This paper tests various theories of the decision on when to call in-the-money convertible bonds by following newly issued convertible bonds over the first ten years of their life and relating the decision to call or not call at each point to the determining characteristics implied by each theory. Our results support the yield advantage and after-tax cash flow hypotheses, as well as a variant of the safety premium hypothesis. We find no evidence to support the signaling hypothesis, and no evidence that the desire to extinguish the bondholder's option is an important element in the call decision.