Contextual influences on real-time processing of linguistic stimuli: Traditional and dynamical approaches

This work examines contextual influences on processing of linguistic stimuli. The traditional (symbolic) models of context influences are reviewed and their shortcomings pointed out. The dynamical approach, recently emerging in the area of behavioral sciences, is suggested as a viable alternative. Two studies follow. In the first one we use the case of perception of ambiguous sentences to show that perception of linguistic stimuli is the outcome of an underlying dynamical process. Thus it may be better described in dynamical terms, employing notions such as multistability and differential coherence of patterns, than in the traditional, symbolic framework. The second study is an on-line investigation of contextual adaptation. We studied general category names embedded in neutral or biasing sentential contexts. The results obtained indicate that the initial lexical access is context independent. The relative availability of particular members of category suggests that the initial state is best captured by a multistable representation, which may be essential for the flexibility of linguistic processing. Contextual adaptation seems to occur later in the unfolding sentence. A more detailed investigation into the timing and nature of contextual adaptation suggests that this adaptation takes the form of rapid reorganization of conceptual information rather than just facilitation of relevant category members. The results of the studies presented have implications both for dynamical and psycholinguistic approaches. The main implication for the dynamical approach is the importance of using on-line methods in studies of perception. Dynamical studies that use off-line methods perhaps miss an important stage of processing: a transition from locally invariant to contextually congruent organization of information. For psycholinguistics the characterization of language processing as pattern formation has at least three major advantages: (1) capturing the timing of the processes allows for including distinctions between fast/slow, linear/nonlinear processing, (2) conceptualization of the initially available lexical information as a constraint on possible meanings rather than meaning itself allows for accounting for apparently contradictory psycholinguistic data, (3) adding the dimension of stability of the patterns generated during language processing makes possible new predictions regarding speed and variability of performance on various psycholinguistic tasks.