^{page 36 note 1} As an aid to investigation, fourteen records were made, and a number of kymograph tracings taken. The records are of (a) selected words uttered in isolation, and (b) readings of phrases, sentences, and a passage of narrative.

^{page 37 note 1} In certain contexts a long i: and an extra long a: may be heard. For the former, see the description of the pronunciation of i below. For the latter, see the discussion of the vowel junction ‘a’ + ‘a’ in paragraph 5.

^{page 37 note 2} The Roman orthography used in the texts studied shows the on-glide to 0 by writing uo, but omits to show it before u where it is indeed not so readily audible as distinct from the vowel itself. Initial i and e are in the orthographic texts normally written ji and je, thus jimse, jeske, but my informant omitted the j in writing and disliked an over-vigorous pronunciation of the palatal semi-vowel before front vowels, especially before i. He rejected a pronunciation jejke in which the j was pronounced as in the English word “Yes”.

^{page 37 note 3} See paragraph 6 on Syllabic Rhythm and Stress.

^{page 38 note 1} A certain distortion of the general pattern might be said to arise out of this pronunciation since the initial is not palatalized (see Table of Initials). It is as if the centralized e has demanded a middle rather than a front initial.

^{page 38 note 2} The accepted phonetic term “back vowel” is here used as part of the technical description of the realization of what is called a “middle vowel“ in the Digoron vowel system.

^{page 38 note 3} In the words uoj and uoliæj an unrounded centralized vowel was regularly used, which I wished to transcribe as ә, writing ^woj, ^wәliәj respectively. My informant, however, felt them to be as the spelling implies, namely ^woj, ^woliәj, although elsewhere he pronounced the group oj with a very close and rounded o. I am unable to decide whether these pronunciations are to be regarded as the phonetic realization of ‘WO’ + front consonant, or whether the orthography attempts to show a realization of ‘W’ + ‘8’ which does not take effect in my informant's pronunciation.

^{page 39 note 1} The occurrence of front or back vowels in initial position implies a preceding semi-vowel glide which is usually realized in absolute initial position, and invariably so in junction initial position. The lack of a potential semi-vowel on-glide to middle vowels has an important bearing on their treatment in junction position. (See paragraph 5.)

^{page 39 note 2} See Table of Initials.

^{page 39 note 3} See Table of Finals.

^{page 39 note 4} What is meant in this paper by such terms as “junction final”, “simple”, “compound”, etc., is fully set out when the consonant system is discussed in paragraph 3.

^{page 39 note 5} See paragraph 2.

^{page 40 note 1} By “word” is here understood “a word uttered in isolation”. When a word occurs in connected speech it may form part of a phonetic entity greater than itself, so that the first or last syllables of the isolated word may cease to be absolutely initial or final in a phonetic group. In general, the system of absolute and junction initials and finals proposed for isolated words holds good also for these larger phonetic groups.

^{page 41 note 1} Except, of course, in the case of W.

^{page 42 note 1} The term ‘post-velar’ is preferred here to ‘uvular’, since the absence of a separate set of velar fricatives may mean that χ and Ε do duty as fricatives for the velar as well as the uvular class. Apart from their place of articulation, they appear to be more closely related linguistically to k than to q. The function in the phonological system of q, which appears to be something of an “odd man out”, calls for investigation.

^{page 43 note 1} See Catford, J. C., “On the Classification of Stop Consonants”, in Le Maitre Phonétique, 1939.Google Scholar

^{page 43 note 2} See paragraph 4.

^{page 51 note 1} This word may also be pronounced .

^{page 52 note 1} See paragraph 1.

^{page 54 note 1} Professor Bailey writes: “The Dictionary has dzsehun as a secondary form beside zsehun ‘say’. [Our informant] believed himself always to say dzsehun ‘I say’, dzaxta ‘he said’. In the printed texts I have, however, regularly seen zaxta.“

^{page 56 note 1} For comments on the use of a true velar fricative in this word, see p. 61.

^{page 57 note 1} See p. 42 on distribution of j, W, q, and

^{page 58 note 1} In paragraph 1.

^{page 58 note 2} See earlier discussion of j as Front Initial, on p. 52.

^{page 58 note 3} My informant's spelling. Printed texts show khepp, thæpp.

^{page 59 note 1} For the onomatopœic word for a hoopoe bird, my informant wrote either ubnbub or ububup.

^{page 59 note 2} -v is very rare as the spelling of a simple absolute final and, in my informant's usage at least, appears to be an alternative to the spelling -f. He wrote sometimes istuf, sometimes istuv.

^{page 60 note 1} The term “unvoiced sound” is here used to mean one in which such features'as force of articulation, lack of aspiration, etc., are those appropriate to a voiced rather than to a voiceless sound.

^{page 60 note 2} See Tables VII and VIII on pages 62 and 63.

^{page 65 note 1} This distribution is not confirmed by the spelling of the Pamyatniki, where a number of the words pronounced by the speaker with i are spelt with æ.

^{page 67 note 1} The speaker felt very strongly that he kept the pair art “fire” and aid “oath” apart in pronunciation. As far as my observation went, distinction was often, but I believe not always, maintained by aspiration after the final t. In connected speech, where the word ard forms part of a sequence like ard in, the d is, of course, fully voiced. My informant admitted, however, that the plural forms ardtәardtæ and artәaruæ might be indistinguishable. This is as one would expect, since the phonetic context makes full voicing of the d unlikely and rules out the possibility of distinction by aspiration, as the first element of the group -tt- is not exploded.

^{page 68 note 1} Other than those marked in the Tables of Initials as not being regular members of the consonant system.

^{page 69 note 1} The gemination which may occur here is of type (b).

^{page 69 note 2} For length of final p, see p. 58.

^{page 69 note 3} See p. 78.

^{page 70 note 1} Professor Bailey was good enough to give me the historical grammarian's explanation of the gemination after fæ- ni-, with which students of Indo-Iranian philology are doubtless familiar. He also pointed out that historically speaking one must postulate two fæ- prefixes and two ni- prefixes, one of each pair having the geminating property here described while the other has not.

^{page 70 note 2} In many of these words the speaker felt the final of the first syllable to be g, d, etc., rather than k, t, etc.—perhaps for etymological reasons. When he used a slow and“careful“style of pronunciation for such words, the beginning of the medial stop was sometimes accompanied by voice. In a quicker and probably more natural style, the whole stop was voiceless. Only the voiceless pronunciation is shown here. The orthography in many cases shows the first part of the stop as voiced but recognizes that there is no voice at the moment of release. I cannot account for the apparently reverse procedure sometimes adopted towards the affricates, as when is written xulutdzag (cf. xuludz).

^{page 71 note 1} Junctions where the second vowel is a are special cases, and are dealt with as such later.

^{page 71 note 2} This spelling is used in the texts consulted for this study, but Professor Bailey tells me that other texts also write -juoncæ, -iuoncæ.

^{page 72 note 1} ‘ә’ + ‘o’ and ‘ә’ + ‘u’ are realized as ә^wo, ә^wu, e.g. dә^w obaw dæ obau, dә^wurn dæ uru, and so do not call for special mention.

^{page 72 note 2} See realization of ‘ba’ + ‘avәrun’ as bavәrun. If the correct formula were ‘ba’ + ‘ivәrun’, we should expect the phonetic outcome to be bajvәrun. avәrun, bavәrun and nivәrun may take their places with perfect regularity as terms of a series of the kind set out on p. 69.

^{page 73 note 1} Junction (vi) prompts a digression of a certain linguistic interest, which illustrates some of the snares in the path of the investigator who has only one speaker to draw upon.

My informant's speech contains the following series, comparable to those set out on p. 69: afun, fәjjafun, rajafun, bajafun, jijjafun. The four last terms of this series were interpreted by the speaker and by myself as realizations of the junctions ‘fә’ + ‘afun’, ‘ra’ + ‘afun’, ‘ba’ + ‘afun’, ‘ni’ + ‘afun’. Of these, only the last, which shows gemination of the appropriate linking glide (see p. 71) is what might have been anticipated as the regular outcome of the junctions postulated. fәjjafun appeared to be irregular since ‘fә’ + ‘afun’ should give fafun in utterance [see Junction (xii)], and there was no precedent for a linking j between a and a. rajafun and bajafun led me to propose aja as a possible realization of ‘a’ + ‘a’ The implications of this hypothesis were far-reaching and full of apparent contradictions. It would be tedious to describe the various lines of approach that were taken up in a fruitless attempt to find an explanation of the linking j in accordance with other observed phonetic facts. When the matter was brought to his notice, Professor Bailey shed light on it at once by pointing out that the recognized form of the verb which is pronounced af un in the speech of my informant is әjafun, and that the first term of this series must be regarded as unorthodox, and perhaps individual. My informant himself adhered stoutly to afun, but conceded that it would be possible for him to pronounce the form with prefixed ‘ra’ as raifun, thus bringing his realization of ‘ra’ + ‘afun;’ into line with that of ‘ba’ + ‘amontse’, etc.

^{page 73 note 2} All the examples collected of this pronunciation (about half a dozen in all) occur after χ, which may be signficant. The speaker himself was not prepared to admit that he prounounced u in these words; hefelt he was saying wæ. An alternative transcription would be -Wә.

^{page 74 note 1} In the little work I have so far been able to do on the Iron dialect of Ossetic, I have found that the stress is much more clearly defined.

^{page 74 note 2} My informant arranged words and phrases of four syllables under the following given “accent” headings: ////, / X X X, X X/X, X/X/, /x/X, X//X. His examples are omitted for the sake of brevity. In attempting to interpret these lists, allowance must, of course, be made for the relatively small number of examples, the fact that the arrangement was made on one occasion only, and may not be constant, and the fact that my informant probably never fully understood the distinction it was attempted to draw between stress and quantity rhythms.

^{page 75 note 1} I have deliberately chosen as examples nouns of a specific class which take the same inflexional forms, since this brings out more clearly the differences between them that must be ascribed to differences of phonetic structure. The points of phonetic interest that arise out of these paradigms were common to those of all classes of noun examined.

^{page 75 note 2} The names of the cases are as given by Professor Bailey.

^{page 77 note 1} The names of the tenses are taken from Miller's Die Sprache der Osseien.

^{page 78 note 1} See paragraph 2, Syllabic Analysis and Syllabic Structure.

^{page 78 note 2} This fable appears in A. Christensen's Teoctes Ossetes in both Iron and Digoron versions. The text of the transcribed passage is based on Christensen's Digoron version, but certain minor alterations were made by my informant.

^{page 78 note 3} The printed texts show zaxla for this word, which was always written dz- and pronounced dz by my informant. See footnote on p. 54.