Stimulus Representation and Processing in Human Associative Learning: An evaluation of current elemental and configural associative learning theories
Theories of associative learning describe learning about the relationship between two events, e.g. the eating of an apple and subsequent stomach ache. One important classification of these models is based on the stimulus representation they suppose. Whereas elemental models assume that the represent...
Keine Tags, Fügen Sie den ersten Tag hinzu!
|Zusammenfassung:||Theories of associative learning describe learning about the relationship between two events, e.g. the eating of an apple and subsequent stomach ache. One important classification of these models is based on the stimulus representation they suppose. Whereas elemental models assume that the representations of a stimulus compound consist of representations of its components establishing associations, configural models propose that stimulus compounds are represented and associated as a whole. However, as the empirical results have not consistently favoured one class of models, it was suggested that humans and animals can switch between both modes of stimulus representation. Alternatively, modern theories as the elemental model of Wagner (2003) and Harris (2006) or the extended configural model of Pearce (Kinder & Lachnit, 2003) postulate flexibility within either elemental or configural principles and question the existence of a second, fundamentally different mode of stimulus representation. In order to evaluate these two hypotheses, we contrasted the predictions of the elemental and configural models. Firstly, a user-friendly environment for computer simulation of the models was written. Then, contradicting predictions were tested in two series of human learning experiments. Furthermore, it was investigated whether manipulations of the experimental setting would influence stimulus processing and on which mechanism these effects are based. The results demonstrate that models that comprise configural principles are not made superfluous by modern elemental models. Instead, the extended version of Pearce’s configural model was able to account for all behaviour observed both during learning of two feature-negative discriminations as well as in generalisation tests after learning. Together with the “elemental” results of other studies, this corroborates the position that there are two modes of stimulus processing during associative learning. Further tests and experiments, however, are necessary concerning factors that influence which mode is chosen in a learning situation. Manipulations of the stimulus material (perceptual grouping by motion; spatial arrangement) as well as manipulations of the experimental procedure (learning paradigm; time pressure during stimulus presentation; causal nature of the cover story) had no effects. This indicates that more controlled research is needed investigating not only the factors but also the circumstances controlling these factors’ relevance. Furthermore, additional theoretical analyses revealed that the observed advantage of the configural model is not based on the configural representation itself but on a normalisation of the representation’s activation. Therefore, elemental models incorporating this mechanism could compensate their present shortcoming.|