ABSTRACT

Sometimes psychologists want to perform an experiment with more than one block of observations of each subject and with each block having as many stages as there are treatments to compare (McBurney, 1983; Shaughnessy & Zechmeister, 1990). Then each treatment is given once in each block for each subject. This approach is attractive because it ensures that each subject has the same amount of experience with each treatment in each block, possibly minimizing contamination of estimates of treatment effects by time-related effects. It is tempting to find a mean for each treatment for each subject and treat these means as raw data for analyzing treatment effects. One might also find a mean for each stage of the experiment and examine those means for evidence of practice effects. Cotton (1994) has warned, however, that failure to do a full analysis of data from this design may cause distortions in the conclusions one draws from those data.