In the past experimental studies on mill hold-up were done using tracers and residence time distribution analysis (RTD). In one or two studies the entire content of the mill was weighed after the experimental run. Either as a result of inflexible experimental setup or difficulties in measuring hold-up in closed-circuit industrial mills via the RTD approach, the published correlation between hold-up measurements and mill feed rate disagree with each other. In this study a carefully constructed pilot mill enabled hold-up measurements in real time while the mill was operating. Therefore, extremely accurate hold-up data and transient hold-up data were obtained. In light of this data the correct trends in hold-up as a function of operating conditions are established. Explanations for the observed trends are provided throwing much light on anomalies on previously published data.