The variation of breakage rates with grinding environment is important in the analysis and simulation of various grinding processes. In particular, computer simulation of the labor-intensive experimental Bond grindability test requires reliable estimates of breakage rate and distribution parameters from a minimum amount of experimental work. This paper was aimed at studying the variation of breakage rate parameters in the Bond mill. Batch kinetic experiments with monosize feeds and standard locked-cycle Bond grindability tests with natural size feeds were carried out on cement clinker and trass samples. Using the cumulative-basis, size-discretized population balance model (PBM) of batch grinding and invariant breakage distribution parameters determined from monosize short-time grinding experiments, the breakage rate parameters were back-calculated by minimizing the sum of squares of relative deviations between the simulated and experimental product size distributions. The back-calculated rate parameters were found to be varying depending on either grinding time or the feed size distribution. No single set of parameters was found to be representative of the whole grinding path in the Bond grindability test, and therefore attempts to simulate the Bond test with 1 set of parameters resulted in incorrect computed work index values.