Previous tests of P300 in deception detection have focused mainly on amplitude analysis. Since countermeasures for such tests have been reported, we looked here at other possible variables as deception indices: P300 scalp distribution and amplitude variance, both across 30 sites. We were also concerned, for the first time, in testing for recognition of more than one guilty item in a mock crime scenario. There were three groups: (1) two-probe group, two of six items were guilty knowledge (GK) items; (2) three-probe group, three of six items were GK items; (3) control group, zero of six items were GK items. In group analyses, in the two-probe group, P300s for lies were significantly greater than P300s for truthful responses. There were significant interactions of condition (Lie vs Truth) by site, suggesting different scalp profiles for deceptive versus truthful responding. Amplitude variance across sites was also greater in Lie than in Truth blocks. These results did not obtain in the three-probe and control groups. In terms of amplitude variances in probe conditions across groups, two-probe group was larger than three-probe and control groups. Regarding individual diagnostics, the variance method yielded a greater-than-chance detection rate of 71% versus 28% false positives. Regarding amplitude at multiple nonfrontal sites, 71% of guilty subjects were detected versus 14% false positives. Grier's (1971) A′ indices of various test discrimination efficiencies varied from .76 to .87. Results of the present study suggested further investigation of the variance method as a diagnostic tool for lie detection.