Figures and data
Task and recording sites.
(A) Behavioral task. A central cue was first presented to indicate the category of the target to be searched for. This was followed by a search array consisting of eleven stimuli, including two target stimuli and nine distractors. The cue and the two targets belonged to the same category, although the targets were visually distinct from the cue. Monkeys were rewarded for maintaining fixation on either target for at least 800 ms. The white trace indicates eye gazes. In this example, the monkey started from the center, fixated on one face target for less than 800 ms, moved to another face target and fixated for less than 800 ms, then moved to a distractor, and finally shifted back to the face target, fixating for 800 ms to complete the trial. (B) Stimuli. Four categories of visual objects (40 images per category) were used for neural recordings. (C) Probability of fixation transitions as a function of fixation duration. (D) MRI images showing representative recording sites in areas V4, IT, and LPFC. Red arrows indicate the directions of electrode penetrations. (E) The 20 possible stimulus locations used in the search array. Peripheral receptive fields (RFs) were mapped using the visually guided saccade task, which had the same 20 possible stimulus locations as the visual search task. (F-H) Histogram of the number of stimuli that activated the peripheral RF. (I-K) The aggregated tuning regions of the peripheral units. Color bars show the number of units with tuning regions in a given stimulus location. The right side of the brain was recorded for both monkeys. (F, I) V4 units. (G, J) IT units. (H, K) LPFC units.
Summary of saccade probabilities, saccade counts, and saccade amplitudes (mean ± SD).
Foveal feature-based attentional modulation in V4 and IT.
(A) V4 face-selective units. (B) IT face-selective units. (C) V4 house-selective units. (D) IT house-selective units. (E) V4 non-selective units. (F) IT non-selective units. Shown are normalized population firing rates during fixations on face targets, face distractors, house targets, and house distractors, respectively (see Methods). Shading around the mean firing rates indicates ±SEM across units. Vertical lines and numbers indicate population-level latencies. For selective units, firing rates were normalized to the maximum response to the attended stimulus from the preferred category. For non-selective units, fixations on face and house targets were combined into target fixations, and fixations on face and house distractors were combined into distractor fixations. Firing rates were normalized to the maximum response during target fixations.
Response latency of attentional modulation in foveal and peripheral receptive fields.
Shown are cumulative distributions of feature-attention effect latencies, computed from individual foveal face-, house-, and non-selective units in V4 and IT, and from peripheral non-selective units in V4, IT, and LPFC. (A) V4 versus IT. (B) IT versus LPFC. (C) V4 versus LPFC.
Peripheral feature and spatial attentional modulation in V4, IT, and LPFC.
(A–C) Population responses of peripheral non-selective units in V4, IT, and LPFC, respectively, to target stimuli and to the same stimuli when they appeared as distractors. (D–F) Population responses of the same units to stimuli followed by saccades directed into their receptive fields (attention in) versus away from their receptive fields (attention out), in V4, IT, and LPFC, respectively. (A, D) V4. (B, E) IT. (C, F) LPFC. Shaded area denotes ±SEM across units.
Feature and spatial attention in peripheral non-selective units.
Left panels show average population responses to target stimuli and distractor stimuli (i.e., feature attention). Right panels show responses of the same units to stimuli followed by saccades into their RFs (attention in) or out of their RFs (attention out) (i.e., spatial attention). (A–C) The current fixation is on a distractor and the subsequent fixation is also on a distractor. (D–F) The current fixation is on a distractor and the subsequent fixation is on a target. (G–I) The current fixation is on a target and the subsequent saccade is on a distractor. (A, D, G) V4 (n = 727). (B, E, H) IT (n = 216). (C, F, I) LPFC (n = 463).
Feature attentional modulation and saccade transitions around target fixations.
(A) Four saccade sequences around target (T) fixations. Tpre and Dpre represent the target and distractor fixations immediately before the target fixation, respectively. Tpost and Dpost represent the target and distractor fixations immediately after the target fixation, respectively. Arrows (→) indicate saccades. In the Tpre→T→Tpre sequence, the pre- and post-target fixations were on the same stimulus. The proportions of trials/saccades for each sequence are shown along the lines/arrows (left panel). A summary of the four saccade sequences is shown in the right panel. (B–G) Average population responses to the fixated target stimulus in the Tpre fixation of the Tpre→T→Tpre (‘Tpre re-fixated’, red solid line) and Tpre→T→Dpost (‘Tpre not re-fixated’, blue solid line) sequences, and to the same distractor stimulus in the Dpre fixation of the Dpre→T→Tpost/Dpost sequences (the red dashed line indicates ‘Dpre away’ fixations matched with ‘Tpre re-fixated’ fixations; the blue dashed line indicates ‘Dpre away’ fixations matched with ‘Tpre not re-fixated’ fixations). (B) IT foveal face-selective units (n = 476). (C) IT foveal house-selective units (n = 336). (D) IT foveal non-selective units (n = 556). (E) V4 foveal face-selective units (n = 246). (F) V4 foveal house-selective units (n = 304). (G) V4 foveal non-selective units (n = 1051). The corresponding behavioral conditions are illustrated on the right side of the plots. Shading around the mean firing rates indicates ±SEM. (H–J) Population responses of peripheral non-selective units to target stimuli in their receptive fields in Dpre fixations of the Dpre→T→Tpost (‘Tpost’, red solid line) and Dpre→T→Dpost (’T not fixated’, blue solid line) sequences, and to the same distractor stimulus in the Dpre fixation of the Dpre→T→Tpost/Dpost sequences (the red dashed line indicates ‘Dpre away’ fixations matched with ‘Tpre re-fixated’ fixations; the blue dashed line indicates ‘Dpre away’ fixations matched with ‘Tpre not re-fixated’ fixations). (H) V4 (n = 566). (I) IT (n = 103). (J) LPFC (n = 414). Behavioral conditions for the peripheral units are illustrated on the right side of the plots.
Influence of stimulus category on feature attentional modulation.
(A–D) Feature attention effects on responses to four subsets of house stimuli for IT house-selective units (n = 339). The subsets were sorted from low to high based on the amplitude of the responses to house distractor stimuli. The shaded areas indicate the time window (150–225 ms after fixation onset) used for analyzing attentional effects. (E–H) Feature attention effects on responses to four subsets of face stimuli for the same IT house-selective units (n = 339). (I–L) Attentional effects as a function of visual response amplitudes to subsets of face and house stimuli. (I) IT house-selective units (n = 339). (J) IT foveal face-selective units (n = 480). (K) V4 foveal house-selective units (n = 301). (L) V4 foveal face-selective units (n = 266). Normalized visual response amplitudes to subsets of house and face distractors (calculated in a window from 50–225 ms after fixation onset) are plotted against the amplitude of attentional effects (attended – unattended, calculated in a window from 150–225 ms after fixation onset).
Response of foveal and peripheral units.
(A–D) V4 foveal units (n = 1898). (E–H) IT foveal units (n = 1511). (I–L) V4 peripheral units (n = 765). (M–P) IT peripheral units (n = 239). (Q–T) LPFC peripheral units (n = 507). (A, E, I, M, Q) Mean population response aligned at cue onset. (B, F, J, N, R) Mean population response aligned at array onset. (C, G, K, O, S) Mean population response aligned at fixation onset. Firing rates were normalized to the maximum response. Shaded areas denote ±SEM across units. The foveal units responded to the foveal cue stimulus but not to the peripheral stimuli in the search array before the monkeys fixated on them. In contrast, the peripheral units responded to the peripheral stimuli in the search array but not to the cue stimulus at the center of the screen. (D, H, L, P, T) Distribution of the category selectivity index for each group of units.
Attentional modulation in foveal units.
(A–F) Distribution of feature attention index. (A) V4 face-selective units. (B) IT face-selective units. (C) V4 house-selective units. (D) IT house-selective units. (E) V4 non-selective units. (F) IT non-selective units. For category-selective units, we used stimuli from their preferred category. For non-category-selective units, we used both face and house stimuli. The vertical blue line indicates zero, and the blue arrow indicates the mean index. Asterisks indicate a significant difference from zero (Wilcoxon signed-rank test, P < 0.05). (G, I) Normalized population responses of foveal non-selective units during fixations on face target, face distractor, house target, and house distractor. Error shade denotes ±SEM across units. The gray shaded rectangle indicates the time window (150–225 ms after fixation onset) used to calculate the feature attention index. (H, J) Comparison of feature attention index for house (x-axis) and face (y-axis) stimuli in the foveal non-selective units. Each circle represents a unit. The insets at the top right show the distributions of the projections of all units onto the diagonal, which were not significantly different from zero. (G, H) V4 non-selective units. (I, J) IT non-selective units. (K–P) Foveal feature-based attentional modulation in V4 and IT with target and distractor fixation durations equalized. Legend conventions as in Fig. 2.
Peripheral feature and spatial attentional modulation in V4, IT, and LPFC, with responses aligned to saccade onset.
Legend conventions as in Fig. 4.
Feature and spatial attention in peripheral non-selective units, with responses aligned to saccade onset.
Legend conventions as in Fig. 5.
Foveal responses during saccade transitions around target fixations.
Legend conventions are the same as in Fig. 6, but responses are aligned to the onset of the second saccade.
Influence of stimulus category on feature attentional modulation.
Legend conventions as in Fig. 7. (A–H) V4 house-selective units. (I–P) V4 face-selective units. (Q–X) IT face-selective units. (A–D, M–P, U–X) House stimuli. (E–H, I–L, Q–T) Face stimuli.
