Les etymons en arabe: Analyse formelle et semantique. By Georges Bohas and Karim Bachmar. Recherches, vol. 23. Beirut: Dar el-Machreq Editeurs, 2013. Pp. 214.

As in the book under review, Georges Bohas has devoted much energy, often in collaboration with his pupils, to lending support to the idea that Semitic roots, as attested in Arabic, can be reduced to semantic core units consisting of two radicals or, on an even deeper level, to a (non-ordered) set of articulatory features called a "matrix." In his longstanding theory of "matrices et etymons," such matrixes are said to be the base ("invariant notionnel") for a whole array of semantically related roots. The "etyma" then are specific phonemic instantiations of these abstract matrixes. The matrix {[labial], [coronal]}, for instance, represents the semantic core "to hit," which can be instantiated by "etyma" as [b, t], [b, d}, etc., the elements of which are roots such as (core consonants marked in bold) batta 'to cut', badaha 'to hit someone with a stick', tabara 'to break', tabala 'to annihilate', and many more. When such roots cannot be semantically related, Bohas and Bachmar argue that they actually represent different "matrixes." The root mata'a, for instance, has the unrelated meanings 'to hit someone with a stick' and 'to stretch a cord'. While the first meaning matches the matrix ([labial], [coronal]}, the second meaning derives from the matrix {[nasal], [coronal]) with the semantic base ("invariant notionnel") "traction," the consonant m being both labial and nasal.

By sheer combinatorics, Bohas and Bachmar arrive at the number of (26 x (26-1)) : 2 = 325 possible "etyma"--w and y are excluded from the total of 28 Arabic consonantal phonemes as "extensions" of the labial b and the coronal t, respectively. The product 26 x 25 is divided by two since, according to this theory, the order of the consonants in the etyma plays no role. Not all of the combinatorically possible etyma are attested or meaningful, due to, e.g., co-occurrence restrictions (41 theoretically possible etyma), but Bohas and Bachmar arrive at the number of 192 fairly "productive" etyma. Additionally, 21 etyma are said to be somewhat working, 10 etyma are only attested in onomatopoetic context (and are therefore of problematic status), and 36 etyma are only attested with one root.

The volume...