Purpose: The purpose of this paper is to examine whether the taxonomy of two conceptual models of leadership roles for principals--instructional and transformational leadership (IL and TL)--can be traced empirically in a sample of Norwegian school leaders. Design/methodology/approach: The participants consisted of 149 school leaders attending a national training program for newly appointed principals at the University of Oslo. Leadership preference was measured according to a theoretically grounded self-report scheme representing the most important aspects of both leadership models. Estimations were conducted by means of principal component analyses and confirmatory factor analyses. Findings: The principal component analyses revealed seven meaningful factors: demand-supportive leading, outcome control and loyalty to school owner, management by objective, trust in standard programs for development, trust in professional community, emphasis on collective responsibility, and distributed leadership. Furthermore, by using these seven factors as inputs, the authors attempted to model second-order factors representing IL and TL. The results showed that the structure of IL and TL could not be replicated. Instead, a more complex cross-model structure was revealed. Practical implications: The results suggest that even though the concepts of IL and TL are valuable analytic tools, they may be too simplistic to represent the reality of school leaders' thoughts and actions. Originality/value: The study contributes to the field by challenging the established models of school leadership and by generating insights into Norwegian school leaders' leadership beliefs and preferences.