Our reconstruction of a proto-language is theoretical and partial; but the language itself was necessarily real and whole. As a real language, it shared with all real languages the characteristic of non-uniformity. Some of its speakers spoke differently from others; they spoke differently from their linguistic predecessors and successors. And at any time, in any dialect, the phonemes of a real language display variations in different phonetic environments. In short, the language which our reconstructions adumbrate had linguistic change; it had isoglosses; and it had allophones.

The element -ti-, -sti-, -stri- in Latin words like rūsticus, domesticas, terrestris goes back to a -ti-/-tri-formation of adverbs or prepositions denoting space and time; thus, *domesti-: *entesti-, *nemestri-: *ent(e)ri-. The terms formed on these adverbs and prepositions refer not only to location, but also to family, household, and social organization in general. They express a sharp distinction between terms designating persons or objects within the household and those outside it. In a general way, this contrast is expressed by the element r on the one hand and the element i on the other; no special function can be ascertained in this connection for the element t. Adverbs of space appear partly with t, partly without it. The r or tr came to be used in comparative forms; originally its meaning was spatial. It seems that the function of designating a contrast, which preceded the function of designating a comparative, developed in expressions meaning such things as ‘above’ and ‘below’: an object that is above another is ‘more above’ than the object with which it is contrasted.

The Sui language is spoken in the southeastern part of Kweichow, China. While the language can be shown to have certain relations with the Tai group, it exhibits many peculiarities and divergences from the ordinary Tai languages. The material for this study is taken chiefly from an analysis of over 2600 lines of folk songs, gathered by the writer in 1942 from two small villages, Li and Ngam, in Lipo, Kweichow.

1. The principle of descriptive analysis of constructions on the basis of immediate constituents has been recognized by most descriptive linguists, and most of them have employed immediate constituents in analytical procedures, though in some instances not directly so. Linguists have also recognized that there are certain significant features about such constituent elements of constructions. Bloomfield formally treated some of these features and called them taxemes. But not all the significant features of constructions were presented by Bloom-field, and not all the interrelationships of these features have been adequately handled. For example, Bloomfield's four taxemes are not of coordinate value in the description and classification of constituent units.