Syntactic classification of acquired structural regularities

摘要: Syntactic Classification of Acquired Structural Regularities Christian Forkstam (christian.forkstam@cns.ki.se) Cognitive Neurophysiology Research Group, Karolinska Institutet hospital N8, 171 76 Stockholm, Sweden. F.C. Donders Centre for Neuroimaging Radboud University Nijmegen, The Netherlands. Karl Magnus Petersson (karl.magnus.petersson@fcdonders.ru.nl) CSI, Center intelligent systems, Universidade do Algarve, Faro, Portugal. from artificial grammars (e.g., Seger, Prabhakaran, Poldrack, & Gabrieli, 2000; Skosnik et al., 2002). For example, al. (2004) investigated a grammaticality classification task using an implicit acquisition paradigm without feedback in which the participants were only exposed to positive examples (i.e., well-formed consonant strings) generated by Reber grammar. results showed that syntactic violations activated Broca’s region (Brodmann’s area (BA) 44/45). In current study we tested validity this finding modified experimental design, strings balanced substring familiarity relative string-set, independent grammatical status; and sequential instead whole string presentation strings. Abstract paper investigate neural correlates acquired sequence structure event-related FMRI study. During acquisition, engaged short-term memory performance feedback. We manipulated statistical frequency-based rule-based characteristics stimuli independently order their role grammar acquisition. performed reliably above chance on task. observed partly overlapping corticostriatal processing network both manipulations including inferior prefrontal, cingulate, parietal regions, caudate nucleus. More specifically, left frontal BA 45 nucleus sensitive endorsement, respectively. contrast, these structures insensitive manipulation. Implicit learning Keywords: Artificial Grammar; Functional Neuroimaging; FMRI; Inferior Frontal Cortex; Caudate Nucleus. A complementary perspective AGL views as model investigating (Forkstam Petersson, 2005). (1967) defined process individual comes respond appropriately inherent input. Thus, he argued, capacity generalization show is based structural regularities reflected input sample. suggested humans acquire knowledge underlying through inductive put use during classification. Support character example lesion studies amnesic patients. Knowlton Squire (1996) patients normal controls classical transfer version similarly tasks while no explicit recollection whole-item or fragment bi- tri-gram) information. Based they argued depends abstract exemplar-specific latter indicates distributional information local acquired, former suggests Introduction Humans possess adaptive mechanisms capable implicitly extracting solely observation (Stadler Frensch, 1998), indicated (AGL). can learn abstraction intrinsic natural language Chomsky, following von Humboldt, ‘infinite finite means’. simplest relevant formal incorporating idea represented family right-linear phrase grammars, be implemented finite-state architecture (FSA), are typically used AGL. It has recently been aspects infants (Gomez Gerken, 2000), second adults (Friederici, Steinhauer, Pfeifer, Recent functional magnetic resonance imaging (FMRI) indicate related brain regions (Petersson, Forkstam, Ingvar, 2004) number have material