The theory of planned behavior suggests attitudes are a product of salient beliefs. This study examined whether aggregating salient beliefs was plausible within a more biologically centered information processing environment. A neural network was used to examine associations among beliefs relating to exercise intention. Data on intentions and behavioral, normative, and control beliefs from 114 respondents were used to train (by error backpropagation) a neural network to associate beliefs with intention. The R 2 between the network's estimated and self-reported intention was .66. The network's representation comprised 6 belief profiles associated with high, moderate, or low behavioral, intention. The neural network accommodated complex relationships among beliefs and belief-intention associations and indicated how high-level constructs such as attitudes may be viewed as the best fit (compromise state) between aroused beliefs.