Experimental design.

A) Overall procedure. Forty participants were randomly assigned to one of the two functional magnetic resonance imaging (fMRI) sequences, one fMRI sequence covering the whole brain and the other covering only the medial temporal lobe. Participants took part in two (f)MRI sessions, separated by 24 hours including one night of sleep. B) Day 1 procedure. Participants learned naturalistic face-object pairs (shown here as pictograms for illustration) and immediately retrieved the pairs with the category retrieval (imm. retrieval: immediate category retrieval) to assess whether successful learning occurred. Next, participants had a short pause of 5 minutes, during which they remained in the scanner and performed an odd-even task preventing them from actively rehearsing the face-object pairs. Following the short pause, participants took the 30-minute category retrieval. They chose the face-associated object category (organic or inorganic) and rated the confidence of their response (sure, unsure, guess). Due to the short pause and the pseudo-random shuffling of the face-object pairs, the interval between the one-trial learning and the 30-minute category retrieval was constant at 30 minutes for each pair. C) Day 2 procedure. Repetition of the category retrieval at 24 hours including the confidence rating, followed by an object recognition task and another confidence rating.

Behavioral results.

A) Relative frequencies (bar sections) of confidence judgements across retrievals. Frequency of sure responses drops at the 24-hour category retrieval and rises again during recognition. Frequency of guess responses shows the opposite pattern of change. Frequency of unsure responses drops at the 24-hour category retrieval and remains constant between the 24-hour category retrieval and recognition. B) Guessing accuracy (percentage correct of all guess responses) on the category retrievals and the recognition. Above-chance accuracy was observed only for guess responses given on the recognition task (p = 0.005). These results include all 40 participants. Abbreviations: ret = category retrieval; recog = recognition. Note: dotted line represents chance level of 0.5

Right hippocampal activity increased for both correct sure and correct guess responses at the 30-minute and the 24-hour category retrieval and correlated for guess responses with guessing accuracy at both testing time points.

A) Common right hippocampal activity increases underlying correct sure and correct guess responses at the 30-minute category retrieval. We isolated activity increases underlying associative retrieval from activity increases underlying mere face recognition, task characteristics, and retrieval effort by computing the contrast correct sure responses > incorrect guess responses and the contrast correct guess responses > incorrect guess responses. These contrasts were then statistically conjugated to find common brain activation underlying correct sure and correct guess responses. For a table of cluster sizes and t-values, see Table S2. Results presented in this panel stem from the small FOV fMRI sequence. B) Common right hippocampal activity rises underlying correct sure and correct guess responses at the 24-hour category retrieval. Statistical analysis analogous to 3A. For a table of cluster sizes and t-values, see Table S3. Results presented in this panel stem from the small FOV fMRI sequence. C) Right hippocampal activity underlying correct guess responses on the category retrieval correlated with the guessing accuracy on the category retrieval at 30 minutes and at 24 hours. Upper panel: positive brain-behavior correlation in the right hippocampus. We show the contrast of correct guess > incorrect guess responses correlated with the guessing accuracy on the category retrieval at 30 minutes (for cluster sizes and t-values, see Table S4). Lower panel: positive brain-behavior correlation in the right hippocampus. We show the contrast of correct guess > incorrect guess responses correlated with the guessing accuracy on the category retrieval at 24-hours (for cluster sizes and t-values, see Table S5). Outliers (above absolute z-value of 3) were removed. Results presented in this panel stem from the small FOV fMRI sequence.

Hippocampal functional connectivity measured during learning and during the 30-minute category retrieval correlated with the 30-minute category retrieval guessing accuracy.

A) Hippocampal-medial prefrontal functional connectivity during learning correlated with the later recorded 30-minute category retrieval guessing accuracy. We computed the functional connectivity between the seed region in the right hippocampus head (see box inlet) with neocortical regions during learning using a gPPI. In participants with a higher guessing accuracy on the 30-minute category retrieval, the functional connectivity was stronger between the right hippocampus head and the right anterior cingulate gyrus and the right medial prefrontal cortex. Outliers (above absolute z-value of 3) were removed. For a table of cluster sizes and t values see Table S6. Results presented in this panel stem from the whole-brain fMRI sequence. B) Hippocampal-medial prefrontal functional connectivity during the 30-minute category retrieval correlated with the 30-minute category retrieval guessing accuracy. We computed the functional connectivity between the seed region in the right whole hippocampus (see box inlet) with neocortical regions during the 30-minute category retrieval using a gPPI. Participants with a higher guessing accuracy on the 30-minute category retrieval, exhibited a stronger functional connectivity between the right hippocampus and the right medial prefrontal cortex. Outliers (above absolute z-value of 3) removed. For cluster sizes and t-values see Table S7. Results presented in this panel stem from the whole-brain fMRI sequence.

Overnight right-left hippocampal functional connectivity increases improved guessing accuracy at the 24-hour category retrieval.

We computed a gPPI analysis with the seed region in the right hippocampus head (see box inlet) using the small FOV fMRI sequence to examine overnight changes in the functional connectivity within the medial temporal lobe underlying correct guess responses on the category retrieval. We contrasted the functional connectivity underlying the correct guess responses given on the 24-hour category retrieval with the functional connectivity underlying correct guess responses given on the 30-minute category retrieval and then correlated the result with the 24-hour guessing accuracy on the category retrieval. Connectivity estimates and guessing accuracy were z-scored. Outliers (above absolute z-value of 3) removed. See Table S11 for all results.

Encoding – retrieval and retrieval – retrieval similarity.

A) Left panel: Similarity of voxel patterns between encoding and the 30-minute category retrieval were relevant for retrieval success. Significant between-subjects correlation of encoding-retrieval similarity values with guessing accuracy on the 30-minute category retrieval in bilateral middle temporal gyrus. Results presented in this panel were acquired with the whole-brain fMRI sequence. Right panel: small FOV ROI overview for B) and C) B) Similarity of voxel patterns between encoding and the 24-hour category retrieval were relevant for retrieval success. Encoding-retrieval similarity values were significant in left parahippocampal gyrus for both correct guess responses (left) and correct sure responses (right) (t(19) = 2.65, p = 0.015, BF10 = 3.5; t(19) = 2.35, p = 0.029, BF10 = 2.1, respectively). Results presented in this panel were acquired with the small FOV fMRI sequence. C) Dissimilarity of voxel patterns between the 30-minute and the 24-hour category retrieval were relevant for retrieval success. Left panel: 30-minute – 24-hours category retrieval dissimilarity in the right hippocampus head for correct sure responses. Results presented in this panel were acquired with the small FOV fMRI sequence. Right panel: 30-minute – 24-hours category retrieval dissimilarity values correlated between-subjects with the delta of correct sure responses (24-hours sure accuracy minus 30-minute sure accuracy) in bilateral anterior cingulate gyrus (B. ant. cing. g.). Results presented in this panel were acquired with the whole-brain fMRI sequence. *p < 0.05, **p < 0.01 by Student’s t test. Abbreviations: ROI, region of interest; R., right; L., left; hipp., hippocampus; parahipp., parahippocampus; g., gyrus.

Encoding – recognition pattern similarity.

A) Encoding-recognition pattern similarity in the right hippocampal body and right parahippocampal gyrus for correct sure responses on the recognition task (t(19) = 3.91, p = 0.001, BF10 = 34.0; t(19) = 2.11, p = 0.048, BF10 = 1.5, respectively). Results presented in this panel were acquired with the small FOV fMRI sequence. B) Encoding-recognition pattern similarity in bilateral inferior temporal gyrus for correct guess responses on the recognition task (p = 0.038, t(17) = 2.25, BF10 = 1.8). Results presented in this panel were acquired with the whole-brain fMRI sequence. Right panels: ROI overview. *p < 0.05, **p < 0.01 by Student’s t test. Abbreviations: R., right; L., left; B., bilateral; parahipp., parahippocampus; hipp., hippocampus; entorh., entorhinal; inf., inferior; mid., middle; ant., anterior; cing., cingulate; sup., superior; front., frontal; s., sulcus, g., gyrus c., cortex.