Rater subjectivity has long been an intriguing topic. The use of discussion as a resolution method is a practical way to reduce this subjectivity. However, the efficacy of discussion depends on whether different raters get equally engaged in it or one rater tends to dominate others. This study investigated whether and how rater dominance occurs in discussion. To this end, three discussion sessions in which five Iranian raters negotiated to resolve discrepancies in rating were analyzed. Findings indicated raters were "unequally" engaged in discussions, so rater dominance obviously existed. However, it did not necessarily display itself in more turn-takings, a higher amount of speech or changes in scoring. The joint construction of discourse was found to affect raters' understanding of the rating criteria and scoring method. This in turn played a key role in how dominance was realized. The findings illuminate the complexity of rater dominance as a highly context-dependent issue.