In vivo reflectance confocal microscopy shows promise for the early detection of malignant melanoma. One diagnostic trait of malignancy is the presence of pagetoid melanocytes in the epidermis. For automated detection of MM, this feature must be identified quantitatively through software. Beginning with in vivo, noninvasive confocal images from 10 unequivocal MMs and benign nevi, we developed a pattern recognition algorithm that automatically identified pagetoid melanocytes in all four MMs and identified none in five benign nevi. One data set was discarded due to artifacts caused by patient movement. With future work to bring the performance of this pattern recognition technique to the level of the clinicians on difficult lesions, melanoma diagnosis could be brought to primary care facilities and save many lives by improving early diagnosis.