Human subjects typically keep about seven items (plus or minus two) in short-term memory (STM). A theoretical neuronal model has been proposed to explain this phenomenon with physiological parameters of brain oscillations in the gamma and theta frequency range, i.e., roughly 30-80 and 4-8 Hz, respectively. In that model, STM capacity equals the number of gamma cycles (e.g., 25 ms for 40 Hz), which fit into one theta cycle (e.g., 166 ms for 6 Hz). The model is based on two assumptions: (1) theta activity should modulate gamma activity; and (2) the theta/gamma ratio should correlate with human STM capacity. The first assumption is supported by electrophysiological data showing that the amplitude of gamma oscillations is modulated by the phase of theta activity. However, so far, this has only been demonstrated for intracranial recordings. We analyzed human event-related EEG oscillations recorded in a memory experiment in which 13 subjects perceived known and unknown visual stimuli. The paradigm revealed event-related oscillations in the gamma range, which depended significantly on the phase of simultaneous theta activity. Our data are the first scalp-recorded human EEG recordings revealing a relationship between the gamma amplitude and the phase of theta oscillations, supporting the first assumption of the above-mentioned theory. Interestingly, the involved frequencies revealed a 7:1 ratio. However, this ratio does not necessarily determine human STM capacity. Since such a correlation was not explicitly tested in our paradigm, our data are not conclusive about the second assumption. Instead of theta phase modulating gamma amplitude, it is also conceivable that focal gamma activity needs to be downsampled to theta activity, before it can interact with more distant brain regions.