Background: Sleep-wake states (SWS) affect the expression of interictal epileptiform discharges (“spikes”), which affects resultant source localization calculations used in epilepsy evaluation. We hypothesize that spike localizations from non-REM sleep 1-3 are most concordant with one another. Methods: We used Standardized low-resolution brain electromagnetic tomography (sLORETA) in Curry 8 software to calculate source localization voxels of spikes in N1-3, REM, or wakefulness (W). We assessed voxel concordance between N1-N2-N3/N1-N2-W/N1-N3-W/N2-N3-W/REM-N1-N2/REM-N1-N3/REM-N2-N3/REM-N1-W/REM-N2-W/REM-N3-W. We classified concordances into those containing and not containing a SWS (e.g. N1 vs. not-N1 = N1-N2-N3/N1-N2-W/N1-N3-W/REM-N1-N2/REM-N1-N3/REM-N1-W vs. REM-N2-W/REM-N3-W/REM-N2-N3/N2-N3-W) for comparison. Results: Concordances did not differ for N1-3 or W. However, concordances with REM were lower than those without REM as a fraction of source localization space (median 32.1% vs. 56.1%, p<0.001) and cortical grey matter (median 20.4% vs. 27.3%, p=0.003). Conclusions: As expected, source localizations from spikes in N1, N2, and N3 did not significantly differ from one another because these three states are constituent members of non-REM sleep. Surprisingly, however, source localizations derived from awake spikes – not a constituent of non-REM sleep – also did not differ. In contrast, REM was most different by reproducibly exhibiting the least three-way concordance. These findings reinforce the unique localizing ability of REM sleep.