This paper studies the incentives that arise in a two-period agency relationship with moral hazard when agents are subject to limited liability. Since the existence of limited liability creates rent the principal can motivate an agent by credibly threatening him to be fired. It is shown that a combination of a two-period contract, where the agent is fired after period one in case of poor performance and retained otherwise, and a one-period contract optimally implements high effort. In particular, this combination is strictly better than a two-period contract, where the agent is retained in period two for sure. Moreover, there is a combination of one-period contracts that is equivalent to the optimal combination. While the second-period contract is the same as the optimal contract in the static model, the first-period contract pays a lower bonus in case of success. In an extension of the model "learning by doing" is considered. It turns out that the ranking of contracts is reversed if the increase in revenues due to "learning by doing" is sufficiently strong. In addition, a commitment problem arises which makes short-term contracting strictly worse than long-term contracting.