Reward association affects neuronal responses to visual stimuli in macaque te and perirhinal cortices

Abstract

To study the roles of the perirhinal cortex (PRh) and temporal cortex (area TE) in stimulus-reward associations, we recorded spike activities of cells from PRh and TE in two monkeys performing a visually cued go/no-go task. Each visual cue indicated the required motor action as well as the availability of reward after correct completion of the trial. Eighty object images were divided into four groups, each of which was assigned to one of four motor-reward conditions. The monkeys either had to release a lever (go response) or keep pressing it (no-go response), depending on the cue. Each of the go and no-go trials could be either a rewarded or unrewarded trial. A liquid reward was provided after correct responses in rewarded trials, whereas correct responses were acknowledged only by audiovisual feedback in unrewarded trials. Several measures of the monkeys' behavior indicated that the monkeys correctly anticipated the reward availability in each trial. The dependence of neuronal activity on the reward condition was examined by comparing mean discharges to each of the 40 rewarded stimuli with those to each of the 40 unrewarded stimuli. Many cells in both areas showed significant reward dependence in their responses to the visual cues, and this was not likely attributable to differences in behavior across conditions because the variations in neuronal activity were not correlated with trial-by-trial variations in latency of go responses or anticipatory sucking strength. These results suggest the involvement of PRh and TE in associating visual stimuli with reward outcomes.

Figure 1.

Behavioral task. A, Eighty images of objects were used as visual cues. They were divided into four groups, each representing a different motor and reward condition. The visual cue indicated the required motor action (go or no-go response) and also the availability of liquid reward (R+, reward; R−, no reward) at the end of trial after a correct motor response. B, Time sequence of events in a trial. The monkey's lever press started a trial. The go response was to release the lever and press it again within a 600 ms window starting 200 ms after the visual cue onset. The no-go response was to hold the lever down. The monkey had to keep pressing the lever and fixating the eyes on the fixation point until the time of reward delivery, which was 1300 ms after the visual cue onset.

Figure 2.

Recording positions shown in lateral and ventral views (top two lines) and frontal sections (bottom four lines) of the brain. The vertical lines in the lateral and ventral views indicate the posterior and anterior limits of recordings. The gray regions in the frontal sections indicate areas from which the recordings were made. The medial regions were within the perirhinal cortex, and the lateral regions were in area TEad. The recordings were conducted in the right hemisphere in both monkeys. sts, Superior temporal sulcus; amts, anterior middle temporal sulcus; rs, rhinal sulcus. The numbers to the left of the drawings indicate the distance (in millimeters) of the section from the ear bar position.

Figure 3.

Responses of a PRh cell. A–D, Individual mean responses (black) to 20 rewarded go stimuli (A), 20 rewarded no-go stimuli (B), 20 unrewarded go stimuli (C), and 20 unrewarded no-go stimuli (D), together with the mean (red) and the SD (blue) within each motor–reward condition. The number of trials included in producing the plots of each type were 116, 111, 102, and 111, respectively. E, Root of the variance (blue), which represents the magnitude of stimulus selectivity, and (black), which was used to normalize r(t) and m(t) for t test. S = 20. F, The averaged responses in the rewarded conditions x_{r = +} (red) and unrewarded conditions x_{r = −} (green). r(t) is the difference between the two lines. G, The averaged responses in the go conditions x_{m = +} (red) and no-go conditions x_{m = −} (cyan). m(t) is the difference between the two lines.

Figure 4.

Two examples of cells that showed significant dependence on the reward condition. A, C, The red and green lines indicate averaged responses in the rewarded and unrewarded conditions, respectively. The blue line indicates the magnitude of the stimulus selectivity. The cyan line at the bottom indicates the histograms of the times of start (upward) and end (downward) of the eye fixation. The red and green lines at the bottom indicate the histograms of the times of bar release (downward) and re-press (upward) in rewarded (red) and unrewarded (green) trials. The period of visual cue presentation is indicated by the light blue shading, and the time of reward delivery is indicated by the red vertical lines. B, D, Rastergrams of spike occurrence in individual trials of unrewarded (green) and rewarded (red) conditions for the cell shown in A and C, respectively. A vertical short line segment indicates a spike, and a horizontal line represents a trial. The trials shown here were consecutive trials in each condition with different stimuli. Cell shown in A and B was recorded from PRh, and that in C and D was from TE. The cell in A and B is the same cell as that in Figure 3.

Figure 5.

Comparison of the magnitude of reward dependence with that of stimulus selectivity. A, The ratio of the variance attributable to reward conditions (r(t))²/2 to the sum of the variance attributable to stimulus selectivity and that attributable to reward dependence (r(t))²/2 + v(t) are displayed as a cumulative frequency histogram. The number of cells was cumulated and normalized by the total number of cells in each area. The half of the cells with smaller ratios in each area is not shown, because the distribution among these cells was very close to the theoretical distribution of pseudo dependence expected from visual stimulus selectivity in both areas. The variances of responses shown here were calculated for the mean firing rate in a window 250–350 ms from the stimulus onset. The broken line indicates the theoretical distribution of false reward dependence originating in visual stimulus selectivity. Approximately of PRh cells showed reward dependence comparable with stimulus selectivity (that is, r > 0.7v). TE cells showed smaller reward dependence than PRh cells. B, An example of a PRh cell with strong dependence on the reward condition. The cell is indicated as “cell 1” in A. The magnitudes of responses to rewarded stimuli (red line) essentially segregated from those of responses to unrewarded stimuli (blue line). C, An example of a PRh cell with nonsignificant reward dependence. The cell is indicated as “cell 2” in B.

Figure 6.

Comparison of the visual stimulus selectivity between rewarded and unrewarded stimuli. The magnitude of the stimulus selectivity in rewarded trials was defined by the root of mean variance of firing rates in go rewarded and no-go rewarded conditions, which can be written as . The magnitude of the stimulus selectivity in unrewarded trials was calculated similarly but from firing rates in the go unrewarded and no-go unrewarded conditions, that is . The mean firing rates in the window from 100 to 500 ms from the stimulus onset were used. The magenta dots represents PRh cells, and the cyan dots represent TE cells. The stimulus selectivity was comparable between the rewarded and unrewarded conditions in both areas.

Figure 7.

Time courses of reward dependence and visual stimulus selectivity. A, B, The blue lines indicate the time course of averaged normalized stimulus selectivity u_S/u_E in PRh cells (A) and in TE cells (B). All of the cells (201 PRh cells and 139 TE cells) were included. The unbiased estimate of the variance attributable to stimulus selectivity was normalized by the trial variance and then averaged across cells in each area. The red and cyan lines indicate the time course of averaged normalized reward dependence u_R/(u_S + u_E). The unbiased estimate of the variance attributable to reward dependence was normalized by a sum of the unbiased estimate of the stimulus selectivity variance and the trial variance and then averaged across cells in each area. All of the lines in A and B were scaled by the peak value of each line to facilitate the comparison of their time courses. The light blue area indicates the period of visual cue presentation, and the red vertical line indicates the time of reward delivery. The peak value of the reward dependence was 4.8% and that of stimulus selectivity was 73% in A, and they were 2.9 and 114% in B, respectively. The stimulus selectivity preceded the reward dependency in both areas. C, The time courses of reward dependence in PRh cells (cyan) and TE cells (magenta). D, The time courses of stimulus selectivity in PRh cells (purple) and TE cells (blue). The two lines are scaled by the peak value of each line. Both the stimulus selectivity and reward dependence appeared and developed earlier in TE.

Figure 8.

Comparison of the strength of motor dependence with that of reward dependence in PRh (A) and TE (B). The green lines show the cumulative frequency of the cell in terms of the strength of motor dependence, calculated by the ratio of the variance attributable to motor conditions to the total variance but excluding the trial variance (m(t))²/2[(m(t))²/2 + (r(t))²/2 + v(t)]. The red lines show the cumulative frequency of the cell in terms of the strength of reward dependence (r(t))²/2[(m(t))²/2 + (r(t))²/2 + v(t)]. The variances of responses were calculated in a window from 100 to 500 ms after the stimulus onset. The broken lines indicate the theoretical distribution of false dependence originating in visual stimulus selectivity. The motor dependence was smaller than the reward dependence in both areas.

Figure 9.

Two models for translating visual information (white part) to reward information (black part). A, Feedback model. The visual-to-reward conversion (indicated by the white downward arrow) occurs in a higher center, and the reward information is sent along the feedback pathway. According to this model, the onset latency of the reward signal should be longer in TE than in PRh. B, Progressive model. The visual-to-reward conversion occurs progressively along the forward pathway. This model predicts a longer onset latency of reward signal in PRh than in TE, which is what we found in the present study.