Proceedings of The Physiological Society

University of Manchester (2010) Proc Physiol Soc 19, PC230

Poster Communications

Information supply and neural decision time

B. Pearson1, R. H. Carpenter1

1. Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom.

  • Figure 1. Cumulative distributions of observed and simulated correct responses (probit scale) as a function of reciprocal latency for two subjects, for the indicated number of lines of each orientation.

Behavioural response times reflect the neural mechanisms underlying decision, and are often modelled neurophysiologically by a decision variable S representing the activity of a population of neurons. In response to a stimulus, S rises at a rate that depends on the information being supplied, until it crosses a threshold that initiates the response: the key features of this LATER (Linear Approach to Threshold at Ergodic Rate) model1 are seen in single neuron activity in monkey cortex2. In humans, the dependence of latency on information has conventionally been investigated with random dot kinetograms (RDK), in which a background of dots moving randomly is overlaid with a proportion of dots moving coherently in one direction3. Cumulative distributions of saccadic latency for subjects trained to saccade in the direction of the apparent motion demonstrate the parallel shift characteristic of a change in the rate of rise of S. In monkeys performing such a task, direction-selective MT neurons encode local motion energy4; this information appears then to be pooled over a wider area, by cells in the lateral intraparietal area. As coherence increases, the rate of rise of their activity also increases, peaking sooner and producing saccades of shorter latency5. Although such results link behavioural responses quantitatively to an underlying neural mechanism, the details may be specific to the parietal visual motion pathway. Is there an equivalent mechanism for distributed but static visual stimuli? Instead of a field of dots we used a centrally fixated 12.5 deg diameter annulus containing twelve lines (each subtending 0.5 deg). In each trial, they were initially vertical but after a random foreperiod (0.5-1.5 secs) became obliquely-oriented at ±45 deg; the proportions tilted clockwise or anticlockwise provided a static equivalent of motion energy, that would be expected to excite different numbers of cortical line detectors tuned to each orientation. With local ethical committee approval, we recorded latencies for five subjects told to manually rotate a switch in the prevailing direction. As for RDK stimuli, median latency increased as coherence was reduced, and subjects made more errors. We used Monte Carlo simulations - in which two LATER units, one for each response direction, competed to initiate the manual response - to model the observed distributions of latency, a more stringent requirement than simply predicting means. For all subjects, the simulated distributions were an excellent fit to the observed distributions (Kolmogorov-Smirnov p > 0.15), though - unlike what was previously reported for human RDK experiments - it was necessary to alter the SD of the rate of rise, as well as the mean rate, as a function of the afferent information supply (Figure 1).

Where applicable, experiments conform with Society ethical requirements