Trilled vocalizations, wherein notes are repeated in rapid succession, are found in a variety of taxa including oscine birds, singing mice and nonhuman primates. Previous work on birds and singing mice has provided evidence of vocal performance constraints in trills, where there is a trade-off between the rate of the note repetition and the bandwidth (or frequency range) of each note. Here, we investigate vocal performance constraints in the trilled portion of the female contribution to the duet in the Bornean gibbon, Hylobates muelleri, recorded from seven sites in Sabah, Malaysia. We used two approaches. First, to ensure that our results were comparable with previous studies on vocal performance constraints, we used a 90% quantile regression to examine the relationship between trill rate and bandwidth. We found that there was a significant negative correlation between bandwidth and trill rate. Second, we formally compared multiple hierarchical models to identify the best predictors of bandwidth and trill rate. Our top model predicting bandwidth showed that trill rate and location within the trill were reliable predictors of bandwidth. With trill rate as the response variable, our top model included location within the trill as well as trill duration. We found that there were no important site-level differences in bandwidth but that trill rate varied predictably among sites. Our analyses provide strong evidence for performance constraints in the production of trills in Bornean gibbon females. Further research is needed to determine whether higher-performance trills provide honest signals of caller quality and whether gibbons respond differently to low- and high-performance calls.