To normalize out any overall differences in activity between neurons and task conditions, we z-scored spike rates to the fixation baseline. Rates were mean-pooled across the 200 ms before sample object onset separately for each neuron and condition, and the mean and standard deviation across all within-condition trials was computed. These were used to z-score rates across all time points and trials for that neuron and condition. The same transformation was used for analysis of LFP power, except it was also computed separately for each frequency, in order to also normalize out the typical 1/f distribution of power across frequency.
