Åse Helene Kvist Innes-Ker
Postdoc in the Department of Psychology
University of Notre Dame
Abstract:
Emotional expressions emerge from combinations of expressive features that themselves rarely convey
an emotion. In addition, the valence of an expression can be conveyed at below-threshold exposures.
Both finding suggest that expressions are processed as a gestalt, although little previous work has
been done directly testing this conjecture. The present investigation into gestalt processing of
emotional expressions relies on a definition of gestalt developed within a stochastic process model
framework. Stochastic process models are considered meta-theoretical models, developed to infer
underlying cognitive processes from reaction-time data. Processing varies on four logically
distinct, but interacting dimensions: architecture (serial, parallel or coactive), independence
(preservation or violation), capacity (limited, unlimited or super capacity) and stopping-rule
(self-terminating or exhaustive). Gestalt processing suggests either a parallel or co-active
architecture, positive dependence between channels, super capacity and an exhaustive stopping-rule.
Participants made decisions about expressive features in rapidly presented arrays in a search
paradigm adapted from Wenger and Townsend. The features were the eyes, nose and mouth isolated from
two separate photographs of the same individual expressing different emotions. One expression
arbitrarily served as target. The arrays consisted of 2 to 4 features that were either placed in a
biologically correct or scrambled configuration. The features were either all from one or the other
source photograph, or a mixture from both sources Responses were made under either an AND
instruction or an OR instruction. Between subjects conditions compared happy and neutral; fear and
neutral or happy and fear expression. Both gestalt and scrambled stimuli revealed an underlying
parallel architecture for both gestalt and scrambled stimuli, although Gestalt stimuli were responded
to faster than scrambled. In direct comparison between gestalt and scrambled stimuli containing
identical features, gestalt overall afforded more efficient (super-capacity) processing.
Participants employed a self-terminating stopping rule whenever possible, suggesting an analytic and
strategic approach towards the task. When exhaustive stopping rule was apparent, gestalt configured
stimuli were processed with super-capacity while scrambled showed evidence for limited capacity.
Minimum time processing uniformly resulted in limited to extremely limited capacity.