In English, voicing alternations (e.g. knife ~ knives) impact mostly monosyllables, while polysyllables are rarely impacted. The opposite is true of French: most monosyllables that end in [al] keep their base faithful under affixation (e.g. bal ~ bal ‘ball(s)’), while most polysyllables tolerate a stem change (bokal ~ boko ‘jar(s)’). In this talk, I examine the two types of languages, and show that the symmetry is only superficial. The French trend is accessible to the grammar and extends readily to novel words, whereas English speakers treat novel words the same regardless of size. In other words, English speakers fail to find the generalization (the surfeit of the stimulus, Becker et al. 2011).
Positional faithfulness, and in particular, initial syllable faithfulness explains this asymmetry: the [al] in bal is protected by initial syllable faithfulness and by general faithfulness, while the [al] in bokal is protected by general faithfulness only. English goes against the Universal bias, requiring monosyllables to be less faithful than polysyllables. But with general faithfulness highly ranked, the ranking of initial syllable faithfulness is irrelevant, and the speakers are blocked from forming the required generalization.
Having established the asymmetry in the novel word tasks, we press English speakers further and ask them to learn unfamiliar morphophonological alternations (e.g. mi:p ~ mi:b-ni). Unencumbered by the counter-typological nature of actual English, speakers revert to Universal Grammar, and exhibit the French pattern.
This line of investigation, which goes from real words to novel words and from novel words to novel alternations, allows us to trace the biases that humans use in the phonological organization of their lexicon, and allows us to expose behavior that roundly contradicts the ambient language, yet conforms to the trends we see in the world’s languages.