- In this paper, quotation marks are used to enclose letters or symbols, as opposed to phonetic or phonemic units. Thus “e” means ‘the letter e’.
- In a recent paper, “The Phonemic Structure of Zographensis,” Language 28.360–5, G. H. Fairbanks undertakes to find the phonemes of the “language represented by the Zographensis manuscript,” without ever defining what he means by this phrase. It emerges from his procedure that he means the original translation, since he ignores nearly all data contradicting the normalized standard. However, his final sample transcription is based directly on the spelling of Zogr, including without explanation certain forms which are spelled differently in other parts of the text. His stated purpose is “to see what kind of analysis can be made of an Old Bulgarian document by descriptive techniques alone, and to note the problems suggested by historical considerations on which there is no descriptive evidence” (footnote 2). He operates with numerous unstated assumptions, some of which inevitably affect his conclusions, and not only does he fail to state the historical evidence on which he makes certain significant choices (and to note the historical problems), but he passes by much of the descriptive evidence without notice. The article gives the impression that he has made a thorough analysis of an original text written by an expert phonetician who has, however, been slightly inconsistent.
- Of the 288 surviving folios, 17 are a later addition using a different spelling system. V. Jagić's publication of the text in cyrillic transcription, Quattuor Evangeliorum Codex Glagoliticus olirn zographensis nunc petropolitanus, Berlin 1879, marked the beginning of modern OCS studies. Zogr has since been the principle foundation for all descriptions of OCS.
- It is necessary to assume that the alphabet was essentially systematic. We are assuming that a native speaker would most likely follow the principle of one-phoneme-one- symbol, not bothering with phonetic detail. Exceptions and inconsistencies could come from other writing systems he knew or simply from his lack of inventiveness and care. It is not important here whether Cyril was by race a Slav or a Greek, since all evidence indicates that his knowledge of Slavic was that of a gifted and imaginative native.
- Jagić himself states that the letter does not occur in “the oldest documents” (Glago- ličeskoe pis'mo, SPb 1911, p. 205), and he gives no example from Zogr.
- Fairbanks states (Language 28.360) that there are 21 consonant symbols, and he includes as units Št and žd. Here he has been misled by his prejudices into a wrong assumption: neither does a unit ši-letter occur in the portion of text he based his analysis on (something less than one-fifth of the older part of Zogr), nor does a unit žd occur anywhere in OCS, except as a ligature which is clearly a space-saving device, less frequent than “ml” or “tv”. It is, to be sure, convenient for grammatical description to treat št and žd as units, for they are in morphological alternation with T and d, respectively. But there is no descriptive evidence to prove them unit phonemes any more than is the case in modern Bulgarian. Fairbanks does not mention “ǵ” or “phi”; “theta” does not occur in the text he analyzed.
- Exceptions: ǵazofilakiὅ ‘treasury’ Luke 21:1, ǵolъǵota ‘Golgotha’ John 19:17 ∼ gazo-fulikˆiὅ Mark 12:43, golъgota Mark 15:22.
- Fairbanks excludes foreign words from the discussion entirely, citing vowel successions unfamiliar to the Slav. He fails to mention the problems of consonantism involved. Since it is difficult in some cases to determine the position of foreign phonemes or phoneme groups even in modern languages where we have native informants, it is justified to put them aside here. But surely such words as ‘angel’ angelъ and ‘gospel’ evangelie must have been current among all baptized Slavs, and the careful distinction of kˆesaŕъ ‘the Roman emperor, Kaisar’ from the older borrowing cěsaŕъ ‘king, basileus’ (e.g. in John 19) suggests that the Slavs in contact with Byzantium used both terms. Seliš čev, in his posthumous work Staroslav. jazyk, confidently talks about the “city dialect” of the Salonika Slavs. It is quite probable that there was such a dialect, but the texts do not permit us to say that we know anything about it.
- This is the symbol which consists of the letter “¸” modified by a small hook, cf. Diele, Altkirchenslavische Grammatik I, §6, note 4; Leskien Handbuch, p. 6. Fairbanks fails to mention this letter, although it occurs in the text he studied (e.g. gr¸dqi Luke 6:47). His claim to be working “by descriptive techniques alone” is compromised by frequent omissions of facts which must be accounted for. One must deal with all the evidence, not just select bits.
- Here Fairbanks drops the question, with a reference to general works treating all the mss. He departs from descriptive analysis in favor of an untested assumption.
- “i3,” has no numerical function, a fact apparently first observed by Seliš šev, Staroslav. jazyk, I, p. 47. It does occur, rarely, after a consonant; but in some cases it is joined as a ligature to the consonant-letter, e.g. in b[o] ši3i1 (Luke 1:35; see photograph of folio 133v in Sbornik old. rus. jaz., 83) the lower loops of the “ž” serve as the upper loops of “i3”. Jagić does not mention ligatures in his edition, so that their full usage cannot be analyzed without the original. The Acc sg enclitic pronoun ‘him’ is normally spelled “i3”.
- Fairbanks does not mention this fact, but gives the impression that there is a single symbol for y. It is only in some cyrillic texts that the two elements of the digraph are joined together by a line. In some glagolitic texts (including Zogr2), y is spelled by “ъ” plus any of the three i-letters.
- The signal is often emphasized by the use of a diacritical mark placed over a vowel not after a consonant.
- A tiny number of exceptions, e.g. deseNte (Luke 8:1, 9:1), soNdeNšte (Luke 22:30), vъždeNždetъ (John 4:13). The anomalous “q” (see above) could be represented graphically by “+N”.
- Josef Kurz gives the figures (Rocznik Slawistyczny 10.151), which I convert to percentages: “n⁀” written for etymological *nj in 82+% of cases; “1⁀” for *lj 66% “correct”; but “r”⁀” for *rj slightly less than 19%. Kurz did not include in his figures for *lj those cases where the “epenthetic l” after labials has been omitted, sometimes with and more often without the diacritic, e.g. pristav⁀enbe (Luke 5:36), zem⁀i (Matth 25:25), pristǫp⁀ ъ (Matth 25:24). Vondrák gives the complete material, Altkirchenslavische Grammatik2 (Berlin, 1912) 323–4. The meaning of these spellings is surely not significant for the Cyrillo- Methodian dialect; they must, however, be mentioned in any study which purports to be based on descriptive evidence.
- The behavior of root-initial vowels in compound words is like that of word initial in all of Slavic in the historical period. The spellings like měroǫ (Luke 6:38) and ǫtτъǫdu (Luke 11:7) are rare (Kurz, loc. cit., says 8), and doubtless have no meaning for Cyril's dialect. They may, however, be of interest for the scribe's dialect: it must not be forgotten that the shape of a glagolitic “ǫ” is entirely different from “ǫ̈”, and a purely mechanical slip, possible in a cyrillic text, seems unlikely here. Fairbanks does not mention such deviations from his theory, but notes that only “a” and front vowels may follow other vowels. He failed to notice υъorǫžъ s¸ (Luke 11:21), nauči (e.g. Luke 11:1) and zakonoučitel⁀e (5:17) in his material, and even states specifically (fn. 4) that “except in foreign words, the symbol o does not occur after any vowel.” Root-initial ǫ is rare, and not attested in any compounds in Zogr; but cf. blagoǫxanie ‘fragrance’ in Euch.
- By Kurz's count (RS 10.151–2), “ü” in pos. 2 74 X ∼ “u” 5X; after “št”/”žd” 101 ∼ 7. For pos. 3, we give the complete data for cases where we expect the consonant-letters + diacritic; that is, we are employing comparative material to help interpret the confused spelling: rˆǫ̈ 26 ∼ rˆ ǫ 2 ∼ rǫ̈ l ∼ rǫ 13; lˆǫ̈ 211 ∼ 1ǫ 180 ∼ lˆǫ 25 ∼ lǫ 5; nˆǫ̈ 30 ∼ nǫ̈3 ∼ nˆ ǫ 3. Fairbanks noted that “lˆǫ̈” and “lǫ̈” occur in the same forms, and interprets it as a failure of the scribe to write the diacritic. Fairbanks did not note “1ˆǫ” at all, although it occurs in his sample (e.g. priemr ǫtъ Luke 8:13). “nˆǫ” does not happen to be written there.
- These variant spellings have been studied at length, usually with the aim of determining the “degree of softness” of the preceding consonant, as though the scribe was trying to make careful phonetic distinctions. An impressionistic sampling of usage is given by Kul'bakin in Le Vieux slave (Paris 1929) 198–206. Fairbanks does not mention the exceptional cases.
- utro/ütro ‘morning’, uže/'uze ‘already’, ne u/ne ü ‘not yet’; for specific occurrences and discussion see van Wijk, Slavia 6.233–8 and Kurz, Slavia 9.9 f. Etymologically we expect üže, ne ü but utro, and this may well have been the case in Cyril's dialect. Most dialects, however, equalized the three one way or the other.
- Kurz notes only 2 “ě” ∼ some 600 “a”, RS 10.151–2. The exceptions, taken together with the 4 cases of “ě” after “št”/”žd” (∼ 471 “a”), are of interest for the scribe's dialect. Fairbanks did not notice the ištěte (Luke 12:29) in his material.
- See van Wijk's article in Slavia 6.233–8 and Kurz Slavia 9.681 ff. for data and discussion. It seems probable that the older texts had only the type ěviti.
- Here is another example where Fairbanks failed to record the facts from the material he was analyzing. He specifically states that “3ě” does not occur in his data (fn. 9). In Luke 7:46 there is noʒě, and it is opposed to nozě twice each in verses 38 and 44. Fairbanks has found it necessary to interpret the contrast of the graphs lˆǫ̈ and lǫ̈ “in the same forms” (364), but by omitting mention of the numerous other instances where “the same form” is found with different spellings, he gives a false impression of the problems involved.
- The cases are listed by Grunskij, in comparison with the usage of Marianus (Sbornik Otd. rus. jaz. 83.30–2), and again by van Wijk (Juěnosl. filolog, 5.43–5). “ʒ” appears in less than a quarter of expected cases: 42 “ʒ” ∼ 156 “z” for * ʒ.
- On this point Fairbanks has blundered in stating that “the writing cë occurs in stems, the writing ca in inflectional endings.” Any student of OCS has to learn that the suffix -ě in a number of declensional forms and the imperative entails the change of stem k/g/x to c/ʒ/s, and that stems ending in -c-/-ʒ- frequently take inflectional suffixes beginning with -a-. In the pages Fairbanks was dealing with there is no lack of examples; e.g. rçcě (Acc du fem) Luke 13:13, měcě (L sg fem) 13:21, ocě (L sg n) 6:42, sυěštъnicě (L sg m) 11:33, rъcěte (imv) 10:10; also the no noʒě noted above. In the adverb which gives its name to the letter, ʒělo is written 10 times, but it does not happen to occur in Fairbanks' sample. These facts would add two new phonemes to Fairbanks’ inventory; palatalized cj and 3j.
- Cf. e.g. Vondrák, Altkirchenslav. Gram.2 (1912) 297, Fortunatov, Lekcii po fonetike starosl. jaz. (1919) 13 ff., Leskien, Altbulg. Gram. (1919) 3, Łoś, Gramatyka Staroslowiańska (1922) 17–8, Kul'bakin, Vieux slave (1929) 41, Weingart, Rukovět jazyka staroslovënského, I (1937) 115, Vaillant, Manuel du vieux slave, I (1948) 60, Słoñski, Gramalyka języka starosło- wiańskiego (1950) 19–20.
- Hamm, Gramatika starocrkvenoslavenskog jezika, I (1947) 65 ff., firmly states that these consonants are pronounced as in Serbo-Croat, i.e. without palatalization. Diels, Altkirchenslavische Grammatik (1932) 70, sidesteps the question (“die Frage… ist hier nicht zu betrachten”), but van Wijk, Geschichte der altkirchensl. Sprache (1931) 43, calls palatalization unproved. I hasten to state that I am converting the statements of all of these authors into phonemic terms; the usual terminology contrasts the “strong softening” of the hushing sibilants, j, and the palatal ḷ, ń, ŕ to the “weak softening” of the labials and dentals. Certainly one must assume for many dialects of late PSI a pronunciation with some degree of palatalization before front vowels in the latter type; but it is a long way from that admission to the assumption that palatalization is a fundamental, distinctive feature with phonemic value. The allophonic character of the “half-softness” is recognized now by S. D. Nikiforov, Staroslav. jaz. (Moscow 1952) 35.
- Fairbanks states (362) that “the hypothesis which will most simply account for these facts of distribution” is to assume that “the symbols p, b, m, v, t, d, s, z”—and we must add, following his line of thought, c, ʒ (cf. note 24 above)—”represent palatalized consonants before the vowel symbols ě, e, i, ü, ъ, and “non-palatalized” before others. It most certainly is not “simpler” to assume that ten consonant symbols and 14 vowel symbols are ambiguous. Nor is it simpler to end up with 38 phonemes (33 + 5) than with 35 (22 + 13). His statement (364) that “since the symbol ę does not occur after the consonants k, g, x, it may be assumed that ę… indicates that the preceding consonant is palatalized” is a complete non-sequitur. It indicates that “ę” doesn't follow velars, i.e. that the velars are restricted in their occurrence. His next two statements “this conclusion is supported by historical evidence, but can be arrived at on purely descriptive grounds” are both false. There is no historical evidence, and the descriptive grounds do not substantiate the conclusion he desires. He has assumed that there is phonemic palatalization, and by a circular argument which skims lightly over the descriptive facts has proved it to himself. One is reminded of Leskien's judgment of a similar work: “das ist papierne Sprachforschung und ihr Resultat ist ein papiernes” (AslPh. 27.28).
- To my ear there is a yod before /i/ in Mac. only after a consonant (zemji ‘lands’). The phonetic j after a front but before a back vowel (e.g. pijam ‘I drink’) is likewise weak, and, being automatic, non-phonemic. A morpheme whose underlying form ends in /j/ is then subject to change: /j/ before zero and /a, o/, but /Ø/ before /e, i/: e.g. ‘my’ (m) moj, (with def. art.) mojot; (f) moja ∼ (n) moe, (pl) moi. Deciding where to write j occasioned much argument among the authors of Mac. orthography, and the use of the letter causes a great deal of trouble to students.
- The historical group *jb, whatever it was in PSl, then became j: * mojъ > moj, *υojъna > vojna. From the evidence of the texts, perhaps it is preferable to refer to an OCS “weak i,” in Trubetzkoy's terminology, since both *mojъ and *moji are spelled moi, and only comparative data justifies separating them.
- It must not be forgotten that once the writing system was established, the scribes were saddled with it—if there was no letter for j, no means to distinguish ě from ja, and a superflous letter or two (i2, i3, sometimes 3), the scribe simply had to be more careful in his copying and to spend more time learning spelling rules before he could compose on his own. The invention of the cyrillic letter “ja” marks a real break in the tradition, and it took the scribes some time to learn how to deal with it, as is evidenced by the orthographies of OCS, “Middle Bulgarian” and early Russian mss.
- For instance, in Suprasliensis the N du noun eleni ‘2 deer’ is distinct from the L sg adj elenˆi ‘deer's’, and there would be contrasts -nъ, -ne ∼ -nˆъ, -nˆe if the appropriate forms happened to be attested.
- N. Trubetzkoy sets up an “indefinite nasal consonant” /N/, and points to the spelling “aNgˆl-” ‘angel’ in Psalt. Sin. as evidence both of the independence of the letter “N” and of a fourth nasal vowel a. [Trubetzkoy lists the phonemes of OCS and discusses their qualities in his mimeographed lectures, Altkirchenslawische Sprache, Vienna 1934. A fuller discussion, including this particular point, is found in his Altkirchenslavische Grammatik, which is being published in the 228th volume of the Sitzungsber. of the Vienna Academy.] This seems to base too much on isolated spellings of a foreign word in a ms. which is in general carelessly written. However one wonders about the relation of the native nasal vowels to such foreign words as Konstantin and Pentēkostē (Euch, p¸tikost- ∼ Sav. penti- kost-; cf. Diels §6 note 12). Fairbanks identifies the nasal element with /n/. He then assigns “¸” to /a/ “due to historical considerations” (364). It would be interesting to know what they are, for the Macedonian and Bulgarian dialects which are likely to furnish some pertinent information all have identified ę (insofar as it did not become identified with ǫ̈ and then ǫ) with a front mid vowel, generally e, but occasionally, apparently, ë (cf. Mladenov, Geschichte d. bulg. Spr. 119). One suspects that his chief ground is that he has argued himself out of a phoneme e and has no other place to assign the presumedly allophonic ¸.
- Although “u” in Zogr has a second element which is not entirely identical with “ižica”, the combined evidence of the glagolitic texts shows that originally “u” was indeed made up of two units, “o” + “ižica”, exactly parallel to Greek omicron-upsilon and the cyrillic digraph.
- See Durnovo, “Mysli i predpoloženija o proisxoždenii staroslavjanskogo jazyka i slavjanskix alfavitov,” Ryzantinoslavica I (1929) 48–85, and Trubetzkoy, “Die altkirchen-slavische Vertretung der urslav. *lj, *dj,” ZslPh 13 (1936) 88–97. Trubetzkoy and his followers use special symbols for the supposed phonemes in reconstructing OCS texts (“urkirchenslavisch”), e.g. Isačenko in Jazyk a pôvod frizinských pamialok (Bratislava 1943) and Horálek in “Studie o slovanském verši,” in Sborník filologický, XII (1940–6).
- The norm is based on the usage of Zogr itself, controlled partly by the spelling of other OCS texts but principally by the usage of early Russian mss. like the Arxangel'skoe evangelic of 1092 (first scribe) and the Uspenskij Sbornik of the twelfth century.
- V. Jagić, “Studien über das altslovenisch-glagolitische Zographos-evangelium,” AslPh 1.1–55,2.201–269; A. Leskien, “Die Vocales ъ, ь in den Codices Zographensis und Marianus,” AslPh 27.321–349; N. van Wijk, “Zu den altbulgarischen Halbvokalen,” AslPh 37.330–377. See also Š čepkin, Rassuždenija o jazyke Sav. Kn. (= Sbornik 67 no. 9). Further comment and complete literature in J. Kurz, “K Zografskému evangeliu,” Slavia 9.465–489, 673–696.
- The figures are based on my own examination of eight sections of 16 folios each, comprising half of the ms. Samples were taken from all Gospels, since scholars (particularly van Wijk) have made much of the “archaic” quality of the later portions of the text as compared to the beginning. Since I cannot see that my results (for weak jers) show any significant difference between samples, I do not cite the figures in detail, but give approximate averages, noting particularly large ranges.
- The total number of strong jers is relatively small. In my sample of half the ms. only 259 ъ and 279 ь appeared, whereas nearly as many medial weak jers are to be found in any section of about 18 folios. The relatively small total means that individual cases weigh more heavily in the percentages. Two of my sample sections had no “errors” in the strong jers, and two more had almost none: these were in the last half of the ms.
- Only one instance of the loss of “ъ” was found; it is cited above, “ь” however is omitted 21× (= 7+%); “ъ” > “ь” 4× (out of a possible 259) but “ь” > “ъ” only once (out of 279); “ъ” > “o” 12× (= 4+%) and “ь” > “e” 28× (10%). Variation in samples is large. In the first, no ъ > o but 35+% ъ > e; in the second once (= 2+%) ъ > o and 24+% ъ > e; in the third ъ > o 2 (= 10%) and ъ > e only 4%. The exact percentages cannot be considered of great significance, since too many variables enter into the interpretation of given spellings as containing “strong jers”, but they do serve as a kind of measurement. The most confused usage was found in Mark, the most consistent in Luke and John. Vondrák lists many (not all) examples of “o” for “ъ” and “e” for “ь”, Altk. Gr.2 193.
- Kul'bakin objects to this (RESl 2.199) as being unlikely in so early a text, but it seems to me that he is being over dogmatic. Because the phonetics of Zogr seem old it is assumed that the ms. is equally old, and numerous conclusions are drawn from this. But no one can prove that Zogr is not a later ms. copied by a conservative scribe. It must never be forgotten that in OCS there are no specifically dated or localized mss., and that the theories are built up on a structure of assumptions. The evidence is frequently ambiguous and often outright contradictory, and the scholar cannot afford to lose sight of the fact, however passionately he may believe in the plausibility of his assumptions and conclusions.
- See Kul'bakin, Vieux slave 44 ff for the evidence, and Kurz, RS 10.142 ff for a few additional details and explanations.
- Kurz cites the textual evidence and the literature (Slavia 9.483 ff). The most likely explanation was advanced by Roman Jakobson, who posits for various PSl dialects a nasal schwa, Remarques sur l'évolution phonologique du russe comparée à celle des autres langues slaves (= TCLP 2, Prague 1929) 26. This would mean a back unrounded nasal vowel which brings symmetry into the system: .
- Cf. Jakobson, Remarques, 61.
- On the basis of paleography, Jagić places Zogr2 near the end of the twelfth century (Glagoličeskoe pis'mo, 134), that is, a hundred and fifty to two hundred years after the writing of the older section.
- It even occurs once in a root where it is found in various Mac., Bg., and eastern Sb. dialects: ʒizdete ‘you build’ Matth 23:29. This is one of a number of places where a secondary ʒ, not attested in OCS, appears in the medieval texts and modern dialects.
- “ǫ” is written etymologically 47×, but stands for *jg 28×. “ǫ̈” is “correct” 115×, but represents *ǫ 125×.
- eǫže Matth 18.19, sǫ 18:34; υъ Vifaniję 21:17.
- Jagić makes this statement twice, in the “Prolegomena” to the edition, p. XVIII, and in Glag. pi'mo, 134. We must assume that the cases where “ь” appears, without comment in the notes, are misprints in the edition. At least one, υъsě (Matth 19:27), proves to be “ъ” in the picture published by Geitler, Die albanesischen und slavischen Schriften, Vienna 1883. It is this sort of question which makes the publication of the whole ms. in photographic reproduction desirable.
- “ъ” for *ь 81+% “correct”, for *ъ 72+%. In one of the samples of Zogr1, “ъ” omitted 6+%, “b” 21+%, in another “ъ” omitted 5+%, “ь” 35+%.
- For the first part of Matthew, “e” for strong ъ is about 30%; “o” for ъ never represents a large percentage. It must be noted that the percentage for ъ in Zogr2 is misleading: I found only 6 cases, four of which are in the root smokov- ‘fig’, one is the “article” in rabotъ ‘this slave’, and one a “tense jer”, věčъnoi (Acc sg m). Only the last seems to have specific dialect coloring, and its isolation weakens its value as evidence.
- υьnide, zьlě and sъzъda are examples of “mutated jers” which are in accord with Jagić's rule. The expected υъnide and zъlě are found elsewhere in Zogr, but sъzьda does not happen to occur, sъnъmište is one of the exceptions to the rule, with ъ > ъ before a front vowel; but this root is nearly always spelled with “ъ” in Zogr.
- Fairbanks, who transcribed verses 2–5 of this text, took zъlě and sъnъmište at face value, but somehow decided to omit the controversial jer in sъzъda. It would be interesting to see what he would do with some word like koždo, which violates his rules. He fails to point out the omission of phonemes in the abbreviated words.
WORD
Volume 8, 1952 - Issue 4: Slavic Word No. 1