The purpose of this study was to examine the differences of student-athletes' preferred leadership behavior for their coaches based on gender, competition level, task dependence, and task variability. Four hundred and eight male and female student-athletes from four NCAA Division I and six Division II universities expressed their preferences using the Revised Leadership Scale for Sport (RLSS) (Zhang, Jensen, & Mann, 1997). The preference version of the RLSS included six behavior dimensions, autocratic, democratic, positive feedback, situational consideration, social support, and training and instruction behaviors. A split-plot ANOVA was performed on the individual preference scores grouped by gender, competition level, task dependence, and task variability for the six dimensions of coaching behavior. The ANOVA also computed interactions. Fisher's LSDs were performed on all significant interactions. Among genders, the ANOVA demonstrated a significant gender by level interaction for democratic behavior. Fisher's LSD failed to detect any significant interactions. Among competition levels, the ANOVA demonstrated a significant task dependence by level interaction for autocratic behavior. Division I independent sport student-athletes had significantly higher preferences than Division II independent sport student-athletes. A significant task variability by level interaction revealed Division I open sport student-athletes had significantly greater preferences for autocratic behavior than did Division II open sport student-athletes. The results also demonstrated a significant task dependence by level interaction for democratic behavior. Division I independent sport student-athletes showed significantly greater preferences for democratic behavior than Division I interdependent sport student-athletes and Division II independent sport student-athletes showed significantly greater preferences than did Division II interdependent sport student-athletes. Independent sport student-athletes, regardless of gender or competition level, showed significantly greater preferences for democratic, positive feedback, situational consideration, and social support behaviors. The results also indicated a significant task variability by level interaction for autocratic behavior. Division I open sport student-athletes had significantly greater preferences for these coaching behaviors than Division I closed sport student-athletes. Open sport student-athletes, regardless of gender or competition level, had significantly greater preferences for democratic, positive feedback, and social support behaviors. The results demonstrate support for a portion of the multidimensional model of leadership (Chelladurai, 1979; 1990) with differences in behavior preferences based on student-athlete characteristics of competition level, task dependence, and task variability. The results may aid in the evaluation of coaching behavior and coaching method and in defining training preparation programs that would enhance the congruence between student-athlete behavior preferences and actual coaching behaviors. The results suggest the use of the multidimensional model of leadership and the related instruments for future investigations of sport leadership behavior.