This article extends the characteristics-based choice framework of technology adoption to account for decisions taken by boundedly-rational individuals in environments where traits are not fully observed. It is applied to an agricultural setting and introduces the concept of ambiguity in the agricultural technology adoption literature by relaxing strict informational and cognition related assumptions that are implied by traditional Bayesian analysis. The main results confirm that ambiguity increases as local conditions become less homogeneous and as computational ability, own experience and nearby adoption rates decrease. Measurement biases associated with full rationality assumptions are found to increase when decision makers have low computational ability, low experience and when their farming conditions differ widely from average adopter ones. A complementary empirical paper (Useche 2006) finds that models assuming low confidence in observed data, ambiguity and pessimistic expectations about traits predict sample shares better than models which assume that farmers do not face ambiguity or are optimistic about the traits of new varieties.