Acquisition and manipulation of mental structures : investigations on artificial grammar learning and implicit sequence processing

Abstract

This thesis introduces repetitive artificial grammar learning as a paradigm in the investigation of sequential implicit learning, in particular as a model for language acquisition and processing. Implicit learning of sequential structure captures an essential cognitive processing capacity of interest from a larger cognitive neuroscience perspective. We investigate in this thesis the underlying neural processing architecture for implicit learning/acquisition to acquire and process non-motor sequences, an implicit non-motor procedural learning ability present in the human cognitive system. In doing this, we validate and explore the repeated artificial grammar learning paradigm as a laboratory model to investigate the acquisition and processing of structural aspects of language, e.g. (morpho-) syntax processing, to further our understanding of the specific neural processing architecture subserving the syntax processing ability of the language faculty. A theoretical background on sequential procedural learning and formal grammars in cognitive processing is presented together with a general outline of the neuronal implementation of the cognitive functions involved. We suggest a lexical view on the processing and acquisition of artificial grammars to be beneficial to understand the nature and representation of the acquired knowledge. From this perspective we suggest that formal grammar acquisition and processing of the (regular) grammar type commonly studies in artificial grammar learning can be used as a model to investigate the neuronal infrastructure supporting language acquisition and processing, including to characterize the neuronal infrastructure supporting syntax processing and unification (cf. e.g., Hagoort, 2003; Jackendoff, 1997; Jackendoff, 2007; Kaan & Swaab, 2002; Shieber, 1986; Vosse & Kempen, 2000).

In study 1 we describe the neuronal implementation using a setup based on the seminal study on implicit learning by Reber (1967), and report an overlap in the neural activation on artificial syntax violation and similar natural syntax violation. In study 2 we replicate this finding using a more elaborated model with repeated acquisition sessions to simulate a prolonged acquisition period, and using a sequential presentation forcing the cognitive processing into a sequential processing mode. A neuronal activation pattern is reported which suggests that frontostriatal circuits are at play during artificial grammar classification, specifically the left inferior frontal region Broddmann s area 44/45 and the head of the caudate nucleus. In study 3 we repeate the behaviour performance, introducing a preference classification instruction to further the cognitive system into an implicit learning mode, and report a clear and increasing preference for grammatical structure over repeated sessions. In study 4 we investigated the basal ganglia component in Huntington patients with specific caudate head lesions. While the patients did not show any deficit in their behaviour performance, structures in the basal ganglia including the caudate head showed abnormal activation patterns compared to their matched normal controls. Also, a cooperative activation between basal ganglia and hippocampus typically involved in declarative memory was found. We interpret this to reflect attempts of the cognitive system to compensate the damaged procedural processing with declarative knowledge processing.

In summary, in the studies of this thesis we have gained an initial characterization of the neural infrastructure subserving artificial grammar processing. We have done so by characterising the end-state of the learning process as well as characterizing the learning curves reflecting the outcome of acquisition at different time points. This thesis reports findings supporting the view that the extended artificial grammar learning model is useful to capture structural aspects in language acquisition processing in the laboratory.