YEARBOOK OF MORPHOLOGY 2001 Yearbook of Morphology Editors: Geert Booij Jaap van Marie Consulting Editors: Stephen Anderson (Yale) Mark Aronoff (Stony Brook, N.Y.) Mark Baker (New Brunswick, N.J.) Laurie Bauer (Wellington) Rudie Botha (Stellenbosch) Joan Bybee (Albuquerque, N.M.) Andrew Carstairs-McCarthy (Christchurch) Ore ville Corbett ( Guildford, U.K.) Wolfgang Dressler (Wien) Martin Haspelmath (Leipzig) Jack Hoeksema (Groningen) Rochelle Lieber (Durham, N.H.) Peter Matthews (Cambridge, U.K.) Franz Rainer (Wien) Sergio Scalise (Bologna) Henk Schultink (Utrecht) Andrew Spencer (Colchester. U.K.) Editorial address: Editors, Yearbook of Morphology Faculteit der Letteren, Vrije Universiteit De Boelelaan 1105 1081 HV Amsterdam, The Netherlands e-mail: [email protected] YEARBOOK OF MORPHOLOGY 2001 Edited by GEERT BOOIJ Vrije Universiteit, Amsterdam, The Netherlands and JAAP VAN MARLE Open University, Heerlen, The Netherlands SPRINGER-SCIENCE+BUSINESS MEDIA, B.V. A C.I.P. Catalogue record for this book is available from the Library of Congress. ISSN 0922-3495 ISBN 978-90-481-6061-7 ISBN 978-94-017-3726-5 (eBook) DOI 10.1007/978-94-017-3726-5 Printed on acid-free paper All Rights Reserved © 2002 Springer Science+B usiness Media Dordrecht Originally published by Kluwer Academic Publishers in 2002 No part of the material protected by this copyright notice may be reproduced or utilized in any form or by any means, electronic or mechanicaL including photocopying, recording or by any information storage and retrieval system, without written permission from the copyright owner. Table of Contents PETER ACKEMA and AD NEELEMAN I Morphological selection and representational modularity 1 JONATHAN BOBALJIK I Syncretism without paradigms: remarks on Williams 1981, 1984 53 BRIAN D. JOSEPH I Defining "word" in Modern Greek: a response to Philippaki- Warburton & Spyropoulos 1999 87 CEMIL ORHAN ORGUN and SHARON INKELAS I Reconsidering Bracket Erasure 115 GREG STUMP I Morphological and syntactic paradigms: arguments for a theory of paradigm linkage 14 7 Theme: Affix ordering and productivity (guest editor: Harald Baayen) HARALD BAAY EN I Affix ordering and productivity: a blend of phonotactics and prosody, frequency, and lexical strata 181 GEERT BOOIJ I Prosodic constraints on stacking up affixes 183 JENNIFER HAY and HARALD BAAY EN I Parsing and productivity 203 ANDREA KROTT, ROBERT SCHREUDER and HARALD BAAYEN I A note on the function of Dutch linking elements 237 ANKE LUDELING, TANJA SCHMIDT and SAWWAS KIOKPASOGLOU I Neoclassical word formation in German 253 INGOPLAGI The role of selectional restrictions, phonotactics, and parsing in constraining suffix ordering in English 285 Morphological selection and representational modularity* PETER ACKEMA AND AD NEELEMAN 1. MODULAR SELECTION In much work on word formation, the morpho-phonological properties of mor phemes are strictly separated from their semantic and morpho-syntactic proper ties. A morpheme is not a unit taken from the lexicon and combined with other morphemes. Rather, its semantic and morpho-syntactic features are inserted in the semantic and morpho-syntactic components respectively, while only the morpho-phonological component contains its overt form. The three representa tions thus formed must of course be related, something which is achieved by a set of mapping principles. A range of proposals along these lines can be found in Sproat (1985), Anderson (1992), Halle & Marantz (1993), Beard (1995) and Jackendoff (1997), amongst others. We will refer to models of this type as representationally modular (borrowing a term from Jackendoff).1 In this paper we explore the consequences of the separation of morpho syntax and morpho-phonology for morphological selection. It is traditionally assumed that affixes select for an X0 of a particular category (abstracting away from semantics). For example, the affix -able selects a V0, while the affix -ize selects an N° (or A0). Notice that such statements mention two different proper ties of the selected element. One is its category, the other its bar-level. The former type of selection varies per affix, but the latter type holds of all affixes indeed, it partially defines the notion 'affix' as commonly understood. Given that the two types of selection are qualitatively different, it would be desirable to distinguish them formally. That such a distinction must be made is in fact implied by the separation of an affix's morpho-syntactic and morpho-phonologi cal properties. Representational modularity implies that there is no such thing as the selectional requirements of 'an' affix, since affixes are not monolithic entities, but rather the combination of morpho-syntactic and morpho-phonological properties. This means that a distinction must be made between what the morpho-syntactic part of the affix selects in the morpho-syntactic component and what its morpho-phonological part selects in the morpho-phonological representation. We propose that selection for category is associated with the morpho-syntactic part of an affix, while the selection for bar-level is a by-product of its morpho-phonology. The morpho-phonological part of an affix is a dependent category, which, much like a simple clitic in the sense of Zwicky (1977), requires a base to form a phonological word with. In other words, morphological selection does not exist as such. It is the combination of two types of independently motivated selection: categorial selection in the Geert Booij and Jaap van Marie (eds), Yearbook of Morphology 2001, 1-51. © 2002 Kluwer Academic Publishers. 2 Peter Ackema and Ad Neeleman morpho-syntactic component and what one might call phonological selection in the morpho-phonology. The idea that selectional requirements must be distributed across compo nents is familiar from selection at the sentence level. The separation between the syntactic and the semantic component assumed in most models of syntax makes it possible to distinguish between the syntactic and semantic selectional requirements of a head. As Grimshaw (1979) has shown, it is desirable that this distinction be made. For example, both ask and wonder select for an interroga tive complement in the semantics, but wonder in addition selects for a CP in syntax. Hence the contrast between John asked the time and *John wondered the time. The purpose of this paper is to argue that the separation of morpho-syntac tic and morpho-phonological selection is not only conceptually desirable, but has a number of empirical advantages as well. In particular, it explains under which circumstances constituents larger than heads (that is, phrasal constitu ents) can be parts of words. For example, the morpho-phonological part of certain exceptional affixes can be shown to be a phonological word in its own right. Exactly such affixes are expected not to impose morpho-phonological selectional requirements, with the consequence that they may attach to phrases. In the same vein, representational modularity allows certain affixes not to have a morpho-phonological part at all (zero derivation). In that case, too, no morpho-phonological requirements will be imposed, with the consequence that phrasal derivation is allowed. Before we turn to the empirical advantages of modular selection in sections 4 through 9, we need to clarify the details of the model that underlies the analysis. The nature of the morpho-syntactic and morpho-phonological compo nents is discussed in section 2, while the mapping principles that relate the two are the topic of section 3. 2. MORPHOLOGY AND MODULARITY Representational modularity assumes that phonology, semantics and syntax are independent generative systems associated by mapping principles. In other words, a sentence has a semantic, syntactic and phonological representation, whose wellformedness is determined by conditions particular to the respective components. In addition, grammaticality requires successful association of these representations. At the sentence level, then, the model of grammar we assume (following Jackendoff 1997) is as below. Morphological selection and representational modularity 3 (1) Semantics Syntax Phonology ! ! ! Semantic structure ~ Syntactic Structure ~ Phonological structure The motivation for an independent semantic component is not immediately relevant for the argumentation in this paper. The evidence for a separation of syntactic and phonological structure is straightforward: phonological represen tations are not isomorphic to syntactic representations, and phonological and syntactic primitives are members of disjoint sets. A simple example illustrating this, borrowed from Jackendoff (1997:26), is given in (2). In syntax, a big house is a DP that consists of a determiner and a complex NP complement. In phonol ogy, it consists of two phonological words, the first of which is formed by the determiner and the adjective. So, both constituency and labels differ. (2) a. [oP a [NP [AP big] house]] b. ["' [w a big] [w house]] There does not seem to a plausible continuation of the syntactic derivation that connects (2a) and (2b)- familiar syntactic derivations operate under informa tional monotonicity: by hypothesis they can add, but not destroy or change structure and labels (compare Chomsky's (1995) inclusiveness condition; see also Neeleman & Van de Koot 2002). The implication is that (2a) and (2b) are generated separately and associated by mapping principles. Indeed, this is the predominant view in the literature on the syntax-phonology interface (Selkirk 1984, 1986, Nespor & Vogel1986, Inkelas & Zec 1990, amongst others). How does morphology fit into this picture? Let us begin by observing that it is not a component on a par with syntax, phonology and semantics. A sentence has a syntactic, phonological and semantic representation, but not a morpholog ical one. Rather, morpho-syntactic representations are embedded in phrasal syntactic representations. Similarly, morpho-phonological representations are embedded in prosodic structures. It seems, then, that the situation at word level mirrors the situation at sentence level in that there are independent generative systems which define well-formed morpho-syntactic, morpho-phonological and lexical-semantic representations. As is the case at the sentence level, these are associated by mapping principles. The model of grammar in (1) should hence be extended as in (3), following Jackendoff (1997) and Ackema (1999a). 4 Peter Ackema and Ad Neeleman (3) Semantics Syntax Phonology l l l Semantic structure Syntactic structure Phonological structure EMBEDDING <--> EMBEDDING <--> EMBEDDING M -Semantic structure M-Syntactic Structure M-Phonological structure i i i Morpho-Semantics Morpho-Syntax Morpho-Phonology We assume that the morpho-syntax manipulates hierarchical representations, containing distinct structural positions for the morpho-syntactic part of affixes and other morphemes. We will refer to affixal constituent parts of morpho syntactic representations as AFFIXes, to be distinguished from /affix/es, which represent the overt form of an AFFIX as inserted in the morpho-phonology. This notation will be generalized to other morphemes where relevant. In case there is a one-to-one, left-to-right mapping of AFFixes to /affix/es, the model sketched above is empirically indistinguishable from models assum ing joint insertion of morpho-phonological and morpho-syntactic (and lexical semantic) material. However, differences arise in circumstances in which the mapping is not regular. Consider combinations of causative and applicative morphology. Presumably, the morpho-syntactic structures of a causative appli cative and an applicative causative are distinct, with the causative affix c-commanding the applicative AFFIX in the former, while being c-commanded by the applicative affix in the latter. The linear order will reflect this if there is one-to-one, left-to-right mapping (see (4a,a') and (Sa,a')). The result by neces sity obtains in models assuming joint insertion (see (4b) and (Sb)). ( 4) a. [[V APPLICATIVE) CAUSATIVE) <--> a'. [[/v/ -/applicative/) -/causative/) b. [[ < V, /v/ > < APPLICATIVE, /applicative/ >) <CAUSATIVE, /causative/>) ( 5) a. [[V CAUSATIVE) APPLICATIVE) <--> a'. [[/v/-/causative/)-/applicative/) b. [[ <V, /v/> <CAUSATIVE, /causative/>) < APPLICATIVE, /applicative/ >) The advantage of representational modularity is that it can handle cases in which mapping is not one-to-one and left-to-right, while maintaining the appro priate syntactic structures. An example is provided by Chimwi:ni (Abasheikh 1978, Hyman 2001). Post-verbal morphology is templatic in this language. Crucially, the template refers to the /affix/es that spell out applicative and Morphological selection and representational modularity 5 causative, not to the affixes representing these categories in morpho-syntax. In particular, the template requires that the causative /affix/ precedes the applica tive /affix/. This does not affect the mapping of applicative causatives, but it has the consequence that a causative applicative surfaces as in (6b'). Hyman dis cusses evidence based on passive that the syntactic representation must never theless be as in (6b). Such data cannot be handled by the traditional model of joint insertion. (For similar cases from Quechua, see Muysken 1988). (6) a. Template: /causative/-/applicative/ b. [[V APPLICATIVE] CAUSATIVE]+-+ b'. [[lv/-/causative/] -/applicative/] The assumption that morpho-syntactic representations are generated by merger of morpho-syntactic constituents is in contrast with two major alternative approaches. Both of these in fact deny the existence of specifically morpho syntactic structure, but in different ways. In much of the syntactic literature inspired by Baker (1988), the internal structure of morphologically complex heads is derived by syntactic head-to-head movement (modulo possible post syntactic readjustment rules). In Beard's (1995) Lexeme-Morpheme Base Morphology, there is no such thing as a complex head in morpho-syntax. Rather, where other theories assume complex structures, Beard assumes rules which add features to a base. It is beyond the scope of this paper to discuss these models in detail. Let us instead give some general considerations that in our opinion support the model adopted here. Deriving morphological constructs by syntactic head-to-head movement faces difficulties in two domains. Empirically, if a structure [X AFFIX] is the result of head movement of X in a structure [AFFIX [xP ... X ... ]], then we would expect material that can or must be present in XP to be stranded by movement of X. Although a possible analysis for inflection, such stranding is hardly attested in the case of derivation, if at all (see Bresnan & Mchombo 1995 and Ackema & Neeleman 2001). To give an example, the material contained in the VP in (7a) cannot be present if a synthetic compound is formed, as in (7b ). 2 (7) a. to [v P reluctantly drive [NP a rusty truck] [PP to Arizona]] b. *[NP a [N [v truck drive] er] [vP reluctantly tv [NP a rusty tN] [PP to Arizona]]] Word formation though head-to-head movement also has a conceptual disad vantage. Syntactic selection holds between chain roots (Chomsky 1981, Brody 1995, Jackendoff 1997). For example, there is no raising to 9-positions. If an AFFIX selects a host after movement, that type of selection would have to be exceptional in this regard (Ouhalla 1991, Lieber 1992).