Fig 1.
This viewing sphere shows the images of faces produced by rotating the viewer/camera in yaw or pitch by 0° (the front view), 45° (three-quarter views) or 75°.
Table 1.
Participant demographics by group (each n = 20).
Fig 2.
This figure shows images of one identity from each of the 21 viewpoints used as test views for all participants. The centre image is a front view and nominally the 0° view. Views along each axis are in 15° increments from 0° (15°, 30°, 45°, 60° and 75°). The views identified by dotted lines are the comparison views for our four different groups (front face, ¾ yaw right, ¾ pitch-down, and ¾ pitch-up).
Fig 3.
Face 1 was always the comparison view (in this case the participant was in the ¾ pitch-up group). Face 2 could be any one of the 21 views rotated in yaw or pitch. The identity of Face 2 could either be the same as Face 1 (as it is here) or different (i.e., it could be an image of one of the 8 other identities). Masks were presented in between and directly after these two face stimuli.
Fig 4.
Overall matching performance is shown for each of the comparison view groups. Error bars represent +/-1 SEM.
Fig 5.
Matching performance as a function of test view axis and test view angle is shown for each of the comparison view groups: front view (top left panel), ¾ yaw view (top right panel), ¾ pitch-down view (bottom left panel), and ¾ pitch-up view (bottom right panel). Error bars represent +/-1 SEM.
Fig 6.
Reaction time (ms) on the matching task as a function of test view axis and test view angle of rotation for the group comparing: front view faces (top left panel), ¾ yaw view faces (top right panel), ¾ pitch-down view faces (bottom left panel), and ¾ pitch-up view faces (bottom right panel). Error bars represent +/-1 SEM.
Table 2.
Unstandardised (B) and Standardised (β) regression coefficients and squared semi-partial correlations (sr2) for each predictor in multiple regression analysis predicting matching performance d’.