Detecting subtle differences in behavior using waveform display analysis

We have devised a method, behavioral waveform display analysis, to analyze complex ethological information by measuring behavior in real time and visualizing it as a time-dependent, multistate waveform. To facilitate the generation and statistical analysis of behavioral waveform displays, we have designed a simple Macintosh-based software program. When keystrokes coded to particular behavioral states are entered in real time, this software measures and collates the time, frequency, and duration of each behavioral state. These data can then be displayed either in a tabular format for statistical analysis of behavioral duration and frequency or as graphical coordinates for creating waveform displays by direct importing into graphing programs. An illustration of the use of waveform display analysis to detect anomalous behaviors in cocaine- and amphetamine-treated mice, some of which are not detectable by a standard time-sampling assay, is shown. Both waveform display and time-sampling analysis detected drug-induced changes in sniffing, bar hanging, digging, and rearing. However, unlike time-sampling analysis, waveform display analysis also detected. changes in the total duration, frequency, and average duration of these behaviors as well as additional changes in gnawing and locomotion. Additionally, visual scanning of behavioral waveform displays detected drug-induced changes in the patterns of behavior not detectable by time-sampling, including 1) a staged progression to a limited behavioral repertoire consisting of sniffing, locomotion, and rearing; 2) rapid switching between these remaining few behaviors; 3) a delayed onset of postinjection I:earing relative to sniffing and locomotion; and 4) the absence of other transient stereotypies during the onset of drug action. These data Indicate that behavioral waveform display provides an approach for the detection, visualization, and statistical analysis of aspects of complex behavior not amenable to detection by time-sampling methods. (C) 1998 Elsevier Science Inc.