Performance expression of Western music is characterized by complex phenomena. Performers themselves find it difficult to objectively describe the causality of score directions and instrument operation. Thus, in this study, we propose a method that can help identify the factors contributing to the expression of performances. Specifically, we use a modeling method to describe the relationship between an expression and the factors causing it by associating the feature quantities of an expression with a combination of simple directions retrieved from the score. We then systematically structure contributory factors by classifying the expressive tendencies of these models using criteria related to score directions. The method also defines the degrees to which these directions contribute to performance expressions. In addition, it allows for the application of standard multivariate analysis techniques to comparative analyses of the relationship between expressions and directions in performances. The findings contribute to the objective understanding of factors emphasized by performers to control expressions in their performances and can be extended to future research with diverse conditions.