The main objective of this research was to test the hypothesis of whether family agriculture, as defined herewith, would indeed be predominant in the Taquara Branca watershed, in the State of São Paulo, Brazil, site of a larger sustainable development assessment study. Data on 33 of 106 landholdings in the watershed were obtained by directly applied cross-section questionnaires. The variables for multivariate analysis were first selected through a unidimensional analysis, when some questions were selected with basis on answer variability. Multivariate statistics was used (PROC CLUSTER, SAS) to define and analyze the groupings. Twenty two respondents (out of the 33) were involved in agricultural activities, comprising, then, the universe of the study. Cluster analysis allowed the identification of three natural groups among the 22 landholdings. It was possible to classify each group, according to their characteristics, as follows: group 1 -- twelve typical family landholdings; group 2 -- six non-family, small landholdings; and group 3 - four non-family, large landholdings. Although not predominant in terms of area, typical family agriculture was found to be the more frequent (36%) in the watershed.
family agriculture; cluster analysis; sampling
