¹ The kind of Persian described here is the standard modern language; unless otherwise indicated, the transcription of the examples represents the style of pronunciation heard on the radio and in formal speech. The statements about stress hold also for the colloquial style, which differs from the formal standard chiefly in having certain assimilations and elisions and in showing contracted forms of certain common verbs. The transcription of Persian words quoted from grammars or dictionaries is adjusted to the essentially phonemic transcription used in the forms cited directly. All cited words were checked with at least one native speaker of Tehrani Persian; in many instances several speakers were consulted. My principal informant was Hosein Nasr, a graduate student at Harvard University.

² The best accounts known to the writer are H. Jensen, Neupersische grammatik 21–3 (Heidelberg, 1931), and V. S. Rastorgueva, Kratkij očerk grammatiki persidskogo jazyka 620–1, appendix to B. V. Miller, Persidsko-russkij slovar' (Moscow, 1953). Jensen made full use of previous scholarly publications: books are listed in his bibliography 6–8, and articles are referred to in footnotes. More recent discussions may be found in E. Rossi, Grammatica di persiano moderno 17–8 (Rome, 1947), and in A. K. S. Lambton, Persian grammar (Cambridge, 1954), scattered through the book in appropriate grammatical sections; most of these can be located by the index s.v. stress. The only textbook of modern Persian known to the writer which has consistent marking of stress throughout is Spoken Persian [by C. T. Hodge and others] (Washington, [1956]; multilith).

³ The older dictionaries ignore the matter of stress almost completely. Of modern dictionaries, S. Haïm, New Persian-English dictionary (Tehran, 1934–6), gives indication of nonfinal stress fairly frequently, while A. K. S. Lambton, Persian vocabulary (Cambridge, 1954) gives none. Only Miller (op.cit. in fn. 2) has consistent indication of nonfinal stress throughout the dictionary.

^3a An excellent article, accessible to me only after the present study was in proof, is M. Lucidi, L'accento nel persiano moderno, Ricerche linguistiche 2.108–40 (1951). Lucidi covers much of the same ground as my paper, suggesting that some features of Persian sentence stress are ancient and that the whole question of accent in Iranian deserves further study. On some points he offers more detailed treatment; on the other hand, he does not attempt full coverage of word stress, regards length as a primary feature of the vowel system, disregards the vocative stress, and in several other matters differs from the position taken here.

⁴ Rastorgueva 616. The interesting and important problem in diachronic phonology posed by the replacement of quantitative by qualitative differences in the vowel systems of Iranian languages is treated at some length in the appropriate chapters of V. S. Sokolova, Očerki po fonetike iranskix jazykov, 2 vols. (Moscow, 1953).

⁵ Hodge apparently regards the two levels as determined, with the conditioning factors being either order (e.g. later stress in the same phrase is stronger) or intonation (e.g. stress is weaker when coincident with low pitch). See Hodge, note 1.2.3.

⁶ The term ‘word’ as used in this study is a morphological term to be understood pretty much in its traditional sense in Western scholarship about Persian. Undoubtedly a rigorous definition could be devised on the basis of morpheme distribution, and it may be assumed that the word as defined by such functional criteria would coincide with the traditional word with only a few points of disagreement or indeterminacy. Traditional Persian orthography, however, is not completely consistent in showing a space at the word boundary. A given affix may be joined under certain circumstances and separated under others, even though its phonological and morphological status remains identical. An important factor in this is the previous history of the affix.

The term ‘word’ cannot be used here in a purely phonological sense. Word boundaries in Persian tend to be marked by phonological junctures; but in many instances no clear juncture is audible, and on the other hand many words, often Arabic loanwords with original 'ayn or glottal stop, show the phonetic characteristics of juncture embarrassingly present where they are not useful for morphological analysis. Obviously, careful analysis of the juncture system of spoken Persian is in order, but the results are unlikely to affect the major points of the present study.

⁷ J. Kuryłowicz, L'accentuation des langues indo-européennes 446 fn. (Krakow, 1952), offers a generalization about the stress of monosyllables in languages of various stress systems, relevant here.

⁸ C. F. Hockett, Two models of grammatical description, Word 10.210–33 (1954).

⁹ In the classical language mi- or hæmi- and be- occur with other verb forms, and other combinations of two prefixes occur; see Jensen 134–7. Also, in modern dialects other prefixes () and other combinations occur. In general the first of two prefixes is stressed, except that in some dialects the negative prefix is stressed even in second position; e.g. A. Christensen, Le dialecte de Sämnān 240 (Copenhagen, 1915).

^9a Present-stem forms without mi- or be- are relatively rare in the modern language. The following instances occur. The aorist of many compound verbs is regularly formed without be-; the be- forms of budæn ‘to be’ (present stem hæst- or zero) are bašæm etc.; the present of daštæn ‘to have’ is dapæm etc.; and prefixless forms of xastæn ‘to want’ are combined with the apocopated infinitive (= the past stem) to yield a periphrastic future tense xahæm ræft etc. Since the prefixless forms with compound verbs are normally stressless (see §3.1 below), it is difficult to determine the position of the word stress; the forms bašæm etc. and daræm etc. have word stress on the stem, not the ending. The forms of xastæn in the future are normally stressed on the ending and the following infinitive is unstressed. A parallel to the last stress pattern can be found in the compound forms with daštæn ‘to have’, such as daræm mikonæm ‘I am (in the process of) making’, which seem to be stressed either on the mi-, with daræm unstressed, or on the ending of daræm, with mikonæm unstressed. The forms of xastæn and daštæn in these constructions may be regarded either as in some sense preverbs or else as parallel to the past participle in the perfect tenses, kærdé sem, kærdé budæm, kærdé bas̆æm, etc.

¹⁰ The double pp in this word is surprising since the stem of the verb is pa-, but this pronunciation is given in Haïm and confirmed by Persian speakers. No other example of such doubling is known to the writer.

¹¹ For -o Rastorgueva 653, 661; Jensen 188–9; Lambton, Grammar 138–40. The morpheme -hæm is not usually regarded as a suffix in grammars, more attention being given to its independent and prefixal uses, but it is mentioned briefly in Rastorgueva 653 and by Hodge note 4.5.3.

¹² Lengthy discussions of the syntactic roles of -ke may be found in standard grammars ; there is apparently no discussion in print about the morphological status of -ke or even about the conditions under which it is stressed. Rastorgueva 620 fn. remarks merely that it ‘is not always enclitic’.

¹³ A fairly full listing and brief explanation of these alternants are given by Hodge, note 4.2.3.

¹⁴ Hodge, note 5.2, apparently feels that there is always a vowel difference in addition to the stress difference.

¹⁵ This -í suffix derives historically from several sources, but these have clearly fused into one morpheme in modern Persian, and attempts to separate them on a historical basis are futile.

¹⁶ The position of stress in Arabic is assumed to be that most widely used in modern reading of the classical by Arabs, without further inquiry into the history or phonological status of such stress. Cf. H. Birkeland, Stress patterns in Arabic (Oslo, 1955), and A. Satterthwait, Classical Arabic stress (unpubl. M.A. thesis University of Pennsylvania).

¹⁷ Lambton, Vocabulary 176, gives vædey for vælí, but Haïm, Miller, and the Persian speakers consulted admit only vælí; væléy is archaic or dialectal.

¹⁸ Most of these words are of verbal origin : s̆áyæd is a prefixless 3rd person singular present or aorist form (cf. s̆nayestæn 'be fitting'), -tævan is a present stem without personal endings (cf. tævanestæn 'be able'), gúya is an old optative (cf. goftæn 'say').

¹⁹ Of these words five , dæríqa, , xodáya, and xóša are old vocatives. In earlier Persian this unstressed vocative suffix -a (Jensen §56, §70) was added freely to nouns (including personal names), less often to adjectives; these five words seem to be the only vestiges of this usage.

Many verb forms, but is now limited to these words, which today are isolated and no longer part of the verb system. It is probable that these two -a suffixes, vocative and optative, were one morpheme in classical Persian. Their morphemic status in modern Persian is precarious. Jensen §§169 (Anm. 1), 184, 187, 254, 255.

²⁰ The nonfinal stress on these two words is not noted in either M or H except in the compound gær čænd Hl.130, but in Rastorgueva 620 they are listed among examples of non-final stress. Lambton, Grammar 74, lists mægær with initial stress and notes that in ægær stress falls ‘on the initial or final syllable’.

²¹ In the colloquial style some titles are drastically shortened and form contractions with the following name. They still show the shift of stress, however, when vocative. For example, kerbela'í hæs'n, voc. kérbela'i hæsæn, has colloquial forms kælbæsn, voc. klbæsæn. Other examples of short forms: aqaye > aye, mæšhædi > mæšdi, mæšti.

²² Examples may be found in Hodge, e.g. áqa 36, xánom 51, aqa (unstressed) 55, fátme 108. A nice example of contrast appears in a conversation 106, áqaye pakrævan (vocative in first sentence) : aqáye pakræván (subject of sentence later).

²³ At least one instance is recorded in Hodge 107 (colloquial style). One speaker says némidunæm ‘I don't know’; the other replies čiye 'What do you mean by “I don't know”?'. This may be the same phenomenon referred to in A. Bausani, Di una possibile origine dell'accentuazione sull'ultima sillaba in persiano moderno, Oriente moderno 27.123–30 (1947), esp. 124–6. Most of Bausani's examples, however, seem to be instances of the 'drawl' feature of spoken Persian, which consists of overlengthening clause-final (including sentence-final) syllables under certain conditions. This drawl, which Hodge symbolizes with a superscript colon, is independent of stress and is not discussed in the present article.

²⁴ A brief description, with examples, in C. A. Ferguson, Spoken Damascus Arabic, Unit 8, Pronunciation 2 (Washington, 1949; mimeogr.). It can be observed in the sentences of the course as revised by F. A. Rice, Spoken Damascus Arabic (Washington, 1954; multilith), in which stress and intonation are indicated in the transcription.

²⁵ E. V. Sevort'an, Fonetika tureckogo literaturnogo jazyka 142 (Moscow, 1955). N. A. McQuown and S. Koylan, in Spoken Turkish (Madison, 1943), do not mention this phenomenon, but many examples occur in transcribed sentences, e.g. Ne arzu ediyorsunuz? (9), Sinema saat kaçta başhyor (13).

²⁶ G. Bergsträsser, Zum arabischen Dialekt von Damaskus 45 (Hannover, 1924).

²⁷ Apparently not described for Turkish but verified by observation of native speakers and indicated by the stress markings in McQuown, e.g. beş kuruş (11), çok dogru (129), her türlü (268), etc.

²⁸ McQuown 246–7.

²⁹ Bergsträsser, loc.cit. Most Lebanese Arabic, however, has unstressed ma bá'rif, etc.

³⁰ For a convenient summary of Greek and Sanskrit evidence see H. Hirt, Indogermanische Grammatik 5.335–8 (Heidelberg, 1929). Cf. also Kuryłowicz, L'accentuation 112, 135.

³¹ Hirt 293–319, 338–65; Kuryłowicz 105–20.

³² Hirt 400.

³³ A. Meillet, La déclinaison et l'accent d'intensité en perse, Journal asiatique 9.15.254–77 (1900).

³⁴ R. Gauthiot, De l'accent d'intensité iranien, MSL 201–25 (1918), Grammaire sogdienne 29–38 (Paris 1914–23); J. Kuryłowicz, Traces de la place du ton en gāthique (Paris, 1925), L'accentuation 438–51; G. Morgenstierne, Indo-Iranian frontier languages I 30–2 (1929) and elsewhere; S. Konow, Saka studies 12–6 (Oslo, 1932) and elsewhere. A strong criticism of this principle, especially in verb forms, appears in the important article by P. Tedesco, a-Stämme und aya-Stäunme im Iranischen, ZII 2.280–315 (1923) ; the development of Persian verb forms is discussed 302–5. I owe this reference, as well as several helpful criticisms and suggestions, to Warren C. Cowgill of Yale University.

³⁵ G. Morgenstierne, Archaisms and innovations in Pashto morphology, NTS 12.95–7 (1942) suggests that certain features of Pashto stress preserve original Iranian accent. For brief summaries of stress in Pashto see H. Penzl, A grammar of Pashto (Washington, 1955) ; D. A. Šafeev, Kratkij grammatičeskij očerk afganskogo jazyka 1043, in P. B. Zudin, Russkoafganskij slovar' (Moscow, 1955). The stress contrast in Pashto has much greater yield than in Persian; there are lexical pairs such as 'knot' : ‘kind of duck’, and shift of stress plays a considerable role in the noun declension.

³⁶ The most detailed treatment of stress known to the writer for any modern Iranian language is V. I. Abaev, Ritmika osetinskoj reči in Osetinskij epos 96–134 (Moscow, 1939), repr. in Osetinskij jazyk i fol'klor 529–60 (Moscow, 1949). Brief summaries of stress in Ossetic are also available elsewhere. The stress system of Ossetic shows many interesting features, including a shift morpheme of definiteness bælás ‘tree’ : ‘the tree’.