Temporal stability of a given measurement within individual participants is a desirable property of psychophysiological measures. The present study aims to examine the reliability of high-frequency oscillatory activity in the human electroencephalogram (EEG) across 4 sessions. Convolution of the EEG time series with Morlet wavelets yielded time-frequency representations of the signal for each session. Stability of both topography and time course of gamma-band activity (GBA) was determined for two participants performing a feature-based selective attention task in four separate sessions, spaced at weekly intervals. We found high temporal stability of non-phase-locked GBA typically occurring in time ranges between 200 and 500 ms following presentation of a stimulus, both in terms of topography and time course. Early phase-locked GBA (80-120 ms) showed higher variability with respect to topography, but was consistent in terms of time course. We conclude that measures of high-frequency oscillatory activity as used in the cognitive neurosciences meet stability requirements necessary for meaningful interpretation of this parameter of brain function.