A) A representative comparison of recorded action potential waveforms, before and after the first lesion in Monkey U (top) and the sixth lesion in Monkey H (bottom). The location of the lesion electrodes are marked by black dots, showing visible changes in waveforms with low spatial specificity across the array. Signal was observed in the recording sessions immediately before and after lesioning (left, right). An action potential detection rate was determined from both periods of rest and task engagement (gray-scale shading, capped at 5 Hz for visualization). B) The multineuron activity recorded on each electrode was analyzed for changes in the proportions of activity. Action potential waveforms from a selected day (black) were compared to 1000 waveforms from a previous day (grey) by projecting the selected day’s waveforms into the top two principal components determined from the previous day. The median radius of all waveform points relative to the origin in these two dimensions are represented as circles for both the comparison and original day. The difference in radii, Δr, was computed for each channel and for all pairs of recording days, separated by no more than four days to ensure minimal rates of spontaneous turnover noted in the literature Gallego et al. (2020). Changes in radius could arise through changes in the relative proportion of activity among the recorded neurons, as well as putative neuron loss (r ≪ 0) and gain (r ≫ 0). C) The day-to-day pairwise comparisons fall within three groups: pre-lesion days (pre-pre), pre-lesion versus post-lesion days (pre-post; up to three days post-lesion), and post-lesion days (post-post; four to seven days after a lesion). In all comparison cases, 24, 48, 72, or 96 hours separated the recording sessions. Distribution of the Δr values for all channels and days are shown for each group from Monkey U. Central panels Gaussian mixture models were fit to the data and the optimal number of components for each was determined using the Bayesian information criterion (BIC, normalized arbitrary units). An entirely new cluster is identified for the pre-post group that is hypothesized to largely represent the loss of neurons from the local population, beyond the usual rate of turnover observed in pre-lesion conditions. Right panel The three groups were then compared by looking at the percentage of the 96 electrodes that matched across comparison days. The percentage of matching neurons dropped significantly after a lesion (median test; * < 0.017, corresponding to the conservative Bonferroni corrected significance threshold of 0.05 for the three comparisons). D) The same analysis is performed for Monkey H, which yielded two cases: lesions consistent with Monkey U (lesions 2, 4, and 7; left) and those with high levels of turnover before and after injury (lesions 5, 6, 8, 9, and 10; right). Note, lesions in Monkey U were well-spaced out over three months and considered as independent samples. Lesions in Monkey H were performed in much quicker succession, which likely contributes to the discrepancy.