We tested the hypothesis of a role of gamma-range synchronized oscillatory activity in visual feature binding by recording evoked potentials from 12 subjects to three stimuli: two coherent ones (a Kanizsa triangle and a real triangle) and a non-coherent one (a Kanizsa triangle in which the inducing disks had been rotated so that no triangle could be perceived). The evoked potentials were analysed by convoluting the signal for each subject and each stimulation type by Gabor wavelets centred from 28 up to 46 Hz, providing a continuous measure of frequency-specific power over time. A first peak of activity was found around 38 Hz and 100 ms with a maximum at electrode Cz in each experimental condition. A second peak of activity occurred around 30 Hz and 230 ms, with a maximum at O1 in response to the real triangle and a maximum at Cz in the case of the illusory triangle. At 100 ms we did not find any variations of the gamma-band component of the evoked potential with stimulation type, but the power of the 30 Hz component of the evoked potential between 210 and 290 ms differed from noise only in the case of a coherent triangle, no matter whether real or illusory. We thus found a 30 Hz component whose power correlates with stimulus coherency, which supports the hypothesis of a functional role of high-frequency synchronization in feature binding.(ABSTRACT TRUNCATED AT 250 WORDS)