In current constraint-based models of phonological learning, learning the legal (and illegal) sound sequences in one’s language (i.e. phonotactics) facilitates learning alternations. However, there is still little empirical evidence for this. I first show that alternation learning in an artificial grammar experiment is facilitated when the phonotactics in the lexicon match the alternation. This supports current models of phonological learning. What does this mean for languages with a mismatch between phonotactics and alternations? In the second part of this talk, I present corpus analyses as well as computational learning simulations of two such languages – Korean and Turkish. I show that in both languages the reported mismatches between phonotactics and alternations are superficial. This undermines assumptions in previous analyses of these patterns. Instead, I argue that there is a bias to maintain similar generalizations in phonotactics as well as alternations. This further supports the claim that learning phonotactics aids in learning alternations, ultimately arguing in favor of constraint-based models of phonology that account for these two generalizations using a single mechanism.