The track-to-track association problem is to determine the pairing of sensor-level tracks that correspond to the same true target from which the sensor-level tracks originated. This problem is crucial for multisensor data fusion and is complicated by the presence of individual sensor biases, random errors, false tracks, and missed tracks. A popular approach to performing track-to-track association between two sensor systems is to jointly optimize the a posteriori relative bias estimate between the sensors and the likelihood of track-to-track association. Algorithms that solve this problem typically generate the K best bias-association hypotheses and corresponding bias-association likelihoods. In this paper, we extend the above approach in two ways. First, we derive a closed-form expression for computing “pure” track-to-track association likelihoods, as opposed to bias-association likelihoods which are weighted by a unique relative bias estimate. Second, we present an alternative formulation of the track-to-track association problem in which we optimize solely with respect to marginal association likelihoods. Finally, we provide two algorithms that find the K provably best track-to-track associations with respect to our new likelihood function. These results facilitate what is commonly known as system-level track ambiguity management.