Single - Case Experimental Designs

Single-case experimental designs are characterized by repeated measurements of an individual's behav­ ior, comparisons across experimental conditions imposed on that individual, and assessment of the measurements' reliability within and across the con­ ditions. Such designs were integral to the develop­ ment of behavioral science. Early work in the field of psychology depended on the analysis of the expe­ riences of one or a few individuals (Ebbinghaus, 1885/1913; Thorndike, 1911; Wertheimer, 1912). The investigator identified a phenomenon (e.g., learning and memory, the law of effect, the phi phe­ nomenon) and pursued experimental arrangements that assessed its reliability and the functional rela­ tions among the pertinent variables (e.g., the rela­ tion between the length of a series of nonsense syllables and learning curves, recall, and retention; the relation between the consequences of behavior and the rate of the behavior; the relation between an observer's distance from blinking lights and appear­ ance of movement). Because the research was con­ ducted on the investigators themselves (e.g., the memory work of Ebbinghaus) or on just a few par­ ticipants (e.g., Thorndike's cats and Wertheimer's human observers), the experimental arrangements often involved intensive study, with numerous mea­ surements of behavior recorded while each individ­ ual was studied under a variety of conditions. Only after the development of statistical methods for analyzing aggregate data did the focus shift to comparisons across groups of participants, with each group exposed to a single condition (see also Chapter 8, this volume). In the original case, the "participants" were plants in fields split into plots. The statistical methods were developed to assess the significance of differences in yields of plots of plants treated differently. R. A. Fisher's (1925) Statistical Methods for Research Workers set the course for the field. Fisher began development of his methods while employed as the statistician at an agricultural experiment station early in his career. The fact that data on large numbers of participants tend to be normally distributed (regardless of whether the par­ ticipants are plants, people, or other animals) led to the easy adaptation of group statistical methods to research with humans. The standard practice came to emphasize the importance of group means, differ­ ences in these means, and the use of statistical tests to draw inferences about the likelihood that the group differences were representative of differences in the populations of interest (e.g., Kazdin, 1999; Perone, 1999). Despite the rise of group statistical methods, Single-case designs continued to be used in some important work because they allowed the investiga­ tor to study the details of relations among variables as expressed in the behavior of individuals (e.g., Bijou, 1955; Skinner, 1938; Watson, 1913), which resulted in reasonably clear demonstrations of func­ tional relations among the variables being studied (e.g., conditioned startle responses, reinforcement, and schedules of reinforcement). Articulation of the necessary elements of Single-case designs, notably in Sidman's (1960) seminal Tactics of Scientific Research, helped make the designs practically de rigueur in basic research on free-operant behavior