Linking Perception, Cognition, and Action: Psychophysical Observations and Neural Network Modelling

• Published in: PloS one Vol. 9; no. 7; p. e102553
• Main Authors: Méndez, Juan Carlos, Pérez, Oswaldo, Prado, Luis, Merchant, Hugo
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 16.07.2014; Public Library of Science (PLoS)
• Subjects: Analysis; Analysis of Variance; Artificial neural networks; Biology and Life Sciences; Brain; Classification; Cognition; Cognition - physiology; Decision making; Decision Making - physiology; Delay; Human subjects; Humans; Linear Models; Memory; Models, Neurological; Motion detection; Neural networks; Neurophysiology; Perception; Perception - physiology; Psychomotor Performance - physiology; Psychophysics; Reaction Time; Short term memory
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0102553

It has been argued that perception, decision making, and movement planning are in reality tightly interwoven brain processes. However, how they are implemented in neural circuits is still a matter of debate. We tested human subjects in a temporal categorization task in which intervals had to be categorized as short or long. Subjects communicated their decision by moving a cursor into one of two possible targets, which appeared separated by different angles from trial to trial. Even though there was a 1 second-long delay between interval presentation and decision communication, categorization difficulty affected subjects' performance, reaction (RT) and movement time (MT). In addition, reaction and movement times were also influenced by the distance between the targets. This implies that not only perceptual, but also movement-related considerations were incorporated into the decision process. Therefore, we searched for a model that could use categorization difficulty and target separation to describe subjects' performance, RT, and MT. We developed a network consisting of two mutually inhibiting neural populations, each tuned to one of the possible categories and composed of an accumulation and a memory node. This network sequentially acquired interval information, maintained it in working memory and was then attracted to one of two possible states, corresponding to a categorical decision. It faithfully replicated subjects' RT and MT as a function of categorization difficulty and target distance; it also replicated performance as a function of categorization difficulty. Furthermore, this model was used to make new predictions about the effect of untested durations, target distances and delay durations. To our knowledge, this is the first biologically plausible model that has been proposed to account for decision making and communication by integrating both sensory and motor planning information.