| Abstract: | In social psychological experiments, the manipulations of interest are often presented to subjects along with, or as part of, some stimulus. An example would be a manipulation of the verbal label associated with a stimulus photograph, in a person perception study. Ordinarily it is not possible for stimuli to be completely crossed with treatments because subjects cannot be exposed to any stimulus more than once. In such designs it is wise to counterbalance the assignment of stimuli to treatment conditions, but this gives rise to difficulties in the analysis of the experimental data. Data from such designs have frequently been misanalyzed in the published literature. This paper presents a method of constructing a counter-balancing scheme to simplify the analysis, and appropriate methods of analysis for both the simplified and the general case. It is emphasized that stimuli are generally best treated as a random factor in such designs, permitting increased generalizability, but the case where the stimulus factor is fixed is also considered. |
