Article Figures & Data
Figures
- Figure 1.
A, Sample depth maps and random-dot stereograms (RDS). Top, The depth map of the upright face was presented in an oval region that cropped the ears and the neck. Center, An inverted face was created by rotating the upright face by 180° in the picture plane. Bottom, A depth-power matched random shape was generated by randomizing the phases of the Fourier transform of the upright face while preserving the amplitudes. Anaglyphic examples of the RDS in “near” and “far” conditions, at 100% SNR, are also shown for illustration purposes only. B, In the signal-noise (SNR) task (top), SNR varied between 0% and 100%. In the feature task (bottom), two stimuli were presented consecutively on each trial. Disparity difference varied within 150 arcs.
- Figure 2.
A, B, Depth discrimination thresholds in the signal-noise (SNR) and feature tasks. In the SNR task, thresholds were higher (i.e., poorer) for the upright face than the inverted face and the random shape. In the feature task, thresholds for the upright face were significantly lower (i.e., better) than those for the random shape, whereas thresholds for the inverted face did not differ significantly from those of the other two stimulus conditions. C, Sample staircases from a randomly-sampled observer, showing threshold estimates across trials within one experimental run, for the upright face condition. D, Behavioral thresholds obtained in-bore (SNR task). Thresholds were higher for the upright face than the random shape. Individual subject thresholds are overlaid in A–C. Error bars represent ±1 SEM.
- Figure 3.
GLM β-weights (% signal change) of the three stimulus conditions in the (A) left and (B) right hemispheres. Univariate signals were globally higher in the left hemisphere than the right hemisphere. Signals were generally higher in V1, V2, and LO. Error bars represent ±1 SEM (N = 22).
- Figure 4.
A, SVM accuracies for all ROIs in the three exhaustive classifications comparing all stimulus conditions. Classification accuracies (Bonferroni-corrected) that are significantly above-baseline (0.527; dashed line) are indicated by asterisks. Error bars represent ±1 SEM (N = 22). B–D, Brain-behavior correlations in V4 for upright-face versus random-shape (B); V3A for inverted-face versus random-shape (C); and V3 for upright-face versus inverted-face (D). The horizontal axis represents a behavioral index computed as the threshold difference between conditions. Additional results from the whole-brain classification (searchlight) and from an additional brain-behavior correlation following the concatenation of the inverted and non-face conditions (FFA) are presented in Extended Data Figures 4-1, 4-2, respectively.
Tables
Analysis Figure Data structure Type of test Variables Observed power Experiment 1, SNR task 2A Assumed normal One-way repeated-
measures ANOVAIndependent: stimulus condition
Dependent: depth threshold0.843 Experiment 1, feature task 2B Assumed normal One-way repeated-
measures ANOVAIndependent: stimulus condition
Dependent: depth threshold0.722 Experiment 2, behavior 2D Assumed normal One-way repeated-
measures ANOVAIndependent: stimulus condition
Dependent: depth threshold0.610 Experiment 2, GLM 3 Assumed normal Three-way repeated-
measures ANOVAIndependent: hemisphere
Stimulus condition
ROI
Dependent: GLM β-weight0.580 0.372 >0.999 Experiment 2, MVPA 4 Assumed normal One-sample t test SVM accuracy against baseline Upright face vs random
shape classificationV1 0.690 V2 >0.999 V3 0.972 V3A 0.744 V3B 0.422 V7 0.919 hMT+ 0.703 V4 0.880 LO 0.905 FFA 0.762 Inverted face vs random
shape classificationV1 0.959 V2 0.997 V3 0.639 V3A 0.999 V3B 0.479 V7 0.998 hMT+ 0.175 V4 0.999 LO 0.764 FFA 0.396 Upright face vs inverted
face classificationV1 0.437 V2 0.418 V3 0.064 V3A 0.999 V3B 0.875 V7 0.504 hMT+ 0.088 V4 0.903 LO 0.540 FFA 0.838 Experiment 2, MVPA-
behavior correlation4 Assumed normal Pearson correlation Variable 1: Threshold difference Variable 2: SVM accuracy Upright face vs random shape V1 0.050 V2 0.378 V3 0.288 V3A 0.071 V3B 0.055 V7 0.074 hMT+ 0.051 V4 0.690 LO 0.258 FFA 0.075 Inverted face vs random shape V1 0.312 V2 0.087 V3 0.087 V3A 0.579 V3B 0.117 V7 0.342 hMT+ 0.053 V4 0.053 LO 0.075 FFA 0.072 Upright face vs inverted face V1 0.533 V2 0.449 V3 0.984 V3A 0.058 V3B 0.353 V7 0.250 hMT+ 0.202 V4 0.488 LO 0.193 FFA 0.544 Upright face vs non-upright-face V1 0.057 V2 0.129 V3 0.170 V3A 0.060 V3B 0.110 V7 0.065 hMT+ 0.061 V4 0.121 LO 0.064 FFA 0.562
Extended Data Figure 4-1
Sample searchlight maps for SVM discriminations between the upright face and random shape conditions. Results are superimposed onto the representative flattened surface maps of a single participant with ROIs delineated. Gyri are colored in light grey and sulci in dark grey. Download Figure 4-1, TIF file.
Extended Data Figure 4-2
Brain-behavior correlation in FFA for the upright-face versus non-face (i.e., concatenating the inverted face and random shape conditions) comparison. The horizontal axis represents a behavioral index computed as the threshold difference between the upright face condition and the mean of the two non-face conditions. Download Figure 4-2, TIF file.
In this issue
