Will Bennett has filed his dissertation, Dissimilation, Consonant Harmony, and Surface Correspondence (678pp.), and posted it as ROA-1173. He defended in November before an appreciative full house that included his local committee (Akin Akinlabi, Alan Prince, Bruce Tesar) and the Skype-delivered vocal presence of his outside member, Sharon Rose (UCSD). Bennett develops a comprehensive theory of consonant dissimilation through deep analysis of patterns in 10 languages (Kinyarwanda, Sundanese, Cuzco Quechua, Obolo, Chol, Ponapean, Zulu, Yidiny, Latin, and Georgian) and a broader survey of about 130 more.
Bennett’s theory is built on a striking insight into the logic of Surface Correspondence, an approach to vowel and consonant harmony developed by Rachel Walker, Sharon Rose, and Gunnar Hansson. Correspondence between segments is used to support constraints on identity, a notion that plays a role in many linguistic modules and structural relations (input–output, base–reduplicant, output–output, within output, even between inputs in Tesar’s recent work). Surface correspondence holds entirely within a single output.
In moving from harmony to disharmony, the naïve mechanical idea would be to treat it as a twisted version of harmony: because dissimilating items interact, the thought would run, they must be formally linked so that constraints demanding difference can be stated over the linkage. Bennett’s theory has nothing of the sort: dissimilating consonants do not correspond and dissimilation is not due to constraints that specifically demand it.
Surface correspondence falls under the control of two types of constraints: those basic to earlier work which penalize the occurrence of similar elements that fail to correspond (as high vowels with high vowels, or stridents with stridents); and those, developed as a class by Bennett, that structurally limit the extent or membership of the correspondence relation, for example by prohibiting it from crossing a stem–prefix boundary or relating elements in different syllabic roles. Correspondent elements provide the grist for harmony, which demands further similarities. Bennett’s crucial move is to break out of the harmonic frame of mind and look afresh at the logic of the theory: the key observation is that the constraints are also satisfiable by lack of correspondence. As long as the noncorresponding consonants are not similar in the relevant respect, no penalty is incurred for failure to correspond, and without a correspondence relation present, no limitations on the relation can be transgressed. Dissimilation offers a viable alternative route to satisfying the very same constraints that are involved in setting the scene for harmony.
Bennett finds a deep and thoroughly unexpected consequence of the theory: a predicted asymmetry between harmony and disharmony, which he calls the ‘mismatch’ property. In harmony systems, there is a substantive distinction between the class of features that precondition harmony and those that harmonize when the precondition is met. Disharmony occurs under correspondence theory to avoid the precondition; it is not some kind of anti-harmony among features that would otherwise harmonize. Bennett’s large-scale empirical work strongly supports the basic prediction that harmonizing and dissimilating features differ, and it is consistent with the predicted relation between disharmony and the preconditions for harmony, opening the way for further detailed scrutiny.
Bennett has joined the faculty of Rhodes University in Grahamstown, South Africa, where he is teaching linguistics and pursuing his research interests, which range over many areas of phonology and syntax, with special focus on African languages.