Technical Program

Paper Detail

Paper: PS-2B.28
Session: Poster Session 2B
Location: H Fl├Ąche 1.OG
Session Time: Sunday, September 15, 17:15 - 20:15
Presentation Time:Sunday, September 15, 17:15 - 20:15
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Unifying Neural Delay Representations in Cognitive Tasks: A Joint Human Behavioral and Recurrent Neural Network Study
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
DOI: https://doi.org/10.32470/CCN.2019.1343-0
Authors: Daniel Ehrlich, John D. Murray, Yale University, United States
Abstract: In this study we define contingency representations, a representational schema for delay tasks in which neural states encode prospective choice points, and demonstrate how such a representation unifies seemingly contradicting sensory-, action- and rule-based representations reported for prefrontal cortex neurons in different delay tasks. Further, we describe a novel experimental paradigm, the conditional delayed logic (CDL) task, in which we investigate competing theories of representational structures as they are utilized to perform varied working memory tasks. We trained a recurrent neural network to perform the CDL task, identifying a contingency representation subspace and testing its functional and mechanistic properties. Human subjects tested on the CDL task demonstrated behavior consistent with the contingency-based representational schema and inconsistent with many leading models of working memory. Contingency representations, in addition to clarifying neuronal delay tuning, provide a novel hypothesis for mixed selectivity as well as dynamic tuning observed during many working memory tasks. Lastly, we present a set of falsifiable predictions and analyses for neural data sufficient to differentiate contingency representations from alternative representational theories.