¹ An earlier version of this paper was presented on 28 Dec. 1964 at the New York meeting of the Modern Language Association of America. The formulas offered in the course of the paper were developed in discussions with my colleagues at Ohio State University, particularly D. Terence Langendoen and Robert C. Channon. I am also grateful to Alo Raun and Robert T. Harms for many valuable comments and suggestions, not all of which have been followed in the final version of this paper.

The research upon which this work is based was supported by the National Science Foundation of America.

² Exceptions are some recent loanwords and certain native compounds in Estonian, and words to which a special type of emphatic stress has been applied in Finnish.

³ For a description of Finnish quantity, see for example Antti Sovijärvi, Über die phonetischen Hauptzüge der finnischen und der ungarischen Hochsprache, Ural-Altaische Bibliothek II (Wiesbaden, 1956), and Paavo Ravila, ‘Quantity and phonemic analysis’, Proceedings of the Fourth International Congress of Phonetic Sciences, Helsinki 1961 490–3 (The Hague, 1962).

⁴ The intervocalic behavior of certain clusters consisting of plosives and resonants requires special treatment at the segmental-quantity level in Finnish as well as in Estonian. At the level of syllabic quantity, such clusters fit within the general patterns.

⁵ The idea of expressing the suprasegmental structure of a word by a formula showing the number of prosodic possibilities as a function of the number of syllables was suggested by a recent article by Pavle Ivić, ‘Broj prozodijskih mogućnosti u reči kao karakteristika fonoloških sistema slovenskih jezika’, Južnoslovenski filolog 25.75–113 (Beograd, 1963).

⁶ For a summary of the history of the problem, cf. my ‘Segmental and syllabic quantity in Estonian’, American studies in Uralic linguistics 21–82 (Bloomington, 1960), and Robert T. Harms, Estonian grammar 143–67 (Bloomington, 1962).

⁷ These arguments are developed more fully in my paper ‘Juncture’, to appear in the Proceedings of the Fifth International Congress of Phonetic Sciences (Münster, 1964; in press).

⁸ The Estonian words are quoted in their standard orthography. In the case of plosives, the letters b d g stand for a short voiceless lenis consonant. Intervocalic p t k indicate a so-called short geminate; pp, tt, kk stand for overlong geminates. One vowel symbol is used for a short vowel; standard spelling does not distinguish between the long and over-long durations of vowels, both symbolized by a sequence of two vowel letters.

⁹ These generally known facts are outlined at greater length in my paper cited in footnote 6. Cf. also Paul Ariste, Eesti keele foneetika 89–95 (Tallinn, 1953).

¹⁰ Cf. my ‘The function of consonant quantity in establishing phonological units in Estonian’, to appear in American studies in Uralic linguistics II (Bloomington, 1966; in press).

¹¹ The formula applies to noncompound words. In the case of compounds and certain derivative suffixes that share many characteristics of words, the pattern starts again after the word or suffix boundary. Such derivative suffixes have not yet been exhaustively studied in Estonian. For a similar phenomenon in Finnish, cf. Martti Sadeniemi, Metriikkamme perusteet 43–9 (Helsinki, 1949).

¹² Vowel quantity 2 is never followed by consonant quantity 3, and vowel quantity 3 is never followed by consonant quantity 2. The problem is discussed in more detail in the article quoted in footnote 10.

¹³ The dissyllabic units here postulated were derived on the basis of observations of possible quantity contrasts in present-day Estonian. The significance of dissyllabic units in the historical development of Balto-Finnic languages and of Lappish has been recognized before. In a different context, Erkki Itkonen has written (Struktur und Entwicklung der ostlappischene Quantitässysteme 17 [Mémoires de la Société Finno-Ougrienne 88; Helsinki, 1946]: ‘In bezug auf den Stufenwechsel müssen wir also konstatieren, dass die gegenseitigen Beziehungen des Vokals und des nachfolgenden Konsonantismus auf der Grenze der ersten und zweiten Silbe prinzipiell andere als auf der Grenze der 2. und 3. Silbe sind. Hätten im letzteren Fall Vokal + Konsonantismus ein ähnliches festes Ganzes wie im ersteren Fall gebildet, so hätte der auf den Vokal der 2. Silbe folgende Konsonant immer im Stammkonsonantismus die schwache Stufe ausgelöst, also auch dann, wenn die 2. Silbe offen war. In die Ursachen dieser Doppelheit wollen wir uns hier nicht vertiefen—die Aufhellung der Frage setzt eine Lösung aller mit der ursprünglichen finnisch-lappischen Akzentuation verknüpften Probleme voraus—; jedenfalls dürfen wir es wohl als ausgemacht betrachten, dass die starke Stufe vor einem auf der Grenze der zweiten und dritten Silbe stehenden Einzelkonsonanten zunächst darauf beruht, dass der genannte Konsonant, wie man es aufgefasst oder empfunden hat, den Vokal der dritten Silbe verdeckt und den Vokal der zweiten Silbe offen lässt, mit anderen Worten auf der Silbenteilung, die nach der Ansicht Panconcelli-Calzias eine blosse Erdichtung ist.’ Itkonen recognizes also that longer words are combinations of dissyllabic or trisyllabic ‘speech measures’ (ibid.).