The overhead associated with training new staff brought onto software development teams has long been a matter of concern for both information systems (IS) researchers and practitioners. The projected global shortfall in qualified IT professionals in coming years only serves to emphasize the importance of this issue. This is particularly so, given that it is generally recognized that training overhead may outweigh any benefits derived from adding new members to a project team. Thus, the issue of on-the-job training has received plenty of attention from IS researchers. There is, however, a notable lack of useful prescriptive guidelines in the literature and practitioners tend to rely very much on heuristics (developed largely from individual experience). Our own view is that most commonly-used estimates for veteran training efforts and rookie assimilation are too low. In addition, we believe that a much finer- grained level of detail is required in any model proposed as a useful decision support aid in this area. At the very least, we contend that such a model should encompass project and organizational characteristics; rookie experience, ability, and skills match with project requirements; and, following from this, training needs. To test this, we have embarked on a series of case studies. Stage 1 is exploratory and developmental, major aims being the development of a system dynamics model of on-the job programmer training and the generation of a set of hypotheses. Stage 2 involves the testing of these hypotheses. Essentially, the systems dynamics model is a representation of Stage 1 findings. Typically, in systems dynamics models, most variance is caused by a limited number of parameters. Thus, by identifying these and employing replication logic, we can construct a set of Stage 2 field studies that we can use to validate, refute, or refine our Stage 1 findings.