The objectives of this study are to examine both the significance of the impact of farmers' perceptions regarding new technology for the adoption decision and how perceptions themselves are influenced by the decision to adopt new technology. The study is based on data from 96 wheat farms in the Moret and Jiru woreda (district) of Ethiopia. The probit approach is used to analyse the adoption decision, while the variables relevant to farmers' perceptions are modelled using the ordered probit methodology. A simultaneous equations model combining the probit and ordered probit approaches provides a useful approach to modelling the two-way relationship between perception and adoption. Variables such as farm size, farm income and soil type have a key role to play in the model. Perception is measured by eight different components and as there is a strong collinearity among these various measures, a principal components analysis is attempted to draw the best possible linear combination of variables. © 1999 Elsevier Science B.V. All rights reserved.