DCM Analysis

Effective connectivity analysis was conducted using DCM (30), a hypothesis-driven Bayesian approach that describes the biophysical nature of directed interactions among distinct brain regions by incorporating two forward models: one at the neural and one at the hemodynamic level. By combining a priori knowledge of a biologically plausible neural model (input) with the measured BOLD response (output), it is possible to infer on underlying hidden states such as regional causal interactions.

Identical to our earlier analyses (18, 20), we used deterministic, bilinear, one-state DCM to assess the effective connectivity among five regions of the WM network [Owen et al. (3) and Table ​Table2].2]. These regions comprised the left and right inferior parietal cortex (IPC), left anterior cingulate cortex (ACC), as well as left and right dorsolateral prefrontal cortices (DLPFCs). The activation pattern evoked by the contrast 2-back vs. 0-back provided evidence for the choice of intrinsic connections between the five ROI. Furthermore, the differential effect of WM load (as expressed by the 2-back vs. 1-back contrast) motivated the choice of task-modulated connections.

Behavioral results from the working memory task (mean and SD).

For each participant, time series from each of the five ROIs were extracted using the fixed coordinates from the second-level activations identified in the one-sample t-test and adjusted for the effect of interest (F-contrast). No statistical threshold was used within each ROI, which allowed for the time series extraction of the same set of voxels in all participants. The motivation for this approach is based on previous literature (31) and is advantageous for this study because it ensured that there was no overlap of subject-specific spheres in neighboring brain regions. Furthermore, participants having ROIs with weak activations do not have to be excluded but at the expense of potentially including condition-independent noise (31). This is an issue particularly in small sample sizes but potentially less so in our relatively large study.

The extracted time series of all five ROIs were included in one fully connected DCM model, and intrinsic connections were modeled among these regions (see Figure ​Figure2A).2A). The fully connected DCMs were then reduced using the post hoc optimization procedure for approximating model evidence, proposed by Friston and Penny (32). This approach optimizes only the fully connected model, while the evidence for any sub-model is obtained using generalization of the Savage–Dickey density ratio (33). In addition, post hoc diagnostics of each participant’s DCM were conducted using in-house MATLAB routines (adapted from https://sites.google.com/site/jeandaunizeauswebsite/code/explore-dcm) to ensure that model inversion had converged, requiring at least 10% of variance explained.

Dynamic causal model for working memory (WM). (A) Task-independent, intrinsic connections (blue arrows) and driving input (white arrows). (B) WM-modulated connections (red arrows).

Dynamic causal modeling model specification, estimation, and post hoc optimization were carried out with DCM12, as implemented in SPM12b. Statistical inference on model parameters was conducted in SPSS, Version 22.0 (IBM Corporation, NY, USA). Random-effects inference at the connection level was assessed using ANCOVA analysis after covariate adjustment. Between-group differences were considered significant at a threshold of p < 0.001 after accounting for the number of connections (i.e., 21 intrinsic and 14 modulatory). Pairwise comparisons were used for post hoc analyses of significant between-group differences, applying Bonferroni correction for the three groups.