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.
Carnie, Andrew, and Eithne Guilfoyle, eds. (2000) The Syntax of Verb Initial Languages. Oxford University Press, 256pp. Hardback ISBN: 0-19-513222-X, $45.00; paperback ISBN: 0-19-513223-8, $24.95.
Reviewed by: Mark Campana, Kobe City University of Foreign Studies
1. Introduction (Andrew Carnie & Eithne Guilfoyle)
The editors of this book have collected a fine assortment of essays from various theoretical perspectives. From Irish to Austronesian, many different languages are represented, with new data and conclusions for the interested reader/researcher to compare. The introduction sets the tone, giving brief descriptions of each essay and drawing attention to the major issues in the study of verb initial languages.
The book combines the references cited in each essay. Many papers are heavily footnoted (i.e. refereed), and in most cases it is recommended that they be read after the text has been digested. There is a lot of cross-referencing between papers, an indication of genuine collaboration towards finding out what makes these languages tick.
2. Celtic Initials (Randall Hendrick)
SYNOPSIS: This paper provides an overview of Celtic languages (Irish, Breton, Welsh, Gaelic), and is designed to show how variation can be acccounted for. At the same time, it emphasizes that several properties of these languages are unrelated to their being verb inital per se. One major question has to do with the structural position of the subject--internal to the VP or in some other category outside of it. Preverbal particles are divided up by function (tense, mood, negation, subordinators) and allotted separate positions within COMP; these combine differently in each of the languages. Synthetic vs analytic agreement and its correlation with covert vs overt pronouns is seen to follow from the interaction of three principles (Identify, Avoid Agreement, and Avoid Pronoun) which have different rankings in Welsh/Breton and Irish/Gaelic. This (basically OT) account effectively divorces the issue of agreement from verb movement, in contrast to previous treatments.
COMMENTARY: Hendrick demonstrates convincingly that Celtic subjects do not remain in VP, despite some evidence to the contrary. The question of whether objects also raise is unresolved but unavoidable, given his subsequent comparison of verb movement to that of nouns (Massam addresses this issue head-on in her article). His treatment of COMP as a three-tiered category makes good sense given the specific facts of different languages. In this regard, he also succeeds in de-emphasizing the 'cluster-approach' to verb-initial languages. On the other hand, it seems unlikely that agreement is totally unrelated to verb movement; generally-speaking, the status of OT like principles is unclear in a purely Minimalist account.
Chapter 3: 'VSO order as raising out of IP? Some evidence from Old Irish' (Andrew Carnie, Heidi Harley, and Elizabeth Pyatt)
SYNOPSIS: This paper weighs the comparative evidence of verb-raising to COMP (as in traditional analyses) vs raising to a lower projection of INFL in Old- and Modern Irish (OI, NI, respectively). Somewhat surprisingly, both are attested in OI. The authors first consider the V-to C approach in light of Germanic V2 phenomena, the major difference being the obligatory presence in SPEC, CP of a sentential XP constituent in the latter. Embedded VSO order in NI rules out this approach, but data from OI (absolute vs conjunct verb-forms, prosodically separate preverbs, the positioning of object enclitics) indicates that COMP must indeed be filled with lexical material. When a preverb satisfies this requirement the verb still moves to INFL, however, raising various issues which are then addressed. The paper concludes with some speculation as to how languages may differ in terms of what counts as 'lexical', and other independent parameters.
COMMENTARY: The arguments for V-to-C raising in IO are both exhaustive and compelling. At the same time, there is little beyond theoretical considerations and the ultimate verb-initial 'fact' to convince the reader that movement also takes place to INFL. The prosodic evidence relating to syntactic structure is particularly persuasive. While comparisons with other languages received careful attention, the history of Irish did not; a plausible scenario of change from OI to NI in terms of the theoretical constructs of the paper would have made it more complete.
Chapter 4: 'Tense and N-features in Irish' (Eithne Guilfoyle)
SYNOPSIS: In this paper, Guilfoyle seeks to account for word order in Irish tensed (VSO) and English (SVO) clauses, as well as Irish non-finite clauses (SVO) by means of differences in Event Structure. Drawing on work by Van Voorst (1998), she notes that whereas English subjects can be non-initiators of events, Irish subjects are more restricted: their features must be checked in Tense Phrase, rather than the higher AGR.s (English). Owing to their noun-like character, infinitives do not project initiator positions per se; their subjects are checked in Aspect Phrase instead. The analysis is geared towards acquisition, where it is observed that children often use truncated verbal expressions, more or less taking event initiators for granted. Given further linguistic evidence, they may easily adopt a broader spectrum of possible subjects, as in English and/or tensed Irish (VSO) clauses.
COMMENTARY: This article is conceptually appealing, both theoretically and in addressing first language acquisition data (often overlooked in Minimalist treatises). Some of the details refer to previous work, however, and this leads to some confusion with regard to properties of non-finite clauses. The mediation of Aspect Phrase (or AGR.o) appears to weaken the claim that infinitives (verbal nouns in Irish) do not project a higher position for event initiators; in addition, this category must be used for checking objects. Still, the path of predicted language selection is clear, and provides a hopeful means of verification.
Chapter 5: 'VSO and left-conjunct agreement: Biblical Hebrew vs. Modern Hebrew' (Edit Doron)
SYNOPSIS: Edit Doron's paper examines a curious feature of Biblical Hebrew, the fact that only the left conjunct of a coordinate subject NP triggers agreement on the verb (LCA). A similar effect can be observed in other languages with VS order, including the English 'there' construction. By selectively applying the EPP feature, LCA follows as a local reflex of Agree, the subject essentially remaining in situ. SVO and OVS orders (as well as non-expletive constructions in Modern Hebrew) do not exhibit LCA, leading to the conclusion that the subject moves to SPEC, TP (as in e.g. Standard Arabic, Irish). In this regard Hebrew patterns with Germanic, where the verb either raises to TP (SVO) or a focus position between TP and COMP.
COMMENTARY: The very existence of LCA is somewhat controversial, but the author argues convincingly that alternative analyses (in particular, that of Aoun et al. on Standand Arabic) must acknowledge it. The essential difference between Biblical and Modern Hebrew--i.e. that the EPP feature is assigned lexically in the latter, but only in certain contexts (selectively) in the former-- seems somewhat arbitrary, but the analysis is internally consistent. Overall, it combines thorough scholarship with skillful utilization of Minimalist theory.
Chapter 6: 'VSO and VOS: Aspects of Niuean word order' (Diane Massam)
SYNOPSIS: Massam proposes that both VSO and VOS orders in Niuean (Polynesian) are derived by fronting the whole predicate (e.g. VP), rather than its head (V). In Section 1 she shows that movement must take place to a position within IP (rather than CP), based on the form and behavior of auxiliary and negative elements. In Section 2, noun- and preposition-initial sentences are considered, which render traditional verb-fronting treatments inadequate. In its place a predicate fronting analysis is proposed, driven by EPP feature-checking; the specifier of IP is the landing site. Indefinite objects appear adjacent to the verb, taken as a sign of incorporation. They move along with it, resulting in a VOS order. Some adverbial particles are also assumed to move along with VP, while others having sentential scope appear in the inflectional complex above it. The residual differences between VP and other predicates are then accounted for by reanalyzing a pre nominal morpheme as a preposition, reducing predicate movement to the single feature [-N] (shades of Remarks on Nominalizations).
COMMENTARY: The main idea of this paper (i.e. that whole predicates raise to IP) is clear enough, but some of the details are vexing. As the author notes, VP's can only be seen as moving if everything contained within them has already cleared out. Indefinite direct objects behave as predicted, but indirect objects and obliques do not--their base position must be posited as separate. Adverbial particles with sentential scope are placed under INFL, their relative order unaccounted for. On one hand the structures Massam proposes are refreshingly spare, on the other too much so--the mirror image of Travis & Rackowski's (following) article; both address the same data. The typological distinction between verb- and subject initial languages ([EPP] vs [D] feature checking, respectively) invites still deeper inquiry.
Chapter 7: 'V-initial languages: X or XP movement and adverbial placement' (Andrea Rackowski and Lisa Travis)
SYNOPSIS: This paper begins by applying the theories of Kayne (1994) and Cinque (1999), setting out to derive word order effects in Malagasy and Niuean. Different adverb-types in these languages are in the opposite order from e.g. Italian, a problem uniquely solved by positing them as functional heads, with iterative movement of complements into specifier positions--a process known as intraposition. Peripheral (sentential) morphemes belong to 'framing categories' higher up the tree, to which the verbal complex (TP) raises. Intraposition and predicate-fronting in Malagasy bypass the subject NP and 'islandize' non-subjects, thus accounting for the pattern of extraction. Objects pattern somewhat differently in Malagasy/Niuean, owing to the fixed position of AGR.O in the latter. Overall though, objects in verb-initial languages are grossly different from their counterparts in SVO/SOV languages, which the authors relate to predicate- vs argument fronting, respectively.
COMMENTARY: This paper is a veritable tour-de-force, turning conventional mechanisms like head movement and adjunction on their side in deriving adverb placement (word order generally). The framing categories proposed by the authors are particularly clever, their two-tiered structures conforming precisely to Koopman's (1996) extended functional categories (required by the current incarnation of the Doubly-filled COMP Filter). Things get a little squishy in their treatment of objects, though: for Malagasy, AGR.O can be inserted randomly to explain their distribution, while for Niuean it is fixed (above TP no less). The rationale behind this move--ostensibly to downplay the role of agreement in feature checking--is oblivious to AGR.S. Aware of these problems, the authors argue vigorously for their view, introducing yet more data from other languages (some of them SVO and Austronesian). If anything, the paper promises more than it can deliver, and someone is going to have to check the myriad predictions that flow from it.
Ultimately one has to wonder if intraposition and the other operations commandeered to produce the desired effects aren't too much for one language to bear (let alone its speakers to acquire); one might just as easily look for variations in Kayne's and/or Cinque's theorems.
Chapter 8: 'VP remnant movement and VSO in Quiavini Zapotec' (Felicia Lee)
SYNOPSIS: One of the major themes of this volume is that VP's--rather than V's--are the constituents that undergo fronting, thus deriving verb-initial order. This is the thesis of Lee's paper as well, where sentential clitics also attract PP's and DP's in Quiavini Zapotec (QZ). As in the contributions of Massam, Travis & Rackowski (both this volume), Cinque's (1999) theory of adverbs and Kayne's (1994) of asymmetry play an important role in the analysis. A 'VP remnant' implies that NP and CP arguments have moved to their licensing positions prior to movement of their containing phrase. Subject 'agreement' provides further evidence for this approach, given that the suffixes are really clitics in argument (SPEC) position.
COMMENTARY: There are some surprising details to be found in this account. Adverb Phrases occupy functional category (SPEC) positions, from which adverbial heads then attract lexical categories (DP, VP, etc.) to their own SPEC positions (technical difficulties are overcome by appealing to a latter-day version of the Doubly-filled COMP Filter). CPs must also vacate argument positions before the VP-remnant raises, but really only to avoid being carried along as well. In the appendix it is claimed that head movement could not produce the same (correct) results while adhering to the strictest tenets of adjunction, etc. Still, the various options introduced in the paper itself do not exhaust all the possibilities. Overall, this paper accommodates the facts of QZ quite nicely, but leaves room for alternative approaches.
Chapter 9: 'Locus Operandi' (Ray Freeze and Carol Georgopoulos)
SYNOPSIS: In this paper, Freeze & Georgopoulos make the case that locative constructions (existentials and possessives--including those with 'have') bear on issues of underlying word order. Specifically, they propose an analysis in terms of head parameters and precedence at odds with Kayne's (1994) theory of asymmetry. Two correlations between construction-type and word are noted: that verb initial languages lack a word for 'have' (as in English 'Mary has a car'), and that SOV languages lack existential proforms (akin to pleonastic 'there'). SVO languages, on the other hand, have both these properties. Essentially these facts relate to how the feature [+loc] is realized: either as a clitic (in the case of proforms), or as 'have' when a preposition incorporates to it. Movement of this sort is further restricted by linear precedence and/or pragmatic factors.
COMMENTARY: In some respects this paper is a rewrite of Freeze (1992a), and several of the key assumptions are 'pre-cooked', with little opportunity to verify independently. While perhaps underappreciated, that work still stands as a challenge to the mainstream (generativist) view that e.g. existential proforms are divorced from co-occurring locatives in most languages of the world. The latest generalizations being gleaned from it may or may not offer the conclusive proof for a head-driven/linear theory of underlying structure, however. Somewhat disappointing is the authors' treatment of Kayne (1994), which--buttressed with a few concepts like lexical strength-- could probably speak to the locative facts, if not in the unified way that is advocated here.
Chapter 10: 'Prosodic conditions on anaphora and clitics in Jakaltek' (Judith Aissen)
SYNOPSIS: This paper addresses one of the more intractable problems of binding theory, the licensing conditions of null vs. lexical pronouns in Jakaltek (Mayan). Originally noticed by Craig (1977), this phenomenon has inspired several accounts, all of which conclude with some form of parameterization, usually in terms of surface constraints like precedence. Aissen holds that the correct domain of null pronoun binding is prosodic, dividing each sentence into three 'chunks' or phrases: topic, body and tail (the latter for extraposed CPs). Simply put, a pronoun is null if its antecedent is contained within the same prosodic phrase; otherwise it must be lexical. Not only does this account nicely for the facts, it helps explain why c-command plays no role at all.
COMMENTARY: The author motivates prosodic structure only after demonstrating how hierarchical accounts have failed to accommodate the content (as well as the spirit) of the null/lexical distinction. This is done via intonation junctures, based on observations made by Day (1973); it is further supported by facts surrounding the distribution of erstwhile sentential clitics. On the one hand this paper represents a challenge to die-hard syntacticians bent on proving that syntactic principles alone are responsible for binding facts; on the other, it is a wake-up call to those same linguists, who might do well to reconsider their favorite paradigms along the lines proposed here.
Chapter 11: 'Animacy hierarchies and sentence processing' (Seth Minkoff)
SYNOPSIS: This paper attempts to explain why, in some languages but not in others, transitive sentences are subject to animacy restrictions. On the one hand, individual languages vary in terms of word order and pro-drop, which can affect how soon the processor recognizes the subject NP. On the other, all languages seem to prefer Agents (prototypical subjects) with a high degree of animacy. Minkoff's study compares Mam (a Mayan language) with English, arguing that verb medial/non-pro-drop properties of the latter enable early recognition of the subject. As a verb initial/pro-drop language however, Mam relies on animacy to resolve potential confusion between subjects and direct objects.
COMMENTARY: Minkoff's treatment of the basic facts is straightforward and compelling. He demonstrates clearly that the notion of agentivity plays an important role in the determination of subjects. The interpretive component is none other than the processor itself, however, whose internal organization and/or genesis is more-or-less left unexplained. Nevertheless, it is a testable construct which would predict that speakers of Mam require more time in resolving potential ambiguity within a transitive sentence than speakers of English.
Chapter 12: 'Predicate raising in Lummi, Straits Salish' (Eloise Jelinek)
SYNOPSIS: This paper is a systematic description of Straits Salish, a NW Coast language with a rich morphology. The major claim is that the positioning of sentential constituents directly reflects information structure, while the order of morphemes within the verbal complex derives from head movement. In short, predicates raise (are fronted) because they represent new information. In keeping with the author's previous work, Straits Salish represents a typical 'predicate argument language'. Much of the evidence is adduced from the absence of type-shifting, whereby NP's may be referential, quantificational or predicative, depending on the context.
COMMENTARY: A lot of theory is packed into this article, much of it no doubt motivated in previous work. Reading through it thus requires a suspension of criticism that would otherwise make for an interesting story. Very little attention is given to the judgements of Salish speakers as compared to those of English, which seem to provide the necessary tools of analysis. Although various interesting phenomena are covered (deictic roots, dative shift, etc.), no allowance is made for alternative ways of viewing them. Terse.
Chomsky, Noam (1995). The Minimalist Program. M.I.T. Press, Cambridge, MA.
Craig, Collette (1977). The Structure of Jacaltec. University of Texas Press, Austin.
Cinque, Guglielmo (1999). Adverbs and Functional heads: A Cross-linguistic Perspective. Oxford University Press, Oxford.
Day, Christopher (1973). The Jacaltec Language. Indiana University Press, Bloomington.
Freeze, Ray (1992a). 'Existentials and other locatives'. Language 68; 553-95.
Kayne, Richard (1994). The Antisymmetry of Syntax. Linguistic Inquiry Monograph 25. M.I.T. Press, Cambridge, MA.
Koopman, Hilda (1996). 'The Spec-head configuration'. In Edward Garrett and Felicia Lee (eds.), Syntax and Sunset. UCLA Working Papers in Syntax and Semantics 1:37-64.
Van Voorst, Jan (1998). Event Structure. John Benjamins, Amsterdam.
About the reviewer: Mark Campana is a professor of English and linguistics at Kobe City University of Foreign Studies (Kobe, Japan). Research interests include the syntax and morphology of Austronesian and Amerindian languages.