Eye movements are systematic in the active but not the passive session. A linear support vector machine was trained to classify stimulus orientation from eye movements. The classification accuracies (i.e., percentage of correctly classified trials) are shown as distributions (top), box plot (middle), and individual data points (bottom) for both active (green) as well as passive (orange) session. The classification accuracies were significantly larger in the active (median = 19.2%) than in the passive (median = 12.1%) session (t = 4.0, n = 28, p < 0.01, Wilcoxon signed-rank test for two related paired samples). The gray dashed line illustrates chance level (12.5 percent). For detailed results see main text.