Various techniques for selecting a pacing vector based on pacing capture thresholds are described. One example method described includes for each of a plurality of vectors, iteratively delivering at least one pacing stimulus at each of a plurality of magnitudes within a predetermined range of magnitudes to a first chamber, determining if a depolarization occurred in a second chamber of the heart within a predetermined threshold time interval after the pacing stimulus that is less than an interval, identifying a pacing stimulus for which a depolarization in the second chamber does not occur within the predetermined threshold time interval, determining a capture threshold magnitude for the vector based on the magnitude of the pacing pulse for which a depolarization in the second chamber does not occur within the predetermined threshold time interval, and recording the capture threshold magnitudes.