Directional causal influences between LFPs in the two prefrontal hemispheres were measured using bivariate nonparametric spectral Granger causality (Dhamala et al., 2008). Traditional Granger causality quantifies how much variance in one signal can be explained by the recent history of another signal, beyond what can be explained by the history of the signal itself. Spectral Granger expresses these causal interactions in the frequency domain. Unlike traditional parametric causality measures, it is estimated directly via factorization of the cross-spectral density matrix, without relying on estimation of a specific autoregressive model. For this analysis only, LFP spectra were computed around the putative trace transfer period (–1 to –0.5 s relative to delay end) via the multitaper method (4 Hz frequency bandwidth, 3 dpss tapers). Trials were balanced across swap and no-swap conditions in each session by sampling a random subset of trials from the larger group. Presented results represent averages across all between-hemisphere electrode pairs within each session, and across all sessions. As a control, this analysis was also performed on data reversed in time (Figure S4). This is expected to reverse
