Feature Integration During Collective Decision-Making in Fish Shoals

An essential part of decision-making is the comparison between available options that can differ from each other in different features (e.g., color, shape, size, etc.). How these features are integrated during the comparison process can strongly affect the decision performance. In human beings for instance, objects that differ in more than one feature are compared by a serial search at the cortical level in which performance decreases quickly with the number of objects to compare. Human and non-human animal groups however are known to make collective decisions based on distributed self-organized mechanisms. In a search task, these groups should therefore be less sensitive to the number of objects to compare because of their potential ability to process in parallel the objects and their different features. To test this hypothesis, we created a modified version adapted to group behavior of the Treisman search test (1), originally designed to identify cognitive search modes in individual human beings.

Shoals of 10 golden shiners (Notemigonus crysoleucas) were presented with semi-transparent plastic shelters under which they tend to aggregate and that they could perceive from their starting area. The shelters could differ from each other in their size (small versus large) or their darkness level (light versus dark), dark and large shelters being preferred by the fish and considered as a target. In a first test, the target was presented against 2 or 4 identical distractors of either a small size or a low darkness level. This test evaluates the group performance in a single feature search. In a second test, the target was presented against 2 or 4 different distractors of a small size and/or a low darkness level. This test evaluates the group performance in a multiple feature search. As a control, all these tests were replicated in the absence of the target.

Following our hypothesis (distributed feature comparison), the performance of the group in the single and multiple feature search tests should be similar, and in both cases little affected by the number of distractors. Any deviation from this prediction would suggest the presence of at least one serial comparison step that can be identified with a detailed analysis of individual fish behaviors. The results of this experiment will broaden our understanding of collective information processing and decision-making in human and non-human animal groups.

(1) Treisman, A.M. Focused attention in the perception and retrieval of multidimensional stimuli. Perception & Psychophysics 22, 1-11(1977).