The Prefunctional Stage of First Language Acquistion (RLE Linguistics C: Applied Linguistics)

A Crosslinguistic Study

Ianthi-Maria Tsimpli

• Publisher: Routledge
• Year: 2014
• Print ISBN:
  9780415724098
• eBook ISBN:
  9781317932987
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

This book provides a theory of first language acquisition in the syntactic framework of the theory of Universal Grammar. It addresses issues related to the earliest stage of development which ends roughly around the child's second birthday. The theory put forward capitalises on the traditional observation that early child grammars characteristically lack lexical and morphological elements which belong to the 'closed-class' system. This book provides an account of the grammatical differences between the set of functional categories and the substantive categories.

The Prefunctional Stage of First Language Acquisition

I

The theoretical framework

1.1. INTRODUCTION

The ultimate aim of any adequate theory of language is to account for two basic problems. The first is how languages, despite superficial differences, are identical at some deep and abstract level. The second is to define the range of possible variation among languages in a principled way. In the Principles and Parameters framework (cf. Chomsky (1981), (1982), (1986a), (1986b), (1991), (1992)), the first issue is dealt with by assuming the existence of a set of predetermined principles called Universal Grammar (UG). These principles govern the structural and grammatical properties of all languages of the world by virtue of biological necessity. The fact that languages exhibit fundamental similarities therefore follows from the assumption that they share a common set of grammatical principles, i.e. UG.

The second issue is dealt with in terms of a finite set of parameters with a number of open values associated with them. The fixing of these parametric values in a (finite) number of different ways, results in linguistic differences which are manifest in terms of (clusters of) surface phenomena. Language variation is therefore reduced to variation in the values of parameters. The requirement that the set of parameters and the values associated with them should be finite guarantees that languages differ only in restricted and (formally) defined ways (cf. Chomsky (1986a)).

These underlying assumptions are shared by all researchers who subscribe to the Principles and Parameters theory. Consensus breaks down when the question of the nature of the elements with which parameters are associated is considered. In this respect, there are basically two major hypotheses in the literature. According to the first (cf. Chomsky (1986a)), parameters are assumed to be associated with UG principles. In the second hypothesis, articulated in Borer (1984), parameters are associated with individual lexical items. These two hypotheses have different implications for the phenomenon of language variation as well as the process of language acquisition. The question of parameterisation (or lack thereof) in early child language is the core issue discussed in this book. As for language variation, I will briefly present some of the syntactic implications that arise under each of the two major hypotheses, concentrating in particular on the second one, which I adopt.

In Chomsky (1986a) parameters are conceived of as some sort of switches with a number of open positions. The selection of a specific value yields a pattern of linguistic properties, while an alternative parametric value yields a different pattern. A representative example of this approach to parameterisation is the Head-Parameter of X’-theory.

X’-schemata are assumed to provide a universal format according to which constructions are structurally represented. Configurational notions, such as Specifier of X and Complement of X are thus represented as in (1):

According to (1), a complement (i.e. Z") is a sister of the head category while a specifier (i.e. Y") is a sister of the single-bar projection of the head category. Note, crucially, that directionality constraints on the position of specifiers and complements are not defined in the schema in (1). The position of complements in a given language is argued to be fixed according to the value of the Head-parameter adopted in the language in question. The Head-initial value characterises head-complement languages like English, while the Head-final value characterises complement-head languages like Japanese. In this way, the Head-Parameter predicts a uniform pattern in the ordering of heads with respect to their complements regardless of the categorial status of the heads involved.

This prediction, however, appears to be problematic in the light of empirical evidence from a number of languages. In particular, in some languages, e.g. Dutch and German, verbs take their complements to the left, thus giving rise to the OV order, while adpositions take their complements to the right (i.e. PO). It thus seems that the general formulation of the Head-parameter fails to account for the variation attested within head-categories in a given language.

In addition, it has been argued on the basis of crosslinguistic evidence that the general formulation of the definition of governing category, i.e. the syntactic domain in which conditions on the distribution of anaphors and pronominals are specified (Chomsky (1981)), is inadequate (Wexler & Manzini (1987)). In particular, the distribution of pronominal and anaphoric elements exhibits not only interlanguage variation but also variation within the same language (e.g. Icelandic (Wexler & Manzini (op.cit.) and Chinese (Huang (1993)). Thus, the values specified in the governing category parameter are argued to be associated with individual lexical items in order for variation in the distribution of pronominal and anaphoric elements to be accounted for. Such considerations give rise to an alternative theoretical approach to the nature of parameterisation, in particular, the Lexical Parameterisation Hypothesis (Wexler & Manzini (1987)).

1.2. PARAMETERS AND FUNCTIONAL CATEGORIES

According to the alternative approach to parameterisation, parameters are associated with individual lexical elements (Borer (1984), Chomsky (1991), Ouhalla (1991a)).¹ More specifically, parameters are assumed to be exclusively associated with functional categories. These include the elements standardly referred to in traditional literature as the ‘closed-class system’ of a language, e.g. Determiners, Complementisers, Inflection (Tense and Agreement) and Negation.²

Recent syntactic developments have given this approach to parameterisation strong empirical support as the formulation of some of the major parameters involves properties of functional heads. For example, the null-subject parameter (Rizzi (1986), Chomsky (1986a)) is assumed to be regulated by the nature of AGR-S. The distinction between French-type and English-type languages in the distribution of VP-adverbs and negation has also been associated with properties of the AGR head (‘weak’ vs ‘strong’ AGR) (Pollock (1989), Chomsky (1991), Belletti (1990)). In addition, the V2 phenomenon as instantiated in languages like German and Dutch has also standardly been associated with the functional category C (cf. den Besten (1977), Tomaselli (1989)).

Returning to the issue of fixing the order of a head-category with respect to its complement, it is clear that the original formulation of the Head-parameter is at odds with this theory of parameterisation. In the framework of Ouhalla’s (1991a) theory, directionality restrictions, being parameterised properties, are associated with functional rather than substantive elements. Thus, directionality constraints in the selectional properties of functional heads are argued to give rise to crosslinguistic variation in word-order.³ In particular, the fixed order of the verb and its complement is argued to be the result of a choice formulated in terms of a Directionality parameter (Ouhalla (1991b)). In his framework, the functional category in question is AGR-O which selects its VP complement either to the left or to the right. This gives rise to two alternative possibilities adopted by VO and OV languages respectively:

Movement of the verb to the head AGR-O and thereafter to inflectional heads higher in the clause structure and of the object DP to (Spec,AGR-O) gives rise to the VO order in (2a), while a similar process in (2b) results in the OV order. In the absence of directionality restrictions associated with substantive elements, in this case verbs, the order inside V’ is assumed to be free (Borer (1993), cf. Kayne (1994)). Hence, this parametric choice is also shown to involve the presence of a functional category (i.e. AGR-O).

Movement of the object to the Specifier of AGR-O is motivated by Case considerations. In particular, the object moves to this position in order to receive (Accusative) Case as a result of the Spec-head agreement process in this configuration. Note, moreover, that the Specifier is assumed to occur in a fixed position, namely to the left of the functional head. This is a general requirement on Spec’s of functional categories formulated in terms of a directionality constraint on their licensing (cf. Ouhalla (1991a), Giorgi & Longobardi (1990)).

I will also assume that one of the basic properties of functional categories is that they project their own X’-structure (Pollock (1989), Chomsky (1991) & (1992)). Thus, a fully-fledged functional structure is represented as in (3) (word-order inside VP irrelevant).⁴

Following current syntactic proposals (cf. Kitagawa (1986), Kuroda (1985), Fukui (1986), Koopman & Sportiche (1988)), I assume that subjects are base-generated inside VP where they are assigned a theta-role and, subsequently, move to (Spec,AGR-S) for reasons to do with Case. I will take this assumption to imply that VP constitutes the thematic domain of the verb. Thus, the representation of all its arguments inside the VP projection can be viewed as a result of the requirement for thematic positions to be represented within the projection of the selecting head. This assumption will be shown to have direct consequences for the nature of early grammars and the structural representation involved at the relevant stages.

Going back to the structure in (3), recall that objects receive an account similar to subjects in that their raising from the position inside VP to (Spec,AGR-O) is motivated by Case considerations. Thus, structural Case-assignment to subjects and objects is understood to operate under the Spec-head agreement process in a parallel way. This unified analysis of subjects and objects in relation to the presence of the respective AGR projections will be argued to give rise to a number of predictions in relation to word-order and null arguments in early child grammars.

In sum, the theory of parameterisation adopted here assumes that parameters are exclusively associated with functional categories. Directionality restrictions are parameterised properties attributed to functional rather than substantive elements. In this respect, the order inside the VP projection is predicted to be free. Word-order becomes fixed as a result of verb-raising to the functional categories higher in clause structure as well as movement of the arguments to the Specifier position of AGR-S and AGR-O respectively. Structural Case-assignment of subjects and objects takes place in the Spec-head configuration within each AGR projection.⁵

Bearing the above framework in mind, I will now turn to the discussion of current theoretical approaches to language acquisition and the presentation of an alternative theory which will be defended on both theoretical and empirical grounds.

1.3. LANGUAGE ACQUISITION

In the context of the Innateness Hypothesis the language faculty is assumed to constitute part of the biological endowment of the human mind/brain. The internal structure of the Language Faculty, the module of the brain specific to language, is considered to be equipped with certain mechanisms referred to as Universal Grammar (UG). Prior to exposure to linguistic input UG is said to be in the Initial State (SI) while the mature state reached following the interaction of UG with ‘primary linguistic data’ is referred to as the Steady State (S_s). The transition from the S_I to the S_S given the relevant input data can be viewed as taking place instantaneously, in that intermediate states do not affect the steady state ultimately attained (Chomsky, 1986:54). This model is, of course, an idealisation of the actual developmental process underlying language acquisition. Recent work concentrating on the grammatical properties attributed to each developmental stage, i.e. S1, S2, S3 etc., has shown that there is a considerable uniformity in the pattern of crosslinguistic data which can lead to interesting generalisations regarding the nature of UG.

Within the Principles and Parameters framework, the standard assumption underlying the theory of language acquisition is that it involves a process of fixing parameters to the values of the target grammar. The range of open values associated with each parameter is defined by UG, whose Principles constrain the nature of the grammars constructed at the various stages of linguistic development. These assumptions are shared by all researchers in the field of language acquisition. Diversity of opinions arises with respect to the availability vs non-availability of UG Principles and/or parameters during the acquisition process.

There are two main theoretical approaches which have been argued to find empirical support from acquisitional data: The Maturation Hypothesis and the Continuity Hypothesis. Maturation assumes that certain grammatical properties are missing in early child speech due to the comparatively late emergence of the grammatical categories or principles regulating their appearance. Within the Maturation Hypothesis there are two distinct approaches: one assumes that maturation affects the Principles of UG (Felix (1984)) and Borer and Wexler (1987)) while the second position argues that maturation affects functional categories (Radford (1988), (1990), Guilfoyle & Noonan (1988), Tsimpli (1991a)). The theory of acquisition I will present later on in this chapter is along the lines of the Maturational approach plus the assumptions made by the theory of parameterisation outlined in the previous sections.

Continuity, as opposed to Maturation, assumes that UG Principles are available throughout the process of language acquisition (Pinker (1984), (1987), Hyams (1986)). In terms of recent theories of parameterisation (see section 1.2.1.) in which parameters are associated with functional categories, the Continuity Hypothesis claims that, even in the initial stages of language acquisition, the clause structure is similar to the clause structure defined by the adult grammar, in that all functional categories and projections are accessible to the learner and not constrained by maturation (Pierce (1989), Weissenborn (1990), Hyams (1994) among others).

In the following section I will outline each of the main approaches and their claims with regard to specific syntactic phenomena in early child grammar.

1.3.1. THE CONTINUITY HYPOTHESIS

The Continuity Hypothesis (Pinker (1984), (1987), Hyams (1986), (1987)) claims that all UG Principles are available right from the start. The grammars constructed at all stages of linguistic development are thus argued to be ‘possible’ grammars in a sense defined by UG. Parameters, on the other hand, are also available, albeit not as yet fixed to their target value. Parameter-setting relies on the availability of an appropriate set of data referred to as ‘triggering’ data. These data are distinct from input data in that it is only at a certain stage of language acquisition that their presence in the linguistic input leads to parameters being set to their target value.

Hyams (op.cit.) argues for an account of pro-drop in child speech in terms of the Continuity Hypothesis. In particular, she claims that the absence of parametric differences between early child Italian and early child English with respect to the availability of null subjects can be accounted for on the assumption that the initial setting of the pro-drop parameter is to the positive value. If [+pro-drop] is the default value the argument goes as follows: the AG/PRO parameter is initially set to the positive value regardless of the target grammar, hence AG=PRO for both child English and child Italian. The structure inside I consists of both an AUX and an AG node as illustrated in the following tree diagram:

Given that AG=PRO at the stage under discussion, it follows that modals which occupy the AUX position in adult grammar are not allowed to occur, because of the UG constraint on the distribution of PRO, namely that it be ungoverned. The absence of modals and auxiliaries at the relevant stage is thus accounted for. The emergence of modals as well as the use of expletive subjects in child English, Hyams argues, lead to the parameter being reset to the negative value, i.e. the target value. In other words, modals and expletives serve as ‘triggering’ data for the child to set the parameter to the appropriate value. The situation for the Italian child, however, is different. On the assumption that, in Italian, modals (potere and dovere) are main verbs (base-generated under a V node) and auxiliaries (essere and avere) are similarly base-generated under V, the AUX position remains empty. Thus, in adult Italian AG=PRO, as no lexical element occupies the AUX position, thus avoiding a violation of the PRO theorem. Moreover, in the absence...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Original Copyright Page
• Dedication
• Table of Contents
• PREFACE
• ACKNOWLEDGMENTS
• 1 THE THEORETICAL FRAMEWORK
• 2 ASPECT
• 3 AGREEMENT AND NULL ARGUMENTS
• 4 WORD-ORDER
• 5 NEGATION AND MODALITY
• 6 SUMMARY AND FURTHER ISSUES
• BIBLIOGRAPHY
• INDEX