^{page 300 note 1} Before doing so it would be well to say something about the method of transcription used. At the present stage it seems desirable to use as loose a system as possible, and I therefore employ the modern Turkish script supplemented by ḫ and ŋ (and occasionally ṭ to represent ). In spite of its known weaknesses this script is sufficient for the purpose, and it has two advantages: it avoids the impression that the writer is attempting any phonetic representation of sounds, and it enables him to evade such thorny questions as that of the close e, here irrelevant. In all cases where the MS. has or kesre I transcribe i, e.g. irişmek, virmek.—As types the front-vowel forms are given throughout this paper: thus -ü represents also -u.

^{page 300 note 2} The words ‘ normal ’ and ‘ exceptional ’ are here used in a purely quantitative sense: no suggestion about which is the basic form is intended.

^{page 300 note 3} This verb also occurs as vermek, with fethe, in some old MSS.

^{page 300 note 4} Vurmak occurs occasionally as far back as the 15th century.

^{page 300 note 5} Sanmak, however, should probably be regarded as an increased stem, sa-n. Eyd-ür (ayd-ur) is also an increased form, ay-d. Birakmak has the muzari biratğ-ur: being a disyllabic stem it has the sound pattern of an increased form.

^{page 301 note 1} These abbreviations are used: KD ═ Kalile ve Dimne; SN ═ Süheyl u Nevbahar; Füt. ═ Fütûhu'ş;-Şâm. For texts used in this study see p. 307.

^{page 301 note 2} It corresponds to the form in Kashgari, tegir.—In KD, 77, 6, Zajączkowski reads degär, which in my mode of transcription would be diğer. I do not believe that the root form dig- can be sound; it seems probable that Zajączkowski has been misled by the copyist's placing of the vowel points, and that the word should be değir. This seems the more probable in view of the fact that he several times gives the word kanki (═ kangi, which), as kanik, a form I have never seen anywhere.

^{page 301 note 3} It may be a variant of değir, by analogy.

^{page 301 note 4} Zajączkowski gives giyersiz, Cevahir, p. 34, fourth line from the bottom. The form giy- is supported by giyüb in SN, 98. 4; but cf. geyüb, ib. 99. 3.

^{page 301 note 5} Kashgari has both kōr-ür and kōr-er.

^{page 301 note 6} Deny, , Gramm., p. 1109Google Scholar, gives distinct muzari forms for the two verbs sormak: sor-ar (asks) and sor-ur (sucks). But sormak ═ suck is no longer used in standard Turkish, and in Old Ott. its muzari was sor-ar (SN 205,14, and 361, 12). He also (p. 393) gives bōl-er as the muzari of bōlmek, tut the standard form is bōl-er: I have no record of the Old Ott. form.

^{page 301 note 7} Noticed in Gabain, Von, Alttürk. Gr., pp. 211–12.Google Scholar

^{page 302 note 1} See Von Gabain, op. cit., p. 121.

^{page 302 note 2} The fact that Kashgari gives no -i variants does not, of course, prove that they did not exist in the speech he was recording.

^{page 303 note 1} In modern Turkish the -i variant is the standard form in combination with vermek, but it follows fourfold harmony. A tendency in this direction is discernible in Old Ott. also (see p. 313).

^{page 304 note 1} The distinction of front and back vowels is assumed. We have evidence that it was strong in Old Ott. (shown by ) though MSS. have discrepancies which are hard to explain.

^{page 304 note 2} The ü forms appear to have been maintained deliberately as ‘ correct ’ speech for a long time. Meninski, in his Grammatica Turcica (1680), gives them as the standard both for writing and for speech. Speaking of the genitive (p. 26) he says that some use i, ‘ sed Doctiores sic scribunt et legunt, ut nos hie tradimus ’, in which statement the significant word is doctiores. The change in the genitive form seems to be the cause of a curious gap in later Turkish grammar. In modern Turkish the infinitive -mek is never used in the genitive case, for no reason valid at the present day. In Old Ott. this genitive, e.g. gitmeğüŋ, is not infrequent, and it certainly continued in use into the 16th century. But the new vowel harmony would require i in this form; and there is some evidence that ŋ was changing to n in the 16th century ( for in some MSS.). As a result of these modifications gitmeğüŋ would become gitmeğin, thus coinciding with the instrumental form then in common use. This could have caused confusion, and is sufficient to explain the disappearance of the genitive form. It was no doubt a gradual disappearance. Meninski, op. cit., gives -meğüŋ as a current form, but one suspects that in his day it was used only by the doctiores. What is very interesting is the fact that the genitive form, once lost, was never restored, in spite of the later disappearance of the instrumental gitmeğin.

^{page 305 note 1} See the long list in Moravesik, , Byzantinoturcica, 2. 257.Google Scholar

^{page 305 note 2} The Greek τ is an unaspirated dental, and Greeks have difficulty in pronouncing and representing the Turkish t and d, which are aspirated. Sometimes τ is used indiscriminately for both, as in some 18th- and 19th-century ‘ Karamanlitika ’ texts (Turkish in Greek characters); ντ is the usual device for d, but a dotted has also been used.

^{page 305 note 3} Camb. Univ. Library, LI. 5,12, recorded as No. 865 in Browne, , Hand-list of Muhammadan MSS., p. 155Google Scholar, where it is inaccurately described.

^{page 306 note 1} Renderings: 1. Do not walk unwarily, open your eyes, for the day has its night. 2. Your life is the price of a kiss from the ruby of her (his) lips. 3. Believe not what he says, for he is mad.

^{page 306 note 2} But the MS. is unpointed and some of the vowels have been supplied by me. Also some of the spellings are conventional, e.g. bûsesi, which was probably bûsasi in Necâtî's time.

^{page 306 note 3} For a good reason: dir was the muzari of dimek (mod. der, demek), so that there must have been strong inducement to maintain the standard ü form in the copula in order to avoid confusion. (But der and deyü, for dir, diyü, occur occasionally in old MSS.)

^{page 310 note 1} From koymak (or kuymak ?), normal gerund form koya.

^{page 311 note 1} From komak.

^{page 311 note 2} See also the remarks about -ememek on p. 318.

^{page 312 note 1} Note in this connexion the apparent persistence of -ü with -ed- stems, mentioned on p. 313 (footnote).

^{page 312 note 2} The partial survival of -ümez in this text (see the -ememek list on p. 316) suggests that the -ü forms here are not orthographical; but as we do not know the history of the transmission of the text we cannot say positively that there has not been standardization at some stage. I do not think this likely, but we must consider all possibilities.

^{page 313 note 1} But eydi is a special case, in which other phonological tendencies may be involved quite apart from the regressive influence of görmek.

^{page 313 note 2} In this paper I assume that the form is yorumak.

^{page 313 note 3} There may also be a tendency towards persistence of -ü in increased stems in -ed: the lists show examples from SN and the Tef. fr. in combination with virmek, and one of these, ōğredü, also occurs in SN combined with gōrmek. But these examples are not sufficient to serve as a basis for any theory.

^{page 314 note 1} Association and combination will be discussed in the second part of this paper.

^{page 314 note 2} Necâtî again supplies evidence: the gerund dolaşi rhyming with mahveşi (═ mahveş + i, 3rd pers. affix).

^{page 316 note 1} Şîrâzî has one example, ide-bilmez. I have also found it in the dictionary of Helîmî, and elsewhere.

^{page 318 note 1} From a moderately good copy of the traditional text in my possession.

^{page 318 note 2} I transcribe ye as the MS. has not

^{page 318 note 3} Also in Gibb, , Hist, of Ott. Poetry, 6, p. 10Google Scholar, with small variations.—The reduced form yana-dur also occurs in a poem by Nesîmî, in the same vol. of Gibb, p. 37.