The objective of this paper is to develop an optimal incentive system for multitasking scientists in universities or professors under repeat contracting. With the aid of a principal agent model under repeat contracting, we show that (i) when a second task is assigned to a professor and the two tasks are related, the size of the optimal incentive rate for the first task is reduced in some situations but not others relative to that of a single task, (ii) with an increase in the noise in the technical relationship of the second task or imprecision in output measurement, the optimal incentive rate for that task is reduced and for the first task may be reduced or increased , (iii) with greater efficiency of the professor in producing the second output, as reflected in ability relative to cost of effort, the optimal incentive rate for the first task generally decreases, (iv) if the output of the professor's two tasks are negatively correlated then the optimal incentive rate on the first task declines as the size of this correlation increases. The size of the guarantee is always reduced as the professor's ability for a task increases, but is increased as his cost of effort, noisiness of the technology or measurement of output, or correlation between the two outputs increases. It is also possible that, as a professor undertakes several difficult-to-measure tasks, the incentive rate will be reduced to the point that an optimal compensation system will involve only a guaranteed salary, which is a very weak incentive for effort. Selective audits may be useful in these situations.