Domain-specificity in the acquisition of non-adjacent dependencies

At the forefront of investigations into the cognitive underpinnings of language acquisition is the question of domain-specificity, i.e. whether the processes involved in learning language are unique to language. Recent investigations suggest that the mechanisms employed in language learning are also involved in sequential learning of non-linguistic stimuli and are therefore domain-general. Non-adjacent dependencies are an important feature of natural languages. They describe relationships between two elements separated by an arbitrary number of intervening items, and thus potentially pose a challenge for learners. As a hallmark of natural languages they are ubiquitous, an example from English being subject-verb agreement: The socks on the floor are red. Here, learners are required to track the dependencies amongst the two underlined elements across an intervening prepositional phrase. Importantly, it has been shown that non-adjacent dependencies can be learned in the linguistic (Gomez, 2002) and non-linguistic (Creel, Newport & Aslin, 2004) domain. The majority of work presented in this thesis is based on Gomez’s (2002) artificial language learning experiment involving non-adjacent dependencies, adapted to directly compare adults’ learning in the linguistic and non-linguistic domain, in order to build a comprehensive map showing factors and conditions that enhance/ inhibit the learnability of non-adjacencies. Experiment 1 shows that the Gestalt Principle of Similarity is not a requirement for the detection of non-adjacent dependencies in the linguistic domain. Experiment 2 aims to explore the robustness of the ability to track non-adjacent regularities between linguistic elements by removing cues that indicate the correct level of analysis (i.e. interword breaks). Experiments 3 and 4 study domain-specificity in the acquisition of non-adjacencies, and show that non-adjacent dependencies are learnable in the linguistic and nonlinguistic domain, provided that the non-linguistic materials are simple and lacking internal structure. However, language is rich in internal structure: it is combinatorial on the phonemic/ orthographic level in that it recombines elements (phonemes/graphemes) to form larger units. When exposed to non-linguistic stimuli which capture this componential character of language, adult participants fail to detect the non-adjacencies. However, when exposed to non-componential non-linguistic materials, adult participants succeed in learning the non-adjacent dependencies. Experiment 5 looks at modality effects in the acquisition of non-adjacent dependencies across the linguistic and non-linguistic domain. Experiment 6 provides evidence that high familiarity with componential non-linguistic patterns does not result in the correct extraction of non-adjacencies in sequence learning tasks involving these patterns. Overall, the work presented here demonstrates that the acquisition of nonadjacent dependencies is a domain-general ability, which is guided by stimulus simplicity.