Data on a sample of 128 face-to-face voluntary groups are analyzed with the multitrait-multimethod approach. The traits measured include four characteristics: size, sex composition, age composition, and educational composition. These four traits are measured with three different methods: reports from a respondent chosen through a probability sample from the community, reports from an official of the group, and estimates from direct measurement of the traits at an organizational meeting. Results of the analysis suggest that the three measures are parallel with one another in the sense of Lord and Novick 1968. This result means that the three methods are equally desirable. Surprisingly, the direct canvass measure does not appear to be more reliable than the others; on the contrary, it appears to be equivalent to other reports for the traits measured. There is a slight suggestion in the data that officials in the group are more accurate than ordinary respondents in reporting information about less salient characteristics of the members, such as years of education. In general, the three methods of gathering data appear to be equally reliable, although they measure slightly different aspects of group composition.