SPATIOTEMPORAL CORRELATIONS IN HUMAN GAMMA-BAND ELECTROCORTICOGRAMS

Animal electrocorticogram (ECoG) studies have shown that spatial patte rns in the gamma band (> 20 Hz) reflect perceptual categorization, Spa tio-temporal correlations were investigated in the 20-50 Hz range in s earch for similar phenomena in human ECoG, ECoGs were recorded in a so matosensory discrimination task from 64-electrode subdural grid arrays , with inter-electrode spacing of 1 cm, overlying somatosensory, motor and superior temporal cortices in 2 patients with intractable epileps y. Bootstrap techniques were devised to analyze the spatial and tempor al characteristics of the correlations. Despite an extensive search, n o evidence was found for globally correlated activity related to behav ior either in narrow (i.e., 35-45 Hz) or broad (i.e., 20-50 Hz) bands, Spatial patterns, extracted using principal component analysis, could not be classified with respect to stimulus type in any time interval. Instead, spatially and temporally intermittent synchronization was ob served between pairs of electrodes in 1 cm X 1 cm regions with high va riability within and across trials, The distribution of correlation co efficients differed substantially from background levels at inter-elec trode distances of 1 cm and 1.4 cm but not 2 cm or more, The minimum d uration of correlation, the decorrelation time, of the ECoG was about 50 msec; the average correlation duration at 1 cm inter-electrode dist ance was about 150 msec; and the recurrence rate of significant correl ation peaks was about 1.3/sec. The findings suggest that the surface d iameters of domains of spatially correlated activity underlying percep tual categorization in human gamma band ECoG are limited to less than 2 cm and that the intermittent synchronization observed across separat ions of 1 cm and 1.4 cm is not solely due to volume conduction. Thus, if such gamma band spatial patterns exist in the human brain, no exist ing technology would be capable of measuring them at the scalp, and su bdural electrode arrays for cortical surface recording would have to h ave spacings under 5 mm.