Six subjects who never wore contact lenses underwent six determinations of corneal cell density using each of two measurement modes: Mode A: the eye was applanted once and pairs of photographs were taken; Mode B: the eye was applanated and a single photograph was taken. For each measurement mode data were collected on two separate sessions (Test, Retest). Chi-square analysis revealed a significantly larger proportion of uncountable photographs from the first measurement session, and analysis of variance revealed that Mode A was less reliable than Mode B upon retesting. It was concluded that Mode B was the preferred measurement mode, and that many more than six determinations be made on the initial session in order to obtain six readable photographs. Finally, estimates of inter and intra-session variation, intra-pair variation, and inter and intra-observer variation are given.