¹ See George K. Zipf, Selected Studies of the Principle of Relative Frequency in Language (Cambridge, Mass., 1932); The Psycho-Biology of Language (Boston, 1935), chapter 2. I am personally indebted to Dr. Zipf also for calling these French word-counts to my attention.

² Psycho-Biology 47.

³ Reference is to the fully inflected words as used in speech, not to lexical units.

⁴ This constant was found by Zipf, in his English and Latin counts, to be approximately equal to 1/10 of the total number of running words in the sample.

⁵ M. L. Hanley, Word Index to James Joyce's Ulysses (Madison, Wis., 1937).

⁶ See G. K. Zipf, Homogeneity and Heterogeneity in Language, Psychological Record 2.350 (1938), where the curve for the Hanley statistics is illustrated.

⁷ For the term ‘articulatory’, with its correlative ‘conceptual’, see Psycho-Biology 229 ff. These terms are roughly equivalent to the commoner terms ‘empty’ and ‘full’, borrowed from Chinese grammar, but are, I believe, more descriptive. Articulatory words are articles, pronouns, prepositions, conjunctions, auxiliary verbs, and some adverbs; in general, they are the most frequent words in any language.

⁸ Psycho-Biology 44–7.

⁹ Dr. Zipf has told me that Professor John C. Whitehorn of Washington University has observed that for the language of personal letters and colloquial discourse the frequency distribution bends down at the top. Though this observation has never been published, Dr. Zipf has assured me that his own investigations substantiate the Whitehorn finding.

¹⁰ It should be emphasized here that all credit for these counts is due to their compilers, who analysed the original samples of French. All I have done is to reduce their results to graphic form so as to compare their frequency distribution with the standard curve. I owe a further debt of gratitude to Mr. Dabbs for putting his data at my disposal, and to Professor Bandy for providing me with additional information regarding his published statistics.

¹¹ Mercure de France, vol. 271 (1936), from voici 129.7 (= page 129, line 7) to plus 136.14.—Nouvelle revue, vol. 146 (1936), from les 241.1 to les 246.11.—Revue de Paris, Sept. 1, 1936, from 56.1 to dans 61.21 (omitting fn. p. 60).—Revue de deux mondes, vol. 35 (1936), from M. 230.1 to ces 234.36.—Revue hebdomadaire, Jan. 2, 1937, from oeuvre 119.3 to le 126.3.

¹² Mr. Dabbs's statistics appear to be highly accurate, to judge by the parts which I have checked through. The material was presented with great fulness of detail, which made possible further experimentation with the criteria of word-definition, as will be brought out below.

¹³ A Word Index to Baudelaire's Poems (Madison, Wis., 1939). A total of 33,268 running words is included. Checking through the list, I arrived at a total of 33,276, a discrepancy of only .024%; this indicates a very high degree of accuracy for the index.

¹⁴ The straight line of the RF = C formula would cut the horizontal lines or treads of the standard curve at a point only slightly to the left of the middle. See Zipf, Homogeneity and Heterogeneity; and also Zipf, National Unity and Disunity, chapter 1 (Bloomington, Indiana, 1941). I am assuming that here the C of the formula is equal to 1/10 of the sample for French as it was for the samples of English and Latin which Zipf analysed.

¹⁵ On stress as one criterion for word-division in English, cf. L. Bloomfield, Language 180 (New York, 1933).

¹⁶ Cf. Charles Bally, Le Langage et la vie² 61 (Zurich, 1935), where he remarks upon the 'agglutination, déjà très avancée, des prépositions grammaticales avec l'article (cf. fr. du, des, au, aux), complice elle-même de l'agglutination de l'article et des “déterminatifs” avec les substantifs (cf. lom' = “l'homme”) action combinée qui a pour effet de créer des blocs tels que duroi, delom', à peine plus analytiques que lat. regis, viri.'

¹⁷ Psycho-Biology 47–8.

¹⁸ This paper has profited by correspondence with Dr. Martin Joos, who made several useful suggestions as to the presentation of the material.