On the Slavic Word for ‘Morning’: *(j)u(s)tro

ABSTRACT

The history of the Slavic word for ‘morning’ constitutes an old problem of Slavic etymology. Forms such as OCS utro, Cr jȕtro, R утро, Cz jitro ‘morning’ as well as the adverb OCS zaustra ‘in the morning’, Bg зáстра etc. certainly continue the root PIE *h₂eu̯s- ‘(morgens) hell werden’. Consequently, an explanation of the attested forms needs to account for the following three issues: (1) the loss of the root final sibilant in most of the Slavic forms; (2) the acute tone on the root syllable; (3) the presence of a word-initial glide j in some Slavic languages. The present article deals with these problems. By starting from a holokinetic r-stem both the loss of the root final sibilant as well as the acute tone on the root syllable are explained as resulting from contamination of different stem or root variants respectively. The occurrence of a word initial glide, on the other hand, is explained by a transfer within syntagmatic expressions. One of the advantages of the proposed scenario is that all the posited changes may be paralleled by developments elsewhere in Slavic or Indo-European.

KEYWORDS:

• Slavic
• Proto-Slavic
• Proto-Indo-European
• Etymology
• Historical phonology
• Historical morphology
• Historical accentology

Acknowledgements

This study was presented at the 4th Indo-European Research Colloquium at Zurich University (April 5–6, 2018). I would like to thank Barbara Sonnenhauser, Jussi Nuorluoto, Georg Holzer, Ranko Matasović, Thomas Olander, Florian Sommer, Ivan Šimko and two anonymous reviewers for their comments on earlier drafts of this paper. All remaining errors are my sole responsibility.

ORCID

Florian Wandl http://orcid.org/0000-0003-4992-176X

Notes

1 Vaillant (1935) rejects the Slavic forms containing a sibilant as evidence. Due to limited space, I cannot discuss this question here. The proposed scenario may be upheld even if Vaillant’s claim should prove correct.

2 The r-stem is often interpreted as a secondary formation based on a locative in *-er of the s-stem PIE *h₂eu̯s-os-/*h₂us-s- (thus already Bartholomae 1889, 15 for the Vedic forms; on Proto-Indo-European cf. especially Nussbaum 1986, 236 ff.; differently Ozoliņš 2015, 63 ff.). The data is not entirely conclusive regarding the inflection of the r-stem (cf. Peters 1980, 34). An original holokinetic paradigm is suggested by Nussbaum (1986, 238 fn. 36) in parallel to the delocatival model case *(d^h)g ̑^h(e)m-on- ‘human being’. The above-mentioned examples from Greek would then have adopted hysterokinetic inflection analogically, possibly on the example of αἰθήρ ‘clear sky, heaven’. If the scenario presented in the present paper is correct, Slavic and Baltic data support this hypothesis (see below). An independent origin of the r-stem in Vedic and Greek is suggested by Hajnal (1992, 60). Since its emergence would not be bound to any specific features of these languages, it could also be assumed for Slavic and Baltic. In any case, I see no reason to reconstruct a non-prototypical hysterokinetic paradigm *h₂eu̯s-er-/*h₂us-r-, as it is done by Pronk (2018, 305).

3 See Nussbaum (1976, 49): “In short, nothing certain here”, and cf. the discussion in Lundquist (2014, 97 f. with references).

4 According to Pronk (2018, 302 ff.) the original meaning of the PIE r-stem was ‘daybreak, (early) morning’ while the s-stem referred to the deity of Dawn. This reconstruction, however, at least partly depends on Lundquist’s (2014) interpretation of the Vsg Ved uṣar as a poetic nonce-creation which is by no means secure.

5 Differently, Derksen (2008, 510).

6 For more details on these two sound laws cf., for example, Shevelov (1964, 127 ff., 200 f.).

7 Pronk (2018, 300) assumes that the ruki-rule affected any sibilant, i.e. PIE *s and *z, irrespective of the following segment. This not only forces him to reckon with a voiced ruki-variant which has left no traces in Slavic, but also with a reversion of the ruki-rule before dental and velar occlusives since here again we find no traces of this change (cf. Vaillant 1950, 29).

8 Pronk’s (2018, 299 f.) assumption according to which word-internal *sr as well as *ṣr resulting from the ruki-rule regularly gave *zr and *ẓr in Slavic is unconvincing. OCS nozdri pl ‘nostrils’ cannot go back to a preform *nh₂es-r-, because the vowel i in the cognate Ukr нíздря presupposes a jer in the following syllable. Most probably it constitutes a compound form consisting of *nosъ ‘nose’ and a derivative of the verbal root *der- ‘tear, flay’, cf. OCS dьrati (see Vasmer 1953, 225 with references). A parallel explanation was put forward to account for Pronk’s second witness CS męzdra ‘inner side (of a hide)’ (cf. Vasmer 1955, 187 f. with references). Therefore, the only clear instance of a word-internal sequence PIE *sr in Slavic is found in OCS sestra (OSC bystrъ and ostrovъ ‘island’ might be of post-PIE age) where we find the same reflex as in word-initial position, which apparently contradicts Pronk’s assumption. Furthermore, in Pronk’s scenario the domain of the t-epenthesis must be expanded to sequences of *ṣ, *ẓ plus *r, which may be conceived as a further complication.

9 Note that the Vedic variant of the ruki-rule likewise did not apply before r, cf. Gsg Ved usraḥ (cf. Whitney 1879, 62).

10 Cf., for example, the recent etymological dictionaries by Derksen (2008, 510 f.) and Matasović et al. (2016, 411 f.). I will not discuss explanatory models proposing two independent PIE etyma merging into the Slavic word for ‘morning’ since I consider a solution which explains the attested forms from only one ancestor preferable. The reasons for this are mainly theoretical: Any reconstruction carries the danger of being incorrect. Consequently, the probability of a certain etymology to be incorrect increases with the number of additional etymologies it presupposes.

11 Nieminen (1956, 23) adds Ru ýтрось to this list.

12 Note that in this case the domain of epenthesis was the phonological word.

13 The only case of simplification in such clusters that I am aware of occurs in Macedonian dialects. In this case, however, the dental is dropped, cf. зрав for здрав, сесра for сестра (see Koneski 1965, 83). Regarding USo wótšy, USo wótry etc., cf. Schaarschmidt (1998, 109 f.).

14 A partly parallel scenario is proposed by Pronk (2018, 300 f.). However, the assumption of a word-medial change of *sr, *ṣr to *zr, *ẓr (see above) forces him to posit an additional contamination, or in his words, “restoration”, to account for the sequence *tr in the Slavic word.

15 Loss of *x which had arisen secondarily before *t in the Slavic word for ‘morning’ was first proposed by Berneker (1899, 157 f. fn. 1). His overall scenario which he himself refers to as “etwas gekünstelt” is, however, unconvincing since it rests upon the operation of the ruki-rule before *r, and the occurrence of contaminations between different lexemes.

16 This may have happened, for instance, when the apophonic differences in the suffix were eliminated.

17 Regular sound change is marked by “>”, morphological change by “→” and the postulated contamination by “X”; “stst” designates the strong stem and “wst” the weak stem.

18 Indirect evidence may come from the scenario proposed in this article. To avoid circularity I shall, however, refrain from using this as an argument here.

19 Pronk (2018, 299) further mentions ū́šras.

20 Nieminen (1956, 21). Wodtko et al. (2008, 358, 365) posit “Neoablaut” for the root, without, however, giving any arguments. Zinkevičius (1966, 93) locates a change of unaccented or circumflexed au to uu̯, u^u̯ or u in the Ukmergė region. However, this is not where Morkūnas et al. (1977, 183) attest zero-grade forms of the word for ‘dawn’.

21 With reference to Nussbaum (1986, 236).

22 Note that later OCS utrěi is likewise derived from a locative (cf. Vaillant 1942, 10).

23 According to Morkūnas et al. (1977, 183) corresponding forms are also attested in Dumpiai (Klaĩpėdos) and Miežionys (Šalčininkai).

24 Regarding the quality of the suffix vowel Nieminen (1956, 21) refers to Li vãsara, vasãris, pavãsaris as opposed to more archaic Žem vase˜ris, pavãseris. Endzelin (1922, 48 f., thus also Pronk 2018, 304 f.) reckons with a presumably Common Baltic replacement of suffixal *e by *a before an immediately following sequence of sonorant plus *a or *ā. Since the quality of the suffix-vowel was *o in holokinetic strong stems, no such change needs to be posited in the scenario presented here.

25 Without arguing the point Nieminen (1956, 21 f.) states that the relevant forms could also have arisen secondarily. A possible source for the vocalism might be vãsara ‘summer’ or maybe the masculine vãkaras ‘evening’. Ozoliņš’ (2015, 48 f.) explanation according to which the suffix -arā was replaced by -rā because it was “reimagined by some speakers as being identical” with the latter is unconvincing, as long as no parallel examples are adduced.

26 Contamination is also assumed, for example, by Fraenkel (1936, 169) and Petit (2004, 70) to account for the Slavic word. Cf. the parallel development in La iecur, iocinoris and iter, itineris (see Brugmann 1889, 326, 331).

27 The contamination must have taken place before the loss of word final consonants since otherwise *l would have been lost in the NAsg.

28 Cf. for a parallel generalization of the zero-grade root variant *su̯ep-ōr/*sup-n-és → *sup-ōr/*sup-n-és (see Schindler 1966, 74 f.). Alternative reconstructions are given in Wodtko et al. (2008, 676 ff. with references).

29 Cf. Wachter (1997, 5 with references).

30 Cf. for the loss of *u̯ Ru копоть as opposed to Li kvãpas (see Shevelov 1964, 198).

31 Cf. for the change *sk ̑ > *sk OCS iskati.

32 Note, however, La mūscerda and cf. OCS govędo, govęžii, čelędь etc.

33 Based on OCS skarědъ and Ru скаред Vaillant (1958, 180) posits a compound form skar-ědъ ‘mangeur d’ordures’. However, forms containing ě may easily be explained as based on folk etymology, whereas the opposite assumption that the nasal vowel resulted from adaption to the rare suffix OCS -ędь would be unmotivated.

34 The same explanation may be applied to account for the Isg OCS otročęmь MedBl. Alternatively, it could also represent an archaism, cf. (pre-)PSl *atračintmi > otročęmь (cf. Mareš 1993, 129). If the latter is true, this might be the only case of such an archaic ending in an nt-stem. In the light of the standard as well as the Old Czech forms (cf. Gebauer 1918, 414 ff.) the same interpretation seems unlikely for dial. Cz raměm.

35 Differently, Pronk (2018, 300) (see above).

36 Treatment of Sl *x as *k may also be observed in the cluster sx, cf. OCS pasxa, DLsg pascě (see Diels 1963, 147 Anm. 4). However, since this sequence occurs only in loanwords, the corresponding examples are not entirely conclusive.

37 The standard language has not preserved the original inflection, cf. vȑći vȑgnēm.

38 Note that the resulting consonant cluster pst must have been rather rare if not exceptional in the language (ps existed after the loss of weak jers; cf., for example, Cr pȁs, G psȁ).

39 There seem to be no reliable cases of a sequence *kt before consonants (cf. Shevelov 1964, 191).

40 Cf. especially Wachter (1997, 14 ff.).

41 Cf. for the preservation of *k before *n Shevelov (1964, 195). Trubačёv, et al. (1981, 216) recognize a further form originally containing *n in dial. U jýкло ‘брюхо’.

42 Cf. the parallel collective noun *(h₁)éu̯Hd^h-ōr, G *(h₁)uHd^h-n-es proposed by Schindler (1975, 7 f.).

43 According to Kortlandt’s (1977, 38 f.) original assumption word-initial *u yielded Latv û- under the stress. This is, however, contradicted by the fact that accented acute syllables regularly developed sustained tone in Latvian (see Olander 2009, 117 ff. with references).

44 Pronk (2011, 318 f.) was obviously not aware of the zero-grade forms of the word for ‘morning’. Apart from apý-ūšris they are, however, mentioned in Pronk (2018, 299). A further candidate might be Li áiža ‘crack’, Latv aĩza, OPr eyswo ‘wound’, Sl *ězva next to the verb Li ìžti, y˜žta, ìžo ‘pod, burst, crack (ice)’ (see Pronk 2011, 314).

45 Note that Pronk (2011, 314) explains the acute tone of Li íeškoti via a zero-grade form that is reflected in Sl *īskàti which contradicts his own hypothesis.

46 Pronk (2018, 299) unlike in his earlier publication speaks of “regular metathesis” in word initial-position as well, not, however, refraining from his accent-based rule and his separating of the word-initial and word-medial laryngeal metathesis. Word-initial metathesis in the Slavic word for ‘morning’ is further suggested by Matasović et al. (2016).

47 Pronk (2011, 319) adds dial. Lith ánka ‘loop, ring’, Latv. añka ‘rope used to fasten a sail to the mast’ to the list. Cf., however, Derksen (2015, 56) on this etymon.

48 Derksen (2015, 204) seems to doubt the reliability of the accentuation.

49 Note further that acute tone sometimes occurs in Latv substantives as well, i.e. ie˜mȩsts ‘Einwurf’ (cf. Endzelin 1922, 503 f.; Derksen 1996, 260; Petit 2011, 245 ff.).

50 Cf. Derksen (2008, 407) on Sl *po-, *pa-, Le Feuvre (2011), Piperski (2013); for alternative accounts see Petit (2011, 259 ff.).

51 The given example may constitute a recent formation. However, in Proto-Indo-European prefixes/preverbs most certainly occurred in contexts where they were immediately followed by sequences of laryngeal plus consonant.

52 Cf. Olander (2009, 149 f. with references) for a brief definition of Hirt’s law.

53 Note the placement of the reflexive particle si between preverb and verb in Lithuanian, cf. suriñkti ‘collect’ → susiriñkti ‘gather’.

54 Young (2006, 239) provides no zero-grade form while Derksen (2015, 200 f.) cites only n̦̦im^t.

55 Kortlandt’s (1975, 3) assumption according to which word-internal tautosyllabic sequences *Hi- and *Hu- must still have been present in post-PIE times since we don’t find the reflex of Hirt’s law in the l-participle Ru пилá is contradicted by cases like Cr žȉva ‘mercury’ (< PIE *g^u̯h₃i-u̯ó-). The corresponding adjective ži ̑v may have adopted its inflection analogically (see Holzer 2009, 172 f.) which may consequently be correct for Ru жилá and пилá as well.

56 Should this be correct the acute tone in the verb dial. Ru ýдить (Dal’) ‘ripen, fill (with)’ would have to be explained secondarily.

57 I am unaware of any special study on PIE laryngeal metathesis in word-initial position.

58 According to Mayrhofer (1992, 239) the Ved form belongs to the r-stem. We may, therefore, propose a parallel case of thematization.

59 The same may be true for Li ūšti, ūšta, ūšo (cf. Ved uccháti). However, no data on the accentuation seems to be available (cf. Nieminen 1956, 16-18 and Derksen 2015, 484, who considers a denominative formation as well). Smoczyński (2016, 71) posits a nasal-infixed preform *unšta < *unš-sta.

60 The mobile accentuation in ūšrà may be explained secondarily, cf. Li jėgà 4 as opposed to Latv jȩ̃ga (see Illič-Svityč 1963, 68), Li dainà 4 ‘song’ as opposed to Latv daĩņa (see Derksen 2015, 112), Li dirvà 2/4 ‘grain field, arable land’ as opposed to Ved dū́rva ‘a type of millet’ (see Illič-Svityč 1963, 70).

61 Differently, Young (2006) and Pronk (2011) assume generalization of glottalization/acute across the paradigm. To me this seems less plausible since one would expect generalization of the entire root variant, not only its prosodic feature. However, analogy with nominal formations showing ablaut but no prosodic differences within the paradigm would be possible as well.

62 Cases like Cr slȕšati, slȉšati which can be traced back to desideratives in *-Hs- show that the loss of laryngeals in the Coda must have occurred before the ruki-rule (cf. Matasović 2008, 93 whose example Cr mȉš is however inconclusive, cf. Mayrhofer 1986, 171 with references).

63 Note that in some Slavic languages word-initial *j was lost before *u, cf. Ru ухо ‘fish-soup’ but Cr jȕha ‘soup, broth’.

64 Cf. also Li jau auštrà ‘już świt’ (see Smoczyński 2016, 71).

65 An example of a case where a reoccurring speech error was conventionalized is given by Buckley (2011, 19 with references). Note further the phenomenon of “syntagmatic assimilation” of morphology described by Andersen (1980, 16 ff.).

66 Differently, Matasović et al. (2016). Note that absolute chronology does not contradict this explanation since jošt(e) occurs already in the 13th century while jùčera appears only since the 15th (see Matasović et al. 2016 with references).

67 The assumption that the preposition za could govern the genitive case as well (see van Wijk 1928) is in my opinion not supported by any substantial evidence (OCS zaprъva ‘initially’ can be explained analogically to isprъva ‘from the beginning’).

68 Note the parallel existence of adjectives in OCS -ii (< *ьjь) and -ьnъ, cf. skotii next to skotьnъ.