Vowel Harmony as Agreement by Correspondence

Initially, it was believed that blocking effects1 simply could not be modeled with Agreement by Correspondence (ABC) (Hansson 2001; Rose & Walker 2004). Hansson, in particular, lays out this assumption very clearly: “Intervening segments do not themselves enter into the agreement relation holding between the trigger-target pair, and therefore they must be irrelevant to that relation: they cannot be opaque” (237, his emphasis). The model’s assumed inability to account for opaque segments was seen as a positive with respect to accounting for consonant harmony2 because blocking effects are extremely rare in consonant harmony systems3. However, since opaque segments are common in vowel harmony systems, it was unclear how applicable ABC would be to vowel harmony. In spite of this, Rose & Walker are optimistic about the possibility of analyzing at least some vowel harmony systems as ABC: