Figures and data
Testing visual perception in monkeys and humans
(A) Schematic of the setup used to test visual perception in monkeys and humans. Monkeys were given a juice reward on making a correct response (indicated by the green juice pipe). An example video of one of the monkeys performing this task is available in Supplementary Video S1. Humans received no such tangible rewards but were instead compensated monetarily for their participation. (B) Schematic of the oddball visual search task used to test visual perception in monkeys and humans. Participants had to touch and hold (red dashed lines are for our reference) a yellow button till a search array appeared. The search array always contained one oddball item among multiple identical distractors. Participants could touch any item to complete the trial; touching the oddball target was taken as a correct response.
Monkeys have similar coarse object representations as humans
(A) Animate and inanimate objects used in Experiment 1. (B) Object dissimilarities measured from visual search for humans, who perceive animate objects to be similar to each other (blue cluster), and dissimilar to other objects. (C) Same as in (B), but for monkeys who also perceive animate objects to be similar to each other (green cluster) and dissimilar to all other objects. (D) Object dissimilarities for monkeys plotted against object dissimilarities in humans across all 28 pairs of objects tested. r is the Pearson’s correlation (**** indicates p < 0.00005). (E) Inter-participant consistency for all pairs of participants (both human and monkey), calculated as the pairwise correlation between participants’ object dissimilarities. Large values indicate high agreement between that pair of subjects.
Monkeys experience Weber’s law like humans
(A) Schematic of an oddball visual search trial for testing luminance discrimination. The stimuli were placed in a line centred on the screen and participants had to touch the brightest stimuli, which was always the target, to complete the trial successfully. (B) Accuracy of humans on oddball search for targets differing in luminance as a function of absolute luminance change for high luminance (blue) or low luminance distractors (gray). Shaded regions depict s.e.m. of accuracy across 16 participants. The best-fitting sigmoid line across both conditions is overlaid in red. The dotted line indicates chance accuracy level. (C) Same data as in panel B but replotted as a function of percent change in luminance of the target relative to the distractor. The best fitting sigmoid line across both conditions is overlaid in ochre. The dotted line indicates chance accuracy level. (D) Mean squared error between the sigmoid fit and the observed data for absolute luminance changes (red) and relative luminance changes (ochre). Asterisks indicate statistical significance obtained from a rank-sum test across 20 accuracy values in both conditions (*** indicates p < 0.0005) (E-G) Same as B-D but for data from 3 monkeys doing the same task.
Monkeys experience rapid amodal completion like humans
(A) Occluded objects are rapidly and automatically completed during visual search. In the key comparison, search for an occluded display (left) among “likely completions” (middle) as distractors is hard, implying that the two displays have low dissimilarity. By contrast, searching for a “mosaic” display (right) among likely completions (middle) is easy, implying that these two displays are dissimilar. The shapes at the bottom show the full set of stimuli tested in the experiment. (B) Average dissimilarities for occluded-likely and mosaic-likely searches for humans (blue, left) and monkeys (green, right). Asterisks indicate statistical significance using a t-test across the dissimilarities of 10 shapes tested (*** is p < 0.0005; **** is p < 0.00005). (C) Effect strength for each condition. Dissimilarities in visual search for monkeys across all searches tested plotted against the corresponding values for humans. The correlation between all points is indicated on the top left (r = 0.72). Correlations for each group of pairs are also shown separately: occluded among likely (red dots) and mosaic among likely (ochre dots). Asterisks represent statistical significance (*** is p < 0.0005). Correlation values without an asterisk are not statistically significant. (D) Effect strength for each participant. Average dissimilarities of each participant for occluded-likely searches plotted against the dissimilarities for the corresponding mosaic-likely searches. Blue dots represent the 16 human participants; green dots represent 3 monkeys. The dashed line indicates the y = x line. It can be seen that nearly all human and monkey participants showed a larger dissimilarity for mosaic-likely searches compared to occluded-likely searches, confirming the amodal completion effect. (E) Inter-participant consistency for all pairs of participants (both human and monkey), calculated as the pairwise correlation between participants’ object dissimilarities. Large values indicate high agreement between that pair of subjects.
Monkeys show no mirror confusion unlike humans
(A) Perception of mirror confusion of natural images in visual search for humans. Below are the 15 stimuli arranged according to the aspect ratio: calculated as the ratio of the width to the height. (B) Mean ± s.e.m. of the dissimilarities for humans and monkeys across the two types of trials, vertical mirror among reference: the target was a vertical mirror image embedded among reference displays, horizontal mirror among reference: the target was a horizontal mirror image embedded among reference displays, asterisks indicate significance of the Student’s t-test checking for differences between the means where **** is p < 0.0005, and n.s. is p > 0.05 (paired samples), data points are unique trials (averaged across all subjects and repetitions). (C) Effect strength for each search. Difference in dissimilarity between horizontal and vertical mirror conditions plotted against the aspect ratio of the 15 stimuli tested for humans (blue) and monkeys (green). The r values are the Pearson correlation and coloured lines are the best-fitting lines for each group. (D) Effect strength for each participant. Mean dissimilarities for vertical mirror among reference and horizontal mirror among reference pairs for each subject, 7 humans and 3 monkeys. (E) Inter-participant consistency for all pairs of participants (both human and monkey), calculated as the pairwise correlation between participants’ object dissimilarities. Large values indicate high agreement between that pair of subjects.
Monkeys show a local advantage unlike humans
(A) Humans show a global advantage when viewing hierarchical stimuli, such as a circle made of diamonds. Specifically, for humans, shapes differing only at the global level are perceived as more dissimilar than shapes differing only at the local level (i.e. dGlobal > dLocal). (B) Hierarchical stimuli used for the visual search experiment in humans and monkeys. An example video of one of the monkeys performing this task is available in Supplementary Video S1. (C) Dissimilarity between pairs of hierarchical stimuli during visual search, in humans (left group, blue) and monkeys (right group, green). Dissimilarity is shown separately for three groups of search pairs: pairs differing only in global shape (left bars), differing only in local shape (middle bars) and differing in both global and local shape (right bars). Dots centred around each bar represent average dissimilarity for each individual participant. Asterisks indicate statistical significance obtained from an unpaired t-test comparing the mean dissimilarities in each group (n = 9 for global change and local change, n = 18 for both change, ** is p < 0.005, *** is p < 0.0005, ***** is p < 0.00005, non-significant comparisons are not marked). (D) Effect strength for each condition between humans and monkeys. Average dissimilarity for each search pair in monkeys plotted against the corresponding values in humans. The overall correlation between all search pairs is shown on the top left (r = 0.09). Correlations for each group of pairs are also shown separately: pairs differing only in global shape (red dots), in local shape (purple dots) and differing in both global and local shape (ochre dots). Asterisks represent statistical significance (** is p < 0.005). Correlation values without an asterisk are not statistically significant. (E) Effect strength for each participant. Average dissimilarity for each search pair with a global shape difference plotted against the dissimilarity for the corresponding local shape difference for individual participants (humans, blue dots; monkeys, green dots). It can be seen that all humans find global shape differences consistently larger than local shape differences, whereas all monkeys show exactly the opposite pattern. (F) Inter-participant consistency for all pairs of participants (both human and monkey), calculated as the pairwise correlation between participants’ object dissimilarities. Large values indicate high agreement between that pair of subjects.
Monkeys show a local advantage even for connected shapes
(A) Hierarchical stimuli used for this experiment. Here, we connected local elements within each global shape with a gray line to facilitate grouping into a global shape. (B) Avergage dissimilarity between pairs of hierarchical stimuli during visual search in monkeys, shown separately for three groups of search pairs: pairs differing only in global shape (left bars), differing only in local shape (middle bars) and differing in both global and local shape (right bars). Dots centred around each bar represent average dissimilarity for each individual participant. Asterisks indicate statistical significance obtained from an unpaired t-test comparing the mean dissimilarities in each group (n = 9 for global change and local change, n = 18 for both change, * is p < 0.05, **** is p < 0.00005, non-significant comparisons are not marked).
