This paper discusses the problems of gathering and analyzing data on classroom behaviors. The difficulties involved in using the common pretest-posttest experimental design are described. To grapple with these problems, a study was conducted on the effect of environmental changes on student behavior, employing a time-series design and time-series analysis. The hypothesis tested was that minor changes in the physical setting of the classroom could produce predictable, desirable changes in students' behavior. Formal observations of students' locations and activities were conducted six times a day for two weeks, using a time-sampling-by-child instrument. The data from the six daily observations of all children were reduced to daily summary statistics indicating the percentage of time children were observed in a particular area or engaged in a specific activity. The data were then analyzed using time-series techniques. Results indicated that most of the desired behavior changes were produced, thus supporting the hypothesis. The same experiment was conducted using the traditional statistical methods. The conclusion was reached that time-series techniques appear to have great potential for educational research. (JD)