Spatial spectral analysis is essential for deriving spatial patterns from simultaneous recordings of electrocorticograms (ECoG), in order to determine the optimal interval between electrodes in arrays, and to design spatial filters, particularly for extraction of information about the dynamics of human gamma activity. ECoG were recorded from up to 64 electrodes 0.5 mm apart in a linear array 3.2 cm long, which was placed on the exposed superior temporal gyrus or motor cortex of volunteers undergoing diagnostic surgery. Visual displays of multiple traces revealed broad spectrum oscillations in episodic bursts having a common aperiodic wave form with recurring patterns of spatial amplitude modulation (AM patterns) on selected portions of the array. The one-dimensional spatial spectrum of the human ECoG was calculated at successive time samples and averaged over periods of up to 20 s. Log power decreased monotonically with increasing log spatial frequency in cycles/mm (c/mm) to the noise level approximately 2 log units below maximal power at minimal frequency (0.039+/-0.002 c/mm). The inflection point at 0.40+/-0.05 c/mm specified an optimal value for a low pass spatial filter to remove noise, and an optimal interelectrode spacing of 1.25 mm to avoid undersampling and aliasing. An 8 x 8 array with that spacing would be 10 x 10 mm.