PFC interneurons, hippocampal cells, and putative VTA GABAergic cells did not show such task-dependent phase locking. The fraction of significantly buy PFT�� theta phase-locked neurons in each brain region was similar in the two tasks (Figure 5). These findings show that the magnitude of 4 Hz modulation of PFC pyramidal cells and VTA dopaminergic cells was task specific. Although changes in 4 Hz and gamma oscillations in both PFC and VTA were reliably correlated with the working-memory component of the task, they did not predict the directional choice of the animal on a given trial (p > 0.05 for

each rat). In contrast, a sizeable fraction of both PFC and hippocampal cells fired at significantly different rates in the central arm on trials with future left and right turns (Figure 6A). Because such goal-predicting, “prospective,”

or “episode” neurons have been suggested to be a part of the working-memory network (Wood et al., 2000, Frank et al., 2000, Fujisawa et al., 2008 and Pastalkova et al., 2008), we examined their relationship with LFP oscillations and compared them with nonpredicting but active neurons in the 0.0–0.3 segment of the central arm, in which movement trajectories and speed during left and right trials were indistinguishable (Fujisawa et al., 2008). For these comparisons, only neurons that fired at least 1 Hz in the central arm were included. The fraction of both 4 Hz and theta-modulated neurons was significantly higher in the goal-predicting PFC pyramidal cell group (Fujisawa Galunisertib mouse et al., 2008), as compared to nonpredicting cells (Figure 6B). Furthermore, below the magnitude of phase locking of the goal-predicting PFC pyramidal cells to 4 Hz and theta oscillations was also significantly higher than phase locking of the nonpredicting neurons (Figure 6C). This difference could not be explained by the significantly lower firing rate of nonpredicting neurons

in the central arm because the phase-modulation differences persisted after equalizing the spike numbers of all predicting and nonpredicting neurons for the analysis, using exhaustive bootstrapping (Vinck et al., 2010; Figure S6). Although the fraction of predicting neurons in the hippocampus and PFC was similar, hippocampal predicting and nonpredicting neurons did not show differential phase locking to 4 Hz or theta oscillations (Figure S6). Because timing of neuronal spikes was biased by both 4 Hz and theta rhythms, we also examined their joint effects. First, we tested whether there is a phase relationship between these oscillations. The PFC 4 Hz trough (180°) was significantly locked to the trough (198.5° ± 3.78°) of CA1 theta waves in each rat (Figure 7A; p < 0.01; shift predictor statistics).