Abstract
The center‐periphery visual field axis guides early visual system organization with enhanced resources devoted to central vision leading to reduced peripheral performance relative to that of central vision (i.e., behavioral eccentricity effect) for many visual functions. The center‐periphery organization extends to high‐order visual cortex where, for example, the well‐studied face‐sensitive fusiform face area (FFA) shows sensitivity to central vision and the place‐sensitive parahippocampal place area (PPA) shows sensitivity to peripheral vision. As we have recently found that face perception is more sensitive to eccentricity than place perception, here we examined whether these behavioral findings reflect differences in FFA's and PPA's sensitivities to eccentricity. We assumed FFA would show higher sensitivity to eccentricity than PPA would, but that both regions' modulation by eccentricity would be invariant to the viewed category. We parametrically investigated (fMRI, n = 32) how FFA's and PPA's activations are modulated by eccentricity (≤8°) and category (upright/inverted faces/houses) while keeping stimulus size constant. As expected, FFA showed an overall higher sensitivity to eccentricity than PPA. However, both regions' activation modulations by eccentricity were dependent on the viewed category. In FFA, a reduction of activation with growing eccentricity (“BOLD eccentricity effect”) was found (with different amplitudes) for all categories. In PPA however, qualitatively different BOLD eccentricity effect modulations were found (e.g., at 8° mild BOLD eccentricity effect for houses but a reverse BOLD eccentricity effect for faces and no modulation for inverted faces). Our results emphasize that peripheral vision investigations are critical to further our understanding of visual processing.
Keywords: BOLD eccentricity effect, eccentricity effect, face inversion, faces, FFA, FMRI, house inversion, parafovea, PPA, visual cortex
Here, we find with parafoveal stimulation that high‐order visual fusiform face area (FFA) and parahippocampal place area (PPA) are differently affected by eccentricity where PPA's modulation shows qualitative different BOLD eccentricity effects for different visual categories. Our results emphasize that peripheral vision investigations are critical to further our understanding of visual processing.
1. INTRODUCTION
The visual system devotes extensive resources to central vision relative to peripheral vision and this center‐periphery imbalance commences at the retina (Wässle et al., 1990) and continues into early visual cortex. The magnitude of this phenomenon was estimated in retinotopic area V1 and is termed the cortical magnification factor (CMF, Horton & Hoyt, 1991). Given these extensive resources dedicated to central vision, it is not surprising that many visual functions—from basic ones (Cowey & Rolls, 1974; Levi et al., 1985) to higher ones (Akselevich & Gilaie‐Dotan, 2022; Carrasco et al., 1995, 2003; Kreichman et al., 2020; Staugaard et al., 2016; Wolfe et al., 1998)—deteriorate significantly (performance‐wise) as eccentricity (distance from central vision) increases (the behavioral eccentricity effect (Carrasco et al., 1995; Wolfe et al., 1998; Xue et al., 2023)). It was also shown in some cases that when the CMF is accounted for (such that peripheral stimuli are proportionally enlarged to elicit similar cortical surface activation as is elicited by central stimuli), peripheral performance can reach central performance (Carrasco & Frieder, 1997; Cowey & Rolls, 1974; Jigo et al., 2023).
However, in natural vision there is no compensation for CMF and an object in peripheral vision occupies the same retinal space as when it appears in central vision. Furthermore, since different visual functions do not deteriorate at a uniform rate (Akselevich & Gilaie‐Dotan, 2022; Kreichman et al., 2020; Xue et al., 2023), additional task‐, attention‐, or function‐specific processes are likely to contribute to the eccentricity‐based reductions. We have recently found that face discrimination is more sensitive to eccentricity and deteriorates faster with eccentricity in the parafovea (≤4°) than house discrimination and that these different visual categories are likely to be supported by dissociated mechanisms (Kreichman et al., 2020). Following these results, here we assumed that this perceptual difference may be attributed to regions that are highly sensitive to these categories in high‐order ventral visual cortex. Two of these regions are among the most investigated in high‐order visual cortex: The “fusiform face area” (FFA (Gilaie‐Dotan et al., 2008; Grill‐Spector & Malach, 2004; Kanwisher et al., 1997; McCarthy et al., 1997)) sensitive to faces and also known to show a foveal bias (Hasson et al., 2002; Levy et al., 2001), and the “parahippocampal place area” (PPA (Epstein et al., 1999; Epstein & Kanwisher, 1998)) sensitive to places and known to show preference to peripheral stimuli (Hasson et al., 2002; Levy et al., 2001, 2004; Malach et al., 2002). We were interested to parametrically assess FFA's and PPA's activation modulation by eccentricity for their preferred (faces in FFA, houses in PPA) and non‐preferred (houses in FFA, faces in PPA) categories. We were also interested to examine how these regions' activations are affected by eccentricity to inverted stimuli (Epstein et al., 2006; Kreichman et al., 2020).
To that end, we ran an fMRI study (n = 32) where we independently localized FFA and PPA and then tested how each of these regions' activation was affected by eccentricity in the parafovea (≤8°) for upright and inverted faces and houses when participants performed category‐insensitive tasks (two separate experiments with different cohorts of participants). To tap into the veridical sensitivity of these regions to eccentricity and parametrically examine eccentricity‐based sensitivity across the visual field, we kept stimulus size constant (not compensating for the CMF).
We assumed that both FFA and PPA would show eccentricity‐based activation reductions (which we refer to here as the “BOLD eccentricity effect”) but each with a different rate (i.e., a different region‐specific magnitude of BOLD eccentricity effect). Specifically, we hypothesized FFA would show a greater BOLD eccentricity effect (steeper reductions of activation with growing eccentricity) than PPA. In addition, we also assumed that each region would show its own “characteristic” BOLD eccentricity effect that would be category‐independent given previous studies, indicating (i) region‐specific processing that shows category dependency predominantly with respect to activation magnitude (Gilaie‐Dotan et al., 2008; Grill‐Spector, Weiner et al., 2017; Weiner et al., 2017) and (ii) that PPA showed a comparable peripheral bias for both faces and houses (Levy et al., 2001, 2004). Therefore, here we analyzed activation magnitudes and how they were modulated by eccentricity and category preference in FFA and PPA.
2. MATERIALS AND METHODS
2.1. Participants
A total of 32 healthy participants (20 women, aged 18–43 years, mean age 28.6 years ±5.5 SD, 29 right‐handed, see more details in Table S1 at https://osf.io/m8czv/ (Kreichman & Gilaie‐Dotan, 2023)) participated in two fMRI studies (16 participants in upright and inverted “Count20” experiments; 16 participants in upright “DBLstml” experiment). Sample sizes were determined based on earlier studies investigating face‐house sensitivities and eccentricity effects using similar cohort sizes (Levy et al., 2001: n 1 = 13, n 2 = 5, n 3 = 6 participants; Hasson et al., 2002: n 1 = 11, n 2 = 6 participants; Levy et al., 2004: n 1 = 11, n 2 = 5, n 3 = 8 participants; Kanwisher et al., 1998: n 1 = 20, n 2 = 5, n 3 = 5; Epstein & Kanwisher, 1998: n 1 = 10, n 2 = 6, n 3 = 5). All participants reported having normal or corrected‐to‐normal vision and provided written informed consent to participate in the fMRI experiments before the experiments began. The Tel‐Aviv Sourasky Medical Center ethics committee approved the experimental protocol.
2.2. MRI setup
Participants were scanned in a Siemens Magnetom Prisma 3 T scanner equipped with a standard 20‐channel head coil at the Weizmann Institute of Science, Rehovot, Israel. Blood oxygenation level‐dependent (BOLD) contrast for the functional scans was obtained with gradient‐echo echo‐planar imaging (EPI) sequence (face/house experiments: TR = 2500 ms, TE = 30 ms, flip angle = 80°, field of view (FoV) = 216 × 216 mm2, matrix size = 72 × 72, 40 axial slices of 3 mm thickness (no gap) with an in‐plane resolution of 3 × 3 mm2, covering the entire cortex; category localizer experiment: TR = 3000 ms, TE = 30 ms, flip angle = 85°, FoV = 210 × 210 mm2, matrix size = 70 × 70, 48 axial slices of 3 mm thickness (no gap) with an in‐plane resolution of 3 × 3 mm2, covering the entire brain). A high‐resolution whole‐brain anatomical T1‐weighted magnetization prepared rapid acquisition gradient‐echo (MPRAGE) sequence was acquired for each participant (TR = 2300 ms, TE = 2.32 ms, flip angle = 8°, FoV = 240 × 240 mm2, 192 slices of 0.9 mm thickness with no gap, 1 × 1 mm2 in‐plane resolution) to allow accurate cortical segmentation, reconstruction, and volume‐based statistical analysis.
Stimuli were generated on a Windows 7 Enterprise PC and projected using an LCD projector onto a screen (31 cm width × 35 cm height) at the back end of the MRI tunnel with a viewing distance of 110 cm; scanner room was darkened during the experiments. Screen was viewed through a tilted mirror positioned over the participant's forehead. Face/house experiments were run using an in‐house developed platform for psychophysical and eye‐tracking experiments (PSY) developed by Yoram S. Bonneh (Bonneh et al., 2015; Kreichman et al., 2020; Masarwa et al., 2022) running on a Windows PC, and the category localizer experiment (Grossman et al., 2019) was run using “Presentation” software (Neurobehavioral Systems, Inc.). In addition, eye movements were recorded during face/house experiments using an “Eyelink−1000,” an MR‐compatible eye‐tracker (SR Research, Ontario, Canada) with a sampling rate of 500 Hz equipped with a long‐range lens. Eye movements of the right eye were recorded. For behavioral performance analyses, button presses were recorded using a Fiber Optic Response Pad (fORP) 4‐Button Curve Right response box (Current Designs, Philadelphia, USA).
2.3. fMRI procedure and experimental design
Before beginning the experimental scans, each participant performed a standard 3‐ or 5‐point eye‐tracker calibration within the scanner (for five participants due to technical reasons, we used previously stored calibration). The upright face/house experiment was always performed first, then the category localizer and anatomical scans (their order could be switched), and the inverted experimental runs were always performed last. All face/house experiments were block design, each block presenting images of one category (face/house) and one eccentricity (0°, 4°, or 8°). Participants were debriefed about the experiments after the MRI scan.
2.3.1. Main face/house experiment (upright “Count20”)
Stimuli
All face and house images were grayscale images presented on a black background. Face images subtending ∼1.9°× ∼ 2.3° (width× height) were full‐front photographs of men with a neutral expression (taken from an earlier study (Gilaie‐Dotan et al., 2010; Gilaie‐Dotan & Malach, 2007) that modified images from 2 databases (CVL Face Database, n.d. [http://www.lrv.fri.uni-lj.si/facedb.html]; AR Face Database [Martinez, 1998]). House images (as those used in Kreichman et al., 2020) were photographs of real houses and subtended ∼2°× ∼ 2.5° (width × height). Overall, we used 10 different face and 10 different house images. Due to screen limitations of the MRI setup, face and house images centered at 8° on the vertical meridian (at the topmost and bottommost locations, see a video demonstration of the stimuli at https://osf.io/m8czv/) appeared cropped such that only the lower half of images presented at the upper part of the screen and only the upper half of the images presented at the lower part of the screen were visible (overall 6 out of 24 images during long blocks and 4 out of 16 images during short blocks of the 8° conditions appeared cropped; images at all other locations appeared in full size).
Procedure
Each run started with a 20‐second fixation block followed by a block of 15 presented textures (200 ms, 300 ms ISI, grayscale noisy Mondrian‐like; since the first block is sometimes accompanied by a strong fMRI BOLD signal, this block was meant to account for it independently of our conditions of interest), after which another 7.5 s fixation block appeared, and then, the main face/house image conditions were presented. The run ended with a 12‐s fixation block. The experiment was a 2 categories × 3 eccentricities block design experiment where each block included stimuli of one condition (one visual category (upright faces or houses) appearing at one eccentricity (center, 4° or 8°)). Each block type was repeated twice in each run (12 blocks per run), and block order was pseudo‐random. Nine blocks presented 24 images (12 s block) and 3 pseudo‐randomly chosen blocks presented 16 images (8 s block) that were both aligned with the TR sampling rate and allowed measuring behavioral performance (16 and 24 images per short/long block were chosen to best match 3 and 4 TR durations given our stimulus presentation parameters). This mixture of different block lengths was meant to increase attention/engagement while reducing predictability of the required response. Importantly, while overall hemodynamic response function (HRF) varies with block duration, our study focused on the activation magnitude that is measured around the HRF peak and occurs at the earlier stages of the block where there are no expected short vs long block HRF differences. Images in each block were presented consecutively (200 ms/image, 300 ms ISI); in the central conditions, images appeared in the same location (screen center), and in the 4° and 8° conditions, images appeared in eight locations of the block's eccentricity (in each quadrant and on the horizontal and vertical meridians, see Figure 1) in a pseudo‐random order. Images could be repeated several times during a block, and same images appeared across blocks. A white fixation circle was present throughout the duration of the run, and at the end of each block, the fixation circle turned green for 2500 ms to indicate the block‐ended and response was expected. Between blocks, the white fixation circle continued to be present for 6 or 5 s (depending on whether it followed a long or a short block). Participants were instructed to fixate throughout the experiment, be aware of the surrounding area in which the targets could appear, and report at the end of each block (when the fixation circle turned green) whether there were more or less than 20 images displayed in that block (by pressing the right or the left buttons on the response box, respectively); no feedback was given (see Figure 1 for experimental paradigm and Figure 2 for behavioral performance). Each participant underwent two runs of the experiment, and each run was of a different version of the experiment (with reversed block order between versions and different conditions chosen for the short blocks). Version order was counterbalanced across participants. Each run took 4 min 35 s.
FIGURE 1.
Experimental design of parafoveal face/house fMRI experiments. (a) Representative parafoveal upright face‐block timeline. Each condition block (one category (faces or houses) at one eccentricity (0°, 4°, or 8°)) sequentially presented 16 or 24 images (200 ms/image, 300 ms ISI, white fixation present throughout) in pseudo‐random order (see (c)), and ended with a green fixation (2.5 s) indicating that response was expected (in “Count20” to report whether more or less than 20 images appeared, in “DBLstml” to report if 0–3 image pairs appeared, see Methods). Participants were instructed to keep fixation and be aware of the surrounding area in which the targets could appear. Each condition block was repeated twice in each experiment. (b) Representative stimuli for the upright and inverted “Count20” and “DBLstml” experiments and experimental stimuli locations (8 locations/parafoveal eccentricity). Inverted experiments had identical design with inverted stimuli. (c) Representative short block order (16 stimuli, “DBLstml” experiment, expected response: 1 pair).
FIGURE 2.
Behavioral performance by eccentricity and category for all experiments. Accuracy in (a) upright “Count20” experiment (n = 12), (b) “DBLstml” experiment (n = 11), and (c) inverted “Count20” experiment (n = 10). Face conditions in shades of red to yellow, house conditions in shades of blue; lighter shades represent growing eccentricity. No differences between face and house accuracy were found across eccentricities and experiments. Behavior was almost at ceiling across eccentricities apart from the 8° long blocks (see Methods and Results). Error bars represent SEM. See Table 2 for further details.
2.3.2. Control face/house experiment 1 (upright “DBLstml”)
To control the possibility that the results of the upright “Count20” experiment were task‐dependent, we ran an additional experiment with upright faces and houses on a different group of participants (“DBLstml” experiment). This experiment was identical in its design and stimuli to the upright “Count20” experiment. Since we employed here a different task (see below), for several of the stimuli an additional same‐category stimulus was added such that there were instances of two images appearing simultaneously on two opposing sides of fixation (see an example sequence in Figure 1 and a video demonstration of the stimuli at https://osf.io/m8czv/) and this double stimulus could happen once, twice, or three times in a block (double stimulus occurred in half of the blocks (6 of 12)). The participants were not notified of the frequency of these pairs across the experiment and were instructed to fixate and be aware of the surrounding area in which the targets could appear, and report whether there were 0, 1, 2, or 3 simultaneous image pairs in a block (by pressing the corresponding button in the 4‐button response box). This method of simultaneous image‐pair presentation was also meant to facilitate fixation throughout the experimental session (we followed the logic that looking at the center of the screen shall lead to best performance and making saccades was likely to result in missing image pairs). There were two versions of the experiment differing in block order and number of paired‐images stimuli in each block, and each participant ran each version once. In the first version, there were 2, 4, 1, and 1 blocks with 0, 1, 2, and 3 image pairs, respectively; in the second version, there were 2, 3, 2, and 1 blocks with 0, 1, 2, and 3 image pairs, respectively (resulting in a total of 4 blocks with no image pairs, 7 blocks with 1 image pair, 3 blocks with 2 images pairs, and 2 blocks with 3 image pairs). This choice of non‐uniformly distributed block types (pseudo‐random combinations for each block and within each version skewed to one image pair per block) was aimed at keeping the participants' engagement high across the experiment (by reducing response anticipation) while ensuring that for most of the blocks, we were measuring activation to single peripheral stimulus at each time point (11 of 16 peripheral blocks included 0 or 1 image pairs, and throughout all blocks most of the time a single stimulus was presented). Images used in this experiment were the same as those used in the “Count20” experiment. Each run took 4 min and 35 s.
2.3.3. Control face/house experiment 2 (inverted “Count20”)
Experimental design was identical to that used in the upright “Count20” paradigm except that stimuli were inverted upside down.
2.3.4. Category localizer experiment
Each participant underwent a blocked design visual category localizer experiment (Berkovich‐Ohana et al., 2020; Gilaie‐Dotan et al., 2008, 2009, 2010, 2013) to identify the cortical regions preferentially activated by specific visual categories (faces, houses, objects, body parts, or patterns (Grossman et al., 2019); all images were different than those used in the main experiment, for more details see https://osf.io/m8czv/). Each block lasted 9 s and included 9 images (6° × 6°, 800 ms/image, 200 ms ISI) that were all from the same visual category presented on a gray background. Each block was followed by a 6 second blank screen. Blocks of each category were repeated seven times in a pseudo‐random order across the experiment. The experiment lasted 550 s. The first and last blocks of the experiment were fixation blocks that lasted 21 and 9 s, respectively. A central fixation point was present throughout the experiment. Participants were instructed to perform a 1‐back memory task and report via button presses whether the presented stimulus was identical to or different than the previous stimulus. Image repetition occurred once or twice in each block.
2.3.5. fMRI data preprocessing and analysis
fMRI data were analyzed with the BrainVoyager software version 21.4 (Brain Innovation, Maastricht, The Netherlands). The first 2 TRs (volumes) of each functional scan were discarded. Preprocessing of functional scans included 3D motion correction, slice scan time correction, linear trend removal, and high‐pass filtering (filtering out frequencies lower than three cycles across the experiment). Functional data were incorporated into 3D normalized MNI space where all further analyses were performed.
Exclusions performed on the data per participant (see https://osf.io/m8czv/): runs with head motion larger than 1 mm in 1 of the 6 motion directions in both runs were excluded from the analyses (participants with one run excluded based on head motion: upright “Count20”: n = 1, inverted “Count20”: n = 1, “DBLstml”: n = 2) and participants whose both upright runs were excluded based on head motion were discarded from the study (upright “Count20”: n = 2; “DBLstml”: n = 1). On top of this, three additional participants had both of their inverted “Count20” runs excluded based on head motion (so their upright runs but not their inverted runs were included). Four additional participants underwent only 1 run in the upright “DBLstml” experiment. In addition, one participant was excluded due to chance‐level behavioral results during central blocks of upright “DBLstml” experiment. An additional participant's data of one run of upright “Count20” exp. could not be analyzed due to technical issues.
2.3.6. ROI localization
Functional regions of interest (ROIs) in the right and left hemispheres were defined individually for each participant using the category localizer experiment (Gilaie‐Dotan et al., 2008, 2009, 2010). Individual participant category localizer experiment data (Hasson et al., 2003; Levy et al., 2001) were fitted to a general linear model (Friston et al., 1994) implemented in BrainVoyager. This was done separately to the time course of each individual voxel according to the experimental protocol. The model coefficients for each voxel were determined so that the error term between the model's prediction and the measured voxel time course was minimized (based on default settings of BrainVoyager). The t‐test between coefficients of different conditions was applied to determine the voxel's activation pattern. Face‐sensitive area in the lateral fusiform gyrus (“FFA” (Kanwisher et al., 1997; McCarthy et al., 1997)) was defined by preferential activation to faces relative to houses (contrast: faces > houses, all p's < .003, t(177)'s ≥ 3.11 at the voxel level, uncorrected), and place‐sensitive area in the collateral sulcus (CoS) and parahippocampal gyrus (“PPA” (Epstein & Kanwisher, 1998)) was defined by preferential activation to houses relative to faces (contrast: houses > faces, all p's < .003, t(177)'s ≥ 3.11 at the voxel level, uncorrected), see Figure 3 (statistical thresholds follow common practices in the field, see (Chen et al., 2023)). We made sure that the y MNI coordinates (posterior–anterior dimension) of FFA and of PPA were between −35 and −60 (when there were two FFA foci, we sampled the posterior one). In some participants, some ROIs could not be delineated in one of the hemispheres since they did not show the expected category selectivity (see Table S2 at https://osf.io/m8czv/ for a detailed list). For two participants with missing category localizer experimental data (1 was excluded due to head movements, 1 did not complete the localizer experiment), the localization of face‐ and place‐related regions was based on contrasting the central face and central house blocks of our main fMRI experiment (upright face/house experiment; FFA by faceCenter > houseCenter, PPA by houseCenter > faceCenter) in combination with the expected anatomical locations. Table 1 presents average MNI coordinates and size (number of voxels) across participants for each ROI (see Table S2 at https://osf.io/m8czv/ for individual participants' ROI MNI coordinates). Details about an additional external localization of the FFA and PPA with bigger images that was used in complementary analyses is available at https://osf.io/m8czv/. Note that new algorithmic‐based methods have also been developed recently to localize regions of interest in ventral visual pathway (Julian et al., 2012).
FIGURE 3.
Fusiform face area (FFA) and parahippocampal place area (PPA) independent localization (10 representative participants). Demonstration of individually defined FFA (in orange) and PPA (in blue) by the standard category localizer experiment (n = 5 from “Count20” experiment; n = 5 from “DBLstml” experiment; coronal views; see Methods). R/L ‐ right/left hemisphere.
TABLE 1.
Independently defined FFA and PPA ROI details for each of the upright experiments.
| Average | SE | n | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| X | Y | Z | Voxel count | X | Y | Z | Voxel count | ||||
| Participants from “Count20” exp. (n = 14) | FFA | L | −41.08 | −51.78 | −19.44 | 431.25 | 1.02 | 1.50 | 1.14 | 73.93 | 12 (10) |
| R | 41.94 | −53.10 | −17.32 | 713.92 | 0.82 | 1.52 | 0.99 | 90.61 | 13 (10) | ||
| PPA | L | −28.62 | −50.86 | −10.37 | 1513.84 | 0.48 | 0.94 | 0.58 | 243.59 | 13 (10) | |
| R | 29.64 | −50.10 | −10.08 | 1703.07 | 0.77 | 1.06 | 0.67 | 352.25 | 14 (11) | ||
| Participants from “DBLstml” exp. (n = 14) | FFA | L | −38.90 | −50.81 | −19.04 | 612.25 | 1.02 | 1.48 | 0.63 | 88.17 | 12 |
| R | 39.94 | −50.96 | −17.41 | 602.42 | 0.93 | 1.18 | 0.83 | 97.17 | 14 | ||
| PPA | L | −27.45 | −50.05 | −10.54 | 1115.21 | 0.61 | 0.73 | 0.73 | 141.25 | 14 | |
| R | 27.87 | −48.74 | −9.08 | 1350.50 | 0.48 | 0.64 | 0.61 | 169.07 | 14 | ||
Note: For both the “Count20” experimental cohort (n = 14, top) and the “DBLstml” cohort (n = 14, bottom), MNI coordinates and number of voxels are presented for each ROI (average and standard error across participants). n represents number of participants that had each ROI defined by the standard external localizer (see Methods) after exclusion; in parentheses is the number of participants that also underwent the inverted “Count20” experiment (note that the group mean ROI coordinates presented here were defined solely based on the upright experiment cohort). R/L – right/left hemisphere (for a detailed list by participant see https://osf.io/m8czv/).
2.3.7. ROI analysis
For each participant, event‐related time course of activation from each of the main experiments (upright “Count20”, inverted “Count20”, and “DBLstml”) was based on both runs each participant underwent (two versions of the main experiment). For each participant and each ROI (FFA‐R, FFA‐L, PPA‐R, and PPA‐L), we sampled the average event‐related time course (upright “Count20”: Figure 4a; inverted “Count20” and “DBLstml”: Figure 5a,e) after the experimental protocol was fit to the experimental data (all voxels; all experimental blocks were modeled to reduce the unexplained variance). The peak of the hemodynamic response function (HRF) was taken as the average response during TRs 3 and 4. The normalized response was defined as the response of each condition divided by the response of the ROI's preferred category at 0° (Figures 4b and 5b). The slope of the response in each ROI was defined as the change in peak response by eccentricity between central 0° and 4°, and between central 0° and 8°, and this was calculated for the preferred and for the non‐preferred categories separately (Figure 4c,d and 5c,d,g,h). Eccentricity‐specific category selectivity was calculated by subtracting the response of the non‐preferred category from that of the preferred category at each eccentricity (i.e., for FFA: faceCenter vs houseCenter, face‐4° vs. house‐4°, face‐8° vs. house‐8°; for PPA: face conditions were subtracted from house conditions, Figure 7).
FIGURE 4.
Fusiform face area (FFA) and parahippocampal place area (PPA) activation modulation in the parafovea to preferred and non‐preferred categories (upright “Count20” experiment, n = 14). (a) Mean activation (% signal change, y‐axis) for each experimental condition as a function of time (by TRs on the x‐axis) in each region of interest (ROI) across participants (FFA‐L (n = 12), FFA‐R (n = 13), PPA‐L (n = 13), PPA‐R (n = 14)), faces in dark to light orange with growing eccentricity, houses in dark to light blue with growing eccentricity. Category selectivity in central vision is evident in all ROIs by higher activation to the central preferred versus central non‐preferred category. (b) Peak responses (mean of 3rd and 4th TRs from (a)) of each condition relative to that of central preferred category were calculated (normalized response); black borders demarcate each ROI's activation to it central preferred category. (c, d) Response modulation with eccentricity (slopes from 0° to 4° (c) and from 0° to 8° (d), see Methods) for preferred and non‐preferred categories for each ROI; a negative slope represents the BOLD eccentricity effect (activation reduction with growing eccentricity), a positive slope represents a reverse BOLD eccentricity effect, and a zero slope represents no modulation of the activation by eccentricity. Dots on each bar represent individual data for that region and that category. At 4°, both FFA and PPA showed a BOLD eccentricity effect for their preferred categories and much‐reduced eccentricity effects (for PPA no effect) for their non‐preferred category. At 8°, a qualitative difference emerged where FFA showed a BOLD eccentricity effect for both categories (albeit to different extents), but PPA showed qualitatively different modulations with a typical BOLD eccentricity effect for its preferred category but a reverse effect (positive slope reflecting a reverse (negative) BOLD eccentricity effect) for its non‐preferred category. Error bars represent SEM across participants.
FIGURE 5.
Fusiform face area (FFA) and parahippocampal place area (PPA) activation modulation in the parafovea to preferred and non‐preferred categories in control experiments (“DBLstml”, n = 14; inverted “Count20”, n = 11). Notations as in Figure 4 with (a–d) for “DBLstml” results (FFA‐L (n = 12), FFA‐R (n = 14), PPA‐L (n = 14), PPA‐R (n = 14)) and (e–h) for inverted “Count20” (FFA‐L (n = 10), FFA‐R (n = 10), PPA‐L (n = 10), PPA‐R (n = 11)). Since inverted stimuli are not considered optimal for FFA and PPA, in (f) peak responses are not normalized. Note that at 4° in (c), similar to the “Count20” results, both FFA and PPA show a BOLD eccentricity effect for their preferred categories but no eccentricity effect for their non‐preferred category. Furthermore, at 8° in (d), similar to the “Count20” results (Figure 4), in the “DBLstml” experiment – while FFA showed the BOLD eccentricity effect (evident by negative slopes) for both preferred and non‐preferred categories, PPA showed quantitatively different eccentricity‐based modulations by viewed category (negative slope for its preferred (houses) category versus a positive slope to its non‐preferred (faces) category (see Results)). Also, in the inverted experiment at 8°, we found that while FFA showed similar effects as for upright stimuli (f), for PPA there was no modulation by eccentricity to inverted faces. Error bars represent SEM.
FIGURE 7.
Fusiform face area (FFA) and parahippocampal place area (PPA) category selectivity modulation by eccentricity. (a) Upright “Count20” (n = 14). (b) “DBLstml” (n = 14). Eccentricity demarcated by shades of black (central) to gray (parafovea). In both experiments, we found category selectivity was highest for central stimuli and decreased with eccentricity in all ROIs. However, in both FFA and PPA, category selectivity modulations in the parafoveal were not consistent across experiments. Error bars represent SEM.
Additional supplementary ROI analyses based on another localization experiment with bigger images are available at https://osf.io/m8czv/.
2.3.8. Fixation assessment
Since our experiment aimed to evaluate activation to peripheral stimuli, eye movements were recorded to validate fixation was kept throughout the experimental sessions. Fixation analysis was performed offline. To evaluate participant's fixation performance, for each block, we calculated the percentage of time that the participant's gaze was within 1.5° from the fixation point (inter‐stimulus intervals were not included in this analysis). We then computed a fixation score for each eccentricity by averaging across four blocks of each eccentricity (taking into account blocks of faces and of houses together), and this was done separately for each run. We defined good fixators as participants that kept fixation at least 80% of the time during the 8° blocks. We evaluated the effect of fixation on activation by comparing activation of participants with good versus poor fixations (see Tables S1 and S3 at https://osf.io/m8czv/).
2.3.9. Behavioral performance analysis
During each experiment, participants were asked to keep fixation throughout the blocks and perform the behavioral task while being aware of the peripheral area in which stimuli could appear. Both behavioral tasks were aimed to keep participants attentive during the experiments while not demanding category‐specific attention.
The partial visibility of the 8° images on the vertical meridian (extending beyond the visible visual display, see “Main face/house experiment (upright ‘Count20’)” in Methods) may have influenced performance in the long blocks of the “Count20” experiment, so we mostly relied on the short blocks' performance for 8° blocks.
In the upright “Count20”, “DBLstml”, and inverted “Count20” experiments, there were 2, 3, and 1 participants (respectively) whose responses were not recorded due to technical issues resulting in behavioral results reported for n = 12 (upright “Count20”), n = 11 (“DBLstml”), and n = 10 (inverted “Count20”); see Figure 2 and Table 2. Note that one participant in the upright “Count20” experiment did not respond to the central conditions, and for one participant in the upright “Count20” exp. and 2 participants in the “DBLstml” exp., only seven out of eight repetitions were available for the central condition. See https://osf.io/m8czv/ for more details.
TABLE 2.
Group behavioral performance by eccentricity and category for upright “Count20”, “DBLstml”, and inverted “Count20” experiments.
| Upright “Count20” (n = 12) | “DBLstml” (n = 11) | Inverted “Count20” (n = 10) | |||
|---|---|---|---|---|---|
| Center | Faces | 97.7 ± 2.3 | 95.5 ± 4.5 | 97.5 ± 2.5 | |
| Houses | 95.5 ± 2.9 | 95.5 ± 3 | 100 ± 0 | ||
| 4° | Faces | 100 ± 0 | 95.5 ± 4.5 | 100 ± 0 | |
| Houses | 95.8 ± 4.2 | 95.5 ± 4.5 | 97.5 ± 2.5 | ||
| 8° | Short blocks | Faces | 100 ± 0 | 100 ± 0 | |
| Houses | 83.3 ± 11.2 | 100 ± 0 | |||
| Long blocks | Faces | 19.4 ± 9.6 | 10 ± 10 | ||
| Houses | 16.7 ± 8.7 | 0 ± 0 | |||
| All blocks | Faces | 37.5 ± 7.8 | 75.0 ± 7.5 | 35.0 ± 7.6 | |
| Houses | 33.3 ± 7.7 | 68.2 ± 7.6 | 22.5 ± 2.5 | ||
Note: Average accuracy (% correct) ± SEM are presented. See Figure 2, “Behavioral performance analysis” in Methods and https://osf.io/m8czv/ for further details.
2.4. Statistical analyses
Statistical analyses (ANOVA, post‐hoc, and Wilcoxon signed‐rank tests) were performed using R studio (version 2021.9.0.351, RStudio Team, 2020). Repeated‐measures ANOVAs were performed using the R get_anova_table() function that automatically applies the Greenhouse–Geisser sphericity correction only to factors violating the sphericity assumption (i.e., Mauchly's test significant p‐value ≤.05). One‐sample, two‐sided Wilcoxon signed‐rank tests were always preformed relative to the zero vector.
3. RESULTS
3.1. Behavioral performance
We deliberately chose a behavioral task that would near ceiling performance so that task difficulty would be comparable across eccentricities and visual categories and thus will allow investigating FFA's and PPA's sensitivity to visual categories regardless of task modulations. Accuracy for the central and 4° conditions was close to ceiling in all experiments (Figure 2 and Table 2). In the 8° condition (“Count20” experiment, upright, and inverted), we found that the performance for the short blocks alone was high (almost as for central and 4° blocks), but for the long (3/4 of the blocks) and short (1/4 of the blocks) blocks combined it was below chance (and even lower for long blocks alone, see Table 2). This reduction in performance in the long blocks can be attributed to the partial visibility of the 8° stimuli on the vertical meridian (potentially leading to missing these stimuli and thus not counting them which resulted in an erroneous response in the long blocks; see Figure 2 and Table 2; full details available online at https://osf.io/m8czv/). Importantly, performance appeared to be comparable for the face and house conditions at all eccentricities.
3.2. FFA's and PPA's activation modulation by eccentricity
3.2.1. Main experiment: Upright “Count20” experiment
FFA and PPA were independently localized using an external localizer (see Figure 3) with anatomical locations (face‐selective areas in the lateral fusiform and house‐selective areas in the collateral sulcus) and coordinates (see Table 1 and Table S2 at https://osf.io/m8czv/) comparable to those found in earlier studies (e.g., Hasson et al., 2002; Ishai et al., 1999). Afterward, the experimental time courses (upright “Count20” experiment, n = 14) of each of our four ROIs (FFA‐R/L, PPA‐R/L) were analyzed. We first verified that these ROIs showed the expected category selectivity for the central experimental stimuli in our experiment as expected from earlier studies (e.g., Gilaie‐Dotan et al., 2008), and indeed, this is what we found for each of the ROIs (paired, two‐sided Wilcoxon signed‐rank test: right/left FFA: faceCenter > houseCenter: p's < 10−3, in right/left PPA: houseCenter > faceCenter: p's < 10−3, see Figures 4 and 5 and Table 3). Furthermore, as can be seen in Figures 4a and 5a, activation levels to the preferred categories of each of the ROIs (for faces in FFA‐R and FFA‐L, for places in PPA‐R and PPA‐L) appeared to decrease with eccentricity as would be expected from the behavioral eccentricity effect (Carrasco et al., 1995; Wolfe et al., 1998).
TABLE 3.
Complementary statistical comparisons.
| Experiment | Statistics performed on | Comparison (contrast) | Test purpose | Result | Test performed |
|---|---|---|---|---|---|
| Upright “Count20” exp. (n = 14) | Peak response | FFA‐L: FaceCenter, HouseCenter | Testing typical category selectivity | p = 4 × 10 −4 | Paired Wilcoxon Signed‐Rank (2‐sided) |
| FFA‐R: FaceCenter, HouseCenter | p = 2 × 10 −4 | ||||
| PPA‐L: FaceCenter, HouseCenter | p = 2 × 10 −4 | ||||
| PPA‐R: FaceCenter, HouseCenter | p = 1 × 10 −4 | ||||
| PPA‐L: FaceCenter, Face‐8° | Testing increase in activation for 8° vs. center | p = 4 × 10 −4 | |||
| PPA‐R: FaceCenter, Face‐8° | p = 1 × 10 −4 | ||||
| PPA‐L: FaceCenter | Testing activation for FaceCenter < baseline | p = .008 | One sample Wilcoxon Signed‐Rank (2‐sided) | ||
| PPA‐R: FaceCenter | p = 2 × 10 −4 | ||||
| PPA‐L: Face‐8° | Testing activation for Face‐8° > baseline | p = .008 | |||
| PPA‐R: Face‐8° | p = .04 | ||||
| Slope 0°–4° | FFA‐L: Faces | Test preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .002 | One sample Wilcoxon Signed‐Rank (2‐sided) | |
| FFA‐R: Faces | p = 2 × 10 −4 | ||||
| PPA‐L: Houses | p = 4 × 10 −4 | ||||
| PPA‐R: Houses | p = 1 × 10 −4 | ||||
| FFA‐L: Houses | Test non‐preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .006 | |||
| FFA‐R: Houses | p = .047 | ||||
| PPA‐L: Faces | p = .83 | ||||
| PPA‐R: Faces | p = .90 | ||||
| Slope 0°–8° | FFA‐L: Faces | Test preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = 4 × 10 −4 | One sample Wilcoxon Signed‐Rank (2‐sided) | |
| FFA‐R: Faces | p = 2 × 10 −4 | ||||
| PPA‐L: Houses | p = .02 | ||||
| PPA‐R: Houses | p = 2 × 10 −4 | ||||
| FFA‐L: Houses | Test non‐preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = 4 × 10 −4 | |||
| FFA‐R: Houses | p = .001 | ||||
| PPA‐L: Faces | p = 4 × 10 −4 | ||||
| PPA‐R: Faces | p = 1 × 10 −4 | ||||
| “DBLstml” exp. (n = 14) | Peak response | FFA‐L: FaceCenter, HouseCenter | Testing typical category selectivity at central vision | p = 4 × 10 −4 | Paired Wilcoxon Signed‐Rank (2‐sided) |
| FFA‐R: FaceCenter, HouseCenter | p = 1 × 10 −4 | ||||
| PPA‐L: FaceCenter, HouseCenter | p = 1 × 10 −4 | ||||
| PPA‐R: FaceCenter, HouseCenter | p = 1 × 10 −4 | ||||
| PPA‐L: FaceCenter, Face‐8° | Testing increase in activation for 8° vs. center | p = .024 | |||
| PPA‐R: FaceCenter, Face‐8° | p = .020 | ||||
| PPA‐L: FaceCenter | Testing activation for FaceCenter < baseline | p = .005 | One sample Wilcoxon Signed‐Rank (2‐sided) | ||
| PPA‐R: FaceCenter | p = .006 | ||||
| PPA‐L: Face‐8° | Testing activation for Face‐8° > baseline | p = .807 | |||
| PPA‐R: Face‐8° | p = .760 | ||||
| Slope 0°–4° | FFA‐L: Faces | Test preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = 4 × 10 −4 | One sample Wilcoxon Signed‐Rank (2‐sided) | |
| FFA‐R: Faces | p = 1 × 10 −4 | ||||
| PPA‐L: Houses | p = .057 | ||||
| PPA‐R: Houses | p = .008 | ||||
| FFA‐L: Houses | Test non‐preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .26 | |||
| FFA‐R: Houses | p = .24 | ||||
| PPA‐L: Faces | p = .54 | ||||
| PPA‐R: Faces | p = .90 | ||||
| Slope 0°–8° | FFA‐L: Faces | Test preferred category Slope ≠ 0 (i.e. BOLD eccentricity effect) | p = 4 × 10 −4 | One sample Wilcoxon Signed‐Rank (2‐sided) | |
| FFA‐R: Faces | p = 1 × 10 −4 | ||||
| PPA‐L: Houses | p = .02 | ||||
| PPA‐R: Houses | p = .002 | ||||
| FFA‐L: Houses | Test non‐preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .002 | |||
| FFA‐R: Houses | p = .004 | ||||
| PPA‐L: Faces | p = .02 | ||||
| PPA‐R: Faces | p = .02 | ||||
| Inverted “Count20” exp. (n = 11) | Slope 0°–4° | FFA‐L: Inverted faces | Test preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .019 | One sample Wilcoxon Signed‐Rank (2‐sided) |
| FFA‐R: Inverted faces | p = .064 | ||||
| PPA‐L: Inverted houses | p = .001 | ||||
| PPA‐R: Inverted houses | p = .006 | ||||
| FFA‐L: Inverted houses | Test non‐preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .048 | |||
| FFA‐R: Inverted houses | p = .27 | ||||
| PPA‐L: Inverted faces | p = .27 | ||||
| PPA‐R: Inverted faces | p = .08 | ||||
| Slope 0°–8° | FFA‐L: Inverted faces | Test preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .003 | One sample Wilcoxon Signed‐Rank (2‐sided) | |
| FFA‐R: Inverted faces | p = .001 | ||||
| PPA‐L: Inverted houses | p = .37 | ||||
| PPA‐R: Inverted houses | p = .03 | ||||
| FFA‐L: Inverted houses | Test non‐preferred category slope ≠ 0 (i.e. BOLD eccentricity effect) | p = .009 | |||
| FFA‐R: Inverted houses | p = .009 | ||||
| PPA‐L: Inverted faces | p = .43 | ||||
| PPA‐R: Inverted faces | p = .1 |
Note: For each statistical contrast, its full details are provided and the outcome in Result column. Significant results in bold. See Results for specific details.
To examine whether these apparent BOLD eccentricity effects were significant for each ROI, whether any BOLD eccentricity effects were modulated by stimulus category (preferred vs non‐preferred stimuli), and whether there were any right–left hemispheric differences for FFA and/or PPA, we subjected the event‐related peak activations (see Figures 4a and 5a and Methods) in right and left FFA and in right and left PPA to three‐way ANOVAs (eccentricity (0°, 4°, 8°) × category (preferred/non‐preferred) × hemisphere (R/L hemisphere)). As expected, for both FFA and PPA, we found a significant BOLD eccentricity effect (all p's < .01; Table 4), a significant stimulus category effect (all p's < 10−5), and, importantly, an interaction between category and eccentricity (FFA: p = 6 × 10−4, PPA: p = 1.3 × 10−6). This indicates that for both FFA and PPA, their eccentricity‐based activation modulations were dependent on their stimulus category preference such that the modulations for their preferred and non‐preferred categories were different. No differences between right and left hemispheres were found; however, a significant three‐way interaction (hemisphere × category × eccentricity) was found in PPA (F (2,24) = 5.82, p = .009, Table 4). This may reflect eccentricity‐based activation modulation by category in a different manner for the left and right hemisphere and may reflect the fact that for houses in PPA‐L, there was an average enhancement of the signal from 4° to 8°, but for PPA‐R, there was an average reduction of the signal from 4° to 8° (average % signal change for 4° to 8° in PPA‐L 0.3% to 0.41%, in PPA‐R 0.35% to 0.27%).
TABLE 4.
Statistical examination of hemisphere, category, and eccentricity effect for fusiform face area (FFA) and parahippocampal place area (PPA) in each experiment.
| Experiment | ROI | Factors | Main effects | Interactions |
|---|---|---|---|---|
| “Count20” exp. | FFA (n = 11) | Hemisphere | F (1,10) = 0.17, p = .68 | Hemisphere × Category |
| R, L | F (1,10) = 0.27, p = .61 | |||
| Category | F (1,10) = 140.86, p = 3 × 10 −7 | Hemisphere × Eccentricity | ||
| Pref., Non‐pref. | F (2,20) = 0.41, p = .66 | |||
| Eccentricity | F (2,20) = 48.43, p = 2 × 10 −8 | Category × Eccentricity | ||
| 0°, 4°, 8° | F (2,20) = 10.78, p = 6 × 10 −4 | |||
| Hemisphere × Category × Eccentricity F (1.2,12.01) = 0.43, p = .56 | ||||
| PPA (n = 13) | Hemisphere | F (1,12) = 0.88, p = .36 | Hemisphere × Category | |
| R, L | F (1,12) = 0.43, p = .52 | |||
| Category | F (1,12) = 72.94, p = 1 × 10 −6 | Hemisphere × Eccentricity | ||
| Pref., Non‐pref. | F (2,24) = 2.9, p = .07 | |||
| Eccentricity | F (2,24) = 7.84, p = .003 | Category × Eccentricity | ||
| 0°, 4°, 8° | F (2,24) = 25.05, p = 1 × 10 −6 | |||
| Hemisphere × Category × Eccentricity F (2,24) = 5.82, p = .009 | ||||
| “DBLstml” exp. | FFA (n = 12) | Hemisphere | F (1,11) = 2.91, p = .1 | Hemisphere × Category |
| R, L | F (1,11) = 6.1, p = .03 | |||
| Category | F (1,11) = 57.189, p = 1 × 10 −5 | Hemisphere × Eccentricity | ||
| Pref., Non‐pref. | F (2,22) = .83, p = .4 | |||
| Eccentricity | F (2,22) = 32.8, p = 2 × 10 −7 | Category × Eccentricity | ||
| 0°, 4°, 8° | F (2,22) = 9.43, p = .003 | |||
| Hemisphere × Category × Eccentricity F (2,22) = 3.49, p = .04 | ||||
| PPA (n = 14) | Hemisphere | F (1,13) = 4.35, p = .057 | Hemisphere × Category | |
| R, L | F (1,13) = 4.94, p = .04 | |||
| Category | F (1,13) = 45.81, p = 1 × 10 −5 | Hemisphere × Eccentricity | ||
| Pref., Non‐pref. | F (2,26) = 2.04, p = .1 | |||
| Eccentricity | F (2,26) = 2.94, p = .07 | Category × Eccentricity | ||
| 0°, 4°, 8° | F (2,26) = 5.53, p = .02 | |||
| Hemisphere × Category × Eccentricity F (2,26) = 2.41, p = .1 |
Note: For each ROI and each experiment, the results of a repeated‐measures, three‐way ANOVA on peak response with hemisphere (R, L), category preference, and eccentricity as factors are presented (see Results for further details). Significant results in bold.
Abbreviations: Non‐pref, non‐preferred category; Pref, preferred category; R/L, right/left hemisphere.
Although a fixation point was present throughout each experiment to assist participants in keeping fixation throughout, we examined whether our results were potentially affected by poor fixations. To that end, we compared brain activations between participants with good vs those with poor fixations for each of the ROIs separately. While only half of the participants in the upright “Count20” experiment (7 of 14) were classified as having good fixation behavior (see Methods), we found that activation levels were not affected by fixation quality in any of the ROIs (3‐way mixed‐design ANOVAs with fixation quality (good/poor) × eccentricity (0°, 4°, 8°) × category (preferred/non‐preferred) on peak responses revealed no significant effect of fixation quality and no significant interaction between fixation quality and condition in all ROIs; see Table S3 at https://osf.io/m8czv/). We assume that this may reflect a spatio‐temporal stimulation tradeoff, where participants that kept fixation throughout parafoveal conditions received continuous peripheral stimulation, whereas participants moving their eyes to peripheral locations may have benefitted from more central stimulation but in a temporally interleaved fashion.
To quantify the proportion of eccentricity‐based activation modulation relative to each region's peak response, we normalized the responses within each ROI relative to the activation to its preferred category at central vision and this was done to the preferred and non‐preferred categories (see Methods and Figure 4b). In FFA, as can be seen in Figure 4b, we found a more gradual decrease in activation with growing eccentricity for its preferred (faces) than for its non‐preferred (houses) category such that the activation to central and parafoveal faces remained above baseline (0° to 8° activity reduction for faces: FFA‐L: 76% ± 18% (SD), FFA‐R: 76% ± 28% (SD)), whereas activation to parafoveal houses at 8° declined to below baseline activity. In PPA, the decrease in activation with eccentricity for its preferred category (from 0° to 8° for houses: PPA‐L: 28% ± 65% (SD), PPA‐R: 60% ± 34% (SD)) appeared to be more moderate than that found for faces in the FFA (∼76% decrease), as we anticipated. However, more importantly, in PPA we observed that for its non‐preferred category (faces), there was a substantial deviation from the anticipated BOLD eccentricity effect. Specifically, while the typical BOLD eccentricity effect is evident by reduced activation with growing eccentricity, in PPA we found an unexpected increase (rather than decrease) of activation with growing eccentricity for the non‐preferred category (faces) such that faces at 8° significantly activated PPA above baseline (one‐sample, two‐sided Wilcoxon signed‐rank test: face‐8°: PPA‐L: p = .008, PPA‐R: p = .04, see Table 3), while central faces significantly deactivated it (one‐sample, two‐sided Wilcoxon signed‐rank test: faceCenter: PPA‐L: p = .008, PPA‐R: p = 2 × 10−4, see Table 3). This eccentricity‐based rise of activity was significant (paired, two‐sided Wilcoxon signed‐rank test: faceCenter vs. face‐8°: PPA‐L p = 4 × 10−4; PPA‐R p = 1 × 10−4, see Figure 4b and Table 3).
To estimate and quantify the eccentricity‐related activation modulations, we defined an index that reflected the slope of response change with eccentricity (from 0° to 4° and from 0° to 8°) for each ROI, for each category (see Figure 4c,d and https://osf.io/m8czv/ Tables S4 and S5).
For the slopes between 0° and 4°, both FFA and PPA showed a significant BOLD eccentricity effect for their preferred category (significantly negative slopes: FFA‐L: p = .002, FFA‐R: p = 2 × 10−4, PPA‐L: p = 4 × 10−4, PPA‐R: p = 1 × 10−4, one‐sample, two‐sided Wilcoxon signed‐rank tests; see Table 3), which were comparable in magnitude (repeated two‐way ANOVA with ROI (FFA/PPA) × hemisphere (R/L) on the preferred category slopes revealed no effect of ROI or hemisphere and no interaction; see Table 5). For the non‐preferred categories in FFA, there was a significant BOLD eccentricity effect (FFA‐L: p = .006, FFA‐R: p = .047; one‐sample, two‐sided Wilcoxon signed‐rank tests), but for PPA, there was no BOLD eccentricity effect (the slopes were not significantly different than zero: PPA‐L: p = .8, PPA‐R: p = .9; one‐sample, two‐sided Wilcoxon signed‐rank tests, see Figure 4c and Table 3).
TABLE 5.
Fusiform face area (FFA) vs parahippocampal place area (PPA) preferred category slope differences.
| fMRI experiment | Factor | Main effect | Interactions | |
|---|---|---|---|---|
| Upright “Count20” exp. (n = 10) | Slope 0° to 4° | ROI | F (1,9) = 2.4, p = .15 | F (1,9) = 0.01, p = .91 |
| FFA, PPA | ||||
| Hemisphere | F (1,9) = 0.1, p = .75 | |||
| R, L | ||||
| Slope 0° to 8° | ROI | F (1,9) = 6.9, p = .027 | F (1,9) = 1.61, p = .23 | |
| FFA, PPA | ||||
| Hemisphere | F (1,9) = 1.28, p = .28 | |||
| R, L | ||||
| “DBLstml” exp. (n = 12) | Slope 0° to 4° | ROI | F (1,11) = 5.2, p = .043 | F (1,11) = 1.52, p = .24 |
| FFA, PPA | ||||
| Hemisphere | F (1,11) = 2.4, p = .14 | |||
| R, L | ||||
| Slope 0° to 8° | ROI | F (1,11) = 16.79, p = .002 | F (1,11) = 4.77, p = .051 | |
| FFA, PPA | ||||
| Hemisphere | F (1,11) = 0.38, p = .54 | |||
| R, L | ||||
Note: For each upright experiment, the results of a two‐way repeated‐measures ANOVA (with ROI and hemisphere as factors) on the preferred category slope are presented (see Results). Significant results in bold. Note that in both experiments, a significant difference between FFA and PPA was found in their preferred category slope reflecting FFA's higher sensitivity to eccentricity.
We further examined the slopes between 0° and 8° for FFA and PPA according to their preferred and non‐preferred categories. For their preferred categories, both PPA and FFA showed an anticipated BOLD eccentricity effect [significant negative slopes with one‐sample, two‐sided Wilcoxon signed‐rank test: houses in PPA (PPA‐L: p = .02, PPA‐R: p = 2 × 10−4) and faces in FFA (FFA‐L: p = 4 × 10−4, FFA‐R: p = 2 × 10−4); see Figure 4d and Table 3] where PPA showed a quantitatively smaller BOLD eccentricity effect (smaller slope) than that found in the FFA [repeated two‐way ANOVA with ROI (FFA/PPA) × hemisphere (R/L) on the preferred category slopes revealed a significant ROI (but not hemisphere) effect (p = .02); see Table 5]. Importantly, for its non‐preferred category, PPA showed an unexpected reverse BOLD eccentricity effect (enhancement of activation with growing eccentricity). This reverse BOLD eccentricity effect for PPA's non‐preferred category (faces) was qualitatively different than that found for its preferred category [(significantly negative slope for houses PPA‐L: p = 2 × 10−4, PPA‐R: p = .02, one‐sample, two‐sided Wilcoxon signed‐rank test) but significantly positive slope for faces (PPA‐L: p = 4 × 10−4, PPA‐R: p = 1 × 10−4), see Table 3 and Figure 4d] and from that found in the FFA for the non‐preferred category (houses) where slopes were significantly negative in line with the anticipated BOLD eccentricity effect (significantly negative slope for houses: FFA‐L: p = 4 × 10−4, FFA‐R: p = .001, one‐sample, two‐sided Wilcoxon signed‐rank test; see Table 3).
We also directly tested our hypothesis that the reduction in activation within each ROI is category‐independent. To that end, we ran two‐way ANOVA with category and hemisphere on each region's 0° to 8° slopes and found, in contrast to our hypothesis, that for each of the ROIs, the slopes were category‐dependent (see Table 6).
TABLE 6.
Effects of category preference on slope for each region, by experiment.
| fMRI experiment | Factor | Main effect | Interactions | |
|---|---|---|---|---|
| Upright “Count20” exp. | FFA (n = 11) | Category | F (1,10) = 17.47, p = .002 | F (1,10) = 1.15, p = .30 |
| Pref., Non‐pref. | ||||
| Hemisphere | F (1,10) = 0.79, p = .39 | |||
| R, L | ||||
| PPA (n = 13) | ROI | F (1,12) = 53.12, p = 9 × 10 −6 | F (1,12) = 8.63, p = .01 | |
| Pref., Non‐pref. | ||||
| Hemisphere | F (1,12) = 3.18, p = .1 | |||
| R, L | ||||
| “DBLstml” exp. | FFA (n = 12) | Category | F (1,11) = 15.22, p = .002 | F (1,11) = 0.22, p = .64 |
| Pref., Non‐pref. | ||||
| Hemisphere | F (1,11) = 0.63, p = .44 | |||
| R, L | ||||
| PPA (n = 14) | Category | F (1,13) = 34.92, p = 5 × 10 −5 | F (1,13) = 5.43, p = .03 | |
| Pref., Non‐pref. | ||||
| Hemisphere | F (1,13) = 2.13, p = .16 | |||
| R, L | ||||
| Inverted “Count20” exp. | FFA (n = 9) | Category | F (1,8) = 13.89, p = .006 | F (1,8) = 0.39, p = .54 |
| Pref., Non‐pref. | ||||
| Hemisphere | F (1,8) = 1.21, p = .30 | |||
| R, L | ||||
| PPA (n = 10) | Category | F (1,9) = 4.43, p = .06 | F (1,9) = 0.005, p = .94 | |
| Pref., Non‐pref. | ||||
| Hemisphere | F (1,9) = 9.05, p = .01 | |||
| R, L |
Note: For each ROI, repeated two‐way ANOVAs with category preference × hemisphere on the slope (change of peak response between central and 8° stimuli, see Methods). Significant effects are indicated in bold. Note that for both regions, there was a significant effect of category preference on the slope in the upright experiments. Pref./Non‐pref. indicate for each region its preferred and non‐preferred categories.
We also complemented these slope‐based analyses with additional similar analyses that were based a different external localization with bigger images of the FFA and PPA, which is available at https://osf.io/m8czv/ and those results replicated the qualitative differences in the slopes of FFA and PPA to their non‐preferred categories reported here. Importantly, in both analyses, PPA showed an unexpected reverse BOLD eccentricity effect at 8°.
In summary, in the upright “Count20” experiment, we found in line with our hypothesis (i) the expected face‐house selectivity in each of the ROIs, (ii) that FFA activation declined more rapidly with eccentricity than PPA did only at 8°, and in contrast to our hypothesis that (iii) in both regions, the eccentricity‐based decline was modulated by visual category (see Table 6). Importantly, PPA showed a qualitative difference in its decline such that for faces, it showed no BOLD eccentricity effect at 4° and a reverse BOLD eccentricity effect (growing activation with eccentricity) at 8°.
3.2.2. Control experiment 1: Upright “DBLstml” experiment
To examine whether the effects we found in the “Count20” experiment were task‐dependent (Harel et al., 2014), we ran an additional control experiment (“DBLstml”) with a new cohort of participants where we kept the same experimental paradigm but employed a different task (see Methods). We first replicated the expected category selectivity in FFA and PPA with higher activation to the central preferred relative to the central non‐preferred category (see Figure 5a and statistical results in Table 3). As in the “Count20” experiment, we also found here that for its preferred category, PPA's eccentricity‐based modulations appeared more moderate than those found in the FFA (Figure 4). Specifically, the normalized response analysis (see Methods and Figure 5b) showed that in FFA, there was a decrease of ∼82% for faces from 0° to 8° (FFA‐L: 81% ± 28% (SD), FFA‐R: 84% ± 44% (SD); similar to “Count20” exp. with ∼76% reduction), while in PPA, there was a more moderate decrease of ∼13% ± 91% (SD) in PPA‐L and ∼44% ± 47% in PPA‐R for houses from 0° to 8° (cf. with PPA‐L: ∼ 28%, PPA‐R: ∼ 60% in the “Count20” exp.).
In the slope‐related analyses, we found similar results to those found in the “Count20” experiment. For slopes from 0° to 4°, we found an effect of region but not of hemisphere [repeated two‐way ANOVA with ROI (FFA/PPA) × hemisphere (R/L) on the preferred category slopes at 4° revealed a significant ROI (but not hemisphere) effect (p = .043); see Table 5]. The effect of region reflected steeper average slopes in FFA (more negative) relative to those found in PPA. Importantly, while in both FFA and PPA, there was a significant BOLD eccentricity effect for the preferred category (significantly negative slopes: FFA‐L: p = 4 × 10−4, FFA‐R: p = 1 × 10−4, PPA‐L: p = .057 (only trending), PPA‐R: p = .008, one‐sample, two‐sided Wilcoxon signed‐rank tests), for the non‐preferred categories, there was no BOLD eccentricity effect in FFA and in PPA (the slopes were not significantly different than zero; in FFA‐L: p = .26, FFA‐R: p = .24; PPA‐L: p = .54, PPA‐R: p = .90; one‐sample, two‐sided Wilcoxon signed‐rank tests, see Figure 5c and Table 3).
For slopes from 0° to 8°, for their preferred category, both PPA and FFA showed negative slopes consistent with our results in the “Count20” exp. (the BOLD eccentricity effect, one‐sample, two‐sided Wilcoxon signed‐rank test: for houses in PPA (PPA‐L: p = .02, PPA‐R: p = .002) and for faces in FFA (FFA‐L: p = 4 × 10−4, FFA‐R: p = 1 × 10−4), see Table 3) with PPA showing quantitatively smaller slopes than those found in the FFA (two‐way repeated‐measures ANOVA with ROI (FFA/PPA) × hemisphere (R/L) on preferred category slope: effect of ROI (FFA/PPA): F (1,11) = 16.79, p = .002, Table 5). Importantly, for its non‐preferred category, PPA showed here too (as in the “Count20” exp.) a positive slope (opposite to the expected BOLD eccentricity effect; one‐sample, two‐sided Wilcoxon signed‐rank test for faces: PPA‐L (p = .02), PPA‐R (p = .02)) that was qualitatively different than that found in FFA for its non‐preferred category (significantly negative slopes for houses: FFA‐L (p = .002), FFA‐R (p = .004), see Figure 5d and Table 3). Here too, as in the “Count20” experiment, we found that the slopes within each region were category‐dependent (see Table 6).
In summary, here, with a different task and a new group of participants, we found again results that are consistent with our original “Count20” experiment.
3.2.3. Control experiment 2: Inverted “Count20” experiment
In our behavioral study (Kreichman et al., 2020), we found that inversion exposed qualitatively different eccentricity‐based perceptual modulation for houses than for faces. Therefore, here we were also interested to investigate whether this behavioral difference can be attributed to processing differences between FFA and PPA (i.e., whether stimulus inversion affects eccentricity‐based modulations differently in FFA and PPA). Since upright and inverted stimuli were not presented in the same runs, we were not able to directly compare their activation levels, and since inverted stimuli do not typically optimally activate these ROIs (e.g., (Gilaie‐Dotan et al., 2010; Yovel & Kanwisher, 2005)), we did not calculate normalized responses.
The slopes between 0° and 4° were similar to those found in the upright “Count20” experiment although noisier (see Table S4 at https://osf.io/m8czv/).
The slopes between 0° and 8° in the FFA for both preferred and non‐preferred categories were similar (albeit noisier) to those with the upright stimuli, with negative slopes (i.e. BOLD eccentricity effect) for both categories (see Figure 5e–h, Table 3, and Table S5 at https://osf.io/m8czv/; one‐sample, two‐sided Wilcoxon signed‐rank test: significantly negative slopes for inverted faces (FFA‐L: p = .003, FFA‐R: p = .001) and were significantly negative for inverted houses (FFA‐L: p = .009, FFA‐R: p = .009)). Here too, we found that the slopes were modulated by visual category such that eccentricity‐based activation decline was faster for the preferred category than for the non‐preferred category (see Table 6). For the inverted non‐preferred category (inverted faces), we found that the activation levels were around baseline, and importantly, there were no modulations of activity by eccentricity in both PPA‐R and PPA‐L (slopes not significantly different than zero: PPA‐L (p = .43), PPA‐R (p = 1); one‐sample, two‐sided Wilcoxon signed‐rank test, see Table 3).
In summary, for the slopes between 0° and 8° for inverted stimuli, FFA showed a BOLD eccentricity effect for both categories, while PPA was much less affected by eccentricity.
3.3. Summary of slope analyses
Table 7 presents a summary of the eccentricity‐based reductions (i.e., slopes) for each of the ROIs according to their stimulus category preference for all the participants in each experiment, and Figure 6 presents a within‐participant quantitative visualization of the slopes from 0° to 8° for participants that participated in both the upright and the inverted “Count20” experiments (n = 11). As can be seen in Table 7, a qualitative difference between PPA and FFA emerges at 8° and this is also reflected by stimulus inversion. Interestingly, both FFA and PPA activation modulation by eccentricity was influenced by the visual category being processed. As can be seen in Figure 6, while for the FFA, a BOLD eccentricity effect was consistently found for preferred and non‐preferred categories regardless of whether they were presented upright or inverted, for PPA there were qualitative changes of the eccentricity‐based modulations (i.e., negative, zero and positive eccentricity‐based modulations) found according to the visual category being viewed. While we found the anticipated BOLD eccentricity effect in PPA for its preferred category, we found no modulation by eccentricity to inverted faces, and a reverse eccentricity effect for its non‐preferred category of faces.
TABLE 7.
A conceptual summary of the main slope findings by ROI and category preference across experiments.
| Exp. | Category | Slopes 0° to 4° | Slopes 0° to 8° | ||||||
|---|---|---|---|---|---|---|---|---|---|
| FFA | PPA | FFA | PPA | ||||||
| “Count20” upright | Preferred | BEE, BEE | (√) | BEE, BEE | (√) | BEE, BEE | (√) | BEE, BEE | (−) |
| Non‐preferred | BEE, BEE | (−) | noBEE, noBEE | (√) | BEE, BEE | (√) | revBEE, revBEE | (√) | |
| “DBLstml” | Preferred | BEE, BEE | BEE, BEE | BEE, BEE | BEE, BEE | ||||
| Non‐preferred | noBEE, noBEE | noBEE, noBEE | BEE, BEE | revBEE, revBEE | |||||
| “Count20” inverted | Preferred | BEE, BEE | BEE, BEE | BEE, BEE | BEE, noBEE | ||||
| Non‐preferred | noBEE, BEE | BEE, noBEE | BEE, BEE | noBEE, noBEE | |||||
Note: BEE – BOLD eccentricity effect represents significantly negative slopes, noBEE represents slopes not significantly different from zero (i.e., no BOLD eccentricity effect), and revBEE signifies a reverse BOLD eccentricity effect (i.e. significantly positive slopes, increase in activation with growing eccentricity). Italic denotes borderline significance, (√) indicates that the results from the additional bigger images localizer replicated and (−) indicates that there was no effect of eccentricity on activation in the bigger images localizer (i.e. noBee).
FIGURE 6.
Regions of interest (ROI) eccentricity‐based reductions (slopes from center (0°) to 8°) by category preference (n = 11). For each region, eccentricity‐based reductions (y‐axis, as measured by slopes) are presented by category preference (x‐axis) from most preferred (left) to non‐preferred (right). Faces in orange, houses in blue. Inverted categories are presented between these assuming inverted stimuli would be less preferred and less non‐preferred as they may recruit non‐dedicated networks. Negative slopes indicate a typical BOLD eccentricity effect, and positive values a reverse (negative) BOLD eccentricity effect. Data of “Count20” experiment presented for all n = 11 participants with both upright and inverted data. Error bars represent SEM. Asterisks represent slopes significantly difference than 0.
3.4. Category selectivity analysis
While category selectivity is often viewed as the “gold standard” for functionally partitioning high‐level visual cortex into different category‐selective areas (Epstein & Kanwisher, 1998; Gilaie‐Dotan et al., 2009, 2010; Grill‐Spector, Weiner et al., 2017; Kanwisher et al., 1997; Saxe et al., 2006), this method is also criticized (e.g., for being condition/experiment‐dependent (Friston et al., 2006)). Furthermore, it is typically based on centrally presented stimuli. As FFA and PPA are two of the most investigated regions defined by functional localization criteria, here we were interested in also examining whether category selectivity in FFA and PPA are modulated by eccentricity (i.e., whether preferential category activations exist in parafoveal vision, see Figure 7).
Our main finding, which was consistent across experiments, was that there was a significant effect of eccentricity on category selectivity (one‐way repeated‐measures ANOVAs on selectivity by eccentricity (0°/4°/8°) for each ROI: FFA‐L: p = .003, FFA‐R: p = 5 × 10−5, PPA‐L: p = 1 × 10−5, PPA‐R: p = 1 × 10−7) and this reflected highest category selectivity for central stimuli in both FFA and PPA (R and L) and significant reductions in category selectivity with eccentricity (see Figure 7 and Table 8). While the modulations of the selectivities within the parafovea (4° to 8°) were not consistent across our two experiments (see Figure 7), it is clear that across the two experiments for each of the ROIs, the selectivity at 4° was not smaller than that found for 8° (see Figure 7, Tables 8, and S6 at https://osf.io/m8czv/).
TABLE 8.
Eccentricity influence on category selectivity.
| fMRI experiment and ROI | Factor | Main effect | Post‐hoc Bonferroni/Dunn | |
|---|---|---|---|---|
| Upright “Count20” Exp. | FFA – L (n = 12) | Eccentricity | F (2,22) = 7.8, p = .003 | |
| 0°, 4° | p = .02 | |||
| 4°, 8° | p = 1 | |||
| 0°,8° | p = .01 | |||
| FFA – R (n = 13) | Eccentricity | F (2,24) = 15.14, p = 5 × 10 −5 | ||
| 0°, 4° | p = 1 × 10 −4 | |||
| 4°, 8° | p = 1 | |||
| 0°,8° | p = .001 | |||
| PPA – L (n = 13) | Eccentricity | F (2,24) = 18.8, p = 1 × 10 −5 | ||
| 0°, 4° | p = .003 | |||
| 4°, 8° | p = .29 | |||
| 0°, 8° | p = 5 × 10 −5 | |||
| PPA – R (n = 14) | Eccentricity | F (2,26) = 31.12, p = 1 × 10 −7 | ||
| 0°, 4° | p = 7 × 10 −5 | |||
| 4°, 8° | p = .03 | |||
| 0°, 8° | p = 9 × 10 −6 | |||
| “DBLstml” exp. | FFA – L (n = 12) | Eccentricity | F (2,22) = 10.91, p = 5 × 10 −4 | |
| 0°, 4° | p = .003 | |||
| 4°, 8° | p = 1 | |||
| 0°, 8° | p = .008 | |||
| FFA – R (n = 14) | Eccentricity | F (2,26) = 9.33, p = 8 × 10 −4 | ||
| 0°, 4° | p = .01 | |||
| 4°, 8° | p = 1 | |||
| 0°, 8° | p = .003 | |||
| PPA – L (n = 14) | Eccentricity | F (2,26) = 2.74, p = .1 | ||
| 0°, 4° | ||||
| 4°, 8° | ||||
| 0°, 8° | ||||
| PPA – R (n = 14) | Eccentricity | F (2,26) = 8.6, p = .001 | ||
| 0°, 4° | p = .14 | |||
| 4°, 8° | p = .5 | |||
| 0°, 8° | p = 6 × 10 −5 | |||
Note: For each region, results of one‐way repeated‐measures ANOVA (with Eccentricity as factor) on category selectivity are presented (see Results). Significant effects in bold. Note that overall there were significant influences of eccentricity on category selectivity (highest at the center, see Figure 7).
4. DISCUSSION
Here, we investigated how eccentricity modulates activations in FFA and PPA, face‐ and place‐selective regions in human ventral visual cortex, given our recent behavioral findings, indicating that perception for faces declines faster with eccentricity than for houses (Kreichman et al., 2020). We hypothesized that (i) FFA will be more sensitive to eccentricity than PPA given its foveal bias and PPA's peripheral bias (Levy et al., 2001) and that (ii) each region will show a characteristic region‐specific BOLD eccentricity effect mirroring the behavioral eccentricity effect (Carrasco et al., 1995; Wolfe et al., 1998) and will be independent of the viewed category. In a set of fMRI experiments, we found, in line with our hypothesis, that FFA showed bigger BOLD eccentricity effects in the parafovea (≤8°) than those found in PPA. However, within each region, we found that the reduction rate was modulated by visual category, and importantly, that the effect of category on PPA's modulation was more profound (qualitative) than that found for FFA (showing only quantitative changes in the BOLD eccentricity effect). Specifically, we found that PPA's activity was modulated by eccentricity from reductions to enhancements depending on the viewed category. In addition, both regions' characteristic category selectivity reduced in the parafovea.
4.1. BOLD eccentricity effect modulated by visual category in FFA and PPA
FFA showed a more rapid decline of activation with eccentricity than PPA did, even when considering the category‐related differences, and this was in line with our initial hypothesis. This finding is line with our behavior results that led to this study (Kreichman et al., 2020) where we found that face discrimination performance decreased more rapidly than house perception (at ≤4°). Regardless of these different sensitivities to eccentricity, for each of these regions, we found, in contrast to our assumption, that their eccentricity‐based activation modulations were modulated by visual category. This finding challenges our hypothesis for several reasons. Firstly, some studies show that processing within each region appears region‐specific rather than category‐specific (i.e., display similar computational properties across different visual categories (e.g., Gilaie‐Dotan et al., 2008)). Furthermore, neuroanatomical cytoarchitectural investigations reveal different compartments within ventral cortex ((Weiner et al., 2017) FG1‐FG4, likely corresponding to face‐sensitive (FG4 to mFus faces likely corresponding to our FFA ROI) and place‐sensitive (FG3 to CoS places likely corresponding to our PPA ROI) regions). Such cytoarchitectural characteristics may suggest that the computation each region is performing is hardwired and not category‐specific. While our results that the BOLD eccentricity effect is modulated by the viewed category may be surprising, they could be in line with the findings that processing in ventral cortex may reflect interactions between different properties (e.g., disparity and shape (Gilaie‐Dotan et al., 2002) or disparity, shape, and viewpoint (Janssen et al., 2000)). This may suggest that computational processes within these ventral regions are sensitive to eccentricity on top of their category preference as we have found in the PPA and to a much lesser extent in the FFA. Furthermore, these results may support the assumption that representations in ventral visual cortex are distributed such that multiple regions are responsive to multiple categories (e.g., Ishai et al., 1999).
4.2. PPA activation modulations show qualitative differences at 8° but not at 4°
In both FFA and PPA, we found that eccentricity‐based activations were differentially modulated by visual category, but at 8°, qualitative differences emerged between PPA and FFA and these were driven by PPA's responses to its non‐preferred category. At 4°, we observed in both regions a clear BOLD eccentricity effect for the preferred categories and small to non‐existent eccentricity effect for the non‐preferred categories. However, at 8°, different patterns of modulation emerged. In FFA, as expected, the activation underwent reduction with eccentricity for both preferred and non‐preferred categories (i.e., negative slopes for both visual categories reflecting a BOLD eccentricity effect), which was consistent with previous studies (Finzi et al., 2021; Hasson et al., 2003; Levy et al., 2001). In PPA, however, the activation changes by eccentricity were profoundly affected by the visual category – from reduction in activation with growing eccentricity (a BOLD eccentricity effect for houses) to enhancement in activation with growing eccentricity (reverse BOLD eccentricity effect for faces). This reflects qualitatively different eccentricity‐based modulations in PPA at 8° (not evident at 4°) for its non‐preferred category than that observed in FFA. Importantly, these findings were also substantiated when we examined the data by an additional external localization method using bigger images covering larger portions of the visual field (see Methods, Results, and Supplementary Materials at https://osf.io/m8czv/). The reverse BOLD eccentricity effect in PPA for faces was driven by deactivation to central faces, and significantly less deactivation to peripheral (8°) faces that even appeared somewhat above baseline (e.g., in the “Count20” experiment). Faces have been considered the least preferred category for activating PPA, and therefore, many studies have used this category to assist in localizing PPA by contrasting activation to places vs to faces (Epstein & Kanwisher, 1998; Gilaie‐Dotan et al., 2008, 2010; Levy et al., 2004) regardless of the actual activation level to faces in PPA. However, PPA has a preference for peripheral stimuli (the “periphery effect” (Hasson et al., 2003; Levy et al., 2004)), and therefore, here we found that when combining the two factors (least preferred stimuli at a more preferred (peripheral) location), the overall activation was higher than that found for centrally presented faces. This may mean that the “antagonism” to central faces in PPA may wane with growing eccentricity; however, this requires more direct investigation and quantification. Importantly, we hypothesize that the antagonism to central faces in PPA may be driven by the fact that the processing of central faces recruits and relies on centrally biased processes and perhaps not necessarily on the genuine content they convey (Epstein et al., 1999; Epstein & Kanwisher, 1998; Hasson et al., 2003). In line with this hypothesis, earlier studies show that the category of objects that likely recruits mid‐peripherally biased networks (Epstein et al., 1999; Epstein & Kanwisher, 1998; Gilaie‐Dotan et al., 2010; Hasson et al., 2003) activates PPA to a medium extent (Epstein et al., 1999; Epstein & Kanwisher, 1998; Gilaie‐Dotan et al., 2010; Hasson et al., 2003). Furthermore, objects and houses activate PPA to a lesser extent than full scenes (Epstein et al., 1999; Gilaie‐Dotan et al., 2010; Hasson et al., 2003; Henderson et al., 2008) and this may further suggest that objects and houses that can be considered as the figure in an image recruit lesser peripheral resources than those recruited to process backgrounds in scenes occupying the most peripheral information in a scene. This hypothesis is also partly supported by our findings that inverted faces (recruiting resources beyond the face‐sensitive network (Gilaie‐Dotan et al., 2010)) led to less antagonistic effects in PPA than those observed for faces (i.e., reverse BOLD eccentricity effect observed only for upright faces).
4.3. Investigating peripheral processes in high‐order visual cortex with and without accounting for CMF
Some previous studies investigating parafoveal processing in high‐order visual cortex compensated for the CMF (Levy et al., 2001, 2004; Malach et al., 2002; Schwarzlose et al., 2008) finding that FFA shows a significant foveal bias and PPA shows a strong peripheral bias. However, it is unclear how activation levels are modulated when the stimulus keeps its world size and thus its retinal size (e.g. during a shift of gaze across the visual field). Furthermore, some of the peripheral stimuli employed in these studies included multiple peripheral elements versus a single foveal element potentially confounding the visual field findings. Here, to parametrically investigate how eccentricity influences visual activation, we kept stimulus size constant. We deliberately chose this to mimic the constancy in stimulus retinal size following a shift of gaze across the visual field. Our design enabled us to examine more closely modulations within the parafovea and importantly examine whether they are region‐specific (as we anticipated) or sensitive to the viewed category as well as to inversion. A study focusing on FFA while not accounting for CMF and using upright face stimuli (Yue et al., 2011) found a significant BOLD eccentricity effect (i.e., activation reduction as a function of distance from central vision), which is in line with our findings in the FFA for faces. Interestingly, while we found eccentricity‐based enhancement of activation in PPA for faces (i.e., a reverse BOLD eccentricity effect), they find no eccentricity‐based modulations in the PPA for faces, and their results may also be considered as a qualitative deviation from an anticipated BOLD eccentricity effect.
4.4. Limitations
While our results clearly indicate that eccentricity in the parafovea modulates activation levels in FFA to a stronger extent than in PPA, and that within each of the regions, activation modulation is category‐dependent, there are certain limitations of our study that must be taken into account. First, the fact that we did not directly map pRFs (Dumoulin & Wandell, 2008; Finzi et al., 2021; Kay et al., 2015; Silson et al., 2015) prevents us from determining whether the sensitivity of each region is uniform or varies according to visual field sensitivities. FFA has shown consistent preference to central versus peripheral vision (Hasson et al., 2002; Levy et al., 2001). In addition, most population receptive fields (pRFs) in FFA cover the fovea for varying pRF sizes (Gomez et al., 2018; Grill‐Spector, Kay, et al., 2017; Kay et al., 2015), and thus, while foveal stimuli are likely to activate most of the FFA, stimuli at eccentricities beyond central vision (in our case the 4° and 8° conditions) are likely to activate a smaller part of the FFA. Therefore, the observed rapid reduction in FFA's BOLD magnitude with eccentricity may have resulted from reduced activation in the activated voxels and/or from a reduction in the number of activated voxels. PPA on the other hand is more sensitive to visual field organization (Silson et al., 2015, 2016) with a more retinotopic‐like representation where pRFs seem to be centered across different eccentricities in the visual field. Therefore, in the PPA, we may be sampling different groups of neurons when we compare activations to the different eccentricities. Relatedly, as our paradigm included for each eccentricity mixed locations across the visual field (e.g., right/left/up/down), our paradigm prevents revealing potential location‐specific modulations, which may be present (e.g., PPA showing upper visual field bias (e.g., Arcaro et al., 2009; Silson et al., 2015)). Second, while the results we report here for 8° (i.e. qualitative differences between PPA's and FFA's eccentricity‐based modulations for their non‐preferred category) may be criticized due to the limited partial coverage by the standard visual localizer we used (selectivities up to 6°), our findings at 4° show that there are quantitative differences between FFA and PPA when considering the parafovea up to 4°. In an effort to address this shortcoming, we also carried out an additional independent localization of PPA and FFA by bigger images (sized 12° × 12°) that replicated our main findings (qualitative differences between PPA's and FFA's eccentricity‐based modulations at 8° for the non‐preferred category) and assisted in further addressing this potential limitation. Third, the fact that we have not employed fMRI adaptation methods (Gilaie‐Dotan et al., 2008, 2010; Gilaie‐Dotan & Malach, 2007; Grill‐Spector & Malach, 2001) prevents us from precisely determining whether the category‐based reductions within each region reflect modulations within the same population of neurons or whether they reflect within‐voxel modulations of different sub‐populations of neurons. Fourth, our stimuli were not optimal for activating PPA. This is due to the fact that our localizer was limited to ∼6° × 6°, and therefore, our analyses may have not captured PPA's sensitivity to eccentricities beyond that, and our stimuli (sized ∼2° × 2.5°) may have been too small to address PPA's peripheral bias (Levy et al., 2004). While we substantiated our findings using localization with bigger images (see above), more specific investigations of PPA across the visual field may allow more precise characterization of its functional sensitivity relative to those evident in our study. Lastly, since our study focused on FFA and PPA, which are specialized ventral stream regions with apparent antagonistic functional sensitivities (to faces and houses/places), our results may not generalize to other visual categories and/or to other category‐sensitive regions (e.g., OFA and OPA/TOS (Gilaie‐Dotan et al., 2008, 2010; Silson et al., 2015, 2016)).
4.5. Summary
While further investigations are required to address additional aspects not investigated in our study, our results highlight important properties of high‐level visual cortex. As perception changes with growing distance from central visual field (e.g., Akselevich & Gilaie‐Dotan, 2022; Kreichman et al., 2020), here we found that two of the most prominent regions in ventral high‐order visual cortex (FFA and PPA) show significant eccentricity‐based modulations of BOLD activity within the parafovea (a BOLD eccentricity effect). Importantly, these modulations are not only driven by eccentricity but also by the viewed category. In FFA, BOLD modulations (that we term a BOLD eccentricity effect) were found for all investigated categories. For PPA on the other hand, the BOLD eccentricity effect was substantially and qualitatively affected by the viewed category, from a BOLD eccentricity effect (to its preferred category) as we anticipated, to an unexpected reverse BOLD eccentricity effect (to its non‐preferred category). We propose that PPA's activity may reflect not only a peripheral bias, but also antagonism to central visual field processes (such as those that are recruited by central upright faces but not by central inverted faces eliciting activation beyond central‐vision‐related networks (Gilaie‐Dotan et al., 2010)).
FUNDING INFORMATION
The study was funded by ISF grants 1485/18 and 1462/23 to SGD and Council for Higher Education Levtzion scholarship to OK.
CONFLICT OF INTEREST STATEMENT
The authors declare they have no conflicts of interest.
Supporting information
Data S1. Supporting information.
ACKNOWLEDGMENTS
We thank Rafi Malach and Michal Harel for their support and Kalanit Grill‐Spector, Yoni Pertzov, and Vadim Axlerod for their comments and suggestions.
Kreichman, O. , & Gilaie‐Dotan, S. (2024). Parafoveal vision reveals qualitative differences between fusiform face area and parahippocampal place area. Human Brain Mapping, 45(3), e26616. 10.1002/hbm.26616
Contributor Information
Olga Kreichman, Email: oplaksin88@gmail.com.
Sharon Gilaie‐Dotan, Email: shagido@gmail.com.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in the OSF repository at https://osf.io/m8czv/.
REFERENCES
- Akselevich, V. , & Gilaie‐Dotan, S. (2022). Positive and negative facial valence perception are modulated differently by eccentricity in the parafovea. Scientific Reports, 12(1), 21693. 10.1038/s41598-022-24919-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Arcaro, M. J. , McMains, S. A. , Singer, B. D. , & Kastner, S. (2009). Retinotopic organization of human ventral visual cortex. Journal of Neuroscience, 29(34), 10638–10652. 10.1523/JNEUROSCI.2807-09.2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Berkovich‐Ohana, A. , Furman‐Haran, E. , Malach, R. , Arieli, A. , Harel, M. , & Gilaie‐Dotan, S. (2020). Studying the precuneus reveals structure–function–affect correlation in long‐term meditators. Social Cognitive and Affective Neuroscience, 15(11), 1203–1216. 10.1093/scan/nsaa137 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bonneh, Y. S. , Adini, Y. , & Polat, U. (2015). Contrast sensitivity revealed by microsaccades. Journal of Vision, 15(9), 11. 10.1167/15.9.11 [DOI] [PubMed] [Google Scholar]
- Carrasco, M. , Evert, D. L. , Chang, I. , & Katz, S. M. (1995). The eccentricity effect: Target eccentricity affects performance on conjunction searches. Perception & Psychophysics, 57(8), 1241–1261. 10.3758/BF03208380 [DOI] [PubMed] [Google Scholar]
- Carrasco, M. , & Frieder, K. S. (1997). Cortical magnification neutralizes the eccentricity effect in visual search. Vision Research, 37(1), 63–82. 10.1016/S0042-6989(96)00102-2 [DOI] [PubMed] [Google Scholar]
- Carrasco, M. , McElree, B. , Denisova, K. , & Giordano, A. M. (2003). Speed of visual processing increases with eccentricity. Nature Neuroscience, 6(7), 699–700. 10.1038/nn1079 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chen, X. , Liu, X. , Parker, B. J. , Zhen, Z. , & Weiner, K. S. (2023). Functionally and structurally distinct fusiform face area(s) in over 1000 participants. NeuroImage, 265(May 2022), 119765. 10.1016/j.neuroimage.2022.119765 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cowey, A. , & Rolls, E. T. (1974). Human cortical magnification factor and its relation to visual acuity. Experimental Brain Research, 21(5), 447–454. 10.1007/BF00237163 [DOI] [PubMed] [Google Scholar]
- CVL Face Database . (n.d.). CVL Face Database. http://www.lrv.fri.uni-lj.si/facedb.html, http://www.lrv.fri.uni-lj.si/facedb.html, http://www.lrv.fri.uni-lj.si/facedb.html
- Dumoulin, S. O. , & Wandell, B. A. (2008). Population receptive field estimates in human visual cortex. NeuroImage, 39(2), 647–660. 10.1016/j.neuroimage.2007.09.034 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Epstein, R. , Harris, A. , Stanley, D. , & Kanwisher, N. (1999). The Parahippocampal place area. Neuron, 23(1), 115–125. 10.1016/S0896-6273(00)80758-8 [DOI] [PubMed] [Google Scholar]
- Epstein, R. , & Kanwisher, N. (1998). A cortical representation of the local visual environment. Nature, 392(6676), 598–601. 10.1038/33402 [DOI] [PubMed] [Google Scholar]
- Epstein, R. A. , Higgins, J. S. , Parker, W. , Aguirre, G. K. , & Cooperman, S. (2006). Cortical correlates of face and scene inversion: A comparison. Neuropsychologia, 44(7), 1145–1158. 10.1016/j.neuropsychologia.2005.10.009 [DOI] [PubMed] [Google Scholar]
- Finzi, D. , Gomez, J. , Nordt, M. , Rezai, A. A. , Poltoratski, S. , & Grill‐Spector, K. (2021). Differential spatial computations in ventral and lateral face‐selective regions are scaffolded by structural connections. Nature Communications, 12(1), 2278. 10.1038/s41467-021-22524-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Friston, K. J. , Holmes, A. P. , Worsley, K. J. , Poline, J.‐P. , Frith, C. D. , & Frackowiak, R. S. J. (1994). Statistical parametric maps in functional imaging: A general linear approach. Human Brain Mapping, 2(4), 189–210. 10.1002/hbm.460020402 [DOI] [Google Scholar]
- Friston, K. J. , Rotshtein, P. , Geng, J. J. , Sterzer, P. , & Henson, R. N. (2006). A critique of functional localisers. NeuroImage, 30(4), 1077–1087. 10.1016/j.neuroimage.2005.08.012 [DOI] [PubMed] [Google Scholar]
- Gilaie‐Dotan, S. , Gelbard‐Sagiv, H. , & Malach, R. (2010). Perceptual shape sensitivity to upright and inverted faces is reflected in neuronal adaptation. NeuroImage, 50(2), 383–395. 10.1016/j.neuroimage.2009.12.077 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gilaie‐Dotan, S. , Hahamy‐Dubossarsky, A. , Nir, Y. , Berkovich‐Ohana, A. , Bentin, S. , & Malach, R. (2013). Resting state functional connectivity reflects abnormal task‐activated patterns in a developmental object agnosic. NeuroImage, 70, 189–198. 10.1016/j.neuroimage.2012.12.049 [DOI] [PubMed] [Google Scholar]
- Gilaie‐Dotan, S. , & Malach, R. (2007). Sub‐exemplar shape tuning in human face‐related areas. Cerebral Cortex, 17(2), 325–338. 10.1093/cercor/bhj150 [DOI] [PubMed] [Google Scholar]
- Gilaie‐Dotan, S. , Nir, Y. , & Malach, R. (2008). Regionally‐specific adaptation dynamics in human object areas. NeuroImage, 39(4), 1926–1937. 10.1016/j.neuroimage.2007.10.010 [DOI] [PubMed] [Google Scholar]
- Gilaie‐Dotan, S. , Perry, A. , Bonneh, Y. , Malach, R. , & Bentin, S. (2009). Seeing with profoundly deactivated mid‐level visual areas: Non‐hierarchical functioning in the human visual cortex. Cerebral Cortex, 19(7), 1687–1703. 10.1093/cercor/bhn205 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gilaie‐Dotan, S. , Ullman, S. , Kushnir, T. , & Malach, R. (2002). Shape‐selective stereo processing in human object‐related visual areas. Human Brain Mapping, 15(2), 67–79. 10.1002/hbm.10008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gomez, J. , Natu, V. , Jeska, B. , Barnett, M. , & Grill‐Spector, K. (2018). Development differentially sculpts receptive fields across early and high‐level human visual cortex. Nature Communications, 9(1), 788. 10.1038/s41467-018-03166-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Grill‐Spector, K. , Kay, K. , & Weiner, K. S. (2017a). The functional neuroanatomy of face processing: Insights from neuroimaging and implications for deep learning. In Bhanu B. & Ajay K. (Eds.), Deep learning for biometrics (pp. 3–31). Springer International Publishing. 10.1007/978-3-319-61657-5_1 [DOI] [Google Scholar]
- Grill‐Spector, K. , & Malach, R. (2001). fMR‐adaptation: A tool for studying the functional properties of human cortical neurons. Acta Psychologica, 107(1–3), 293–321. 10.1016/S0001-6918(01)00019-1 [DOI] [PubMed] [Google Scholar]
- Grill‐Spector, K. , & Malach, R. (2004). The human visual cortex. Annual Review of Neuroscience, 27(1), 649–677. 10.1146/annurev.neuro.27.070203.144220 [DOI] [PubMed] [Google Scholar]
- Grill‐Spector, K. , Weiner, K. S. , Kay, K. , & Gomez, J. (2017b). The functional neuroanatomy of human face perception. Annual Review of Vision Science, 3(1), 167–196. 10.1146/annurev-vision-102016-061214 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Grossman, S. , Gaziv, G. , Yeagle, E. M. , Harel, M. , Mégevand, P. , Groppe, D. M. , Khuvis, S. , Herrero, J. L. , Irani, M. , Mehta, A. D. , & Malach, R. (2019). Convergent evolution of face spaces across human face‐selective neuronal groups and deep convolutional networks. Nature Communications, 10(1), 4934. 10.1038/s41467-019-12623-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Harel, A. , Kravitz, D. J. , & Baker, C. I. (2014). Task context impacts visual object processing differentially across the cortex. Proceedings of the National Academy of Sciences, 111(10), E962–E971. 10.1073/pnas.1312567111 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hasson, U. , Harel, M. , Levy, I. , & Malach, R. (2003). Large‐scale mirror‐symmetry organization of human Occipito‐temporal object areas. Neuron, 37(6), 1027–1041. 10.1016/S0896-6273(03)00144-2 [DOI] [PubMed] [Google Scholar]
- Hasson, U. , Levy, I. , Behrmann, M. , Hendler, T. , & Malach, R. (2002). Eccentricity bias as an organizing principle for human high‐order object areas. Neuron, 34(3), 479–490. 10.1016/S0896-6273(02)00662-1 [DOI] [PubMed] [Google Scholar]
- Henderson, J. M. , Larson, C. L. , & Zhu, D. C. (2008). Full scenes produce more activation than close‐up scenes and scene‐diagnostic objects in parahippocampal and retrosplenial cortex: An fMRI study. Brain and Cognition, 66(1), 40–49. 10.1016/j.bandc.2007.05.001 [DOI] [PubMed] [Google Scholar]
- Horton, J. C. , & Hoyt, W. F. (1991). The representation of the visual field in human striate cortex. Archives of Ophthalmology, 109(6), 816. 10.1001/archopht.1991.01080060080030 [DOI] [PubMed] [Google Scholar]
- Ishai, A. , Ungerleider, L. G. , Martin, A. , Schouten, J. L. , & Haxby, J. V. (1999). Distributed representation of objects in the human ventral visual pathway. Proceedings of the National Academy of Sciences, 96(16), 9379–9384. 10.1073/pnas.96.16.9379 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Janssen, P. , Vogels, R. , & Orban, G. A. (2000). Three‐dimensional shape coding in inferior temporal cortex. Neuron, 27(2), 385–397. 10.1016/S0896-6273(00)00045-3 [DOI] [PubMed] [Google Scholar]
- Jigo, M. , Tavdy, D. , Himmelberg, M. M. , & Carrasco, M. (2023). Cortical magnification eliminates differences in contrast sensitivity across but not around the visual field. eLife, 12, 1–23. 10.7554/eLife.84205 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Julian, J. B. , Fedorenko, E. , Webster, J. , & Kanwisher, N. (2012). An algorithmic method for functionally defining regions of interest in the ventral visual pathway. NeuroImage, 60(4), 2357–2364. 10.1016/j.neuroimage.2012.02.055 [DOI] [PubMed] [Google Scholar]
- Kanwisher, N. , McDermott, J. , & Chun, M. M. (1997). The fusiform face area: A module in human extrastriate cortex specialized for face perception. The Journal of Neuroscience, 17(11), 4302–4311. 10.1523/JNEUROSCI.17-11-04302.1997 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kanwisher, N. , Tong, F. , & Nakayama, K. (1998). The effect of face inversion on the human fusiform face area. Cognition, 68(1), B1–B11. 10.1016/S0010-0277(98)00035-3 [DOI] [PubMed] [Google Scholar]
- Kay, K. N. , Weiner, K. S. , & Grill‐Spector, K. (2015). Attention reduces spatial uncertainty in human ventral temporal cortex. Current Biology, 25(5), 595–600. 10.1016/j.cub.2014.12.050 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kreichman, O. , Bonneh, Y. S. , & Gilaie‐Dotan, S. (2020). Investigating face and house discrimination at foveal to parafoveal locations reveals category‐specific characteristics. Scientific Reports, 10(1), 8306. 10.1038/s41598-020-65239-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kreichman, O. , & Gilaie‐Dotan, S. (2023). Parafoveal vision reveals qualitative differences between FFA and PPA. Open Science Framework. https://osf.io/m8czv/ [DOI] [PubMed] [Google Scholar]
- Levi, D. M. , Klein, S. A. , & Aitsebaomo, A. P. (1985). Vernier acuity, crowding and cortical magnification. Vision Research, 25(7), 963–977. 10.1016/0042-6989(85)90207-X [DOI] [PubMed] [Google Scholar]
- Levy, I. , Hasson, U. , Avidan, G. , Hendler, T. , & Malach, R. (2001). Center–periphery organization of human object areas. Nature Neuroscience, 4(5), 533–539. 10.1038/87490 [DOI] [PubMed] [Google Scholar]
- Levy, I. , Hasson, U. , Harel, M. , & Malach, R. (2004). Functional analysis of the periphery effect in human building related areas. Human Brain Mapping, 22(1), 15–26. 10.1002/hbm.20010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Malach, R. , Levy, I. , & Hasson, U. (2002). The topography of high‐order human object areas. Trends in Cognitive Sciences, 6(4), 176–184. 10.1016/S1364-6613(02)01870-3 [DOI] [PubMed] [Google Scholar]
- Martinez, A. , & Benavente, R. (1998). T. A. face database. (n.d.). CVC Technical Report nr 24. http://rvl1.ecn.purdue.edu/∼aleix/aleix_face_DB.html
- Masarwa, S. , Kreichman, O. , & Gilaie‐Dotan, S. (2022). Larger images are better remembered during naturalistic encoding. Proceedings of the National Academy of Sciences, 119(4), e2119614119. 10.1073/pnas.2119614119 [DOI] [PMC free article] [PubMed] [Google Scholar]
- McCarthy, G. , Puce, A. , Gore, J. C. , & Allison, T. (1997). Face‐specific processing in the human fusiform gyrus. Journal of Cognitive Neuroscience, 9(5), 605–610. 10.1162/jocn.1997.9.5.605 [DOI] [PubMed] [Google Scholar]
- RStudio Team . (2020). RStudio: Integrated development environment for R.
- Saxe, R. , Brett, M. , & Kanwisher, N. (2006). Divide and conquer: A defense of functional localizers. NeuroImage, 30(4), 1088–1096. 10.1016/j.neuroimage.2005.12.062 [DOI] [PubMed] [Google Scholar]
- Schwarzlose, R. F. , Swisher, J. D. , Dang, S. , & Kanwisher, N. (2008). The distribution of category and location information across object‐selective regions in human visual cortex. Proceedings of the National Academy of Sciences, 105(11), 4447–4452. 10.1073/pnas.0800431105 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Silson, E. H. , Chan, A. W.‐Y. , Reynolds, R. C. , Kravitz, D. J. , & Baker, C. I. (2015). A Retinotopic basis for the division of high‐level scene processing between lateral and ventral human Occipitotemporal cortex. Journal of Neuroscience, 35(34), 11921–11935. 10.1523/JNEUROSCI.0137-15.2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Silson, E. H. , Steel, A. D. , & Baker, C. I. (2016). Scene‐selectivity and retinotopy in medial parietal cortex. Frontiers in Human Neuroscience, 10(August), 17. 10.3389/fnhum.2016.00412 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Staugaard, C. F. , Petersen, A. , & Vangkilde, S. (2016). Eccentricity effects in vision and attention. Neuropsychologia, 92, 69–78. 10.1016/j.neuropsychologia.2016.06.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wässle, H. , Grünert, U. , Röhrenbeck, J. , & Boycott, B. B. (1990). Retinal ganglion cell density and cortical magnification factor in the primate. Vision Research, 30(11), 1897–1911. 10.1016/0042-6989(90)90166-I [DOI] [PubMed] [Google Scholar]
- Weiner, K. S. , Barnett, M. A. , Lorenz, S. , Caspers, J. , Stigliani, A. , Amunts, K. , Zilles, K. , Fischl, B. , & Grill‐Spector, K. (2017). The Cytoarchitecture of domain‐specific regions in human high‐level visual cortex. Cerebral Cortex, 27(1), 146–161. 10.1093/cercor/bhw361 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wolfe, J. M. , O'Neill, P. , & Bennett, S. C. (1998). Why are there eccentricity effects in visual search? Visual and attentional hypotheses. Perception & Psychophysics, 60(1), 140–156. 10.3758/BF03211924 [DOI] [PubMed] [Google Scholar]
- Xue, S. , Fernández, A. , & Carrasco, M. (2023). Featural representation and internal noise underlie the eccentricity effect in contrast sensitivity. Journal of Neuroscience, 44(3), e0743232023. 10.1523/JNEUROSCI.0743-23.2023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Yovel, G. , & Kanwisher, N. (2005). The neural basis of the behavioral face‐inversion effect. Current Biology, 15(24), 2256–2262. 10.1016/j.cub.2005.10.072 [DOI] [PubMed] [Google Scholar]
- Yue, X. , Cassidy, B. S. , Devaney, K. J. , Holt, D. J. , & Tootell, R. B. H. (2011). Lower‐level stimulus features strongly influence responses in the fusiform face area. Cerebral Cortex, 21(1), 35–47. 10.1093/cercor/bhq050 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1. Supporting information.
Data Availability Statement
The data that support the findings of this study are openly available in the OSF repository at https://osf.io/m8czv/.