Compensatory Lengthening: Phonetics, Phonology, Diachrony (review)

This book, the author’s Ph.D. dissertation completed at the University of California, Berkeley, is the latest addition to the ‘Outstanding dissertations in linguistics’ series, edited by Laurence Horn. This most stimulating volume identifies two types of compensatory lengthening (CL). Citing fifty-eight languages from twenty families (documented in Appendix A, 191–96), Kavitskaya depicts CL due to consonant loss as CVC → CV: (Type 1). The example of Lithuanian is given in which nasals are deleted if followed by voiceless fricatives. The second type of CL, known as CVCV CL, occurs as a result of vowel loss, that is, two open syllables with a short vowel in each developed into a closed syllable with a long vowel. The example furnished here is (Late) Common Slavic *borŭ > Serbo-Croatian bo:r ‘pine forest’. This may be formalized as CVCV → CV:C. Appendix B contains twenty-one languages and dialects from seven language families with Type 2 CL (117–19). One third of these are Slavic. Quite interestingly, the author affirms that Type 1 is far more common than Type 2 (5), and only two languages were found that display CVCV → CV:C synchronic alternations (Lama [Gur] and Baasaar [Voltaic]; 5, n. 2). Now that the groundwork has been laid, typologists will surely want to hypothesize why this is so.

The work assumes, as the author points out, a listener-oriented view of sound change (following the research of John Ohala): ‘internal phonetic properties of the speech signal can be misparsed and reinterpreted, yielding phonologization’ (10). The examination of the fifty-eight languages reveals that the phonetic motivations for CVC CL always involve either only certain consonants which delete, or, most often, ‘lengthened vowels were phonetically longer in the environment of lost consonants and thus could be reanalyzed as phonemically long with the loss of the consonant through the mechanism of hypercorrection’ (12–13).

It is fascinating to see that the loss of the glottal stop causes CL in ten of the fifty-eight languages— languages as diverse as Akkadian and Bella Coola. In the latter, *CVʔC > CV:C or CVC’ (191, n. 2). Colloquial Arabic dialects as representative of Semitic are excellent examples of the former development, as Old Arabic raʔs ‘head’ > ra:s. Biblical and Modern Hebrew both have ro:š as the cognate, which exemplifies *a: > o: (the so-called Canaanite vowel shift). Tiberian Hebrew is mentioned in the appendix without any notation that the loss of the glottal stop triggered CL (192). Rather, K states that it displays ‘morphological (templatic) alternations’ (192, n. 5). When comparing Tiberian Hebrew qa:ra: ‘he read’ with Old Arabic (pausal) qaraʔ, one notes that Hebrew lost the glottal stop, and CL of the preceding vowel occurred. This is quite different from the morphological CL discussed, in which the five ‘guttural’ consonants (ʔ, ʕ, h, ħ, r) do not geminate, and therefore CL takes place instead. Consequently, Hebrew should be added to the author’s list of Persian, Ket, and so on, where the glottal stop triggers CL.

When one deals with dozens of languages in a work such as this, mistakes inevitably creep in. First, let us mention that Hebrew ne:tse: ʔ is written with final glottal stop; however, the glottal stop is not pronounced (88). The word ‘people’ is ʕa:m, with the definite article ha:ʕa:m, and not with a voiceless pharyngeal fricative (89). The word ‘beginning’ is ro: š, no ro (89). Second, let us turn to Persian. It is surprising to see it listed as both Farsi (220) and Persian (222) in the index. In fact, they are even both cited in a single paragraph (perhaps making some linguists think they are distinct languages) (85). For colloquial Tehrani Persian, te:ran ‘Tehran’ should be replaced by the correct te:run (83). Gurage (192) is not one language but rather a language and dialect cluster with at least a dozen members (Ethio-Semitic). Finally, the name of the Polish linguist Jerzy Kuryłowicz is misspelled (208).