Since perceptual and neural face sensitivity is associated with a foveal bias, and neural place sensitivity is associated with a peripheral bias (integration over space), we hypothesized that face perception ability will decline more with eccentricity than place perception ability. We also wanted to examine whether face perception ability would show a left visual field (LeVF) bias due to earlier reports suggesting right hemisphere dominance for faces, or would show an upper or lower visual field bias. Participants performed foveal and parafoveal face and house discrimination tasks for upright or inverted stimuli (≤4°) while their eye movements were monitored. Low-level visual tasks were also measured. The eccentricity-related accuracy reductions were evident for all categories. Through detailed analyses we found (i) a robust face inversion effect across the parafovea, while for houses an opposite effect was found, (ii) higher eccentricity-related sensitivity for face performance than for house performance (via inverted vs. upright within-category eccentricity-driven reductions), (iii) within-category but not across-category performance associations across eccentricities, and (iv) no hemifield biases. Our central to parafoveal investigations suggest that high-level vision processing may be reflected in behavioural performance.
One of the main coding principles in the visual cortex is eccentricity, where the foveal representation is significantly magnified in the visual cortex relative to its size on the retina, and as distance from the fovea grows, the cortical representation is reduced1,2,3. This foveal magnification is assumed to be the main factor contributing to reduced performance with growing eccentricity for multiple (but not all) visual tasks4,5,6. Studies also show that this foveal enhancement can be overcome if peripheral information is scaled to match the cortical representation (aka M-scaling7).
Another coding principle is hierarchical processing, where visual information is initially processed according to the physical aspects in early visual areas (e.g. in V1, V2) and as information progresses in the visual hierarchy it is integrated and becomes associated with perceptual aspects (specialization) rather than the physical aspects of the stimulus in higher visual areas (e.g. in LOC, FFA, PPA)8,9,10,11,12,13,14. This high-level visual specialization is predominantly manifested in two anatomically distinct visual processing pathways, the ventral perception visual pathway processing aspects related to shape and form, and the dorsal spatial visual pathway processing aspects related to action preparation and location in space15,16,17,18,19.
Multiple neuroimaging and electrophysiological studies investigating the ventral perception visual pathway have revealed that face processing is associated with foveally biased regions and critically rely on them while place and house related processing are associated with and rely on peripherally biased regions20,21,22,23,24,25,26,27,28,29,30 sensitive to integration over space20,31. In addition, population receptive field (pRF) studies with faces show that most of the pRFs in the ventral face areas include the foveal area, and extend to cover parafoveal locations during peripheral face task32,33. Recent studies further suggest that the structure of these category selective regions is also distinct25,26.
Reasoning that neural coding preferences likely reflect behavioural performance, and based on earlier studies and on reduced visual performance at peripheral locations6, we hypothesized that face related tasks, that are associated with a cortical foveal bias would show reduced performance in more peripheral locations relative to house related tasks that are associated with a cortical peripheral bias. Thus we expected that place related tasks would be less affected by eccentricity and would show, on top of the expected eccentricity reductions6, higher performance than faces as eccentricity increases (as presented in the middle and left models in Fig. 1a). While we would like to compare face and house related performance, accuracy levels may be affected by across category task difficulty differences. To account for potential across categories performance differences at the center (Fig. 1a on the left) each category’s performance can be normalized according to central performance (Fig. 1a middle panel). Such normalization allows examining the effects of the eccentricity-related performance reductions relative to central performance. However, we cannot rule out the possibility that despite differences in neural representations, face and house performance will not show significant differences across foveal to parafoveal locations (see Fig. 1a right model). We also wanted to examine whether face perception ability may show a left visual field (LeVF) bias due to earlier reports suggesting right hemisphere dominance for faces34,35,36,37,38,39,40,41, and whether there would be any upper-lower visual field differences as suggested earlier42. Lastly, due to the evident VF biases in scene/place-related regions in pRF studies (UVF bias in ventral PPA and LoVF bias in dorsal TOS43), we also hypothesized that there may be an UVF vs LoVF difference in house performance. To test these hypotheses we measured face and house discrimination performance for upright and inverted stimuli at foveal to parafoveal locations (up to 4°, see Fig. 1b–e) in each of the four visual field quadrants in a group of normally sighted individuals while participants held fixation and their eye movements were being monitored. We reasoned that since inversion has been used to expose holistic processing of central faces (e.g44,45,46,47), measuring how eccentricity modulates inversion may expose characteristics reflecting category-specific mechanisms. To minimize potential task difficulty related differences between face and house performance, on top of examining performance in the same visual field locations, we also examined normalized performances, reductions in performance levels as a function of eccentricity (i.e. slopes), and how inversion is modulated by eccentricity. In addition, we reasoned that if category-related performance relies on category-specific mechanisms we should expect that despite the eccentricity-related reductions in performance, performance across eccentricities would be correlated for a specific category, but not across categories.
In line with the eccentricity expected reductions in performance, we found that for all visual categories accuracy and d-prime were highest at the centre of the visual field and dropped significantly as eccentricity increased (see Fig. 2, Table 1 for detailed results, and Table 2 for statistical results). The superior performance at the centre of the visual field was also evident in longer RTs for growing eccentricities for all categories and orientations (see Fig. 2c,d, Table 1 for detailed results, Table 2 for statistical results, and Supp. Mat.).
On top of the expected eccentricity-related reductions we found (see Fig. 2 and Table 1), we directly assessed whether faces and houses are qualitatively different in a series of analyses. First, in Fig. 2a and b we can see the comparison between upright and inverted performance for faces (Fig. 2a) and for houses (Fig. 2b). While performance was significantly higher for upright than for inverted faces (see statistical results of a 2-way ANOVA in Table 2) corresponding to the face inversion effect, for houses, following a 2-way ANOVA analysis, we found that inverted houses’ performance was significantly higher than that of upright houses (see Table 2), and for both categories this was not a result of speed-accuracy trade-off (see Fig. 2 and compare panels c with a and d with b). Second, since performance as a function of eccentricity may be relative to performance at the center, we normalized accuracy measurements according to performance at the center that served as a baseline, and this was done for each participant and each category, creating performance indices. These results can be seen in Fig. 2e, reflecting the opposite effects that inversion had on face (reduction) and on house (increase) discrimination (see Table 1). It can also be seen in Fig. 2c and d that reaction times were differently affected by inversion for each category (see Table 1 and for statistical results see Table 2). The next set of analyses focused on a group of participants (n = 15) that participated in the 4 main experiments (upright/inverted faces/houses) allowing us to perform within-participant and between-category analyses (Figs. 3 and 4). Since we wanted to estimate how performance of each category is modulated by eccentricity, and since the reductions in performance from central vision to parafovea were apparently linear, we calculated for each participant the slope of the eccentricity-driven performance decline for upright and inverted faces and houses. As can be seen in Fig. 3a that presents the average eccentricity-driven slopes for each category and each orientation, we found a significant effect of category and an interaction between category and stimulus orientation, also reflecting the results in Fig. 2 (see Table 2 for statistical details), and this indicates that inversion affects performance differently for faces than for houses. Figure 3b displays individuals’ slopes for upright and inverted faces (dark gray) and houses
(light gray). Since we hypothesized that the differences between the upright and inverted slopes may be a possible category distinguishing feature, we calculated a single characteristic measure for each category that took into account both the upright and the inverted performance modulation by eccentricity and their relations. This measure was a subtraction of the inverted slope from the upright slope for each category (see Fig. 3c). This measure was found to be significantly different between the two categories (see Table 2). Individual data of this analysis are presented in Fig. 3d. Lastly, following the logic of Haxby et al.48, we reasoned that if faces and houses are supported by distinct category-specific mechanisms, then performance would be associated across eccentricities for each category but not across categories. On the other hand, if performance is only driven by eccentricity and not by category-specific mechanisms, then we should expect that performance across categories would be correlated. We found a significant correlation between performance index in 3° and in 4° for faces (R2 = 0.607, p = 0.0006, non-directional, Fig. 4a) and for houses (R2 = 0.470, p = 0.004, non-directional, Fig. 4b) supporting category specific mechanisms. Furthermore, when comparing faces at 4° and house performance at 3° (R2 = 0.095, p = 0.262, non-directional, Fig. 4c) or face performance at 3° and house performance at 4° (R2 = 0.089, p = 0.278, non-directional, Fig. 4d) we found that these were not significantly correlated (see Fig. 4c, d) which further supports category-specific mechanisms.
Estimating visual acuity
VA related measurements were also estimated by the ‘VA’ tumbling E experiment at different locations to provide an estimate for basic vision performance decline with eccentricity. As can be seen in Fig. 2f, at 2° visual acuity seemed to decline relative to the centre by an average of 0.2 LogMAR units which correspond to 2 lines on the ETDRS chart, and an additional decline of another line on the ETDRS chart for 4°. Typically, a value of 0 LogMAR corresponds to static VA of 6/6, and a 0.1 difference reflects one line in the static chart.
Visual field comparisons
As can be seen in Fig. 5 and in Supplementary Tables ST2 and ST3, no significant differences between upper visual field (UVF, in blue) and lower visual field (LoVF, in green) or between left visual field (LeVF, in red) and right visual field (RVF, in yellow) were found for upright faces or for any other category.
Estimating physical differences’ contribution to performance
We calculated the physical differences between sets of images used in our experiments to examine whether these could possibly account for performance. To that end, we estimated for each pair of different images used in the discrimination house or face experiments the physical difference between them (Euclidean distance) and examined whether bigger physical differences would be associated with higher performance (see Methods for more details). We reasoned that a bigger physical difference would facilitate distinguishing between different images (and thus improving accuracy on the ‘different’ condition). Figure 6 shows the physical difference between pairs of images for each experiment (Fig. 6a), accuracy performance (Fig. 6b) and RTs (Fig. 6c) for faces and houses for the different house discrimination experimental versions (see details in Methods) as a function of eccentricity. As can be seen, accuracy of house version 1 was higher than that of house version 2 (post-hoc p = 0.004), but house version 2 was not significantly different than face accuracy (post-hoc p = 0.241, see also Supplementary Tables ST4 for details and ST5 for statistical results). However, despite the significant physical differences between the house versions, there was no significant difference between the RTs of the two house versions (see ST5). The physical difference between the house versions can explain the differences in accuracy between the house versions (version 1 with higher physical differences between the images may have made house discrimination easier, and indeed showed higher accuracy than version 2 with the smaller physical differences). However, physical differences were not likely to account for the face vs house performance differences since despite significant different physical differences between the face and house version 2 (p < 0.0001), there was no significant difference between face and houses version 2 performance (p = 0.241, see also Supplementary Table ST5).
A group of 41 neurotypical participants aged 18-47 (mean age 26.6 years ± 6 (SD)) with normal or corrected-to-normal vision participated in this study. After exclusion (see criteria below) 29 participants took part in the upright face discrimination experiment, 22 of them participated in the inverted face discrimination experiment, 18 of them and 4 additional participants (altogether 22 participants) participated in the upright house discrimination experiment, 15 of them participated in the inverted house discrimination experiment, 18 of them and an additional participant (altogether 19 participants) participated in the tumbling E experiment49, and 6 of them and additional 4 participants (altogether 10 participants) participated in the shape discrimination experiment. All participants signed an informed consent form before their participation. The Bar Ilan University ethics committee approved all the experimental protocol, and the methods were carried out according to the ethics committee guidelines and regulations. Participants whose accuracy or RTs exceeded 2 SD from the mean of the group in any of the experiments were excluded from all the study analyses.
All experiments were conducted on an Eizo FG2421 24” HD LCD monitor (1920×1080 pixels resolution) running at 100 Hz, using an in-house developed platform for psychophysical and eye-tracking experiments (PSY) developed by Yoram S. Bonneh50 running on a Windows PC.
In order to ensure fixation during all experimental sessions, eye movements were recorded with an EyeLink infrared system (SR Research, Ontario, Canada) with a sampling rate of 500 Hz, equipped with a 35 mm lens while head movements were limited by chin rest. Eye movements were recorded binocularly, only left eye data were analyzed. A standard 5-point calibration was performed before each session. In addition, in-session calibration trials were incorporated into the beginning and end of each session to estimate the deviation from fixation for each session. This was computed as the difference between the eye position during the in-session position calibration (baseline) and the average eye position during that session. Trials with a deviation from fixation greater than 0.5° during a time window around the onset of the target stimulus were excluded from the analysis.
Participants were instructed to look at the fixation point and be alert to and aware of the surrounding area in which the target could appear. Each experimental session included eye position calibration trials before and after the main experiment (see more details below). Throughout all experimental sessions participants started each trial at their own pace by pressing a key after fixation appeared at the centre of the screen. Experiments were performed in a dark room, no feedback was given, and the viewing distance was 60 cm.
Upright faces discrimination experiment
In each trial, participants were asked to judge whether two faces presented sequentially were the same or different. A study face always appeared at the centre of the screen for 200 ms, and after an ISI of 250 ms, a target face (either same or different than the study face) appeared for 200 ms (to eliminate the possibility of succeeding in the task if performing a saccade towards the target) at a location chosen randomly from 13 possible locations in the visual field (centre or one of 4 locations at 2°, 3° or at 4°, as depicted in Fig. 1d). The face images for the same condition (same image served as the study and the target face) were chosen randomly from a set of 5 different faces, each appeared once in each location per session. For the different condition, each of the two face images was chosen randomly from two different sets of 5 face images each, 5 of the face images used for the different condition were the images used in the same condition. All face images were full-front colored photographs of real men with a neutral expression cropped and aligned to each other (see full details at51,52 with the original images taken from 2 databases (CVL Face Database [http://www.lrv.fri.uni-lj.si/facedb.html]53; AR Face Database [Martinez and Benavente 1998]54). The faces were presented on a black background and subtended 1.6° × 2.2° (width x height). Trials were mixed randomly in terms of condition (same/different) and location in the visual field. Each participant underwent 5 runs of the experiment with each condition (same/different) repeated 5 times in each of the 13 locations. Overall, there were 25 ‘same’ face trials and 25 ‘different’ face trials for each of the 13 locations for each participant. Face discrimination performance was measured as accuracy (percent correct) per location. Experimental procedures are illustrated in Fig. 1b–d.
Upright houses discrimination experiment
The experimental design was identical to that used in the face discrimination paradigm except for the use of grayscale images of real houses rather than the face images. We ran two versions of the experiment, each using a different set of house images. The first version was built to equate central performance between faces and houses. However, since the physical difference between the house and face experiments was big, we created an additional house version with a different set of houses to reduce the physical between stimuli differences. Each participant performed the first version (V1) twice and the second version (V2) 3 times; order of sessions was V1-V2-V2-V1-V2. House images subtended 2.4° × 2° (width x height). Experimental procedures are illustrated in Fig. 1b–d.
Inverted faces discrimination experiment
The experimental design was identical to that used in the face discrimination paradigm except that faces were inverted (both study and target faces). Experimental procedures are illustrated in Fig. 1b–e.
Inverted houses discrimination experiment
The experimental design was identical to that used in the upright houses discrimination paradigm except that houses were inverted (both study and target houses). Experimental procedures are illustrated in Fig. 1b–e.
VA tumbling ‘E’ experiment
A tumbling E test was used to measure visual acuity (VA) threshold49 at 9 different visual field locations (centre, 2° and 4°, 4 locations at each eccentricity). Separate staircase procedures were applied for each of the 9 locations; trials of all locations were mixed randomly in each session. The stimuli were a black E on gray background that in each location subtended initially 0.5° and faced 1 of 4 optional directions. Participants’ task was to determine the E’s facing direction (4AFC), while the E’s size was reduced in a 3:1 staircase procedure with 0.1 log unit steps according to performance (stopping after 6 reversals). Experimental procedures are illustrated in Fig. 1d,e and Supplementary Figure SF1. Results are reported in log MAR units (minimum angle of resolution (MAR) needed to correctly discriminate (at 79% accuracy performance) the E’s facing direction, values closer to 0 indicate better VA).
Note that our measurements of dynamic VA are not precisely comparable to the standard static ETDRS measurements since (i) our viewing distance (60 cm) was not standard (for near 40 cm or for far 3 m), (ii) exposure time was limited (vs. unlimited exposure in static VA examinations), (iii) divided attention across the VF (vs. focusing all attention on the centre of VF), (iv) contrast was lower than used in standard VA tests. However, the dynamic tumbling E measurements have been shown to correspond to VA standard measures55,56.
Shape discrimination experimental details are provided in Supplementary Material.
All analyses reported were calculated individually and then the average over participants and SE are reported.
For each of the discrimination experiments performance was measured as accuracy (percent correct) per location and for the VA experiment performance was reported in log MAR units per location (as presented in Fig. 2). For each location we calculated individual performance and we present the average results over all participants. For each experiment and each eccentricity, we averaged the performance of all 4 locations (upper/lower left/right locations) of that eccentricity. RT analyses were done in the same way as for accuracy measures.
In order to assess the eccentricity effect (i.e. the drop in visual performance with eccentricity) and compare it across all discrimination tasks, we computed a performance index (PI), see Fig. 2e. PI was calculated as the performance (accuracy) at each eccentricity divided by the performance at the centre. In the group of the n = 15 that performed all 4 main experiments (upright/inverted faces and houses) we measured the eccentricity related reductions in performance by calculating the slopes of the decline from central vision (0 eccentricity) to 4° of eccentricity by subtracting the performance at 0° from that of 4° (the average performance across all 4 locations in that eccentricity), and these are presented in Fig. 3a,b. Slope differences were calculated for each visual category (faces and houses) as the difference between the upright slope and the inverted slope (slope difference = upright slope – inverted slope). As before, this was calculated for each participant and then averaged across participants. Results are presented in Fig. 3c,d.
To compare upper vs. lower VF and right vs. left VF performances, the two locations at each eccentricity and each hemifield were averaged (e.g. to compare UVF vs. LoVF, for the 2°, 2° right UVF and left UVF were averaged and compared to the average of 2° right LoVF and left LoVF), see Fig. 5 and Supplementary Figure SF2.
To estimate physical distances (or differences) that may account for differences in performance, we calculated the Euclidean distance between 2 different face images (for faces) or 2 different house images (for houses). Euclidean distance between 2 images was estimated as the mean luminance level absolute difference over all pixels in these images. These are presented in Fig. 6a.
Statistical analyses (ANOVA and post-hoc) were performed with StatView5.0 software for Windows (SAS Institute Inc, Cary, NC) and presented in Table 2, Supplementary Tables ST3 and ST5.
Error bars for accuracy, RTs, and VA represent standard errors across participants calculated using the Cousineau method (see57 for details). For the performance indices and slope-related analyses (see below) error bars represent standard error of the mean (SEM).
Our investigations into face and house-related processing in central to parafoveal locations (up to 4°) revealed that, as expected, all tested categories, showed an eccentricity effect (i.e. reduced accuracy and d-prime, and slower RTs with growing eccentricity6). Our findings also support our hypothesis that place-related processing would be less affected by eccentricity than face related processing (as proposed in both left and middle models in Fig. 1a), based on earlier neuroimaging findings that faces are associated with foveal bias and places with a peripheral one20,21,23,24,25,26,29,30. Our hypothesis was supported by larger slopes of eccentricity related reductions for faces than for houses. In addition, for each category the slope of eccentricity related reductions was modulated quantitatively (i.e. magnitude) and qualitatively (i.e. direction) different by inversion suggesting different processing for faces and for houses in the parafovea. Specifically, accuracy of inverted faces was worse than that of upright faces at all eccentricities, which extended the central vision face inversion effect45,58,59 to the parafovea. In contrast, for houses, we found an opposite effect with higher performance for inverted stimuli. Furthermore, we found that within-category but not between-category performance was associated across eccentricities, again supporting dissociated category-related mechanisms which is in line with functional and structural brain studies25,48. In addition, we found no significant performance biases for LeVF, RVF, upper or lower visual fields (see also Supplementary Material).
Our study aimed to assess whether high-level visual behavioural performance reflects brain organization as evident from functional and anatomical brain studies of human visual cortex. Multiple neuroimaging and electrophysiological studies20,23,24,25,26,30 show that faces are associated with foveal processing and houses and places with peripheral processing and integration over space31. We anticipated that this will be reflected in behaviour according to the model proposed in Fig. 1a left and middle panels. In earlier neuroimaging studies that investigated eccentricity biases in face and place related regions, the goal was to determine whether the overall eccentricity biases coexisted or were evident in category sensitive regions (using category specific stimuli of up to ~6° from the fovea, and overall general visual stimuli of up to ~10° from the fovea)20,30. In our study we were interested to investigate how category specific behaviour is modulated by eccentricity within parafoveal domain (≤4°). Some earlier neuroimaging studies have also addressed the relation between category related areas and the effects of stimulus size on their activation levels (e.g20,31,60.). Some of these studies show that the activity in face-related areas is not affected by image size changes for central stimuli while the activity in place-related areas is influenced by size. Another study shows that for peripheral stimuli (at 10°) image size does seem to enhance the face-related N170 component, and this may be due to the larger images penetrating the central receptive fields of the face areas33. In our study we kept the stimulus sizes constant across the parafovea reasoning that as in natural vision, objects’ world size does not change. Under these experimental conditions and while refraining from direct between-category comparisons, our behavioural results support high level vision brain organization principles for faces and places (see above).
When comparing visual categories in brain imaging research most often less emphasis is given to controlling task difficulty across categories (e.g20,21,23,48.). However, in our study we wanted to investigate whether and how the extensively studied high-level visual brain mechanisms of faces and places are reflected in behaviour. To study this behaviourally we wanted to compare between categories. Therefore, we needed to control for differences in task difficulty between the categories that despite our attempts to equate them may arise from the different processing characteristics of each category (foveal vs peripheral bias, differences in spectral features or in physical world size) or from our experimental design. To that end we refrained from directly comparing between accuracy levels of the two categories. Instead, we exploited inversion since stimulus inversion is known to adversely affect the holistic nature of face perception44,45,46,47 and since both categories are typically perceived in an upright orientation. We used inversion as a category-related measure to assess within-category inversion effects across eccentricities, and these were compared between faces and places. This enabled us to relate behaviour to neuroimaging studies investigating faces and places. Our results are in line with the model suggesting that faces and places are differentially modulated by eccentricity in the parafovea as depicted in Fig. 1a left and middle panels. In line with neuroimaging results, we found that faces are more adversely affected by eccentricity than houses. This was evident in a series of analyses that we performed (see Results) resulting in a single characteristic measure (slope difference, see Results and Methods) that indicated that the magnitude of modulation for faces by eccentricity is bigger than that of houses. The fact that the faces’ slope difference was in an opposite direction to that of houses (see Fig. 3c), and that within category (but not across categories) performance was correlated across eccentricities support the idea that the mechanisms underlying face and place processing are dissociated (for review see25), and our study extends this to parafoveal vision.
Since faces are known to show a foveal bias, it is yet unclear how face inversion will be modulated by eccentricity. Although inverted faces keep the local features and spatial relations of the faces intact, the stimuli and their holistic structure are unfamiliar to the visual system and this leads to the face inversion effect45,58,59, where people are worse when performing a task on inverted faces relative to upright faces. Earlier studies show that processing of inverted faces recruits both face specific and non-face related mechanisms that include parietal attention-related foci52,58,61,62,63. An earlier study examining face inversion at right or left peripheral locations compensated for the foveal cortical magnification by enlarging peripheral faces (by CMF) reported on an inversion effect in the periphery without any RT differences46. However, we were interested to investigate how the face inversion effect is modulated by eccentricity when face size is kept constant, as is typical during natural vision. One possibility is that face inversion would be evident in peripheral vision if upright and inverted faces performance would both proportionally degrade with growing eccentricity. Another possibility is that in peripheral vision upright face perception will reduce and reach levels similar to those of inverted faces, thus weakening the central face inversion effect. Our design allowed us to examine these two possibilities in both accuracies and response times. Interestingly, while we did not find any effects of inversion on RTs, we found that the face inversion effect in central vision (that was evident in accuracies) did not diminish with eccentricity, and in fact seemed to grow (see the bigger slope for inverted faces relative to that of upright ones).
While inversion is known to affect central face perception and we have also found that is extends to parafoveal vision, it is unclear how inversion affects other visual categories. Here we examined how inversion affects an additional category of houses, which is considered in the neuroimaging literature to be an opposing category to faces (e.g13,14,64). We found, in contrast to face inversion, that accuracy was higher and RTs were faster for discriminating inverted houses relative to upright houses, and this was similar to what was found earlier for central faces and houses65. Thus this inversion effect was different than that found for faces. Several possible explanations may account for this. First, since the house inversion experiments in our study were performed after the upright houses experiments, one could argue that learning effects may underlie the improved performance for inverted houses. However, since this was also the case for faces (with inverted faces being always after upright ones), this explanation may not be the only factor explaining the differences between house and face inversion effects. A second explanation is related to the peripheral bias associated with places. A direct comparison of upright vs inverted central houses did not reveal any significant accuracy difference, and the significant effect we found between upright and inverted houses stems from differences in parafoveal eccentricities (see Fig. 2b). In contrast to faces, mechanisms supporting houses show a periphery related effect20,24,31, and therefore may be more sensitive to inversion (manipulation of stimulus orientation) at the periphery rather than in central vision. This may explain why faces show an inversion effect at the centre of the visual field which grows with eccentricity, while houses are sensitive to inversion only with peripheral stimuli which does not seem to grow with eccentricity (no significant slope difference between the slopes of upright and inverted houses). Thirdly, inversion may reduce holistic processing, and this in turn may lead to local low level feature comparisons, which are assumed to be quicker. This would clearly disrupt holistic face processing but may simplify house discrimination, and the faster RTs for inverted houses relative to upright ones may provide an indirect indication for this. We observed a different pattern of inversion effects on houses than on faces, and thus we anticipate that inversion may influence additional visual categories in a category specific manner (for a review see66). It is possible that face and place mark the limits of this behavioural effect despite possibly relying on different brain mechanisms that come into play under inversion52,58,59,61.
The fact that house related performance was faster than that of faces, even at the fovea, may seem counterintuitive. However, as we mentioned above, task difficulty may play a role and thus can perhaps explain why RTs for houses may be faster. Furthermore, in addition to house/place related processing relying on different cortical mechanisms and different computations than those of faces25,26,64,67, house/place related mechanisms also show a tendency for transient rather than sustained activity64,68. One study shows that house related areas as the house related ventral PPA and dorsal TOS areas also exhibit transient BOLD responses while the face related areas (ventral FFA and also OFA) show sustained BOLD activity and this is independent of the preferred or non-preferred category64. Another recent study shows that different areas in the ventral stream receive different contributions of transient and sustained inputs69. If transient related activity is associated with faster processing than that associated with sustained activity, then these differences could contribute to differences in response timing. Another possibility is that the differences in physical low level aspects of the stimulus could contribute to the speed of processing and this again may relate to task difficulty differences. As we discussed above (and see also Fig. 6a), there were bigger physical differences between pairs of house images than between pairs of face images. Thus, the house task may have been easier and thus a decision may have been reached sooner.
While we found multiple indications that face and house behaviour reflect brain mechanisms supporting face and house processing, we cannot rule out some possible confounding factors that may limit the generalization of our findings. First, our paradigm enabled us to investigate parafoveal eccentricities up to 4° and perhaps at further peripheral locations these effects may be greater or diminish for faces and for houses when peripheral mechanisms start to kick in. A second possibility relates to the small size of our stimuli with an emphasis on the house stimuli. Houses and places in our daily life are typically big and take up a bigger portion of the visual field27. The small houses used in our experiment may have forced the system to rely on additional mechanisms that are non-typical to house related processes, and these non-typical mechanisms contributed to the house-related performance we found. Nevertheless most neuroimaging investigations, like our study, have been carried out with stimuli that are not directly reflecting world size. A third possibility is that low-level aspects of our specific stimulus sets could have contributed to the effects we found for faces and houses. Indeed, the analysis we carried out examining whether physical differences may account for the performances we observed (see Fig. 6) indicated that physical differences may partially but not fully explain performance. Furthermore, it could be argued that all that our experimental design recruited predominantly shape discrimination mechanisms. However, if this were the case then inversion should have resulted in comparable performances. Since there were clear and significant inversion effects for faces and for houses, and they were different from each other, it is unlikely that we merely measured shape discrimination processes. A fourth possibility is that top-down attentional mechanisms influence performance according to visual field location and visual category5,70,71,72. In fact, the face pop out effect73,74 seems to point to enhanced top down influences for faces at peripheral locations relative to other visual categories (but see75). Lastly, despite the fact that our results seem to be in agreement with neuroimaging results, it could very well be that the relationship between brain function and behaviour is much more complex and that our results do not reflect a one to one correspondence with functional activity in high level visual cortex. Hence the effects of eccentricity on face and house discrimination could be partially due to the demands imposed on the system by low-level aspects or by top-down attentional mechanisms rather than being attributed solely to face or house related perceptual mechanisms.
In conclusion, our investigations of face and house discrimination at foveal and parafoveal regions revealed that as expected, performance for all visual categories reduced as a function of eccentricity. Our central to parafoveal investigations of upright and inverted faces and houses allowed us to carry out multiple analyses together suggesting that high level visual processing may be reflected in behavioural performance. Furthermore, our behavioural results support the notion that faces are associated with a foveal bias while houses are less affected by eccentricity.
No datasets were generated or analysed during the current study.
Daniel, P. M. & Whitteridge, D. The representation of the visual field on the cerebral cortex in monkeys. J. Physiol 159, 203–221 (1961).
Virsu, V. & Rovamo, J. Visual resolution, contrast sensitivity, and the cortical magnification factor. Exp. Brain Res. 37, 475–494 (1979).
Rovamo, J. & Virsu, V. An estimation and application of the human cortical magnification factor. Exp. Brain Res. 37, 495–510 (1979).
Carrasco, M., McElree, B., Denisova, K. & Giordano, A. M. Speed of visual processing increases with eccentricity. Nat. Neurosci. 6, 699–700 (2003).
Carrasco, M., Evert, D. L., Chang, I. & Katz, S. M. The eccentricity effect: Target eccentricity affects performance on conjunction searches. Percept. Psychophys. 57, 1241–1261 (1995).
Staugaard, C. F., Petersen, A. & Vangkilde, S. Eccentricity effects in vision and attention. Neuropsychologia 92, 69–78 (2016).
Carrasco, M. & Frieder, K. S. Cortical magnification neutralizes the eccentricity effect in visual search. Vision Res. 37, 63–82 (1997).
Felleman, D. J. & Van Essen, D. C. Distributed Hierarchical Processing in the Primate Cerebral Cortex. Cereb. Cortex 1, 1–47 (1991).
Markov, N. T. et al. Anatomy of hierarchy: Feedforward and feedback pathways in macaque visual cortex. J. Comp. Neurol. 522, 225–259 (2014).
Malach, R. et al. Object-related activity revealed by functional magnetic resonance imaging in human occipital cortex. Proc. Natl. Acad. Sci. 92, 8135–8139 (1995).
Grill-Spector, K. et al. Differential Processing of Objects under Various Viewing Conditions in the Human Lateral Occipital Complex. Neuron 24, 187–203 (1999).
McCarthy, G., Puce, A., Gore, J. C. & Allison, T. Face-specific processing in the human fusiform gyrus. J. Cogn. Neurosci 9, 605–610 (1997).
Epstein, R. & Kanwisher, N. A cortical representation of the local visual environment. Nature 392, 598–601 (1998).
Kanwisher, N., McDermott, J. & Chun, M. M. The Fusiform Face Area: A Module in Human Extrastriate Cortex Specialized for Face Perception. J. Neurosci. 17, 4302–4311 (1997).
Ungerleider, L. G. & Mishkin, M. Two cortical visual systems. in Analysis of Visual Behavior 549–586 (MIT Press, 1982).
Goodale, M. A. & Milner, A. D. Separate visual pathways for perception and action. Trends Neurosci. 15, 20–25 (1992).
Gilaie-Dotan, S. Visual motion serves but is not under the purview of the dorsal pathway. Neuropsychologia 89, 378–392 (2016).
Kravitz, D. J., Saleem, K. S., Baker, C. I., Ungerleider, L. G. & Mishkin, M. The ventral visual pathway: an expanded neural framework for the processing of object quality. Trends Cogn. Sci. 17, 26–49 (2013).
Freud, E., Plaut, D. C. & Behrmann, M. ‘What’ Is Happening in the Dorsal Visual Pathway. Trends Cogn. Sci. 20, 773–784 (2016).
Levy, I., Hasson, U., Avidan, G., Hendler, T. & Malach, R. Center–periphery organization of human object areas. Nat. Neurosci. 4, 533–539 (2001).
Hasson, U., Harel, M., Levy, I. & Malach, R. Large-Scale Mirror-Symmetry Organization of Human Occipito-Temporal Object Areas. Neuron 37, 1027–1041 (2003).
Parvizi, J. et al. Electrical Stimulation of Human Fusiform Face-Selective Regions Distorts Face Perception. J. Neurosci. 32, 14915–14920 (2012).
Hasson, U., Levy, I., Behrmann, M., Hendler, T. & Malach, R. Eccentricity Bias as an Organizing Principle for Human High-Order Object Areas. Neuron 34, 479–490 (2002).
Malach, R., Levy, I. & Hasson, U. The topography of high-order human object areas. Trends Cogn. Sci. 6, 176–184 (2002).
Grill-Spector, K. & Weiner, K. S. The functional architecture of the ventral temporal cortex and its role in categorization. Nat. Rev. Neurosci. 15, 536–548 (2014).
Weiner, K. S. et al. The Cytoarchitecture of Domain-specific Regions in Human High-level Visual Cortex. Cereb. Cortex 27, 146–161 (2017).
Konkle, T. & Oliva, A. A Real-World Size Organization of Object Responses in Occipitotemporal Cortex. Neuron 74, 1114–1124 (2012).
Dilks, D. D., Julian, J. B., Paunov, A. M. & Kanwisher, N. The Occipital Place Area Is Causally and Selectively Involved in Scene Perception. J. Neurosci. 33, 1331–1336 (2013).
Verhoef, B.-E., Bohon, K. S. & Conway, B. R. Functional Architecture for Disparity in Macaque Inferior Temporal Cortex and Its Relationship to the Architecture for Faces, Color, Scenes, and Visual Field. J. Neurosci. 35, 6952–6968 (2015).
Lafer-Sousa, R. & Conway, B. R. Parallel, multi-stage processing of colors, faces and shapes in macaque inferior temporal cortex. Nat. Neurosci. 16, 1870–1878 (2013).
Levy, I., Hasson, U., Harel, M. & Malach, R. Functional analysis of the periphery effect in human building related areas. Hum. Brain Mapp. 22, 15–26 (2004).
Kay, K. N., Weiner, K. S. & Grill-Spector, K. Attention Reduces Spatial Uncertainty in Human Ventral Temporal Cortex. Curr. Biol. 25, 595–600 (2015).
Grill-Spector, K., Weiner, K. S., Kay, K. & Gomez, J. The Functional Neuroanatomy of Human Face Perception. Annu. Rev. Vis. Sci 3, 167–196 (2017).
Behrmann, M. & Plaut, D. C. Bilateral Hemispheric Processing of Words and Faces: Evidence from Word Impairments in Prosopagnosia and Face Impairments in Pure Alexia. Cereb. Cortex 24, 1102–1118 (2014).
Yovel, G., Tambini, A. & Brandman, T. The asymmetry of the fusiform face area is a stable individual characteristic that underlies the left-visual-field superiority for faces. Neuropsychologia 46, 3061–3068 (2008).
Landis, T. & Regard, M. Hemianopsie und Agnosie. Klin. Monbl. Augenheilkd 192, 525–528 (1988).
Pancaroglu, R. et al. Seeing the eyes in acquired prosopagnosia. Cortex 81, 251–265 (2016).
Bao, J. Y., Rubino, C., Taylor, A. J. G. & Barton, J. J. S. The effects of homonymous hemianopia in experimental studies of alexia. Neuropsychologia 70, 156–164 (2015).
Yovel, G., Levy, J., Grabowecky, M. & Paller, K. A. Neural Correlates of the Left-Visual-Field Superiority in Face Perception Appear at Multiple Stages of Face Processing. J. Cogn. Neurosci 15, 462–474 (2003).
Gilaie-Dotan, S., Saygin, A. P., Lorenzi, L. J., Rees, G. & Behrmann, M. Ventral aspect of the visual form pathway is not critical for the perception of biological motion. Proc. Natl. Acad. Sci. 112, E361–E370 (2015).
Gilaie-Dotan, S. et al. The role of human ventral visual cortex in motion perception. Brain 136, 2784–2798 (2013).
Previc, F. H. Functional specialization in the lower and upper visual fields in humans: Its ecological origins and neurophysiological implications. Behav. Brain Sci. 13, 519–542 (1990).
Silson, E. H., Chan, A. W.-Y., Reynolds, R. C., Kravitz, D. J. & Baker, C. I. A retinotopic basis for the division of high-level scene processing between lateral and ventral human occipitotemporal cortex. J. Neurosci. 35, 11921–11935 (2015).
Van Belle, G., de Graef, P., Verfaillie, K., Rossion, B. & Lefèvre, P. Face inversion impairs holistic perception: Evidence from gaze-contingent stimulation. J. Vis. 10, 1–13 (2010).
Farah, M. J., Tanaka, J. W. & Drain, H. M. What causes the face inversion effect? J. Exp. Psychol. Hum. Percept. Perform 21, 628–634 (1995).
Kovács, P., Knakker, B., Hermann, P., Kovács, G. & Vidnyánszky, Z. Face inversion reveals holistic processing of peripheral faces. Cortex 97, 81–95 (2017).
Rossion, B. Picture-plane inversion leads to qualitative changes of face perception. Acta Psychol. (Amst) 128, 274–289 (2008).
Haxby, J. V. Distributed and Overlapping Representations of Faces and Objects in Ventral Temporal Cortex. Science (80-) 293, 2425–2430 (2001).
Bonneh, Y. S., Sagi, D. & Polat, U. Local and non-local deficits in amblyopia: acuity and spatial interactions. Vision Res. 44, 3099–3110 (2004).
Bonneh, Y. S., Adini, Y. & Polat, U. Contrast sensitivity revealed by microsaccades. J. Vis. 15, 11 (2015).
Gilaie-Dotan, S. & Malach, R. Sub-exemplar Shape Tuning in Human Face-Related Areas. Cereb. Cortex 17, 325–338 (2007).
Gilaie-Dotan, S., Gelbard-Sagiv, H. & Malach, R. Perceptual shape sensitivity to upright and inverted faces is reflected in neuronal adaptation. Neuroimage 50, 383–395 (2010).
Martinez A, B. R. 1998. T. A. face database. CVC Technical Report nr 24. http://rvl1.ecn.purdue.edu/~aleix/aleix_ face_DB.html.
Yehezkel, O., Sterkin, A., Lev, M. & Polat, U. Crowding is proportional to visual acuity in young and aging eyes. J. Vis. 15, 23 (2015).
Lev, M. et al. Training improves visual processing speed and generalizes to untrained functions. Sci. Rep 4, 7251 (2015).
Cousineau, D. Confidence intervals in within-subject designs: A simpler solution to Loftus and Masson’s method. Tutor. Quant. Methods Psychol 1, 42–45 (2005).
Rossion, B. & Gauthier, I. How does the brain process upright and inverted faces? Behav. Cogn. Neurosci. Rev 1, 63–75 (2002).
Yovel, G. & Kanwisher, N. The Neural Basis of the Behavioral Face-Inversion Effect. Curr. Biol. 15, 2256–2262 (2005).
Rousselet, G. A., Husk, J. S., Bennett, P. J. & Sekuler, A. B. Spatial scaling factors explain eccentricity effects on face ERPs. J. Vis. 5, 755–763 (2005).
Rosenthal, G., Sporns, O. & Avidan, G. Stimulus Dependent Dynamic Reorganization of the Human Face Processing Network. Cereb. Cortex 27, 4823–4834 (2016).
Pitcher, D., Duchaine, B., Walsh, V., Yovel, G. & Kanwisher, N. The role of lateral occipital face and object areas in the face inversion effect. Neuropsychologia 49, 3448–3453 (2011).
Sadeh, B. & Yovel, G. Why is the N170 enhanced for inverted faces? An ERP competition experiment. Neuroimage 53, 782–789 (2010).
Gilaie-Dotan, S., Nir, Y. & Malach, R. Regionally-specific adaptation dynamics in human object areas. Neuroimage 39, 1926–1937 (2008).
Yovel, G. & Kanwisher, N. Face perception: domain specific, not process specific. Neuron 44, 889–898 (2004).
Rezlescu, C., Chapman, A., Susilo, T. & Caramazza, A. Large inversion effects are not specific to faces and do not vary with object expertise, https://doi.org/10.31234/osf.io/xzbe5 (2016).
Avidan, G., Hasson, U., Hendler, T., Zohary, E. & Malach, R. Analysis of the Neuronal Selectivity Underlying Low fMRI Signals. Curr. Biol. 12, 964–972 (2002).
Gerber, E. M., Golan, T., Knight, R. T. & Deouell, L. Y. Cortical representation of persistent visual stimuli. Neuroimage 161, 67–79 (2017).
Stigliani, A., Jeska, B. & Grill-Spector, K. Encoding model of temporal processing in human visual cortex. Proc. Natl. Acad. Sci. 114, E11047–E11056 (2017).
Carrasco, M. & Barbot, A. Spatial attention alters visual appearance. Curr. Opin. Psychol 29, 56–64 (2019).
Carrasco, M. Covert attention increases contrast sensitivity: psychophysical, neurophysiological and neuroimaging studies. in. Progress in Brain Research 154, 33–70 (2006).
Yeshurun, Y. & Carrasco, M. Spatial attention improves performance in spatial resolution tasks. Vision Res. 39, 293–306 (1999).
Hershler, O., Golan, T., Bentin, S. & Hochstein, S. The wide window of face detection. J. Vis. 10, 21–21 (2010).
Hershler, O. & Hochstein, S. At first sight: A high-level pop out effect for faces. Vision Res. 45, 1707–1724 (2005).
VanRullen, R. On second glance: Still no high-level pop-out effect for faces. Vision Res. 46, 3017–3027 (2006).
This work was supported by ISF grant no. 1485/18 to SGD. We thank Inbal Ziv, Oren Kadosh and Uri Polat for their assistance, the Computer Vision Laboratory, Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia, and the Secondary School Centre, Velenje, Slovenia, for allowing us to use the CVL Face Database (http://www.lrv.fri.uni-lj.si/facedb.html).
The authors declare no competing interests.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Kreichman, O., Bonneh, Y.S. & Gilaie-Dotan, S. Investigating face and house discrimination at foveal to parafoveal locations reveals category-specific characteristics. Sci Rep 10, 8306 (2020). https://doi.org/10.1038/s41598-020-65239-y
Vision as oculomotor reward: cognitive contributions to the dynamic control of saccadic eye movements
Cognitive Neurodynamics (2021)