[The paper delimits the field of descriptive phonology, formulates a system thereof, and briefly points out some of its implications.]

[Armenian, like Hittite, Luwian, and Lycian, retains the third laryngeal initially, and has no inherited long vowels, no palatal-velar distinction, and no feminine gender. These and other archaisms lead to the conclusion that Armenian is an Anatolian language and can be compared to more advantage with Lycian and Hittite than with the IE languages proper.]

[The so-called passive of videō is in reality an old middle, corresponding to Homeric ∊ἴδ∊ται, ἰδέσθαι and meaning from prehistoric times ‘seem, appear’. The meaning ‘be seen’ is late, secondary, and very rare. This view, besides being in consonance with the evidence of other IE languages having verbs derived from the base *weid-, is confirmed by the occurrence in early Latin of a deponent form of videō (Curculio 260–1; Epidicus 61–2). The middle verb vidērī therefore illustrates and supports the writer's theory (AJP 48.157–75) that ‘the Latin passive is essentially a middle voice’.]

Jellinek has presented lists of the Gothic consonant-clusters (excluding accidental clusters resulting from compounding, such as tg), with notations of which clusters are found in only one word, but with no other remarks on frequency of occurrence. Dewey has gone a step further with English consonant-clusters: from his tables we learn how many times each cluster was found in his 100,000 words of sample texts. A different kind of frequency is reported in the Twaddell study of one- and two-syllable German words: having analysed the uncompounded words in a lexicon, he reports the number of different forms in which each cluster was found. The Jellinek report gives us no statistical picture in the sense intended here; the Dewey count gives us a statistical picture of TEXT frequencies; the Twaddell count, of LIST frequencies.