The measurement and integration of behavioral variables: aggregation and complexity as important issues.

The measurement of motor activity in animals is central to the description of animal behavior in general. A wide variety of measurement procedures have been used in the past and various researchers have suggested that multiple measures or "behavioral profiles" be used to describe the effects of various types of experimental manipulations. However, many of the measures used have been found to be unreliable. The principle of aggregation, which states that the sum of a set of multiple measurements is a more stable and representative estimator than any single measurement, allows for construction of variables which are more reliable and have greater generalizability. Examples are provided which show that the application of aggregation can substantially increase correlations within variables over time as well as between variables. Application of factor analytic procedures to multivariate motor activity data is suggested to be a useful method of reducing many variables into fewer, more complex variables, which have greater phenomenon realism. Examples of the application of such factor analytic approaches are provided, both for the traditional two-way factor analytic methods as well as a more recent procedure (PARAFAC model) developed for analysis of three-way longitudinal data sets. It is suggested that multivariate analytic procedures are appropriate for data reduction and description in the area of behavioral neuroscience.