In human social interaction, the notions of commitment and trust are strongly interrelated. A formal model for this interrelation will enable artificial agents to benefit from the associated reduction of complexity and improved possibilities for cooperation. We argue that the notion of living up to one's commitments, rather than actual fulfillment, plays the key role in the trust-commitment interrelation, and we propose a formal analysis of that notion in the framework of agents and strategies in branching time. The main technical innovation is a stringency ordering on an agent's strategies that allows one to classify what an agent does as more or less appropriate with respect to a given commitment, leading to a fine-grained assessment of trustworthiness. Key words: promises, contracts, stit logic, strategy, commitment, trustworthiness