Three arguments for abstraction in phonology

Much recent research in cognitive and brain sciences links speech processing to the auditory and articulatory systems. These facts notwithstanding, here, I argue that phonology is abstract, algebraic, and substance-free. I first review challenges to this hypothesis and explain why it remains a viable possibility. I next suggest that an algebraic, substance-free phonology confers adaptive benefits; and finally, I support this proposal by evidence from three case studies. Using the restrictions on onset structure (e.g., bnif>bdif>lbif), I show that the phonological constraints on syllable structure are inexplicable by auditory and articulatory pressures; hence, they are abstract. A second case, from identity restrictions, suggests that phonological constraints are algebraic, as speakers generalize them to features that are unattested in their language; this is the case in both Hebrew and American Sign Language (ASL). Finally, I show that sign-language-naïve speakers—both adults and infants—freely extend their knowledge of spoken-language phonology to ASL signs, suggesting that phonological knowledge is partly amodal. I conclude that phonological grammar is substance-free.