Spike rates were computed by binning spike timestamps in non-overlapping 50 ms windows. Spike rates were square-root transformed prior to analysis to convert their Poisson-like distributions to approximately normal. LFPs were re-referenced offline to remove any common-source noise by subtracting off the mean across all electrodes in each array. Evoked potentials were removed by subtracting off the mean signal across trials within each condition (object, upper/lower location, and visual hemifield). Thus, all of our analysis is on the remaining induced component. For most analyses, the resulting signals were convolved with a set of complex Morlet wavelets (wavenumber 6). LFP power was log-transformed to render its distribution approximately normal.
