This paper investigates the effect of mobility among a firm’s R&D personnel on its innovation performance. Since knowledge created within an R&D department of a firm is largely embodied in the person of the individual researcher, firms might gain access to external knowledge via hiring new staff. Consequently, innovation should be enhanced. However, firms may suffer from a simultaneous outflow of R&D workers. Our empirical analysis falls back on theories developed by Jovanovic (1979) and Cooper (2001). According to these theories, labor mobility increases the employer-employee matching quality within the R&D department (Jovanovic, 1979). Moreover, as long as the R&D knowledge of a leaving worker can be still used by the worker's former employer, overall labor mobility results in net inflows of R&D knowledge and thus should stimulate innovation (Cooper, 2001). To disentangle the effects of net gains or losses in R&D personnel and replacements within the R&D department is the major aim of this study. We employ, besides net growth in R&D personnel, churning as a measure for R&D worker mobility. Churning depicts the number of workers which are replaced by new ones, i.e. it is an employment-neutral measure of labor turnover. Furthermore, we distinguish between different types of innovations, namely between process and product innovations as well as between market and product-range novelties. Using two sets of bivariate probit regressions, we estimate various knowledge production functions. Labor mobility is found to enhance innovation performance up to a certain threshold. Interestingly, this threshold differs between the types of innovation. The optimal amount of churning among R&D works is lower for process innovations than for product innovations. This result is intuitive since process innovation will require much more firm-specific knowledge and less external knowledge. Our regressions further indicate that a positive effect of labor mobility on innovation performance depends crucially on the appropriability regime. If R&D knowledge is characterized by rivalry in its application, i.e. it is non-duplicative, the threshold up to which a positive effect of labor mobility can be observed is considerably lower.