The term ‘ergativity’ is, in its most generally accepted sense, used to describe a grammatical pattern in which the subject of an intransitive clause is treated in the same way as the object of a transitive clause, and differently from transitive subject. The term was first used to refer to the case marking on constituents of a noun phrase: ‘ergative’ is the case marking transitive subject, contrasting with another case – originally called ‘nominative’ but nowadays ‘absolutive’ – marking intransitive subject and transitive object.

Ergativity is thus complementary to the familiar grammatical pattern of accusativity, in which one case (nominative) marks both intransitive and transitive subject, with another case (accusative) being employed for transitive object.

Use of the terms ‘ergative’ and ‘absolutive’ has been extended to the marking of syntactic functions by particles or adpositions, by pronominal cross-referencing markers on a main or auxiliary verb, and by constituent order. The term ‘ergative’ has been used in a further, syntactic, sense to apply to coreferentiality constraints on the formation of complex sentences, through coordination and subordination; if these constraints treat intransitive subject and transitive object in the same way the language is said to have ‘ergative syntax’, and if they treat intransitive subject and transitive subject in the same way there is said to be ‘accusative syntax’.

Introduction

Studies on VSO languages have played an important part in exploring the role of V-movement in deriving the order of the subject in relation to the verb. On the assumption that all languages, irrespective of their surface word order peculiarities, derive from an underlying SVO structure, the VSO order can be derived simply by assuming that the verb moves to a position preceding the (canonical) position of the subject. This is in essence the view originally outlined in Emonds (1980) and later applied to various languages (cf. Koopman (1984), Travis (1984), Emonds (1985) and Sproat (1985a), among many others).

Obviously, movement of the verb to a position preceding the subject is not the only way of deriving the VSO order from an underlying SVO structure. Another logical possibility is for the subject to move to a position immediately following the verb. This view has been suggested by Choe (1987) in relation to Berber, and adopted by Chung (1990) for Chamorro. This view attributes the derivation of the VSO order to movement of the subject rather than movement of the verb, although both views share the assumption that all languages derive from a unique underlying structure. Presumably, subject-lowering is sufficient to derive the VSO order, at least in the languages mentioned, irrespective of whether the verb moves out of VP to a higher position in the structure.

Certain theoretical developments have recently opened up other possibilities for deriving the VSO order from an underlying SVO structure.

The problem of inflection

It has long been observed in the study of developmental linguistics that at an early age the child often omits verbal inflection or provides the “wrong” inflection (one that is wrong for the adult grammar that is developing). There are two traditional views concerning the explanation of this phenomenon. One view assumes that young children do not know about inflection and that they have to “learn” to “add” inflection. Knowledge of inflection is measured by the proportion of times that the inflection appears in obligatory contexts. This proportion increases over time, indicating greater knowledge. Possibly children learn to add inflection on a verb by verb basis. One variant of the view claims that at a young age even the verbs that a child produces with inflection are not truly inflected verbs. Rather they are unanalyzed “wholes.” On this view, the child does not understand the inflectional processes. Rather, inflection simply becomes stronger, as an associative element of a verb. We can call this the “Growing Strength” (GS) view.

An altogether opposite view is that the child does know the grammar of inflection. What she does not know are the forms of inflection. On this view the child knows not only the UG that underlies inflection, but also knows all the properties of the language that is developing except one set of properties—namely, the particular set of morphological forms that express the inflectional properties.

Introduction

Among the Germanic languages, Icelandic and Yiddish are only distantly related and have had virtually no contact, yet they share a striking number of important (morpho) syntactic properties: head-initial phrase structure, rich subject-verb agreement, overt case morphology on full noun phrases, the availability of empty expletive subjects, and the productive use of verb-first declarative clauses in narrative contexts. Both languages also exhibit the verb-second (V2) phenomenon—not only in root clauses, but in subordinate clauses as well. While the similarities between Icelandic and Yiddish have been the subject of some discussion in the literature (Platzack & Holmberg 1990, Santorini 1989, Sigur∂sson 1990a, Vikner 1991), less attention has been paid to the differences between them. In this paper, I attempt to right this balance by focusing on these differences.

The paper is organized as follows. Section 2 sets the stage for the discussion of the differences between the grammars of Icelandic and Yiddish by reviewing an important similarity between them: the productivity of embedded V2. Following much recent work, my analysis of this shared property relies on the VP-Internal Subject Hypothesis, according to which subjects originate in a position dominated by a maximal projection of the verb. The remainder of the paper is devoted to differences between the two languages. Section 3 shows that although empty expletives are licensed in both Icelandic and Yiddish, their distribution is not identical: empty expletives can occupy Spec of IP in Icelandic, whereas they are barred from this position in Yiddish.

Koopman offers a concrete implementation of some ideas which are independently introduced in Uriagereka (1988). Her proposal has interesting consequences for binary branching, the Projection Principle, X'-shells, and the Uniformity Condition, and it offers a motivation for head movement. Since some of these results are still tentative, the assumptions behind them relying on work in progress, I will not comment on them. I want to concentrate on the notion of “licensing heads.” In particular, I want to discuss whether it is a head licensing another head that determines the properties of incorporation, whether it is the licensed head that does, or whether both licenser and licensee may be relevant.

Uriagereka (1988) arrived at the conclusion that incorporation is a universal process that serves to implement argument substitution. That work explored government relations and the process of cliticization in Romance. Following Kayne (1987) and Torrego (1988), it analyzed clitic placement as D(eterminer)-incorporation, for which independent evidence was presented in the process of generalized D-incorporation from standard DP complements in archaic Romance.

In the light of languages that say vimo-lo home ‘we saw-the man’ (Galician), the reason why the determiner in these languages attaches to the verb was explored. Having shown that the process is not phonological, I concluded that incorporation is involved. The proposal is that D-incorporation is universal at LF, and we see it overtly in some languages. The determiner (which takes a set of phi features and, as such, arguably a referential index) is analyzed as representing a variable that substitutes for a thematic slot in the theta-grid of the verb.

Introduction

In this paper, I will examine a very influential analysis of the structure of the English and French auxiliary systems, an analysis which posits verb raising or movement as a central device. The authors' assumptions about very basic aspects of linguistic structure lead inevitably to this description. I, in fact, am more interested in these assumptions than in whether or not verb movement is a device in the repertoire of syntactic description, as it surely is, for these assumptions are widely shared and reasonable ones, but I think incorrect.

The analysis is Emonds' (1978) and Pollock's (1989). The reasonable assumption lying behind the analysis is that adverbs, and other modifiers, such as negation, occupy fixed positions, so that any observed alternation of their positioning with respect to verbs must therefore be due to verb movement. What is of most interest in this context is to determine more precisely what the rules of distribution and interpretation for these elements are. I will reject the idea that there are universal “slots” in which adverbs of various kinds can appear, one slot for each type of adverb; nor do I think that adverbs are distributed in the same way in different languages. The modes of modification are more various than that. On the other hand, I do think there is a simple elegant set of notions of scope and modification which gets the adverbs right, interacting in the popular way with language-wide parameter settings.

The papers in this volume all deal with the movement of verbs to various functional categories, pursuing a line of research which has been of widespread interest in recent years. They were originally presented at a workshop on verb movement held at the University of Maryland in October 1991. Each paper was subjected to a prepared commentary and then to open discussion. The coherence of the conference theme and the liveliness of the participants made for extraordinarily productive discussion. Now we present the product of that discussion, and we introduce it by identifying some overarching themes and issues.

The conference was the culmination of a three year research project funded by the National Science Foundation (Project BNS 8812408), and we are grateful to the NSF for supporting the research and the workshop in particular. We also thank the Graduate Research Board at the University of Maryland for its support of the workshop, and Peggy Antonisse for making things work.

More immediately we thank Keiko Muromatsu and Ania Pelc for help in preparing the volume that you are now holding, Julie Perrotta for her technical wizardry, and Judith Ayling and the staff at Cambridge University Press for their enthusiasm, support and assistance. Because of the help of these people and the responsiveness of the contributors, we are going to press within eighteen months of the time of the workshop.

Ken Wexler's paper is an important contribution both to the theory of verb movement and to the theory of language acquisition. I will briefly discuss the significance of this work with respect to the view that linguistic analyses are carried out at early stages without benefit of functional categories. The bulk of the commentary will be devoted to exploring the details of Wexler's analysis of the optional infinitive stage in English. Wexler proposes two analyses for this stage. The first treats the child and adult's grammar on a par, claiming that verb lowering applies in each case. We will see that while this analysis is descriptively adequate, it carries the cost of having to explicitly mark matrix indicative clauses as untensed. This flies in the face of the fact that matrix indicatives are always related to a moment of speech in the tense structure of adult grammars. The second analysis treats child English as a verb-raising language like French, Dutch, or German. We will see that this analysis faces some descriptive problems that can be solved by claiming that the projection of object agreement (AGRo) is missing at the early stages of language development in English. We will show that the simple assumption that AGRo is missing at early stages explains Borer & Wexler's (1987) observation that the child cannot form A-chains while dispensing with the extra theoretical assumptions that they invoke.

Introduction

The principles governing the structure of syntactic representations can be seen as a collection of licensing conditions that categories of various types have to meet. Arguments, for example, need to be licensed by theta-role assignment, overt NPs by Case, small pro by being identified, negative polarity items by negative elements, anaphors by being appropriately bound to antecedents, expletive pronouns by being replaced by an NP at LF. Licensing conditions are stated in terms of a number of primitive features (nominal or verbal, phonetic or covert, etc.) and relations (government, c-command, Spec-head agreement, coindexing, etc.).

In recent work, Sportiche (1992) has suggested that an entire set of licensing conditions could be reduced to one, namely Spec-head licensing. As illustration, consider two major cases of syntactic XP-movement: wh-movement and NP movement. They represent movement to Spec positions (Spec of CP and Spec of AGRsP, Spec of AGRoP). This type of movement represents a way to license particular XPs, either because the XP is a wh-phrase and has to be licensed in Spec of CP, or because the XP needs Case. Spec positions thus typically function as “licensing” positions, and (overt or covert) movement represents the means to fulfill the licensing requirement: a particular XP is licit because it “counts” as being in the Spec position of a particular (licensing) head.

This general idea brings together a number of results from recent work that has established the privileged status of Spec positions as licensing positions.

Introductory remarks

In her paper, Santorini presents an ambitious and imaginative proposal, the first one in the literature, to unify certain syntactic differences between Icelandic and Yiddish. These differences have to do with long subject extraction across a complementizer, empty expletives in Spec of IP, and stylistic fronting, all of which are only possible in Icelandic. In commenting on the proposed account, I have deemed it most useful to stick to the perspective of a native speaker of Icelandic, not venturing too far into the intricacies of Yiddish, of which I know little. Further, I have endeavored to stay within the general theoretical framework assumed in the paper.

Santorini proposes to reduce the various differences discussed to a single factor, the location of the feature [+I] (from Rizzi 1990a). She assumes that this feature is in INFL in Yiddish, whereas it can be in either COMP or INFL in Icelandic. Nominative is assigned under head government by the highest [+I] category in the clause. The case-assigning property of the highest [+I] is also assumed to enable the head that bears it to license empty expletives. A third function of [+I], having nothing to do with nominative assignment, is voiding barrierhood, to account for long subject extraction.

It is presupposed in Santorini's paper that Icelandic and Yiddish are quite similar syntactically. It should be noted at the outset, however, that the similarity is probably less striking than one would gather from the list of common features at the beginning of the paper.

Introduction

The basic point of this paper is to argue that there are two distinct kinds of head movement. One kind is triggered by morphological properties of the host head, while the other kind is not, and in fact often appears to be triggered by some property of the moved head. Adopting and extending the terminology of Chomsky & Lasnik (1991), we refer to the former as L-related head movement and the latter as non-L-related head movement.

Both types of head movement are subject to the ECP, but, since the nature of the target of movement is different in each case, the antecedent-government requirement manifests itself in different ways. This gives the appearance of differing locality conditions; in particular, only L-related head movement obeys the “classical” Head Movement Constraint of Travis (1984). By a revision of the Relativized Minimality Condition of Rizzi (1990b), however, we see that both the cases which obey this condition and those which do not are in conformity with the ECP.

We assume a conjunctive formulation of the ECP, as in Rizzi (1990b: chapter 2). Moreover, we assume that traces of head movement are subject to a uniform head-government requirement. For non-L-related head movement, this raises the possibility that the head-governor and the antecedent-governor may be distinct. Our main empirical argument for the framework to be adopted relies on this fact; we will show that there is diachronic evidence from French that non-finite AGR ceased to be a head-governor for head traces in the 17th century, with the result that a number of instances of non-L-related head movement disappeared together.