Project summary
Stimulus generalization refers to the situations where animals respond similarly to a broad set of novel stimuli as the familiar ones, which has now been widely accepted as a key mechanism driving clinical problems such as anxiety disorder and stress disorder (Dymond et al., 2015). Most of the existing generalization models and theories, such as gradient-interaction theory ( Spence, 1936, 1937 ; Hull, 1949), overlap theory (Ghirlanda and Enquist, 1999), the model by Blough (1975) and Shepard (1987), have a focus on identifying and predicting the pattern of generalized behaviour without further exploring its underlying processes. Such models, with a pursuit of an universal mechanism of generalization, lack ability to understand individual differences in generalization gradients that have been emphasized on several recent studies (see Zaman et al., 2020 as a review).

The fundamental aims of the current project is to conduct computational modelling approaches to investigate and disentangle different mechanisms of stimulus generalization. The project will adapt the present similarity-based generalization models from two perspectives. Firstly, we postulate that dynamic stimuli similarity modulated by perceptual variations can better account for large variations of responding to the same stimulus compared with the assumption of stimuli similarity being constant. Secondly, we argue that generalized responding should be modelled in the absolute rather than relative manner because learning and memory problems would influence associative strength on CS (determine the maximum responding of generalization) and further modulate the steepness of generalization gradients. Our models will treat perception, learning, memory, and generalization tendency as independent mechanisms contributing to generalized responding, and apply different parameters to infer specific causes of overgeneralization.

Supervisors
Jonas Zaman
Francis Tuerlinckx
Wolf Vanpaemel

Financed by

Period
November 2020 – November 2024