No Indo-Europeanist can fail to be struck by the numerous cases in which the same root seems clearly to appear sometimes with initial s plus consonant, sometimes with the same consonant but without the s. This is called ‘s mobile’, or in English ‘s movable’. Often the two forms appear in the same language, as in Skt. paś : spaś- ‘see’. In other cases one form appears in one Indo-European language, or in several, the other in other languages. But there is no regularity; all Indo-European languages seem to be unpredictable in this respect. As Wackernagel observes (1.§231), the facts, it seems, Verbieten an ein durchgehendes Lautgesetz zu denken'. In the next sentence he writes: ‘... muss solcher Schwund [of initial s before consonant] schon in der Grundsprache stattgefunden haben, und zwar ... wenn bestimmte Konsonanten im Auslaut des vorausgehenden Wortes standen’.

J. A. Rea's note in IJAL 24.145–50 (1958), Concerning the Validity of Lexicostatistics, is a telling example of partisanship on problems which are still open but on which attitudes tend too often to be taken either wholly favorable or wholly unfavorable.

Using the higher of two retention rates (85.4% against 80.5%), and therefore the one operating more slowly toward change, Rea tests eight Romance languages with 100 words, and finds the time depths of separation absurdly low and recent, as matched against knowledge of the actual history of the languages: less than one-half as great. He therefore denies that lexicostatistics has 'any usefulness or validity even for the languages upon which it is based'.

With the help of the work so far done in comparative Dravidian studies, it is possible to reconstruct the following phonemes for Proto-Dravidian (PDr.).

Consonants:

Vowels:

It is well known that at least two conflicting forces contribute to the development of languages and dialects: regular sound change, which is studied by the techniques of comparative linguistics, and diffusion of linguistic innovations beyond their original source, which is studied by the techniques of linguistic geography. The present paper offers a detailed study of these two forces and their interaction in the dialects of Mazatec.

A procedure for the subgrouping of related languages or dialects on the basis of exclusively shared innovations was proposed by Brugmann and recently elaborated by Dyen (in a review of Otto Dahl's comparative study of Malagasy and Maanyan.) In the present study of the Mazatec dialects, the procedure is applied to an Amerindian language.

In the series bum, bun, bung, the final /m/, /n/, and /ŋ/ are distinguished by oral closure respectively at the lips, at the alveolar ridge, and at the soft palate. When these words are spoken in isolation, the three final consonants share the features of open nasal cavity, voice, and a gradual decrescendo to silence. With these observations in mind, consider the three words bump, bunt, bunk. In these words the final stops are phonetically released or unreleased, in free variation. What is of interest here is the precise articulatory nature of the unreleased allophones and the preceding nasals. The articulatory formation of the nasals is as in bum, bun, bung. But there is no gradual decrescendo to silence: speech is terminated by an abrupt cessation of voice and an abrupt closure of the nasal cavity by an upward movement of the velum. It is evident that in these articulations the closure of the velum cannot be heard as such. The only acoustic and auditory cue for final /p/,/t/,/k/ in the pronunciations described is the abrupt ending of the preceding nasal. If we adopt the letter [I] for this sudden interruption of a nasal, the following symbolism can be interpreted either auditorily or acoustically: [m] + [I] = /mp/, [n] + [I] = /nt/, [ŋ] + [I] = /ŋk/.