While modern grain yield monitors are able to harvest variety and hybrid trials without imposing bias, cotton yield monitors are affected by varietal properties. With planters capable of site-specific planting of multiple varieties, it is essential to better understand cotton yield monitor calibration. Large-plot field experiments were conducted with two southeast Missouri cotton producers to compare yield monitor-estimated weights and observed weights in replicated variety trials. Two replications of multiple varieties were planted in 12-row plots with 0.97 m row spacing. Plots were harvested with a module-building spindle picker equipped with a yield monitor. A separate module was built for each plot and weighed. Yield monitor data were used to calculate an estimated weight for each module. Significant differences in seed cotton yields were detected between the observed (weighed modules) and estimated (yield monitor) values. In addition to a significant variety main effect, a significant location by variety interaction was present in the error, both in terms of yield (absolute) and as a percentage of the observed yield (relative). Some HVI properties were significantly correlated with the absolute and relative error. Data from additional site-years will be analyzed and other factors will be investigated to try and achieve a better understanding of the factors affecting cotton yield monitor calibration.