The statistical analysis of data from a single-subject design is somewhat controversial. The procedure most often chosen to examine data from a single-subject design involves the visual inspection of the graphed outcome variable over time. Problems associated with this procedure are discussed, and questions about the use of traditional statistical tests are reviewed. To supplement visual inspection of the data, the non-parametric smoother proposed by J. W. Tukey (1977) is presented as an appropriate and useful technique for interpreting such data. Two working examples are presented with single-subject data from: (1) baseline observations of the level of a particular hormone; and (2) observations of the behavior of a subject. Four graphs illustrate these working examples. A 46-item list of references is included. (SLD)