This book presents a new theory of grammatical categories - the Universal Spine Hypothesis - and reinforces generative notions of Universal Grammar while accommodating insights from linguistic typology.
Date: Thu, 8 Apr 2004 20:59:24 -0700 (PDT) From: Brady Zack Clark <email@example.com> Subject: Syntactic Change: A Minimalist Approach to Grammaticalization
Roberts, Ian and Anna Roussou (2003) Syntactic Change: A Minimalist Approach to Grammaticalization, Cambridge University Press.
Brady Zack Clark, Department of Linguistics, Stanford University.
Roberts and Roussou seek to provide a general analysis of a robust diachronic phenomenon, grammaticalization. Grammaticalization is standardly defined (e.g., Lehmann 1985) as the creation of new functional material through the reanalysis of lexical material or existing functional material; e.g., the development of English auxiliary verbs from lexical verbs. The central claim of the book is that grammaticalization is a regular case of parameter change (Lightfoot 1991, Lightfoot 1999), not a separate and unique type of change. Consequently, grammaticalization is claimed to be epiphenomenal. This claim dovetails with other recent work on grammaticalization; see e.g. Newmeyer (1998) and Traugott (2003).
There has traditionally been a rift between historical syntacticians working in the Principles and Parameters framework and the grammaticalization community. The former tradition (see e.g. Lightfoot 1999) has viewed diachronic change as a random walk around the space defined by a set of parameters, rejecting the idea that there are tendencies or pathways in diachronic change, a common theme of grammaticalization studies. Roberts and Roussou's book is an important contribution to the historical syntax literature, particularly for its attempt at reconciling these two approaches to change, as well as for its rich set of case studies. As I discuss in the critical evaluation below, some open questions remain, and more work needs to be done to provide a fully specified account of grammaticalization. In the first part of the review, I describe the content of the book. In the second part, I look critically at some of Roberts and Roussou guiding assumptions.
After a brief introduction, a chapter on the formal framework for cross-linguistic variation and change sets the stage for the three empirical chapters that follow.
Chapter 1 Parameters, functional heads and language change
The goal of Roberts and Roussou's book is twofold. First, to address the question of syntactic change in the context of the Minimalist Program in order to provide a general analysis of grammaticalization. Second, to explain the existence of variation and change when language is claimed to be (in some sense) a perfect system (the 'Strong Minimalist Thesis'; Chomsky (1995,2000,2001), more on this below). Along the way, they explore the nature of functional categories, using grammaticalization as a probe. In the first chapter, they present their model of parametric variation. Functional heads (e.g., T(ense), D(eterminer), etc.) are present in all languages (following e.g. Cinque 1999), although they may not be realized morphophonologically. Each functional head may be associated with a P(honetic)F(orm) realization. This realization can be achieved by Merge (i.e., the lexicon provides a morphophonological matrix for the functional head) or Move (i.e., material from elsewhere in the clause structure is moved to the functional head). Given this background, language change consists of change in the PF realization of functional heads. Change occurs when the trigger experience for a parameter setting (e.g., whether a functional head is realized by Merge or Move) is ambiguous or obscure. The next few chapters explore Roberts and Roussou's model of parametric variation and change, and the implications for approaches to grammaticalization.
Chapter 2 T elements
Chapter 2, like the two empirical chapters that follow it, aims to provide evidence for the idea that grammaticalization involves the reanalysis of functional categories. For Roberts and Roussou, clause structure roughly conforms to the hierarchy CP-TP-VP, where CP dominates TP and VP, and TP dominates VP. The principal idea is that reanalysis of functional heads always involves reanalysis of movement; e.g., a functional head that was previously realized by movement is realized by a morphophonological matrix provided by the lexicon. This chapter deals with the grammaticalization of T(ense) elements. The case studies investigated are familiar from previous work on grammaticalization: the development of English modals, Romance futures, and the future particle "tha" in Greek. For each of these, the account is roughly similar. Grammaticalization is always upward reanalysis (alternatively, syntactic scope expansion; Tabor and Traugott 1998) along the clausal hierarchy CP-TP-VP: a small, unproductive, subclass of a lexical category is reanalyzed as a subclass of a functional category. To illustrate, for English modals, a subclass of lexical verbs (Warner 1993: 147) which were once moved to T are directly merged in T; i.e., reanalyzed as elements of T.
Chapter 3 C elements
The next set of case studies focus on the grammaticalization of C(omplementizer) elements like English "that". As with the grammaticalization of T elements, the reanalysis of C elements is upwards. The first three case studies focus on the development of the Greek subjunctive particle "na", the Southern Italian particle "mu", and the English infinitival marker "to". These case studies, in contrast to those in Chapter 2, are not associated with loss of movement steps. Rather, the selectional properties of certain lexical or functional heads change such that certain features associated with a lower head become associated with a higher head. The grammaticalization of "na" and "mu" both involve the transfer of Mood features from T to M(odal) (where M is in the C domain), as a consequence of the loss of inflection (i.e., subjunctive morphology). The final two case studies both involve loss of movement steps: the development of Germanic complementizers and the development of complementizers from serial verbs.
Chapter 4 D elements
The final empirical chapter concentrates on a third functional category: the grammaticalization of D(eterminer) elements. As with the grammaticalization of C and T elements, grammaticalization in the D domain is argued to involve upward reanalysis; e.g., the loss of movement steps or the transfer of features associated with DPs to functional heads in the clausal domain. The case studies include the development of Romance definite determiners out of demonstratives (e.g., the Romance article out of the Latin demonstrative "ille"), French n-words (e.g., "rien"), Greek wh-words (e.g., "dhen") from indefinites, and universal quantifiers. The development of Romance definite determiners is claimed to involve the loss of movement within the DP, triggered by the loss of morphological case. The final two case studies focus on the development of clitic systems in Northern Italian dialects and the development of Welsh agreement affixes. The latter development provides an example where a lexical item associated with DP becomes the realization of a functional head in the clausal domain.
Chapter 5 Theoretical consequences
The final, theoretical chapter returns to the issues raised at the beginning of the book, in light of the case studies discussed in Chapters 2-4. In the first part of the chapter, Roberts and Roussou aim to give a general characterization of grammaticalization. They argue that each of the case studies discussed in previous chapters reduces to a single pattern: upward reanalysis giving rise to a new exponent for a higher functional head (pg. 200). In the second part, they address the tension between the observation that there are pathways of language change and the Principles and Parameters (random walk) approach to syntactic change. Their solution is to define the markedness of parameter values in terms of a simplicity metric, where, for example, Merge is less marked than Move. In the absence of cues (e.g., inflectional morphology) for marked parameter values, the less marked option is taken in acquisition. In the final part of the chapter, Roberts and Roussou address the nature of functional categories. Their central claim is that functional categories are "defective" at the interfaces (Phonetic Form and Logical Form; Chomsky 1995 et seq), in the sense that they lack non-logical content (e.g., argument structure) and are prosodically subminimal. For example, for Roberts and Roussou, the reduced auxiliary "'ll" (as in "Kim'll go to the party") has no argument structure and no prosodic structure.
As discussed in the introduction, Roberts and Roussou's book is exceptional in that it attempts to give a fully general, formal account of grammaticalization, thus bridging the divide between formal and functional accounts of this phenomenon; see van Kemenade (2000) and von Fintel (1995) for similar earlier attempts. The empirical net of the book is cast wide: eighteen case studies from a variety of languages, each with a synchronic description, as well as a formal account of the development. The account for each of these case studies is similar: upward reanalysis alongside semantic and phonological change. The book lays the foundation for future research on other grammaticalization phenomenon along the same lines, as well as work developing the formal framework and the theory of acquisition. Some potential problems and open questions remain, though. I address these in the remainder of the review.
A. The WYSIWIG approach to functional categories
Roberts and Roussou adopt (pg. 28-29) an approach to clause structure in which there is no parametric variation in the set of functional heads (Q, WH, Neg, T, D, etc.) that appear in clause structure. Languages differ only in whether or not these heads are given a PF- realization. I call this approach the universal architecture account of clause structure.
Roberts and Roussou consider an alternative to the universal architecture account: (what they call) the `What you see is what you get' (WYSIWYG) analysis (pg. 24-25). In the WYSIWYG analysis, the only functional categories postulated as present in a given language, or even a given sentence, are the ones for which we see some kind of realization. The only instance they give of this approach is Grimshaw (1997), which argues that there is no fixed structure for clauses. They argue against the WYSIWYG account on conceptual and empirical grounds. Conceptually, the WYSIWYG account has no advantage over the universal architecture account because the distinction between syntax and phonology leads us to expect that certain elements will be realized at one level and absent at another. Empirically, they point out a flaw in Grimshaw's account of embedded clauses in sentences like "I think it rained" and state that it is hard to account for grammaticalization if change and variation involves structural change rather than simple category change - an argument from ignorance.
Roberts and Roussou's curt dismissal of the WYSIWYG approach to clause structure fails to do justice to the large body of literature that argues for that type of account. For example, a widely-accepted view of phrase structure in the Lexical-Functional Grammar tradition (Bresnan 2001, Dalrymple 2001) is that the inventory of syntactic categories is not universally fixed. On this approach, the existence of a functional head position can be motivated in two ways. First, a functional head position can be motivated by the special syntactic elements that appear there; see e.g. Kroeger (1993) on evidence for a distinguished position in English, German, Warlpiri, and Tagalog in which only finite main verbs and auxiliaries appear. Second, a functional head can be motivated by the special positioning of certain elements; see e.g. King (1995) on the special positioning of tensed verbs in Russian. There is a large body of work in this tradition that argues for or against the presence of functional categories in a given language on empirical grounds. For example, Austin and Bresnan (1996) argue against the existence of a CP in Warlpiri. Sells (1995) argues that Korean and Japanese lack functional categories altogether. Roberts and Roussou's failure to grapple in any serious way with the considerable body of literature that assumes a WYSIWYG-like approach to clause structure seriously weakens their claims about the superiority of the universal architecture account.
Another lacuna in Roberts and Roussou's account is detailed discussion of the body of work on the rise of new functional categories; see Condoravdi and Kiparsky (2001), Kiparsky (1995, 1997) and Vincent (1997) (also unpublished work by Ashwini Deo (2001)). Condoravdi and Kiparsky (2001) discuss the rise of a composite functional projection in Greek. Vincent (1997) (cited by Roberts and Roussou) shows that DP emerged historically in Romance. Kiparsky (1995) (whose analysis is reanalyzed by Roberts and Roussou) proposes that CP developed historically in Germanic. A theme of this work is the examination of changes in which the realization of functional categories shifts from inflectional morphology to syntax. This literature provides empirical evidence that structural change does occur and has important implications for what comes under the purview of grammaticalization studies. For Roberts and Roussou, grammaticalization is defined as the creation of new functional material through the reanalysis of lexical material or existing functional material. Just like onomasiological categories (e.g., relational noun) in the model of grammaticalization in Lehmann (1985), the universal architecture of functional projections constitutes the preexisting landscape across which grammaticalization drags functional and lexical elements. If Kiparsky et al. are correct, though, the definition of grammaticalization must be extended to include the establishment of new functional categories (as in Lehmann 1993).
What about Roberts and Roussou's learnability argument; i.e., that structural change is hard to account for? First, this is an argument from ignorance, a mode of argumentation that is weak at best. Second, there is no real reason to think that existing learning algorithms could not deal with structural change, if we put Roberts and Roussou's assumptions about a universally invariant clausal architecture aside. For example, much like the LFG literature described above, Bobaljik and Thrainsson (1998) provide evidence that languages vary along the Split- IP Parameter: languages with a positive setting for this parameter have an AgrSP and TP as separate functional projections whereas languages with a negative settting are characterized by an unsplit IP. Third, Roberts and Roussou do not propose a formal learning algorithm themselves, apart from e.g. some brief discussion of Clark and Roberts (1993) in Chapter 1 (pgs 14-15) and some discussion of how the simplicity metric for parameter values accounts for unidirectionality in Chapter 5. Until this kind of work is done, it is difficult to assess the viability of Roberts and Roussou's arguments against structural change or their arguments for parametric change.
B. Morphology and syntactic acquisition
The causal factor in many of the changes that Roberts and Roussou examine is claimed to be morphological. For example, they key the development of English modals to the loss the infinitive marker "-en" in early English (pg. 42). Likewise, the development of the English infinitival marker "to" is said to be caused by the loss of subjunctive morphology (pg. 107). For Roberts and Roussou, morphology is a cue for marked syntactic structures, where a marked structure is one where a lexical item spells out more than one feature. Movement is an example of a marked syntactic structure. In (1), the lexical item Y realizes the features associated with a low head (V) and a higher functional head (T):
(1) [_TP Y + T [_VP ... t_Y ...]]
Change occurs when this cue for this marked structure has become obscure or ambiguous; e.g., by morphophonological change, as in the case of the development of English modals. Roberts and Roussou assume that the language learner is the locus of change and variation. Do language learners really use overt morphology as cues for the acquisition of syntax? Work on acquisition suggests that the claim that morphology drives syntactic acquisition is, at least, questionable. For example, there is some evidence that children know whether verbs raise or not before they acquire morphological distinctions; see e.g. Lardiere (2000) (cited in Bobaljik 2002).
Let's assume, for the sake of argument, that morphology cues syntax. In order for this to work, we need to have a model of the morphology- syntax interface. Roberts and Roussou seem to be advocating a morphology-driven approach in which variation in morphology is the cause of syntactic variation (Borer 1984). However, there is no explicit connection made between their approach and work like Rohrbacher (1999), which gives a related account. In contrast, in certain realizational approaches to morphosyntax (e.g., the Late- Insertion model of Bobaljik and Thrainsson 1998, Bobaljik 2002, Thrainsson 2003), morphology is a reflection, rather than a determinant, of syntactic structure, and is licensed by the same syntactic elements that determine, for example, (lack of) movement. In this type of approach, morphology may provide a cue to syntax, but there may also be other non-morphological cues (e.g., the position of certain elements). Unfortunately, Roberts and Roussou do not address this type of model. In order to have a truly explanatory model of the relationship between morphology and the acquisition of syntax, it will be necessary to more fully spell out their model of the syntax- morphology interface, and deal directly with the evidence from acquisition cited above.
C. Lexical splits
A lexical split refers to a situation where a functional head is paired with a formally identical (lexical or functional) head. For example, in Standard English, "need" and "dare" function as either auxiliaries or transitive verbs (pg. 43). Roberts and Roussou derive lexical splits in two ways. For functional heads paired with other functional heads (e.g., the complementizer "that" and the demonstrative "that"), different readings attributed to a single lexical item correspond to different positions in which the lexical item can be merged in clause structure. For functional heads paired with lexical heads (e.g., the auxiliary "need" and the transitive verb "need"), the account is roughly similar. Differences in interpretation correspond to position in phrase structure, as well as the presence or absence of argument structure: lexical heads have argument structure but functional heads do not. This analysis has a couple of nice properties. First, it satisfies the desire for parsimony: for functional heads, it is not necessary to posit a separate lexical item for each interpretation. Second, Roberts and Roussou account for the gradualness of grammaticalization by positing that lexical clines are a consequence of lexical splits.
It is not clear how Roberts and Roussou's analysis works out formally, though. Consider the difference between lexical and functional heads. Roberts and Roussou do not give a detailed description of argument structure at the level of individual lexical items. This is an unfortunate gap given that the presence vs. absence of argument structure is one of the key distinctions between lexical and functional heads in their approach. More generally, Roberts and Roussou give neither a formal specification of what features lexical entries consist of nor even a single complete lexical entry. The closest they come is an informal discussion of the makeup of the lexical entries for "want" and the reduced auxiliary "'ll" (pg. 230).
The lack of formal specification for lexical entries also makes it difficult to compare the consequences of Roberts and Roussou's approach to lexical splits with that of other approaches; e.g., Construction Grammar (Fillmore 1999, Fillmore and Kay 1999). In Construction Grammar, it is possible to allot aspects of meaning to a construction (a pairing of form and meaning) rather than to the words making it up. In that way, it is not necessary to say that all functional elements are assigned a meaning (Zwicky 2000) or to posit spurious homophonic elements. The Construction Grammar account of lexical splits is roughly similar to Roberts and Roussou's account in which it is also not necessary to posit a separate lexical item for each interpretation of a functional head. Rather, as discussed above, differences in interpretation arise from different placements of the functional head in clause structure. An important project would be to explore the the different empirical ramifications of these two approaches to lexical splits. Unfortunately, Roberts and Roussou neither provide enough detail about the lexicon to make that comparison possible nor do they explore frameworks beyond the one that they are working in.
D. Language as a perfect system
As discussed above, the goal of Roberts and Roussou's book is to address two questions. First, they seek to address the question of syntactic change in the context of the Minimalist Program. Second, they intend to address a question raised by the Minimalist Program itself: how do we explain the existence of parametric variation and change when "language is in some sense a perfect system (the strong minimalist thesis: Chomsky (1995: 1-10), (2000:96f.), (2001:1-2))" (pg. 1). Roberts and Roussou's answer to the second question, in short, is that parametric variation and change is a consequence of the imperfect mapping from syntax to PF.
Syntactic change and variation can only be understood as an imperfection if the assumption that language is perfect system is shared. Recent discussions of this assumption appear in Johnson and Lappin (1999: 124-133) and Natural Language and Linguistic Theory (Volume 18, Issue 4 and Volume 19, Issue 4). Roberts and Roussou do not refer to any of these discussions. This is unfortunate given that one of the two central questions of their book is predicated on the assumption that language is a perfect system. Without some kind of examination of this assumption, the question about how to explain the apparent imperfection of language change and variation will likely be a non-issue for some readers.
Of course, even if the a priori idea that language is a perfect system is not shared, we still need to provide an empirically accurate, simple, and nonredundant account of change and variation. For readers that do not share Roberts and Roussou's assumption about the optimal design of language, Roberts and Roussou's theory of parametric change and variation will be judged according to these criteria. The perfection or imperfection of language will ultimately play no role in deciding if Roberts and Roussou's theory makes any lasting contribution.
I am grateful to Peter Sells and Elizabeth Traugott for their comments. All remaining errors are my own.
Austin, Peter and Joan Bresnan. (1996) Non-configurationality in Australian Aboriginal languages. Natural Language and Linguistics Theory. 14(2): 215-268.
Bobaljik, Jonathan D. (2002) Realizing Germanic inflection: Why morphology does not drive syntax. Journal of Comparative Germanic Linguistics. 6(2): 129-167.
Bobaljik, Jonathan D. and Hoskulder Thrainsson. (1998) Two heads aren't always better than one. Syntax. 1: 195-236.
Chomsky, Noam. (1995) The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam. (2000) Minimalist inquiries: the framework. In Roger Martin, David Michaels, and Juan Uriagereka (eds), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik (pp. 89-155). Cambridge, MA: MIT Press.
Chomsky, Noam. (2001) Derivation by phase. In Michael Kenstowicz (ed), Ken Hale: A Life in Language (pp. 1-52). Cambridge, MA: MIT Press.
Cinque, Guglielmo. (1999) Adverbs and functional projections: A cross- linguistic perspective. Oxford: Oxford University Press.
Condoravdi, Cleo and Paul Kiparsky. (2002) Clitics and Clause Structure. Journal of Greek Linguistics. 2: 1-39.
Deo, Ashwini. (2001) The development of IP in Indo-Aryan. Unpublished manuscript. Stanford University.
Fillmore, Charles. (1999) Inversion and Constructional Inheritance. In Gert Webelhuth, Jean-Pierre Koenig, and Andreas Kathol (eds), Lexical and Constructional Aspects of Linguistic Explanation (pp. 113-128). Stanford, CA: CSLI Publications.
Fillmore, Charles and Paul Kay. (1999) Grammatical constructions and linguistic generalizations: the what's x doing y? construction. Language. 75: 1-33.
von Fintel, Kai. (1995) The formal semantics of grammaticalization. NELS 25: 175-189.
Grimshaw, Jane. (1997) Projections, heads and optimality. Linguistic Inquiry 28: 373-422.
Heine, Bernd, Ulrike Claudi, and Friederike Hunnemeyer. (1991) Grammaticalization: A Conceptual Framework. Chicago, IL: University of Chicago Press.
Hopper, Paul J. and Elizabeth Closs Traugott. (2003) Grammaticalization (Second Edition). Cambridge: Cambridge University Press.
van Kemenade, Ans. (2000) Jespersen's Cycle Revisted: Formal Properties of Grammaticalization. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds), Diachronic Syntax: Models and Mechanisms (pp. 51-74). Oxford: Oxford University Press.
King, Tracy H. (1995) Configuring Topic and Focus in Russian. Stanford, CA: CSLI Publications.
Kiparsky, Paul. (1995) Indo-European origins of Germanic syntax. In A. Battye and Ian Roberts (eds), Clause Structure and Language Change (pp. 140-170). Oxford: Oxford University Press.
Kiparsky, Paul. (1997) The Rise of Positional Licensing. In Ans van Kemenade and Nigel Vincent (eds), Parameters of Morphosyntactic Change (pp. 460-494). Cambridge: Cambridge University Press.
Kroeger, Paul. (1993) Phrase structure and grammatical relations in Tagalog. Stanford, CA: CSLI Publications.
Johnson, David E. and Shalom Lappin. (1999) Local Constraints Vs. Economy. Stanford, CA: CSLI Publications.
Lardiere, Donna. (2000) Mapping features to forms in second language acquisition. In John Archibald (ed), Second Language Acquisition and Linguistic Theory (pp. 102-129). Oxford: Blackwell.
Lehmann, Christian. (1985) Grammaticalization: Synchronic Variation and Diachronic Change. Lingua e Stile. 20: 303-318.
Lehmann, Christian. (1993) Theoretical implications of grammaticalization phenomena. In William A. Foley (ed), The Role of Theory in Language Description (pp. 315-340). Berlin: Mouton de Gruyter.
Lightfoot, David W. (1991) How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press.
Lightfoot, David W. (1999) The Development of Language: Acquisition, Change, and Evolution. Malden, MA: Blackwell.
Martin, Roger and Juan Uriagereka. (2000) Some Posible Foundations of the Minimalist Program. In Roger Martin, David Micahels, and Juan Uriagereka (eds), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik (pp. 1-29). Cambrdige, MA: MIT Press.
Newmeyer, Frederick J. (1998) Language Form and Language Function. Cambridge, MA: MIT Press, Bradford Books.
Roberts, Ian and Anna Roussou. (1999) A formal approach to "grammaticalization". Linguistics. 37(6): 1011-1041.
Rohrbacher, Bernhard. (1999) Morphology-Driven Syntax: A theory of V to I raising and prodrop. Amsterdam: John Benjamins.
Sells, Peter. (1995) Korean and Japanese Morphology from a Lexical Perspective. Linguistic Inquiry. 26(2): 277-325.
Tabor, Whitney and Elizabeth Closs Traugott. (1998) Structural scope expansion and grammaticalization. In Anna Gicalone Ramat and Paul J. Hopper (eds), The Limits of Grammaticalization (pp. 229-272). Amsterdam: John Benjamins.
Traugott, Elizabeth Closs. (2003) From etymology to historical pragmatics. In Donka Minkova and Robert Stockwell (eds), Studies in the History of the English Language: A Millenial Perspective (pp. 19-29). Berlin: Mouton de Gruyter.
Vincent, Nigel. (1997) The emergence of the D-system in Romance. In Ans van Kemenade and Nigel Vincent (eds), Parameters of Morphosyntactic Change (pp. 149-169). Cambridge: Cambridge University Press.
Warner, Anthony R. (1993) English Auxiliaries: Structure and history. Cambridge: Cambridge University Press.
Zwicky, Arnold. (2000) Constructions and morphological categories. Handout. Stanford University.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Brady Clark is a PhD student in linguistics at Stanford University, where he is involved in several research projects in historical linguistics, semantics, pragmatics, and tutorial dialogue systems. He received his BA in linguistics from the University of Washington in 1997.