[Linguistic relationship has been measured statistically by means of the product-moment correlation coefficient, r. The linguistic meaning of various forms of this coefficient is discussed on the basis of a simplified model. It is maintained that the most satisfactory statistic measures degree of correspondence or similarity rather than relationship in the genetic sense. When applied to Indo-European data, the statistic yields results in good agreement with common philological judgment. Problems of significance are discussed. Finally it is concluded that the statistical technique will both require and help to establish a taxonomy of languages.]

A generation or two ago the writers of handbooks of Indo-European comparative grammar were for the most part considerably more definite in their conclusions regarding the form of the parent speech than their successors are today. This is particularly true with regard to the verbal system. In those days the repertoire of verb categories was based largely on the agreements between Greek and Sanskrit, some of which were indeed striking, and the deviations of the other Indo-European languages were explained as developing out of that skeletal system, or as innovations, or were left without explanation.

This paper presents in summary form the chief results of an extensive lexicostatistic study of Germanic dialect evolution. I undertook the study, on the one hand, in order to put the theory and technique of glottochronology, as set forth in the main by Swadesh and Lees, to a searching test within a sizeable system of related dialects, where historical, literary, and phonological data were relatively abundant. On the other hand, I hoped in this manner to secure new evidence bearing on the controversial genesis of the Germanic dialects. If the results fell into a well-marked chronological pattern supported at key points by extraneous evidence, this pattern might then be juxtaposed and critically correlated with the leading nonstatistic doctrines in the field, notably the tripartite theory of quasigenetic descent, and the modern concepts of diffusion and synthesis represented chiefly by Friedrich Maurer and Ernst Schwarz.

Modern surgery has legitimized exploratory operation with the understanding that it should provide such otherwise inaccessible information as may guide medical scientists in embarking on the least hazardous course of the decisive final action. The present paper is likewise exploratory, inasmuch as the study of the Latin suffix -āceu in Romance has for decades been obstructed by a thick and ever-expanding growth of side-issues, through which a preliminary path must now be cleared. Not all of these by-problems will come up here for detailed discussion, still less for solution, but most will be squarely stated and assigned to a specific place in an intricate web of multiple relationships. More important, an attempt will be made to isolate the core of the issue and to provide a tentative new answer to a crucial question of identity where older accounts have turned out to be unsatisfactory. The immediate aim, then, is to create on slippery and overgrown terrain a temporary base for future investigations.

Serbo-Croatian is an Indo-European language, and like many languages of that family possesses a complex system of nominal and verbal inflection. The forms of the standard language have been fairly well described, though not from the structural point of view, in a number of handbooks. As a source of forms I have used August Leskien's Serbokroatische Grammatik (Heidelberg, 1914); but I have checked all forms against the speech of at least one informant—my wife, a native speaker of the Serbian regional standard, born in Valjevo, long resident in Belgrade.

The question of the nonuniqueness of phonemic solutions, first raised by Chao in 1934, remains one of the most intriguing problems of structural linguistics. It is therefore of interest to note that the attempts of various American and European linguists to arrive at a structural analysis of the nonsyllabic phonemes of present-day standard Bulgarian have resulted in three significantly different descriptions of the system in question. The phonemic inventory most frequently advocated, henceforth referred to as system I, includes (1) a series of hard, nonpalatalized, or plain consonants /p b t d k g f v s z x c m n r l/; (2) a series of corresponding soft, palatalized, or sharp consonants /p' b' t' d' k' g' f' v' s' z' x' c' m' n' r' 1'/; and (3) a glide or nonconsonantal-nonvocalic phoneme /j/. System II, less frequently encountered, admits only groups (1) and (3) as independent phonemes, and regards the palatalized consonants as clusters of consonant plus /j/. System III recognizes the phonemic status of (1) and (2), but considers [j] an allophone of unstressed /i/.