This paper advances Williams and Thompson (1984) by updating their work and by explicitly accounting for price and yield risk in the analysis of acreage response in Brazil for soybeans and by assessing model specification. Empirical equations were estimated using seemingly unrelated regression (SUR). The robustness of the model was evaluated in the battery of misspecification tests suggested by McGuirk et al. (1993) and McGuirk et al. (1995). The results of the testing procedure suggest that the model is fairly robust in terms of normality, heteroscedasticity and functional form. The results point to parameter instability in the soybean model. The approach to the problem of parameter instability involved dividing the data in two periods and estimating regressions for each period. The signs of the significant coefficients were consistent with expectation, particularly for the second period. Soybean acreage is explained mainly by past acreage, expected prices of soybeans and land competing crops (cotton, rice, and corn), and price and yield risk. Results suggest that market signals played a reduced role in the soybean acreage growth in early years. In contrast, in recent years producers in Brazil became more sensitive to changes in prices and risk. Measures of short-run price elasticity of soybean acreage response are similar to the one obtained by William and Thompson (1984) for soybean supply. Long-run elasticities are significantly smaller.