This study was designed to compare the performance of liquid-based preparation from the AutoCyte PREP with the conventional cervicovaginal smear in masked split-samples. In randomly selected 840 cases, the conventional smear was always prepared first, and the AutoCyte PREP used the resldual cells on the collecting device. Parallel AutoCyte PREP slides and matched conventional smears were screened in a blind fashion. All abnormals and 10% random normal cases were reviewed by two pathologists in a blind fashion. The Bethesda System was used for reporting the diagnosis and specimen adequacy. The diagnoses from the two methods were agreed exactly in 767(91.3%) of 840 cases. The AutoCyte PREP demonstrated a 25% overall improvement in the detection of squamous intraepithelial lesion(SIL). The ratio of ASCUS to SIL was decreased as 0.45 compared with 1.00 of conventional smear. The AutoCyte PREP produced excellent cellular preservation and superior sensitivity for detection of atypical cells as compared to the conventional smear. It makes us to be able to subclassify ASCUS into from WNL to HSIL. We thought that the AutoCyte PREP method might contribute to increase the detection rate of abnormal cells than conventional methods.