We aim in this paper to replace by a unified, systematic theory the heterogeneous and incomplete methods hitherto offered for determining IMMEDIATE CONSTITUENTS (hereafter abbreviated IC, plural ICs). The unifying basis is furnished by the famous concept of patterning, applied repeatedly and in divers special forms.

[Skt. adáḥ ‘illud’ is in reality adó = *adáu with the same *-au that is contained in asáu (= ). Appendix: Pāli dhītā- ‘daughter’ and ghara- ‘house’.]

In the treatment of IE s, the insular Celtic languages share in general the peculiarities of Iranian and Greek, where the s-sound was characteristically changed into an aspirate in certain specified positions, and was further liable to disappear. In Irish, however, initial s remained (lenited ṡ, eventually h). The rule for insular Celtic may be stated briefly as follows: (1) Initially before a vowel, IE s remained in Irish and became h in Brittonic. (2) Between vowels, it was eventually dropped in Irish and became h in Brittonic, but was lost here also if the preceding vowel was stressed (cf. further below). (3) In final position, it probably resulted in aspiration, as appears from later sandhi forms, although no trace of it appears in the orthography of the older insular dialects. In other cases, such as before or after another consonant, the s-sound as a rule remained or was fused with such a consonant into a new sound; e.g. -st- > -ss-, -si- > -ll-.

English phonetic theory has always recognized a dichotomy between simple or short vowels on the one hand and diphthongs and long vowels on the other. More recently phonemic theory has posed the problem whether the non-short syllabics are unit phonemes or sequences of two phonemes each. Both possible answers have been given: Trubetzkoy, for example, treated them as units; Trager and Bloch held that they are groups. I myself once considered the analytic treatment inconsistent with the facts, at least with reference to ‘General American’, but have since discovered that my previous conviction was based on an error. While I now agree with Trager and Bloch on the fundamental correctness of analyzing the diphthongs and long vowels, there are some aspects of the problem that they failed to consider or, in my estimation, treated incorrectly. It is therefore worth while to take up again the different phases of the question in order to determine what conclusions can now be drawn.

This paper presents evidence to show that in certain dialects of American Speech—in ‘General American‘—phonetic [Ii], [Uu], [eI], [oU] are not structurally parallel to [aI], [aU], [ɔI] (or [oI]), but that the first group act as phonetically complex single units (single phonemes), whereas the second group function as sequences of two units (two phonemes).