This paper presents the application ofan overlapping sampling technique usedwith a fast Fourier transform on electroencephalographic data. The use of this sampling and analysis technique is demonstrated on data in which steady state, rate of change, and critical event hypotheses are tested. The benefit of applying computer-generated three-dimensional graphics to the results of using this innovative sampling and analysis technique is emphasized.We determined several years ago that there was an inordinate amount of variance in electroencephalographic (EEG) data (Bremner, Yost, & McKenzie, 1982). Although we believed that the data were gathered using valid techniques, we suspected that the large variances were a function of large inter-and intrasubject differences. We determined that there were large variances in the data gathered on a subject under a given experimental condition and that, when the data gathered on a set of 20 subjects under the same experimental condition were composited in a statistical analysis, the variance for the group more than doubled. Even though the differences in the means of two different groups of subjects under two different experimental conditions were "large," the variances kept us from obtaining statistically significant differences when we believed that the subjects' reactions to the different experimental conditions were significantly different. We were unsuccessful in overcoming the problem oflarge variances using different univariate, multivariate, and nonparametric techniques. Also, after examining several different methods of compositing and transforming data, we determined that each of these tended Address correspondence to Michael Yost, Office of the President, Trinity University, 715 Stadium Dr., San Antonio, TX 78284.to remove some of the information that was contained in the original data.We solved the problem oflarge variances by developing a single-subject experimental design and statistical analysis approach for our research (Bremner, Yost, and Pike, 1984;Bremner, Yost, and Zintgraff, 1985;Yost, Bremner, & Fox, 1985). This approach permitted us to gather data in which the variance was the variation in the response of a single subject under a given experimental condition. When we made comparisons of responses to different experimental conditions, the variances were a function of differences in the response of a subject and not differences among subjects. Using these approaches, we were able to reduce the variances and obtain what we considered to be valid and reliable experimental results.The fast Fourier transform (FFT) and a variety of smoothing functions are used in analyzing neuropsychological data. We examined the effect of these techniques in two studies (yost, Bremner, Helmer, and Chino, 1986;Yost, Cooper, & Bremner, 1983). One of the major conclusions of these studies was that the results obtained from the FFI of EEG data are influenced by where in the data the FFI is initiated. Moving the starting point for the FFI analysis by as little as 16 data points (wher...