Discusses propositions relevant to assessing the relationship between teaching and hospital costs. A proper measure of hospital costs will include physicians' fees for services. A number of cost studies do not include house staff charges and therefore are not useful in identifying the cost effect of teaching. Two analytical examples are developed that show that teaching hospitals may be less expensive. The first example, in which students are exogenous, suggests an econometric specification where student variables are interacted with a measure of patient care output. The second example suggests that both a measure of teaching and a measure of patient care are required in order to estimate a cost function. It is also a particular feature of that example that no cost savings are expected unless the student-physician ratio exceeds unity. The resulting models can help in isolating theoretical relationships between teaching and patient care costs, and can guide subsequent empirical work.