Four experiments were carried out to examine the effects of statistical properties of discrete time series graphically displayed on a computer display on human prediction. Experiment 1 showed that instead of the standard deviation, the normalized integrated absolute autocorrelation function, which was a measure of the periodicity of time series, influenced human prediction of future values of a time series. Errors of human prediction increased with decreasing values of normalized integrated absolute autocorrelation function. The results of the paired comparison in Experiment 2 and 3 showed that values of normalized integrated absolute autocorrelation function influenced substantially the predictability of future values of time series, while the standard deviation affected secondarily human prediction, when the values of the function were the same. The statistical properties of the time series produced by subjects in Experiment 4 supported the relevance of normalized integrated absolute autocorrelation function as a measure of the effects of statistical properties of time series on human prediction.