This chapter consists of an investigation of an optional operation in English. The experimental part of this chapter is based on two studies with English- and with Italian-speaking children, reported in Guasti et al. (1995) and Guasti (1996), respectively. The optional nature of this operation will be questioned and the more economical option will be identified. In the second part of the chapter, the experimental studies will be reviewed and support for the current proposal will be drawn from the discussion.
1.1 Variation in the target grammar
English negative interrogatives allow for variation with respect to the position of the negation element. This is demonstrated in (1) and (2) below.
(1) a. What do you not like to eat? b. What dont you like to eat?
(2) a. Did John not eat? b. Didnt John eat ?
The element under investigation here is the negation element. It can appear in a post subject position (presumably NegP) in its full form not, or in a pre-subject position (presumably C) in its clitic contracted-- form nt. The structures in (1), as well as in (2), are both grammatical and appear to be identical in meaning; therefore they appear to answer our requirements of optionality. The structures in (1b) and 154 Summary and conclusions (2b) are the result of a movement of the negation element from NegP to C. Under the minimalist assumption that every movement is either motivated or blocked, these alternatives constitute a problem. The usual explanation for the existence of this optionality is that the negation in its clitic form adjoins the auxiliary in its I- to-C movement as a freerider and is thus costless and spared by economy. However, under this analysis we would expect to find a similar alternation for the two negation forms (the full and the contracted) in all cases that exhibit subject-auxiliary inversion. This does not seem to be the case, however. Consider the following sentences.
(3) a. Had I not gotten there in time, he would be already dead b. * Hadnt I gotten there in time, he would be already dead
(4) a. Never have I not attended a meeting b. * Never havent I attended a meeting
Under the analysis that the movement of Neg to C is a result of a freeridewith the auxiliary, we would have expected (3b) and (4b) to be grammatical, as the auxiliary had also undergoes a subject auxiliary inversion here to create the subjunctive form (in 3b) or to satisfy the requirements of a negative item in Spec CP (in 4b). The fact that in these cases this movement is blocked prevents us from assuming a free-rider analysis for (1b) and (2b) and thus calls for a motivation for this clitic-movement. McClowsky (1997) observes that movement of a negation element to C is obligatory in the presence of a negative polarity item in subject position. This is demonstrated in (5), below.
(5) a. * Which of the kids does anyone not like? b. Which of the kids doesnt anyone like ?
In (5a), the negative element remains in I and thus fails to establish a C-command relation with the negative polarity item anyone, while in (5b) the moved negation C- commands the polarity item and thus the sentence converges. This example again demonstrates that the free distribution exhibited in (1) and (2) does not hold in all contexts. The movement of the negative element in (5b) can thus be described as scope related. Scope of the negation seems to also play a role in the following Summary and Conclusions 155 examples, in which this Neg-movement to C creates a difference of interpretation, rather than of grammaticality:
(6) a. Did anybody not bring a present? b. Didnt anybody bring a present? (7) a. Is John not really a spy ? b. Isnt John really a spy ?
While the examples in (3)-(5) demonstrated a difference in grammaticality created by the Neg-movement to C, the examples in (6)-(7) demonstrate a difference in meaning between the base and the derived forms. In (6), the negative polarity item anybodyis licensed through the yes/ no-question; thus both variants are grammatical. However, the scope-effect of the position of the negation still plays a role leading to different readings for the two variants. In (6a) the question presupposes that most people have brought a present, while the question in (6b) presupposes the opposite, that most people did not bring a present. In (7) there is again a difference of presupposition, where in (7a) the questions assumes that John pretends to be a spy and questions him and in (7b) assumes that John pretends not to be a spy. In both (6) and (7) a no answer has different truth conditions for each variant. Further research is undoubtedly required in order to account for the differences presented in (3)-(7), and to establish the exact nature of the feature(s) that are responsible for this Neg-movement to C. For our purposes, it is sufficient to note that the existence of examples demonstrating differences of both grammaticality and interpretation casts a serious doubt on the view of Neg-movement to C in negative interrogatives in English as a case of a freerider or as manifestation of optionality.
Another difference between the two variants in (1) is the register. In the spoken language, the contracted form of negation (1b) predominates, but in the formal written language the full negation in base-position predominates. As in the case of triggered inversion in Hebrew, distinguishing these variants in register seems to be accurate but not exhaustive; the examples in (3)-(7) demonstrate that factors other than register play a role. An important difference between the current case and the triggered inversion case, is that in triggered inversion the formal variant was the less economical one (the one that did not include inversion) while in the current case the formal variant (1a) is the more economical one. This difference creates an excellent 156 Summary and conclusions testing environment for the proposal made in this dissertation, since the current proposal predicts that children will show a preference for the formal, less-frequent variant (1a, above), i.e., the more economical one. This prediction opposes the common assumption that children always prefer forms that are more frequent and colloquial. The conclusion of this section is that movement of Neg to C in English joins the structures investigated in the previous chapters in reflecting an apparently optional operation that in fact depends on a subtle difference between the two variants. The current proposal, based on the assumption that, in the early stages, children miss this subtle difference (and the dialectal distinction as well), predicts that children will show a preference for the structure in which the Neg-element remains in a lower position, because it is the more economical alternative.
2. Review of experimental investigation
Guasti, Wexler and Thornton (1995) investigated the production of negative interrogatives in ten English speaking children, aged between 3;8 and 4;7, with a task of elicited production. The authors used a shy puppet, and encouraged the children to ask the puppet negative questions; as a control, they encouraged negative declaratives and positive questions as well. The negative questions were divided into three categories: object and adjunct questions that require subject-auxiliary inversion, and subject questions that require no inversion. The stimuli presented to the children to elicit a negative question included a contracted form of negation in a declarative structure, shown in (8), below.
(8) Stimulus to elicit a negative question (from Guasti etal., 1995): Experimenter: I heard the snail doesnt like to eat some things, Ask him what. Target: What dont you like to eat?
The results reported in Guasti etal.(1995) are surprising. While children performed in a close to adult fashion in respect to positive questions and negative declaratives, they did not do so when posing negative questions. More specifically, their performance was poor when dealing with object- and adjunct- negative questions, yet they produced subject negative questions that were adult-like in form. The non- adult-like structures produced by the children for the negative questions were of Summary and Conclusions 157 four types. These appear below in order of the frequency with which they were produced.
(9) Aux-doubling structure What do you dont like to eat? (40%)
(10) No subject-aux inversion (no SAI) What you dont like to eat? (23%)
(11) Not-structure What do you not like to eat? (10%)
(12) Neg+Aux doubling What dont you dont like to eat (8%)
Adult-like responses were produced in 19% of the questions. The authors analyzed the results through the criterial approach (Rizzi 1991, Haegeman and Zanutini 1991). This view considers interrogatives to be controlled by the requirements of the Wh-criterion and negation to be controlled by the requirements of the Neg-criterion. From their results, the authors concluded that children go through an intermediate stage characterized by a mastery of the Wh- criterion (as shown by the production of positive questions) yet at the same time mistakenly think that the Neg-criterion must be satisfied within the IP. The authors show that such a restricted Neg-criterion is possible in UG; in fact, such a restriction is manifested in the Paduan language. This assumption, of a conflict between the requirements of the Wh-criterion and those of the (restricted) Neg- criterion, yields the 4 response types shown in (9)-(12); all of them reflect a failure to raise the negation from I to C. Their analysis leads to the following proposal:
We suggest that the Neg-criterion is subject to parametric variation. Broadly speaking in some languages such as Paduan and child English it must be satisfied in the IP; in others it may be satisfied in the CP, as in adult English and standard Italian. Children learning English initially adopt the most restrictive hypothesis. That is, they assume that the Neg-criterion must be satisfied in the IP, hence the prolonged period of non-adult questions observed in their negative questions. Guasti et al. (1995), p.237. 158 Summary and conclusions Although the authors do not refer to the notion of economy in their analysis, we could classify their proposal as belonging to the local approach presented in the first part of this dissertation. That is, they refer to an initial setting of the relevant parameter, which is more restricted than the target setting. This proposal is local in the sense that childrens intermediate stage is assumed to be independent of other alternatives children might think they have for the target structure. I wish to propose a different explanation for Guasti et al.s results. Consistent with the main proposal of this dissertation, presented in chapter 3, this is an economy- based solution, global in nature, which assumes that children do in fact compare the target structure to another alternative they think exists in the target grammar. In the first part of this chapter we saw that the freerider concept does not explain Neg-movement from I to C, despite this movements dependency on the auxiliarys I to C movement. We also saw that in several contexts the Neg-movement is forbidden even though SAI has occurred and that in other contexts, such movement leads to a semantic difference. We concluded that the motivation of the negations movement is independent of the motivation of the auxiliarys movement. Although the precise nature of this motivation, as well as the feature(s) responsible for it, are not yet clear, let us assume here a tentative neg-feature, located in C, that requires (in English) overt movement of negation to C. This assumption is essentially identical to the assumption of a parameterized neg-criterion made in Guasti et al. although Guasti et al. it is formulated in the terms of the criterial rather than the minimalist approach. The difference between the current proposal and Guasti et al.s, is the view of the relevance of the input to the acquisition of the negative interrogative structure. Based on the analysis in the first part of this chapter, it is concluded that children are exposed to two types of negative questions; one type has negation in C, the other has negation in I, as described in (1) above. Children fail to distinguish these two forms in their interpretation, considering them to stem from the same numeration and to be identical in meaning, and therefore see them as contradictory evidence for the relevant parameter. Based on the proposal of the current dissertation, children reject the possibility that both alternatives are allowed and show a preference for the one in which negation remains in a lower position, because it is more economical. The findings of the study presented in Guasti et al. (1995) are explained, thus, through the claim that children consider the Neg-in-I (1a) option to be identical to Summary and Conclusions 159 what the authors call the target-structure with the Neg in C (1b); therefore, children opt for the former because it is more economical. However, the children produced structures of the (1b) type in only 10% of their responses, as shown in (10) above. Although this percentage is relatively high and cannot be dismissed (the authors specifically note that children do not use this not-structure in situations where adults would generally prefer the Neg-in-C variant), why children produce the other 3 types of non-adult structures, especially the predominant aux-doublingstructure ((9) above), must be explained. The explanation seems to be related to the childrens preference for the contracted form of negation 41 . Recall that the stimuli in Guasti et al.s experiment included a contracted form of negation, by far the predominant form in childrens input. For this reason children, although they show a preference for the negation in I, were forced to produce an extra auxiliary to support the contracted negation, because the true auxiliary moved to C to check its features. This resulted in the aux-doubling response 42 . In principle this is similar to the explanation of the conflicting requirements of the Wh and Neg criteria proposed by Guasti et al.
3. Conclusion
The findings of Guasti et al.s (1995) can be explained by the assumptions made in the current study, and can be understood to support the main proposal of this dissertation. The difference between the proposal made here and Guasti et al.s involves the distinction between global and local. Guasti et al.s proposal claims that there is no alternative to the Neg-in-C structure in the childrens input --hence it is local--, while the current proposal connects childrens disfavoring of this structure to the fact that a more economical alternative, with Neg in I, appears in their input (and is thus global). An empirical prediction that distinguishes the two proposals is formulated in a language in which no alternatives to Neg-in-C exist (that is, a language with obligatory Neg in C in negative interrogatives). In such a 41 This is not to say that the two negation forms represent different lexical entries. Our assumptions require us to assume that the choice of realizing negation as not or as nt is made later in the derivation (but prior to the I-C movement of the neg-element). 42 Questions with the auxiliary bewere omitted from the analysis in Guasti et al. (1995) although some of the questions produced by the children included this auxiliary. In a footnote the authors mention that in this type of questions children have used the not-structuremuch more often and refrained from the aux-doubling structure. This fact supports the analysis proposed in the current chapter as it reveals the relation of childrens preferences to the input they receive. (In the bestructures in the input the not form is more frequent then the nt, as the auxiliary is contracted with the subject). 160 Summary and conclusions language, Guasti et al.s proposal predicts an intermediate stage characterized by a failure to raise negation similar to the one revealed in childrens English, but the current proposal predicts no such failure, since no alternatives exist. Guasti (1996) presents an experimental investigation of negative questions in Italian, similar to the one performed in Guasti et al. (1995). In Italian, there is an obligatory movement of Neg to C in negative questions. Guasti et al. (1995) predict that Italian-speaking children will fail to perform this operation, and thus expose the intermediate stage of restricted Neg-criterion. However, the author has found no errors in the negative questions of Italian- speaking children. This finding supports the current proposal, in which a failure with movement operations is related to the existence of an apparently more economical alternative in the input. As such alternatives do not exist in Italian, but do exist in English, the current proposal accounts for the early negative questions of both English and Italian speaking children.
The combination of the findings in Guasti et al. (1995) and Guasti (1996) has been shown to support the general hypothesis in this dissertation. Furthermore, if we compare the phenomena of the current chapter to triggered inversion in Hebrew, two additional observations are relevant. First of all, the local and the global approach to the non-adult like stages in childrens grammars are distinguishable and the global approach is supported. The second point is that in both chapters the optionality involved was claimed to be related to register, but with a major difference. In triggered inversion the more economical option was also the colloquial one, but in negative questions in English, the more economical option is the formal one. The fact that in both cases economy, and not register, correctly predicted childrens preferences supports the current proposal and shield it from possible objections which are based on notions of canonical form and frequency as predictors of childrens preferences.
Summary and Conclusions The purpose of this dissertation was to answer the following empirical question:
(1) How do children deal with syntactic input that seems to be optional?
In order to answer this question the minimalist framework was adopted as a theoretical basis. It was shown that within the Minimalist Program, optionality of word-order cannot be accounted for. The fundamental problem of optional word- order is its clash with a central principle of the MP, the principle of economy. In the first part of this dissertation we have reviewed the relations between optionality and economy from a theoretical perspective. Solutions proposed in the literature for resolving the optionality-economy tension were discussed, and a proposal was made according to which, true optionality (with respect to word-order) does not exist in natural languages. It was claimed that optionality is undesired from several perspectives, independently of its clash with economy, and that economy is valuable to the theory, independently of its clash with optionality. Therefore, the best way from a theoretical point of view is to preserve the principles of economy in their full and to assume that structures that seem to be optional (i.e., identical in meaning) are in fact distinct from each other. Before turning to empirical evidence to support the no-true-optionality assumption, the notions of economy and optionality were discussed from the point of view of language acquisition. It was claimed that also in acquisition, economy can be seen as an extremely valuable tool while optionality is problematic and hazardous to the process of acquiring a language. The conclusion was that the language-acquiring child rejects the possibility of optional movement and opts for a "one form - one meaning" hypothesis and for unique relations between word-order and interpretation with respect to the target grammar. Furthermore, and maybe most importantly, it was claimed that when the child encounters in her input structures that appear to be the result of an optional operation, she will make use of economy principles as a tool for selecting between the two alternatives, opting for the more economical one. This last claim is in fact the answer proposed for the question in (1) above. 162 Summary and Conclusions The general proposal of this dissertation, formulated in chapter 3, consisted therefore of two related claims. The first is that true optionality does not exist in natural language and that for each case of apparent optionality, it should be possible to identify a difference between the two variants. The second claim was that when such apparent optionality appears in the input to children, they reject the possibility of accommodating both variants in their grammar and choose to prefer the more economical order. The connecting idea behind these two claims is that the principles of economy have a dual function in the language component: it serves the computational process in selecting between competitive derivations and it serves the acquisition process in selecting among alternative input-strings. In part II of the dissertation these two claims were put into empirical examination. Five constructions in four different languages were presented. These constructions were shown to fall under the definition of "appear to be optional", as for each of them several alternatives exist which seem at first sight to be indistinguishable with respect to their interpretation. Each of the chapters in part II was dedicated to one such "optional" construction, and in each the two claims above were examined. Namely, in the first part of each chapter the alternative word-orders were examined from a theoretical perspective, aiming to show that they in fact point to distinct interpretation, presumably as a result of distinct features the relevant elements carry. Further the underlying structure was established and the more economical alternative was identified. In the second part of each chapter, these alternatives were investigated with respect to child language. It was shown that indeed children show a preference toward the more economical word-order. The preference of the children were established on the basis of a comparison of children's and adults' production of each of the alternatives, and based on a repetition task through which children were shown to convert the costly word-order into the more economical one, but not vice versa. Apart from supporting the proposal made in part I, the experimental results in each chapter made an additional contribution to the topic under investigation, by illuminating various aspects of the manner children make use of economy. For example: chapter 4 in which embedded participle-construction in Dutch were investigated, and chapter 6 in which auxiliary+infinitivestructures were examined, have shown that the phenomenon of omission of elements is related to economy considerations. The experimental data in these chapters have shown that omission of auxiliaries and main verbs by children is more probable to occur in structures that Summary and Conclusions 163 require overt movement of these elements. The data in chapter 6, 7 and 8 have indicated that children disfavor a structure only if a more economical alternative is present in their input. That is, structures that involve movement but are obligatory in the target grammar were shown to be acquired with ease. Chapters 5 has shown that the economy considerations are interacting with the proportion of each alternative in the input, while chapter 5, 6 and 8 have shown that such influence of the input, in terms of frequency, colloquiality and 'canonisity' are limited and different for different structures. The influence of economy, in comparison, was shown to be consistent and significant in all the structures investigated. The proposal made in this dissertation with respect to acquisition is clearly falsifiable and easy to test. Once alternative derivations are identified in a language, children acquiring this language are predicted to go through a stage in which their production of the more economical alternative exceeds its proportion in the input. In any case children are predicted to never produce this alternative in a proportion lower than its proportion in the input. Once such case is observed, the current proposal must be reevaluated.