Loanword adaptation in Hungarian: unexpected vowel harmony in material of Latin and Slavic provenance

Abstract

In a 1992 article, Eugen Helimski showed that the accented vowel of Old and Middle High German, Medieval Latin and Pannonian Slavic lexemes governed whether they acquired front-vowel or back-vowel harmony when entering Hungarian as loanwords. While the German material appears exceptionless, some words of Slavic and Latin provenance exhibit unexpected back-vowel harmony. The present article submits that if a labial sound follows the originally accented vowel, front-vowel harmony is blocked. This conditioned rule applies without exception to both Slavic and Latin loanwords; it is thus an economical solution. It follows that variation in Slavic loanwords in Hungarian cannot serve as a testimony of Old Slavic accent shifts, but merely of the place of the original (pitch) accent; and that the Slavic language that provided loanwords to Medieval Hungarian must have been fairly uniform. As for Latin loanwords, it likewise renders an appeal to late accent shifts unnecessary. Helimski also discovered that a subset of Latin words containing a medial cluster *-CiV- could trigger front-vowel harmony even if the original accent fell on a back vowel. Here, I argue that the distribution of front and back vocalism in this type is further governed by the vowel of the initial syllable. This minor rule possibly applies to Slavic as well.

Keywords:

• Vowel harmony
• loanword nativization
• adaptation strategies
• accent
• labials
• Hungarian
• Pannonian Slavic
• High German
• personal names

Notes

¹ These are sometimes called “neutral” vowels but it makes sense to classify them as a kind of front vowels not only because of their phonetic articulation, but also because words which contain roots with these vowels usually select suffixes with front vowels (szín “color”, dat. szin-nek and not †szín-nak; szél “wind”, dat. szél-nek and not †szél-nak; there is however, a large minority of exceptions). The vowel ë /e/ is no longer distinguished from e /ε/ in the standard language but occurs in dialects.

² According to Rédei (1988, 47, fn. 1) there is a general tendency in languages with vowel harmony to preserve velar (back) variants rather than palatal (front) ones.

³ Mollay (1992, 112) notes that some of the earliest borrowings from around 1000 AD specifically entered from Old Bavarian (which had no contrast between voiced and unvoiced stops); this explains the variation between b and p. See also püspök below.

⁴ This word was possibly borrowed via OHG kamara, kamera (Mollay 1982, 337–339).

⁵ As Helimski (1992, 50) notes, other apparent exceptions are in fact not direct borrowings from Latin, but have been taken over from Old High German, as in the case of monostor “monastery” where the immediate source was OHG munusturi, munsturi and Latin monastḗrium only the ultimate source. This is most likely also the case of several proper names like Márton (< HG Martin rather than Lat. Martī́nus and Ágoston (< HG Augustin, not Lat. Augustī́nus).

⁶ On the reconstruction of this language and its manifestations in Hungarian, see, e.g., Richards (2003), Greenberg (2004), Laakso (2006) and Nuorluoto (2010).

⁷ Apart from backness harmony, Hungarian also possesses a limited roundness harmony whereby ö is normally replaced by e if immediately followed by a syllable containing an unrounded front vowel, except in compounds (Vago, Kenesei, and Fenyvesi 1998, 423–424). For csahol and szalonna, I therefore provide the hypothetical front-vowel alternatives †csehel and †szelenne rather †csehöl and †szelönne which would appear phonotactically illicit.

⁸ The same Slavic word entered Hungarian later via Romanian slanină, at a stage when the adaptation rules had ceased to operate, and became szlanina (Bakos 1977, 132).

⁹ It is conceivable that ketrëc and verse, vörse as well as further exceptions like gërënda “beam; garden bed” (1321 Guerenda for expected †garanda) < Sl. grędà, cf. Ru. grjadá (Kniezsa 1974, 191–192; Benkő et al. 1967–1976,1050–1051) are the result of other conditioned rules. Alternatively in some cases, they may be influenced by related derivatives with similar meaning and front vocalism in the source language, e.g., gërëndëly (1363 gerendel-) < *grędíl’ь, cf. Ru. grjadíl’, gredíl’ (Kniezsa 1974, 192–193; Benkő et al. 1967–1976 III, 1051). Note that in this case the eight-vowel dialects show the presence of ë which, contrary to e proper, is a “tolerant” vowel.

¹⁰ This example was suggested to me by Seán Vrieland.

¹¹ Latin -um is rendered as -om in bazsalikom, but as -em (dial. also -ëm) and not -öm in petreselyem because of roundness harmony; cf. fn. 7. The earliest attestations from the 12th c. onwards exhibit back vocalism in all syllables but the first, but at least until the 19th c. forms keep occurring with at least one back vowel combined with tolerant front vowels. Even for Hungarian, this word shows an exorbitant degree of variation through the ages as well as among the dialects.