Neural Pattern Classification Tracks Transfer-Appropriate Processing in Episodic Memory

Averaged accuracy of the 20 cross-validated pattern classifiers trained to discriminate between faces, landmarks, and objects during the encoding phase of the visual and of the verbal memory task. The colored line represents the accuracy of the classifiers trained during the presentation of the target picture. The black line shows the baseline accuracy of the classifiers trained during word cue presentation. Highlighted (*) are the times bins for which classification performance survived multiple comparison correction. In gray is highlighted the time bins for which classification accuracy was highest.

Contribution of each channel to the performance accuracy of the 20 pattern classifiers during the encoding phase of the visual and the verbal memory task. The gray frame highlights the time bins for which classification accuracy was highest.

Relationship between the accuracy of the pattern classifier trained at 170 ms after picture onset during study and the proportion of exemplar hits in the visual and in the verbal memory task.

Pearson r values showing the temporal profile of the association between visual processing (tracked by EEG-based decoding) and memory performance in the visual and verbal memory task. Highlighted (•) are time bins for which the relationship with behavioral performance survived correction for multiple comparisons (for details, see Materials and Methods). A, The green line shows the association for retrieval success (i.e., exemplar hits) and the red line shows the association for retrieval failure (i.e., when participants fail to retrieve any information about the target picture). B, Association for retrieval success (green lines) separated for faces (solid green line), landmarks (dashed green line) and objects (dotted green line), and for retrieval failure (red lines) separated for faces (solid red line), landmarks (dashed red line), and objects (dotted red line).

A, Topography of the SME (5–7 Hz) observed between 120 and 320 ms for the visual and the verbal memory task. Electrodes that reached significance in the cluster-based permutation test are highlighted (*). B, TFRs from a representative channel for the visual (FC5) and verbal (F3) memory task.

For the visual memory task, the classifiers trained between 120–320 ms after picture onset during study and tested around 450 ms after word cue onset at retrieval decode the stimuli category previously associated with the cue. For the verbal memory task, the classifiers trained between 350–450 ms after picture onset during study and tested around 700 ms after word cue onset at retrieval decode the stimuli category previously associated with the cue. A, Accuracy of the decoding at retrieval for both tasks. The black outlines show p = 0.05 significance thresholds generated by a permutation test. B, Results of comparing the accuracy of the replay against chance (33%). * Denotes critical t-value. C, Contribution of each channel for the accuracy of the significant replay of the early visual pattern classifiers at retrieval for the visual and verbal memory task.