Chomsky proposed to study what kind of functions, or more generally, computations, are capable of simulating the mental processes of language, starting with some coarse criteria like whether they can put elements correctly together into sentences and words (models of syntax and morphology) or describe the patterns governing the pronunciation of words (models of phonology).

Over the last sixty years, we have sometimes managed to specify these computations in extremely precise mathematical terms, and thereby learn a lot about human language. One thing we have learned is that, in terms of what broad classes of computations they belong to, there is a marked difference between syntactic and morphological computations, on the one hand, and phonological computations, on the other. This suggests that there might be problems when these two cognitive sub-systems communicate with each other.

In particular, phonology is limited to finite state transductions, which are very simple “Markovian” computations (in fact, it is even more restricted than that). In contrast, syntax and morphology do more complex computations that operate over tree-like structures.

However, these subsystems do not operate in a vacuum. Morphemes, combined by morphology and syntax into trees, correspond to phonological pieces, which the phonological grammar deals with. If there is even the slightest possibility of phonology being sensitive to any of the tree-like structure, it raises the very real possibility that the computation would be outside the capacity of the phonological computation device.

We will examine the relevant morphophonological phenomena, including cyclic segmental and stress patterns and reduplication, and make clear why they threaten finite-state phonological computation; we will find, however, that, strikingly, the precise types of phenomena that would be problematic in fact (very narrowly) fail to be observed. We will explore how and why this might be, and what it might imply for how the interface between morphology and phonology works.


A good grasp of the basics of generative phonology and morphology is essential for this course (at a minimum, some experience doing theoretically motivated segmental phonology problems, and a basic understanding of Distributed Morphology). No particular mathematical or computer science background is expected. The logic of the theoretical arguments will be discussed largely without use of mathematical notation and vocabulary. Interested students with experience doing mathematical proofs will be given opportunities to explore the issues more deeply.

Course slides

Course slides


Background (this would be good to read in advance)

Sound Pattern of English, Chapter 1 and Chapter 2, up to the end of section 5

Heinz and Idsardi (2011): Sentence and word complexity

Heinz (2011a): Computational Phonology I

Hopcroft et al., Chapter 1, Introduction, Sections 1.1.1–1.1.3, 1.5.1–1.5.3

For reading ahead on the theoretical contents

Bjorkman and Dunbar (2014): Cyclic stress in Chamorro revisited

(More to come)

For details on technical matters

Kaplan and Kay (1994): Regular models of phonological rule systems