Abstract
Our sense of every-day visual experience is continuous and seamless, yet neurophysiological and psychophysical evidence suggests that visual processing operates with an intrinsically paced regularity. In the laboratory, signatures of rhythmic oscillations emerge from analyses of EEG recordings and from analysis of behavioral performance (e.g., detection accuracy) measured at parametrically varied time points. Recently, we devised a procedure based on conventional spectral analysis that can disclose rhythms within incidences of perceptual dominance durations during binocular rivalry. With these kinds of data, it is impossible to control time elapsed until the terminal behavioral event. Instead, we relied on an enormous data set (average of 4,364 durations/condition) to recover probabilities of behavioral events at parametrically varied time points, allowing generation of probability density histograms. Such large amounts of data, however, are typically not achieved in human behavioral studies. To overcome this limitation, we have developed a computational procedure for analysis of rhythms (CPAR) that uses the differences in the amount of variance explained by a cumulative distribution function (CDF) and by a transformed CDF which incorporates rhythmic oscillations. To confirm the validity of CPAR, we generated two durations data sets using a modified random walk model: one set containing 200 durations/condition for CPAR, and the other set containing 1,000 durations/condition for spectral analysis. With these simulated data, CPAR achieved comparable results to those obtained using spectral analysis with a considerably smaller data set. We then analyzed publicly available human behavioral data (https://osf.io/q6vmd) with CPAR and with spectral analysis, demonstrating that CPAR outperforms conventional spectral analysis when analyzing human response time data containing just a few hundred durations per condition. We conclude that CPAR offers an efficient, sensitive means for extracting rhythms embedded in durations data.