Article Figures & Data
Figures
- Figure 1.
Behavioral task and performance. A, During each recording session the subject encountered three unfamiliar monkeys and objects (a, b, and c). Each monkey and object was associated with a different amount of juice reward. The three items in each category were paired in the three combinations (ab, ac, and bc). Maximizing the reward was the incentive to discriminate between both monkeys and objects. B, The periods of looking at each of the two videos displayed on the monitor was determined based on the viewer monkey’s eye movement. In this case, monkey a (zero drops) and monkey c (eight drops) are viewed in sequential looks, defined as a succession of fixations and saccades on the same video or same general gaze target (e.g., face or body). C, D, Correct performance was defined as selecting the higher value stimulus on each trial. Both subject animals performed above chance. E, Choice accuracy was highest on trials contrasting stimuli associated with zero to eight drops of juice and the lowest for contrasting three to zero drops of juice. F, Monkeys spent more time fixating on the higher valued stimuli.
- Figure 2.
Raster plot and perievent time histogram for a neuron that responded to the start cue and also the onset and offset of the video, indicated by the vertical red lines. The black arrow indicates the increased firing rate elicited by the start cue that preceded the video onset. Due to a variable (±100 ms) delay between the start cue and video onset, the fast rise in firing rate elicited by the start cue is not as clearly aligned in time as the onset and offset of the video.
- Figure 3.
Raster plots and perievent time histogram for neurons tuned to different stimulus features. A, Category-selective neuron that responded with higher firing rates to objects compared with monkeys (firing rates before the start of each look are offset because during the previous look the monkey was attending to the same category; on each trial, only objects or monkeys but not both were shown on the monitor). B, Value-selective neuron that increased its firing rate in response to the monkey and the object that were associated with the highest reward. C, Identity-selective neuron that increases its firing rate when the subject was looking at a specific monkey. D, Face-selective neuron. Neural activity is aligned to either fixations on the face or eyes (shown in pink), or fixations on the body (shown in purple).
- Figure 4.
Example of a multidimensional neuron, tuned to multiple task and stimulus features. A, Raster plots and perievent time histograms aligned to video onset and grouped by stimulus pairs (monkey pairs shown in shades of blue, object pairs shown in shades of red). Note that this neuron increased its firing rate to the presentation of monkey stimuli, demonstrating both a video onset response and category selectivity. B, The same neuron also differentiated between the three monkeys. Here the neural activity is aligned to the onset of each look. C, The same neuron was selective for fixations on faces, as shown by the neural activity aligned to the onset of fixations on the face or eyes of monkeys (shown in pink) compared with fixations on the body (shown in purple).
- Figure 5.
Distribution of selectivity combinations across the population of recorded neurons. A, The nonproportional Venn diagram illustrates the four types of selectivity as four ellipses, where the intersectional areas of each ellipse correspond to neurons that show selectivity for the overlapping features. Neurons that are responsive to task events (start cue presentation, video onset or end, and presentation of the choice cue) populate the “a” ellipse. Neurons that were category selective or value selective populate the “b” ellipse. Identity-selective neurons populate the “c” ellipse. Finally, the “d” ellipse contains the neurons that were face/eye selective. The number of cells and the percentage of each combination of selectivity in the total population are marked inside each intersectional area. B, A comparison between the theoretical and observed number of neurons in each type of selectivity (area shaded in each color as in A). The theoretical distribution for each selectivity type was generated from a binomial probability distribution centered around the conditional probability calculated by assuming that the probability of a cell to respond selectively for any given level was independent of the selectivity for any other levels. This distribution is shown in the shaded areas (violin plots). The observed number of neurons in each combination of selectivity is indicated by the blue diamonds. A red star indicates when the observed number was outside 99.9% (corrected for two-tailed test and multiple comparisons) of the theoretical distribution.
In this issue
