Can scientists learn to be innovative from their coauthors? Innovative researchers tend to work together, so it is hard to tell whether collaboration transmits a way of doing research or merely reflects who chooses to work with whom. We address this question using the sudden deaths of star scientists, which end collaborations for reasons unrelated to their coauthors' careers. Earlier work using such deaths shows that coauthors' output declines, but this could reflect lost access to status and journals or the end of an ongoing flow of ideas, and it does not tell us whether coauthors durably learned anything from the star. We shift the focus from how much scientists publish and where to how they do research, tracking the novelty of the ideas they combine over carefully reconstructed careers. This design allows us to ask what remains once the star is gone, and thus to distinguish lasting learning from influence that ends when the relationship does.