Functional brain networks for learning predictive statistics

Abstract

Making predictions about future events relies on interpreting streams of information that may initially appear incomprehensible. This skill relies on extracting regular patterns in space and time by mere exposure to the environment (i.e., without explicit feedback). Yet, we know little about the functional brain networks that mediate this type of statistical learning. Here, we test whether changes in the processing and connectivity of functional brain networks due to training relate to our ability to learn temporal regularities. By combining behavioral training and functional brain connectivity analysis, we demonstrate that individuals adapt to the environment's statistics as they change over time from simple repetition to probabilistic combinations. Further, we show that individual learning of temporal structures relates to decision strategy. Our fMRI results demonstrate that learning-dependent changes in fMRI activation within and functional connectivity between brain networks relate to individual variability in strategy. In particular, extracting the exact sequence statistics (i.e., matching) relates to changes in brain networks known to be involved in memory and stimulus-response associations, while selecting the most probable outcomes in a given context (i.e., maximizing) relates to changes in frontal and striatal networks. Thus, our findings provide evidence that dissociable brain networks mediate individual ability in learning behaviorally-relevant statistics.

Keywords: Brain plasticity; Functional Network Connectivity; Individual differences; Statistical learning; fMRI.

Copyright © 2017 The Authors. Published by Elsevier Ltd.. All rights reserved.

Figures

Fig. 1

Trial and sequence design. (a) The trial design: 8–14 symbols were presented sequentially followed by a cue and the test display. (b) Sequence design: Markov models of the two context-length levels. For the zero-order model (level-0): different states (A, B, C, D) are assigned to four symbols with different probabilities. For the first-order model (level-1), diagrams indicate states (circles) and conditional probabilities (solid arrows: high probability; dashed arrows: low probability). Transitional probabilities are shown in a four-by-four (level-1) conditional probability matrix, where rows indicate the context and columns the corresponding target.

Fig. 2

Behavioral performance. (a) Mean normalized performance index (PI) across participants per level during pre-training (gray bars) and post-training (black bars) test sessions. Error bars indicate standard error of the mean across participants. (b) Strategy index boxplots for level-0 and level-1 indicate individual variability. The upper and lower error bars display the minimum and maximum data values and the central boxes represent the interquartile range (25th to 75th percentiles). The thick line in the central boxes represents the median. (c) Scatterplot of strategy index for level-0 against strategy index for level-1.

Fig. 3

Spatial maps of ICA task-related components. 15 task-related components are shown organized into known functional groups (Allen et al., 2011). Spatial maps are thresholded at p < .005 (FWER corrected) and displayed in neurological convention (left is left) on the MNI template. The x, y, z coordinates per component denote the location of the sagittal, coronal and axial slices, respectively.

Fig. 4

ICA components related to matching strategy. Average spatial maps showing significant negative correlation of BOLD change (post minus pre-training) with strategy index for (a) Learning frequency statistics: Precuneus, Sensorimotor and Right Central Executive. (b) Learning context-based statistics: Precuneus and Middle Temporal. Spatial maps are averaged across sessions, thresholded at p < .005 (FWER corrected) and displayed in neurological convention (left is left) on the MNI template. Open circles in the correlation plots denote outliers.

Fig. 5

ICA components related to maximization strategy. Average spatial maps showing significant positive correlation of BOLD change (post minus pre-training) with strategy index for: (a) Learning frequency statistics: Basal Ganglia. (b) Learning context-based statistics: Left Central Executive. Spatial maps are averaged across sessions, thresholded at p < .005 (FWER corrected) and displayed in neurological convention (left is left) on the MNI template.

Fig. 6

Functional Network Connectivity (FNC) change related to strategy. Correlation matrix of FNC change (post minus pre-training) with strategy index for: (a) frequency statistics and (b) context-based statistics. Black dots indicate significant positive, while black diamonds significant negative correlations (at 95% bootstrapped confidence intervals) of FNC change with strategy index. ICA components included in this analysis are: Left Central Executive Network (lCEN), Right Central Executive Network (rCEN), Middle Temporal (MT), Precuneus (PRCUN), Basal Ganglia (BG) and Sensorimotor (SM).