Cross-Disciplinary Perspectives on Lexical Blending 9783110289572, 9783110289237

Table of contents :
Acknowledgements
Introduction: A bird’s-eye view of lexical blending
Blends: Core and periphery
Beyond all reasonable transgression: Lexical blending in alternative music
Blend formation in Modern Greek
Lexical blending in Polish: A result of the internationalisation of Slavic languages
Ukrainian Blends: Elicitation paradigm and structural analysis
Are reduced compounds compounds? Morphological and prosodic properties of reduced compounds in Russian and Mandarin Chinese
Blending between grammar and universal cognitive principles: Evidence from German, Farsi, and Chinese
Quantitative corpus data on blend formation: Psycho- and cognitive-linguistic perspectives
A Combinatory Logic and formal-semantic account of lexical blending
Stress in English blends: A constraint-based analysis
Output-to-output faithfulness in the phonological structure of English blends
Portmanteaus as generalized templates
Contributors
Index

Citation preview

Cross-Disciplinary Perspectives on Lexical Blending

Trends in Linguistics Studies and Monographs 252

Editor

Volker Gast Founding Editor

Werner Winter Editorial Board

Walter Bisang Hans Henrich Hock Heiko Narrog Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume

Volker Gast

De Gruyter Mouton

Cross-Disciplinary Perspectives on Lexical Blending edited by

Vincent Renner Franc¸ois Maniez Pierre J.L. Arnaud

De Gruyter Mouton

ISBN 978-3-11-028923-7 e-ISBN 978-3-11-028957-2 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. ” 2012 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Frank Benno Junghanns, Berlin Printing: Hubert & Co. GmbH & Co. KG, Göttingen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

Contents Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Introduction: A bird’s-eye view of lexical blending. . . . . . . . . . . . . . . . . 1 Vincent Renner, François Maniez and Pierre J. L. Arnaud Blends: Core and periphery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Laurie Bauer Beyond all reasonable transgression: . Lexical blending in alternative music. . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Paula López Rúa Blend formation in Modern Greek. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Angela Ralli and George J. Xydopoulos Lexical blending in Polish: . A result of the ­internationalisation of Slavic languages. . . . . . . . . . . . . . 51 Ewa Konieczna Ukrainian Blends: Elicitation paradigm and structural analysis. . . . . . . 75 Susanne R. Borgwaldt, Tetyana Kulish and Arpita Bose Are reduced compounds compounds? Morphological and prosodic properties of reduced compounds in Russian and Mandarin Chinese . . . 93 Giorgio Francesco Arcodia and Fabio Montermini  Blending between grammar and universal cognitive principles: . Evidence from German, Farsi, and Chinese. . . . . . . . . . . . . . . . . . . . . . . 115 Elke Ronneberger-Sibold Quantitative corpus data on blend formation:. Psycho- and cognitive-linguistic perspectives. . . . . . . . . . . . . . . . . . . . . 145 Stefan Th. Gries A Combinatory Logic and formal-semantic account of lexical blending. 169 Christian Bassac

vi Contents Stress in English blends: A constraint-based analysis. . . . . . . . . . . . . . . 193 Outi Bat-El and Evan-Gary Cohen Output-to-output faithfulness in the phonological structure of . English blends. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Ewa Tomaszewicz Portmanteaus as generalized templates . . . . . . . . . . . . . . . . . . . . . . . . . . 233 Jochen Trommer and Eva Zimmermann Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

Acknowledgements Most of the papers published in this volume were first presented at the International Conference on Lexical Blending, which took place in Lyon in June 2010. We would like to thank all those who took part in the two days of lively discussion. We are also grateful to the contributors for their work and cooperation throughout the publication process, to an anonymous reviewer for his/her comments and suggestions, to Volker Gast for accepting the volume in the series, and to Birgit Sievert, Julie Miess, Frank Benno Junghanns and all the staff at De Gruyter Mouton for their assistance. Lyon, October 2012 Vincent Renner François Maniez Pierre J. L.  Arnaud

Photos Introduction: A bird’s-eye view of lexical blending Vincent Renner, François Maniez and Pierre J. L. Arnaud 1. A brief retrospective view Lexical blends have been popularized in English by the Victorian author Lewis Carroll, who not only elaborated many new formations made up of word fragments, but also pondered on the process of lexical blending in his writings: Well, “slithy” means “lithe and slimy.” “Lithe” is the same as “active.” You see, it’s like a portmanteau – there are two meanings packed up into one word. Alice’s Adventures in Wonderland (1865) For instance, take the two words “fuming” and “furious.” Make up your mind that you will say both words, but leave it unsettled which you will say first. Now open your mouth and speak. If your thoughts incline ever so little towards “fuming,” you will say “fuming-furious;” if they turn, by even a hair’s breadth, towards “furious,” you will say “furious-fuming;” but if you have the rarest of gifts, a perfectly balanced mind, you will say “frumious.” Preface to The Hunting of the Snark (1876)

Blends, however, long predate the latter half of the nineteenth century. Early examples which have survived until the present day include, for instance, Middle French écornifler ‘to nose about and steal’ (< écorner ‘to cut off’ + nifler ‘to sniff’), Early Modern English foolosopher (< fool + philosopher) and Middle Polish pstrokaty ‘motley’ (< pstry ‘multicolored’ + srokaty ‘piebald’). Lexical blending is attested in a variety of typologically different languages – e.g. French (Fradin, Montermini and Plénat 2009), Hebrew (Berman 1989), Indonesian (Dardjowidjojo 1979), Mandarin Chinese (RonnebergerSibold, this volume) –, but much of the specialized literature has dealt with English, both because it is assuredly the most researched language of modern times, and also because of the relative productivity of the process in Late Modern English. English blends notably multiplied in the course of the twentieth century (Ayto 2003), which concurrently led to an increasing number of scholarly works investigating the phenomenon, among which one can mention the widely quoted articles by John Algeo, Blends, a Structural

2 Vincent Renner, François Maniez and Pierre J. L. Arnaud and Systemic View (1977), and by Garland Cannon, Blends in English word formation (1986). Several influential papers – Kubozono (1990), Bat-El (1996), Lehrer (1996) and Kelly (1998) – appeared in the 1990s and undoubtedly gave a new impetus to research in the field, providing fresh insights from a phonological and psycholinguistic perspective. This has resulted in a wide-ranging body of research spanning several linguistic fields – from phonology (e.g. Cutler and Young 1994) to psycholinguistics (e.g. Lehrer 2003), to cognitive (e.g. Kemmer 2003; Gries 2006) and computational (e.g. Cook and Stevenson 2010; Veale and Butnariu 2010) linguistics –, frameworks, notably Optimality Theory (e.g. Piñeros 2004; Hong 2005), and languages (e.g. Bertinetto 2001; Brdar-Szabó and Brdar 2008). 2. Terminological and definitional dissonance The term lexical blend has nowadays mostly displaced Carroll’s portmanteau (word), which is however still used either as an occasional synonym for lexical blend (e.g. Veale and Butnariu 2010; Konieczna, this volume) or to refer to a subcategory of blends (e.g. Algeo 1977; Piñeros 2004; Trommer and Zimmermann, this volume). The words which give the fragments from which a blend is built are usually termed source words; less frequent synonyms include base word (e.g. Bat-El and Cohen, this volume), constituent (word) (e.g. Kelly 1998; Konieczna, this volume; Ralli and Xydopoulos, this volume) and etymon (e.g. Cannon 2000). The constitutive fragments of a blend are commonly designated as splinters (e.g. López Rúa, this volume; Ronneberger-Sibold, this volume). Splinter may also refer to bound morphemes such as Franken‑ (from Frankenstein) and ‑nomics (from economics), which originated in blend formations (Bauer 2004); this has led to the use of the term fracto-lexeme to refer to word fragments which do not have morphemic status (Renner and Lalić-Krstin 2011). Authors who do not consider that complex words made up of two initial submorphemic splinters (e.g. napalm < naphthenate + palmitate) should be labelled blends use the term complex clipping (e.g. Gries, this volume), clipped compound (e.g. Bat-El and Cohen, this volume) or clipping compound (e.g. Bauer, this volume). Linguists have never quite agreed on the exact delimitation of the concept of lexical blending, and the contributors to the present volume are no exception. Some authors consider that the clipping of one source word is enough to qualify for membership in the category (e.g. Lehrer 2007; Brdar-Szabó and Brdar 2008; Konieczna, this volume; López Rúa, this volume), but others exclude various types of complex words, as the following sample of criteria present in the literature shows:

Introduction: A bird’s-eye view of lexical blending 3

– For Bat-El (2006: 66) and Bauer (this volume), a complex word is unambiguously a blend only in case of truncation at the “inner edges”, i.e. if the left source word has been back-clipped and the right source word fore-clipped. – For Ralli and Xydopoulos (this volume), a complex word is a blend only if no source word remains intact and they exclude words of the slanguage (< slang + language) type, when n final segments of the left source word overlap with n initial segments of the right source word. – For Dressler (2000), complex words whose source words are not in a semantically coordinate relation (i.e. are in a modifier-head relation), such as motel (< motor + hotel), are excluded from the category; for Plag (2003: 123), “proper blends” are those which semantically “resemble copulative compounds”, i.e. are coordinate. – For Arcodia and Montermini (this volume), complex words which do not manifest overlapping of part of their source words are excluded and categorized as reduced compounds. – For Tomaszewicz (this volume), if there are no output-to-output correspondence relations at the level of metrical structure between the source words and the resulting complex word, the complex word is not, strictly speaking, a blend. In order to accommodate the many diverging views, one may resort to adopting a prototypical approach and consider that the most inclusive definition is the one to be retained and that the above characteristics are not to be taken as defining features, but as typicality features. The corollary of a prototypical approach is the existence of fuzzy boundaries, which is not in itself a problem for descriptivists – López Rúa (2004) for instance shows how such an approach can work –, but Bauer (this volume) underlines that the uncertain limits of the phenomenon may be an obstacle in the quest for the elusive singularity of blends. 3. Outline of the volume This collection of articles has its origin in a conference on lexical blending which was organized at the University of Lyon in June 2010. The aim of the conference was to bring together linguists working in various languages and different disciplines or frameworks in order to encourage debate and cross-fertilization of ideas. Eleven papers from the conference, together with one invited contribution, make up the contents of the volume, whose goal is to combine perspectives in order to give a broad overview of the

4 Vincent Renner, François Maniez and Pierre J. L. Arnaud current research and act as a catalyst for further cross-linguistic and crossdisciplinary investigation. The subject is approached from a variety of disciplinary vantage points (morphological, semantic, phonological, cognitive), methodological approaches (experimental, statistical, corpus-based, contrastive) and theoretical frameworks (Combinatory Logic, formal semantics, Natural Morphology, Optimality Theory and its Output-to-Output Correspondence Theory and Coloured Containment versions). A first group of seven articles provides theoretically informed descriptions of salient aspects of lexical blending in a variety of understudied and typologically diverse languages (the ­Germanic, Slavic, Greek, Iranian and Chinese subfamilies are represented). In the introductory chapter, Laurie Bauer underlines that lexical blending is best grasped if it is understood as a prototypical category whose description centers around a number of defeasible constraints. He examines a series of constraints which are thought to influence the output form of a blend, discusses the semantics of coordinate blends, and concludes by making a plea for a firmer delineation of the category, stressing that these formations have not been accounted for in an entirely satisfactory way yet and that further investigation and a “flash of insight” are still needed to adequately capture the essence of lexical blending. In Beyond all reasonable transgression: lexical blending in alternative music, Paula López Rúa approaches the issue of blend formation by analyzing a corpus of blends naming bands and singers in the field of alternative music, from the point of view of form, underlying structure and purpose. Her corpus reveals a high frequency of items formed out of two bases and of combinations which retain the left source word or even both source words in their entirety. She also observes a tendency towards high constituent integration, and the occasional presence of an underlying grammatical modification relationship and of what she calls “transgressive semantic combinations”. She concludes that while lexical blending is primarily used to catch the audience’s attention, it also conveys information about the artists’ views on life, society or their art and thus represents the subculture to which they belong, reinforcing group bonds. In the third chapter, Angela Ralli and George Xydopoulos examine blending in Modern Greek, which constitutes a recent phenomenon and is still limited to slang and some dialects. They make a case for blends being structurally similar to the stem-word compounds of Greek, with differences being the absence of a linking element and form reduction predominantly on the non-head. Like compounds, blends must combine the denotations of their constituents and the authors show that some recent jocular formations are purely formal and their lack of meaning disqualifies them as blends.

Introduction: A bird’s-eye view of lexical blending 5

Beyond the discussion of the morphological creativity involved in blends vs. compounds, the descriptive side of the paper shows how blending appears in a language that did not have it before. In Lexical blending in Polish: a result of the internationalisation of Slavic languages, Ewa Konieczna echoes Ralli and Xydopoulos by explaining that lexical blending in Polish had been extremely rare until English started penetrating the Slavic languages, first in the 1960s, and then, more visibly, once the Cold War ended. She collected 234 blends from present-day Polish (a majority of which are nonce-formations) and her study details their main formal and semantic characteristics. Especially worthy of note is the fact that, in Polish, overlapping of part of the source words significantly correlates with pejorative connotation, and also the fact that graphic blends, i.e. blends which can only be identified through graphic means (e.g. przeNIKanie < przenikanie + NIK) form a sizable formal subtype. Investigating lexical blending in another Slavic language, Susanne R. Borg­­waldt, Tetyana Kulish and Arpita Bose analyze the structural characteristics of Ukrainian spoken blends that were produced in a hybrid-object naming task. Whereas previous research has mostly dealt with deliberate word-formation (written blends) or unintentional speech errors (spoken blends), their corpus consists of spoken blends that were intentionally coined in an experiment for which they used digitally manipulated color photographs of hybrid objects and asked native Ukrainian speakers to give them a name. After a brief explanation of the principles of word-formation in Ukrainian, they present the results of their hybrid-object naming study, focusing on the structure of blends and other subtractive word-formations in their data, and use their corpus to test prevalent theories of preferred blend structures. Giorgio Francesco Arcodia and Fabio Montermini’s article first looks at Russian “stump compounds” like filfak ‘phil(ological) fac(ulty)’, and views them as resulting from a combination of constraints, such as the need for each element to correspond to the minimal word pattern and also the need for the first element to be identifiable. They appeal to the Cognitive linguistics concept of conventionalization to conclude that, due to their highly constrained properties, Russian stump compounds are close to “canonical compounds”. The authors then turn to Mandarin Chinese blends. Given the nature of the language and that of their underlying forms, these blends manifest few constraints in formal or categorial terms, contrary to the situation in Russian. The fact that they are usually dissyllabic, however, corresponds to the minimal Prosodic Word of Mandarin, which is a sign of conventionalization. This leads to the conclusion that what blends in the two languages have in common is the role of the minimal Prosodic Word.

6 Vincent Renner, François Maniez and Pierre J. L. Arnaud Finally, Elke Ronneberger-Sibold proffers a contrastive analysis of lexical blending in German, Farsi and Mandarin Chinese and makes use of an original typology of blends based on relative degrees of transparency of the output (from the point of view of the producer). She shows that while the telescope blends found in German (e.g. Kamelefant < Kamel + Elefant) are absent in both Farsi and Mandarin, contour blends (e.g. German Tomoffel < Tomate + Kartoffel) occur in all three languages, which she explains by the fact that the latter are based on a universal ability to identify words by their abstract sound shapes (i.e. mainly their rhythmical contour and the vowel which carries main stress). The last five contributions provide a theoretically original focus on English and Spanish data. First, Stefan Th. Gries discusses a variety of case studies concerned with the effects of similarity and recognizability on the formation of blends, focusing on the degree of similarity between the two source words that are blended, the ordering of source words in blends and the ways in which source words are split up and merged into blends. His contribution draws comparisons between speech-error blends (both authentic and induced) and intentional word-formation blends, as well as between the latter and complex clippings, and extends previous work by proposing new corpus-linguistic operationalizations of relevant concepts and increasing the sample sizes from previous studies in the field. Some of the variables that are quantified are the length, frequency, stress pattern, semantics, ordering and locus of similarity of source words, as well as the overall similarity of source words to each other and to the resulting blend. In Chapter 9, Christian Bassac, a formal linguist, accounts for the form of blends, applying three combinators from Combinatory logic. Interestingly, his operations are on letters, but phonemes would be dealt with in the same way. He then shows how the representations of lexical units in the Generative lexicon can be combined to produce the meanings of blends in a way similar to an earlier treatment of compounds. In Stress in English blends: a constraint-based analysis, Outi Bat-El and Evan-Gary Cohen study the constraints that govern stress placement in English blends. Starting from the observation that the stressed syllable of the blend is generally identical to that of one of the source words, they attempt to identify factors that determine which of the source words provides its stressed syllable and arrive at a theory that combines the position-based view (which argues that the stress provider is the right-hand source word) and the size-based view (which argues that the stress provider is the source word whose size is identical to that of the blend). Their corpus consists of English blends originating from two source words, which involve truncation in at

Introduction: A bird’s-eye view of lexical blending 7

least one of the source words and where truncation is located in the middle of the blend, thus excluding clipped compounds such as sitcom (< situation + comedy). The authors conclude that both size and position play a role in determining stress placement in blends. Lexical blends have often been considered as products of word creation, rather than word formation, on account of their unpredictability. To overcome the lack of usefulness of rigid rules, Ewa Tomaszewicz applies Optimalitytheoretic methods, in particular those of Output to Output Correspondence Theory, which was initially developed to account for resemblances between morphologically related words. Her investigations show that it is less the segments than the prosodic structures that determine which parts of the source words are preserved in the blend. In the final contribution to the volume, Jochen Trommer and Eva Zimmermann present an Optimality-theoretic account of a specific type of Spanish portmanteau, where the shorter source word is inserted into the prosodic structure of the longer one, which also happens to be the head of the blend (e.g. dedo + democracia > dedocracia). Contra Piñeros (2004), they claim that this type of blend is not created extra-grammatically, but can be derived in the regular morphological component of the grammar. The authors regard this type of portmanteau as a case of templatic shortening aimed at matching a prosodic template. References Algeo, John. 1977. Blends, a structural and systemic view. American Speech 52 (1): 47–64. Ayto, John. 2003. Newspapers and neologisms. In Jean Aitchison and Diana M. Lewis (eds.), New Media Language, 182–186. London: Routledge. Bat-El, Outi. 1996. Selecting the best of the worst: The grammar of Hebrew blends. Phonology 13 (3): 283–328. Bat-El, Outi. 2006. Blend. In Keith Brown (ed.), Encyclopedia of Language and Linguistics, 2nd edn., Vol. 2, 66–70. Amsterdam: Elsevier. Bauer, Laurie. 2004. A Glossary of Morphology. Edinburgh: Edinburgh University Press. Berman, Ruth A. 1989. The role of blends in Modern Hebrew word-formation. In Paul Wexler, Alexander Borg and Sasson Somekh (eds.), Studia Linguistica et Orientalia Memoriae Haim Blanc Dedicata, 45–61. Wiesbaden: Harrassowitz. Bertinetto, Pier Marco. 2001. Blends and syllabic structure: A four-fold comparison. In Mercè Lorente, Núria Alturo, Emili Boix, Maria-Rosa Lloret and Lluís Payrató (eds.), La Gramàtica i la Semàntica en l’Estudi de la Variació, 59–112. Barcelona: Promociones y Publicaciones Universitarias / Secció de Lingüística Catalana de la Universitat de Barcelona.

8 Vincent Renner, François Maniez and Pierre J. L. Arnaud Brdar-Szabó, Rita and Mario Brdar. 2008. On the marginality of lexical blending. Jezikoslovlje 9 (1/2): 171–194. Cannon, Garland. 1986. Blends in English word formation. Linguistics 24 (4): 725–753. Cannon, Garland. 2000. Blending. In Geert E. Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology: An International Handbook on Inflection and Word-Formation, Vol. 1, 952–956. Berlin & New York: Mouton de Gruyter. Cook, Paul and Suzanne Stevenson. 2010. Automatically identifying the source words of lexical blends in English. Computational Linguistics 36 (1): 129–149. Cutler, E. Anne and Duncan Young. 1994. Rhythmic structure of word blends in English. In ICSLP 94: 1994 International Conference on Spoken Language Processing: September 18-22, 1994, Yokohama, Japan, Vol. 3, 1407–1410. Kobe: Acoustical Society of Japan. Dardjowidjojo, Soenjono. 1979. Acronymic Patterns in Indonesian. Pacific Linguistics Series C 45: 143–160. Dressler, Wolfgang U.  2000. Extragrammatical vs. marginal morphology. In Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 1–10. Munich: Lincom Europa. Fradin, Bernard, Fabio Montermini and Marc Plénat. 2009. Morphologie grammaticale et extragrammaticale. In Aperçus de Morphologie du Français, ­Bernard Fradin, Françoise Kerleroux and Marc Plénat (eds.), 21–45. Saint-Denis: Presses Universitaires de Vincennes. Gries, Stefan Th. 2006. Cognitive determinants of subtractive word formation: A corpus-based perspective. Cognitive Linguistics 17 (4): 535–558. Hong, Sung-Hoon. 2005. An optimality theoretic analysis of English blends. Korean Journal of Linguistics 30 (3): 525–557. Kelly, Michael H. 1998. To ‘brunch’ or to ‘brench’: Some aspects of blend structure. Linguistics 36 (3): 579–590. Kemmer, Suzanne. 2003. Schemas and lexical blends. In Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.), Motivation in Language: Studies in Honor of Günter Radden, 69–97. Amsterdam & Philadelphia: Benjamins. Kubozono, Haruo. 1990. Phonological constraints on blending in English as a case for phonology-morphology interface. Yearbook of Morphology 3: 1–20. Lehrer, Adrienne. 1996. Identifying and interpreting blends: An experimental ­approach. Cognitive Linguistics 7 (4): 359–390. Lehrer, Adrienne. 2003. Understanding trendy neologisms. Rivista di Linguistica 15 (2): 369–382. Lehrer, Adrienne. 2007. Blendalicious. In Judith Munat (ed.), Lexical Creativity, Texts, and Contexts, 115–133. Amsterdam & Philadelphia: Benjamins. López Rúa, Paula. 2004. The categorial continuum of English blends. English Studies 85 (1): 63–76. Piñeros, Carlos-Eduardo. 2004. The creation of portmanteaus in the extragrammatical morphology of Spanish. Probus 16 (2): 203–240.

Introduction: A bird’s-eye view of lexical blending 9 Plag, Ingo. 2003. Word-Formation in English. Cambridge: Cambridge University Press. Renner, Vincent and Gordana Lalić-Krstin. 2011. Predicting stress assignment in lexical blends: the case of English and Serbian. In Nenad Tomović and Jelena Vujić (eds.), ELLSIIR Proceedings, Vol. 1, 265–273. Belgrade: Publications of the Faculty of Philology of the University of Belgrade. Veale, Tony and Cristina Butnariu. 2010. Harvesting and understanding on-line neologisms. In Alexander Onysko and Sascha Michel (eds.), Cognitive Perspectives on Word Formation, 399–420. Berlin & New York: Mouton de Gruyter.

Blends: Core and periphery Laurie Bauer This chapter presents a review of literature about blending with particular reference to English, focusing on questions of definition and subcategorization. It is pointed out that much of what is taken to identify the category of blend is also found in other types of word-formation, and a plea is made for a firmer definition of the term blend, with clearer delineation of categories. Keywords: blend, English language, word-formation, recognizability.

1. What is striking about blends? Blends (perhaps particularly in English,1 but also in other languages) have been intensively studied in recent years, because their structure is so fascinating. You take two words and telescope them together into a new word: this seems like an immensely powerful procedure, and one which linguists ought to be able to describe. The result is that in one way, we know a great deal about the structure of blends. At the same time, it is not clear what we can predict. Those who start with a body of data and apply rules to it are almost always pleasantly surprised as to how much of their data is covered by their generalizations. However, they never seem to succeed completely, and often it is the case that we are left with quite a lot of processes/rules which apparently apply or not at the whim of the analyst. Not only that, blends are poorly defined, a problem to which I shall return. We are happy to say that smog and motel are blends, but there is a large area where people disagree about the classification, and we have to ask what the precise limits of blends are. We seem to have not cut-and-dried categories, but prototypical or canonical categories. So we expect to find not criteria that say that something is or is not a blend, but defeasible constraints, which may or may not be met 1. The research for this paper was supported by the Royal Society of New Zealand through a Marsden research grant to the author for the project “The morphology of current English”. My thanks go to colleagues and ICLB 2010 conference attendees for their feedback on the original version.

12 Laurie Bauer in individual cases. This in fact gives us a very good approach to what we know about blends. 2. What might we know about blends? We can divide blends into two major groups: coordinate (smog) and determinative (motel) (Dressler (2000) prefers the terms paradigmatic and syntagmatic, although he includes only the former under the heading of “blend”). Determinative blends are headed, and are endocentric (Bat-El 2006). Coordinate blends may not be headed semantically (though they may have a phonological head); they are sometimes termed exocentric (Bat-El 2006). Some blends arise as speech errors, others as deliberate or unconscious ways of expanding the vocabulary. There is a question of whether the two are formally identical, though the speech-error type probably combines antonyms more often than deliberate formations do. Kubozono (1990) looks at parallels between errors and word-formation and suggests their formal identity, but Gries (this volume) suggests formal distinctiveness as well. Many blends are clearly deliberate formations, but whether this happens more with blends than with other types of word-formation is hard to tell, because we do not have good data on either side. 2.1. Ordering In determinative blends, the order is more likely to be externally determined, especially in English: identi(ty)(Ø)kit, mo(tor)(ho)tel, jazz(Ø)(ex)ercise all follow the ordering of the corresponding phrases or compounds, for example. In coordinate blends, the order of elements cannot be determined by syntax. Relevant factors in determining the order may be: – Length (shorter first) – Frequency (more frequent first) – Semantics (more prototypical first) – Pragmatics (temporal order) These factors are taken from Kelly (1998), where determinative and coordinate blends are treated together. Examples to make the point include:

Length: Frequency: Prototypicality: Temporal order:

donk(ey)(el)ephant sm(oke)(f)og sp(oon)(f)ork br(eakfast)(l)unch

Blends: Core and periphery 13

Notice that length, frequency and prototypicality would be expected to cooccur anyway because of general principles of economy, Zipf’s Law, and the like. Note however that Lalić-Krstin and Halupka-Rešetar (2010) report that prototypicality is not a factor at least in Serbian. It is not clear whether ordering is related to other ordering principles in related constructions. Relevant points of comparison might include any of the following: – Ordering of elements in co-compounds with similar semantics: singersongwriter is fine, but ?songwriter-singer is odd; this could be a matter of length. – Ordering of elements in irreversible binomials (see Malkiel 1959): knife and fork as opposed to ?fork and knife might involve prototypicality; on the other hand, not all binominals are really irreversible – eggs and bacon and bacon and eggs are both heard. – Ordering of elements in coordination with and: if you travel by train and taxi, does the ordering reflect temporal ordering, or is it due to some other principle? At the zoo, would you go to see the seals and penguins or the penguins and seals, and would length of the elements or order of expected visitation determine which? Is an Englishman, and Irishman and a Scotsman fixed because of idiomatization (at least in the telling of jokes), or is the order determined by more fundamental principles? 2.2. Recognizability Both elements in the blend must be recognizable if the blend is to be successful. Cold as a blend of cold and old is not a successful blend – at least not in the spoken medium. It is generally accepted in the psycholinguistic literature that recognizability is easier for word beginnings than for word ends. That being the case, we would expect to find that you need less of the first word in a blend than of the second for recognition to occur. This is apparently what happens (see Gries 2004 and more generally on recognizability Fradin et al. 2009). Many of the other factors which will be discussed below in terms of formal matching and the like may be there because they contribute to recognizability. Nevertheless, whether one sees recognizability as a fundamental notion or a derived one, it is important, and too infrequently discussed as such.

14 Laurie Bauer 2.3. Formal expectations In this section, I will list a number of the defeasible constraints on form that have been discussed in the literature, give examples, and also point out instances of counter-examples. While the counter-examples are important, in the sense that the easy availability of counter-examples suggests that the constraint may need some re-working, or that it may be a relatively low constraint in a hierarchy of constraints, the fact that such constraints have counter-examples is not in itself a problem. (1) In a blend ab from words ag and db, the number of syllables in ab ≤ the number of syllables in the longer of ag and db. (That is, the longest base word sets up a maximum length for the blend.) Examples: beef(Ø)(buff)alo (= db), ball(oon)(parach)ute (< db) But: baro(que)(ro)coco (= db + 1) (2) In a blend ab from words ag and db, where ab is not a monosyllable, a and b are each at least one syllable long. Example: dies(el)(alc)ohol But: ch(annel)(t)unnel This constraint cannot work with monosyllables, since it would give a minimal disyllabic blend, which is not the case. With monosyllables, we need to make appeal to Onsets and Rhymes: smog, brunch (see below, in number (7)). Examples like palimony, where there is phonological overlap between the two elements of the blend, make it difficult to know how to apply this: does the al belong to the first word or not? Note, however, that while this question is crucial for input-oriented theories, in models based on output constraints it becomes irrelevant. The example of chunnel is an interesting one, in that it suggests that similarity between the two words which make up the blend may be more pervasive than just some phonological matching at the point of overlap. Various other contributors take up this point. (3) The stressed syllable from at least one of the two elements will be retained in the blend. Both may be. There is a preference for the stress pattern of the righthand word to be retained. Examples:  posi(tively)(abso)lutely, vodka(Ø)(mar)tini, Ox(ford)(Cam)bridge, alpha(betic)(nu)meric But: ball(oon)(parach)ute

Blends: Core and periphery 15

This is what suggests right-headedness of blends in a phonological sense. Bat-El and Cohen (this volume) take this idea much further. (4) In a blend ab from words ag and db, g and d are not null elements. Examples: edu(cation)(enter)tainment, tig(er)(li)on But: key(Ø)(con)tainer (g is null), jazz(Ø)(ex)ercise, identi(ty)(Ø)kit (d is null) In other words, this constraint postulates that something is deleted from both base words. The question arises whether this is part of the definition of a blend: it is not clear whether keytainer should be analysed as a blend or as a compound with a foreclipped second element. I will retain such examples here (and give a reason for this later), but the definitional problem underlines yet again the difficulty in determining the boundaries of blending. (5) In a blend ab from words ag and db, where a ends in a consonant and b begins in a consonant, the second consonant will be less sonorant than the first. Example: slaŋ(Ø)(lan)guage But: West(ern)(Aust)ralia Again we have the superficial problem of determining which element the belongs to, and again the answer is that this constraint may be better reformulated as an output condition. The constraint is taken from Bat-El (2000), who bases it on Venneman (1988), which is really about sequences of syllables in any words, so the original inspiration here is not specific to blends. Fundamentally it says that blends have the same kind of phonological structure as ordinary words have, something which is made more overt in the next potential constraint. (6) The blend ab must meet all relevant phonotactic requirements. Smoke + drink cannot give /*sməʊŋk/ This example is from Kuzobono (1990), and it is not clear how accurate it really is. English does have the occasional word like boink with a diphthong before /ŋk/, but such words are all onomatopoeic. However, whether the particular example holds or not, the general principle is clear. This particular example – to the extent that it can be said to hold – shows a breach of another constraint, namely the preference for a break between syllables or between

16 Laurie Bauer onset and rhyme (see below, (7)). There is a more general rule in operation here, namely that few phonotactic constraints hold between onset and rhyme, while many constraints hold within each of these constituents. In the case of the example given in (6), we would, for other reasons, expect to find /smɪŋk/, which is phonotactically unexceptionable. In more general terms, though, a potential blend of stutter and quaver as *stwaver would presumably be ruled out by phonotactic constraints. There is a question as to how much of the first word is accepted into the blend, and how much of the second. We can call the point at which the switch from one base word to another occurs as the “crossover point”. There are some constraints which act to define this crossover point in blends. (7) has already been mentioned above. (7) In a blend ab from words ag and db, where ab is a monosyllable, a is a syllable onset and b is a syllable rhyme. Examples: br(eakfast)(l)unch, sm(oke)(f)og, d(ove)(h)awk But: (no counter-examples??) Bat-El (2006) gives some possible counter-examples like slosh < slop + slush, boost < boom + hoist, but these etymologies are not generally agreed ones, and her example of moorth < moon + earth is not easily confirmed (it is not findable through Google, for example). (8) In a blend agb from words ag and gb, where g is a phoneme/letter or phoneme/letter string in common between the two base words, this overlap defines the crossover point. Examples: Chicago(go)rilla (letters in common), gu(ess)estimate (phonemes in common) But: h(orse)(z)ebra, tig(er)(li)on, cam(era)(re)corder where different phoneme or letter strings are deleted from . each of the two words In other words, there is phonological or orthographic overlap between the elements in a blend. There may be both, there needs only to be one. Note that the overlap is orthographic as often as phonological, so that the overlap in diabesity may be the . Note also the example glasphalt, which is based on something which is often /ˈæʃfɔːlt/ despite the spelling, and the spelling wins out.

Blends: Core and periphery 17

(9) In a blend ab from the words ag and db where g and d share no phonemic/orthographic material, the break between a and b will fall at a syllable break or, failing that, at an onset/rhyme break. Example: posi(tively)(abso)lutely But: tig(er)(li)on Occasionally, we find infixed blends. An example from Lewis Carroll is slithy < slimy + lithe, but this construction is rare in English beyond Lewis Carroll. Some examples are found in French, though:

S’embellemerder < s’emmerder ‘be bored’ + belle-mère ‘mother-in-law’ Ridicoculiser < ridiculiser ‘to ridicule’ + cocu ‘cuckold’

The rarity of such examples may be due to no more than the rarity of infixes, that is the processing problems caused by discontinuity. 2.4. Semantics Renner (2006) gives four main semantic relationships between elements in coordinate blends, with a few exceptional cases. These are dealt with here in order of frequency, following Renner, and then the semantics of determinative blends is briefly considered. Because Renner includes clipping compounds in his data, the figures may not be reliable if you reject the notion that clipping compounds and blends form part of the same category. 2.4.1. Hybrid The first meaning is where the blend denotes a hybrid of the things denoted by its elements. Examples are: Tig(er)(li)on Spa(nish)(E)nglish Edu(cation)(enter)tainment English in particular likes this meaning of blends, and uses a blend to create this meaning. Although it seems at first glance that the two elements here are on an equal footing, there may actually be some semantic headedness in items like tigon. For some a tigon is the offspring of a lioness, while a liger is the offspring of a tigress. The two animals are, apparently, significantly different. A similar distinction may underlie the difference between hebra and zorse.

18 Laurie Bauer For those who make the distinction, the order may be predetermined by the meaning, the compounds may be right-headed, and these may not be like co-compounds at all, but be determinative. Consider the range of mock language names such as Frenglish, Hinglish, Japlish, Russlish, Singlish, Spanglish and so on. There is some disagreement in the literature on whether these are types of English or relexified versions of the other language but nobody seems to suggest that they are 50-50 hybrids with neither taking priority. So while there might be discussion about which element is the head, it seems probable that (at least most of the time) these are headed blends. While there is clearly a problem of interpretation involved here, it is not necessarily any different from the problem of interpretation involved in so-called coordinative N+N compounds such as gardener-handyman or fighter-bomber. 2.4.2. Addition This category is supposed to be made up of the instances where the blend denotes the addition of the two parts. Examples are: Seman(tics)(syn)tax Sk(irt)(l)eggings Stag(nation)(in)flation 2.4.3. Polyvalence This category is supposedly made up of words where the blend denotes something which has features of both elements, so that this category is the blend equivalent of the co-compound forms like singer-songwriter. Examples are: Sp(oon)(f)ork Cafet(eria)(audit)orium Trans(mitter)(re)ceiver The difference between having features which come from the two elements and being a hybrid of the two elements is not always clear, and it seems to me that these two categories could probably be coalesced. Whether the addition category could also be added into the mix is perhaps less clear, but the difference between saying that semantax is all of semantics and all of syntax and saying that it has both syntactic and semantic features is sometimes moot.

Blends: Core and periphery 19

2.4.4. Tautologous Finally in Renner’s categorization, we have instances of the coordination of synonyms, as in: Ruc(tion)(rump)us = ruckus W(imp)(p)uss(y) Both these etymologies are uncertain, but the example of posilutely, used earlier, would also fit in here. The category is much more numerous in other languages (see Wälchli 2005: 143–146 on synonymic co-compounds), and probably also in speech-error blends. 2.4.5. Determinative blends The first four semantic categories dealt with here have all belonged to co­ ordinate blends. Determinative blends have a semantic structure more similar to endocentric compounds, with modification of one element by another. Examples are: Cre(mate)(re)mains Mo(tor)(Ø)camp 3.

What is the limit of blending?

The trouble with prototypical categories is that as you move away from the prototype you cannot tell whether you are now in another area of description or only in a less prototypical part of the same domain. Accordingly, we can look at a number of different formations, and still not be sure whether they really count as blends or not. This, as I see it, is one of the major problems facing the study of blends at the moment: without some way of circumscribing the object of description, we have no way of guaranteeing that our description is an accurate one. Let me take some examples, some of which readers will consider to be blends, and some of which they will not consider to be blends, and in each case I will ask, in effect, whether these could be seen as more or less prototypical blends. If we accept them all, we have a real problem of description; if we do not accept them all, we need to know why these fall outside the blending category.

20 Laurie Bauer – Mot(or)(ho)tel: as has been said, this is determinative and some exclude determinative instances from the category of blends by definition (Dressler 2000). – Posi(tively)(abso)lutely: formally, this looks like a typical blend, with some­ thing deleted from both; yet the tautologous blends are a minor category semantically in English, so this is not completely prototypical, though probably uncontroversially a blend. – Gu(ess)(Ø)estimate: this is a typical blend, with some overlap of the elements determining the crossover point. Again, it is probably best classified as tautologous. – Key(Ø)(con)tainer: here we have an entire word combined with an element­ which looks typical of those found in blends. The next example provides a similar example, but with the entire element in a different position. It is tempting simply to exclude these by definition, but in cases like guestimate (above) it seems that we must take one of the elements as being phonologically complete, but it is impossible to know which. To exclude one would probably mean excluding the other. – Identi(ty)(Ø)kit – Pseudo(Ø)(sui)cide: Here we appear to have a case of prefixation, but the base is clipped. Pseudo- is not a potentially free morph, and so requires a base, but pseudo-suicide would have provided a potential base. In this case, though, we might not know whether to treat it as prefixation or as a case of neo-classical word-formation. With pseudo-cide it is not clear whether it is prefixation or blending. – E(lectronic)(Ø)mail: the form here fits the definition of a clipping compound or of a blend, but the pronunciation is odd for a blend since it seems to be the pronunciation for the letter. – Electrophilic: this will generally be taken as an example of a neo-classical compound, but it might, after all, be considered as a blend of, say, electron and nucleophilic. – Breatharian: this is ostensibly an instance of suffixation, but again, it might be possible to re-interpret it as a blend of, say, breath + (veget) arian or breath + (disciplin)arian. – Sci(ence)Fi(ction): this is a prototypical clipping compound, excluded by definition by some, e.g. Bauer and Huddleston (2002), but allowed by Renner (2006). There is a real question as to whether there is an independent category “blend” or whether there is only a category of “phonetically reduced compound”. Gries (this volume) suggests that there are, indeed, two categories.

Blends: Core and periphery 21

– Arm candy: this is a relatively recent compound meaning “an attractive woman being escorted by an (older) man”, but we need to ask what makes it such rather than a blend of arm + (eye) candy or arm (rest) + (eye) candy. (I would suggest that few would deny the influence of eye candy on the formation of arm candy.) – Mole(cule)ch(emical)(organ)ism: finally I include an example which is about as confused as it can get. This is on the border of word manufacture (example and comment from Bauer 1983). 4. Conclusion We know a lot about blending but paradoxically this leads us to a position where we have a fuzzy category of blends. Given that the category is a fuzzy one, it is not necessarily clear why more types should not belong to it. If the term blend is to be meaningful, we need to restrict it more and make sure we know just what it is we are trying to describe. It seems to me that virtually everything I have said provides a hypothesis which requires further testing. But I do not believe that just continuing along the same lines will lead us to a good outcome. We need more than new experiments on an ill-defined set of words. We also need a flash of insight, which will allow us to capture the essence of blending and separate it out from everything else (if, indeed, that is the appropriate solution); blends are still a descriptive problem. The papers collected in this book may not get us that far. But they certainly expand the data-base, and lead us to consider the advantages of different theoretical lenses when considering the problems that blends pose to the analyst. They also open up the question of the ways in which blends are used in different languages, both in terms of their semantics and in terms of the pragmatics of their use – especially in languages where blending is a relatively new process, not yet fully integrated into the system of word-formation. They confirm the relevance of blending to the morphologist-linguist, but also the fascination of this way of forming new lexical items. References Bat-El, Outi. 2000. The grammaticality of “extragrammatical” morphology. In Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 61–81. Munich: Lincom Europa. Bat-El, Outi. 2006. Blend. In Keith Brown (ed.), Encyclopedia of Language and Linguistics, Vol. 2, 66–70. Oxford: Pergamon.

22 Laurie Bauer Bauer, Laurie. 1983. English Word-formation. Cambridge: Cambridge University Press. Bauer, Laurie and Rodney Huddleston. 2002. Lexical word-formation. In Rodney Huddleston and Geoffrey K. Pullum (eds.), The Cambridge Grammar of the English Language, 1621–1721. Cambridge: Cambridge University Press. Dressler, Wolfgang U. 2000. Extragrammatical vs. marginal morphology. In Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 1–10. Munich: Lincom Europa. Fradin, Bernard, Fabio Montermini and Marc Plénat. 2009. Morphologie grammaticale et extragrammaticale. In Bernard Fradin, Françoise Kerleroux and Marc Plénat (eds.), Aperçus de morphologie du français, ­21–45. Saint-Denis: Presses Universitaires de Vincennes. Gries, Stefan Th. 2004. Shouldn’t it be breakfunch? Linguistics 42 (3): 639–667. Kelly, Michael. 1998. To “brunch” or to “brench”: Aspects of blend structure. Linguistics 36 (3): 579–590. Kubozono, Haruo. 1990. Phonological constraints on blending in English as a case for phonology-morphology interface. Yearbook of Morphology 3: 1–20. Lalić-Krstin, Gordana and Sabina Halupka-Rešetar. 2010. The order of constituents in Serbian blends: some experimental data. Paper presented at ICLB, Lyon, June 2010. Malkiel, Yakov. 1959. Studies in irreversible nominals. Lingua 8: 113–160. Renner, Vincent. 2006. Les composés coordinatifs en anglais contemporain. Lyon: Université Lumière Lyon 2 dissertation. Vennemann, Theo. 1988. Preference Laws for Syllable Structure. Berlin & New York: Mouton de Gruyter. Wälchli, Berhard. 2005. Natural Coordination and Co-compounds. Oxford: Oxford University Press.

Beyond all reasonable transgression: Lexical blending in alternative music Paula López Rúa This paper approaches the issue of blend formation by focusing on a subcorpus of blends naming bands and singers belonging to alternative musical styles. Therefore, the study puts forward an analysis of the main trends in blend formation in this subfield from the point of view of form, underlying structure and purpose. In general terms, items formed out of two bases (­Medievil) are the most common, and so are combinations that keep the initial source word intact, or even both source words (Obscenery). Likewise, there is a tendency towards high constituent integration (Apocalipstick), and the sources seem to favour an underlying grammatical relationship of modification (Sexorcist). Finally, the analysis shows that transgressive semantic combinations are used to reinforce the images of violence, sacrilege or chaos which define a band’s ideology (Atheretic, Assassinner). In conclusion, the study reveals that lexical blending is a powerful aesthetic device to catch the audience’s attention by striking the eye and the ear. However, it also arises as a handy tool to satisfy intellectual and social needs, since the items thus formed convey valuable information about the artists’ views on life, society or their art, and contribute to representing the subculture those artists belong to, thus reinforcing group bonds. Keywords: blending, morphology, music, splinter, transgression. 1. Introduction: blends and blending Blending is a very productive source of words in present-day English, although, as Bauer (1983: 26) remarks, the limits of the category are not easy to draw, since “blending tends to shade off into compounding, neo-classical compounding, affixation, clipping, and […] acronyming”. Although there is indeed a great deal of variation as regards the definition of blends, it is generally acknowledged, on the one hand, that the process of blending involves more than one base word (usually two, and occasionally three), which explains why authors such as Lehrer (2007) describe blends as “underlying compounds”. On the other hand, blending also implies the

24 Paula López Rúa loss of material (i.e. shortening) from at least one of the source words. Thus Algeo (1991), for instance, points out that blends are obtained by joining “two or more forms” and that at least part of one must be shortened. Lehrer (2007) calls this “word part” a “splinter”, a term which will also be used in this paper and which is described as a clipping which “cannot occur alone as a word” (2007: 116), for example dram- or -edy in dramedy (‘drama’ + ‘comedy’). As regards the process of blending, a simple but useful description of what happens when the source words are “joined” is provided by Kelly (1998: 579), who defines these items as words “formed by snipping components from existing words and stitching the components together either through simple concatenation or through concatenation coupled with overlap of shared phonological segments”. In this respect, it must be noted that, as Plag (2003) remarks, blending is a type of word-formation process where phonology and morphology interact. Therefore, it should be accounted for (together with “truncation”, i.e. clipping) from the perspective of prosodic morphology. On these grounds, and in agreement with Lehrer’s (2007) remark that the most frequent domains for blends are the media and commercial sectors, this paper approaches the issue of blend formation by turning to a sphere of human interaction clearly marked by creativity and freedom, namely the field of music, and more specifically, a miscellaneous collection of musical trends which can be encompassed by the term alternative music. Alternative music is often associated with small independent companies, and alternative musicians are usually characterized by their will to keep away from commercial pop music and mainstream culture. As will be shown in the analysis, through their choice of names artists ascribed to alternative musical trends (indie rock and pop, alternative metal – doom metal, death metal –, anarcho-punk, gothic rock, geek rock, grunge, etc.) turn the language into a suitable instrument to convey their desire to go against the prevailing trends, thus being different and authentic. In other words, they use language to break away from the norm and also from what is foreseeable or even politically correct. As their names are part of their image, in most cases bands and singers try to come up with denominations which define their music, or convey their ideas about politics, society, sex, religion, or their immediate context. In many cases the names are purposely devised to catch the audience’s attention by appealing to the senses; in other words, they try to strike the eye and the ear by resorting to all types of linguistic deviation: typographical, phonological and morphological (López Rúa 2010). As will be shown in the following section, this paper reports on the main trends in blend formation found in this subfield by considering the

Beyond all reasonable transgression: Lexical blending in alternative music 25

parameters of form, underlying structure (both grammatical and semantic) and motivation. 2.

The analysis of the corpus

The corpus gathered comprises 190 names of bands and singers. For the most part, these names belong to current or recently disbanded groups. The corpus was gathered from on-line sources (see References). The items provided by these sources were carefully examined, and those illustrating the process under study were selected for further analysis and classification. When necessary, etymological data were obtained from the Oxford English Dictionary, the Collins English Dictionary and the Merriam-Webster’s Online Dictionary. As already mentioned, the analysis of the items was carried out by considering their form and structure and the possible reasons for their creation. The results of the analysis are summarized in the subsections that follow. 2.1. The parameter of form When analysing the form of the items, two criteria were taken into account: the number and type of constituents, and their degree of integration. Both criteria have been adapted from López Rúa’s (2004) study on the category of English blends. 2.1.1. Number and type of constituents This parameter considers both the number of source forms and the type of morphosyntactic units which are shortened. Therefore, the corpus registers three-constituent items (Doominhated: ‘doom’ + ‘dominated’ + ‘hated’; Stormental: ‘storm’ + ‘torment’ + ‘mental’) and two-constituent items, the latter being far more frequent than the former. On the other hand, the following combinations of morphosyntactic units were found to operate in items with two sources: a) full word + splinter: for example, Bloodphemy ‘blood’ + ‘blasphemy’; Sintury ‘sin’ + ‘century’; Skillzophrenic ‘skill’ + ‘schizophrenic’; Soulitude ‘soul’ + ‘solitude’; or Vaultage ‘vault’ + ‘voltage’. b) splinter + full word: as in Apocalipstick ‘Apocalyptic’ / ‘Apocalypse’ + ‘lipstick’; Euthanausea ‘euthanasia’ + ‘nausea’; Medievil ‘Medieval’ + ‘evil’; and Silength ‘silence ‘ + ‘length’.

26 Paula López Rúa c) splinter + splinter (two splinters with different degrees of orthographic or phonic fusion): for example, Blore ‘blood’ + ‘gore’; Enertia ‘energy’ + ‘inertia’; or Musilosophy ‘music’ + ‘philosophy’. d) word + word (two words overlapping, both being identifiable in the item): for instance, Goreality ‘gore’ + ‘reality’; Humanimal ‘human’ + ‘animal’; Morbidoom ‘morbid’ + ‘doom’; and Sexorcist ‘sex’ + ‘exorcist’. e) word + embedded word or embedded splinter (a particular type of blending where the word which is embedded – in boldtype – shares some material with the recipient word): for example, Assoult ‘assault’ + ‘soul’; Insatanity ‘insanity’ + ‘Satan’; Insicknia ‘insignia’ + ‘sick’; Messendger ‘messenger’ + ‘send’; or Mushitcians ‘musicians’ + ‘shit’. f) word/splinter + other types of constituents: for instance, UFOetry ‘UFO’ (an acronym) + ‘poetry’; Ill Niño ‘ill’ + ‘El Niño’ (a phrase); St Lunatics ‘St Louis’ (a phrase) + ‘Lunatics’; or Destroys’r’us ‘Destroy’ + ‘Toys’r’us’ (a clause that functions as a trademark). g) atypical combinations and items including combining forms. One instance of an atypical two-constituent combination is RJD2, which joins two initialisms (combinations of initials read out one by one): the alphanumeric item ‘R2D2’ (one of the famous Star Wars robots) and ‘RJ’ (‘Ramble John Krohn’, the singer’s name). As defined by Lehrer (2007: 124), combining forms are “bound bases that combine with full words or with other combining forms”, the most typical being “those appearing in neo-classical compounds”. Neoclassical combining forms spotted in the corpus are, for instance, bio- (‘life’) in Biolence: ‘bio-’ + ‘violence’; -(ec)tomy (‘excision, surgical removal’) in Embryectomy (‘embryo removal’); -naut (‘navigator’) in Unfortunaut: ‘unfortunate’ + ‘-naut’; or necro- (‘dead, death’), from Greek necros (‘corpse’), in Necronoclast (‘necro-’ + ‘iconoclast’). The corpus also includes less typical bound morphemes from Greek and Latin which are recovered and used recurrently. Since in these cases we come across blends created by analogy with a reference item (euthanasia, autopsy), the resulting forms could be alternatively analysed as ­combinations of word + word, or word + bound morpheme, in which one of the constituents is a recognizable, though unusual, combining form. A few instances of less typical bound morphemes are the following: -caust, from holocaust < Greek holo- (‘whole’) + kaustos (‘burn’), in Atomicaust (‘atomic’ + ‘holocaust’), Iconocaust (‘iconoclast’ + ‘holocaust’), or Genocaust (‘genocide’ + ‘holocaust’); -thanasia, from euthanasia < Greek eu- (‘easy’) + tanathos (‘death’), in Christhanasia or Youthanasia; and -topsy, from autopsy < Greek auto- (‘self’) + -opsy (‘sight’), in Foetopsy,

Beyond all reasonable transgression: Lexical blending in alternative music 27

Cryptopsy, Cuntopsy, or Vomitopsy. Lastly, -holic (roughly, ‘addict’) is a splinter which can be categorized as a new “combining form” (Lehrer 2007: 124), which was found to occur in items like Mytholic (‘myth’ + ‘-holic’), although it is the item which originated the splinter (‘alcoholic’) that occurs as one of the sources in the blend Diskoholics (‘disk’ + ‘alcoholics’) Anonymous. 2.1.2. Degree of integration of constituents Adams (1973: 150) uses the term haplology to talk about “the overlapping of vowels, consonants or syllables” which, in her view, characterizes most blends. Three main stages in the degree of integration can be distinguished: a) Low integration (that is, clustering of splinters or combining forms from the sources). There is no overlap of common material, for instance, in Con-Dom (‘control’ + ‘domination’, and also a play on ‘condom’), a combination of splinter + splinter; Sunpocrisy (‘sun’ + ‘hypocrisy’), a word + a splinter; or Vomicaust (‘vomit’ + ‘(holo)caust’), a splinter + a combining form (see 2.1.1.g. above). From the point of view of constituent integration, this can be regarded as a marginal type of blends on the borderline with compounds and clippings, a position evinced, for example, by the similarity between a compound like Alsace-Lorraine and a clipped compound – an abbreviated form of a compound – such as Alcan (‘Alaska’ + ‘Canada’). Those items are described by Adams (1973: 137) as “compounds of clipped elements” that nevertheless fall “within the range of blends” (for instance, biopic: ‘biographical’ + ‘picture’). Bauer (1983: 233) calls them “compounds made of clippings” (for example sci-fi: ‘science’ + ‘fiction’, or sitcom: ‘situation’ + ‘comedy’) and points out that sometimes it is difficult to know whether the item “should be treated as a clipping or as a blend”. Although a discussion on the exact location of these items would fall outside the scope of the present study, it should at least be mentioned that, in order to generate an overall measure of typicality for the blend, this parameter should be combined with others such as simple or compound stress (suggested by Bauer 1983), spelling, degree of shortening of the sources, underlying grammatical structure, semantic integration, or the ex­­­ tent to which one source conforms to the shape of the other (in Sunpocrisy, the shape of ‘hypocrisy’ is generally maintained, since sun- replaces hy-). The number of items of this type in the corpus is quite low. There seems to be a preference for some degree of overlap (see below), for

28 Paula López Rúa witty malapropisms, such as Prole Position (‘pole position’), Malice in Wonderland (‘Alice in Wonderland’), and Goatbless (‘God bless’), or for the straight clustering of complete words or even combining forms, which occasionally implies word play with a third word, as in the following examples (which fall outside the scope of the study since there is no shortening involved): DamnNation (‘damn’ + ‘nation’, and at the same time ‘damnation’), SinKing (‘sin’ + ‘king’, and also ‘sinking’), or Soundshine (‘sound’ + ‘shine’, and also ‘sunshine’). b) Medium integration. The process involves overlap of graphemes, phonemes or just phonic features, and the more material overlapping at both levels (phonic and graphic), the higher the degree of integration. The shared material concentrates in the borderline between the sources, which is the central part of the resulting blend. The following are examples of increasing constituent integration, with shared graphic material in boldface: from lower medium (Angstorm: ‘angst’ + ‘storm’; Deepain: ‘deep’ + ‘pain’; Gonorrhoeaction: ‘gonorrhoea’ + ‘reaction’; Gorgasm: ‘gore’ + ‘orgasm’; Sicklown: ‘sick’ + ‘clown’; Stigmatheist: ‘stigma’ + ‘atheist’; Stormageddon: ‘storm’ + ‘armageddon’) to higher medium (Decrepitaph ‘decrepit’ + ‘epitaph’; Funerealm ‘funereal’ + ‘realm’; Nocternity: ‘nocturnal’ + ‘eternity’; Obscenery: ‘obscene’ + ‘scenery’; Trespassenger: ‘trespass’ + ‘passenger’; Twilightning ‘twilight’ + ‘lightning’). As can be noticed, the examples show the overlap of only graphic material (as in Stigmatheist, where final /ə/ and initial /eɪ/ do not coincide), and the overlap of both phonic and graphic material in different degrees: for example, Obscenery, where the items totally share the spelling and the pronunciation of the overlapping element (/siːn/); Nocternity, where phonic overlap (/tɜːn/) is higher than graphic overlap; or Stormageddon, where the overlap of graphic material is paired with the phonic overlap of a consonant sound (/m/) preceded by an overlap of phonic features, since the vowel sounds of the two sources coincide in length and backness. c) High integration (fusion). In cases of high integration, one of the constituents is directly mapped onto the other, and the recipient base determines the overall shape and size of the item. The pronunciation of the whole item is thus similar or even identical to that of one of the constituents. The high degree of phonic integration of these items, since one of them is practically assimilated to the other, contributes to their overall degree of integration. The following are a few cases of relatively high integration, in which it can be noticed that the pronunciation of the whole item is similar, though not identical, to that of one of the constituents. The recipient base is in boldface, and once again, the more material they share,

Beyond all reasonable transgression: Lexical blending in alternative music 29

the higher the degree of integration: Axegressor ‘axe’ + ‘aggressor’; Blastard: ‘blast’ + ‘bastard’; Deatholation: ‘death’ + ‘desolation’; Deifecation: ‘deification’ + ‘defecation’; Ethernity: ‘ether’ + ‘eternity’; Heavylution: ‘heavy’ + ‘evolution’; Pantommind: ‘pantomime’ + ‘mind’; Sintury: ‘sin’ + ‘century’; Skillzophrenics: ‘skill’ + ‘schizophrenics’; or The Beatscuits: ‘beat’ + ‘biscuits’. Within this subtype it is worth mentioning the existence of several examples of wordplay based on the possibilities of reanalysis of the items, for instance, Transfear: ‘transfer’ + ‘fear’, but also trans- (a prefix meaning ‘across, beyond’) + ‘fear’; or The Foxymorons: ‘foxy’ + ‘oxymorons’ (an oxymoron is an epigrammatic effect by which contradictory terms are used in conjunction, so note the possible reanalysis as ‘foxy’ (‘cunning’) + ‘morons’ (‘stupid people’), which is itself an oxymoron). A further development from this subtype is represented by those cases in which the pronunciation of the item is similar to that of a third item, which also contributes to wordplay and may reinforce the semantic connotations of the blend. For example, Suffereign (‘suffer’ + ‘reign’), whose pronunciation is similar to ‘sovereign’; or Disseized (‘disease’ + ‘seized’), whose pronunciation is similar to ‘deceased’. In the latter example it should be noted that, despite the phonic integration of the constituents, spelling (the double ) draws attention to the different sources that compose the blend. In this respect, and as noted by Lehrer (2007: 120), typographic devices are sometimes used “to call attention to part of the blend” and to the sources of that blend. Other examples from the corpus that illustrate this point are, for instance, Deep-pression (‘deep’ + ‘pression’, whose joint pronunciation is similar to ‘depression’), or SinTuition (‘sin’ + ‘intuition’, and also, using the clue of spelling, ‘sin tuition’). One step forward in the degree of constituent integration is illustrated by “orthographic blends”, because, as pointed out by Lehrer (2007: 120), most blends “are identical in speech and writing”. However, in orthographic blends the degree of phonic integration of the constituents is so high that the blend can only be appreciated when seen in writing. The following are a few examples from the corpus: Abwhore (‘abhor’ + ‘whore’); Evilution (‘evil’ + ‘evolution’); Frontears (‘frontiers’ + ‘tears’); Knightmare (‘knight’ + ‘nightmare’); Ricanstruction (‘Puerto Rican’ + ‘reconstruction’); Skullptor (‘skull’ + ‘sculptor’); The Amplifires (‘amplifiers’ + ‘fires’); Whorizon (‘whore’ + ‘horizon’); and Wonderbrass (‘Wonderbras’, plural trademark + ‘brass’). As in the previous subtype, there are cases in which the pronunciation of the item is identical to that of a third item, as in Reigncarnation (‘reign’ + ‘incarnation’, pronounced

30 Paula López Rúa as ‘reincarnation’), and also cases where spelling draws attention to the components of the blend (for instance, SIN.thetic: ‘sin’ + ‘synthetic’) 2.2. The parameter of the underlying structure 2.2.1. Grammatical structure As regards the lexical category the items belong to, for the most part, the blends of the corpus are nouns composed of other nouns (Sindustry is a noun from ‘sin’ + ‘industry’), which is logical if we consider that the items are devised to function as names of bands and therefore to designate. There are, however, a few items which, in spite of their obvious nominal function, seem to be primarily adjectives (Loadead: ‘loaded’ + ‘dead’), or verbs (Recrehate ‘recreate’ + ‘hate’). Such items follow the creative tendency to use atypical lexical or syntactic categories as names of bands (consider, for example, Better Than Ezra, Die Happy, or INXS). The parameter of grammatical structure takes into consideration the lexical categories of the items functioning as sources and the relationship established by those items (coordination or subordination). Two main subtypes are identified: a) noun + noun or adjective + noun forming a “syntactic compound” (Lehrer 2007). In the expanded form of these items, one element modifies the other. For instance, Breakestra is “a kind of orchestra”. Blends with a modifying element were found to be more frequent than those where both elements are at the same level. Combinations of noun + noun are, for instance, Breakestra (‘break’, a style of music + ‘orchestra’); Sexorcist (‘sex’ + ‘exorcist’); Sindustry (‘sin’ + ‘industry’); or Warchetype (‘war’ + ‘archetype’). Combinations of adjective + noun or noun + adjective occur, for example, in Hollowgram (‘hollow’ + ‘hologram’); Infinight (‘infinite’ + ‘night’); or Soulitary (‘soul’ + ‘solitary’). b) noun + noun or adjective + adjective forming a “coordinate compound” or “copulative compound” (Plag 2003), in which both elements are at the same level, and the entity denoted by the blend shares properties of the referents of both elements. For example, Flametal refers to the type of music played by the band, which combines both ‘flamenco’ and ‘metal’. Other examples of noun + noun combinations are Melodisaster (‘melody’ + ‘disaster’), Musilosophy (‘music’ + ‘philosophy’), or Sinocence (‘sin’ + ‘inno­cence’). Combinations of adjective + adjective are, for instance, Blindead (‘blind’ + ‘dead’), Darkane (‘dark’ + ‘arcane’), Loadead (‘loaded’ + ‘dead’), Medievil (‘Medieval’ + ‘evil’), or Sarkaustic (‘sarcastic’ + ‘caustic’).

Beyond all reasonable transgression: Lexical blending in alternative music 31

c) other constituents. As already mentioned, the corpus also registers a few cases which include other constituents as sources, for instance verbs or adverbs: Attackill (‘attack’ + ‘kill’), Incinerhate (‘incinerate’ + ‘hate’), Suckceed (‘suck’ + ‘succeed’), Insidead (the adverb ‘inside’ + the adjective ‘dead’), etc. These items were found to illustrate both coordination and subordination relationships. 2.2.2. Semantic structure and purpose These parameters will be discussed together, since the semantic structure was found to be related to the creative purpose. In most of the items analysed, there is an underlying modifier + head or head + modifier structure, and a shocking combination of items which conveys a message of rebelliousness against the rules. As already mentioned, the rules of language are broken to show an anti-establishment spirit, and names become illustrations of a philosophy of life. In general, the names tend to revel in recalling unpleasant images of evil, destruction, violence, sacrilege or obscenity (Cuntopsy, Hellixxir, Ludichrist, Warchitect, Vomitrocity), sometimes including references to a particular style of music that the band represents (Gorevenge, Metalmorphosis, Thrashfusion, Superfunktion). Besides this general trend, three subtypes of items are worth commenting on from the point of view of the effects of the formal combination of items on semantic structure. a) items where there is a formal combination of bases that complement each other: the result is a complementary effect. What is intended in these cases is simply a combination of meanings, as in Flametal (‘flamenco’ + ‘metal’); Sentimetal (‘sentimental’ + ‘metal’), or Stratovarius (‘Stratocaster’ + ‘Stradivarius’: the band’s style is a combination of heavy metal and classical music). b) items with a formal combination of redundant or synonymous bases: the result is a hyperbolic effect. The intended effect is the reinforcement of images that define the band’s ideology: negativity, sacrilege, madness, death, sickness, violence, chaos or destruction; for example, Assassinner (‘assassin’ + ‘sinner’); Atheretic (‘Atheist’ + ‘heretic’); Massacaust (‘massacre’ + ‘holocaust’); Revengeance (‘revenge’ + ‘vengeance’), or Warmageddon (‘war’ + ‘armageddon’). c) items with a formal combination of opposite or contradictory bases: the result is an antithetical effect. Such combinations contribute to making clear the band’s anti-establishment attitudes. In other words, the wish to go against the system is reflected in the breach of the semantic rules

32 Paula López Rúa of language. A few instances are Enertia (‘energy’ + ‘inertia’); Lullacry (‘lullaby’ + ‘cry’); Mercynary (‘mercy’ + ‘mercenary’); Sinocence (‘sin’ + ‘innocence’), or Wreckreation (‘wreck’ + ‘creation’; note also the same pronunciation as ‘recreation’). 3. Graphic summary The tables below display the percentages of items with respect to the parameters of form (number and type of constituents and degree of integration) and grammatical structure. Shaded cells indicate the highest percentages. Table 1. Distribution of items according to number and type of constituents No. of constituents Type of constituents

Example

3 2 2 2 2 2 2 2

Stormental Skillzophrenic Silength Blore Insidead Insatanity Necronoclast UFOetry

word + splinter splinter + word splinter + splinter word + word word + embedded word items with combining forms word/splinter + other constituents

Total

Percentage 4.7 % 28.4 % 12.6 % 7.4 % 21.1 % 8.9 % 11.6 % 5.3 % 100 %

Table 2. Distribution of items according to degree of integration Degree of integration

Example

Low Medium High

Sunpocrisy  Angstorm, Decrepitaph Thrashfusion

Total

Percentage 7.4 % 40.5 % 52.1 % 100 %

Table 3. Distribution of items according to grammatical structure Structure

Example

Percentage

N+N / A+N (syntactic compound) N+N / A+A (coordinate compound)

Sexorcist Flametal

67.4 % 32.6 %

Total

100 %

Beyond all reasonable transgression: Lexical blending in alternative music 33

In general terms, it can be observed that items formed out of two bases clearly outnumber those with three sources. Besides, combinations that keep one source word at the beginning of the item or both source words intact are more frequent than combinations with splinters. There is also a tendency towards high integration of the constituents. Lastly, an underlying grammatical structure where one of the sources modifies the other is more frequent than a relationship of coordination. 4. Conclusion Alternative artists share a number of cultural and philosophical traits rather than a specific musical style. Together with their music, their lyrics, image and names show their wish to go against the mainstream and be different. This study evinces that language is a valuable instrument to shape this spirit of independence: language users are thus allowed to be different by playing with the rules and occasionally breaking them. Furthermore, the analysis carried out in this area of human interaction seems to confirm that blending as a word-formation process goes beyond a combination of forms that simply designates a combination of meanings (as in tangelo, which names a hybrid of ‘tangerine’ + ‘pomelo’). In fact, the device succeeds in satisfying not only pragmatic needs (the practical need to name a new entity), but also aesthetic needs (as language is an endless field for play and an outlet for human wit and imagination), ideological needs (as those new items can say something about the artists’ views on life, music, society, culture, or their art), and social needs (because the names contribute to representing the subculture those artists belong to and thus reinforce the bonds between their members). Acknowledgements The research reported in this paper has been funded by the Galician Ministry of Innovation and Industry (INCITEG08PXIB204033 PRTT-206). This grant is hereby gratefully acknowledged.

34 Paula López Rúa References a) General references Adams, Valerie. 1973. An Introduction to Modern English Word-Formation. London: Longman. Algeo, John (ed.). 1991. Fifty Years among the New Words. Cambridge: Cambridge Uni­­versity Press. Bauer, Laurie. 1983. English Word Formation. Cambridge: Cambridge University Press. Butterfield, Jeremy (ed.). 2003. Collins English Dictionary (6th edn.). Glasgow: HarperCollins Publishers. Kelly, Michael H. 1998. To ‘brunch’ or to ‘brench’: some aspects of blend structure. Linguistics 36 (3): 579–590. Lehrer, Adrienne. 2007. Blendalicious. In Judith Munat (ed.), Lexical Creativity, Texts and Contexts, 115–133. Amsterdam & Philadelphia: John Benjamins. López Rúa, Paula. 2004. The categorial continuum of English blends. English Studies 85 (1): 63–76. López Rúa, Paula. 2010. Making and breaking the rules: lexical creativity in the alternative music scene. Language Awareness 19 (1): 51–67. Merriam-Webster’s Online Dictionary. http://www.merriam-webster.com (accessed February 4, 2010). Plag, Ingo. 2003. Word-Formation in English. Cambridge: Cambridge University Press. The Oxford English Dictionary (2nd edn.). 1989. Oxford: Clarendon Press.

b) References consulted to gather the corpus All Music Guide. http://www.allmusic.com (accessed April 4, 2012). Encyclopedia Metallum. The Metal Archives. http://www.metal-archives.com (accessed April 4, 2012). Indyrock Magazine. http://www.indyrock.es (accessed April 4, 2012). Mondo Sonoro Magazine. http://www.mondosonoro.com (accessed April 4, 2012). My Space. http://www.myspace.com/music (accessed April 4, 2012). Punk Music. http://www.hcpunk.com (accessed April 4, 2012).

Blend formation in Modern Greek Angela Ralli and George J. Xydopoulos In this paper we discuss lexical blending as a rather novel but developing word-formation process in Modern Greek. We understand blends as deliberate creations that follow the structure of [stem word] compounds. We claim that the formation of blends is subject to headedness restrictions and, unlike compounds, both constituents of blends are reduced, with the nonhead assuming a stronger form reduction. In our view, blend formation is situated at the boundary of linguistic competence and creativity: if blends share a certain structure with compounds, blending could be considered as part of the native speaker’s linguistic competence. However, blends differ from compounds in that speakers create them consciously and intentionally, for specific communicative purposes. Keywords: lexical blends, compounds, endocentricity, exocentricity, reduction, competence, creativity, subvarieties, marginal vocabularies. 1. Definition and main claims It is generally accepted that a blend is created from the structural fusion of two words. The byproduct of this fusion is the truncation of segmental material from the inner edges of the two constituents, or from only one of them. As claimed by Bat-El (1996), semantically, a blend delivers the concept of the two base words, and its meaning is contingent upon the semantic relation between them. Similarly, Hock and Joseph (1996) describe blending as a process delivering both a new signifiant and a new signifié. A blend resembles a compound, since both formations involve the combination of two (or more) lexemes. However, while the form of blends results from conscious thinking, as stated by Ronneberger-Sibold (2006: 157), a compound is generated effortlessly according to word-formation mechanisms. In this paper, following Dressler (2000) and Ronneberger-Sibold (2006), we accept that blending is subject to a general pragmatic constraint, which is imposed by the speakers’ willingness to form a blend with a certain degree of semantic transparency, depending on various communicative purposes. However, we claim that although Greek blends may be deliberate creations, their structure follows the structure of a certain type of Greek compounds,

36 Angela Ralli and George J. Xydopoulos namely that of [stem word] (Ralli 2007, 2009), where the first constituent is always a stem, and the second constituent a word. More specifically, we show that: a) blend formation is subject to headedness considerations, according to which the head is the righthand element, as is the case with Greek endocentric compounds; b) while compounds are subject to a morphologically constrained form reduction with respect to the first constituent, in blends, both constituents may be reduced, with the non-head assuming a stronger form reduction, and the semantically empty compound marker -o- never surfaces. Thus, we agree with Soudek (1978: 466) that even if Greek blends are not fully identical to compounds, they are basically built according to normal rules of word formation, in our case, Modern Greek compound formation. We finally propose that blend formation is situated at the boundary of linguistic competence and creativity: if blends share a certain structure with compounds, blending could be considered as part of the native speaker’s linguistic competence. However, as already said, blends differ from compounds in that speakers create them intentionally, for specific communicative purposes. Claims and proposals are exemplified on the basis of a corpus comprising ca. 200 blends, which were collected from dialectal and general dictionaries of Greek and, mostly, internet term bases and blogs (see e.g. www.slang.gr). 2. Blending vs. compounding Blends are one-word units, which may be compared to compounds with respect to the following properties. Both types of constructions are phonological words, that is they display only one stress, while each member has its own stress, when taken as an independent word: (1) a.

Compounds limnoθálasa < ‘lagoon’ psarósupa < ‘fish soup’

límn(i)1 θálasa ‘lake’ ‘sea’ psár(i) súpa ‘fish’ ‘soup’

1. Inflectional endings and other segments which do not surface in compounds and blends are put in brackets.

Blend formation in Modern Greek 37

b.

Blends aerajitó < ‘food served on airplanes’ aγnigenís < ‘a relative you have never met’

(f)ajitó aér(as)2 ‘air’ ‘food’ áγn(ostos) (s)igenís ‘unknown’ ‘relative’

According to Ralli (2007, 2009), the majority of Greek compounds are built on the basis of two structural patterns, [stem stem] and [stem word] (see (2a) below).3 Those which have the [stem stem] structure display a different inflectional ending from that of the second constituent, when taken in isolation, and are always stressed on the antepenultimate syllable. In contrast, compounds of the [stem word] pattern do not differ from the second constituent with respect to stress and inflection. Since blends also share the inflection and the stress with their second member, we suppose that their formation belongs to the [stem word] type: (2) a.

Compound [stem stem] [stem word] kuklóspito < kúkl(a) spít(i) domatosaláta < domát(a) saláta ‘doll house’ ‘doll’ ‘house’ ‘tomato salad’ ‘tomato’ ‘salad’

b. Blend [stem stem] n/a

[stem word] vláma < vlá(kas) (vlí)ma ‘very stupid’ ‘stupid’ ‘thick’ jiroíni < jír(os) (ir)oíni ‘kebab-addiction’ ‘kebab’ ‘heroine’

Blends and compounds combine the same grammatical categories. There are [noun noun] or [adjective noun] instances, [noun verb], [adverb verb] and 2. Greek masculine nouns in -as participate in word formation with a bare stem, that is without the ending -as. Since blends share with compounds a [stem word] structure (see below), aer- is considered to be the first constituent of the formation aerajitó. 3. A small number of Greek compounds may also be built according to two other patterns: [word stem] and [word word]. However, these patterns involve uninflected constituents at the left-hand side, such as adverbs. E.g. eksoxórafo ‘οutfield’ (< ékso ‘out’ xoráf(i) ‘field’), ksanamiló ‘talk again’ (< ksaná ‘again’ miló ‘talk’).

38 Angela Ralli and George J. Xydopoulos [verb verb] ones, as well as [adjective adjective] examples. Moreover, the constituent members of both categories bear the same functional relations, that is subordinative, attributive and coordinative (Bisetto and Scalise 2005). For an illustration, consider the following cases: (3) a. NN subordinative . Compound Blend

psarósupa < ‘fish soup’ aγapúδi < ‘love song’

b. NN coordinative . Compound jiδoprόvata < ‘goats and sheep’ Blend krópoli < ‘wax and propolis’ c. NN attributive . Compound vroxónero < ‘rain water’ Blend vermuδjáris < ‘man with hairy legs wearing bermuda shorts’

psár(i) súpa ‘fish’ ‘soup’ aγáp(i) (traγ)úδi ‘love’ ‘song’ jíδ(ja) prόvata ‘goats’ ‘sheep’ k(erí) (p)rópoli ‘wax’ ‘propolis’ vrox(í) nerό ‘rain’ ‘water’ verm(úδa) (ark)uδjáris ‘bermuda’ ‘bear’

d. AN attributive . Compound mavropínakas < ‘blackboard’ Blend vlaksitzís < ‘stupid taxi driver’

mávr(os) pínakas ‘black’ ‘board’ vlaks (taks)itzís ‘stupid’ ‘taxi driver’

e. AA coordinative . Compound ksinóγlikos < ‘sweet and sour’ Blend psidrós < ‘tall and fat’

ksin(όs) γlikόs ‘sour’ ‘sweet’ psi(lós) (xo)drós ‘tall’ ‘fat’

Blend formation in Modern Greek 39

f. NV subordinative . Compound afisokoló < ‘stick posters’, lit. ‘poster stick’ Blend sidirázo < ‘buy a newspaper to get the CD’ g. AdvV subordinative . Compound krifokitázo < ‘peek’ Blend ipulegízo < ‘approach in an insidious manner’ h. VV coordinative . Compound aniγoklíno < ‘open and close’ Blend vrexalízi < ‘it rains in small drops’

afís(a) kolό ‘poster’ ‘stick’. sidí (aγo)rázo ‘CD’ ‘buy’.

krif(á) kitázo ‘stealthily’ ‘look’ ípul(a) (pros)egízo ‘insidiously’ ‘approach’.

aníγ(o) klíno ‘open’ ‘close’ vréx(i) (psi)xalízi ‘it rains’ ‘it drizzles’

Subordinative and attributive blends are subject to rightward-headedness, exactly like compounds. In both categories, heads transmit their category and specific meaning to the mother nodes, as in the examples (3a,c,d,f,g) above. Headedness does not generally apply to coordinative structures (in both compounds and blends, see (3b,e,h)), since the two constituents are juxtaposed, and neither is responsible for the category and the basic meaning of the construction as a whole. Crucially though, and as opposed to compounding, where exocentric formations are productively built, Greek blending does not show any exocentric structures. In our view, this is due to two reasons. First, exocentricity induces semantic unpredictability in that there is a meaning which does not follow from that of the constituent parts. Since semantically a blend is also less transparent than a compound, as a result of its considerable form reduction, it would be impossible for the speakers to calculate the meaning of the blend as it would have become completely opaque.4 Second, as claimed by Ralli (2007) and Ralli and Andreou 4. Although this observation should have a cross-linguistic validity, according to Vincent Renner (p.c.) some exocentric blends are attested in English (e.g. ebonics, humiture).

40 Angela Ralli and George J. Xydopoulos (2010), in Greek exocentric compounds a derivational suffix is added to the composition of two lexemes, as shown in (4), which gives the compound its grammatical category and the specific meaning ‘which/who has the property of’: (4) anixtoxéris ‘generous’ tu anixtoxéri s tu anixtoxér i tu anixt- xér- ‘open’ ‘hand’

Exocentric formation I nflection Derivation Compounding

The presence of this suffix makes the final output of the compound a derived structure, something impossible with blends, which combine only two lexemes and are not subject to further word formation, i.e. derivation. On the basis of these remarks, it is legitimate to claim that blend formation in Greek starts as a type of [stem word] compound formation, but it further assumes a form reduction, which affects both constituents. Moreover, while the presence of a marker/linking element -o- is obligatory in compounds (5a), in blends, it never surfaces (see 5b):5 (5) a. Compound psil-ό-liγnos < psil(όs) liγnόs ‘tall and thin’ ‘tall’ ‘thin’ b. Blend psidrós < psi(lós) (xo)drós ‘tall and fat’ ‘tall’ ‘fat’ 3. Form reduction As stated above, blends are subject to a form reduction affecting both constituents. However, since compounds also seem to undergo a certain type of form reduction, we should make a distinction between reduction in blending and reduction in compounding. First, compounds do not have a word-internal inflectional ending. In other words, the first constituent of 5. See Ralli (2008) for details on compound marking.

Blend formation in Modern Greek 41

compound structures is generally a stem, that is, a word deprived of its inflectional ending. As opposed to compounds though, and as shown in (2) above, segment deletion in blends is not restricted to the inflection of the first constituent, but may affect both constituents and involve parts of their stems. Second, it has been shown by Ralli and Karasimos (2009) that a constraint, the so-called Bare-stem constraint, systematically forbids derivational suffixes to be overtly realized within compounds. According to this constraint, com­ pounds keep their internal structural cohesion by requiring the first stem constituents to be as bare as possible, i.e. without any overt suffixal material, as shown below: (6) a. kliδabaróno < kliδ-όn(o) abarόno ‘padlock’ ‘lock’ ‘bar’ b. krifokitázo < krif-á kitázo ‘peek’ ‘stealthily’ ‘look’ Thus, formations containing a stem and a word, where only a derivational suffix is absent from the first stem, are true compounds and not blends.6 With respect to the phonological reduction that blends are subject to, we assume a prosodic structure as put forward in the relevant literature, e.g. in Bat-El (1996) for Hebrew, Arvaniti (1998) for Greek, Kubozono (1990) for Japanese, Bertinetto (2001) for French, German and Italian, and Trommer and Zimmermann (this volume) for Spanish. In this line of thought, while the first constituent of Greek blends can be reduced to the point of keeping only the onset of the first syllable, the second constituent, i.e. the head, keeps the largest amount of material. Since blends, like compounds, are subject to right-headedness, we believe that maximization of the size of the head makes its identification easier, and facilitates the semantic recoverability of the formation. The crucial role of the head in blending is also shown by the syllabic length of the construction, which, in the majority of cases, follows that of the head, as also shown by Kubozono (1990) for Japanese and English.7 However, there are also exceptions, where the size of a Greek blend surpasses 6. This is not a real deletion though, but only a superficial absence of the derivational suffix, since its properties are still active. See Ralli and Karasimos (2009) for more details. 7. As suggested by Bauer (1983), the size of blends, their syllable structure and the general segmental makeup are predictable from the base constituents.

42 Angela Ralli and George J. Xydopoulos that of the second constituent by one syllable at most, and also displays an alteration of the onset of its second syllable: po.tí.zo (7) a. pe.zo.tí.zo < pe.zós ‘splash a pedestrian’ ‘pedestrian’ ‘splash’ vs. possible blend *pezízo and compound *pezopotízo b.

kor.na.lá.kas < kor.ná.ro ma.lá.kas ‘a “wanker” driver ‘blow the horn’ ‘wanker’ who blew the horn’ vs. possible blend *kornákas and compound *kornaromalákas

c. a.γa.pú.δi < a.γá.pi tra.γú.δi ‘love song’ ‘love’ ‘song’ vs. possible blend *aγúδi and compound *aγapotráγuδo d. a.e.ra.ji.tó < a.é.ras fa.ji.tó ‘food served on airplanes’ ‘air’ ‘food’ vs. possible blend *aritό and compound *aerofajitό Generally, the following instances of segment reduction can be frequently identified: i) The syllabic length of the second constituent is maintained, while the first constituent contributes two syllables to the blend: (8) si.di.rá.zo < si.dí a.γo.rá.zo ‘buy a newspaper to get the free CD’ ‘CD’ ‘buy’ ii) The first syllable of the second constituent is replaced in its entirety by that of the first constituent: (9) a. vlá.ma < vlá.ka vlí.ma ‘extremely stupid’ ‘stupid’ ‘thick’ b. psi.drós < psi.los xo.drós ‘tall and fat’ ‘tall’ ‘fat’ c. vla.ksi.tzís < vlaks ta.ksi.tzís ‘stupid taxi driver’ ‘stupid’ ‘taxi driver’ d. vre.xa.lí.zi < vré.xi psi.xa.lí.zi ‘it rains in small drops’ ‘it rains’ ‘it drizzles’

Blend formation in Modern Greek 43

iii) The syllabic structure of the first constituent is almost entirely reduced except for the onset of its first syllable, which replaces the onset of the first syllable of the second constituent (cases termed acro-blends by KoutitaKaimaki and Fliatouras 2001): (10) a. kró.po.li < ke.rí pró.po.li ‘wax and propolis’ ‘wax’ ‘propolis’ b. tza.má.tos < tzá.mi γa.má.tos ‘very cool person’ ‘cool’ ‘great chap’ c. pan.tó < pal.tó man.tó ‘a semi-casual coat’ ‘overcoat’ ‘light coat’ iv) The entire form of the second constituent is kept, and only the onset of the first one is added to it, producing voicing, where applicable: (11) a. kré.vo.me < ku.ré.vo.me ré.vo.me ‘burping while ‘have a haircut’ ‘burp’ having a haircut’ b. mba.tá.ta8 < ma.la.kía pa.tá.ta ‘a rubbish thing ‘junk’ ‘crap’ (e.g. a film)’ c. zno.pós < ze.stós no.pós ‘hot and sweating’ ‘hot’ ‘sweating’ d. zar.xí.δis < zá.ri ar.xí.δis ‘a player taking a long ‘dice’ ‘a pain in the ass’ time to throw the dice’ It should be noted that the extent of form reduction varies, depending on the speaker’s willingness to communicate a smaller or greater part of the meaning of the combination. As also noticed by Ronneberger-Sibold (2006), speakers resort to lesser or greater form reduction for communicative purposes. As a result, the form and meaning of Greek blends are not easily recoverable, especially when the first constituent has been drastically reduced to the initial sound, as in the examples of (11). In contrast, the items of (12) are relatively easy to recover as, in spite of the segment deletion, both constituents are fully recognizable: 8. This blend contains a word-initial cluster /mb/ which is contrary to Greek phonology, which does not allow such a cluster in this position.

44 Angela Ralli and George J. Xydopoulos (12) a. vla.ksi.tzís < vlaks ta.ksi.tzís ‘stupid taxi driver’ ‘stupid’ ‘taxi driver’ b. vre.xa.lí.zi < vré.xi psi.xa.lízi ‘it rains in small drops’ ‘it rains’ ‘it drizzles’ 4. Pseudo-blends All data discussed so far display a compound-like morphological structure, and an interpretation deriving from the meaning of the constituent parts, while the only semantic deviation which they may assume is that produced by metaphor.9 However, Arvaniti (1998) as well as Koutita-Kaimaki and Fliatouras (2001)10 report data which they consider as blends only because they appear to follow the structural constraints of blend formation, where most of the first constituent is lost while the second loses no more than one or two syllables. Crucially, though, these data lack the meaning criterion to be true blends, since they are semantically empty (see also Katsouda and Kritikou 2009: 440). For an illustration, consider the examples in (13) that constitute phonological paronyms either of the first or the second constituent (see Xydopoulos 2008: 157–158): (13) a. krókotas < krókos kókotas no meaning ‘yolk’ ‘Kokotas’ (surname) b. bekáltsa < bekátsa káltsa no meaning ‘woodcock’ ‘sock’ c. maimúθ < maimú mamúθ no meaning ‘monkey’ ‘mammoth’ d. aitóst < aitós tost no meaning ‘eagle’ ‘toast’ e. melitsúla < melisúla tsúla no meaning ‘tiny bee’ ‘slut’ f. tsiguínos < tsigúnis piguínos no meaning ‘meanie’ ‘penguin’ 9. According to Ralli and Andreou (2012), semantic deviation is not sufficient to denote exocentricity. 10. Following the same line of thought, Katsouda (2009) discusses the phenomenon of blending in Greek, as opposed to that of contamination, from a terminological point of view.

Blend formation in Modern Greek 45

g. skulikó < no meaning

skulíki sxolikó ‘worm’ ‘school bus’

h. periptéri < no meaning

períptero peristéri ‘kiosk’ ‘pigeon’

i. nixterjíδa < no meaning

nixteríδa jíδa ‘bat’ ‘goat’

j. kokinoskupítsa < kokinoskufítsa skupítsa no meaning ‘Little Red Riding Hood’ ‘whisk’ Given the absence of meaning, the process of building these formations aims solely at the formation of jocular/ludling expressions as a lead on to a joke or anecdote. Structure-wise, this ultimately leads to paronymic formations that can be phonologically similar either to the first or to the second constituent of the formation. For this reason, we call them “false blends” or “pseudoblends”.11 Following our line of thought, only few of these formations could, at some point, qualify as blends on the condition that they manage to obtain a semantic content through specific pragmatic circumstances and to become lexically established. For an illustration, consider the following examples: < tzip tsipúra (14) a. tzipúra12 ‘an expensive 4x4 car’ ‘Jeep’ ‘sea bream’ b. piθikótsis < píθikos Biθikótsis ‘a folk singer who ‘monkey’ ‘Greek singer’ looks like a monkey’ In semantic terms, the item tzipúra, in (14a), has acquired the meaning of “expensiveness” from the fact that the sea bream is quite an expensive 11. In this category we also include what Soudek (1978: 465) calls “graphic blends”. These are formed on the basis of effects created by spelling (i.e. small vs. capital letters, combination of graphemes from different alphabets, use of punctuation marks, etc.) and are made exclusively for advertisement / communicative purposes, e.g. SYNéδrio “conference of ‘Synaspismos’ party” < SYN “[acronym of the] Greek left party ‘Synaspismos’”] + synéδrio “conference” (see Katsouda and Kritikou 2009). These too are “pseudo blends” since only one of their forms (the written one) can contribute to their recoverability. 12. Tzipura is a left-headed blend because it is not based on a morphological compound but on a phrasal appositive formation (Tzip tsipura ‘a Jeep which is a tsipura’). See Ralli (2012) for relevant information.

46 Angela Ralli and George J. Xydopoulos fish. Similarly, the item piθikótsis, in (14b), refers to the Greek folk singer Biθikótsis, whose appearance was said to be reminiscent of a monkey. Other cases that should be excluded from the category of blends are those which are built in accordance with the compounding processes, but are reduced via haplology,13 as shown below: (15) a. vlaxorjátis < ‘rural peasant’

vláxos xorjátis ‘rural’ ‘peasant’

b. panoleθríamvos < ‘disaster and triumph’

panoleθría θríamvos ‘disaster’ ‘triumph’

c. peripteréiban < períptero réiban ‘cheap RayBan-like sunglasses’ ‘kiosk’ ‘RayBan’ Finally, there are a couple of cases that again Koutita-Kaimaki and Fliatouras (2001) consider as “infixed blends”, assuming some sort of infixation: (16) a. tsa.kla.kí.δja < ‘items used in dancing’

tsa.kí.δja kla.ké.tes ‘scat’ ‘tap-dance’

b. skar.fa.ló.no < ‘climb up’

ska.ló.no kar.fó.no ‘climb’ ‘nail’

We doubt that this is a case of blending as we were not able to find other examples in Greek that follow the same pattern and because these appear to be very rare formations in other languages too (see e.g. Soudek 1978: 464 for English and Ronneberger-Sibold 2006: 178–179 for German). Furthermore, the suggested recovery of the original constituents looks rather paretymological (see also Katsouda 2009: 893 for a similar conclusion). 5. Blending at the confine of competence and creativity In languages like Greek, a basic question may arise as to why there should be blend formation if there are productively built lexeme combinations resulting in compounds. A tentative answer would be that speakers form 13. However, several linguists consider that haplologic formations are blends (e.g. “slanguage”. See, for instance, Ronneberger’s (2006: 167) “complete blending”). However, to our understanding, blending involves loss of material from both constituents contrary to what happens in haplology. As shown in the formations illustrated in (15a,b,c), identical segments are adjacent, thus, they are clipped for the sake of haplology.

Blend formation in Modern Greek 47

blends in order to create a special effect in specific situations, something which is not possible with regular compounds. For instance, with the use of “lighter” or “stronger” form reduction, they may express irony or mystery, define a playful situation, convey an allusive message, etc. Thus, although reduction is governed by linguistic laws, as we showed earlier (in Section 3), its motivation is clearly extragrammatical. This is substantial enough to lead us to propose that blending is a process at the boundary between linguistic competence and creativity. According to Schultink (1961) and Lieber (1992), morphological creativity is the process under which there is a conscious coinage of a new word, as opposed to morphological competence and productivity, which involve words that are created by applying automated word-formation rules (Bauer 1983, 2001; Plag 1999). Extending the notion of morphological creativity, Baeskow (2004: 78) assumes that it can also imply a superficial reanalysis of items, which may be done for specific purposes, but without bringing any real change to their categorial status. Adopting these views, we would like to suggest that it is possible to account for the peculiar status of blends. As already pointed out, there are properties which could characterize them as instances of compounding, and properties that make them different. In all the examples given so far, the constituents seem to be deprived of their status of stems (or roots), since they miss substantial parts of their form. We suggest that this picture is only superficial, since the full stem forms, from which the constituents originate, still keep their status as far as their lexical entry is concerned, in that there is no change in their category and meaning. In other words, we imply that blend formation is a special type of compounding: structurally, blends belong to compounds, but superficially, in particular contexts, they lose part of their form. This is done intentionally, for special communicative purposes. As such, blends can be considered as belonging to the confines of morphological competence and creativity. The claims we put forward in this work are also supported by the fact that, in Greek, blends are only found in special vocabularies belonging to subvarieties or dialects. In fact, Greek blends, in their vast majority, form part of the vocabulary of slang, which is mostly used by young people and other well-defined social groups. In this area, blending is becoming quite systematic and productive, and speakers create blends for different types of informal communicative situations as nonce formations (Xydopoulos 2008: 260–263). Relatively fewer and less systematic blends (no more than ca. 20 items) are found in some dialects of Greek, the majority of which can be located on the island of Samos (northeastern Aegean sea) and in Messinia (southwestern Peloponnese) (see Koutita-Kaimaki and Fliatouras 2001).

48 Angela Ralli and George J. Xydopoulos It follows then that blending in Greek is a rather novel process compared to English, where a substantial number of blends are part of the general vocabulary (even in the scientific/technical jargon), as for instance, motel, smog, quasar, medicare, arcology, aniseed etc., and recoverability is not a problem for speakers. 6. Final remarks In the previous sections we have shown that genuine blends in Greek have to be structurally transparent for users to recover their constituent elements. To this end, we claimed that they are based on endocentric (subordinate and attributive) and coordinate compounds, which are reduced to a variable degree. In Greek, they can be found (at least for the time being) only in subvarieties (e.g. youth language, marginal varieties/slang), where abbreviation through the process of reduction is generally allowed. Given that most, if not all, of the data that we were able to collect belong to marginal/slang vocabulary we are tempted to suggest that blending is a novel process in Greek, which will probably become more productive in the long term, and be applicable in general neology, as in English. This development may occur because the structure of blends does not differ from that of a certain type of compounds, and pragmatic and cultural reasons make English-dominated languages, like Greek, borrow lexical items with relative ease. Acknowledgements We would like to thank the editors of the volume and the audience of the International Conference on Lexical Blending, held on 10–11 June 2010 in Lyon, for their comments and suggestions. All mistakes remain our own. References Arvaniti, Amalia. 1998. Endeikseis prosodiakis domis se prosfata leksika migmata tis Ellinikis [Indications for prosodic structure in recent Greek lexical blends]. Studies in Greek Linguistics 18: 68–82. Baeskow, Heike. 2004. Lexical Properties of Selected Non-native Morphemes in English. Tübingen: Gunter Narr Verlag. Bat-El, Outi. 1996. Selecting the best of the worst: The grammar of Hebrew blends. Phonology 13: 283–328. Bauer, Laurie. 1983. English Word-Formation. Cambridge: Cambridge University Press.

Blend formation in Modern Greek 49 Bauer, Laurie. 2001. Morphological Productivity. Cambridge: Cambridge University Press. Bertinetto, Pier Marco. 2001. Blends and syllable structure: A four-fold comparison. In Mercè Lorente, Núria Alturo, Emili Boix, Maria-Rosa Lloret and Lluis Payrató (eds.), La gramática i la semántica en l’estudi de la variació, 59–112. Barcelona: Promociones y Publicaciones Universitarias. Bisetto, Antonietta and Sergio Scalise. 2005. The classification of compounds. Lingue e Linguaggio 4 (2): 319–332. Dressler, Wolfgang U. 2000. Extragrammatical vs. marginal morphology. In Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 1–10. Munich: Lincom Europa. Hock, Hans Henrich and Brian D. Joseph. 1996. Language History, Language Change and Language Relationship: An Introduction to Historical and Comparative Linguistics. Berlin & New York: Mouton de Gruyter. Katsouda, Georgia. 2009. The terms contamination and blending in Greek bibliography: Problems and solutions. In Mary Baltazani, Georgios K. Giannakis, Tasos Tsangalidis and George J. Xydopoulos (eds.), Proceedings of the 8th International Conference of Greek Linguistics (Ioannina, 29 August – 2 September 2007), 885–896. Ioannina: University of Ioannina. Katsouda, Georgia and Paraskevi Kritikou. 2009. Optikoi symfyrmoi [Visual blends]. In Anastasia Christofidou (ed.), Deltio Epistimonikis Orologias ke Neologismon, Vol. 9–10, 439–460. Athens: Academy of Athens. Koutita-Kaimaki, Myrto and Asimakis Fliatouras. 2001. Blends in Greek dialects: A morphosemantic analysis. In Angela Ralli, Brian Joseph and Mark Janse (eds.), Proceedings of the First International Conference of Modern Greek Dialects and Linguistic Theory, 117–130. Patras: University of Patras Publications. Kubozono, Haruo. 1990. Phonological constraints on blending in English as a case for phonology-morphology interface. Yearbook of Morphology 3: 1–20. Lieber, Rochelle. 1992. Deconstructing Morphology: Word Formation in Syntactic Theory. Chicago: Chicago University Press. Plag, Ingo. 1999. Morphological Productivity: Structural Constraints in English Derivation. Berlin & New York: Mouton de Gruyter. Ralli, Angela. 2007. I Sinthesi Lekseon: Morphologiki Diaglossiki Prosengisi [Compounding: A Morphological Cross-linguistic Approach]. Athens: Patakis Publishers. Ralli, Angela. 2008. Compound Markers and Parametric Variation. Language Typology and Universals (STUF) 61: 19–38. Ralli, Angela. 2009. I. E. Hellenic: Modern Greek. In Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 453–464. Oxford University Press. Ralli, Angela. 2012. Compounding in Modern Greek. Dordrecht: Springer. Ralli, Angela and Marios Andreou. 2012. Revisiting Exocentricity in Compounding: Evidence from Greek and Cypriot Greek. In Ferenc Kiefer, Maria Ladanyi and Peter Siptar (eds.), Current Issues in Morphological Theory, 65–82. Amsterdam & Philadelphia: John Benjamins.

50 Angela Ralli and George J. Xydopoulos Ralli, Angela and Athanasios Karasimos. 2009. The Bare-Stem Constraint in Greek Compound Formation. Gengo Kenkyu 135: 29–48. Ronneberger-Sibold, Elke. 2006. Lexical blends: Functionally tuning the transparency of complex words. Folia Linguistica XL (1–2): 155–181. Schultink, Hendrik. 1961. Produktiviteit als Morfologisch Fenomeen. Forum der Letteren 2: 110–125. Soudek, Lev I. 1978. The relation of blending to English word-formation: Theory, structure and typological attempts. In Wolfgang U. Dressler and Wolfgang Meid (eds.), Proceedings of the 12th International Congress of Linguists, 462–466. Innsbruck: Institut für Sprachwissenschaft. Xydopoulos, George J. 2008. Leksikologia: Eisagogi stin analysi tis leksis kai tou leksikou [Lexicology: Introduction to the analysis of the word and the lexicon]. Athens: Patakis Publishers.

Lexical blending in Polish: A result of the ­internationalisation of Slavic languages Ewa Konieczna This paper investigates the nature of lexical blending in contemporary Polish, arguing that the process in question has been triggered by the internationalisation of Slavic languages, understood as, among other things, the adoption of foreign (predominantly English) derivational patterns such as, for example, compounding without interfixation or clipping. Even though the dispute as to the place of blending in the Polish morphological system remains unsettled, an attempt has been made to produce both a morphological and a syntactic classification of Polish blends derived from contemporary mass media with the aim of determining their productivity and comparing it with that in English. The study also includes a discussion of the correlation between expressiveness and phonetic motivation, the principle of economy in blend encoding and decoding as well as the use of typographic devices and the occurrence of homophonic and homonymic elements. Finally, it is stated that a vast majority of Polish blends are highly dependent on the extralinguistic context and unlikely to be institutionalised. Keywords: Polish, blending, clipping, compounding, internationalisation. 1. Introduction The purpose of this paper is to shed some light on lexical blending in contemporary Polish, which started to be more productive only recently from a position of virtual non-existence several decades ago. The reason for the growing number of blends in Polish can be attributed to a tendency towards the internationalisation of Slavic languages (see the next section for an explanation of this term), which had its humble beginnings as early as the 1960s, and which has only recently become very influential. This paper aims to discuss the motivation of Polish blends, the status of the splinters used in them, their syntactic structure and the ratio of native to foreign source elements. The discussion to follow is based on a corpus of 234 journalistic and colloquial blends from the contemporary Polish mass media and several dictionaries of Modern Polish.

52 Ewa Konieczna 2. The internationalisation of Polish The internationalisation of Slavic languages started as far back as the 1960s (Danes 1999); however, it became particularly apparent only after the year 1989/1990, that is, after the fall of communism in Eastern Europe, which triggered an attitude of openness and susceptibility to the influence of languages traditionally associated with Western Europe, especially English. The majority of linguists (e.g. Buttler 1981; Gutschmidt 1998) define internationalisation as the process of adopting not only new lexical items but also non-native affixes and derivational patterns which – in order to deserve this name – should take place in at least three languages. The literature on the subject (e.g. Bosák 1999; Waszakowa 2005) is unanimous in stating that it is a one-sided process, because Slavic languages have assumed the role of the recipient, whereas the role of the donor language has been taken up by English and occasionally by other Western languages.1 The internationalisation of Slavic languages is basically rooted in two sources: firstly, Classical Greek and Latin, which provide an inventory of affixes, combining forms and lexical items, and secondly English, which is a source of novel lexemes as well as word-formation patterns alien to the Slavic morphological tradition. Thus, as has often been remarked, internationalisation, seen as a multifaceted process, exerts a significant influence on the typological tendencies of the languages in question. Due to space limitations, it is not possible to analyse the internationalisation of Slavic languages in detail here (for more information see e.g. Gutschmitt 1998; Bosák 1999). In the case of Polish, internationalisation manifests itself not only in the borrowing of whole words, which come predominantly from English, e.g. komputer, drink, dealer, etc., or in the reliance on foreign affixes that presently enjoy an international status, e.g. -acja ‘-ation’, -yzm ‘-ism’, ekstra‘extra’, etc., but also in the use of new typological models, which have not been very productive in Polish until recently, but are common in English, such as compounding, clipping and blending. As it is often argued that clipping and compounding pave the way for blending (Brdar-Szabó and Brdar 2008), I would now like to concentrate on the increasing productivity of these two morphological processes in Modern Polish. 1. Tulloch (1991) has observed that borrowings from Slavic languages are relatively rare in other languages. Examples of such borrowings are glasnost and perestroika, taken from Russian, which have been adapted to the morphological system of the English language, as can be exemplified by the following derivatives: glasnostian, glasnostic, glasnostified and perestroikan.

Lexical blending in Polish 53

Waszakowa (2005), in her study of Polish nominal neologisms of foreign origin from the years 1989–2004, found that compounds constitute 65 percent of the whole sample, with derivatives representing merely 35 percent. What is more, a significant growth in the number of compounds is not the only change that Polish is undergoing at the moment; another is the typological shift in compounding patterns. This means that, first of all, compounding without interfixation,2 which was not very productive in Polish, is becoming more and more common: (1) a. budzik-radio alarm-clock ‘alarm clock’ b. bluzeczka-nic blouse-nothing ‘scanty blouse’ Another example of the possible influence of English on Polish with respect to compounding patterns is the compound gry (gen.) plan ‘game plan’, in which the order of the constituents has been reversed by putting the noun in the genitive case in the initial position, whereas the unmarked word order is characterised by a genitive occurring in the final position, namely plan gry (gen.). A proliferation of hybrids is another distinct tendency. Recent years, that is, from the end of the twentieth century onwards, have been marked by an upsurge in combinations consisting of native and foreign morphemes, both of classical (2a) and English origins (2b, 2c and 2d), e.g.: (2) a. minimiasto mini-town ‘a very small town’ b. topstudentka top-student ‘the best student’ c. ciuchland garment-land ‘second-hand clothes shop’ 2. In Polish compounding is usually accompanied by interfixation, e.g. list-o-nosz ‘postman’ < list ‘letter’ + -o- + nos-i-ć ‘bring’.

54 Ewa Konieczna d. Rywingate Rywin-gate ‘political scandal caused by Lech Rywin’ Besides classical and English elements, in hybrid combinations one may also come across Russian (ero-)3 or Italian ones (kato-), as in eromania ‘erotic mania’, or katofeminizm ‘Catholic feminism’. Another morphological process that can also be seen as a sign of the progressing internationalisation of Polish is clipping. This type of shortening became more productive in Polish only in recent decades, as amply evidenced by Kaproń-Charzyńska (2005), who refers to it as reductive irregular derivation, following both the well-established tradition of Polish word-formation studies (Grzegorczykowa 1979; Puzynina 1998) and more recent research (Jadacka 2001; Wróbel 2001): (3) a. stoma < b. rela < c. biblia < d. ogólniak <

stomatologia ‘dentistry’ religia ‘religious instruction’ biblioteka ‘library’ (liceum) ogólnokształcące ‘secondary school’ ogólnie ‘generally’ + -o- (interfix) + kształcące ‘educating’.

Note that in (3b), (3c) and (3d) clipping is accompanied by suffixation. As for (3b) and (3c) there is the addition of the feminine morpheme -a, since, according to Kaproń-Charzyńska (2005), the gender of the clipped word is usually (in 75 percent of cases) the same as the gender of the base (religia and biblioteka are also feminine). With regard to (3d), the suffix -ak is added, which is an example of morphological condensation (Szymanek 2010), understood as the formation of a derivative (based on the corresponding compound or noun phrase) in which the head noun (in this case, liceum) is replaced by a nominal suffix like -ak, -owiec, or -ka.

3. What is meant here by saying that ero- (from эротичный ‘erotic’) has been borrowed by Polish from Russian is that it is a clipped form of the lexeme that has been borrowed, not the lexeme itself, as it evidently comes from Greek.

Lexical blending in Polish 55

When attempting to look at this process diachronically, one must admit that in the past clipping in Polish was basically limited to the creation of diminutive forms of personal names (Szpyra-Kozłowska 2000). Nowadays, owing to the prevalent linguistic fashion, it has become a popular means of creating expressively loaded tautological coinages (denoting the same concept as the base) used mainly in slang and professional jargon. Clipped forms are especially common in school and youth slang and they are used to refer mainly to the names of school subjects, see (4a) and (4b); pieces of modern equipment, see (4c) and (4d); events, see (4e) and (4f); and in some informal phrases, see (4g) and (4h): (4) a. hista < b. gimna < c. komp < d. procek < e. impra < f. dyska < g. nara < h. gratki <

historia ‘history’ gimnastyka ‘physical education’ komputer ‘computer’ procesor ‘processor’ impreza ‘party’ dyskoteka ‘disco’ na razie ‘see you’ gratulacje ‘congratulations’

When it comes to professional jargon, clippings constitute a major source of combining forms used in journalese, as in euroobywatel ‘European citizen’, infokonto ‘electronic account’, or narkobiznes ‘drug business’ (see (5a), (5b) and (5c)). It should be emphasised at this point that euro, info and narco should be regarded as borrowings from English rather than native Polish clippings and, apart from euro (or rather Euro standing for the European Football Championship) they function merely as combining forms, not lexicalised clippings.

56 Ewa Konieczna (5) a. euro < b. info < c. narko <

europejski ‘European’ informatyczny ‘electronic’ narkotykowy ‘(of) drug’

As a result of adopting the above-mentioned derivational models, typical of analytic languages, Polish is becoming less and less synthetic, while its analytic and agglutinating tendencies are on the rise (Gutschmidt 2003). Being themselves the manifestation of a typological shift, clipping and compounding are also considered to be the necessary prerequisites for the occurrence of portmanteau words (Brdar-Szabó and Brdar 2008). Since the productivity of compounding, clipping and blending has grown, it can be tentatively assumed that there exists a correlation between the number of clippings and compounds on the one hand and the number of blends on the other. 3. Blending and its place in the Polish morphological system At the outset it needs to be stated that blending is not an entirely new word formation process in the Polish morphological system, as it dates back to Old Polish: (6) a. dmucha < ‘[it] blows’ b. pstrokaty < ‘motley’

dmie + duch ‘blows’ ‘wind’ pstry + srokaty ‘multicoloured’ ‘piebald’

Blending was also involved in creating some more recent coinages, now opaque to a vast majority of native speakers: (7) a. rajstopy < ‘tights’ b. harcmistrz < ‘scout instructor’ c. spodnium < ‘trouser suit’

rajtuzy + ‘leggings’ harcerski + ‘scout’ spodnie + ‘trousers’

stopy ‘feet’ mistrz ‘master’ kostium ‘suit’

Lexical blending in Polish 57

Nonetheless, never before has Polish witnessed such an upsurge of blends and, consequently, never before have Polish linguists faced the task of describing this process, which is quite alien to the Polish morphological tradition due to its low productivity in the past. What is worth emphasising at this point is that Polish does not use the term blending, but the axiologically loaded notion kontaminacja, which derives from Latin contaminatio ‘spoiling, staining’, definitely entailing negative connotations. As this morphological process seems to be treated with reserve by prescriptive grammar, the English terms blending and portmanteau did not catch on; neither did the German Kofferwort or Wortverschmelzung. What presently constitutes an urgent challenge for Polish morphology is the need to formulate a consistent definition of this term, which has not yet been done. However, this view is not shared by all linguists, as some Polish scholars are of the opinion that blending should not be labelled as a separate word formation process. For example, we can find no mention of blending at all in Grzegorczykowa et al. (1998). Kaproń-Charzyńska (2005) discusses words composed of one or two clipped elements; however, she treats them as compounds consisting of disintegrated lexemes (złożenia z ucięciem), e.g.: (8) a. gimbus < ‘school bus’ b. muzytor < ‘actor-musician’

gimnazjum + autobus ‘secondary school’ ‘bus’ muzyk + aktor ‘musician’ ‘actor’

Ochmann (2004), discussing novel compounds consisting of a “disintegrated” base, makes a distinction between compounding and blending and notices that the boundary between these two word-formation processes is not clearcut in the Polish morphological system. She seems to be of the opinion that blending is a process in which the final part of the first constituent and the initial part of the second constituent are clipped, which must be accompanied by phonological and graphic overlapping (in Polish referred to as węzeł kontaminacyjny). Thus, in her opinion, cygaros ‘thick cigarette’  3 %) 11 (> 1 %) 1 (> 1 %) 7 (> 1 %)

As Table 2 shows, most blends and complex clippings were coined from two nouns. In this group, 427 words contained linking elements. Thirty-nine of the 1148 words contained additional diminutives or other suffixes (for a comprehensive treatment of the morphological structures, see Kulish 2009). Of the 1065 words that were created by merging two nouns, 444 were linear blends, 115 were complex clippings, and the remaining 506 words belonged to various types – blends and clippings with interfixes and intercalative blends (e.g. rybananba < ryba + banan, ‘fish’ + ‘banana’). On average, the blends in our data set were 3.4 syllables long, the complex clippings were 2.8 syllables long, which means that the Ukrainian blends were relatively long compared to English blends (cf. Gries 2004c), and that they were longer than the Ukrainian complex clippings. 3.2. Hypotheses about the structure of English blends The rather large number of Ukrainian blends produced in this experiment allows us to test hypotheses put forward about the structure of blends. Most structural observations in the literature, however, are based on English blends; in order to attenuate language-specific biases, we therefore limit our analysis to the 444 Ukrainian linear NN blends without any linking elements in our data, corresponding to 322 types.

Ukrainian Blends: Elicitation paradigm and structural analysis 85

If the structures observed in English blends reflect universal cognitive and/or linguistic tendencies, we will expect them to be realized across languages and also to emerge in the Ukrainian data; if the structures previously observed in corpora of English blends do not emerge in the Ukrainian data, this could be due to either different linguistic characteristics (which might, in turn, impact on the planning and selection stages during the word production process) or to the fact that our corpus is made up of intentionally coined spoken blends, i.e. blends that were produced under different production modes than most existing data collections of blends. We will first test tendencies related to the order of constituents, then proceed to observations concerning operations performed on the constituents. Finally, we will investigate the phonological structure of the resulting blends and focus on the cross-over point. 3.2.1. Order of constituents Tendencies that have been proposed relating to the preferred ordering of constituents cover several domains: In blends created from two source words the first source word is in general shorter (Kelly 1998; Gries 2004b, this volume), more frequent (Kelly 1998), more prototypical (spork < spoon + fork, Kelly 1998) than the second source word and comes first in temporal order (brunch < breakfast + lunch, Kelly 1998). Under ideal circumstances we would be able to test several tendencies and possible interactions between them (see e.g. Kelly 1998; Gries 2004b). However, as we did not have good frequency information for Ukrainian, we could not test for frequency effects, and as we did not have any prototypicality ratings for Ukrainian, we could not test for semantic effects either (testing for temporal order was not applicable). This left us with only one remaining observation to be tested, namely that the first element is on average shorter than the second element. Surprisingly, in our data the opposite pattern emerged. The first source words had an average syllable length of 2.6 syllables (type: 2.57), the second source words a length of 2.41 syllables (type: 2.45). One possible explanation is that the observed tendency might mainly hold for written blends, and not for spontaneously produced spoken blends (both intentional and speech error blends); for German speech-error blends the same pattern was found, with the first source words being longer than the second source words (MacKay 1973, but see Gries, this volume). Another possibility is that our production experiment did not allow for the generation of unbiased data: as our stimuli were visual in nature, the

86 Susanne R. Borgwaldt, Tetyana Kulish and Arpita Bose visual dominance of one constituent in the hybrid object pictures may have influenced the ordering of constituents (for a related discussion on the ordering of compound constituents, see Borgwaldt and Benczes 2011). 3.2.2. Operations performed on constituents Other structural observations concern the operations performed on the blends. Out of 444 blends, 104 showed an overlap of the source words, ranging from one phoneme to complete word overlap. The remaining 340 blends were created without overlap. Table 3 displays the percentages of the four types of blend in our dataset: splinter and splinter, splinter and word, word and splinter, and word and word (= complete overlap) (for similar classifications, see Lehrer 2007). Table 3. Operations performed on the source words Type of . blend

1st constituent

2nd constituent

Token. occurrence

Type. occurrence

Spl. + Spl. Spl. + Wd Wd + Spl. Wd + Wd

Reduced Reduced Preserved Preserved

Reduced Preserved Reduced Preserved

194 (44 %) 152 (34 %) 81 (18 %) 17 (4 %)

144 (45 %) 106 (33 %) 62 (19 %) 10 (3 %)

In our corpus of blends the dominant pattern was that both constituents were reduced. If only one constituent was reduced, it was more often the first constituent than the second one. If the goal of creating a lexical blend is to coin a word which, on the one hand, is semantically transparent, i.e. whose parts are recognizable (Bat-El 2006; Bauer, this volume) and, on the other hand, as short and thus economical as possible, it might be more important for the second constituent to be preserved, as word beginnings are more perceptually salient than word endings. The data pattern is also in line with the observation that the shorter source word will on average contribute more to the blend than the longer source word (e.g. Gries 2004c). In our corpus the first source word was on average longer than the second source word (see 3.2.1); accordingly the first source word was a better candidate to be shortened and still to be recognized than the second source word.

Ukrainian Blends: Elicitation paradigm and structural analysis 87

3.2.3. Syllable length One characteristic that has been proposed in the literature (e.g. Bat-El 2006; Rainer 1993 for Spanish blends; Neef 2009 for German blends) is a presumed tendency for blends to have the same number of syllables as the longer source word. A token analysis of the number of syllables of the 444 blends and their source words in our corpus, presented in Table 5 below, showed that the majority of produced blends did not adhere to this constraint. Table 4. Number of syllables in blends and their longest source words Number of syllables

Occurrences (token)

more than longest source word same as longest source word less than longest source word

229 (52 %) 166 (37 %) 49 (11 %)

We proceeded to investigate possible frequency effects by splitting the blends into 252 types that showed up once in the corpus (i.e. hapax legomena) and 69 types that were created by at least two participants and analyzing them separately. If there were a strong structural tendency for blends to obey this rule, we might expect “optimal blends”, that is, blends created according to this rule to be coined by more language users than blends with a less optimal structure. However, in the separate analyses the pattern remained the same. Table 5. Number of syllables in blends and longest source words, divided into blends produced once and blends produced more than once Number of syllables

more than longest SW same as longest SW less than longest SW



Occurrences (type) blends produced by . only one participant

blends produced by . more than one participant

134 (53 %) 90 (36 %) 28 (11 %)

37 (54 %) 25 (36 %) 7 (10 %)

Kubozono (1990) proposed a different tendency, claiming that the number of syllables in a blend should be identical to the number of syllables in the second source word. The data in Table 6 reveals that this was the case in less

88 Susanne R. Borgwaldt, Tetyana Kulish and Arpita Bose than a third of the Ukrainian blends, and again the data pattern for blends that were coined only once and those that were coined several times looked very much alike. Table 6. Number of syllables in blends and their second source words Number of syllables

more than 2nd SW same as 2nd SW less than 2nd SW

Occurrences . (token)

309 (70 %) 124 (28 %) 11 (2 %)

Occurrences . (type) blends produced by only one participant

blends produced by more than one participant

171 (68 %) 70 (28 %) 11 (4 %)

49 (71 %) 20 (29 %) 0 (0 %)

To summarize, more than half of the blends in our corpus were longer than their longest source words. Neither the proposed tendency that the optimal blend should have the same number of syllables as the longer source word, nor the requirement that the number of syllables in the blend should equal the number of syllables in the second source word seems to have been a strong rule during the creation of the Ukrainian blends. Had that been the case, we would have seen more differences between blends only coined once, and more frequent, i.e. “optimal” blends. Possibly, in ad-hoc creations like the ones produced in the above experiment, the length of a neologism is not that important, as the advantage of a shorter word-form might rather emerge over time, leading to shorter blends in corpora of written blends that are at least to some degree established in the language. 3.2.4. Cross-over point Another set of observations concerns the optimal cross-over or switch point. Generally, two words can combine in different ways to form a blend (for an example see the different ways to blend fish and banana listed in 2.3). One constraint that seems to hold is a blocking constraint (Bauer 1983) – the resulting blend form should not be phonologically or orthographically identical to an existing word. Gries (2004b) proposed that the shorter source word would on average contribute more to the blend than the longer source word, and that the second source word would on average contribute more than the first source word.

Ukrainian Blends: Elicitation paradigm and structural analysis 89

To test this prediction, we investigated the contributions of the two words by counting the number of phonemes in the blend that belonged to the first and second source word, respectively; in the case of overlap blends, overlapping phonemes counted for both parts. For example, in lymomat (< lymon + tomat, ‘lemon’ + ‘tomato’) lymon contributed 4 phonemes (lymo) to the blend, and tomat contributed 4 phonemes (omat) as well. The analysis of the 322 blend types in our corpus revealed that the first source word contributed 4.25 phonemes to the blend on average, while the second source word’s contribution was on average 4.08 phonemes. As the first source words were longer than the second ones in our corpus (see discussion in 3.2.1), Gries’ prediction that the shorter source word should contribute more to the blend was supported by the data, but his prediction that the second source word should contribute more than the first one was not. 4. Conclusions The goal of this paper was two-fold. First we wanted to present the results of a novel object-naming task designed to explore Ukrainian word formation patterns. In that experiment Ukrainian participants had to name novel hybrid objects that were composed of two identifiable parts (such as an animal that is part chicken and part fox, i.e. chicken fox, or a cloud in the shape of a horse, i.e. horse cloud); data on German and Hungarian labels for these objects had been previously collected. The data pattern that emerged showed that Ukrainians favoured blends and complex clippings as labels for the hybrid objects, in contrast to Germans’ and Hungarians’ preferences for compound labels to refer to the hybrid objects. It remains to be seen if these contrasting data patterns exclusively reflect different word-formation preferences or if the high percentage of blends and clippings could be related to factors that are not language-specific, like participants’ motivation and differences in task instructions. The corpus of blends obtained with the above paradigm is made up of deliberately uttered spoken blends. Possibly, they differ structurally both from spoken blends produced as speech errors and from written blends that might have been more carefully coined than our spontaneously produced hybrid object labels. The second goal of this paper was to test whether prevalent hypotheses of preferred blend structures were supported in our Ukrainian data. Surprisingly, in our corpus the first source word was on average longer than the second source word – a characteristic that had been observed before in

90 Susanne R. Borgwaldt, Tetyana Kulish and Arpita Bose speech error blends, but not in deliberate blends. In most blends both constituents were reduced, and if only one constituent was reduced, it was the first one rather than the second one. The majority of the blends were longer than the longest source word; in general the blends in our corpus were rather long in terms of syllable length. Possibly neologisms improve their chances of survival in the language when they are short and economical. Ad-hoc creations might on the other hand turn out to be longer, as economy might not be such an important property in the naming task we employed. Replications of the experimental paradigm introduced in the first part in other typologically different languages could provide us with more data on intentionally produced lexical blends. They could be used for cross-linguistic investigations into structural preferences for this type of blends, which might turn out to be different from both speech error blends and written blends, and contribute to deeper insights into spoken word production processes. References Algeo, John. 1977. Blends, a structural and systemic view. American Speech 52 (1–2): 47–64. Azarova, Larisa. 2004. Structural and Phonetic Organization of Compound Words (the concept of ‘gold’ proportion). Kyiv: Pedagogical Press. Bat-El, Outi. 1996. Selecting the best of the worst: The grammar of Hebrew blends, Phonology 13 (3): 283–328. Bat-El, Outi. 2006. Blend. In Keith Brown (ed.), Encyclopedia of Language and Linguistics, 66–70. Oxford: Elsevier. Bauer, Laurie. 1983. English Word-Formation. Cambridge: Cambridge University Press. Berko, Jean. 1958. The child’s learning of English morphology. Word 14 (2–3): 150–177. Berman, Ruth and Eve Clark. 1989. Learning to use compounds for contrast: data from Hebrew. First Language 9 (27): 247–270. Bertinetto, Pier Marco. 2001. Blends and syllable structure: A four-fold comparison. In Mercè Lorente, Núria Alturo, Emili Boix, Maria-Rosa Lloret and Lluís Payrató (eds.), La gramática i la semántica en l’estudi de la variació, 59–112. Barcelona: Promociones y Publicaciones Universitarias. Borgwaldt, Susanne and Réka Benczes. 2011. Word-formation patterns in a crosslinguistic perspective: Testing predictions for novel object naming in Hungarian and German. In Doris Schönefeld (ed.), Converging Evidence: Methodological and Theoretical Issues for Linguistic Research, 221–246. Amsterdam & Philadelphia: John Benjamins. Brdar-Szabó, Rita and Mario Brdar. 2008. On the marginality of lexical blending. Jezikoslovlje 9 (1–2): 171–194.

Ukrainian Blends: Elicitation paradigm and structural analysis 91 Friedrich, Cornelia. 2008. Kontamination – Zur Form und Funktion eines Wortbildungstyps im Deutschen. Erlangen & Nuremberg: University of  ErlangenNuremberg dissertation. Fromkin, Victoria. 1973. The non-anomalous nature of anomalous utterances. In Victoria Fromkin (ed.), Speech Errors as Linguistic Evidence, 215–242. The Hague: Mouton. Garrett, Merrill. 1980. Levels of processing in sentence production. In Brian Butterworth (ed.), Language Production, Vol. I: Speech and Talk, 177–220. London: Academic Press. Gottfried, Gail. 1997. Using metaphors as modifiers: Children’s production of metaphoric compounds. Journal of Child Language 24 (3): 567–601. Gries, Stefan. 2004a. Shouldn’t it be breakfunch? A quantitative analysis of blend structure in English. Linguistics 42 (3): 639–667. Gries, Stefan. 2004b. Some characteristics of English morphological blends. In Mary Andronis, Erin Debenport, Anne Pycha and Keiko Yoshimura (eds.), Papers from the 38th Regional Meeting of the Chicago Linguistics Society, Vol. II: The Panels, 201–216. Chicago: Chicago Linguistics Society. Gries, Stefan. 2004c. Isn’t that fantabulous? How similarity motivates intentional morphological blends in English. In Language, Culture, and Mind, Michel Achard and Suzanne Kemmer (eds.), 415–428. Stanford, CA: CSLI. Herman, Nataliya. 2009. Neologisms in Modern Ukrainian Mass-media. Berdyansk: Berdyansk State Pedagogical University. Horpynych, Volodymyr. 2004. Morphology of the Ukrainian Language. Kyiv: Akademiya. Kemmer, Suzanne. 2003. Schemas and lexical blends. In Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.), Motivation on Language: Studies in Honor of Günter Radden, 69–97. Amsterdam & Philadelphia: John Benjamins. Kelly, Michael. 1998. To ‘brunch’ or to ‘brench’: Some aspects of blend structure. Linguistics 36 (3): 579–590. Klymenko, Nina. 1984. Word-building Structure: Semantics of Complex Words in Modern Ukrainian. Kyiv: Naukova dumka. Korunets’, Il’ko. 2004. Comparative Typology of English and Ukrainian. ­Vinnytsya: Nova knyha Kubozono, Haruo. 1989. The mora and syllable structure in Japanese: Evidence from speech errors. Language and Speech 32 (3): 249–278. Kubozono, Haruo. 1990. Phonological constraints on blending in English as a case for phonology-morphology interface. Yearbook of Morphology 3: 1–20. Kulish, Tetyana. 2009. Novel object naming in Ukrainian. Potsdam: University of Potsdam MSc thesis (Institute of Linguistics). Lehrer, Adrienne. 2007. Blendalicious. In Judith Munat (ed.), Lexical Creativity, Texts and Contexts, 115–133. Amsterdam & Philadelphia: John Benjamins. López Rúa, Paula. 2004. The categorical continuum of English blends. English Studies 85 (1): 63–76.

92 Susanne R. Borgwaldt, Tetyana Kulish and Arpita Bose MacKay, Donald. 1973. Complexity in output systems: Evidence from behavioral hybrids. American Journal of Psychology 86 (4): 785–806. Manova, Stela. 2011. Understanding Morphological Rules – With Special Emphasis on Conversion and Subtraction in Bulgarian, Russian and Serbo-Croatian. Dordrecht: Springer. Mellenius, Ingmarie. 1997. The Acquisition of Nominal Compounding in Swedish. Lund: Lund University Press. Misyats’, Nataliya. 2007. Peculiarities of Ukrainian Technical Terminology Formation. L’viv: Zhitomir State Ivan Franko University. Neef, Martin. 2009. IE, Germanic: German. In Rochelle Lieber and Pavol Stekauer (eds.), The Oxford Handbook of Compounding, 386–399. Oxford: Oxford University Press. Neliuba, Anatoliy. 2004. Stem- and Word-composing in the Context of Word-formative Nomination. Kharkiv: Kharkiv National University. Neliuba, Anatoliy. 2007. Phenomena of Economy in the Word-building Nomination of the Ukrainian Language. Kharkiv: Kharkiv National University. Nicoladis, Elena. 2003. What compound nouns mean to preschool children. Brain and Language 84 (1): 38–49. Onufriyenko, Galyna. 1994. Typology of hybrid constructions in East Slavic lan­­guages (on the material of Russian and Ukrainian technical terminology). Ukrainian Journal 4: 12–16. Rainer, Franz. 1993. Spanische Wortbildungslehre. Tübingen: Niemeyer. Ronneberger-Sibold, Elke. 2006. Lexical blends: Functionally tuning the transparency of complex words. Folia Linguistica 40 (1–2): 155–181. Schmitt, Norbert and Cheryl Zimmermann. 2002. Derivative word forms: what do learners know? TESOL Quarterly 36 (2): 145–171. Semenza, Carlos, Claudio Luzzatti and Simona Carabelli. 1997. Morphological representation of compound nouns: A study on Italian aphasic patients. Journal of Neurolinguistics 10 (1): 33–43. Stemberger, Joseph. 1982. The nature of segments in the lexicon: Evidence from speech errors. Lingua 56 (3–4): 235–259.

Are reduced compounds compounds? Morphological and prosodic properties of reduced compounds in Russian and Mandarin Chinese Giorgio Francesco Arcodia and Fabio Montermini * Blends and, more generally, reduced compounds are often considered as marginal phenomena in the grammars of languages, as opposed, for instance, to “canonical” compounding. In this paper, we propose to distinguish two types of compounding in the world’s languages, one which is the reflection of a general, and possibly universal, human cognitive ability to connect two words by simply juxtaposing them, and one which corresponds to morphological patterns that are codified (conventionalized) in the grammar of a language. The production and interpretation of the first, we claim, are determined by cross-linguistic constraints, while those of the latter are determined by language-specific factors. In this paper, we consider reduced compounds in two genetically unrelated and typologically distinct languages, Russian and Mandarin Chinese, in order to show that the degree of conventionalization of the process has a direct impact on the formal and semantic properties of this type of construction. Keywords: compounding, reduced compounds, blends, grammaticalization, Russian, Mandarin Chinese. 1. Introduction Blends and, more generally, reduced compounds are often considered as marginal phenomena in the grammars of languages, as opposed, for instance, to “canonical” compounding. We consider blending as a particular phenomenon that creates a new word by combining two or more words that have one or more segments in common, and for which the common string serves as the conjunction point (Eng. motel < motor hotel). Reduced compounds, on * Although this work is the outcome of a joint project, Sections 1 and 2 were written by Fabio Montermini and Sections 3 and 4 by Giorgio Francesco ­A rcodia. We shall be using simplified Chinese characters throughout the article.

94 Giorgio Francesco Arcodia and Fabio Montermini  the other hand, are complex words that are made up of two or more words, at least one of which is shortened, usually according to prosodic principles (Russian imčistka < imičeskaja čistka ‘dry-cleaner’s’). While blends are also reduced compounds, the reverse is not necessarily true. In this paper, we propose to distinguish two types of compounding in the world’s languages, one which is the reflection of a general, and possibly universal, human cognitive ability to semantically connect two words by simply juxtaposing them, and one which corresponds to morphological derivational patterns that are codified in a language’s grammar. The construction and interpretation of the first, we claim, is determined by cross-linguistic constraints, while those of the latter are determined by language-specific factors. In this paper, we consider reduced compounds in two genetically unrelated and typologically distinct languages, Russian and Chinese, in order to show that the degree of conventionalization of the process has a direct impact on the formal and semantic properties of this type of constructions. Compounding is often considered as the most widespread word-formation process across languages. In linguistics texts it is not rare, for instance, to find statements such as the following: There are a considerable number of languages without inflections, perhaps none without compounding and derivation.  (Greenberg 1963) Compounds are important objects of morphological investigation, because compounds are present in all languages of the world (as far as described in grammars). Thus, compounding is the widest-spread morphological technique. (Dressler 2006: 23)

However, it is not certain that compounding is a mechanism which belongs to the grammars of all languages (cf. Bauer 2009). Rather, we should probably distinguish Compounding1, i.e. the human cognitive capacity of semantically linking two words without any grammatical element marking this relation, from Compounding2, a lexical enrichment process encoded in the morphology of individual languages (cf. also Bauer 2006: 721; Guevara and Scalise 2009). Compounding1 may be considered as a property of human linguistic ability (and maybe beyond, as experiments on ape language sometimes seem to suggest, cf. Kenneally 2007: 42–43), whereas Compounding2 is a restrictive notion, subject to language-specific parameters. On the other hand, Compounding1 is probably subject to universal constraints pertaining to linguistic ability. Let us take a first example: in French, compounding is restricted to some combinations of categories and the position of the head is parameterized; as in other Romance languages French compounds are left-headed (heads are underlined):

Are reduced compounds compounds? 95

(1) NN poisson chat ‘fish+cat, catfish’ VN monte-charge ‘lift+load, freight elevator’ *VV *monte-descend ‘lift+bring down’   (vs. lexicalized entre-sort ‘walk in+walk out, fair stand’) *NV *main-porter ‘hand+carry’ (vs. lexicalized colporter ‘neck+carry, peddle’) On the contrary, French blends (Fr. mots-valises or amalgames) may violate the above-mentioned principles; for instance, they often display “illegal” category combinations (2a), and may be left- or right-headed (2b): (2) a. VV NV

s’étrangueuler (étrangler + s’engueuler = ‘strangle + argue’) pleurire (pleurer + rire = ‘cry + laugh’) giraffoler (giraffe + raffoler = ‘giraffe + be crazy about’) télépater (télépathie + épater = ‘telepathy + amaze’)

b.

élevache (élevage + vache = ‘breeding + cow’) nostalgérie (nostalgie + Algérie = ‘nostalgia + Algeria’) musictionnaire (musique + dictionnaire = ‘music + dictionary’) coca-colonisation (Coca-Cola + colonisation = ‘id. + colonization’)



(data from Grésillon 1984; Fradin 2000; Bertinetto 2001)

The main constraint on blends of this type is that there is a common phonological string between the two components that may function as a juncture (e.g. [ʁaf] in giraffoler), which dominates all other, language-specific constraints, for instance that on category combination or on head position, and also constraints that are active for “canonical” constructed words, such as base-derivate faithfulness. All the other properties of the blends in (2), we claim, do not derive from language-specific properties, but from more general, possibly universal, constraints, both for their semantic interpretation and for the identification of the head, for which there is no positional cue.

96 Giorgio Francesco Arcodia and Fabio Montermini  We consider Compounding2 as the product of the tendency of languages to encode multi-word expressions into constructions. Recent developments in morphological theory, such as Construction Morphology (cf. Booij 2009, 2010), tend to consider compounds as constructions, i.e. as “abstract schemas that generalize over sets of existing complex words with a systematic correlation between form and meaning” (Booij 2009: 201). From a diachronic point of view, we may see this process (i.e., roughly, the development of Compounding1 into Compounding2) as a kind of conventionalization. In fact, we consider conventionalization as a particular kind of grammaticalization, although we prefer to avoid using this label in this paper, since it is commonly reserved for the transformation of a lexical item into a functional one (but for recent studies on the topic which have tried to extend this concept to constructions see Bybee 2003; Traugott 2003). Under this view, at the extreme point of conventionalization, we argue, reduced compounds no longer distinguish themselves from the “real” compounds of a language. In fact, the simple juxtaposition of two elements with no phonological modification of one or both of them is just one of the possibilities found for compounding in the world’s languages. Indeed, compound elements may undergo various types of modifications, including historical, partially unpredictable variation (3a), the choice of a particular form of the base (usually a “bare root”, 3b), or systematic phonological modifications (3c): (3) a. b. c.1

French: col+porter Danish: jom+fru German: Schwimm+bad Russian: pyle+sos Greek: psaro+kaiko Armenian: gəlxa+c’av

‘neck+carry, peddle’ (vs. cou ‘neck’) ‘young+lady, virgin’ (vs. ung ‘young’) ‘swim+bath, swimming pool’ ‘dust+suck, vacuum cleaner’ ‘fish+boat, fishing boat’ (vs. psari ‘fish’) ‘head+pain, headache’ (vs. gəlx ‘head’)

Often, the modifications exemplified in (3c) are systematic for all or for a subset of the compounds of a language. So, the adjunction of an [o] to the first element of compounds is a pervasive strategy in Greek (cf. Ralli 2007). These modifications also include purely prosodic changes, such as stress or tone shifts (cf. Montermini 2010: 84 for some examples). We think that reduced compounds may, at least in some cases, be analysed as (prosodically based) modifications of the same nature as those exemplified in (3c). 1. Vowels marked in bold correspond to linking elements systematically found in the compounds of Greek and Armenian.

Are reduced compounds compounds? 97

In this paper we present two case studies of reduced compounds in different, genetically unrelated languages, showing three different degrees of encoding of these structures into the grammars of particular languages. First, in Section 2, we analyse the so-called stump compounds of Russian, which display the highest degree of encoding. In Section 3 we turn to Chinese, whose reduced compounds are, apparently, less constrained. 2. Stump compounds in Russian The creation of stump compounds (henceforth SCs) in Russian is a construction strategy that was already used before World War I, had a great development after the Russian revolution and during the entire Soviet era, and is still very productive (cf. Comrie et al. 1996: 139–140). It consists, essentially, in the compounding of two (or more) lexemes, at least one of which is reduced, often to a monosyllable. Russian SCs may be divided into four types, according to two criteria:2 (i) both members are clipped (c+c) / the first member is clipped, the second member is not (c+f); (ii) both members are based on stems (s+s) / the first member is based on a stem, the second member is based on an inflected word form (s+w).3 Table 1 sums up the situation (most of the examples are from GRJa 1960; Comrie et al. 1996; Billings 1998, and/or have been collected via Internet searches): Table 1. types of SC in Russian 

(c = clipped; f = full)

stem+stem

stem+word

c+c

filfak filologičeskij + fakultet ‘philology faculty’

pomrež pomoščnik + režisseragen ‘assistant director’

c+f

kapstrana kapitalističeskaja + strana ‘capitalist country’

zavsektorom zavedujuščij + sektorominstr ‘sector director’

2. Abbreviation explanation: c = clipped word; f = full word; s = stem; w = word. 3. The terms stem and word are borrowed from Billings (1998). In our approach, stems correspond to lexemes, abstract units which are represented by their quotation forms (i.e. the nominative singular), words correspond to inflected word forms.

98 Giorgio Francesco Arcodia and Fabio Montermini  Note that we only present binary SCs, but SCs made up of more than two elements are far from uncommon (e.g. gorvoenkom < gorodskoj + voennyj + komitet ‘city war committee’). Other patterns than those exemplified in Table 1 are attested, but they are rarer. For instance, there are examples of. SC in which the first member is an inflected form (e.g. beruši < beregiimper + ušinom pl ‘earplug’), or SCs in which the second element is not clipped but syncopated (e.g. narkomat < narkotičeskij + kommissariat ‘drug police station’). Sometimes, apparently, only the second element is clipped (e.g. sport-mag < sport + magazin ‘sporting goods shop’; kinorež < kino + režisser ‘film director’). However, as we will see later, these cases may not be considered as different from the other SCs mentioned above. We consider (contra Billings 1998) that all the different types of SC illustrated in Table 1 are not different in substance, and that they are the output of the same kind of process. According to Billings, in fact, “real” SCs (our type c+f) are subject to prosodic constraints (“monosyllabic if consonant-initial and disyllabic if vowel-initial”). However, this observation is invalidated by facts: (4)

politotdel proletkul’t oblsud jurkomitet učbilet

političeskij + otdel proletarnaja + kul’tura oblastnyj + sud juridičeskij + komitet učeničeskij + bilet

‘political division’ ‘proletarian culture’ ‘regional court’ ‘legal committee’ ‘student identity card’

Rather, we consider that SCs of both types illustrated in Table 1 are formed following the same prosodic constraint, namely that the first or both elements are mapped onto the format of the minimal prosodic word in Russian, which can be defined as a bimoraic (closed) syllable. Some observations support this characterization of the Russian minimal word. The Russian lexicon does not contain lexical words which are not at least bimoraic. The rare exceptions are constituted by some inflected forms, such as ldu (‘iceprep sg’), rta (‘mouthgen sg’). However, (i) all these forms display onsets that violate the universal sonority scale; (ii) the forms in question are always in alternation with forms containing a full vowel (cf. lëd ‘icenom sg’, rot ‘mouthnom sg’); (iii) diachronically the consonantal group is the outcome of the fall of a short vowel (a yer, cf. Vaillant 1950: 126–131; Pesetsky 1979). Moreover, the minimal word we have characterized is pertinent for describing other phenomena, such as the formation of hypocoristics or of expressive clippings (Soglasnova 2003; Montermini 2007). Cases such as sportmag may thus be reinterpreted as MinW+MinW SCs, i.e. not different from other c+c SCs.

Are reduced compounds compounds? 99

Simply, here the first element already possesses the desired output format, and the same ambiguity may be observed for second elements, as in futklub ( futbol’nyj + klub ‘football club’), terakt (terrorističeskij + akt ‘terrorist attack’). The behaviour of vowel-initial lexemes deserves attention. As illustrated in (4), a vowel-initial base may sometimes give a closed monosyllabic base to the SC. However, as observed by Billings (1998), these lexemes more often give disyllabic bases: (5) informbjuro informacionnoe + bjuro ‘information agency’ upravdom upravljajuščij + domom ‘manager of a block of flats’ We consider that the disyllabic format for vowel-initial bases is the outcome of a universal semiotic constraint, requiring that a base contain enough phonological material to be recognized. This is demonstrated by the fact that when the first element tends to become a fixed element (i.e. a sort of affix, or a splinter, in Bauer’s (2005) terms), and base recognition is no longer an issue, this constraint may be violated: (6)

in- < inostrannyj inotdel inturist injaz(yk)

‘foreign’ ‘foreign affairs section’ ‘foreign tourist’ ‘foreign language’



jur- < juridičeskij jurfirma jurfak jurkom(itet)

‘legal’ ‘legal company’ ‘law faculty’ ‘legal committee’

From a morphological point of view, SCs display different inflection properties depending on their construction type. Table 2 sums up the situation: Table 2. gender / declension class of Russian SCs  stem+stem c+c c+f

(c= clipped; f = full) stem+word

masculine / Class I gender / class inherited from W2

invariable

Class I-membership for c+c SCs is an effect of their construction strategy: as we have seen, the minimal word, onto which SC members are mapped, has the form of a closed syllable, and nouns that are consonant-final in the

100 Giorgio Francesco Arcodia and Fabio Montermini  nominative singular belong to Class I. The same effect is observed with gender assignment: the SCs in question are assigned the gender which is most compatible with Class I, i.e. masculine. Of course, in this case gender base-derivate faithfulness is violated when the base has a different gender — and may, see (7b), or may not, see (7a), be an inflected noun: (7) a. proletkul’t proletarnaja + kul’tura ‘proletarian culture’ texizdat texničeskoe + izdatel’stvo ‘technical publishing house’ specnaz special’noe + naznačenie ‘special forces’ b. likbez likvidacija + bezgramotnostigen sg ‘abolition of illiteracy’ stroemol stroenie + molekulgen pl ‘molecule construction’ We did not accomplish a systematic corpus analysis, but we have the strong feeling that c+c combinations are disfavoured when the second element is not a Class I masculine noun. Comrie et al. (1996: 142), for instance, cite the case of stengaz (stennaja + gazeta, ‘wall newspaper’), which was formed under this form in the 1930s, but was later replaced by stengazeta, which seems to be the preferred form today.4 Finally, SCs of the zavsektorom type are generally invariable both for case and number: (8) upravdelami (upravliauščij + delamiinstr pl ‘on-site manager’) dlja upravdelami ‘for the on-site manager’ etot / eti upravdelami ‘this / these on-site manager(s)’ Let us now turn to the constituents of SCs, an aspect that the studies on the topic have often neglected. In most cases, the SC corresponds to a complex syntactic construction, and in particular to a noun phrase. In the majority of cases the phrase the SC is “derived” from is a “phrasal lexeme” (cf. Benigni and Masini 2009, 2010), a (Rel) Adj. + Noun construction, or, more rarely, a Noun + complement construction, where the complement corresponds to an inflected noun (see the data in Table 1, and in 7b). (An even rarer type of SC takes name + patronymic complexes as bases, e.g. Nikvas < Nikolaj Vasil’evič.) When the base is constituted by an Adj. + N phrase, cases in which the adjective is qualifying and not relational are very rare, but nevertheless attested (cf. legprom < legkaja + promyšlennost’ ‘light industry’, 4. An informal Google search (June 2010) gives a proportion of 150 / 1 for stengazeta vs. stengaz.

Are reduced compounds compounds? 101

starlet < staršij + lejtenant ‘senior lieutenant’). Moreover, since the construction of relational adjectives is very productive and systematic in Russian, some cases may be considered as ambiguous (cf. sportmag or futklub cited above), and in fact there seems to be no justification in distinguishing these compounds from N+N compounds in other languages. The semantic relation between the two elements is almost always that of a subordinate compound. However, since different SCs may come from different syntactic structures, headedness may not be unambiguous. To sum up, we consider that Russian SCs correspond to two constructions with the following characteristics: Table 3. Russian SCs as constructions SC1 formal morphological semantic

SC2

MinW + MinW masculine Class I

MinW + lexeme gender / class inherited from W2 invariable (mostly) subordinate

The formal and morpho-categorial properties of both types of SC appear to be perfectly conventionalized. As far as semantics is concerned, however, interpretation mostly depends on pragmatics, due in particular to the fact that head position may not be unambiguously determined. This is a consequence of the proximity of SCs to syntactic constructions. As for the French blends presented in (1), some of the properties of Russian SCs (in particular the semantic ones) depend on extragrammatical factors (in the sense of Dressler 2000), such as pragmatics, and in general the construction of reference. However, by virtue of their highly constrained formal and morpho-categorial properties, on a scale of conventionalization they should be considered as closer to “canonical” compounds. 3. Reduced compounds in Mandarin Chinese The topic of abbreviations in the Chinese lexicon has been extensively studied (see Ceccagno and Basciano 2009: 115 and the works cited there), possibly because abbreviation is an extremely productive pattern for coining new lexical items in Mandarin; according to the figures presented in Sawer (1995), 18.7% of neologisms are abbreviations, almost one word in five. Two different (albeit connected) phenomena are subsumed under the label “abbreviation” in Chinese lexicology. According to Packard (2000: 268), “[a]

102 Giorgio Francesco Arcodia and Fabio Montermini  new word is considered an abbreviated form if (a) there is a clear preexisting word or phrase that contains all the constituents of the abbreviation, and (b) the abbreviation is considered to be derived directly from that longer word or phrase”. So, according to such a definition, the input of an abbreviation could be both “bare” compounds, i.e. combinations of two or more lexemes, or phrases, built according to syntactic rules (examples adapted from Packard 2000: 271): (9)

劳动 保险 → láodòng bǎoxiǎn labour-act protect-risk ‘labour’ ‘insurance’

(10)

环境 与 发展 → huánjìng yǔ fāzhǎn surround-border and emit-expand ‘environment’ ‘and’ ‘development’

劳保 láobǎo labour-protect ‘labour insurance’ 环发 huánfā surround-emit ‘environment and development’

An observer who is not acquainted with Mandarin would probably infer that 保险 bǎoxiǎn (protect + risk ‘insurance’) is a compound word, since it is made of two lexical morphemes; but what about 劳动保险 láodòng bǎoxiǎn (labour + insurance ‘labour insurance’)? Here two complex words are combined, so a compound is combined with another compound. One could propose arguments supporting either a morphological or a syntactic treatment for 劳动保险 láodòng bǎoxiǎn, and never come up with indisputable evidence in favour of one or the other hypothesis. It could be argued that the full forms in (9) and (10) are both phrasal lexemes (see above), which may contain relational elements (such as the conjunction 与 yǔ ‘and’); however, phrasal lexemes typically have some sort of stable, idiomatic meaning, as the Russian form železnaja doroga (‘railway’, lit. ‘iron road’) (Benigni and Masini 2009: 177), while the meaning of (9) and (10) appears to be fully compositional. Here we shall rather draw a borderline between abbreviations such as that in (9), the input of which are just words, and those such as that in (10), the input of which are phrases containing functional elements and, accordingly, lie outside the domain of morphology stricto sensu. Before moving to the issue of how the reduction of compounds works in Mandarin, we must briefly illustrate some facts on the relationship between morpheme, syllable, character and word in the language. The most striking characteristic of the Chinese lexicon is the almost perfect correspondence between the syllable and the written character.

Are reduced compounds compounds? 103

(­汉字 Hànzì) which, in turn, represents a morpheme 90% of the time: we thus see a very strong tendency in Chinese to a matching between a certain unit of phonology (the syllable), a unit of meaning (the morpheme) and a unit of writing (the character): (11) 狗 gǒu ‘dog’, 水 shǔi ‘water’, 书 shū ‘book’ Note, moreover, that there are no morphs smaller than a syllable in Mandarin Chinese, except for the nominalizing (weakly diminutive) suffix -儿 -r. In most cases, thus, the minimal unit of speech (the syllable) and the minimal unit of writing (the character) are endowed with a meaning. Moreover, the inventory of syllables in Modern Mandarin is limited to little more than 400 types, not all of which are attested in all of the four tones. No new syllables types are allowed, and the existing syllables may not be interrupted; hence, as pointed out by Ronneberger-Sibold (this volume), a blend such as English brunch from breakfast and lunch is impossible in Chinese. Blending, thus, is allowed only on the basis of compounds; for this reason, such shortened forms are referred to here as reduced compounds. What does a “typical” Chinese word look like? It is estimated that more than 80 % of the lexical items of Modern Chinese are disyllabic (Shi 2002: 70–72): (12)

电脑 diànnǎo electric-brain ‘computer’

足球 zúqiú foot-ball ‘football’

饮料 yǐnliào drink-material ‘beverage’

Since the vast majority of syllables/characters have a meaning of their own (i.e. they correspond to morphemes), most Mandarin words are complex, made up of two or more morphemes. It is the 字 zì, the character, rather than the 词 cí ‘word’, which matters for the average speaker of Chinese (cf. Myers 2006: 171–172). As Chao (1968: 138) puts it: Whatever conception of the syntactic word we shall find scientifically justifiable to define, it plays no part in the Chinaman of the street’s conception of the subunits of the Chinese language. Thus, if one wishes to ask what the syntactic word xianzai ‘now’ means, one would say: ‘现 在’这 两 个 字 是 什 么 意 思? ‘Xianzai’ zhe liangge zi shi shenme yisi? ‘What is the meaning of these two zi ‘xianzai’?’.5 5. Italics in the original text. The romanization of Chinese words has been altered by the author to be consistent with the 拼音 pīnyīn system used throughout the

104 Giorgio Francesco Arcodia and Fabio Montermini  So, when one builds a reduced form in Chinese, the constituents will be morphemes, rather than mere phonological strings (Yang 2003: 138), as in examples (9–10). Compare: (13)

Italian: Ministero della Cultura Popolare ministry of culture popular ‘Ministry of Popular Culture’

(14)

Chinese: 北 京 Běijīng North-capital ‘Beijing’

→ MinCulPop

图书馆 → 北图 túshūguǎn Běitú picture-book-building North-picture ‘library’ ‘Beijing library’ (adapted from Packard 2000: 271)

In the Italian reduced compound MinCulPop, which used to stand for Ministero della Cultura Popolare (Ministry of Popular Culture) in the Kingdom of Italy, the three “stumps” making up the abbreviation (Min, Cul and Pop) have no meaning of their own.6 The same is true for all the stumps found in Russian SCs, as exemplified in Section 2 above. On the other hand, in 北图 Běitú ‘Beijing Library’ the constituents 北 běi ‘north’ and 图 tú ‘picture’ are actually endowed with a meaning. However, this does not entail that Mandarin reduced compounds will necessarily be compositionally transparent: a reduced form such as 北图 Běitú could have other meanings. The opposite might hold: the reduction of compounds can result in increased opacity and lexicalization. We shall go back to this point later. How are reduced compounds built in Chinese? Basically, two (or, less frequently, three or more) morphemes from the words to be abbreviated are chosen. Several patterns are attested and, according to Ceccagno and Basciano (2009: 116), they are partly motivated by the need for meaning transparency, i.e. preventing (when possible) the opacification described above, and by the need to avoid homonymy with existing words (examples adapted from Yuan 2002 and Ceccagno and Basciano 2009): present article (the official romanization in the P.R.C.). Chinese characters had to be modified as well, since the “official” shape of some of those characters at the time of writing is different from that of 1968. 6. Actually, pop is used nowadays in Italian as a loanword from English meaning ‘popular’, but apparently this was not common before WWII.

Are reduced compounds compounds? 105

(15) 电影 评论 → diànyǐng pínglùn electric-picture criticize-discuss

影评 yǐngpíng picture-criticize ‘film review’

(16) 同济 大学 → Tóngjǐ Dàxué Tongji great-study

同济 Tóngjǐ Tongji ‘Tongji University’

(17) 交通 大学 → 交大 jiāotōng dàxué jiāodà join-communicate great-study join-great ‘Jiao Tong (communication) .  University’ In (15) the characters chosen for the reduced form are 影 yǐng ‘shadow, image, picture’ and 评 píng ‘criticize, review’, which are more relevant for the definition of the meaning of the compound than the “deleted” morphemes, 电 diàn ‘electric, electricity’ and 论 lùn ‘discuss, dissertation, theory’. In (16), since 同济 Tóngjǐ is not a common noun of the Chinese lexicon, it can be used freely to indicate Tongji University; in (17), however, the first part of the name of the institution, 交通 jiāotōng, is also a noun, meaning ‘traffic, communications’ and would thus be ambiguous. To avoid such ambiguity, one morpheme from each of the two components of the name is chosen, 交 jiāo ‘join’ and 大 dà ‘great’, even though they do not contribute to meaning transparency. Note that, unlike the reduced compounds of French and Russian described above, Chinese reduced compounds are typically built from other compounds, rather than from simplex (or derived) words. Such a process has been termed metacompounding by Ceccagno and Basciano (2007: 225); according to their definition, in metacompounding “at least one of the constituents refers to an underlying compound that does not appear on the surface”. We may take (15) as an example: both 电影 diànyǐng ‘film’, lit. ‘electric picture’, and 评论 pínglùn, lit. ‘criticize-discuss’, are compound words, and the reduced compound (metacompound) 影评 yǐngpíng is thus made of truncated forms of the “underlying” compounds. This is especially relevant when one compares reduced compounds to the “ordinary” compounds of Mandarin, as far as meaning and headedness are concerned. Just as in French (cf. ex. 1–2), Mandarin reduced compounds can exhibit anomalous behaviour (ex. adapted from Ceccagno and Basciano 2007: 225; see also Ceccagno and Basciano 2009: 125–126):

106 Giorgio Francesco Arcodia and Fabio Montermini  (18)

分保 fēnbǎo divide-protect/guarantee ‘re-insurance’

The compound in (18), 分保 fēnbǎo, is a noun which is apparently made up of two verbs in a coordinate relationship, and is therefore exocentric. Also, the meaning does not seem to be compositional, i.e. is not easily understood from the meaning of the constituent morphemes. However, if one makes reference to the underlying full forms, it clearly appears that the whole structure is actually endocentric: (19) 分 保险 fēn bǎoxiǎn ‘divide’ ‘insurance’ The underlying structure consists of a verb and a noun, to the right, and is thus an endocentric right-headed noun, which is consistent with all productively formed Mandarin compound nouns. The right-hand constituent, 保险 bǎoxiǎn, is the head also for semantic reasons, since it means ‘insurance’, and the whole structure is (somewhat) compositionally transparent. Thus, an important difference between Chinese reduced compounds and the French blends in (2) is that the latter are structurally odd, and the underlying full words are oddly combined as well, whereas Mandarin reduced compounds seem anomalous on the surface, but once one looks at the underlying full forms, it clearly appears that they are in no way different from “ordinary” compounds. A phenomenon which is well explained by the strong association between syllables/characters and meaning is the “creation” of new morphemes through abbreviation (examples adapted from Packard 2000; Myers 2006: 173): (20) 飞机 → fēijī fly-machine ‘airplane’

机场 jīchǎng machine-field ‘airport’

(21)

来台 台语 láitái Táiyǔ come-platform platform-language ‘come to Taiwan’ ‘Taiwanese (lang.)’

台湾 → Táiwān platform-bay ‘Taiwan’

In (20), the morpheme 机 jī ‘machine’ stands for ‘airplane’ in the compound 机场 jīchǎng ‘airport’; it acquired a new meaning as part of the compound

Are reduced compounds compounds? 107

飞机 fēijī. In (21), the morpheme 台 Tái ‘platform’ acquired the meaning ‘Taiwan’ as part of the compound 台湾 Táiwān and can be used to form new compounds such as 来台 láitái ‘come to Taiwan’. As Myers (2006: 173) puts it, “a Chinese compound defines its component morphemes as much as they define it”. It might be interesting to remark that in Russian, where, as seen above, the reduction of compounds is quite common, a similar phenomenon is found (Billings 1998): (22) professional’nyj sojuz → profsojuz professional union ‘trade union’ In (22), the abbreviated form prof stands for professional’nyj ‘professional’ in the reduced compound profsojuz ‘trade union’. The abbreviated form prof is used in other reduced compounds: (23) prof-sobranije prof-meeting ‘trade union meeting’ In profsobranije, prof- does not stand for professional’nyj, but for profsojuz ‘trade union’. Just like the Chinese examples seen above (20–21), a morpheme has acquired a new meaning from being part of a compound. This phenomenon can also be linked to a similar tendency found in (pseudo-)learned compounds in European languages. It is the case, for instance, of the element auto- in Italian, which has the meaning ‘automobile’ in such words as autostrada (‘motorway’), autoscuola (‘driving school’) or autocisterna (‘tanker’). As pointed out before, one cannot deny that compound reduction may cause a decrease in transparency and, thus, ambiguity. An often quoted example is that of (24) (cf. Chao 1968: 493): (24) a. 人民 代表 大会 → 人大 rénmín dàibiăo dàhuì réndà people representative great-conference person-great ‘General Assembly of the People’s . Representatives’ b. 人民 大学 → 人大 rénmín dàxué réndà people university person-great ‘Renmin (People’s)  University’

108 Giorgio Francesco Arcodia and Fabio Montermini  Two different words or phrases, thus, may be abbreviated into the same form. Let us now turn to the analysis of the phonological and prosodic properties of reduced compounds in Mandarin. As mentioned above, in statistical terms, the typical Chinese word is made up of two syllables. This has not always been the case in the history of the language. The lexicon of Old Chinese (ca. 1200 B.C. – 300 A.D.) had a strong tendency towards monosyllabicity: before 200 B.C. disyllabic words represented roughly 20% of the lexicon, at least in the written style (Guo 1997: 150; Shi 2002: 72). In the evolution towards the modern language, the lexicon has undergone a massive process of disyllabification. Why did this happen? An explanation has been put forward by Feng (1998, 2001) in the framework of Prosodic Morphology (McCarthy and Prince 1998). In Old Chinese (around 1000 B.C.), the minimal syllable was CVC and the maximal one was CCCMVCCC (M = medial), with at least ten possible final consonants. In Middle Chinese (around 800 A.D.), the syllable structure was simplified to CV (minimal) and {C, S} V {C, S} (maximal; S = semivowel), with no consonant cluster allowed in the coda and only two types of final consonants, three nasals and three stops; the final stop consonants did not survive in Old Mandarin (10th–14th centuries; Feng 1998: 224). In the modern language, the minimal syllable can be a vowel, which may be preceded by a semivowel and/or a single consonant; no consonant clusters are allowed, and the only possible codas are [n] and [ŋ] (Yip 2000: 20). In the framework of Prosodic Morphology, the Prosodic Word is the minimal independent unit of prosody, and it is realized by the foot. It is given as a rule that a foot must be at least binary, either under syllabic or moraic analysis (McCarthy and Prince 1993: 43, quoted in Feng 1998: 227). In the transition from Ancient Chinese to Middle Chinese, the minimal syllable became CV, having only one mora, and the bimoraic foot was no longer possible. Words made of one minimal syllable only could not constitute a foot any longer, and the new model for a foot (and, therefore, a prosodic word) was a twosyllable combination (adapted from Feng 1998: 228). This started a process of disyllabification of the lexicon; since each syllable mostly corresponded to a morpheme, the lexicon was consequently flooded with complex words. It does not come as a surprise, therefore, that compounds and phrases are mostly reduced to two-syllable forms, regardless of the “size” of the original expression (Lin 2001: 83; examples adapted from Wang 1998: 61–63): (25) 民用 mínyòng people-use ‘civil’

航空 → 民航 hángkōng mínháng fly-air ‘aviation’ ‘civil aviation’

Are reduced compounds compounds? 109

(26) 高级 中学 → 高中 gāojí zhōngxué gāozhōng high-level middle-study ‘high level’ ‘middle school’ ‘secondary school’ (27) 北 大西洋 公约 组织 → 北约 běi Dàxīyáng gōngyuē zǔzhī Běiyuē north great-west-ocean public-agreement organization ‘North Atlantic’ ‘convention’ ‘organization’ ‘NATO’ So, the output of a reduction process is, in most cases, a minimal Prosodic Word. This is to be taken as a tendency, rather than as a rule, since we have several examples of reduced compounds of three or more syllables (ex. from Yuan and Ruan 2002): (28) 环境 保护 委员会 → 环委会 huánjìng bǎohù wēiyuánhuì huánwēihuì surround-territory protect committee member-union ‘environment’ ‘protection’ ‘committee’ ‘environment pro tection committee’ Thus, the same trends are apparent both in “ordinary” word formation and in abbreviations: two-syllable words are preferred, but words with more syllables are not ruled out. What is most interesting, from our perspective, is that the tendency towards disyllabism in abbreviation often causes an opacification of meaning, somehow in violation of the principle according to which speakers of Chinese languages associate each syllable/character with a meaning. So, for instance, the ambiguity between (24a) 人民代表大 会 ‘General Assembly of the People’s Representatives’ and (24b) 人民大学 ‘Renmin (people’s) University’, both shortened to 人大 réndà, is not resolved by abbreviating the first to, say, *人代大 *réndàidà (person-represent-great), in which all first syllables in the three words are used (Lin 2001: 83). Prosodic trends, apparently, can be stronger than semantico-pragmatic considerations. To sum up, there appear to be fewer constraints on the input of reduced compounds in Chinese than in Russian (cf. Table 3). As far as the output is concerned, however, Mandarin seems to have a strong preference for disyllabic forms, which coincide with the minimal Prosodic Word, in accordance with the general tendency of the language. It remains to be explained, however, why in many cases three syllables rather than two are chosen.

110 Giorgio Francesco Arcodia and Fabio Montermini  4. Concluding remarks In this paper, we presented data on reduced compounds from two non-related languages, namely Russian and Mandarin Chinese, in order to assess the degree of conventionalization of reduced compounding in languages where this phenomenon is very productive. Starting from the distinction between Compounding1, i.e. the human cognitive capacity of semantically linking two words without any grammatical element marking this relation, and Compounding2, i.e. a lexical enrichment process encoded in the morphology of individual languages, we proposed that reduced compounds (at least in certain languages) may be prosodically-based modifications of a nature analogous to that, say, of the adjunction of a linking [o] to the first constituent in Greek compounds, cf. ex. (3c). When such modifications are encoded in the language, reduced compounds do not differ from the “ordinary” compounds of a language, i.e. they are to be regarded as conventionalized, encoded in the morphology of the language. In Russian, reduced compounds appear to be perfectly conventionalized, in terms of their formal and morpho-categorial properties; semantic interpretation, however, mostly depends on pragmatics, due in particular to the fact that head position may not be unambiguously determined. This is explained by the proximity of Russian reduced compounds to syntactic constructions. In short, SCs in Russian appear to be similar to “canonical” compounds. In Chinese, also because of the isolating nature of the language, which has virtually no obligatory morphological markers, the input and output of reduction do not show strong restrictions in terms of formal and morphocategorial properties. Since the reduced compounds of Mandarin mostly coincide with the so-called “metacompounds”, i.e. they are the reduction of other compounds rather than simplex words, the oddness of their structural and semantic properties can be easily explained by making reference to the underlying full forms (see ex. (18)–(19)). However, prosodic constraints operate on the format of the output: regardless of the size of the full form, reduced compounds are most often made up of two syllables, a format that corresponds to the minimal Prosodic Word of Modern Mandarin Chinese. In this sense, Mandarin Chinese reduced compounds are somehow conventionalized, although to a lesser degree, since disyllabism is a tendency rather than a rule. Our data also highlighted the role of the (minimal) Prosodic Word in reduction: whereas in Chinese prosodic constraints operate on the output of a reduction process, in Russian they operate on the input, i.e. on the shape of the “stumps”. Incidentally, we also remarked that the minimal Prosodic

Are reduced compounds compounds? 111

Word is relevant for other phenomena, such as the formation of hypocoristics or of expressive clippings in Russian and, even, the “restructuring” of the Modern Chinese lexicon, where the preferred word-form just coincides with the minimal Prosodic Word. References Bauer, Laurie. 2005. The borderline between derivation and compounding. In Wolfgang U. Dressler, Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and its Demarcations, 97–108. Amsterdam & Philadelphia: John Benjamins. Bauer, Laurie. 2006. Compound. In Keith Brown (ed.), Encyclopedia of Language and Linguistics. 2d ed., 719–726. Oxford: Elsevier. Bauer, Laurie. 2009. Typology of compounds. In Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 343–356. Oxford: Oxford University Press. Benigni, Valentina and Francesca Masini. 2009. Compounds in Russian. Lingue e Linguaggio 8 (2): 171–193. Benigni, Valentina and Francesca Masini. 2010. Phrasal lexemes and reduction strategies in Russian. Poster presented at the 14th International Morphology Meeting. Budapest, 13–16 May 2010. Bertinetto, Pier Marco. 2001. Blends and syllable structure: a four-fold comparison. In Mercé Lorente, Núria Alturo, Emili Boix, Maria-Rosa Lloret and Lluís Payrató (eds.), La Gramática i la Semántica en l’Estudi de la Variació, 59–112. Barcelona: Promociones y publicaciones universitarias. Billings, Loren. 1998. Morphology and syntax: Delimiting stump compounds in Russian. Proceedings of the First Mediterranean Conference of Morphology (Mytilene, Greece, Sept. 19–21 1997), Geert Booij, Angela Ralli and Sergio Scalise (eds.), 99–111. Patras: University of Patras. Booij, Geert. 2009. Compounding and Construction Morphology. In Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 201–216. Oxford: Oxford University Press. Booij, Geert. 2010. Compound construction: Schemas or analogy? A Construction Morphology perspective. In Sergio Scalise and Irene Vogel (eds.), CrossDisciplinary Issues in Compounding, 93–107. Amsterdam & Philadelphia: John Benjamins. Bybee, Joan. 2003. Mechanisms of change in grammaticization: The role of frequency. In Brian D. Joseph and Richard D. Janda (eds.), The Handbook of Historical Linguistics, 602–623. Oxford: Blackwell. Ceccagno, Antonella and Bianca Basciano. 2007. Compound headedness in Chinese: An analysis of neologisms. Morphology 17 (2): 207–231. Ceccagno, Antonella and Bianca Basciano. 2009. Shuobuchulai. La formazione delle parole in cinese. Bologna: Serendipità.

112 Giorgio Francesco Arcodia and Fabio Montermini  Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley & Los Angeles: University of California Press. Comrie, Bernard, Gerald Stone and Maria Polinsky. 1996. The Russian Language in the 20th Century. Oxford: Clarendon Press. Dressler, Wolfgang U.  2000. Extragrammatical vs. marginal morphology. In Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 1–10. München: Lincom Europa. Dressler, Wolfgang U.  2006. Compound types. In Gary Libben and Gonia Jarema (eds.), The Representation and Processing of Compound Words, 23–44. Oxford: Oxford University Press. Feng, Shengli. 1998. Prosodic structure and compound words in Classical Chinese. In Jerome L. Packard (ed.), New approaches to Chinese Word Formation, 196–259. Berlin & New York: Mouton de Gruyter. Feng, Shengli. 2001. 论汉语“词”的多维性 [On the multidimensionality of “word” in Chinese]. Dangdai Yuyanxue 3 (3): 161–174. Fradin, Bernard. 2000. Combining forms, blends and related phenomena. In Ursula Doleschal and Anna Maria Thornton (eds.), Extragrammatical and Marginal Morphology, 11–59. München: Lincom Europa. Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Joseph H. Greenberg (ed.), Universals of Language, 73–113. Cambridge, MA: MIT Press. Grésillon, Almuth. 1984. La règle et le monstre: le mot-valise. Tübingen: Niemeyer. GRJa. 1960. Grammatika russkogo jazyka. Tom I. Fonologija i morfologija. Moskva: Izdatel’stvo Akademii Nauk SSSR. Guo, Xiliang. 1997. 先秦汉语构词法的发展 [Development of word formation in pre-Qin Chinese]. In Xiliang Guo (ed.), 汉语史论集 [Collected essays in the History of Chinese], 131–152. Beijing: Shangwu Yinshuguan. Guevara, Emiliano and Sergio Scalise. 2009. Searching for universals in compounding. In Sergio Scalise, Elisabetta Magni and Antonietta Bisetto (eds.), Universals of Language Today, 101–128. Amsterdam: Springer. Kenneally, Christine. 2007. The First Word. The Search for the Origin of Language. London: Penguin. Lin, Hua. 2001. A Grammar of Mandarin Chinese. München: Lincom Europa. McCarthy, John and Alan Prince. 1993. Prosodic Morphology I – Constraint interaction and satisfaction. Ms., University of Massachussets and Rutgers University. McCarthy, John and Alan Prince. 1998. Prosodic Morphology. In Andrew Spencer and Arnold M. Zwicky (eds.), Handbook of Morphology, 283–305. Oxford: Blackwell. Montermini, Fabio. 2007. Hypocoristiques et minimalité en russe. In Elisabeth Delais-Roussarie and Laurence Labrune (eds.), Des sons et des sens. Données et modèles en phonologie et morphologie, 199–213. Paris: Hermès. Montermini, Fabio. 2010. Units in compounding. In Sergio Scalise and Irene Vogel (eds.), Cross-Disciplinary Issues in Compounding, 77–92. Amsterdam & Philadelphia: John Benjamins.

Are reduced compounds compounds? 113 Myers, Mike. 2006. Processing Chinese Compounds: A Survey of the Literature. In Gary Libben and Gonia Jarema (eds.), The Representation and Processing of Compound Words, 169–196. Oxford: Oxford University Press. Packard, Jerome L. 2000. The Morphology of Chinese. Cambridge: Cambridge University Press. Pesetsky, David. 1979. Russian Morphology and Lexical Theory. Ms., Cambridge, MA: MIT. Ralli, Angela. 2007. Compound marking in a cross-linguistic approach. In Nabil Hathout and Fabio Montermini (eds.), Morphologie à Toulouse. Actes du colloque international de morphologie 4e Décembrettes, 207–220. München: Lincom Europa. Sawer, Michael. 1995. Handling neologisms in teaching and learning Modern Standard Chinese. Australian Review of Applied Linguistics 12: 203–228. Shi, Yuzhi. 2002. The Establishment of Modern Chinese Grammar. The Formation of the Resultative Construction and its Effects. Amsterdam & Philadelphia: John Benjamins. Soglasnova, Svetlana. 2003. Russian Hypocoristic Formation: A Quantitative Approach. Chicago, IL: University of Chicago dissertation. Traugott, Elizabeth C. 2003. Constructions in grammaticalization. In Brian D. Joseph and Richard D. Janda (eds.), The Handbook of Historical Linguistics, 625–647. Oxford: Blackwell. Vaillant, André. 1950. Grammaire comparée des langues slaves. Tome I. Phonétique. Paris: Klincksieck. Wang, Fusheng. 1998. La Formazione delle Parole nella Lingua Cinese Contemporanea. Trieste: Università degli Studi di Trieste – Scuola Superiore di Lingue. Yang, Xipeng. 2003. 汉语语素论 [The Chinese morpheme]. Nanjing: Nanjing Daxue Chubanshe. Yip, Po-Ching. 2000. The Chinese Lexicon. A Comprehensive Survey. London & New York: Routledge. Yuan, Hui. 2002. 前言 [Introduction]. In Hui Yuan and Xianzhong Ruan (eds.), 现 代汉语缩略语词典 [A dictionary of abbreviations in Modern Chinese], 1–14. Beijing: Yuwen Chubanshe. Yuan, Hui and Xianzhong Ruan (eds.). 2002. 现代汉语缩略语词典 [A dictionary of abbreviations in Modern Chinese]. Beijing: Yuwen Chubanshe.

Blending between grammar and universal cognitive principles: Evidence from German, Farsi, and Chinese Elke Ronneberger-Sibold The paper investigates blending in German, Farsi, and Standard Chinese (pŭtōnghuà) on the basis of a typology of blending techniques developed for German in Ronneberger-Sibold (2006). These typologically heterogeneous languages were chosen in order to study the interrelation in blending between (1) individual linguistic systems, (2) certain extralinguistic conditions, such as social conventions, and (3) universal cognitive principles enabling language users to produce and understand blends although these are not accounted for by their grammatical competence. In fact, it can be shown that a technique here named telescope blending (e.g., German Kamelefant < Kamel ‘camel’ x Elefant ‘elephant’), which is most closely related to regular German compounding, is not realized in Farsi and Chinese due to their different grammars, whereas so-called contour blending (e.g., German Tomoffel < Tomate ‘tomato’ x Kartoffel ‘potato’) exists in all three languages, probably because it relies less on German grammar than on the presumably universal human faculty of identifying a word by certain features of its sound shape. This fact shows that these two types are not only structurally different, but also processed in different ways. In addition, it underpins the claim of Natural Morphology that the first stages of language acquisition are governed by universal cognitive principles and that these principles remain at least partly active throughout an individual’s lifetime, showing up, e.g., in deliberate extragrammatical morphological operations such as blends. (Cf., e.g., Dressler and Karpf 1995, Dressler 2000, Kilani-Schoch and Dressler 2005.) Our findings also corroborate the claim made by Kemmer (2003) that classical “building block” approaches to morphological analysis are inappropriate for certain types of blends. Keywords: blending, word creation, contour blend, telescope blend, German, Farsi, Chinese, Natural Morphology.

116 Elke Ronneberger-Sibold 1. Introduction: subject, aims, and material of the study. Definitions This paper summarizes some preliminary results concerning lexical blending from an ongoing research project on word creation in three typologically heterogeneous languages, namely German, Farsi (the national language of Iran), and Standard Chinese (pŭtōnghuà). For illustration, English examples will also be given. In this project, the term word creation refers not only to the coining of entirely new words not based on any previously existing linguistic elements (German Urschöpfung, cf., e.g., Bußmann 2003: 755), but includes all operations for the production of new lexemes which are not covered by regular word formation. Thus, blending is here defined as the deliberate creation of new lexemes from existing ones by creative morphological techniques different from the productive rules (or models) of regular compounding. To put it in a nutshell, blending is extragrammatical compounding. The central aim of this paper is to investigate the interrelation in blending between (1) the linguistic systems of the three individual languages under investigation, (2) certain extralinguistic conditions, such as social conventions, and (3) universal cognitive principles enabling language users to produce and understand blends although these blends are not accounted for by their grammatical competence. The general theoretical background is the claim of Natural Morphology that the first stages of language acquisition are governed by universal cognitive principles and that these principles remain at least partly active throughout an individual’s lifetime, showing up, e.g., in deliberate extragrammatical morphological operations such as blends. (Cf., e.g., Dressler and Karpf 1995; Dressler 2000; Kilani-Schoch and Dressler 2005.) The investigation is performed on the basis of a typology of German blending techniques, in which the different types are arranged according to their relative distance from regular German compounding (cf. RonnebergerSibold 2006). It will be shown that the technique most closely related to regular German compounding is not realized in Farsi and Chinese due to their different grammars, whereas another type, relying less on German grammar than on the presumably universal human faculty of identifying a word by cer­ tain features of its sound shape, exists in all three languages, notwithstanding their grammatical differences. This suggests that at least this type of blend is processed differently from regular compounds. The study thus provides evidence for the extragrammatical character of blending in the sense of Natural Morphology, and at the same time for different grades of extragrammaticality. Moreover, the way in which the processing of the “universal” type of blends is modelled in this paper has much in common with the schema-­based

Blending between grammar and universal cognitive principles 117

approach of Kemmer (2003), although the possibilities of mapping one schema onto another are more constrained in the modelling proposed here. The material for the study has been collected by Mir Kamaladin Kazzazi for Farsi and Stefanie Potsch-Ringeisen for Chinese.1 Their main sources were specialised dictionaries, the Internet, and extensive field work in different cities of Iran and in Beijing.2 The German corpus is described in Ronneberger-Sibold (2006). It consists of 612 blends: 220 are brand names coming from a project on such names, 392 are literary, mostly satirical blends in the widest sense taken from Grésillon (1984). The outline of this paper is as follows: Section 2 contains general observations on the nature and function of blending and the typology of blending techniques mentioned above, applied to the German material. The use made of the different techniques will be explained in Section 3 for Farsi and in Section 4 for Chinese. The results will be compared and the differences explained. Section 5 contains a general conclusion. 2. A typology of blends The typology of blending techniques underlying this study is explained in more detail in Ronneberger-Sibold (2006). For convenience, the most important points are summarized here. 2.1. Preliminary remarks As a rule, blends are less transparent than corresponding regular formations or syntagms.3 In a large number of blends, this is due to shortening or other 1. I am most grateful to both of them not only for the material, but also for their invaluable help in understanding blending in these languages. My thanks are also due to the external experts of the project, Peter-Arnold Mumm (University of Munich) and Wolfgang Behr (University of Zurich), for their advice, as well as to our native Chinese informants, especially to Han Wenting. The responsibility for the contents of this paper is, of course, mine. 2. Their most important sources were Catalogues Groups Iran’s Industries, Exports, and Minerals 1354–1355/1975–1976; Ekrāmi 1385/2006; Farhangestān-e zabān va adabiyāt-e Fārsi 1378/1999; Parchami 2004; Samā‘i 1383/2004 for Farsi, as well as Ling 2002; Liu 2007; Shanghai Daily (ed.) 2007; Xiandai hanyu cidian 2009 for Chinese. 3. Note that transparency is a prerequisite of, but not identical with motivation. Transparency means that a complex word can be semantically interpreted by its constituents and the way these are put together. A transparent word is

118 Elke Ronneberger-Sibold modifications of the blended items. But even if this is not the case, i.e., if the blended words are contained in the blend in full, morphological transparency is reduced because the hearer or reader cannot automatically rely on the regular models of compounding, as, by definition, these are not observed in a blend. This lack of transparency is often implicitly or explicitly considered as a drawback of blends that ought to be kept to a minimum. Contrary to this, in our framework, a controlled reduction of transparency is considered as one of the major aims of blending words instead of compounding them or combining them in a syntagm. Another important motivation is the possibility of creating certain sound shapes which could not be realized by regular compounding. Due to its functional importance, the relative transparency of different types of blends is the leading criterion in the present typology.4 Four types are ordered by decreasing transparency and, by the same token, by increasing distance from regular compounding. They are considered exclusively from the perspective of the producer. This has to be mentioned because, as a consequence of the reduced transparency of word creations, the perspective of the receiver is not simply the converse of the producer’s perspective. For instance, a hearer cannot decide whether a completely opaque, monomorphemic output such as Eduscho, a German brand name for coffee, or Persil, a German brand name for a washing powder, were created by shortening, or blending, or some other technique. In fact, Eduscho is an acronymic shortening of Eduard Schopf, the name of the founder of the firm, whereas Persil is a fragment blend (cf. 2.2.4) < Wasserstoffperoxyd ‘hydrogen peroxide’ x Silikat ‘silicate’, two substances that are active in the washing process. Due to the structural similarity of the outputs for the

motivated if this semantic interpretation, here called its word-formational meaning, matches its lexical meaning. That this is not necessarily the case can be demonstrated by the words blueberry and strawberry. Both are transparent compounds with the literal meanings ‘blue berry’ and ‘berry having to do with straw’, respectively. However, for an average speaker of English, only blueberry is motivated, because he or she knows that blueberries are blue, whereas strawberry is not, unless he or she knows that strawberries are or were often cultivated on beds covered with straw in order to prevent soiling of the fruit. 4. This is an important point in which the typology proposed here differs from others which have influenced it in various other respects. Titles of particular interest for this paper are Marchand 1969, Grésillon 1984, Cannon 1986, Kemmer 2003, Reischer 2008.

Blending between grammar and universal cognitive principles 119

hearer or reader, these types are often subsumed under the same category,5 e.g., under that of clipping-compounds by Marchand (1969: 445–446) or of reduced compounds in Arcodia and Montermini (this volume).6 However, from the producer’s viewpoint, it makes an important difference whether he or she merely optimizes the sound shape of a previously existing source form or combines two words from scratch with respect to meaning and form. Finally, it should be noted that the typology aims at covering only the basic blending techniques, such that every non-basic blend can be interpreted as a variation or combination of the basic techniques. Covering every single blend by a specific type is excluded in principle, given the creative freedom of the techniques involved. Striving for completeness in this sense can only lead to an infinite multiplication of ever smaller classes. 2.2. A typology of blending techniques 2.2.1. Complete blending The most transparent type comprises so-called complete blends, in which both blended words are contained in full. In the simplest case, the blended lexemes are juxtaposed as in a compound, but the end of the first overlaps with the beginning of the second. Blends of this kind are here called telescope blends. They are particularly transparent if the overlapping part is a morpheme of its own, such as Schimmel in Amtsschimmelpilz < Amtsschimmel ‘red tape’ x Schimmelpilz ‘fungus’. If the overlapping part is, however, a sequence of segments without any morphological function, transparency is reduced, e.g., in the example in Figure 1: Kamelefant < Kamel ‘camel’ x Elefant ‘elefant’.

5. Bauer (2003: 47), although keeping them apart in principle, states that “Acronyms tend to merge into blends when more than one letter is taken from each of the words of the title”, adducing German Gestapo (< Geheime Staatspolizei) as an example. 6. From our point of view, only the French examples in this paper are blends, whereas the Russian and Chinese ones are for the most part acronymic shortenings. This is why the internal grammatical structure of the latter is so regular with respect to the distribution of heads and non-heads: it is simply inherited from the regular source forms.

120 Elke Ronneberger-Sibold

Figure 1. Typology of blending

A different, much rarer type of complete blends is characterised by a total overlap, such that one word is entirely included in the other in a so-called inclusive blend, e.g., ComMUNICHation for a communication event taking place in Munich. Of all the techniques contained in our typology, telescope blending is most similar to productive German compounding. In fact, it is the experience gained by decoding regular compounds that provides hearers and readers with strategies for coping with telescope blends. One such strategy is trying to process them in both directions, from the beginning and from the end. For even with regular compounds, both options exist: you can either decode them, so to speak, “online” word by word, or, if they are not too long, you can temporarily store the entire construction as an unanalysed whole and then look for its head first in the final position. This solution suggests itself in particular if, for semantic and/or grammatical reasons, there is a strong expectation regarding a particular head. Note that the nominal head of a German determinative compound determines the gender, number, and case of the entire compound. As these categories are copied on the determiner at the beginning of the NP, its hearers or readers expect a corresponding noun at its end. This is only one manifestation among many others of the German framing principle, pervasive in the entire system of this language and quite frequently requiring that constituents be processed from the end rather than from the beginning. (Cf. Ronneberger-Sibold 1997.)7 7. In this respect, German is a stumbling stone for Hawkins’ “Early immediate constituents” principle; see Hawkins (1994: 402).

Blending between grammar and universal cognitive principles 121

Thus, if hearers or readers attempt to decode, e.g., Kamelefant from the beginning, they will discover Kamel as a possible modifying or coordinate element of a compound, plus an uninterpretable “rest” -efant. If, on the contrary, they try to decode the word from the end, they will discover Elefant as a possible head or coordinate element. This is comparable to the visual perception of picture puzzles, observers of which can activate one shape or another and switch between the two. And just like these observers, after having discovered both shapes, can tell that a dot has, say, the function of an eye in one shape and that of a button in the other, the hearers or readers of Kamelefant, having discovered Kamel and Elefant, will recognize the common stretch -el- and its double affiliation with both constituents. As a test, they can in fact easily expand Kamelefant into a regular compound by spelling out the common element twice, yielding Kamel-Elefant. However, this virtual operation is not indispensable for decoding a telescope blend, for even blends where such a regular expansion is much less acceptable (such as ? Amtsschimmel-Schimmelpilz) can be understood. In sum, German language users approach telescope blends by the same strategies as regular compounds, i.e., they expect a construction made up of two words in a determinative or coordinate relation. This is different with the following type, here termed contour blending. 2.2.2. Contour blending At first sight, a contour blend such as Tomoffel < Kartoffel ‘potato’ x Tomate ‘tomato’ consists of one word only, here called the matrix word. In our example, the matrix word, which is primary for the analysis of the blend, is Kartoffel. Although it is not entirely included in the blend, it can be traced, so to speak, by several phonological features of high importance for its recoverability. In German, these are firstly its overall rhythmical contour defined by its number of syllables and the place of its main stress, i.e. -ˈ--, secondly its stressed vowel, here -o-. Taken together, these features define the abstract sound shape (or gestalt) of Kartoffel.8 While it is true that different 8. It is probably not by chance that the same elements also play a crucial role in other conscious or unconscious operations based on the similarity of sound shapes, such as rhyming, folk etymology (Ronneberger-Sibold 2002), slips of the ear and mondegreens (Bond 1999, Ronneberger-Sibold 2010), and even the imitation of brand names (Ronneberger-Sibold 2001). Even in German inflectional and derivational morphology, important grammatical and lexical information is encoded by a modification of the syllabic shape of base forms

122 Elke Ronneberger-Sibold words may have the same abstract sound shape, e.g., Kartoffel ‘potato’ and Pantoffel ‘slipper’, such ambiguities are normally resolved by the context. Thus, in a text or a conversation about vegetables, Kartoffel will be activated, whereas in a context dealing with shoes or homewear, Pantoffel will be the leading association. The second word, in our example Tomate, is partly inserted into this sound shape, thereby replacing part of the matrix word. Most typically, the beginning of the matrix word is replaced by the beginning of the inserted word.9 In this case, the insertion may maximally fill the entire pretonic part, as in Tomoffel. Possible overlaps between the matrix word and the inserted word enhance the transparency of the blend, but are not indispensable. Another possible point of insertion is the rhyme of the matrix word, as in Konkurz ‘short bankruptcy’ < Konkurs ‘bankruptcy’ x kurz ‘short’. In German, this type is rare, because it impairs the identities of the rhymes of the matrix word and the blend. However, more comparative research is needed in this respect. In spite of this latitude concerning the point of insertion, there seems to be one fundamental constraint: whether the second word is inserted in the pretonic part of the matrix word or in its rhyme, it must not span the boundary between the two, which is immediately before the stressed vowel (unless, of course, the stressed vowels of both words overlap). This can be easily demonstrated by a minimal blend such as brunch (< lunch as the matrix word and breakfast as the inserted word). If we tentatively replace the stressed vowel of lunch by the /ɛ/ of breakfast, the result brench would no longer be transparent with respect to both words, because the rhythmical contour and the stressed vowel of the blend would no longer come from the same constituent. This example also demonstrates the way such blends are decoded by hearers or listeners: once they have identified the matrix word by its sound shape, this word opens up, so to speak, a semantic frame, which allows them to recognize the second word inside this sound shape.

(through syllabic affixes) and of the stressed vowel (through umlaut and ablaut). Apparently, many generations of language users have unconsciously selected these forms from a large number of variants, because of the perceptual salience of grammatical and derivational categories contained in them. 9. In some cases, such as Engl. askillity (< ability x skill), cited by Soudek (1978: 464), the onset of the inserted word does not replace the beginning of the matrix word, but the onset of its stressed syllable. This is what Soudek calls “implanted blends”. I am not aware of any clear example of this type in German, Farsi, or Chinese.

Blending between grammar and universal cognitive principles 123

Thus, the fundamental difference between telescope blends and contour blends lies in the way the blended words are combined: in a telescope blend, they are juxtaposed, as in a compound, though with an overlap, whereas in a contour blend, one word is inserted into the abstract sound shape of the other. This difference entails corresponding differences in the way such blends are de­­coded by language users and hence in their appropriate linguistic description. As pointed out, language users create and interpret telescope blends using the regular models of compounding, though in a somewhat modified manner, in languages where such models are available. Therefore, telescope blending can be described in a classical “building block” model of morphology (the term is taken from Kemmer 2003: 69), e.g., Kamelefant as Kamel + Elefant with an overlap in -el-. Contrary to this, describing Tomoffel in the same “building block” manner, i.e., as Tomate + Kartoffel with back-clipping of Tomate > Tom and fore-clipping of Kartoffel > offel would not only be ad hoc, but also miss the decisive point, which is that Tom- fills the pretonic part of the rhythmic contour of Kartoffel. Such a relation can only be described by using abstract sound shapes or schemas, much as in the approach of Kemmer (2003). The difference between her schema-based analysis of the blend swooshtika < swastika x swoosh (Kemmer 2003: 82) and our analysis of comparable contour blends is that, for Kemmer, all the phonetic features of a word contribute to its phonological schema, whereas in our approach, some features, such as the rhythmic contour and the rhyme, are more important in identifying a sound shape than others. In this sense, our notion of “sound shape” is more abstract than Kemmer’s notion of “schema”, because in our sound shapes certain features are in fact taken away (literally abstracted) from reality. Abstracting the sound shape from a word and using this sound shape to identify the word in oral communication, e.g., in noise, when not every detail can be perceived, seems to be a universal ability.10 Thus, in recovering the matrix word of a contour blend via its sound shape, language users need not rely on their acquired linguistic competence, but on an ability which they have preserved and continuously used from their earliest childhood on. This might be a factor explaining why contour blending seems to be far more widespread than telescope blending in different languages. To put

10. There is evidence that even very young children imitate the typical rhythmical contours of their surrounding linguistic input, long before producing their first words (cf. Friederici, Friedrich, and Christoph 2007).

124 Elke Ronneberger-Sibold it in a nutshell, telescope blending is play with grammar, whereas contour blending is play with sound shapes.11 2.2.3. Semi-complete and fragment blending Semi-complete and fragment blending will play a minor role in the following. Therefore, they are merely presented here without extensive discussion. In a so-called semi-complete blend, one word is contained in full, the other only in fragments, but, unlike the matrix word of a contour blend, the reduced word does not provide the contour and often not even the rhyme of the blend. Therefore, it is less easily recovered, unless the retained string is very long and distinctive. Who would, for instance, recognize German Silber ‘silver’ in Novasil, the name of a product in the class of metalwares, or Assam in Assindia for a tea? The end of the scale is reached with completely opaque, so-called fragment blends such as the above-mentioned Persil < Wasserstoffperoxyd ‘hydrogen peroxide’ x Silikat ‘silicate’. As exemplified by this name, this technique is mostly used for the deliberate creation of pseudo-exoticisms characterized by their typical foreign sound shape. Like semi-complete blends, fragment blends can be relatively transparent only if long and distinctive strings of the blended words are retained in the blend, e.g., in Dynaject, the name of a syringe, from German dynamisch or English dynamic x to inject. 2.3. Incidence of the different types in German In German, all the types of blending listed so far are quite active, preferably in lexical domains in which reduced transparency is communicatively functional, such as brand names, the language of advertising, humorous (especially satirical) texts, children’s literature, and all kinds of jargons which need not or are in fact not meant to be understood by outsiders, e.g., juvenile slang. The degree of transparency aimed at in each blend depends 11. Both types are combined in a variant represented by German Kurlaub (< Kur ‘cure’ x Urlaub ‘holidays’), describing holidays which are combined with a cure. On the one hand, Kur is integrated in the rhythmic contour of Urlaub as in a contour blend; on the other hand, not only the inserted word, but also the matrix word Urlaub is contained in the blend in full, as in a telescope blend. This variant seems to be particularly frequent in English; cf. glasphalt < glass x asphalt, wargasm < war x orgasm etc., cited by Bauer (2003: 46), and glitterati < glitter x literati etc., cited by Kemmer (2003: 73).

Blending between grammar and universal cognitive principles 125

on its communicative function. This is evidenced by the German corpus underlying our study. It is divided between brand names and literary blends in the widest sense, mostly coming from satirical texts. 66.3

  satirical creations  brand names 30.6

31.8

26.8

16.4 2.6 complete . blends

contour . blends

semi-complete . blends

25 0 fragment . blends

Figure 2. Relative frequencies of blending types in German satirical creations and brand names (in percent)

Approximately 30% of the brand names and the literary blends are of the contour type, but the remaining 70% are distributed most unevenly among the two groups: in the literary blends, there is a strong preference (66.3%) for complete blends, especially for the most transparent telescope type. In the brand names, on the contrary, the semi-complete and fragment types are preferred (51.8%) while being practically non-existent (2.6%) among the literary blends. The reason for this difference is that literary blends, in order to fulfil their amusing function, have to be understandable, hence the preference for the most transparent telescope type. In contrast, in brand names, which, for commercial and legal reasons, ought not to describe a product, but only to label it an evocative way, the more opaque types such as Persil or Assindia are functional. Contour blends seem, in a way, to play a role of their own, for in spite of their transparent character, their degree of preference is the same in both groups. 3. Blending in Farsi 3.1. The facts In Farsi, blending is relatively rare, as compared to German. Our entire corpus consists of 27 instances. The most important lexical domains are different kinds of juvenile slang, heavily influenced by English and French, and – to a lesser degree – brand names. A few blends are also included in the general

126 Elke Ronneberger-Sibold lexicon in colloquial use. No blends were found in literary or journalistic texts. The reason is probably a fairly normative attitude of Iranian society towards the national language, which is considered as a cultural heritage in need of preservation rather than as a tool for communication allowing for and even requiring constant adaptation to changing communicative needs. While these extralinguistic restrictions on blending do not come as a surprise, the distribution of the blending techniques is quite unexpected from the German point of view: of the 27 blends, 21 (i.e. 77.8 %) are of the plain contour type, e.g., Tehrānğeles12 /tehrɒnʤeles/, with Los Angeles as the matrix word and Tehrān /tehrɒːn/ as the inserted word, referring to ‘a Los Angeles which resembles Teheran’ (because of the large Iranian minority), or ğāf /ʤɒːf/ ‘prostitute’, with dāf /dɒːf/ ‘girl’ as the matrix word and ğelf /ʤelf/ ‘sexy’ (with negative connotation) as the inserted word in juvenile slang. Among the remaining six blends, there are three inclusive blends, as well as one semi-complete, one fragment, and one modified contour blend. In sum, the analysis of blending techniques in our Farsi corpus has yielded two remarkable results: a) Our corpus does not contain any telescope blends, although all the other types of blends are represented in our material. b) There is a very strong preference for contour blends. In the following, an attempt is made to explain these findings. 3.2. The absence of telescope blends Theoretically, the apparent absence of telescope blends in our material can be explained in three different ways: a) It might be due to chance, given the relatively small size of our corpus. b) It might be due to extralinguistic reasons: as suggested by the German material, telescope blends are primarily used in literary texts in the widest sense. As, for the reasons outlined above, blending – and for that matter word creation in general – is not considered appropriate for such texts by Iranian society, there may simply be no occasion to use telescope blends. (Nevertheless, one could expect at least some instances in the slang of young people, who are usually less hostile to word creation than older generations.) 12. The transcription follows a widespread scientific standard. Cf., e.g., Enzyklopaedie des Islam 1913–1936.

Blending between grammar and universal cognitive principles 127

c) There might be an intralinguistic reason, i.e., some feature in the grammar or lexicon of Farsi which blocks the creation of telescope blends. If a) and b) are correct, language users should in principle be capable of producing or at least understanding telescope blends, even if they regard them as inappropriate, ridiculous, or simply ugly. On the contrary, c) would imply that language users would have extreme difficulties in handling telescope blends both in perception and production. As a preliminary, informal test, we therefore asked our native Iranian collaborator to interpret a small sample corpus of five telescope blends invented by our German linguistic expert for Farsi.13 All blends were of the “easy” type where the common element is a morph, e.g., *ham-zabān-šenās ‘a linguist who is a good friend’ (< ham-zabān ‘good friend’, literally ‘with-language/way of thinking, i.e., somebody who shares my language, my way of thinking’ x zabān-šenās ‘linguist’, literally ‘language-scholar’). The semantic structure of the blends was either determinative, as in this example, or copulative, as in *sang-del-tang ‘hard-hearted and sad’ (< sang-del ‘hard-hearted’, literally ‘stone-heart’ x del-tang ‘sad, disheartened’, literally ‘heart-narrow’). Although our native Iranian informant is a trained linguist working on blends and although he knew what the discussion was about, he felt unable to think of a semantic interpretation for these extremely “easy” blends – from a German point of view. The only cases which seemed interpretable to him (though not as blends) had a possible regular reading in addition to their reading as a blend, such that he could interpret them along the lines of the regular reading.14 An example is *sar-bāz-did ‘control of soldiers’ (< sar-bāz ‘soldier’, literally ‘head-risk, i.e. somebody who risks his head’ x bāz-did ‘control’, literally ‘again-see, i.e. re-view’). Our informant interpreted this blend as a regular formation meaning ‘main control’ with the productive prefixoid sar- ‘main’ (derived from sar ‘head’).15

13. I am grateful to Peter-Arnold Mumm (University of Munich) for inventing the blends and to Mir Kamaladin Kazzazi for acting as an informant. 14. Only in the case of one blend with the very clear and concrete determinative meaning ‘basin for melted snow’ could he imagine that perhaps some hearers or readers might detect the possible reading as a blend. The word is *barfāb-gir < barf-āb ‘melted snow’, literally ‘snow-water’ x āb-gir ‘basin’, literally ‘water-take’. 15. Bāz ‘risk’ and bāz ‘again’ are homonyms with different etymologies.

128 Elke Ronneberger-Sibold Although more native speakers are needed in order to confirm our informant’s impression, his reaction clearly points to a language-internal, systematic reason for the absence of telescope blends in Farsi. As explained in 2.2.1., in processing telescope blends, German language users rely on their strategies developed for regular compounds. Farsi language users, on the contrary, do not develop such strategies, because plain compounds of two juxtaposed words are no longer productive. The lexicon does contain transparent words of this type, e.g., dastband ‘bracelet’ (< dast ‘hand’ + band ‘string, ribbon’ (Majidi 1990: 410)), but such nouns are completely lexicalised, and hence retrieved as wholes. The productive model of a determinative construction is the syntactic so-called ezafe construction, in which the head precedes its modifier, and the two are linked by a subordinating element -e, termed ezafe, which is phonologically attached to the head, e.g., dar-e χāne ‘front door’, literally ‘door-ezafe house’ (Majidi 1990: 70). This construction is, however, not confined to the equivalent of English or German compounds, but has to be applied (recursively, if necessary) in any determinative noun phrase between all heads and their modifiers, e.g., kif-e bozorg-e Sussan ‘Sussan’s big bag’, literally ‘bag-ezafe big-ezafe Sussan’ (Majidi 1990: 71). As evidenced by this example, the ezafe construction is syntactic rather than lexical. The same is, mutatis mutandis, true for the productive equivalents of German or English copulative compounds, the members of which are linked by -o- ‘and’, e.g., kot-o-šalvār, ‘suit’, literally ‘jacket and trousers’. Therefore, Farsi language users cannot rely on any productive model of compounding in the production or interpretation of telescope blends in the same way as German language users do. This is most probably the structural reason why telescope blending is systematically blocked in Farsi. 3.3. The preference for contour blends As explained in 2.2.2, contour blending does not depend on regular compounding in the same way as telescope blending does, but is rather based on the universal ability to identify words by their sound shapes. This makes contour blending available for Farsi language users, contrary to telescope blending. However, semi-complete and especially fragment blending do not rely on any regular model of word formation either. In fact, our material contains some instances of these techniques, which proves that they are possible in Farsi. So why are they so much dispreferred as compared to contour blending? The answer is most probably: because they lack transparency. According to Seiler (1975), languages can be typologically described and ordered

Blending between grammar and universal cognitive principles 129

according to the degree of descriptiveness of their words: there are languages in which things, persons, acts, etc. are preferably designated by descriptive terms, and others where opaque labels are preferred for the same purpose. Although, in any language, both types are normally available and distributed according to their usefulness in different domains in the lexicon, there is a typological bias. In the case of Farsi, this bias is clearly in favour of the descriptive type. In fact, the relative hostility of Farsi language users towards blending (as to word creation in general) might be motivated not only by the conservative attitude mentioned above, but also by this typological bias, although language users are, of course, less aware of this motive than of their metalinguistic reflections and preferences. In any case, if blends are created and accepted by the speech community at all, they should be as transparent as possible. As telescope blending is not possible, the most transparent type available is contour blending. Although descriptiveness is, in principle, a parameter of its own, i.e., one which is not identical to any other typological parameter, there may be functionally or historically motivated preferences for certain clusterings, e.g., between a high degree of descriptiveness and of analyticity. This should be kept in mind in the following discussion of blending in Chinese. 4. Blending in Chinese 4.1. Linguistic and extralinguistic background Contrary to the Iranian speech community, the Chinese one is quite fond of word creation. In humorous texts, in languages for special purposes, especially in the terminology of administration, but also in daily life and above all in the new media – e-mail, online chat, SMS text messaging, etc. – Chinese speakers welcome every technique which allows circumventing or abbreviating the complicated encoding of a message in Chinese logographic characters, especially if the result is witty and playful.16 This need is predominantly fulfilled by shortening, and far less by blending. In fact, blends in the sense defined here are possibly even rarer than in Farsi. So far, we have found only 17 of them, out of approximately 16. Messages in Chinese script are encoded electronically in the following way: the text is first written morpheme by morpheme in pīnyīn, the official transcription system into Latin script also used in this paper. The program then offers a list of possible Chinese characters for each morpheme from which the writer chooses the appropriate one.

130 Elke Ronneberger-Sibold 600 Chinese word creations. There are structural reasons for this in the Chinese language; they will be explained below. From a typological point of view, Chinese is often considered as an isolating language (Norman 1988: 10). Taken in its strictest sense, this would exclude blends from Chinese altogether, because the language would be lacking complex words. Whereas this may have been the case to a greater extent in its historical stages, Modern Chinese does contain complex words. The most common word pattern is a determinative compound of two morphs, e.g., dàxué 大学, ‘university’, literally ‘big learning’. Packard (2000) devotes his entire book to the proof that Chinese syntax operates on such words instead of individual morphemes as in an ideal isolating language (Packard 2000: 1). While this is undoubtedly true in most cases, the analytical character of Chinese nevertheless manifests itself in several respects, two of which are important for blending. These are, first, the internal structure of words, and secondly, the degree of cohesion of compounds. 4.1.1. The internal structure of Chinese words The vast majority of Chinese words is entirely made up of invariable, monosyllabic morphs represented by one logographic character each, e.g. dà ‘big’ 大. These units of one syllable representing one morph, which in turn is represented by one logographic character, cannot be modified by any morphological operation, i.e. (putting aside some tone changes which might be considered as indicative for the functioning of certain morphs as nouns or verbs) there are neither morphophonemic, nor paradigmatic alternations (Packard 2000: 130). Even purely phonologically determined sandhi phenomena are kept to a minimum. Most of them are centered around weakly stressed syllables (Norman 1988: 148–149). Thus, with very few exceptions, the Chinese units of one syllable-morph-character can be neither split up, nor extended, nor otherwise modified for grammatical purposes. A further consequence is that (again with very few exceptions) there are no polysyllabic morphs in the native lexicon. Moreover, due to severe constraints on possible onsets and codas, the number of theoretically possible syllables is fairly low, and as not even all possible syllables are in fact realized, pǔtōnghuà contains only 420 different syllables (Chen 1999: 36). Although they can be further differentiated by four tones (a possibility which, again, is not fully exploited in the lexicon), it is clear that there must be numerous homophonous morphs differing in the written code only. In fact, this is one of the reasons for the maintenance of the Chinese script in spite of its obvious practical disadvantages. In the spoken

Blending between grammar and universal cognitive principles 131

language, the ambiguity of morphs is mostly resolved by their context, i.e., by being part of words and, as far as words are concerned, by the utterance in which they occur (cf. Packard 2000: 287–292). Thus, Chinese hearers and readers normally have to take into account even larger portions of a text in order to decode it than German ones. To put it in a nutshell, in German, you make sense of individual words or sometimes even morphemes, but in Chinese of whole utterances. This difference is nicely expressed by the metaphor of close fit vs. loose fit by Hawkins (1986: 121–127) with respect to German and English: in a close-fit language, the hearer can reconstruct the meaning of an utterance by “locally” proceeding from one relatively autonomous word to the next one, whereas in a loose-fit language, the hearer has to proceed from what the entire utterance presumably means. Clearly, English is more loose-fit than German, but Chinese is far more loose-fit than English. 4.1.2. The cohesion of Chinese compounds The distinction between compounds and syntagms is a notorious problem in Chinese linguistics, especially with respect to certain VO compounds (cf. Chao 1968: 415–434 and the summary of that chapter and the ensuing literature in Packard 2000: 109–115). Among the various criteria which have been proposed, two are particularly relevant to blending. These are the idiomatisation of a compound (such as in dàxué, cited above) and the use of bound morphemes within a compound. As in many other languages, Chinese morphemes can be classified as free or bound. Free morphemes can occur as independent words, e.g., dà 大 ‘big’ in dà fángzǐ 大房子 ‘big house’, or in compounds, as in dàxué 大学. Bound morphemes cannot occur as independent words; they have to be combined with at least one other morpheme in a compound (or a derivation, for some of them have the status of affixes, a question which need not be discussed here). For instance, shēng 生 ‘to give birth, to grow, to live, life, alive, unripe, raw, student, strange’ (an extremely polysemous morpheme) can only be used in compounds, e.g. 学生 xuéshēng ‘student’, literally ‘learning student’. Chinese contains a large and ever-growing number of bound content morphemes (also called bound roots), such as shēng, even in its native lexicon (Packard 2000: 280–283), whereas in German or English, this phenomenon is largely confined to the Classical and Neoclassical part of the lexicon. (Cf. the so-called combining forms such as therm, meaning ‘warm’ or ‘warmth’, but occurring in formations such as thermometer or endotherm only.) Of course, a Chinese construction containing such a bound root must be a compound.

132 Elke Ronneberger-Sibold However, the situation is complicated by a third class of morphemes, which are bound in one function (typically when they are used as nouns), but free in another (typically when used as verbs). For instance, xué is bound when it means ‘learning’,17 but free when it means ‘to learn’, e.g., in the sentence 我学中文 wǒ xué zhōngwén ‘I learn Chinese’. (Of course, like any free morpheme, it is also allowed in compounds, when it is used as a verb.) 4.2. Blending in the native lexicon of Chinese The structural features of the Chinese language explained so far present an obstacle to blending in several respects: first, blends implying the disruption of syllables are systematically excluded, e.g., a blend such as English brunch would be totally impossible, because the syllables lunch and break are disrupted. Secondly, blending must not result in the creation of new syllables. Therefore, brunch would also be excluded on the grounds of being a new syllable (let alone the fact that its initial and final consonant clusters would make it unpronounceable anyway). Thirdly, in the native lexicon, any kinds of blends relying on polysyllabic morphemes are impossible, for there are (almost) no such morphemes. Therefore, even a creation such as German Kamelefant would be out of the question, because it is built on the two polysyllabic morphemes Kamel and Elefant. (In fact, the blend would also be impossible because of the resyllabification of syllable-final /‑l/ in /(ka)mel/ to syllable-initial /l-/ in /le(fant)/). As a consequence, blending in the native lexicon is only possible on the basis of compounds the members of which must stay intact phonetically and – as far as possible – also semantically. Moreover, even the grammatical structure of the resulting blend plays an important role.18

17. In this function, it is even approaching the status of an affix, comparable to German -logie, English -logy in neoclassical formations, such as 心理学 xīnlǐxué‚ ‘psychology’, literally ‘heart reason learning’, 地质学 dìzhìxué ‘geology’, ‘earth nature learning’, etc. (Norman 1988: 156) 18. All three conditions are met in highly transparent constructions named blends in the Chinese literature on word formation (Sun 2009), where a common element occurring in two coordinated compounds is left out. For instance:

Blending between grammar and universal cognitive principles 133

4.2.1. Chinese telescope blends? As the overlapping parts in Chinese telescope blends have to be morphemes, the only possible candidates resemble German Amtsschimmelpilz ‘red tape’, which can be read either as a telescope blend (< Amtsschimmel ‘red tape’, literally ‘office mould’ x Schimmelpilz ‘mould’, literally ‘mould fungus’, with Schimmel belonging to both blended words) or as a regular determinative compound consisting of Pilz or Schimmelpilz as head and Amtsschimmel or Amts- respectively as determining member. Likewise, Chinese compounds such as 大 学 生 dà xué shēng big learning student ‘university’ ‘student’ could theoretically be interpreted either as a regular compound containing shēng ‘student’ as head and dàxué ‘university’ as determining member, or as a telescope blend of 大 学 学 生 dà xué x xué shēng big learning learning student ‘university’ ‘student’ However, according to our informants, Chinese language users concentrate on the regular reading only, not even considering the interpretation as a

早 餐 午 餐 zăo cān x wŭ cān early meal midday meal ‘breakfast’ ‘lunch’

早 午 餐 zăo wŭ cān early midday meal ‘brunch’

Such constructions are felt to be irregular for semantic reasons, i.e., because their readings as blends deviate from their normal readings as left-branching determinative compounds, e.g., ‘an early midday meal’. Even though this regular meaning is not the intended one, the fact that it is structurally possible is crucial for the acceptance of such blends.

134 Elke Ronneberger-Sibold blend as an alternative, let alone as an amusing one, and this although the construction dàxué xuéshēng is possible and even attested (though not as a lexicalised compound). This stands in sharp contrast to German language users, who find the reading as a telescope blend amusing as soon as they have discovered the possible overlap in Schimmel. The most probable reason for this difference is the structural inadequacy of a morpheme belonging to two members of a compound at the same time. Such a possibility is not even taken into consideration by Chinese language users, as soon as a regular reading is available. The importance of a regular reading is confirmed by the following invented example: 学 生 xué shēng x learning student ‘student’

生 活 shēng huó living living ‘life’

学 生 活 xué shēng huó *‘students’ life’

For our informants, this construction was not interpretable as a telescope blend, even if they knew this type of word creations from German. Instead of considering the possibility of connecting shēng with its left and right neighbours at the same time, they chose the only possible regular interpretation as a VO compound with xué as a verb: ‘to learn living’. The reason is that both shēng and huó are bound morphemes. Hence, huó has to be attached to shēng to form the compound shēnghuó, leaving, so to speak, xué standing alone. As explained above, xué can only function as a free morpheme in its verbal meaning. If no regular reading is available, artificially constructed telescope blends are simply not understood by language users. Thus, Chinese, like Farsi, does not allow for telescope blends, if for different reasons: corresponding regular patterns are not lacking, but, on the contrary, they are so strong that they supersede, so to speak, possible readings as blends. 4.2.2. Native Chinese contour blends If contour blends are based on universal cognitive abilities, they should be possible even in Chinese. This is indeed the case. Of course, the integrity of the syllable-morpheme-character and even a superficial structural regularity

Blending between grammar and universal cognitive principles 135

must be respected in contour blends, but their deviant semantic structure signals their status as blends to the hearer or reader. In the creation of Chinese contour blends, as in German, Farsi, and many other languages, pretonic morphs may be replaced by others. As in Farsi, this is easy from a prosodic point of view, because, at least in pausa, Chinese compounds are stressed on the last syllable (Norman 1988: 148).19 A rather intricate and witty example is the following: 旅 友 驴 lǚ yoǔ x lǘ ‘travel’ ‘companion’ ‘donkey’ 驴 友 lǘ yoǔ ‘backpack tourist, i.e., a travel. companion loaded like a donkey’ The matrix word 旅友 lǚyoǔ ‘travel companion’ is a regular compound formed after the model of, e.g., gōngyoǔ ‘workmate’, zhànyoǔ ‘wartime comrade’, etc. In the blend, the pretonic morpheme lǚ ‘travel’ of lǚyoǔ has been replaced by the almost (except for the tone) homophonous morpheme 驴 lǘ ‘donkey’, yielding lǘyoǔ 驴友. As a regular compound, this word would mean ‘companion of a donkey’ without any reference to the concept of ‘tourism’. It can only be interpreted in the intended way as ‘backpack tourist’ by recovering the matrix word lǚyoǔ ‘travel companion’, i.e., by interpreting the construction as a contour blend. Thus, as in Farsi, Chinese language users can easily perform an operation in contour blends which they categorically refuse in telescope blends, namely recovering a part of a word that is covered up, so to speak, by another word. This common preference for contour blends in spite of otherwise important structural differences between Farsi and Chinese points again to a principled difference in processing telescope and contour blends. Following a few remarks on other blending techniques in the native lexicon, this will be confirmed by our findings on non-native blends.

19. A different view, based more on theoretical considerations than on phonetic description, is expressed in Duanmu 2007.

136 Elke Ronneberger-Sibold 4.2.3. Other blending techniques Fragment blending occurs in our native material only if the blend coincides with a previously existing word which assures its structural regularity and helps to motivate it semantically, i.e., if the blend is a so-called prop lexeme, roughly like the English acronym CARE < Cooperative for American Remittances to Europe.20 The following example combines fragment and inclusive blending in quite a humorous way: 白 领 bái lǐng x white collar ‘white collar’

骨 干 精 英 gǔ gàn x jīng yīng bone spine select hero ‘backbone’ ‘elite’



白 骨 精 bái gǔ jīng ‘white bone ghost’ 21

The fragment blend báigǔjīng ‘white bone ghost’ coincides with the name of a female demon in the well-known classical novel The Journey to the West. Therefore, it is a prop lexeme. However, in this particular example, the meanings of the source words are not completely superseded by the meaning of the host word, but are rather included in the following meaning of the blend: ‘a successful ( jīngyīng ‘elite’), tough (gǔgàn ‘backbone’) woman (báigǔjīng ‘white bone ghost’) in a high professional position (báilǐng ‘white collar’)’. Due to this combination of meanings of the source words and the host word, this creation is considered as an inclusive blend with the special feature that the blended words overlap entirely, rather than as a simple prop lexeme. 4.3. Blending on the basis of English loanwords Loanwords from languages with complex syllable structures, such as English, are phonetically adapted to Chinese by expanding unpronounceable consonant clusters into Chinese syllables. By definition, these syllables are morphemes, written by the corresponding characters, but unless semantically appropriate morphemes happen to be available, the meanings of the morphemes, so to speak, do not “count”. That is, they function like the characters of a syllabic writing system. As a consequence, such loan words 20. The term prop lexeme is taken from Ungerer 1991. 21. ‘Ghost’ is another possible textual meaning of jīng.

Blending between grammar and universal cognitive principles 137

are polysyllabic morphemes. These are the basis for contour and fragment blends, as in the native vocabulary. 4.3.1. Contour blending on the basis of English loanwords A simple example is the following. The English word shampoo is adapted to Chinese as 香波 xiāngbō, literally ‘scent wave’. (As luck would have it, it was possible to find these morphemes in the same semantic frame as that of shampoo.) As a Chinese equivalent for English shower gel, a contour blend was created with xiāngbō as the matrix word and 浴 yù ‘to bathe’ as the inserted word, yielding: 香 波 浴 xiāng bō x yù scent wave bathe ‘shampoo’

浴 波 yù bō bathe wave ‘shower gel’

As in the native blend 旅友lǘyoǔ ‘backpack tourist’ cited above, the referential meaning of the blend, in this case ‘shower gel’, is not directly motivated by its literal meaning ‘bathe wave’, but only by the referential meaning of the matrix word xiāngbō ‘shampoo’. In order to make a polysyllabic matrix word maximally recoverable, there is a tendency to choose the place where the other word is inserted in its pretonic part in such a way that there is maximal similarity between the inserted word and the syllable that disappears in the matrix word. (In our material, only monosyllabic words are inserted.) This place is not necessarily the first syllable, as demonstrated by the following example. The English acronym dinks < double income, no kids for a type of well-off young couple who spend their double income exclusively for themselves is expanded into the three Chinese syllables dīng kè zú 丁克族, meaning ‘nail, gram, ethnic group’ respectively. ‘Nail’ and ‘gram’ do not “count” (which is symbolised by the asterisks in the representation below), but zú 族 ‘ethnic group’ is quite appropriate as a suffix characterising a group of people. Now, instead of a child, dinks often have a dog. In English, such a couple is described by simply extending the acronym dinks into dinkwads (< double income no kids with a dog). This would be very difficult to adapt to Chinese

138 Elke Ronneberger-Sibold phonetically. So, the problem was solved by creating a contour blend of the trisyllabic morpheme dīngkèzú as the matrix word and Chinese 狗 gǒu ‘dog’ as the inserted word: 丁 克 族 x 狗 dīng kè zú gǒu *nail *gram ethnic group dog ‘dinks’ 丁 狗 族 dīng gǒu zú *nail dog ethnic group ‘dinkwads’ Here, the inserted word gǒu ‘dog’ replaces the second syllable of the matrix word instead of the first. According to our informants, it is the common seg­­ mental feature [+ velar] in the onsets of kè and gǒu plus the common prosodic feature [+ open syllable] which motivates this point of insertion, which – to my knowledge – is neither attested in Farsi, nor in German or English. The fact that, in Chinese, the syllable of the matrix word which is to serve as the point of insertion can be chosen according to its individual similarity with the inserted word, and not only according to its initial distribution in the matrix word, is well in line with the general, typologically conditioned importance of the syllable as the basic element of the linguistic structure of Chinese. Although Modern Chinese undeniably contains words, these do not play the same fundamental structural role as in an inflecting language, even if, in the course of time, inflection has been considerably reduced, as in Farsi or English. 4.3.2. Fragment blending on the basis of English loanwords Fragment blends created on the basis of polysyllabic loanwords from English are normally as unmotivated as their source words. See, e.g., the following calque of English bobo < bourgeois x bohemian. Note that even in this completely desemanticised sequence of syllables, the superficial structure of a typical Chinese compound is preserved:

Blending between grammar and universal cognitive principles 139

布 尔 乔 亚 波 希 米 亚 族 bù ěr qiáo yà x bō xī mǐ yà zú cloth hence high Asia wave to hope rice Asia ethnic group ‘bourgeois’ ‘Bohemian’ 布 波 族 bù bō zú cloth wave ethnic group  ‘bobos’ 4.4. The structural regularity of Chinese blends An important feature of all Chinese blends, especially as compared to their German counterparts, is their closeness to regular compounds. In part, this similarity is conditioned by the immutable basic units of one syllablemorpheme-character. But even the way these basic units are combined into higher structures is remarkably regular. This preference for structural regularity is particularly evident in regular readings superseding possible readings as telescope blends, and in native fragment blends “propped” onto previously existing regular compounds. The care taken by language users to make regular use of Chinese gram­ ­mar even in the realm of irregularity may be in part a cultural feature, motivated by the high esteem for the national language, but an additional structural reason could be the extremely “loose fit” of Chinese explained above. Compared to German or even English, the Chinese linguistic system produces utterances with a relatively small amount of redundancy. As we have seen, the information provided by the semantic and structural features of the context has to be used in the process of decoding, because individual morphemes and even grammatical structures are ambiguous. In such a situation, where every bit of structural information is needed, “playing around” with linguistic structure is not considered as funny by language users, but simply as dysfunctional. In close-fit languages – and any Indo-European language is “close-fit” as compared to Chinese, even if much of the original inflection has disappeared, as in English or Farsi – there is more room for the playful modification of the basic regular structures. This is why telescope blending as a relatively slight modification of regular compounding is allowed in German, but not in Chinese. Thus, the regular linguistic system not only provides certain patterns for extragrammatical morphological operations, but determines even the degree to which these patterns have to be observed.

140 Elke Ronneberger-Sibold 5. Conclusions In this paper, blending was considered as deliberate extragrammatical compounding: it resembles regular compounding insofar as two existing words are combined into a new one, but this is done by creative techniques which are not covered by the rules or models of regular word formation. On the basis of German blends, four basic types were established, ranked by decreasing morphological transparency and at the same time by decreasing similarity to regular compounding. Whereas in the first type, called telescope blending, regular compounding is only slightly modified by a medial overlap between the blended words, e.g., in German Kamelefant < Kamel ‘camel’ x Elefant ‘elefant’, the second type, called contour blending, is less based on grammar than on the universal ability to recognize words by certain features identifying their sound shape, especially by their rhythmic contour. For example, brunch is relatively transparent, although both blended words, breakfast and lunch, are mutilated, because the rhythmic contour (i.e., a heavy monosyllable) and the rhyme of the so-called matrix word lunch are maintained in the blend. Under this condition, breakfast can be recovered by its onset only, because it is inserted “in the right place” into the sound shape of lunch. This is a question of handling sound shapes, not of grammar. The third and especially the fourth type, called semi-complete and fragment blending respectively, are more artificial and hence even less grammar-based than the first two. They entail the manufacturing of opaque words out of splinters from the blended words. The use of these types was contrastively analysed in three languages exhibiting important differences with respect to their individual grammars, writing systems and social settings, namely German, Farsi and Standard Chinese. The aim was to investigate the interaction, in blending, between these intra- and extra-linguistic factors and certain universal cognitive abilities, such as the recognition of sound shapes. The grammatical and typological differences between the three languages have a much stronger impact on telescope blending than on the other types: in Farsi, telescope blending is completely blocked, because it cannot rely on productive compounding as a regular base. In Chinese, compounding is pro­ ductive, but the regular pattern is so strong that it supersedes blending, so to speak: constructions which could be interpreted as telescope blends are inter­­ preted according to regular patterns instead. Fragment and especially contour blending, on the contrary, are possible techniques in all three languages, irrespective of their grammatical and typological differences. It is only the concrete implementation of these techniques that is influenced by basic structural features such as the impossibility of disrupting syllables in Chinese.

Blending between grammar and universal cognitive principles 141

These findings confirm the claim of Natural Morphology that language is founded on certain very general cognitive principles which not only govern the first stages of language acquisition, but are also available later on, when language users deliberately overstep the limits of their linguistic system in word creation. Acknowledgement I would like to thank the German Science Foundation (Deutsche Forschungs­ gemeinschaft) for supporting the project underlying this article. References Bauer, Laurie. 2003. Introducing Linguistic Morphology. 2nd edn. Edinburgh: Edin­burgh University Press. Original edition 1988. Bond, Zinny S. 1999. Slips of the Ear: Errors in the Perception of Casual Conversation. San Diego, CA: Academic Press. Bußmann, Hadumod. 2003. Lexikon der Sprachwissenschaft. 3rd edn. Stuttgart: Kröner. Original edition 1983. Cannon, Garland. 1986. Blends in English word formation. Linguistics 24 (4): 725–753. Catalogues Groups Iran’s Industries, Exports, and Minerals. 1354–1355 / 1975–1976. Tehrān: Rahnema Organization. Chao, Yuen Ren. 1968. A Grammar of Spoken Chinese. Berkeley: University of California Press. Chen, Ping. 1999. Modern Chinese: History and Sociolinguistics. Cambridge: Cambridge University Press. Dressler, Wolfgang U.  2000. Extragrammatical vs. marginal morphology. In Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 1–10. München: Lincom Europa. Dressler, Wolfgang U. and Annemarie Karpf. 1995. The theoretical relevance of pre- and protomorphology in language acquisition. Yearbook of Morphology 1994: 99–122. Duanmu, San. 2007. The Phonology of Standard Chinese. 2nd edn. Oxford & New York: Oxford University Press. Original edition 2001. Ekrāmi, Mahmud. 1385/2006. Mardomshenāsi: Estelāhāt-e khodemāni [Ethnology: Colloquial Idiomatics]. 3rd edn. Mashhad: Nashr-e ivār. Original edition 1384/2005. Enzyklopaedie des Islam: Geographisches, ethnographisches und biographisches Wörterbuch der muhammedanischen Völker. 1913–1936. Leiden: Brill. Farhangestān-e zabān va adabiyāt-e Fārsi. Goruh-e takhasosi-ye ekhtesārāt [Academy for Persian Language and Literature. Special Group for Abbreviations] (ed.).

142 Elke Ronneberger-Sibold 1378/1999. Osul va zavābet va raveshhā-ye ekhtesārsāzi [Principles, Conditions, and Methods of Abbreviation]. Tehrān: Farhangestān-e zabān va adabiyāt-e Fārsi. Friederici, Angela D., Manuela Friedrich and Anne Christoph. 2007. Brain responses in 4-month-old infants are already language specific. Current Biology 17: 1208–1211. Grésillon, Almuth. 1984. La règle et le monstre: le mot-valise. Interrogations sur la langue, à partir d’un corpus de Heinrich Heine. Tübingen: Niemeyer. Hawkins, John A. 1986. A Comparative Typology of English and German: Unifying the Contrasts. London & Sydney: Croom Helm. Hawkins, John A. 1994. A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press. Kemmer, Suzanne. 2003. Schemas and lexical blends. In Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.), Motivation in Language: Studies in Honour of Günter Radden, 69–97 (Current issues in linguistic theory 243). Amsterdam & Philadelphia: Benjamins. Kilani-Schoch, Marianne and Wolfgang U. Dressler. 2005. Morphologie naturelle et flexion du verbe français. Tübingen: Narr. Ling, Yuanzheng. 2002. Xiandai hanyu suolüeyu [Abbreviations in Modern Chinese]. Beijing: Yuwen chubanshe [Publishing House for Language and Script]. Liu, Yaping. 2007. Wangluo wenxue yuan de “kuanghuanhua” tese [On the Peculiarities of the “Tendency Towards Wit and Craziness” in the Language of Inter­ ­net Literature]. Changchun gongye daxue xuebao. Shehui kexueban [Journal of Changchun University of Technology Social Sciences Edition] 19 (1): 105–106. Majidi, Mohammad-Reza. 1990. Strukturelle Grammatik des Neupersischen (Fārsi). Vol. II Morphologie. Hamburg: Buske. Marchand, Hans. 1969. The Categories and Types of Present-Day English Word Formation. 2nd edn. Munich: Beck. Original edition 1960. Norman, Jerry. 1988. Chinese. Cambridge: Cambridge University Press. Original edition 1936. Packard, Jerome L. 2000. The Morphology of Chinese: A Linguistic and Cognitive Approach. Cambridge: Cambridge University Press. Parchami, Mohebollāh. 2004. Paskuchehā-ye Farhang [Dark Alleys of Culture]. Tehrān: Farhang-e Mahrokh. Reischer, Jürgen. 2008. Die Wortkreuzung und verwandte Verfahren der Wort­ bildung: Eine korpusbasierte Analyse des Phänomens “Blending” am Beispiel des Deutschen und Englischen. Hamburg: Kovač. Ronneberger-Sibold, Elke. 1997. Typology and the diachronic evolution of German morphosyntax. In Jacek Fisiak (ed.), Linguistic Reconstruction and Typology, 313–335. Berlin & New York: Mouton de Gruyter. Ronneberger-Sibold, Elke. 2001. Phonological features identifying the sound shape of morphs: External and internal evidence. In Chris Schaner-Wolles, John Rennison and Friedrich Neubarth (eds.), Naturally! Linguistic Studies in Honour of Wolfgang Ulrich Dressler Presented on the Occasion of his 60th Birthday, 441–450. Torino: Rosenberg & Sellier.

Blending between grammar and universal cognitive principles 143 Ronneberger-Sibold, Elke. 2002. Volksetymologie und Paronomasie als lautnach­ ahmende Wortschöpfung. In Mechthild Habermann, Peter O. Müller and Horst Haider Munske (eds.), Historische Wortbildung des Deutschen, 105–127. Tübingen: Niemeyer. Ronneberger-Sibold, Elke. 2006. Lexical blends: functionally tuning the transparency of words. Folia Linguistica 40: 155–181. Ronneberger-Sibold, Elke. 2010. …und aus der Isar steiget der weiße Neger Wumbaba: Lautgestaltprägende Elemente bei der Schöpfung von Mondegreens. In Rüdiger Harnisch (ed.), Prozesse sprachlicher Verstärkung: Typen formaler Resegmentierung und semantischer Remotivierung, 87–106. Berlin & New York: Mouton de Gruyter. Samā’i, Mahdi. 1383/2004. Farhang-o-loqāte zabāne makhfi [Glossary of the Secret Language]. 3rd edn. Tehrān: Nashr-e markaz. Seiler, Hansjakob. 1975. Die Prinzipien der deskriptiven und der etikettierenden Be­­ nennung. In Hansjakob Seiler (ed.), Linguistic Workshop III, 2–57. München: Fink. Shanghai Daily (ed.). 2007. Chinese Buzzwords with English Explanations. Shanghai: Better Link. Soudek, Lev I. 1978. The relation of blending to English word-formation: theory, structure, and typological attempts. In Wolfgang U. Dressler and Wolfgang Meid (eds.), Proceedings of the Twelfth International Congress of Linguists, 462–466. Innsbruck: Institut für Sprachwissenschaft. Sun, Juan. 2009. Shixi hanyu goucifa zhong shifou cunzai hunchengfa [Discussion of Blends in Chinese Word Formation]. Keti yanjiu [Research Questions] Jan. 2009, Nr. 2: 56–57. Ungerer, Friedrich. 1991. Acronyms, trade names, and motivation. Arbeiten aus Anglistik und Amerikanistik 16: 131–158. Xiandai hanyu cidian [Dictionary of Modern Chinese]. 2009. Zhongguo shehui kexueyuan yuyan yanjiusuo cidianbianji shi [Chinese Academy of Social Sciences. Institute for Linguistic Research. Dictionary Section] (ed.), 5th edn. Beijing: Shangwu yinshuguan [Commercial Press]. Original edition 1973.

Quantitative corpus data on blend formation: Psycho- and cognitive-linguistic perspectives Stefan Th. Gries On the basis of a heuristic characterization of intentional blending as a tripartite word-formation process, this paper discusses a variety of case studies concerned with the effects of similarity and recognizability on the formation of blends. More specifically, the case studies focus on (i) the degree of similarity between the two source words that are blended (on different levels of linguistic analysis), (ii) the ordering of source words in blends, and (iii) the ways in which source words are split up and merged into blends. The case studies include comparisons to supposedly related phenomena, viz. speecherror blends and complex clippings, and extend previous work by proposing new, or improving existing, corpus-linguistic operationalizations of relevant concepts and by increasing the sample sizes from previous studies. Keywords: blends, complex clippings, cut-off point, similarity.

1. Introduction Blending, the process that underlies the creation of brunch or chunnel from breakfast and lunch or channel and tunnel respectively, is one of the most perplexing word-formation processes, given that: – – – –

it is not as rule-governed as derivational processes; it is not as productive as most derivational processes; it is more creative than most derivational processes; it involves conscious effort and word play on the part of the coiner, which often results in “violations” of more rigid morphological rules and includes the “integration” of many kinds of information that are not central to linguistic study (e.g., the interplay between orthography and pronunciation); – it nevertheless exhibits superficial similarity to other intentional wordformation processes (e.g., compounding, (complex) clipping, abbreviation, acronymy); – it has an unplanned counterpart in the form of speech-error blends.

146 Stefan Th. Gries Given the interaction of all these characteristics, it comes as no surprise that some have adopted a somewhat pessimistic stance towards blends: “in blending, the blender is apparently free to take as much or as little from either base as is felt to be necessary or desirable. […] Exactly what the restrictions are, however, beyond pronounceability and spellability is far from clear.”  (Bauer 1983: 225) “we find no discernible relationship between phonology […] and a viable blend. […] This fact helps to make blends one of the most unpredictable categories of word-formation.”  (Cannon 1986: 744)

It is true: blends involve a mind-boggling degree of complexity, and the kind of (near-)categorical rules and processes we often find elsewhere in morphology are hard to come by. On the other hand, just because blends do not exhibit many, if any, categorical rules does not mean that blends are unpredictable. In fact, most, if not all, linguistic phenomena are not categorical in nature, but probabilistic and multifactorial – and so are blends. We should therefore adopt a probabilistic approach to the analysis of blends and their structure, but we need larger samples than those studied in some of the classic studies (e.g., 314 in Pound 1914, 132 in Cannon 1986) and statistical methods that can handle probabilistic distributions better than intuition or hunches alone. For the purposes of this paper, I will adopt the following relatively uncontroversial definition of intentional lexical blends: an intentional fusion of typically two (but potentially more) words where a part of a first source word (sw1) – usually this part includes the beginning of sw1 – is combined with a part of a second source word (sw2) – usually this part includes the end of sw2 – where at least one source word is shortened and/or the fusion may involve overlap of sw1 and sw2. This definition is intended to distinguish such blends from speech-error blends, which are not intentional even though they may be experimentally induced, and complex clippings, which involve the concatenation of the beginnings of two source words. The data to be studied currently include 2329 formations; however, not all of them have already been annotated with regard to all the parameters that will be discussed below. In what follows, I will present several case studies regarding what one might informally consider the three interrelated “temporal stages” of intentional blending: the selection of the source words to be blended, the (related) decision for a particular order of these source words in the blend, and the (related) decision of how exactly to split up the words and blend them. This division of blending into three different stages is of course somewhat

Quantitative corpus data on blend formation 147

artificial and not intended as a characterization that is isomorphic to the actual psycholinguistic processes, but it is nevertheless a convenient heuristic to approach the phenomenon. The case studies to be discussed here involve (non-standard) elements from many different levels of linguistic analysis: – – – –

graphemes and phonemes; graphemic and phonemic n-grams; syllables and their stress patterns; words, their lengths, frequencies (and semantics).

Crucially, I will argue and exemplify that the study of these aspects of intentional blends requires that blends be compared to other intentional word-formation processes as well as (randomly generated) baselines. In addition, I will argue that intentional blend formation involves an interplay of phonemic/graphemic/… similarity on the one hand as well recognizability on the other. Some of the case studies to be discussed are replications of previous work on the basis of a now larger data set, but others will be new or, maybe even more importantly, illustrate that the field is still at a stage where methodological fine-tuning is required by demonstrating that not all previous studies, including my own, have succeeded in operationalizing the relevant parameters optimally. 2. The selection of the source words When it comes to selecting the words to be blended, a variety of studies have shown that the source words speakers choose to blend are similar to each other. This is true of speech-error blends’ phonological characteristics (MacKay 1987; Kubozono 1990), syntactic/POS characteristics (MacKay 1987; Berg 1998), and their semantic characteristics (Levelt 1989; Berg 1998), but it is also true of the intentional blends focused on here most (Kubozono 1990; Kelly 1998; Gries 2004a–c). However, there are many ways in which words can be similar to each other (lengths (of different types of units), frequency/dispersion, phonemes, graphemes, syllables, stress patterns, semantics, etc.) and different ways in which each of these similarities can be operationalized. In addition, when it comes to, say, phonemic/graphemic similarity, source words may be similar to each other in different parts of the words. Finally, similarity measures of all of the above kinds must always be compared against expected/random baselines so as to make sure that whatever similarity value is obtained is not squarely within the range of chance values. In this paper, I will report on several small case studies, each of which

148 Stefan Th. Gries is concerned with a different facet of word similarity and supersedes earlier work on this in terms of the amount of data covered and/or in terms of how the data are studied and evaluated. 2.1. The lengths of source words First, let us explore the lengths of source words of three different kinds of blends: authentic error blends (i.e., unplanned lapses that happened to have been overheard and that were quoted in psycholinguistic studies), induced error blends (i.e., unplanned lapses that were induced in published experimental studies) and intentional word-formation blends (compiled from published studies as well as by the author). Previous studies have argued that, in intentional blends, sw1 is shorter than sw2 (cf. Kelly 1998; Gries 2004c) while, in authentic errors, the reverse tendency was obtained (cf. MacKay 1973 on German error blends). To replicate these findings, I counted for both source words the numbers of syllables, phonemes and graphemes (for intentional blends only) and compared their average lengths (as independent samples); the results are represented in the three panels of Table 1. Table 1. Medians and interquartile ranges (in parentheses) of lengths of blend types authentic error blends (186) sw1 sw2 pU-test

syllables

phonemes

2 (1, 2) 2 (1, 2) > 0.79

5 (4, 7) 5 (4, 6) > 0.47

letters

induced error blends (32) sw1 sw2 pU-test

syllables

phonemes

3 (2, 4) 3 (2.75, 4) > 0.92

8 (6, 9) 8 (6.75, 9) > 0.97

letters

intentional blends (1921) sw1 sw2 pU-test

syllables

phonemes

letters

2 (1, 3) 2 (3, 3)