The analysis of small group data with hierarchical linear models is discussed, concentrating on the usefulness and reliability of such analyses using data reported by N. M. Webb (1982). Results of Webb's analyses for 96 junior high school students in small groups are compared with results obtained with random effects linear models for the analysis of hierarchically nested data with the VARCL computer package. It is concluded that the traditional linear model used by Webb produces regression estimates that are very close to the ones produced by random effects models. Based on these results, it is argued that the use of random effects linear models does not always produce the strikingly different results that some applications of these techniques suggest, as shown by S. H. Raudenbush and D. J. Willms (1991). The regression coefficients of fixed effect linear models are fairly robust in data analysis of small groups, even when assumptions of the fixed effects linear model are clearly violated. The difference between fixed and random effects linear models is mainly in the opportunities to test more complicated models with the latter. The difference between fixed versus random models is not in the estimates of the regression coefficients; it is in the conclusions reached. Using the same data set, different conclusions are reached than those drawn by Webb. Choosing between traditional fixed effects linear models and random effects linear models is discussed. Four figures, 3 tables, and a 21-item list of references are included. (Author/SLD)