Some cautions regarding the phonological continuity hypothesis.

We consider the Phonological Continuity Hypothesis (PCH) of Fitch (2018) in light of a broader range of formal systems. A consideration of the learning and generalization of simple patterns such as AAB from Marcus (Marcus 2000 Curr. Dir. 9, 145-147(doi:10.1111/1467-8721.00080)) shows that finite-state automata defined in the standard way fail to generalize in a compatible fashion. However, pushdown automata with finite-memory limits do show compatible generalization capabilities. The third class of formal systems-tree automata-provide yet another possibility for the processing of words within sentences. We conclude that there are additional possible formal differences between sound patterns and sentence patterns, which will make testing the PCH even more difficult. This article is part of the theme issue 'What can animal communication teach us about human language?'