Figures and data
PRF diagnostics and mean parameter values by group (control, DP) and hemisphere (left, right).
A) Mean coefficient of determination, R2, of all pRF fits included in ROIs B) Mean category selectivity of pRF selections from the functional localizer. Category selectivity was defined as the percent signal change for each ROI’s preferred category with respect to percent signal change for objects. Voxels were chosen by selecting the top 20% highest t values within anatomical masks. C) Counts of surviving ROIs after pRF fitting by group and hemisphere. An ROI was defined as retinotopically driven, and therefore included for analysis, if more than 10 voxels remained after R2 thresholding. D) The proportion of voxels that were retained after thresholding on R2 > .20. E) Mean pRF size, which was defined as 
Measures of visual field coverage (VFC).
VFC was computed as the proportion of pRFs covering each point in the visual field. A) Group averaged plots of VFC by ROI. Data from the right hemisphere has been reflected across the vertical meridian and combined with data from the left hemisphere. The dotted lines show the convex hull polygon at the 50% density threshold. B) Convex hull polygons, computed at the 50% density threshold, displayed for each individual subject by group and ROI. C) Cuts through the horizontal meridian showing normalized group coverage to the left and right of fixation (at 0 degrees) by hemisphere. D) Total area (in degrees of visual angle squared) for the convex hull polygons in panel B. Datapoints represent individual subjects. E) Proportion of the total convex hull area falling above the horizontal meridian. Values above 0.5 indicate an upper visual field bias and values below 0.5 indicate a lower visual field bias. F) Proportion of the convex hull area falling into the contralateral visual field.
Lines of best fit relating pRF eccentricities (in degrees of visual angle) to pRF sizes by group and ROI.
Gray lines show lines of best fit for individual subjects and colored lines show the group average. Lines were constrained to the central 6 degrees of visual angle because data in far eccentricities were often sparse, especially in more anterior face-selective regions.
Fixed effect parameter estimates of pRF size by hemisphere (right, left) for the control and developmental prosopagnosic (DP) groups.
Separate models were created for each region of interest. Note that in each one the intercept has been mapped to the control group in the left hemisphere. Formula (R, lme4 package): size ∼ group + hemisphere + group:hemisphere + (1+eccentricity|subjectID) * p < 0.01; ** p < 0.001
Spatial distributions of pRF center coordinates across ROIs and hemispheres.
A) Contours representing the group density distributions of pRF center coordinates by group and hemisphere. Green colors represent pRFs in the right hemisphere and blue colors represent pRFs in the left hemisphere. B) The proportion of pRFs falling into four distinct eccentricity bins. Datapoints represent individual participants. C) Measures of bivariate dispersion of pRF centers by group and ROI. Datapoints represent individual participants. Bivariate spread was measured by computing the area of the 95% probability contour encircling the distribution of pRF centers.
Scores from each member of the prosopagnosic sample on diagnostic tests consisting of: The Hanover Early Visual Assessment (HEVA), The 20-Item Prosopagnosia Index (PI20), The Cambridge Face Memory Test (CFMT), The Cambridge Face Memory Test with Australian Faces (CFMT Aus), famous faces test (FFT), old/new faces test, and The Subthreshold Autism Trait Questionairre (SATQ).
Asterisks (*) indicates a given score meets the inclusion criterion set in advance of the study.
Schematic diagram of retinotopic mapping runs.
Diagram is for illustration purposes only and not drawn to scale. All original faces have been replaced with AI-generated faces to preserve identifying information of original people. Runs began with a 4 second full-field exposure followed by eight sweeps of bar stimuli in eight different directions and ending with a single clockwise rotation of a wedge. Participants fixated a central dot throughout each run and indicated via button press whenever the dot changed color to red. The carrier images shown inside each aperture changed every 250 msecs (5 Hz), alternating between natural outdoor scenes and collages of various visual objects on backgrounds of pink noise (shown). See Methods for more details.
Polar angle plots displayed on inflated cortical surfaces for all control participants.
Polar angle plots displayed on inflated cortical surfaces for all DP participants.
Two dimensional density plots (heatmaps) of eye position recordings by subject (controls) and run.
Two dimensional density plots (heatmaps) of eye position recordings by subject (DPs) and run.
Eye fixation performance during the retinotopic mapping experiment for controls and DPs.
Eye fixation performance was summarized by computing the area of the 95% probability contour of eye position samples. Points are averages across all runs obtained. Diamonds are averages for each group (Controls, DPs).
Face-selectivity was defined as the percent signal change to blocks of face stimuli minus the percent signal change to blocks of object stimuli.
To quantitatively compare face-selectivity in DPs and controls, we used the variable window method (Norman-Haignere et al., 2013) whereby voxels were selected for the analysis by applying group-defined region of interest (ROI) masks and choosing the top 20% of the voxels with the highest t value for the faces-minus-objects contrast. Voxel selection was fully cross-validated in a leave-one-run-out fashion – voxels were selected based on three out of four runs and face selectivity was measured from the left out run. For each participant, the final measure of face-selectivity was the average across four cross-validated folds. For each ROI and hemisphere, separate Welche two sample t-tests were conducted comparing data from DPs to controls. Due to the number of tests conducted, a more conservative alpha threshold of p < 0.01 was used to establish statistical significance. * p < 0.01; ** p < 0.001
Fixed effect parameter estimates for models evaluating pRF model goodness-of-fit, R2, by hemisphere (right, left) and group (control, DP).
Separate models were created for each region of interest. In each model, the intercept has been mapped to the control group in the left hemisphere. Formula (R, lme4 package): R2 ∼ group + hemisphere + group:hemisphere + (1|subjectID/hemisphere) * p < 0.01; ** p < 0.001
