Fig. 4: Population decoding results.
From: Single neuron responses underlying face recognition in the human midfusiform face-selective cortex
A Decoding of whether the picture was a face (F) or a place (P), using all responsive units (N = 43); (B) decoding of whether the picture was a familiar face (FF) or unknown face (UF), using face-responsive units (N = 36); (C) decoding of whether the picture was a familiar place (FP) or unknown place (UP), using place-responsive units (N = 24). In all cases, the overall decoding performance is shown on top. D decoding of all 4 categories (familiar face, unknown face, familiar place, and unknown place) using all responsive units (left), and decoding of the pictures presented using all responsive units (right). Decoding performance was similar for familiar and unknown stimuli (for both faces and places) and was higher for faces (27%) compared to places (18%). E Cumulative decoder using all responsive units and discriminating: faces vs. places, familiar vs. unknown faces, familiar vs. unknown places (not significant), and the picture identity of the face and place stimuli. The solid lines mark the corresponding logistic regressions (see “Methods” section). The top and bottom dashed horizontal lines mark the chance level for the binary decisions (face vs. place; familiar vs. unknown face; familiar vs. unknown place) and discrimination of face/place identify, respectively. The colored dashed lines below the chance levels show the time periods where each decoder is significantly larger than chance (permutation test, p < 0.05). We observe different time profiles of the decoders, with the decoding of faces vs. places having an earlier inflection point and higher slope compared to the other ones (right panels; two-sided t-test; N = 24). The inflection point was also earlier for the familiar vs. unknown face decoder compared to the face and place picture identification. Source data are provided as a Source Data file.
