Search

The research presented in this PhD dissertation examines the phenomenon of semantic implicit learning, using semantic preferences of novel verbs as a test case. Implicit learning refers to the phenomenon of learning without intending to learn or awareness that one is learning at all. Semantic preference (or selectional preference – as preferred in computational linguistics) is the tendency of a word to co-occur with words sharing similar semantic features. For example, ‘drink’ is typically followed by nouns denoting LIQUID, and the verb ‘chase’ is typically followed by ANIMATE nouns. The material presented here spans across disciplines. It examines a well-documented psychological phenomenon - implicit learning – and applies it in the context of language acquisition, thereby providing insights into both fields. The organisation of this dissertation groups its experiments by their methodology. Chapter 1 provides an overview of the current psychological and linguistic literature. Chapter 2 includes a pen-and-paper study carried out in a classroom environment on Polish learners of English, where awareness is assessed by subjective measures taken at each test question as well as a post-experiment questionnaire. Chapter 3 includes a collection of 5 computer-based experiments based on a false-memory paradigm. After exposure to sentential contexts containing novel verbs, participants are shown to endorse more previously unseen verb-noun pairings that follow the correct semantic preference patterns to the pairings that violate it. The result holds even when participants do not reveal any explicit knowledge of the patterns in the final debriefing. Awareness is additionally assessed using indirect measures examining correlations of confidence judgements with performance. Chapter 4 examines whether implicit learning of novel verb semantic preference patterns is automatic. To this end, a reaction time procedure is developed based on two consecutive decisions (“double decision priming”). The method reveals that semantic implicit learning, at least in the described cases, exerts its influence with a delay, in post-processing. Chapter 5 comprises research done in collaboration with Dr Nitin Williams, University of Reading. It documents an attempt at finding neural indices of implicit learning using a novel single-trial analysis of an electroencephalographic (EEG) signal, based on empirical mode decomposition (EMD) denoising. Chapter 6 presents a final discussion and indications for future research. The main contribution of this dissertation to the general field of implicit learning research consists in its challenging the predominant view that implicit learning mainly relies on similarity of forms presented in training and test. The experiments presented here require participants to make generalisations at a higher, semantic level, which is largely independent of perceptual form. The contribution of this work to the field of Second Language Acquisition consists of empirical support for the currently popular but seldom tested assumptions held by advocates of communicative approaches to language teaching, namely that certain aspects of linguistic knowledge can develop without explicit instruction and explanation. At the same time, it challenges any view assuming that vocabulary learning necessarily relies on explicit mediation. The experiments collected here demonstrate that at least word usage in context can be learnt implicitly. A further contribution of this dissertation is its demonstration that the native language may play a key role in determining what is learnt in such situations. A deeper understanding of the phenomenon of semantic implicit learning promises to shed light on the nature of word and grammar learning in general, which is crucial for an account of the processes involved in the development of a second language mental lexicon.