Although psychologists view bonuses and penalties as very different means of providing incentives for workers, economists have had less success at making sense of the distinction. A rational worker should be indifferent as to whether a payment scheme is called a bonus or a penalty plan if the actual contingent pay stream is identical in the two cases. In this paper we provide a framework for understanding the difference between payment plans that are deemed to be penalty or bonus schemes, and derive implications for when such plans should be implemented as a function of observable features of the manufacturing and monitoring systems. We call a payment plan a "bonus" scheme if the high payment occurs infrequently in equilibrium; a payment scheme entails a possible "penalty" if the low wage occurs infrequently. The frequency of high and low payments is derived in equilibrium in a model with moral hazard and probabilistic monitoring. We focus on the role of commitment and the possibility of false positives in he monitoring technology. It is shown that when the firm can commit to a monitoring intensity the workers will (almost) always be diligent and a penalty scheme will be observed. When commitment is infeasible the optimal payment structure depends on whether the monitoring technology permits false positives. In the absence of false positives the workers will be observed to face a penalty scheme if found shirking, but when false positives are possible there will be considerable shirking by workers in equilibrium, and a bonus scheme will be observed. We then analyze the crucial features of our theoretical monitoring technology in he context of actual employment situations. We find that middle-management and other non-production jobs are appropriate for bonus-type incentives, whereas in unskilled jobs or aspects of highly skilled jobs that require diligence but not skill, such as arriving on the job on time, we predict penalty incentives. We argue that the observed scarcity of penalty-type schemes can be explained by our model, without resorting to psychological justifications. In addition, we interpret the Japanese manufacturing systems as having a particular, built-in monitoring system that can be analyzed in our framework and shown to implement a high level of diligence from factory workers.