UCLA

Animals are confronted with abiotic cues, conspecific signals, and signals eavesdropped from heterospecifics that create some degree of uncertainty as to the state of their world. They must reduce uncertainty to make optimal decisions. Animals can improve the accuracy with which they make decisions by combining stimuli from different modalities, a phenomenon I term “multimodal integration”. However, in some situations an animal does not benefit from this increase in accuracy and the animal does better to ignore certain stimuli. Despite a body of literature that documents such situation-dependent integration, to date, a quantitative approach to understand the conditions in which integration is favored is lacking. My dissertation develops a framework for thinking about the functional significance of integrating stimuli in multiple modalities. First, I conceptually bring together the ideas of uncertainty, costs of mistakes and prior expectation of the state of the world in order to explain why more information sources are not always better. I then present a quantitative model that parameterizes uncertainty, the costs of mistakes, prior expectations and the costs of attending to stimuli in predicting whether or not an animal should integrate sequential stimuli in different sensory modalities. The model applies to multimodal stimuli in that different levels of uncertainty can be specified for each stimulus. This feature importantly captures the property of sensory modalities to be independently affected by uncertainty. For example, the sound of noisy traffic will not likely affect your ability to see a friend walking towards you. Finally, using yellow-bellied marmots (Marmota flaviventris), I field test the extent to which three of the model’s parameters affect integration of olfactory-acoustic predator stimuli. I found that the benefit of alert behavior and uncertainty of the second (acoustic) stimulus does indeed affect integration. Overall, this dissertation establishes a foundation for a new line of inquiry into situation-dependent integration, which will help us understand the evolution of cognitive systems, communication networks, animal signals and the ways in which individuals interact with the abiotic world.