(A) Schematic representing the cross-temporal pattern analysis. Each pixel represents a correlation coefficient between two population vectors. Entries into the vectors contain each neuron’s chosen value regression coefficient at Time T and at Time T + δt. If the chosen value correlates are consistent across the neuronal population at the two distinct time points, there will be a strong cross-temporal correlation (red colour). At two points close in time, chosen value correlates of each neuron will inevitably be similar. If these correlates are consistent for only a short period of time, there will be a transient population code; whereas if each neuron’s chosen value correlate is consistent for a prolonged period, there will be a sustained population code. If each neuron within a population correlates with chosen value at two separate points of a trial (e.g. choice and outcome), in the absence of sustained coding bridging the two, there is a reactivation population code. (B) Cross-temporal pattern analysis of OFC neurons (n = 214). There is clear evidence for sustained coding of the chosen value at choice (top left), as well as before and throughout outcome (bottom right), reflected by strong correlations extending off the diagonal of the plot. Blue lines indicate a significant area of cross-correlation (p<0.05, see Materials and methods). There is also sustained coding of the chosen value signal from choice through outcome, shown by a strong cross-temporal correlation both prior (grey dashed box) and during reward (black dashed box). Within the dashed areas, blue lines indicate a significant area of cross-correlation (p<0.05, see Materials and methods). (C and D) The black dashed inset (bottom left quadrant in B) is then performed in high (C) and low (D) time constant OFC neurons separately. The sustained coding is present specifically in high time constant cells (largest cluster of cross correlation, p=0.0002), but absent in low time constant cells (p=0.2248; permutation test, see Materials and methods). See also Figure 5—figure supplement 1 and 2: Sustained chosen value correlates are present at choice and outcome within DLPFC and ACC, but sustained coding from choice through outcome is absent. Sustained coding between choice and outcome was much stronger in OFC than in DLPFC or ACC (permutation tests; OFC v DLPFC, p=0.0008, OFC v ACC, p<0.0001, see Materials and methods).