Coding of Image Sequences in Macaque V4 Networks

FERNANDEZ LEON, J.A.; HANSEN, B.; DRAGOI, V.

New Orleans

Conferencia; 2012 SfN Neuroscience; 2012

SfN

The serial position effect is the enhancement of memory for early (primacy) and late (recency) items in a sequence and is considered a fundamental and ubiquitous behavioral phenomenon observed in a variety of species, including pigeons, monkeys, and humans. Whether the serial position effect can be observed in the coding of image sequences in visual cortex is currently unknown. We investigated the temporal structure of responses in the presence of a stimulus sequence and in its absence. We recorded extracellular neuronal activity from multiple single-units in visual area V4. Each trial began with a brief fixation (300 ms), followed by a sequence of seven distinct natural images presented for 200 ms each. Each successful stimulus trial was followed by a blank trial, equal in length to the stimulus trial. We computed correlations between stimulus-evoked and blank neuronal activity for the population of cells (n=91) to assess the degree of serial recall and reactivation between the stimulus and blank trials. Correlation analysis yielded an overall tendency of the cells to display negative or low mean correlations between stimulus and blank trials for images presented in the middle of the sequence, i.e., image positions 3, 4, and 5 (mean ± SEM; 0.07±0.06). However, surprisingly, we noticed strong, statistically significant stimulus-blank correlations for the first image in the sequence (0.18±0.02) and last image in the sequence (0.16±0.03). Therefore, V4 neurons exhibit increased, cue-triggered activity in the absence of visual stimulation. To test the robustness of our serial position effect, we shuffled neuronal responses between image positions (200 ms bins) for both the stimulus and blank trials. This led to an overall reduction in mean correlations across all image positions both for the population of cells (0.02±3.63*10-6) as well as for the cells recorded in each session (0.003±0.001). These results indicate that the serial position effect does not depend on the particular images used in our set, but rather on the precise image location in the sequence. Consistent with the serial position effect our results demonstrate for the first time the capacity of V4 networks to recall information, in the absence of the stimulus, after continuous exposure to serially presented natural scenes.