Theme is built on a unique algorithm that searches for hidden repeated patterns in behavior and interactions, based on a model of the temporal organization of behavior. It considers both the order and the time distances between behavioral event types as well as hierarchical organization. The software behaves as an evolution program, in the sense that it detects complex patterns gradually as combinations of simpler ones and deals with combinatorial explosions through competition between patterns such that only the most complete patterns are retained (survive) while partial detections are discarded.
Theme can detect complex repeated patterns that are hidden to observers and very hard or impossible to detect with other available methods. The software includes various tools for the filtering and analysis of detected patterns on the basis of their frequency, complexity, structure, actor identity and behavioral content. Theme has been used extensively in studies of human communication, spoken dialogue, gestures, protocol analysis, etc. Other potential applications are the analysis of human-system interaction and navigation patterns in software user interfaces or web sites, detection of drug effects on behavior, behavioral disorders in human and animals, etc.
Theme detects statistically significant time patterns in sequences of behaviors. These so called T-patterns are based on the timing of events, relative to each other. This can give striking new insights into the structure of behavior that cannot be found with the naked eye or with any other method. Scientists have been conducting research with Theme for over twenty years, and Theme has been successfully used in a wide variety of disciplines (developmental psychology, psychiatry, etc). Theme can be used in conjunction with any of the solutions described on this website, or as a stand-alone package.
Timing and order are crucial to the effectiveness of nearly any kind of action, but they are often hard to capture in statistics. Theme is a software package that was especially developed to deal with this challenge. If two behaviors are linked by a certain time interval more often than expected by chance, Theme defines them as a T-pattern. This T-pattern analysis gives new and unexpected insights into the structure of behavior.
Theme is based on a well-established mathematical model. Due to this model Theme can detect much more structure than other methods (series of 20-30 events are no exception, where most methods stop at 3-5). The basis of Theme is a statistical test that finds significant time intervals between two series of events. The dyads found are tested against other events and dyads, and new connections are made. This is repeated until no new relationships are found. A special algorithm eliminates duplicate and incomplete patterns. In other words, Theme describes behavior in terms of time relationships between activities.
Theme is especially effective in the analysis of social interactions. Most behavior is highly organized in time but can also be quite variable. Contrary to other methods, Theme can detect and quantify the overall time structure in behavioral data without being distracted by the variation in details.
The results found by Theme are very robust. The probability of long T-patterns occurring by chance is extremely low in most data. Theme performs checks against chance effects. A rigorous randomization function checks which proportion of the patterns can be explained by chance. In most data that is a small subset of the shorter patterns.