An empirical assessment is reported of Fishbein’s model (Fishbein, 1967c; Fishbein & Ajzen, 1975) for predicting behavioral intentions (I) from attitudes toward the behavior (Aв) and the subjective (social) norms (SN) perceived to govern that behavior: I = Aв (W₁) + SN (W₂) where W₁ and W₂ are empirically determined weights (usually standardised multiple regression coefficients). The model was tested by measuring behavioral intentions, the components of the model, and other related variables, for two different issues. The first issue was the (future) learning of statistics by introductory psychology students, some of whom were already learning statistics and some of whom were not. Measures were taken on three different occasions, that is, under three different conditions for the learning group: i) immediately prior to the commencement of their statistics course; ii) at the end of the final statistics lecture but prior to examination; and iii) after the examination results were known. For the second issue, a representative sample of the general population who lived close to a proposed nuclear power station site indicated their support or opposition to the introduction of nuclear power into New Zealand and their intentions to engage in behavior in support of their views. It was hypothesised that: i) the various components of attitude discussed by Fishbein -- attitudes toward the object, attitudes toward the behavior, subjective (social) norms, the motivation to comply with these norms, and personal normative beliefs -- should be conceptually and factorially independent from one another; ii) the weight assigned to the components of the model should be different for different groups of people and for different conditions of measurement; iii) other two-components (one attitudinal and one normative) models would sometimes give better prediction of behavioral intentions than Fishbein’s model; and iv) the model would not always be sufficient to explain behavioral intentions, that is, variables external to the model would sometimes have a significant influence on behavioral intentions over and above that made by the components of the model. Generally, these hypotheses were supported, with the results for hypotheses iii) and iv) indicating that Fishbein’s model was not adequate and needs to be further reconsidered and developed. It was also found that: a) personal normative beliefs, which had been dropped from the model by Fishbein and his associates, was often a critical variable for predicting behavioral intentions and was not “simply an alternative measure of behavioral intentions”; b) the motivation to comply was an alternative measure of subjective norms and did not improve the prediction of behavioral intentions; c) other variables which improved prediction of behavioral intentions beyond that achieved by Fishbein’s model were mostly ego-involvement, arousal, or self-relevance type variables, which suggests that the role of this type of variable in influencing behavioral intentions needs to be studied further; and d) the causal links within a model can be different for different groups of people or different conditions, and adding variables to a model may change the relationships between variables and so change the whole model. It was concluded that much more work needs to be done in the attitude-behavior domain in the search for a parsimonious, but nevertheless more complex, model for predicting behavioral intentions and ultimately overt behavior.