific conversational context. By way of illustration, consider (1): (1) I think [it’s cool] (Example taken from Thompson 2002: 132)

Specifically, Thompson challenges the widely spread assumption that the bracketed sequence is a subordinate complement clause functioning as an argument (i.e. the direct object) of think. Moreover, she claims that what has been described as a main clause (i.e. I think) should be best understood as an epistemic/evidential/evaluative formulaic fragment 1

The original label “usage-based, lexical-constructional approach” is nonetheless operational to refer to the works of scholars such as Iwata (2008).

168

Chapter Five

conveying a stance on the part of the speaker toward the content of the alleged subordinate clause. This analysis, in which the matrix verbs and their subjects are stored and retrieved as formulaic stance markers, fits in better with the grammatical, pragmatic, prosodic and phonological data than a complementation analysis. In view of the above-mentioned repertoire of formulations of CxGs, two important observations need to be made: these formulations should be considered variants of the same constructionist theme (Goldberg 2006) and should thus be viewed as complementary. As Goldberg and Suttle (2010: 7) explicitly note, these formulations can be taken to be a family of constructionist approaches in which the shared assumptions prevail over any differences regarding the formalization system used, the type of semantics invoked or the focus of the research program. A second important observation concerns the observation that, regardless of the fact that the Goldbergian formulation of CxG enjoys a higher degree of popularity among linguists and psychologists, there is no mainstream version of CxG. As pointed out by Goldberg (2003a: 223), the central assumption shared by all constructionist approaches is that the knowledge that language users have can be best characterized as a massive web of constructions, known in the CxG literature as the constructicon. In other words, constructionist grammarians assume that a language is constructions all the way down. In addition, constructionist approaches have a number of further points of convergence, to which we turn in the next section.

3. The foundational assumptions of constructionist approaches Constructionist approaches can be characterized as implicitly or explicitly endorsing the following tenets: Constructions, far from being “mere taxonomic artifacts”, are regarded as free-standing theoretical entities in relative independence of the lexical units that fill them in, resulting in specific constructs (Goldberg 1995: 3–4, 2006). In addition, it is claimed that constructions have real psychological plausibility for language users and foreign language learners (Bencini and Goldberg 2000, Martínez Vázquez 2004, Valenzuela and Rojo 2008, inter alia). Therefore, it is argued that all types of lexico-grammatical units, from the morpheme to more complex units such as the conditionalcomparative construction, are taken to be constructions, that is, conventional associations between a given form and its corresponding semantics and/or discoursal function, as shown in Table 5-1:

Words as Constructions Construction Morpheme Word Complex word Idiom (filled) Idiom (partially filled) CovariationalConditional construction Ditransitive construction

169

Form/Example e.g. anti-, pre-, -ing e.g. Avocado, anaconda, and e.g. Daredevil, shoo-in e.g. Going great guns e.g. Jog someone’s memory The Xer the Yer (e.g. The more you think about it, the less you understand) Form: Subj [V Obj1 Obj2] (e.g. He gave her a Coke; He baked her a muffin) (double-object)

Table 5-1. Constructions at varying levels of complexity and abstraction (adapted from Goldberg 2003a: 220) As can be seen in Table 5-1, constructions exhibit varying degrees of morphosyntactic complexity, ranging from simpler units such as the morpheme to complex constructions such as the conditional-comparative construction. In addition, constructions can feature varying degrees of generality or, conversely, of specificity. Thus, for instance, the passive construction and the resultative construction display a considerable degree of abstraction, which contrasts with the fairly specific nature of morphemes. Finally, constructions can also display different degrees of fixation. These range all the way from fully or partially filled in idioms (which allow little or no variation at all) to constructions with a high degree of abstraction, such as the passive or the resultative (which admit a considerable degree of variation). The foundational assumption underlying the inventory of constructions exemplified in Table 5-1 is that grammar and the lexicon, far from being two separate or independent modules (as claimed in Chomskyan linguistics), form a continuum with a soft dividing line (Croft and Cruse 2004: 255). Constructionist approaches assign priority to the semantic and pragmatic facets of a construct over its formal properties in general and its syntactic aspects in particular. In addition, the relationship between semantics and pragmatics is thought of as a continuum, in which there is room for a considerable number of factors ranging from lexical semantics to discourse pragmatics, including textual and/or register considerations. Constructionist approaches, in much the same vein as functionalist models, argue that the basic function of language–and of the constructions of which it is made up–is to construct and convey meaning, thus serving to attain specific communicative goals. In the words of George Lakoff: “The

170

Chapter Five

primary function of language is to convey meaning. A grammar should therefore show as directly as possible how parameters of form are linked to parameters of meaning” (Lakoff 1987: 583). However, this should not be taken to imply that constructionist grammarians claim that all the constructions in a language can be explained (or motivated) in terms of semantico-pragmatic factors and extralinguistic ones (e.g. iconicity, language processing considerations, etc.). In other words, while it is assumed that language is by and large motivated by linguistic as well as extralinguistic factors, it is also acknowledged that there exist pockets of idiosyncrasy which must unavoidably be learned (see Tomasello 1998: xii). Constructionist approaches aim at achieving a satisfactory level of explanatory adequacy. Specifically, the family of construction grammars can be regarded as generative, in the sense that it tries to account for the infinite number of grammatically acceptable expressions in a given language, while also ruling out the ungrammatical or unacceptable ones (Goldberg 1995: 7). Furthermore, (un-)grammaticality and/or (un-) acceptability are regarded in relative rather than absolute terms. In addition, contextual factors are considered to be of pivotal importance in determining the degree of acceptability of a given construct. From a methodological viewpoint, acceptability judgments are empirically based not only on introspective judgments by native speakers but also on a representative sampling of naturally-occurring data retrieved from corpora as well as from psycholinguistic experiments. Unlike Generative Grammar, constructionist approaches posit no transformations or derivations by means of insertion or deletion rules. In other words, the specific semantico-pragmatic properties of a given construction are directly mapped onto its surface form. There is no level of representation beyond surface structure. Specifically, the meaning of a given construct is taken to be the result of the combination of the meaning of the construction with that of the specific lexical items filling in that construction. In addition, constructionist approaches dispense with any syntactic elements without any overt formal realization or abstract categories without any meaning and/or function. In this connection, a case in point is the PRO category, which in Generative Grammar refers to a null pronoun functioning as the subject of an infinitival clause, as in: David tried [PRO to dance] or [PRO To dance with David] was fun. Constructionist approaches are centrally concerned with accounting for all the linguistic phenomena systematically attested in real data, and not just the core grammar alone. Although the original impetus of constructionist approaches was on the study of idiomatic constructions

Words as Constructions

171

(e.g. the let alone-constructions investigated in Fillmore, Kay and O’Connor 1988[2003]), it should be noted that constructionist grammarians firmly believe that it is only by looking at the properties of idiosyncratic constructions that we can really understand the most general, regular and/or productive aspects of language (see further Gries 2013). Finally, specific generalizations about a given language are formalized via inheritance networks, which capture the properties shared by lowerlevel configurations and their higher-level constructions, as well as those properties which are specific to the lower-level configurations (or overrides). Moreover, the vertical and horizontal relations among constructions in the constructicon are represented through inheritance hierarchies (see Kay and Fillmore 1999). Thus far we have been concerned with the central features of constructionist approaches. In what follows, we will examine a number of areas in which subtle differences can be observed among the different strands of constructionist approaches. Some formulations of CxG, such as Goldberg’s Cognitive Construction Grammar, invoke a type of semantics based on non-classical categorization, thus revolving around prototypes and extensions from the prototypes. This type of semantics, dramatically contrasts with the more formal type of semantics employed by Fillmore, Kay and colleagues in Berkeley CxG (Fillmore and Kay 1995) and more recently in Sign-Based Construction Grammar (Sag 2012), known as Minimal Recursion Semantics. In line with monotonic inheritance systems, Minimal Recursion Semantics requires that no conflict exists between the attributes and the values in the matrix. As a consequence, this type of semantics cannot, at least in principle, account for the inherent flexibility and malleability of language use manifested in, for example, coercion effects (see further section 5.2). In addition, some formulations of CxG accept that constructions (e.g. the ditransitive), as well as words (e.g. bank), exhibit polysemy effects (Langacker 2000: 35). The acceptance of constructional polysemy by some practitioners of CxG is grounded on two important premises: first, the fact that constructions, in much the same vein as words, are taken to be form and meaning/function correspondences; second, the assumption that grammar and the lexicon form a continuum, albeit with a soft dividing line (Goldberg 2006, Boas 2008). Let us dwell on the concept of constructional polysemy with special focus on English ditransitive construction. The same form–a word or a construction–may be associated with different though nonetheless semantically related senses. In the words of Goldberg (1995: 31), “[c]onstructions are typically associated with a family of closely related

172

Chapter Five

senses rather than a single, fixed abstract sense” (emphasis added to the original). Thus, in the case of the English ditransitive construction, two related meanings can be discerned: Actual transfer of possession and intended transfer of possession, as exemplified in (2)-(3), respectively. (2) Peter tossed Mary a drink (actual transfer: Mary got the drink) (3) Mary baked Peter a cake (intended transfer: Mary intended Peter to receive the cake)

However, Goldberg proposes that the English ditransitive construction involves six related constructional meanings, five of which are extensions of the first, central sense (Goldberg 1995: 38). As shown in (4), the distinct senses of the construction are associated with specific verb classes. (4)

A. Central Sense: agent successfully causes recipient to receive patient 1. Verbs that inherently signify acts of giving: give, pass, hand, serve, feed,… 2. Verbs of instantaneous causation of ballistic motion: throw, toss, slap, kick, poke, fling, shoot,… 3. Verbs of continuous causation in a deictically specified direction: bring, take,… B. Conditions of satisfaction imply that agent causes recipient to receive patient 1. Verbs of giving with associated satisfaction conditions: guarantee, promise, owe,… C. Agent causes recipient not to receive patient 1. Verbs of refusal: refuse, deny D. Agent acts to cause recipient to receive patient at some future point in time 1. Verbs of future transfer: leave, bequeath, allocate, reserve, grant,… E. Agent enables recipient to receive patient 1. Verbs of permission: permit, allow F. Agent intends to cause recipient to receive patient 1. Verbs involved in scenes of creation: bake, make, build, cook, sew, knit,… 2. Verbs of obtaining: get, grab, win, earn,…

The constructional view reflects the general idea that a set of constructions does not consist of independent entities that exhibit irregular organizational patterns, but is instead a “highly structured lattice of

Words as Constructions

173

interrelated information” which is assumed to “display prototype structures and form networks of associations” (Goldberg 1995: 5). This uncompromising focus on robust generalizations among constructions is one of the major recurring themes in Goldberg’s work to date. Some constructionist approaches, such as Embodied Construction Grammar and in particular Cognitive Construction Grammar, emphasize the psychological reality of constructions. Thus, in the light of results from experimental work as well as of investigations into natural language processing, aphasia or language acquisition issues, it is submitted that constructions have psychological plausibility as constructs in the minds of language users (see further Bencini and Goldberg 2000, Goldberg, Casenhiser and Sethuraman 2005, inter alia). More recent research carried out by Goldberg and her team is geared towards providing neurolinguistic evidence in favor of constructions as free-standing linguistic entities (Goldberg 2011). In other constructionist approaches, such as Berkeley CxG and Sign-Based Construction Grammar, psycholinguistic (or even neurolinguistic) experimentation plays a less prominent role. Learning and experience are deemed to be fundamental in explaining language acquisition, with constructions (whether words or phrases) being learned inductively on the basis of positive input as well as general cognitive mechanisms (see further Tomasello 2003, Diessel 2013). Constructionist approaches generally do not invoke, or at least are quite cautious about positing, the existence of an innate, formal UG that is specific to language. In other words, certain biological prerequisites or cognitive mechanisms are understood to be necessary for language acquisition, but no claim is made that these involve any innate linguistic knowledge. In addition, even if some grammatical categories (e.g. nouns and verbs) are recognized to be universal, it does not necessarily follow that these have to be taken to be innate. Constructionist theories argue that language must be learnable from positive input together with fairly general cognitive abilities, since the diversity and complexity witnessed does not yield to accounts that assume that cross-linguistic variation can be characterized in terms of a finite set of parameters, as claimed in Chomskyan generative linguistics. Finally, most constructionist approaches make the prediction that a high proportion of cross-linguistic generalizations must be languagespecific and possibly also construction-specific (see especially Croft 2001), such generalizations being handled as tendencies which are amenable to explanation in terms of general cognitive constraints in conjunction with the functions of the constructions concerned.

Chapter Five

174

To round off this section, an important observation is in order. While the foundational assumptions of constructionist approaches differ dramatically from those of the Chomskyan program, these differences should not however be overemphasized. Thus, in this connection, it must be borne in mind, as Goldberg (2003a: 223) notes, that constructionist approaches share with mainstream Generative Linguistics the following three premises: (i) the assumption of language as cognitive (mental) system; (ii) the premise that structures may be combined to create novel utterances (e.g. She mopped him something, recognized as an instance of the ditransitive construction by most native English speakers); and (iii) the recognition of the fact that a non-trivial theory of language learning is needed.

4. What exactly is a construction? As will be recalled from the inventory of constructions displayed in Table 5-1, the scope of application of the concept of construction encompasses varying degrees of morphosyntactic complexity, generality (or, by contrast, specificity), as well as transparency (or, by contrast, idiosyncrasy). This follows from the constructionist assumption that grammar and the lexicon, far from being two separate modules, form a continuum. In this section, we will examine in some detail the notion of construction, as currently employed in constructionist approaches. Consider, by way of illustration, the following construct: (5) What did John get Mary?

An actual expression or construct like the one in (5) typically involves the combination of at least half a dozen different constructions. Constructions are combined freely to form actual expressions as long as they are not in conflict. The observation that language has an infinitely creative potential is accounted for, then, by the free combination of constructions. Thus, it can be said that (5) is the result of the superimposition of the constructions listed in (6): (6) a. John, Mary, did, get, what constructions (i.e. words) b. Ditransitive construction c. Question construction d. Subject-Auxiliary Inversion construction e. VP construction f. NP construction

Words as Constructions

175

In the construction grammar literature, at least two different notions of constructions can be discerned.2 The original definition of the term construction runs as follows: C is a CONSTRUCTION iff def n is a form-function pair, such that some aspect of the form or some aspect of the function is not strictly predictable from C’s component parts (Goldberg 1998: 25, emphasis added to the original).

This definition of construction revolves around the criterion of idiosyncrasy or non-predictability as the sine qua non condition for construction status. Thus, consider (7): (7) Pat faxed Bill the letter (Example taken from Goldberg 1998: 206)

A construct like the one in (7) implies that Bill actually received the letter by fax. However, according to Goldberg, the semantic specification of successful transfer cannot be attributed to the lexical meaning of the construct’s components. In other words, the overall interpretation of successful transfer comes from the ditransitive construction, the meaning of which is X CAUSES Y TO RECEIVE Z, rather than from the meaning of the lexical items involved. This is why (7) qualifies as an instance of the ditransitive construction according to the original definition of the term, because its overall interpretation is non-compositional, that is, not strictly predictable from its component parts or from other constructions. However, the original idiosyncrasy (i.e. non-predictability) requirement in the original formulation of the constructions has been recently downgraded, in concert with other cognitively-oriented usage based models such as CG and RCG (see further Goldberg 2006: 224 for discussion). On this view, the extended, usage-based definition of a construction can be established in the following terms: Any linguistic pattern is recognized as a construction as long as some aspect of its form or function is not strictly predictable from its component parts or from other constructions recognized to exist. In addition, patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency (Goldberg 2006: 5, see also Langacker 2005: 139-143 for a similar view) 2

For further details on the treatment of constructions in the present-day functionalcognitive scenario in general and constructionist approaches in particular, the interested reader is referred to Schönefeld (2006), Gonzálvez-García and Butler (2006), as well as Mairal Usón and Gonzálvez-García (2010), inter alia.

Chapter Five

176

An important corollary deriving from this usage-based definition is that stored (typically highly frequent) regularities between form and meaning are considered constructions even if they are fully compositional (emphasis in original) (see Goldberg 2006: 214-215). Thus, consider (8): (8) I love you

The construct in (8) illustrates the case of a fully transparent, compositional instance of the transitive construction. However, given the fact that it is highly frequent and highly entrenched, it qualifies as a construction according to the expanded, usage-based definition of a construction.3 The question then arises as to how to justify this expansion of the definition of the notion of construction. In this connection, it is important to bear in mind Langacker’s observation that Lower-level schemas, expressing regularities of only limited scope, may on balance be more essential to language structure than high-level schemas representing the broadest generalizations (Langacker 2000: 3, emphasis added to the original)

Another compelling argument in favor of this expansion of the notion of construction stems from the following observation made by Goldberg: Facts about the actual usage of linguistic expressions such as frequencies and individual patterns are recorded alongside more traditional linguistic generalizations (Goldberg 2006: 45)

As things stand, this expanded, usage-based definition of a construction seems particularly well-suited to capture both broad generalizations and more limited or item-specific patterns (see further Bybee 2010, 2013).

5. Construction Grammar(s) and Morphology (or the Morphology-Semantics-Pragmatics-Syntax continuum) In this section we explore some of the useful implementations of (Cognitive) Construction Grammar with a view to describing and explaining morphological phenomena, such as (Spanish) compounds and 3

It should be emphasized that Sign-Based Construction Grammar, as formulated in Boas and Sag (2012), does not endorse this usage-based expansion of the notion of construction. Instead, these construction grammarians rely on the original definition of the term, based on the idiosyncrasy requirement (see Kay 2013).

Words as Constructions

177

coercion. Before going into the discussion and analysis of these phenomena, it needs to be emphasized once again that, in keeping with the grammar-lexicon continuum (see Figure 5-1), Morphology is seen as interacting with the other linguistic levels of description (i.e. Semantics, Pragmatics and Syntax) rather than as an independent module.

5.1. Compounds In her analysis of Spanish compounds based on (Cognitive) Construction Grammar, Yoon (2011) posits at least four types of metonymic and/or metaphoric patterns to describe and explain [V + N] compounds in Spanish of the type exemplified in (9)-(12): (9)

a. Agent as a referent: limpiabotas ‘cleans-shoes (shoeshine boy)’, cuidaniños ‘watches-kids (babysitter)’ b. Instrument as a referent: matamoscas ‘kills-flies (fly swatter)’, calientapiés ‘warms-feet (foot warmer)’, lavaplatos ‘washes-dishes (dishwasher)’ c. Place as a referent: guardarropa ‘keeps-clothes (wardrobe, cloakroom)’ (10) Metonymy derived from a metaphor a. Agent as a referent: guardaespaldas ‘guards-backs (bodyguard)’, sacamuelas ‘pulls out-molars (quack, dentist)’ b. Instrument as a referent: matasellos ‘kills-stamps (postmark)’ (11) Metaphor derived from a metonymy a. Plant as a referent: girasol ‘turns around-sun (sunflower)’ b. Animal/insect/bird as a referent: picaflor ‘pecks-flower (hummingbird)’, saltamontes ‘jumps-hills (grasshopper)’ (12) Metaphor involving two metonymies: Causative event: cumpleaños ‘birthday’, pasatiempo ‘pastime, hobby’

As argued in Yoon (2009), the Spanish compounds exemplified in (9) (e.g. matamoscas) exhibit a considerable degree of transparency (or compositionality) from a semantic viewpoint. In cases like these, the original meaning of the predication pattern remains unaltered (‘X mata moscas para alguien’/‘X kills flies for someone’). Moreover, as also noted in Yoon (2009), the meanings of agent person, event, instrument, time, and place are connected with an idealized cognitive model (Lakoff 1987), that is, the prototypical transitive scenario, which plays a crucial role in sanctioning compounds of the [V + N] type, among which those with a metonymic basis are particularly important.

178

Chapter Five

5.2. Coercion The term coercion, which can be traced back to the works of Moens and Steedman (1988) and Pustejovsky (1993), is understood by construction grammarians as the resolution of a conflict between the meaning of the construction and the meaning of its component parts (Michaelis 2003: 264). Specifically, in light of the Override Principle, the relevance of coercion to a constructionist approach can be stated as follows: The Override Principle: “If a lexical item is semantically incompatible with its syntactic context, the meaning of the lexical item conforms to the meaning of the structure in which it is embedded” (Michaelis 2003: 268).

Thus, in accordance with the Override Principle, it is assumed that constructional meaning always wins over the lexical meaning of the construction’s components. As things stand, coercion is taken to be a compelling piece of evidence in favor of the superiority of constructions as overall determinants of sentence meaning. Thus, coercion shows in the last instance that there is soft dividing line between grammar and the lexicon.4 Let us now illustrate how the characterization of coercion provided by Michaelis works for a construct like the one in (13): (13) La Voz de Almería: Más Almería (La Voz de Almería, newspaper) ‘La Voz de Almería: More Almería’

Following Michaelis (2003), it could be argued that in (13) there is a conflict between the meaning of the quantifier más ‘more’, on the one hand, and the proper noun Almería, on the other. Specifically, the quantifier in question requires that the noun with which it combines should denote something which can be quantified. Proper nouns such as Almería are in principle incompatible with the meaning of quantity. However, in (13), the prima facie conflict in the nominal expression más Almería is resolved through metonymy. Thus, by virtue of the PLACE FOR OBJECT

4

In this connection, it should be noted that Ziegeler (2007, 2010) argues that the notion of coercion, as employed in constructionist circles, is superfluous, as it can be explained on the basis of metaphor and/or metonymy alone. The interested reader is referred to Gonzálvez-García (2011) for a different view that shows that coercion is compatible with a metaphor/metonymy account, as advocated in the Lexical Constructional Model (Ruiz de Mendoza and Mairal Usón 2008).

Words as Constructions

179

metonymy, the proper noun Almería is then interpreted as “more news of Almería”. Coercion is operational alongside the grammar-lexicon continuum, especially in argument structure constructions. Consider, by way of illustration, the following instances of the subjective-transitive construction in Spanish (see further Gonzálvez-García 2009, 2011): (14) a. Ramón quizá se piensa líder de ese otro fascismo musoliniano y español ‘Ramón perhaps thinks himself the leader of that other Mussolinean and Spanish fascism’ (CREA, 1995, Francisco Umbral, Leyenda del César Visionario, Novela) b. *La opinión pública piensa a Ramón líder de ese otro fascismo musoliniano y español ‘Public opinion thinks Ramon leader of that other Mussolinean and Spanish fascism’ (15) a. Señor, en este domingo de Adviento nos reconocemos pobres, miserables y pecadores ante ti y ante los hermanos ‘Our Lord, on this Sunday of Advent, we acknowledge ourselves poor, shameful and sinners before you and the brothers’ (CREA, 1998, Efímero, 98103027) b. *Señor, en este domingo de Adviento reconocemos a nuestros hermanos pobres, miserables y pecadores ante ti ‘Our Lord, on this Sunday of Advent, we acknowledge our brothers poor, shameful and sinners before you’ (16) a. Te sabes mía ‘You know yourself to be mine’ (Café Quijano, Dame de esa boca) b. #Sé a tu hermana mía ‘I know your sister to be mine’

The examples in (14)-(16) are instances of what Gonzálvez-García (2007, 2009, 2011) calls the reflexive subjective-transitive construction, on the grounds that the presence of a reflexive pronoun instead of a full lexical NP as direct object of the matrix verb saves an otherwise ungrammatical result (cf. (14b), (15b)) or a marginally acceptable outcome (cf. (16b)). Interestingly enough, those verbs amenable to coercion via a reflexive pronoun in the secondary predication frame end up being drawn into the orbit of evaluative cognitive predicates such as considerar ‘consider’. The reflexive subjective-transitive construction can be characterized as follows:

180

Chapter Five The reflexive subjective-transitive construction: The subject/speaker (NP1) expresses a (non-cancellable) direct, personal, judgmental stance involving a high degree of commitment towards the state of affairs/action encoded about himself/herself (NP2)

However, it should be stressed that the subjective coloring of this construction can be modulated by the lexical semantics of the verb with which the construction fuses. Thus, those verbs which imply a more objective coloring, such as saber ‘know’, impose on the reflexive subjective-transitive construction an additional nuance of introspection by the subject/speaker, as shown in (16)(a). 5 However, upon closer inspection, it turns out that saber, unlike other cognition verbs (e.g. pensar ‘think’), is also felicitous in the secondary predication frame with a full lexical NP in the object slot. These configurations are the object of inquiry of the next section in this chapter.6

5.3. Words as constructions: D¦browska (2009) In the remainder of this chapter, we shall examine the viability of taking words as constructions in their own, drawing on the proposal made in D¦browska (2009) for English. According to this author, English native speakers make a number of robust generalizations regarding collocations with a high degree of frequency, such as: trudge wearily, trudge slowly, trudge through the snow (or even trudge through + an expression encoding a path with a considerable degree of density such as snow, mud or thick vegetation), trudge up the stairs (or, in a more schematic form, trudge UPWARDS, which covers lower-level, more specific variants with up the stairs, upstairs, up the steps, up the hill, or to the top). Further compelling evidence for frequency as a determining factor stems from the possibility that English native speakers do not store much less frequent combinations such as trudge across miles of fitted carpets and sanded floors (D¦browska 2009: 207).7 5

For a discussion of the relevance of coercion for the phenomenon of polysemy involving cases of sensory/physical perception as well as cognitive perception in Spanish, the reader is referred to Horno Chéliz (2004) and references therein. 6 For a fine-grained analysis of coercion within constructionist approaches in general and its impact on the family of the subjective-transitive constructions in particular, see Gonzálvez-García (2011). 7 See Dąbrowska (2009: 207) for a more detailed analysis of the formalization of the collocation trudge through the snow from a cognitive, usage-based perspective, compatible with Cognitive Construction Grammar (Goldberg 2006) or Radical Construction Grammar (Croft 2001).

Words as Constructions

181

In what follows we will examine the validity of the conception of words as constructions in the sense of D¦browska (2009) in the secondary predication frame in Spanish. In this regard, consider the examples reproduced in (17a-b) and (18a-c): (17) a. Le entristecía muchísimo saber a su hermana en la cárcel/trabajando de día y noche/#(una) psicópata/ convertida en una psicópata por falta de afecto ‘It made him/her quite sad to know that her sister was in jail/working day and night/ that her sister was a psychopath/that her sister had turned into a psychopath due to a lack of affection’ b. #Le entristecía muchísimo saber la carta extraviada por Correos/#el vino de mala calidad ‘It made him/her quite sad to know that the letter had got lost in the mail/ that the wine was not a good one’ (18) a. *Ella sabe a su hijo la persona más encantadora del mundo ‘She knows her son *(to be) the most charming person in the world’ b. Le entristecía saber a su hijo en la cárcel (sensory perception) ‘It made her unhappy to know that his son was in jail’ c. Se sabe la persona más encantadora del mundo (cognitive perception) ‘She knows herself *(to be) the most charming person in the world’

In view of the acceptability differences illustrated in (17)-(18), a number of important generalizations can be made. First, the verb saber ‘know’ is sensitive to the human or non-human reference of the lexical NP functioning as direct object of the main verb. Only in the former case can this verb felicitously occur in the secondary predication frame, as shown in the acceptability differences in (17a-b). Moreover, this verb is also sensitive to the sensory or cognitive/evaluative flavor of the state of affairs/event encoded in the XPCOMP. Thus, the verb saber ‘know’ may zoom in on a visual perception interpretation if the object slot is filled in by a lexical NP and the XPCOMP encodes a state of affairs likely to be directly perceived by the subject/speaker, as in (18b) above. This explains why states of affairs with an overwhelmingly evaluative interpretation yield an ungrammatical result (cf. (18a)) after saber, as they clash with the direct perception construal imposed by the secondary predication with this verb. However, if the object slot is realized by a reflexive pronoun, the Spanish verb saber ‘know’ brings into focus a cognitive perception (i.e. self-evaluation) interpretation, roughly equivalent to that of consider-type verbs. In addition, the choice of the reflexive pronoun in (18c) also contributes a further meaning nuance intrinsic to this lower-level reflexive

182

Chapter Five

configuration, namely, that this judgment value is the result of introspection by the subject/speaker (see further Gonzálvez-García 2007). To summarize, Dąbrowska (2009) contends that it is necessary to provide a fine-grained picture of the connections between (i) argument structure, (ii) the lexical representation of a given verb, and (iii) a given collocation between the verb and any other element(s) in the construction. In the light of the cursory analysis of the behavior of saber ‘know’ in the secondary predication frame in Spanish presented here, it can be claimed that Dąbrowska’s conception of constructions as words seems particularly well-suited to accommodate the polysemous behavior of saber ‘know’ between a sensory perception interpretation and a cognitive (evaluative) one. Specifically, this verb is sensitive to a number of factors: (i) the reflexive or non-reflexive nature of the NP in the object slot, (ii) the human or non-human nature of the NP in the object slot, and (iii) whether the XPCOMP denotes a perceivable or evaluative state of affairs. With these observations in mind, the following low-level constructions (or verbspecific constructions in the sense of Croft (2003)) can be posited: (19) a. ‘saber + reflexive + XPCOMP ’ Se sabe español ‘He knows himself to be Spanish’ b. ‘saber + NP + XPCOMP ’ Le entristecía saber a su hijo en la cárcel ‘It made him/her sad to see her son in jail’

6. Closing remarks In the previous pages, we hope to have been able to demonstrate that Morphology in general and the word in particular play a crucial role in the grammar-lexicon continuum, as conceived by construction grammarians. Thus, constructionist approaches enable us to provide a principled descriptive and explanatory account of morphological facts as well of their connections with Semantics, Pragmatics and Syntax. At a higher level of delicacy, the cases of compounding and coercion examined in this chapter provide compelling evidence for the close connections that exist among the components of a given construction, commonly referred to in the literature under the label of collocations. Last but not least, we hope to have succeeded in showing that cognitively-influenced, usage-based constructionist approaches furnish satisfactory ways of analyzing grammatical phenomena of the kind investigated in this chapter in their inherent linguistic complexity. In addition, they also provide adequate formalization systems to capture the analogies and differences among

Words as Constructions

183

constructions at varying levels of specificity (or, conversely, of schematicity) through inheritance hierarchies, which pervade the entire constructicon.

References Bencini, G. and A. E. Goldberg 2000, “The contribution of argument structure constructions to sentence meaning”, Journal of Memory and Language 43, 640-651 Bergen, B. and N. Chang 2005, “Embodied Construction Grammar and simulation-based language understanding”, J-O. Östman and M. Fried (eds.) Construction Grammars: Cognitive Grounding and Theoretical Extensions (Constructional Approaches to Language, 3) Amsterdam and Philadelphia: John Benjamins, 147-190 Boas, H. C. 2003, A Constructional Approach to Resultatives, Stanford: CSLI Publications Boas, H. C. 2006, “A frame-semantic approach to identifying syntactically relevant elements of meaning”, P. Steiner, H. C. Boas and S. Schierholz (eds.) Contrastive Studies and Valency. Studies in Honor of Hans Ulrich Boas, Berlin and New York: Peter Lang, 119-149 —. 2008, “Determining the structure of lexical entries and grammatical constructions in Construction Grammar”, Annual Review of Cognitive Linguistics 6, 113-144 —. 2013, “Cognitive Construction Grammar”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 233-251 Boas, H. C. and I. A. Sag (eds.) 2012, Sign-Based Construction Grammar, Stanford: CSLI Publications Booij, G. 2010, Construction Morphology, Oxford: Oxford University Press —. 2013, “Morphology in Construction Grammar”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 255-273 Borer, H. 2003, “Exo-skeletal vs. endo-skeletal explanations”, J. Moore and M. Polinsky (eds.) The Nature of Explanation in Linguistic Theory, Chicago: CSLI and University of Chicago Press, 31-67 Bybee, J. 2003, “Cognitive processes in grammaticalization”, M. Tomasello (ed.) The New Psychology of Language (Cognitive and Functional Approaches to Language Structure), Vol 2, Mahwah (NJ): Lawrence Erlbaum, 145-167

184

Chapter Five

—. 2010, Language, Usage and Cognition, Cambridge: Cambridge University Press —. 2013, “Exemplars and constructions”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 49-69 Bybee, J. and P. J. Hopper (eds.) 2001, Frequency and the Emergence of Linguistic Structure (Typological Studies in Language, 45), Amsterdam and Philadelphia: John Benjamins Bybee, J. and J. McClelland 2005, “Alternatives to the combinatorial paradigm of linguistic theory based on domain general principles of human cognition”, The Linguistic Review 22: 381-410 Bybee, J. and D. Eddington 2006, “A usage-based approach to Spanish verbs of ‘Becoming’”, Language 82(2), 323-355 Croft, W. 2001, Radical Construction Grammar: Syntactic Theory in Typological Perspective, Oxford: Oxford University Press —. 2003, “Lexical rules vs. constructions: A false dichotomy”, T. Berg, R. Dirven, G. Radden and K. U. Panther (eds.) Motivation in Language. Studies in Honour of Günter Radden, Amsterdam and Philadelphia: John Benjamins, 49-68 Croft, W. and A. Cruse 2004, Cognitive Linguistics, Cambridge: Cambridge University Press Culicover, P. and R. Jackendoff 2005, Simpler Syntax, Oxford: Oxford University Press Chomsky, N. 1995, The Minimalist Program, Cambridge (MA): MIT Press Dąbrowska, E. 2009, “Words as constructions”, V. Evans and S. Pourcel (eds.) New Directions in Cognitive Linguistics, Amsterdam and Philadelphia: John Benjamins, 201-223 Diessel, H. 2013, “Construction grammar and first language acquisition”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 347-364 Fillmore, C. J. and P. Kay 1995, Construction Grammar Coursebook, Unpublished manuscript, UC Berkeley Fillmore, C. J., P. Kay and M. C. O’Connor 1988 [2003], “Regularity and idiomaticity in grammatical constructions: The case of let alone”, M. Tomasello (ed.) The New Psychology of Language (Cognitive and Functional Approaches to Language Structure) Vol. 2, Mahwah (NJ): Lawrence Erlbaum, 243-270 Ford, C., B. A. Fox and S. A. Thompson 2003, “Social interaction and grammar”, M. Tomasello (ed.) The New Psychology of Language

Words as Constructions

185

(Cognitive and Functional Approaches to Language Structure) Vol 2, Mahwah (NJ): Lawrence Erlbaum, 119-143 Goldberg, A. E. 1995, Constructions: A Construction Grammar Approach to Argument Structure, Chicago: Chicago University Press —. 1998, “Patterns of experience in patterns of language”, M. Tomasello (ed.) The New Psychology of Language (Cognitive and Functional Approaches to Language Structure 1), Mahwah (NJ): Lawrence Erlbaum, 203-219 —. 2003a, “Constructions: A new theoretical approach to language”, Trends in Cognitive Sciences 7, 219-224 —. 2003b, “Surface generalizations: An alternative to alternations”, Cognitive Linguistics 13, 327-356 —. 2006, Constructions at Work: The Nature of Generalization in Language, New York: Oxford University Press. —. 2009a, “The nature of generalization in language”, Cognitive Linguistics 20 (1), 93-127 —. 2009b, “Constructions work”, Cognitive Linguistics 20, 201-224 Goldberg, A. E., D. M. Casenhiser and N. Sethuraman 2004, “Learning argument structure generalizations”, Cognitive Linguistics 15, 289-316 Goldberg, A. E. and L. Suttle 2010, “Construction Grammar”, Interdisciplinary Reviews; Cognitive Science 1, 1-10 Goldberg, A. E. 2011, “Argument structure constructions: Items and generalizations”, Keynote Presentation, 44th Annual Meeting of the Societas Linguistica Europaea, Logroño, 10 September 2011 Goldberg, A. E. 2013, “Constructionist approaches”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar. Oxford: Oxford University Press, 15-31 Gonzálvez-García, F. 2007, “‘Saved by the reflexive’: Evidence from coercion via reflexives in verbless complement clauses in English and Spanish”, Annual Review of Cognitive Linguistics 5, 193-238 —. 2009, “The family of object-related depictives in English and Spanish: Towards a constructionist, usage-based analysis”, Language Sciences 31, 663-723 —. 2011, “Metaphor and metonymy do not render coercion superfluous: Evidence from the subjective-transitive construction”, Linguistics 49, 1305-1358 Gonzálvez-García, F. and C. S. Butler 2006, “Mapping functionalcognitive space”, Annual Review of Cognitive Linguistics 4, 39-95 Gries, S. Th. 2013, “Data in Construction Grammar”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 93-108

186

Chapter Five

Hopper, P. J. 1987, “Emergent grammar”, Berkeley Linguistics Society 13, 139-157 —. 1988, “Emergent grammar and the a priori grammar postulate”, D. Tannen (ed.) Linguistics in Context: Connecting Observation and Understanding (Lectures from the 1985 LSA/TESOL and NEH Institutes), Ablex: Norwood, 117-134 —. 2001, “Grammatical constructions and their discourse origins: Prototype or family resemblance?”, M. Putz, S. Niemeier and R. Dirven (eds.) Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin and New York: Mouton de Gruyter, 109-129 Horno Chéliz, M. C. 2004, “Aspecto léxico y verbos de percepción. A propósito de ver y mirar”, R. M. Castañer and J. M. Enguita (eds.) Archivo de Filología Aragonesa LIX, 555-575 Iwata, S. 2008, Locative Alternation: A Lexical-Constructional Approach, Amsterdam and Philadelphia: John Benjamins Jackendoff, R. 2010, Meaning and the Lexicon: The Parallel Architecture 1975–2010, Oxford: Oxford University Press Kay, P. and C. J. Fillmore 1999, “The What’s X doing Y? construction”, Language 75, 1-33 Kay, P. 2013, “The limits of (construction) grammar”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 32-48 Lakoff, G. 1987, Women, Fire, and Dangerous Things: What Categories Reveal about the Mind, Chicago: University of Chicago Press Langacker, R. W. 1987, Foundations of Cognitive Grammar, Vol. 1: Theoretical prerequisites, Stanford (CA): Stanford University Press —. 1991, Foundations of Cognitive Grammar, Vol. 2: Descriptive Application, Stanford (CA): Stanford University Press —. 1999, Grammar and Conceptualization, Berlin and New York: Mouton de Gruyter —. 2000, “A dynamic usage-based model”, M. Barlow and S. Kemmer (eds.) Usage-Based Models of Language, Stanford: CSLI Publications, 1-63 —. 2003a, “Constructions in Cognitive Grammar”, English Linguistics 20, 41-83. —. 2003b, “Explanation in Cognitive Linguistics and Cognitive Grammar”, J. Moore and M. Polinsky (eds.) The Nature of Explanation in Linguistic Theory, Stanford: CSLI Publications, 239-261 —. 2005, “Construction Grammars: Cognitive, radical and less so”, F. J. Ruiz de Mendoza Ibáñez and S. Peña Cervel (eds.) Cognitive Linguistics: Internal Dynamics and Interdisciplinary interaction

Words as Constructions

187

(Cognitive Linguistics Research, 32), Berlin and New York: Mouton de Gruyter, 101-159 —. 2009, “Cognitive (Construction) Grammar”, Cognitive Linguistics 20, 167-176 Mairal Usón, R. and F. J. Ruiz de Mendoza Ibáñez 2009, “Levels of description and explanation in meaning construction”, C. S. Butler and J. Martín Arista (eds.) Deconstructing Constructions, Amsterdam and Philadelphia: John Benjamins, 153-198 Mairal Usón, R. and F. Gonzálvez-García 2010, “Verbos y construcciones en el espacio cognitivo-funcional del siglo XXI”, J. F. Val Álvaro and M. C. Horno Chéliz (eds.) La Gramática del Sentido: Léxico y Sintaxis en la Encrucijada, Zaragoza: PUZ, 123-152 Martínez Vázquez, M. 2004, “Learning Argument Structure Generalizations in a Foreign Language”, VIAL, Vigo International Journal of Applied Linguistics 1, 151-165 Michaelis, L. A. 2003, “Word meaning, sentence meaning, and syntactic meaning”, H. Cuyckens, R. Dirven, J. R. Taylor and R. W. Langacker (eds.) Cognitive approaches to lexical semantics, Berlin and New York: Mouton de Gruyter, 163-209 —. 2009, “Sign-Based Construction Grammar”, B. Heine and H. Narrog (eds.) The Oxford Handbook of Linguistic Analysis, Oxford: Oxford University Press, 155-176 —. 2013, “Sign-Based Construction Grammar”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 134-152 Moens, M. and M. Steedman 1988, “Temporal ontology and temporal reference”, Computational Linguistics 14, 15-28 Pollard, C. and I. A. Sag. 1994, Head-Driven Phrase Structure Grammar, Chicago: University of Chicago Press Pustejovsky, J. 1993, “Type coercion and lexical selection”, J. Pustejovsky (ed.) Semantics and the Lexicon, Dordrecht: Kluwer, 73-96 Ruiz de Mendoza, F. J. and R. Mairal Usón 2008, “Levels of description and constraining factors in meaning construction: An introduction to the Lexical Constructional Model”, Folia Linguistica 42, 355-400 Sag, I. A. 2012, “Sign-Based Construction Grammar: An informal sinopsis”, H. C. Boas and I. A. Sag (eds.) Sign-Based Construction Grammar, Stanford: CSLI Publications, 69-202 Schönefeld, D. 2006, “Constructions”, D. Schönefeld (ed.) Constructions SV1-1/2006, Special volume no. 1: Constructions All Over: Case Studies and Theoretical Implications, available at: http://www.constructions-online.de/articles/specvol1

188

Chapter Five

Steels, L. and J. De Beule 2006, “A (very) brief Introduction to Fluid Construction Grammar”, Proceedings of the 3rd Workshop on Scalable Natural Language Understanding, 73-80 Stefanowitsch, A. 2013, “Collostructional analysis”, T. Hoffmann and G. Trousdale (eds.) The Oxford Handbook of Construction Grammar, Oxford: Oxford University Press, 290-306 Stefanowitsch, A. and S. Th. Gries 2003, “Collostructions: Investigating the interaction of words and constructions”, International Journal of Corpus Linguistics 8(2), 209-243 Thompson, S. A. 2002, “‘Object complements’ and conversation: Towards a realistic account”, Studies in Language 26, 125-163 Tomasello, M. 1998, “Introduction: A cognitive-functional perspective on language structure”, M. Tomasello (ed.) The New Psychology of Language (Cognitive and Functional Approaches to Language Structure), Vol. 1, Mahwah (NJ): Lawrence Erlbaum, vii-xxiii —. 2003, Constructing a Language: A Usage-Based Theory of Language Acquisition, Harvard: Harvard University Press Valenzuela, J. and A. Rojo 2008, “What can language learners tell us about constructions?” S. De Knop and T. De Rycker (eds.) Cognitive Approaches to Pedagogical Grammar – Volume in Honour of René Dirven, Berlin and New York: Mouton de Gruyter, 197-229 Van Valin, R. D. Jr. 2005, Exploring the Syntax-Semantics Interface, Cambridge: Cambridge University Press Yoon, J. 2009, “Constructional meanings of verb-noun compounds in Spanish: limpiabotas vs. tientaparedes”, Language Sciences 31, 507530 —. 2011, “Productivity of Spanish verb-noun compounds: patterns of metonymy and metaphor”, Review of Cognitive Linguistics 9, 83-106 Ziegeler, D. 2007, “A word of caution on coercion”, Journal of Pragmatics 39, 990-1028 —. 2010, “Count-mass coercion, and the perspective of time and variation”, Constructions and Frames 2, 33-73

CHAPTER SIX WORD IN COGNITIVE GRAMMAR RICARDO MALDONADO AND BORIS FRIDMAN-MINTZ

1. Introduction It seems fair to say that the notion of word depends on the theoretical base from which it is defined. Thus we find totally contrasting definitions. An exhaustive revision of this notion drives Haspelmath (2011) to the conclusion that there is not one clean set of criteria that applies uniformly and consistently to all languages of the world and that, in fact, criteria are applied in a non-convergent manner. Haspelmath states that in the absence of a reliable notion of word there is no reliable borderline to separate morphology from syntax. Haspelmath is right in pointing out that the morphology/syntax distinction does nothing but privilege the methodological needs of some theoretical approaches, and certainly he is right in pointing out that the notion of word does not constitute a reliable criterion to separate syntax from semantics. However the need for an operational notion of word is unquestionable, as it is crucial not only for linguistic description but for written and oral expression. While a discrete core characterization of the notion of word cannot be provided, this notion will inevitably require specific adjustments responding to the structure of particular languages. We offer an operational definition in the lines of Cognitive Grammar: The minimal symbolic unit that may be represented as syllabically autonomous linked to a unitary meaning. Such symbolic unit is determined by semantic and phonemic dependencies among their components.

Crucial to this definition is the fact that the notion of word corresponds to a mental representation and not to a bundle of features providing discrete entities. It is a metalinguistic problem in which both speakers and

190

Chapter Six

linguists subsume it as an operational notion (Fridman 1984, Lara 2006). We seek to identify association and linking patterns with which units are formed as the basis to have a general definition of word that may be subject to further typological specifications. We propose that any word in any language is cognitively autonomous at the phonetic level in such a way that it may be identified as an isolatable unit with identifiable content. The ways in which these interdependencies are established follow some fundamental Cognitive Grammar notions that we will explore in this chapter. The chapter is organized as follows. In section 2 we introduce some fundamental notions of Cognitive Grammar as defined by Langacker (1985, 1987, 1991a, 1991b, 2000 and multi alia). Section 3 introduces the basic notions of Cognitive Phonetics (Fridman 2009, 2010). Section 4 introduces the concept of word from a Cognitive Grammar perspective and overviews a range of morphemic issues (separability, cohesion, independence, informativity, compositional path) pertaining to the notion of word. The last section offers the conclusions.

2. Cognitive Grammar Cognitive Grammar as introduced by Langacker (1987, 1991a, 1991b, 2000) is one of a variety of models in Cognitive Linguistics that accounts for the behavior of language in terms of processes and mental operations. To the extent that language is a usage based product Cognitive Grammar (CG) can be associated with functional linguistics models and distinguished from formal and generative approaches to language. Thus, instead of having autonomous models with separate levels of language and discrete units entailing operations that are independent from each other, cognitive approaches emphasize the semiotic basis of language which is grounded in cognition. Mental operations and communicative effects interact in the configuration of language. Thus, language is not seen as independent from other mental processes, and meaning is seen as fundamental to conform linguistic units instead of being a post-factum addendum activated after a formal apparatus has produced some output. Cognitive Grammar offers linguistic explanations that are biologically and psychologically plausible and that correspond to realistic language samples used in natural contexts of social interaction as manifested in discourse. The cognitive basis of this approach consists on finding the conceptual basis of language even in its most simple manifestations. Thus, language operates in accordance with basic cognitive systems such as perception, memory and categorization from which it is inseparable. The innate peculiarities of language are but adaptations of more basic cognitive

Word in Cognitive Grammar

191

phenomena and not independent entities with autonomous operations. Some manifestations of these mental-linguistic abilities are the capacity to create new conceptualizations, to employ one basic structure in order to categorize an alternative element, to establish correspondences among different structures, to conceive a situation in different degrees of abstraction, to combine basic structures in order to conform complex ones and to assign different degrees of prominence to the elements that constitute an event. Language is defined as a structured inventory of symbolic units. Such inventory is not a flat list but a set of constructions that establish links with other components as they share one or more properties. There are phonetic, semantic and symbolic units. A symbolic unit is the inseparable and dynamic association between phonetic and semantic units. This fact has important consequences. On the one hand, there are no empty categories, traces, phrase structure trees nor any other type of abstract notions determining the relations among elements without meaning. For every symbolic unit there must be a phonetic and a semantic representation. On the other hand, two linguistic expressions imply two different conceptualizations and not two ways of saying the “same thing”. The inclusion of the speaker’s conceptualization in the analysis involves considering both referential representations (scanning, image formation, figure/ground, etc.) as well as contextual issues (social, psychological, cultural, emotional, etc.) in such way that syntax, semantics and pragmatics are not independent modules but they form an analyzable continuum that is processed simultaneously. With respect to the same basic situation there are several conceptualization options with the same number of linguistic representations, as can be seen in (1a-b): (1) a. b.

La carretera baja del Popocatepetl al centro de la ciudad ‘The road goes down from the Popocatepetl to the city center’ La carretera sube del centro de la ciudad al Popocatepetl ‘The road goes up from the city center to the Popocatepetl’

More relevant than the actual location of the speaker is the reference point from which the mental scanning of the trajectory is initiated. While the choice of the verb responds to meaning the same procedure holds for morphemes and syntactic constructions as they all respond to the way events in the world are conceptualized. There are no empty structural forms or meaningless elements. In both morphemes and syntactic constructions there is a crucial semantic import such that they not only reflect meaning but also specific cognitive strategies and linguistic

Chapter Six

192

patterns. For instance, first and third person pronouns in Spanish may be more significant than expected. Maradona and Hugo Sánchez seem to have mastered new strategies to represent themselves. In one they are plain human beings, in the other they seem to be the main characters of their own novel. In being interviewed about some soccer issue they will respond using third person la situación del equipo es ideal para esta temporada ‘the state of the team is ideal for this season’ but if they are asked about their performance or their plans, instead of the expected first person, a particular use of third person is obtained: (2) a. b.

Hay Maradona para rato lit. ‘There is Maradona for a long time’ ‘Maradona will last for a long time’ Hugo está dispuesto a jugar con todo, para eso entra a la cancha ‘Hugo is ready to put it all in the game; that is why he goes into the field’

Nothing precludes the use of first person yo ‘I’; however by using third person our soccer heroes share the position of their fans to admire themselves. In order to account for these problems we need a theory that not only acknowledges the importance of meaning but also takes into account the speakers’ conceptualization of the event. Linguistic units are constituted in usage events that are represented in the system as abstract forms. Their value in the system depends on their degree of entrenchment; i.e. the degree to which they are conventionalized in the speech community. Less entrenched units (electrolisis ‘electrolysis’), require more conceptual effort than those deeply established in the linguistic community (mesa ‘table’, perro ‘dog’). An obvious example of this difference is the contrast between the passive (3a) and the periphrastic passive (3b) in Spanish where the former has taken over the majority of passive usages (Cabañas Maya 2005, Melis and Peña-Alfaro 2007). (3) a. b.

El control de las tierras ha sido resuelto históricamente por medio de la violencia (CREA, Prensa, México, 1997). ‘Land control has been resolved historically by means of violence’ Esta denuncia se resolvió con el pago correspondiente del pugilista (CREA, Prensa, México,1997) ‘This demand was resolved by the corresponding payment to the boxer’

The simplicity and directness of the “short passive” accounts for the difference. Most se passive constructions need not express the agent (el caso se resolvió ‘the case was resolved’) while “long passives” tend to require an overt agent (el caso fue resuelto por el juez de distrito ‘The case

Word in Cognitive Grammar

193

was resolved by the district judge’). While some cases correspond to the degree of complexity of the construction, others may respond for alternative motivations. The beneficiary in Spanish tends to be coded in dative (Tachita le preparó la cena a los niños ‘Tachita prepared dinner for the children’) rather than benefactive (Tachita perparó la cena para los niños ‘Tachita prepared dinner for the children’) even with verbs such as comprar ‘buy’ for which the beneficiary is not part of its valence. The dative construction reflects a degree of affectedness that the benefactive construction lacks (Maldonado 2002). The only requirement for the adequate description of a language is identifying the symbolic assemblies that take place both in lexicon and grammar. The total or partial coincidence of usage events is what allows us to establish more abstract representations (schemas) that provide the coherence of the linguistic system. These abstractions are reinforced and entrenched as they are more often used with less and less conceptual effort. More specific CG notions will help us define the levels at which the notion of word operates. Symbolic units are of different degrees of complexity. Words can be mono or poly-symbolic, the former being autonomous morphemes, whereas the latter may be constituted by autonomous/dependent morphemes or exclusively by dependent morphemes. Morphemes form words according to cohesion patterns to which we will refer below. Assume for the time being that an agentive morpheme like ero requires a noun to elaborate the schematic noun already present in its valence. Thus a -ero needs mesa ‘table’ to derive the word mesero ‘waiter’. In a higher level things contrast depending on the degree of compositionality/fusion: (4) a. b. c.

Estirar la mano ‘extend the hand’ [compositional/unit] Estirar la pata ‘die (equivalent to kick the bucket)’ [unit/compositional] Estirar la tela ‘extend the piece of cloth’ [compositional]

In (4a) there are two readings, the compositional one in which we actually extend our hand to receive something or the idiomatic unit referring to the common practice of politicians or policemen to practice extortion. The compositional reading is dominant. In (4b) we have the opposite situation where the dominant reading is the idiomatic unit ‘die’, while the compositional reading is still possible. In (4c) only the literal compositional reading is possible. This set of variations license a higher order schema [V+[DET+N]], an abstract schematization of an array of

Chapter Six

194

coinciding uses. Once established, the schema becomes a productive unit which operates as the basis for unrestricted number of expressions in the language. Thus a simple unit can combine with other units to form a higher order unit with higher degree of schematicity. We can suggest that the grammar of a language responds to the need of expressing complex conceptualizations that are beyond the boundaries of a simple lexical unit, as required by context (un policía gordo ‘a fat policeman’ > un policía que pesa más de cien kilos ‘A policeman that weights more than a hundred kilos’). In CG meaning is equated with conceptualization, which is explained in terms cognitive processing. Thus conceptualization is interpreted quite broadly for it includes novel conceptions as well as fixed concepts; sensory, kinesthetic, and emotive experience; recognition of the immediate social, physical, and linguistic context. Semantic structure is but a special case of conceptual structure. A linguistic common expression belongs to a network of related meanings. Most lexical items are polysemic, having a considerable array of interrelated senses, which define the range of their conventionally sanctioned usage. Certain senses are “schematic” relative to others, as indicated by the solid arrows. Some represent “extensions” from others as indicated by the dashed-line arrows. The nodes and categorizing relationships in such a network differ in their degree of entrenchment and cognitive salience. The heavy-line box in Figure 6-1 corresponds to the category prototype. Mesa redonda ‘round table’ may be an elaboration of the more schematic mesa ‘table’ if it only refers to the shape; however, if it refers to the event where some topic is discussed then it is an extension from the prototype and the meaning of shape remains in the base. mesa

mesa de centro

mesa redonda

mesa redonda

mesa de discusión

Fig. 6-1. Mesa ‘table’. Semantic network sample

Elaborations are represented with solid arrows while extensions correspond to dashed arrows. The degree of cohesion between points in the net may vary, as much as the degree of prominence of each point.

W Word in Cognitiive Grammar

195

Some meannings are eassily activated d while otherrs may requ uire more context. Whhile mesa ‘tablle’ is automatically activateed, mesa redonda in its abstract sensse may require a situation of o discussion tto be brought in. The radiial networks suggest that meaning is eencyclopedic. No clear boundary can distinguissh world kno owledge and linguistic kn nowledge. Language iss based on genneral knowled dge. Any faceet of our know wledge of an entity iss capable in principle of playing a roole in determining the linguistic beehavior of an expression th hat designatess it (e.g. as a semantic extension, oor in combinnation with other expresssions). The semantic network is oorganized in teerms of centraality and peripphery as deterrmined by the domain in which ann expression is i activated. T The word cerdo ‘pig’ evokes an aanimal, but onnce the domaiin of eating oor playing spo orts are at play the inteerpretation off Lucas es un cerdo ‘Lucas is a pig’ is reeferred to his bad mannners or to his h roughness as a soccer player. The notion n of domain is eequitable withh Fillmore’s frame fr (1982). In the same way that cerdo takes different inteerpretations deepending on iits dominion, the word mordelón, liiterally ‘one who w bites’, maay mean nothiing to a Spaniard while for speakerss of Mexican Spanish it deesignates the policeman that extorts drivers wideely in Mexicoo City. The do ominion now activated is th hat of the ARBITRARIN NESS OF THE MEXICAN M LAW. A further exa ample of this notion is provided in Figure 6-2:

Fig. 6-2. Coggnitive domain

For peopple not familiaar with Jewish h culture the iimage may reepresent a chandelier aand the star inn the back may m simply bee ornamental but for a member of the Jewish coommunity thee Star of Daviid activates not n only a menorah buut also Hannuukah, the prizees given to kiids along nine days of this festivityy, the dreidel and a so on. The orgganization of meaning in domains coontrasts radicaally with formal approoaches to sem mantics where the meaning oof a word is reeduced to

196

Chapter Six

a bundle of features. Domains need not be primitives and there is no reason to have a limited number of them. They vary in degree of complexity. Basic domains correspond to bidimensional and tridimensional representations that include scalar organizations such as chromatic, musical, geometrical and temporal scales, sensory, emotional domains and so on, but the vast majority of linguistic expressions deal with more complex levels of conceptual organization. The full description of the meaning of an expression corresponds to three levels: the place it occupies in the network, its cognitive domain as provided by a specific context and its mental imagery, i.e. the way in which speakers construe a mental image of the event. The full description of imagery exceeds our goals in this chapter. We will show some basic principles that will help us define the notion of word. Expressions like Un vaso medio vacío/lleno ‘A half empty/full glass’ do not contrast in its referential value but in the way the image is construed. Perspective determines its mental construal. Other crucial dimensions of imagery are the degree of schematicity/elaboration of events, scalarity, priming, the projection of the physical body onto objects in the world, the degree of subjectivity and the metonymic nature of lexicon and grammar. We will focus on three dimensions here: the profile/base distinction, the compositional path of events, and the determinacy of scope. In every linguistic expression a profile is imposed from a base. The base corresponds to the domain (or to each one of the dominions composing the complex matrix of an expression). The profile is a substructure of that domain receiving special prominence. Depending on the focal adjustment a substructure can be profiled from a more generic one. Thus, ojos ‘eyes’ and boca ‘mouth’ can be profiled with respect to the bigger structure cara ‘face’ and of course cara can be profiled as a substructure of a bigger domain such as cuerpo ‘body’ and so on. Typical examples of this phenomenon are hipotenusa ‘hypotenuse’ as profiled with respect to triángulo ‘ triangle’, esposa ‘wife’ which in the domain of parental relationships is profiled with respect to esposo ‘husband’ and punta ‘tip’ from lápiz ‘pencil’ as shown in Figure 6-3:

Word in Cognitive Grammar

197

Fig. 6-3. Figure/Base

Crucially this distinction is not reduced to lexical forms but applies to morphemes and grammatical constructions as well. Agentive morphemes like -dor share the processual base with verbs like admirar ‘admire’ (Figure 6-4b) but profile only the trajector of that process as in Fig. 6-4c

Fig. 6-4. Profiling in admirar-admirador

In the same vein, salir ‘go out’ (salió de casa ‘He left home’) and salido ‘gone, visible’ (El clavo está salido ‘The nail is out/visible’) share the verbal base as some trajectory must be conceptualized, but they profile different parts of the whole structure. The participle only profiles the end part of that process. Crucially, salido (Figure 6-5b) but not salir (Figure 65a) shares the same profile with the noun fuera ‘outside’ (Figure 6-5c) while the noun lacks the processual base found in the participle:

Chapter Six

198

Fig. 6-5. Profiling in salir-salido-fuera

The base/profile contrast may be insufficient for all grammatical structures where more than one element is profiled. Since a relational predication may involve more than one figure it is necessary to distinguish a main figure, the trajector (TR), from a secondary one, the landmark (LM). Subject and object are notions corresponding to TR and LM respectively which signal the degree of prominence of the noun phrase in a clause. Depending on how the event is conceptualized, lámpara ‘lamp’ can be construed either as TR (5a) or LM (5b) from the same objective situation, as can be seen from Figs. 6-6a and 6-6b below: (5) a. b.

La lámpara está sobre la mesa ‘The lamp is on the table’ La mesa está debajo de la lámpara ‘The table is below the lamp’

Fig. 6-6a. arriba

Fig. 6-6b. abajo

Central to conceptual profiling is the notion of compositionality, a dimension that incorporates not only basic profiling but also the path in assembling a linguistic structure. Simple and complex expressions contrast precisely that way. Lexical and periphrastic causatives illustrate this point.

Word in Cognitive Grammar

199

The generic characterization of matar ‘kill’ and hacer morir as [CAUSE TO DIE] as traditionally done by componential analyses (Bierwisch 1975, Foley and Van Valin 1984, and many others) is misleading in that it loses crucial distinctions: (6) a. b.

El director de Pemex mató a sus empleados ‘Pemex director killed his employees’ El director de Pemex hizo que murieran sus empleados ‘Pemex’s director had his employees killed’

As already underlined by different analyses (Fodor 1970, Wierzbicka 1975) in matar the act of causing and dying constitute one indivisible event (6b), whereas two separate events can be distinguished letting causation have more prominence (6a).

Fig. 6-7a. Hacer matar ‘make kill’

Fig. 6-7b. Matar ‘kill’

200

Chapter Six

The compositional path for these conceptualizations is quite different. As can be seen from Fig. 6-7a, first empleados ‘employees’ and morir ‘kill’ combine to constitute the LM of causar ‘cause’. This output elaborates in detail the LM [X] of cause to constitute the complex event CAUSAR-MORIR-EMPLEADOS (CAUSE-DIE-EMPLOYEES) and finally the TR jefe ‘boss’ is incorporated. In this case, the boss can only be responsible for the causing event not for its result since he does not directly impinge the change of state on his employees. Some other force or someone else may have done the act of killing. In contrast, Fig. 6-7b shows that the boss directly imposes dead on his employees. From the compositional path not only the points of convergence between the two events can be seen but also the way in which matar ‘kill’ and causar morir ‘cause to die’ trigger different conceptualizations. Compositionality is determinant in topicalization, complementation, subordination and a variety of grammatical phenomena. Crucial for this chapter is how it interacts with valence in morphology and syntax. Since Tesnière (1959) the study of valence has been centered in the relation between the verb and its arguments. Slowly the notion has been extended to prepositions but it is evident that there are valence relations between adjective and noun, adverb and verb, noun and noun, and most of all, among morphemes. Let us consider three different cases: (7) a. b. c.

Hija de Nely ‘Nely’s daughter’ Camisa sucia ‘dirty shirt’ Ropero ‘closet’

In (7a) hija ‘daughter’ being a relational noun has in its valence the notion of parent. Thus the prepositional phrase is part of its valence [hija [PP]]. In the same vain, sucia ‘dirty’ being an adjective by default has in its valence a noun [[N] sucio]. This generates the schematic representation by which any adjective has a schematic noun. Thus sucio can combine with a variety of nouns plato sucio ‘dirty plate’, niño sucio ‘dirty kid’ and so on. More obviously morphemes have valence. We can posit that -ero has a schematic verb in its valence [[VB] ero] to derive a locative noun. So we can say ropero ‘closet’, librero ‘bookshelf’, trastero ‘cupboard’, etc. Valence relations guarantee different degrees of tightness among morphemes in word formation. In fact they are partially determined by the degree of conceptual autonomy/dependence of each form. A symbolic form is autonomous if it does not require another element for its mental representation. In the relation camisa sucia ‘dirty shirt’ camisa may exist without the adjective compré una camisa ‘I bought a shirt’. This is not the

Word in Cognitive Grammar

201

case for sucia whose existence depends on the presence of a noun. We may assume (8) as valid: (8) camisa = autonomus; sucia = dependent

In the same vain, libro is conceptually autonomous since it does not require a derivative suffix. In contrast the suffix -ero is dependent since a schematic noun is part of its valence. Now the outcome of each combination is imposed by the profile determinant and this changes at each level of the composition of a symbolic form. In camisa sucia ‘dirty shirt’ the noun is the profile determinant as we get a type of shirt not a type of dirty and the construction is nominal. Modifying relations can be defined as those in which the profile determinant is an autonomous element, as can be seen in (9): (9)

The opposite case is represented by librero ‘bookshelf’ where the dependendent element -ero is the profile determinant as we get a type of shelf, not a type of book. We can define complement relations as those in which the profile determinant is the dependent element. This can be seen in (10): (10)

Chapter Six

202

The notion of scope is also fundamental for grammar. Scope is the amount of information necessary to conceptualize something in alternative ways. It can be defined as the sphere an expression may cover. An evident manifestation of this phenomenon is the conceptualization of body parts. Cabeza ‘head’, pierna ‘leg’, mano ‘hand’ and brazo ‘arm’ have different scope dimensions. For example, brazo is in the immediate scope of elbow and hand, while the scope of mano covers dedos ‘fingers’, nudillos ‘knuckles’ and uñas ‘nails’. Thus we can say la punta del dedo ‘the tip of the finger’ but not *la punta de la mano ‘the tip of the hand’ and we can say las uñas de los dedos ‘finger nails’ but not *las uñas de las manos ‘hand nails’. The immediate scope of an expression contrasts with its maximal scope in that the former includes only the elements directly relevant to focalize the core meaning of an expression while the maximal scope covers its relevant background. Thus for nail the immediate scope is finger but its maximal scope includes not only the hand but also elbow and arm, even the whole trunk. Scope is also a determining factor of syntactic and semantic organization, in a number of ways. The aspectual contrast between perfective (11a) and progressive (11b) is a matter of scope: (11) a. Adrián pintó su casa ‘Adrian painted his house’ b. Adrián está pintando su casa ‘Adrian is painting his house’

The extent of information considered for the expression, its scope, depends on the distance of the conceptualizer with respect to the event. Enough mental distance is necessary for the whole event to be captured as in the perfective (11a) represented in Figure 6-8a. However the progressive in Figure 6-8b corresponds to cases where the conceptualizer is in the immediate scope of predication and the event is actually taking place as the conceptualizer speaks.

Fig. 6-8a. Perfective

Fig. 6-8b. Imperfective

Word in Cognitive Grammar

203

Alternatively an imperfective verb like parecerse ‘resemble’ projects a permanent state that goes beyond the boundaries of the maximal scope of predication as in (12a). However, if we impose a progressive view on a perfective we restrict it to the immediate scope of predication and a change of state reading is obtained, as in (12b): (12) a. Lucas se parece a su papá ‘Lucas resembles his father’ b. Lucas se está pareciendo a su abuelo ‘Lucas is [now] looking like his grandpa’

The behavior of adverbs is also strongly determined by scope. Selective and totalizing adverbs (Vigueras 2000) vary according to scope. In (13a), the selective adverb únicamente ‘only’ choses Juan as its figure among other potential candidates. The scope of únicamente increases as it precedes the verb. In (13b), the VP is now in the maximal scope of the adverb in such a way that ‘speaking with Juan’ contrasts with all other actions he could have done: (13) a. Hablé únicamente con Juan ‘I spoke only to Juan’ b. Únicamente hablé con Juan ‘I only spoke to Juan’

Totalizing adverbs show an even greater scope. When placed after the verb, totalmente has scope over the VP. However in initial position it has scope over the whole sentence, as in (14b): (14) a. Su familia lo ignoraba totalmente ‘His family totally ignored him’ b. Totalmente, su familia lo ignoraba ‘Totally, his family ignored him’

Interestingly when the adverb has scope over the whole sentence the subjective evaluation of the conceptualizer emerges in such a way that the adverb in (14b) now functions as a pragmatic marker.

3. Approaching cognitive phonetics As for the phonetic pole, every linguistic unit that is uttered is, by definition, a usage event that can be re-instantiated, entrenched and is therefore cognitively autonomous, regardless of whether it is meaningful

204

Chapter Six

or not. From the perspective of CG, this is our initial proposal for defining a syllable: a rooted sensory-motor coordination schema of uttering activity entrenched in speakers, such that it may be executed and perceived as conventional by the members of a linguistic community. Each speaker of a language masters a large inventory of syllables. As their acquisition goes on, syllable domains grow and speakers establish correspondences amongst their constitutive gestural categories: edge-syllabic and synsyllabic schemata. Edge-syllabic dependent schemata are syllable initial (onset) or syllable final (coda) constricting gestures that attenuate harmonics or produce white noise in a conventional temporal sequence. For example, the word ropa ‘clothes’ contains the syllables [ro] and [pa]. Although such syllables can be uttered on their own, their corresponding constitutive onsets [r] and [p] cannot. Like all edge-syllabic schemata these onsets depend on further cognitive elaboration. Syllabic bound gestures such as oscillating glottis and raised soft palate encompass all instantiations of [r] vibrating tongue tips. However [r] is instantiated in [ro] with simultaneous rounded lips and retracted tongue root, whereas in re [re] (as in the musical note re) it is accompanied by resting lips and tongue root. The phonetic variable alternations of a supposedly autonomous [r] phone are not the consequence of assimilation with allegedly adjacent [o] or [e] segments, but rather the result of the coordination of their intrinsically overlapping gestures. Hence, the acoustic variability of sub-syllabic onsets and codas is always motivated by the integration of edge-syllabic and syllabic bound gestures. Synsyllabic autonomous schemata are conventionally bound syllable gestures that profile characteristic sequences of acoustic bandwidths, produce the syllable’s harmonic acoustic waves and motivate a nucleic temporal sequence of minimal constriction and maximal harmonic prominence. Synsyllabic schemata are autonomous because the coordination routine that they are based upon may be articulated on their own (accounting for all nucleic or vowel only syllables) and because they afford all the specifications needed by the edge-syllabic schemata that ride on them. For example, on the one hand, a [u] labeled synsyllabic schema has a characteristic harmonic bandwidth that corresponds to several postures across whole syllables including raised oral platform (oral cavity high floor), adduction of the glottis (sonority), raised soft palate (non nasal), rounded lips, and retracted tongue root. On the other hand, an edgesyllabic schema such as [s] determines the presence of a tongue body critical constriction, the abduction of the glottis, very low harmonics (if any) and a wide bandwidth of white noise.

Word in Cognitiive Grammar

205

206

Chapter Six

Fig. 6-9 (previous page). Spectrographic beats of the alveolar voiceless fricative posture of Mexican Spanish, roughly corresponding to the underlined letters of the following written words: 1) desarrollemos, 2) judas, 3) enlaza, 4) todas, 5) naturaleza, 6) está, 7) servicio, 8) labores, 9) desnuca, 10) seguimos, 11) semana, 12) comes, 13) servicio, 14) análisis, 15) exigimos, 16) música, 17) abusa, 18) comemos, 19) desarrollamos, 20) dejamos, 21) fallamos, 22) dolosa, 23) fallemos, 24) hallamos, 25) recibimos, 26) dirigimos.

In Mexican Spanish, the [s] of the syllable [si] is articulated at the anterior section of the alveolar region (alveolar ridge), whereas the [s] in [su] is displaced towards the corresponding posterior section (postalveolar or palatoalveolar) due to the traction exerted on the tongue body by the retraction of the tongue root, as determined by the synsyllabic schema represented by the [u]. Therefore, the so-called [s] allophones can be accounted for by the integration of the [s] dependent edge-syllabic schemata schema with the autonomous [a, e, i, o, u] synsyllabic schemata, as exemplified in the following spectrographic representation of 26 instantiations of [s] edge-syllabic postural schemata in the same number of words. The first 15 postures correspond to the onsets and codas of synsyllabic schemata commonly designated by the letters a, e, and i. The remaining onsets and codas are integrated with type u and o synsyllabic schemata. The lowered limit of the frequency band is caused by the widened resonating oral cavity, as configured by the scope of the instantiated synsyllabic schemata, with a retracted tongue root, as well as rounded and protruded lips. If one considers that syllables are the basic autonomous units of the phonetic pole, a schematic exposition of their cognitively complex internal structure is needed, such that it may not be composed of inherently autonomous segments. Fridman (2009) proposes an initial version of such a model.1 Figures (10a), (10b) and (10c) are models of Onset and Coda Schemata for the initial and final syllables of ro.’pe.ros ‘wardrobes’. The coordinative structures listed on their left margin correspond to the representation of gestures to be produced, from the second to the last row. The alphabetic text in their cells describes postures, or stable articulatory states, whereas special symbols are used for three kinds of Transitions, or articulatory change processes, as follows:

1

To see more examples of this model, the reader may visit the website MorphoPhonetics [http://cognitivephonetics.com].

Word in Cognitive Grammar i.

207

‘>’

Deactivation of a preceding posture, towards the beginning of an ensuing resting posture, ii. ‘

oscillating alveolar

>

Lips Tongue Tip Tongue Body Velum

//////////

Oral Platform Orbicularis Oris Tongue Root Glottis

Fig. 6-10a. [r[Ȉ]] onset sch.

//////////

Fig. 6-10b. [‫[ݐ‬Ȉ]] onset sch.

208

Chapter Six

Fig. 6-10c [[Ȉ]s] coda sch.

Every column represents a Beat in a gesturing process, i.e. the unit of conceived time at which the active postures and transitions (changes in between postures) of several coordinative structures are anchored. Every beat is headed by a set of numbers and characters in the Onset Scale (upper row) representing the intrinsic properties of each beat, as well as their patterns of mutual cognitive integration. The square brackets indicate that a beat is cognitively autonomous, since all the cells within its temporal scope (column) are only elaborated by simultaneous postures in the cells of the corresponding coordinative structures. As a whole, such beats are non-changing complex articulations, and hence constitute Complex Postures, conceived as matrixes composed of several articulatory domains (Fridman 2010). The other non-numeric characters in the upper row indicate the directionality of activations/deactivations undergone by the coordinative structures within their beat scope, as represented by the corresponding column: ‘>’ when all the transitions in the column are deactivations (‘>’), ‘