Vowel Harmony and Correspondence Theory (review)

Krämer’s book begins with an introduction to the themes of inquiry in the study of vowel harmony over the past thirty years. As vowel harmony is an iterative process that seems to ‘spread through a word’, it is important to delimit its domain, especially from this perspective. The processes of transparency (when a vowel allows harmony to pass through, but is unaffected) and opacity (when a vowel interrupts the spread of harmony, and in many cases, initiates its own domain of harmony) are exemplified and discussed. In addition, K introduces trojan vowels, the most familiar of which are Hungarian disharmonic words such as hid-nak, in which a front stem vowel takes a back suffix. K’s view is a departure from treating such items as bearing a lexical diacritic; rather, neighboring vowels take on an underlying feature specification of the trojan vowels, a specification that never surfaces on the trojans themselves. The last category of nonpredictable vowel harmony behavior is in hybrid stem vowels, which, taking Hungarian as an example again, are those that allow either front or back variants of the suffix, vacillating in this behavior. For these, K’s proposal is that both featural variants of the stem are stored in the lexicon.

The implementation of assimilation as syntagmatic agreement is modeled through correspondence theory, which states that two elements (in this case, adjacent vowels) are in a privileged identity-striving relation. This correspondence relation is the force behind fully-agreeing vowel harmony. The behavior of nonparticipating transparent vowels is the result of constraints that simultaneously demand both identity and nonidentity with adjacent vowels (S-Ident and *S-Ident). By either exclusively agreeing or exclusively disagreeing with its two neighbors, a transparent vowel ‘balances’ its violations in the conjunction of these two constraints.

One of K’s goals is to characterize patterns of directionality in vowel harmony. In the correspondence view, when roots direct harmony onto suffixes rather than vice versa (as is the case throughout Altaic), it is due to a constraint against affix features initiating correspondence relations. K also studies the behavior of dominant-recessive harmony systems. These systems (which incidentally almost always involve the advancement or retraction of the tongue root) differ from the familiar unidirectional harmony in root-controlled patterns. In dominant-recessive systems, the presence of the dominant feature value (e.g. [+ATR] in Kalenjin) in apparently any morpheme in the word can induce that alternation in all other vowels, bidirectionally. K employs constraint conjunction between the constraints maintaining underlying specifications and those that assess markedness of the recessive feature; the result is a disequilibrium tending towards losing [−ATR]. One thorough case study is that of Nez Perce, which is dominant/recessive, but in addition, surface /i/ participates in both harmonic series, yielding trojan behavior.

In short, K’s book provides a demonstration of how the attested patterns of harmony may be treated through differences in correspondence relationships, markedness constraints, and constraint conjunction. K offers frequent comparison with models employing underspecification, floating features, strict locality, sympathy theory, derivations, and output-output correspondence, all of which he eschews.