This thesis investigates the role of implicit knowledge in second language acquisition, presenting five experiments and related simulations based on artificial grammar learning. It examines whether second language learners can acquire implicit knowledge of noun–verb agreement. In addition it tests whether their ability to do so is influenced by the number of words that intervene between the relevant noun and the finite verb in the input sentences, as this affects performance in artificial grammar learning, the serial response time task and the statistical learning paradigm. Experimental participants were exposed to a modified version of Persian or Basque while performing a memory task. Next, two grammaticality judgement tests (one timed and one untimed) and a sentence correction task assessed whether the participants had acquired either the target noun–verb agreement or pairs of adjacent words that appeared frequently in the learning items. Performance reflecting implicit and explicit knowledge was distinguished according to three criteria based on R. Ellis (2005) and according to the assumption that the former is not under conscious control. Participants’ performance suggested that they had implicit knowledge of the adjacent word pairs. Similarly, the results indicated that they had implicit knowledge of subject–verb agreement when a single word intervened between the subject and the verb, but not with two intervening words. Connectionist simulations supported the results and indicated that performance was unlikely to improve if more exposure were given. Although the influence of additional factors is discussed, the results generally supported the view that an increase in the number of intervening words reduces learning outcome. The intriguing similarity between the results of this thesis iv and previous research in artificial grammar learning, the serial response time task and statistical learning experiments suggests that future research should directly compare the paradigms to ascertain whether similar learning processes are engaged in each case.

This thesis is a case study on the emergence of functional categories in bilingual first language acquisition. The investigation focuses on the transition from one-word to multiword utterances and the shaping of functional projections of Determiner, Agreement and Tense and their associated formal features. The empirical basis of this work is a corpus of thirty-nine videorecorded observations of Carlo, an English-Italian bilingual child, during free-play sessions with an adult. Data was collected separately for English and Italian for a period of fifteen months from when the child was 1;10 until he was 3;1, and was then transcribed in CHAT format. Four interrelated lines of enquiry inform the analysis presented here. The principal research question concerns the acquisitional strategies adopted by C. in these early stages of development in the two languages. A bilingual child is the closest one can get to a perfect matched pair where a number of variables such as socio-cognitive development, socio-economic status, parents' education, etc. are eliminated, and the two main variables to be investigated are the child's two input languages. This is an ideal situation in which the respective roles of general acquisitional strategies and language particular ones can be teased apart. An analysis of the emergence of the morphosyntactic correlates of Determiner, Agreement and Tense categories in English and Italian reveals a discrepancy between the two languages in the age of acquisition, rate of acquisition and in the language-specific strategies the child adopts. The observation of a significant difference in C.'s acquisitional strategies in English and Italian leads us to the second and third research questions: the way in which the emergence of functional categories differs between the two languages, and the reasons why this should be the case. The most obvious difference is the extent to which morphological correlates of functional categories emerge in the child's speech. In Italian, verbal and nominal morphology emerges earlier than in English and, at least in the nominal system, there is evidence that an Agreement category is part of the child's grammar. In English, verbal morphology is virtually non-existent by the end of the period of observation, and there is no substantial evidence that either Agreement or Tense are realised. Lexically-specific, item-based learning plays a substantial role in both languages, but in Italian there is some evidence that a number of grammatical contrasts are becoming productive by age 3;0, albeit some of them are still limited to a small number of lexical items. Two reasons were identified for the observed differences in the emergence of Determiner, Agreement and Tense in English and Italian: a typological reason, and an environmental reason. The former concerns the richness of Italian morphology, where grammatical contrasts are transparently marked both on nominal and verbal paradigms, as opposed to the relative poverty of English morphology where such contrasts correlate less obviously ans systematically with morphophonological markers. The latter reason concerns the very different input conditions in which C. is exposed to Italian and English: Italian is the home language spoken to him by his family and his babsysitters, while he is addressed in English by the staff at the nursery where one adult is in charge of several children and cannot engage in the one-to-one interaction which is typical of the dyadic situation in which C. finds himself at home. The differences observed in the lead-lag pattern between C.'s Italian and his English also provide sufficient evidence to address the fourth research question concerning the separate developement of the two languages. The analysis of the data did not reveal any systematic interferences from one language to the other. On the contrary there is evidence that C. is sensitive to the different morphosyntactic cues of his two input languages, and that he can treat the two as independent, self-contained problem spaces.

This thesis investigates children’s mental representation of syntactic structure and how their acquisition and production of syntax is affected by lexical and semantic factors, focusing on three- and four-year-old children. It focuses on a construction that has been a frequent subject of language acquisition research: the passive. It is often claimed that English-speaking children acquire the passive relatively late in language development (e.g. Horgan, 1978): previous studies have typically found unreliable comprehension and infrequent production of passives by children younger than five (e.g. Fraser et al., 1963). However, there is some evidence from studies providing an appropriate pragmatic context for passives (e.g. Crain et al., 1987) and studies which increase children’s exposure to passives (e.g. Whitehurst et al., 1974) to suggest that children can produce this structure at a younger age. Converging evidence comes from studies of syntactic priming, or the tendency to repeat syntactic structure (e.g. Bencini & Valian, 2008). Syntactic priming effects are potentially informative about the nature of syntactic representation, as they are assumed to reflect the repeated use of the same syntactic representation across successive utterances. With respect to language acquisition, syntactic priming effects can be informative about the extent to which children have acquired an abstract representation of a structure. Specifically, if children have a syntactic representation of the passive, then it should be possible to prime their production of passives, such that they should be more likely to produce passives after hearing passives than after hearing actives. Furthermore, by examining the conditions under which such priming occurs, it is possible to draw inferences about the nature of their passive representation. This thesis presents seven experiments, six using a syntactic priming paradigm, to examine children’s knowledge of passives. Experiment 1 establishes a syntactic priming effect for actives and passives in three- and four-year-old children, and shows that priming occurs for both structures within an experimental session, using a withinparticipants design. Experiments 2, 3 and 4 examine whether young children’s acquisition of the passive is semantically constrained. Experiments 2 and 3 show that children can be primed to produce passive responses by actional and non-actional passive primes. Experiment 4, a picture-sentence matching task, replicates the results of other studies, however, showing that children find subject-experiencer non-actional verb passives more difficult to understand than actional verb passives; this mis-match between the results from the different tasks suggests that some effects of verb-type may be task-related. Experiments 5 and 6 examine whether the observed priming effect could be a lexically-driven effect that is dependent on the repetition of function words (the preposition by or the passive auxiliary). They show that this explanation can be ruled out: children are more likely to produce passives following both passive primes that do not express the agent using a by-phrase and passive primes involving a different auxiliary verb. Experiment 7 examines the later development of passive structures by testing passive production in six- and nine-year-old children. It finds evidence that at six, they still have difficulties with the construction, however by nine, children have an adult-like representation of the passive. I conclude that by four, children have begun to develop a syntactic representation for the passive which is already common to a range of different possible forms(short, full, get and be), and which is not restricted to particular semantic classes of verb. However, these results also suggest that children do not fully master the passive construction before six: young children make morphological errors and errors mapping thematic roles to syntactic positions, even following passive primes. Hence children may acquire the purely syntactic aspects of the passive, leading to a syntactic priming effect, before they acquire other aspects of this structure, hence the children’s occasional errors producing passives.

Chomsky’s Cartesian Linguistics frames into his linguistic work and the resulting debates between rationalists and empiricists. I focus on the following key aspects: (i) the historic connection of Cartesian Linguistics to previous linguistic theorizing, (ii) the development of Chomsky’s own theorizing, (iii) the empirical work addressing the problem of language acquisition and (iv) the problem of computational modeling of language learning. Chomsky claims that his view is situated within a rationalist Cartesian tradition and that only rationalists will be able to account fully for all aspects of human language. My thesis challenges both claims. I found only remote connections between Cartesian and Chomskyan commitments. Chomsky holds that (i) language is species-specific, (ii) language is domain-specific, and (iii) language acquisition depends on innate knowledge. Descartes accepted (i), but argued that language is an indicator of domain-general intelligence. Innate resources play a different role for language acquisition for Chomsky and for Descartes. Chomsky revived linguistics during the 1950s by promising to make it a rigorous part of the biological sciences. However, his work has not resulted in a better understanding of language acquisition and use. Key concepts like ‘innateness’, ‘Universal Grammar’ and ‘Language Acquisition Device’ remain in need of precise definition, and the Poverty of the Stimulus Argument does not rule out data-driven domain-general language acquisition. Empirical work in developmental psychology has demonstrated that children acquire and practice many language-related cognitive abilities long before they produce their first words. Chomsky’s dictum that language learning is uniform across the species and invariably follows genetically determined stages remains empirically unconfirmed. Computational modeling has accounted for some internal structure of language acquisition mechanisms and simulates the specific conditions under which children learn language. Contemporary models use samples of child-directed-speech as input and have replicated numerous aspects of human performance. Given my findings I suggest that Chomskyan linguistics is not Cartesian in substance or in spirit. Descartes was wary of those “who take no account of experience and think that truth will spring from their brains like Minerva from the head of Jupiter” (CSM I, p. 21). His science relied on sense experience (empiricism) and deduction (rationalism) and a truly Cartesian Linguistics will revive this part of the Cartesian tradition.