The study explored a taxonomy of alternative latent models to understand the weight of expectancy-value and approach-related goal orientations in predicting students' achievement behaviors (scientific career interest and math achievement). A full model where all the factors contribute to explaining the total variance of outcomes and two models in which the contribution of expectancy-value and achievement goals factors was evaluated separately within a structural equation modeling framework. Results, among a sample of 812 Italian high school students (486 males and 326 females, "M"age= 18.3), showed that expectancy-value variables explain a substantial portion of math achievement and career interest variance after controlling for the achievement goals. Moreover, when all path coefficients were allowed (full model), global value and its dimensions of opportunity and emotional cost influenced positively student's career aspirations, expectancy resulted the main predictor of math competence, and mastery and performance goals did not show significant effects on the outcomes. [For the complete proceedings, see ED608557.]