Three methods currently in use in South Africa to estimate television audiences are explained. In order to obtain comparable data , the periods of fieldwork for the three methods had to be matched as closely as possible. The samples used for the comparison of data from the three methods had to be filtered from larger samples where appropriate, so that they represented the same universe. The three sets of data are compared. Validation tests had indicated that the peoplemeter data was close to the 'truth'. Therefore, this data was used as the basis against which other data were compared. Of the three, the peoplemeter data provided the lowest audience estimates. There were substantial differences in the audience levels reported by the different methods, but the patterns of daily viewing were remarkably similar. Most of the differences between the peoplemeter levels on the one hand, and the interview and diary data on the other hand, were the result of the methods used to allocate respondents to viewing time periods.