Beat keeping was long thought to depend on vocal learning, the capacity to imitate novel sounds. Parrots, some songbirds and humans have it; most mammals do not. The vocal learning and rhythmic synchronisation hypothesis predicted that a non-vocal-learner should be unable to entrain to a musical beat.
Ronan, a California sea lion at the University of California Santa Cruz's Long Marine Laboratory, was trained from around 2012 to move her head in time with a metronome. She then generalised to musical tracks and, importantly, to tempos she had not been trained on, adjusting her movements to hold phase with the beat. Later work reported that her timing accuracy remained good years on.
California sea lions are not considered vocal learners, so the result is a real counterexample to the strong form of the hypothesis. What it does not show is that beat keeping is widespread among mammals: Ronan is a single, heavily trained individual, and whether wild sea lions do anything comparable is unknown.