Early visual learning induces long-lasting connectivity changes during rest in the human brain

Abstract

Spontaneous fluctuations in resting state activity can change in response to experience-dependent plasticity and learning. Visual learning is fast and can be elicited in an MRI scanner. Here, we showed that a random dot motion coherence task can be learned within one training session. While the task activated primarily visual and parietal brain areas, learning related changes in neural activity were observed in the hippocampus. Crucially, even this rapid learning affected resting state dynamics both immediately after the learning and 24h later. Specifically, the hippocampus changed its coupling with the striatum, in a way that was best explained as a consolidation of early learning related changes. Our findings suggest that long-lasting changes in neuronal coupling are accompanied by changes in resting state activity.

Fig. 1

Experimental paradigm. Participants were scanned on two consecutive days for about 90 min each day. Two resting state runs were acquired each day, preceding and following the learning task or a retinotopic mapping respectively. A structural scan was acquired on both days.

Fig. 2

Behavioral learning and hippocampal activation. a) Participants learned the motion task. Inverse efficiency (IE) is plotted for every block of the task (n = 25). Data are averaged over all participants (n = 11) who learned the task successfully. Error bars show SEM. b) Learning activated the hippocampus. The fitted inverse efficiency values of the learning task were used for the plotted contrast. Statistics were significant at p < 0.05, FWE corrected. Images show activation at p < 0.001 (uncorrected).

Fig. 3

Winning model and summed (group) log evidence for all models. a) Schematic description of the model with a bidirectional connection between the hippocampus and the striatal region. The graphic shows which connections were modified by a consolidation pattern (see Results for detailed explanation). b) The model plotted in a) showed the highest evidence (marked in red). Plotted is the summed log evidence per model relative to the model with the least evidence. The winning model was the same for almost all participants (10 out of 11). A = adaptation, L = learning, C = consolidation, null = no modification.

Fig. 4

Parameter estimates and model fitting reflected consolidation. Parameter estimates for the modulation of the intrinsic connection from a) hippocampus to striatum and from b) striatum to hippocampus. After a big increase directly after the learning in rest run 2, the change in connectivity was preserved at a lower level on the second scanning day for both rest runs (i.e. rest runs 3 and 4). Plotted are the average values for all participants who learned the task (n = 11), error bars indicate the standard error of the mean (SEM). c) Overlay of observed (gray) BOLD time-series during rest with the time-series as predicted by DCM (blue). The two regions included in all tested models are shown for a representative participant.