| Abstract No.: | B-D2145 |
| Country: | Canada |
| | |
| Title: | POPULATION CODE OF MT NEURONS DURING A MOTION DETECTION TASK |
| | |
| Authors/Affiliations: | 1 Jackson Smith*; 1 Changan Zhan; 1 Martin Aguilar; 1 Nick Masse; 1 Erik Cook;
1 McGill University, Montreal, QC, Canada
|
| | |
| Content: | Objectives: In this study we examined how signals from multiple neurons in visual cortex are combined during a perceptual task. Although many studies have shown that the activity of single neurons in visual cortex are correlated with our perceptual experiences, we know very little about the way populations of neurons work together to form representations.
Methods: We recorded the activity of 50 pairs of neurons from the Middle Temporal (MT) area of visual cortex from two monkeys trained to detect the onset of a coherent motion pulse. The stimulus was a pair of non-overlapping random dot patches and neurons were recorded using two microelectrodes inserted into cortex 1-2 mm apart. Trials began by presenting the animals with 0% coherent motion for a random length of time. A 50 ms pulse of coherent motion occurred in either patch alone or in both patches together and the animals were required to release a lever shortly thereafter to receive a liquid reward. The location and size as well as motion velocity and direction of each patch was matched to that preferred by each neuron in order to generate maximal neural responses.
Results: The monkeys’ performance was best when both patches contained a coherent motion pulse. To analyze neural activity, spike trains were first smoothed with a Gaussian kernel. We found that neural activity occurring after onset of the coherent motion pulse was correlated with the animals’ behavioral performance and reaction time. When the activity of our paired neurons was added together, we found that the correlation between neural activity and behavior substantially increased. Importantly, when the activity of our recorded pairs was multiplied, the correlation between neural activity and behavior was less than that of the individual neurons. This relationship between neural activity and behavior was maintained for all sizes of our smoothing Gaussian kernel.
Conclusions: Behavior was best predicted from the summed neural activity. These results suggest the animals used a population code that did not rely on the temporal synchrony of action potentials in MT to perform the motion detection task. |
| | |
| Back |
|
