Inter-rater reliability [IRR] of sleep scoring is currently based on AASM guidelines stipulating the use of percent agreement comparisons of event counts at the gross epoch level, irrespective of within-epoch event matching, precluding identifying errors as omissions or commissions. The present study compares a new reliability system based on computerized within-event event matching against the current standard on an epoch-wise basis in an apples-to-apples comparison. The hypothesis is that the new method would show the current standard was less reliable then normally assumed. In routine quarterly reliability testing, 50 sleep technologists scored 201 epochs from each of 9 selected polysomnograms for apneas, hypopneas, leg movements, and arousals, as compared to a physician(CR) gold-standard scorer. An event-matching protocol determined whether events were closely matched. Total and event-matched counts per epoch for each scored epoch were compared to the gold standard scorer using McNemar’s test with Obuchowski cluster-adjustments.. The new method also separated omission errors from commission errors. In no case was reliability better in the standard method compared to the new method. Over all the within-epoch events, discordance between methods was observed in at least 1 epoch for 32.2% of studies. Apnea events showed no discordance, but 14.8% of studies had overall hypopnea discordances. Agreement rates differed in 4 of 9 PSGs (p&It0.05) for hypopneas, 6 of those (p<0.01) for arousals, and 1 of those (p=0.04) for limb movements. Descriptively, 12.4% of epochs showed an omission error for at least one event, and 7.9% showed a commission error, with some scorers varying in relative rates in the kinds of error made. The event-matching method documented a key methodological flaw in the current AASM-standard approach that relies only on within-epoch event counting using percent agreements for documenting inter-scorer reliability. The new method improves reliability discrimination while producing output that will give more actionable feedback to scorers about kinds and locations of errors in their scored records, compared to a standard scorer. none