Studies reporting sociometric assessments based on nominations have been characterized by important methodological inconsistencies when conducted in the middle school context. The purpose of this study was to examine (1) the possibility of a response bias when participants are provided with a long roster sorted alphabetically, (2) the impact of including or not other-sex peers in the voting population, and (3) the impact of including or not all the grademates in the voting population. Participants were 664 sixth graders from three middle schools. Peer nominations for sociometric items (i.e., "like most" and "like least"), as well as teacher ratings of antisocial behavior and records of academic performance, were collected. A sequence effect in peer nominations was found, suggesting that students whose names were listed higher on the rosters received more nominations than did students whose names were listed lower on the list. Moreover, results indicated that the nominations received from the other-sex grademates and from the grademates outside the classroom improved the predictive validity of the sociometric measure. The implications of these results for the use of sociometric assessment in middle schools are discussed.