Phonological Regularity and Breakdown. An Account of Vowel Length Leveling in Middle English

In the history of English, vowels in stressed syllables showed a length alternation. For example, the singular noun whal ‘whale’ had a short vowel that was lengthened in the plural whāles ‘whales’, whereas the singular crādel ‘cradle’ had a long vowel that was shortened in the plural cradeles ‘cradles’. This alternation mechanically followed the phonological conditioning factors, but has been subsequently lost: as a result, words such as whale and cradle do not differ in vowel length in present-day English. Previous scholarship (e.g., Lahiri & Dresher, Open syllable lengthening in West Germanic. Language 75: 678–719, 1999) suggests that the loss, an instance of leveling, was due to the language learner’s inability to discover the regular mapping that relates the underlying and derived vowel length after independent phonological changes in the language. In this work, I invoke the Tolerance Principle (Yang, On productivity. Linguistic variation yearbook 5: 333–370, 2005; The price of linguistic productivity. How children learn to break rules of language. Cambridge, MA: MIT Press, 2016) to provide a quantitative analysis of how such loss may have taken place. The Tolerance Principle is a mathematical theory that measures exactly what it means for a phonological process to be systematic and predictable, and when it collapses. I show that the quantitative composition of the learning data in late Middle English, indeed, could not support vowel length alternation across the board, thereby providing a more precise causal explanation for its demise. The present project also illustrates the psychological motivation for the (seemingly implausible) use of historical data to investigate the role of child language acquisition in language change.