Paola Suárez: Disentangling sensory and top-down information during perceptual decision making

BCCN Berlin and TU Berlin

 

Abstract

In the field of perceptual decision making, the relationship between the activity of sensory neurons and decision outcomes is known as choice probability (CP) and has been extensively studied. A problem regarding CP is that the origin of the variability in neuronal activity and decision outcomes is unclear. Hence, it is difficult to assess whether sensory neurons have a causal effect on upcoming decisions or not. A purely bottom-up approach supports a causal interpretation, in which variability in choices arises from random fluctuations in the response of sensory neurons to external stimuli. On the other hand, a top-down approach supports a non-causal interpretation, and states that variability in choices reflects fluctuations in the influence of cognitive factors (e.g., attentional signals or intrinsic biases) on the activity of sensory neurons. Recently, sustained traces of CP were disentangled into an early, bottom-up component and a late, top-down component through sophisticated computational modeling. However, using an indirect measurement of bottom-up and top-down influences on choice, results in an expensive way to study decision net- works under different, more complex conditions. In the present work I will extend a network model for perceptual decision making with two-compartment sensory neurons to create a diverse repertoire of spiking patterns. The theory of multiplexing suggests that sensory neurons use different spiking patterns to transmit information from different streams. Thus, this extension provides a direct method to differentiate between bottom-up and top-down contributions to an upcoming decision. Overall, the present work suggests a biologically plausible, simple and generalizable method to tackle the interpretation problem regarding CP.

Additional Information: Master thesis defense

Organized by

Prof. Henning Sprekeler