Associative learning task.

(A) The four target face stimuli and their associated scenes. (B) Associative learning task to test memory for face-scene pairs. Participants viewed a series of face-scene pairs and made a judgment about whether the face and scene were matched or not. (C) Memory performance for both match and nonmatch conditions was high suggesting a strong association was formed before the face search task.

(A) Face search task. Each trial started with a 1-second search cue indicating the target face for that trial. This was followed by an 8-second blank search delay period, and then the search display for 250 ms. Participants pressed a button to indicate the location of the target on the left or right. (B) Illustration of the separate 1-back task used for cross-task classifier testing. Consistent with the main visual search task, the image was presented for 1 second followed by an 8-second delay period. (C) Classification scheme. Classifiers were trained on the neural response patterns from the search cue face stimulus and delay periods from the visual search task; the classifier was tested on face or scene sample stimulus and delay period from the 1-back task. (D) Visualization of the twelve functional ROIs on the cortical surface of a representative participant. All ROIs were defined in an individual’s native space.

Decoding of face and scene information during the search cue period.

(A) Evidence of face information in a priori defined ROIs. Greater than chance-level classification accuracies were found in the dLPFC, SPL, and FFA. Evidence for scene information was found in vLPFC. (B and C) Significant brain regions revealed by a whole-brain searchlight procedure with information about the face cue (B) or associated scene (C). Note that scene information was never shown during the cue period and therefore decoding of scene information reflects memory-evoked responses to the cued face.

Decoding of face and scene information during the search delay period.

(A) Evidence of face information in a priori defined ROIs was only found in the left IPS. However, scene information was decoded in both IFJ and PPA, reflecting memory-evoked target-associated information in a network that encodes the target template. (B and C) Whole-brain searchlight analyses showed no additional brain regions carried significant information about the face (B), but additional scene information was found in the retrosplenial cortex (C).

Behavioral and brain results from the face search period.

(A) Behavioral accuracy and RT both showed a scene-validity effect, suggesting scene information was used to guide attention during search. **p < .01, ***p < .001. Error bars refer to 95% confidence intervals. (B) Whole-brain group-level univariate contrast results showing significantly greater activations for the scene-invalid than scene-valid conditions are illustrated in blue (cold colors), and the reverse contrast in red (hot colors). (C) Contrast betas from the scene-invalid minus scene-valid conditions within each of the a priori ROIs, *p < .05, **p < .01.