Acquisition of SV and VS Order in Hebrew, European Portuguese, Palestinian Arabic, and Spanish

Abstract

In Hebrew, European Portuguese, Palestinian Arabic, and Spanish, both SV and VS orders are possible. However, when children acquire these languages, they do not use the whole array of word orders in their language at the first stage of sentence construction. Interestingly, their word order preference in the early stage of acquisition differs in the different languages: in Hebrew and European Portuguese they use both SV and VS orders with unaccusative verbs, but only SV with unergative and transitive verbs. In Spanish and Palestinian Arabic, children prefer to use VS order with unaccusative, unergative, and transitive verbs in the first stage. We present 11 experiments and 5 spontaneous speech analyses in the four languages, eliciting these word order patterns in 257 different children, and analyzing the patterns in spontaneous speech in 80 more children. Based on these results we propose an account for these cross-linguistic differences and similarities, according to which children at this stage can already move the verb to I, but cannot move the subject outside of the VP yet. As a result, at this stage the subjects stay within the VP. Whether or not the verb moves to I depends on whether the linear order of SV within the VP can be violated—languages in which children assume IP as the Spell-out domain, Spanish and Palestinian Arabic, allow the verb to appear before the subject, whereas Hebrew and European Portuguese, in which the Spell-out domain is initially taken to be VP, do not allow the verb to move to I until the Spell-out domain widens.

Notes

¹In Hebrew, some types of DP have to precede the unaccusative verb: pronouns, proper names, and names of family relatives (daddy, grandma) cannot follow the unaccusative verb (Friedmann 2007). For the other languages, similar effects appear but just as preferences.

²Our analysis of a large sample of Hebrew children's books read to young children, encompassing 4941 sentences, indicated that 19.4% of the sentences with unergative or transitive verbs appeared in VS order. An analysis in Zuckerman (2001:143) indicates that in more formal and media contexts, 87% of the sentences that start with a possible trigger (nonsubject) are produced in VS order. Namely, children who overhear the news or the sports report on TV may be even more exposed to VS sentences than to SV. In Taub-Tabib's (2009) analysis of 561 declarative sentences, randomly sampled from a large corpus of blogs written in colloquial Hebrew (Linzen [2009] corpus, literary style blogs were excluded from the analysis), 90% of the sentences with unergatives appeared in SV order, and only 10% in VS order. Unaccusatives showed a different pattern, with 72% of the unaccusatives appearing in VS order and 28% in SV order.

³For the sake of this description, we are abstracting away from the fact that some word orders are preferred in some specific discourse or syntactic settings (Shlonsky 1997; Costa 1998; Zubizarreta 1998, among others). Importantly, though, we do not assume, as in Belletti (2004), that the word orders with inversion in which the subject is focalized involve a low focus projection or remnant movement to the left of a preverbal subject (Ordóñez 2007, among others). For arguments against low FocusPhrase in inverted structures, see Costa (2004) and Costa & Figueiredo Silva (2006).

⁴This analysis included definite nouns, as well as personal and demonstative pronouns. When analyzing only the sentences with a definite noun subject, there is only a slight preference for SV, 52%.

⁵Another child who was already 5;07 produced a VS sentence without a trigger with both an unergative and a transitive verb.

⁶The type of evidence discussed by these authors includes the distribution of floating quantifiers and positions made available for the argument in inversion contexts, among other constructions. All evidence confirms the idea that there is an intermediate argument position in between the source object position and the target VP-external subject position. For discussion and detailed argumentation, we refer the reader to Silva (2004) and Friedmann (2007).

⁷The fact that syntax makes the two word orders (SV and VS) legitimate can explain the existence of optionality in children's productions. Since the two strategies, SV and VS, are available, they can both be used.

⁸Clitic doubling is subject to dialectal variation in Spanish. Nevertheless, the experiments were run on Argentinian Spanish speakers, where doubling is more common, and the spontaneous speech analyzed was from the Spanish of Spain, where doubling is common as well (Suñer 1999). All varieties of Spanish exhibit a certain degree of doubling.

⁹By hypothesis, the finding that there is early V-to-I movement in French may show that any type of clitic doubling (object and subject) may serve as relevant triggering experience. As shown in De Cat (2003), French children master subject clitics and doubling very early.

¹⁰This makes it clear that the acquisition of V-to-I movement precedes the acquisition of Subject-movement. As such, acquisition of Subject-movement is a prerequisite for the visibility of the effects of V-to-I, complementing the observations made for VP-ellipsis. As argued in Costa & Santos (in press), the visibility of V-to-I in child language data is more complex, since it depends on the cumulative satisfaction of independent requirements, which may create the illusion that this operation is acquired late, which is not supported by spontaneous speech data.

¹¹Notice that we do not suggest that the children do not project CP. Clearly, at the stage where they do not move the verb to C they already have CP, as witnessed by their ability to do wh-movement before they can move the verb to C, and by the finding that the same children who cannot produce V-to-C can already produce embedded sentences with sentential complements to verbs (see Experiment HE4). For example, all 36 Hebrew-speaking children in Experiment HE6, including the 7 children who could not repeat VS sentences with unergative and transitive verbs, could repeat embedded sentential complements of verbs flawlessly (100% correct), and could produce subject relatives in a preference elicitation task (with an average of 96% correct). The task is described in Friedmann & Szterman (2006) and Novogrodsky & Friedmann (2006).