Consonant Harmony

Abstract

The term consonant harmony refers to a class of systematic sound patterns, in which consonants interact in some assimilatory way even though they are not adjacent to each other in the word. Such long-distance assimilation can sometimes hold across a significant stretch of intervening vowels and consonants, such as in Samala (Ineseño Chumash) /s-am-net-in-waʃ/ → [ʃamnetiniwaʃ] “they did it to you”, where the alveolar sibilant /s‑/ of the 3.sbj prefix assimilates to the postalveolar sibilant /ʃ/ of the past suffix /‑waʃ/ across several intervening syllables that contain a variety of non-sibilant consonants. While consonant harmony most frequently involves coronal-specific contrasts, like in the Samala case, there are numerous cases of assimilation in other phonological properties, such as laryngeal features, nasality, secondary articulation, and even constriction degree. Not all cases of consonant harmony result in overt alternations, like the [s] ∼ [ʃ] alternation in the Samala 3.sbj prefix. Sometimes the harmony is merely a phonotactic restriction on the shape of morphemes (roots) within the lexicon. Consonant harmony tends to implicate only some group (natural class) of consonants that already share a number of features, and are hence relatively similar, while ignoring less similar consonants. The distance between the potentially interacting consonants can also play a role. For example, in many cases assimilation is limited to relatively short-distance ‘transvocalic’ contexts (. . . CVC. . . ), though the interpretation of such locality restrictions remains a matter of debate. Consonants that do not directly participate in the harmony (as triggers or undergoers of assimilation) are typically neutral and transparent, allowing the assimilating property to be propagated across them. However, this is not universally true; in recent years several cases have come to light in which certain segments can act as blockers when they intervene between a potential trigger-target pair. The main significance of consonant harmony for linguistic theory lies in its apparently non-local character and the challenges that this poses for theories of phonological representations and processes, as well as for formal models of phonological learning. Along with other types of long-distance dependencies in segmental phonology (e.g., long-distance dissimilation, and vowel harmony systems with one or more transparent vowels), sound patterns of consonant harmony have contributed to the development of many theoretical constructs, such as autosegmental (nonlinear) representations, feature geometry, underspecification, feature spreading, strict locality (vs. ‘gapped’ representations), parametrized visibility, agreement constraints, and surface correspondence relations. The formal analysis of long-distance assimilation (and dissimilation) remains a rich and vibrant area of theoretical research. The empirical base for such theoretical inquiry also continues to be expanded. On the one hand, previously undocumented cases (or new, surprising details of known cases) continue to be added to the corpus of attested consonant harmony patterns. On the other hand, artificial phonology learning experiments allow the properties of typologically rare or unattested patterns to be explored in a controlled laboratory setting.