CC BY
25
R. Matasovic
The devoicing of final stops before *s and the root nouns in Proto-Indo-European
Резюме. В статье речь идет об альтернациях в трактовке звонких (аспирированних) и глухих смычных в исходе корня в праиндо-европейском. Многие случаи могут быть объяснены, если предположить, что конечный глухой является результатом аналогического распространения на всю парадигму аллофона, возникающего в номинативе единственного числа корневого существительного (перед флексией *s). Эта гипотеза позволяет предложить некоторые новые этимологии (например, лат. focus может быть возведено к и.-е. *dhegwh- 'гореть'), или же лучше понять неправильности в этимологических сопоставлениях (например, между праславянским глаголом *lupiti и существительным
Ключевые слова: Proto-Indo-European, etymology, devoicing, root nouns
There are several apparently irregular correspondences of root-final voiceless stops in one IE language with voiced (or voiced aspirated) stops in others. In Pokorny's dictionary (IEW) they are usually treated as variants of the same root with different «extensions» (Wurzelerweiterungen), the status of which has never been explained. They are clearly not suffixes with determinable functions, and the distribution of voiced and voiceless «extensions» in different languages appears random. However, many of these irregularities disappear if the variants with the voiceless stops are due to devoicing before final *-s in the nominative singular of root nouns, and the subsequent generalization of the voiceless stop in the whole paradigm. In other cases, the final voiced stop may have been generalized from the 3rd person singular of the sigmatic aorist, where the devoicing before word-final *-st would also have been regular. In this paper, we shall see how this very simple process can help us better understand a number of otherwise rather difficult etymologies, and to propose some new etymological equations.
Our first example is Latin focus, which may be derived from the PIE root *dhegwh- 'burn'. This Latin word has not been treated convincingly in etymological dictionaries. The connection with Arm. bocc 'fire' (Walde-Hoffmann's 1930-1950, I: 521, IEW 162) is untenable, since Arm. -cc- points to the PIE cluster *-sk'-. The Armenian word might or might not be related to Arm. bosor 'red', but this connection would not help us understand the origin of Lat. focus. A PIE root of the shape DheT- would contradict the phonotactic rules of PIE root-formation. The connection of focus with Lat. fax 'torch', Lith. zvaké 'candle' is impossible because of the diverging vocalism. Hamp's (1992) attempt to explain focus as a back-formation from foculus 'small stove' is extremely difficult. Hamp assumes that foculus is derived from *dhgwe-tlo-, from the root *dhegwh-'to burn' (Lat. foveo 'to warm'), and that Proto-Italic *xwetlo-developed in Latin first to *xwotlo-, and only subsequently to *fotlo- > Lat. foculu-. However, as Michiel de Vaan (2008: 228) points out, the change of *we- to *wo- is datable to the fifth century BC, and the development of *xw to *f is almost certainly earlier (it is probably common Italic). After briefly reviewing the work of his predecessors, de Vaan follows Ernout and Meillet (1979: 243) in leaving the question of the etymology of focus unresolved.
I would like to propose a new solution to this old problem. I believe that Lat. focus developed from the PIE root-noun *dhogwhs 'flame, fire' from the root *dhegwh- 'to burn' (Ved. dâhati, Lith. degù, degti, OCS zegq, LIV 115f.). The root is attested in Latin in the derivative febris 'fever' (< *dhegwhri-) and in the causative foveo 'to warm' (< *dhogwh-eye-)^ The root-final *-gwh- would have been devoiced and delabialized before the nominative *-s, and the result was the cluster *-ks-, as in Lat. vox Voice' < *wokw-s. Another reflex of the same PIE root
1 The etymological connection of focus and foveo is alluded to in Servius' commentary on the Aeneid, 12.118: quidquid ignem fovet, focus vocatur. This also shows that focus is not a sacrificial altar in Latin (ara), but rather a common fireplace, so that the semantic evolution from the noun meaning „fire" is even more probable. On the semantic opposition between ara and focus see Nagy 1974: 88.
noun is probably preserved in PCelt. *degwi- 'flame, blaze' (OIr. daig, cf. Matasovic 2009: 93), with the transfer to the i-stems. If PCelt. *degwi- generalized the stem of the oblique cases, we may reconstruct to PIE a static paradigm with the Nom. sg. *dh6gwh-s, Acc. sg. *dhogwh-m, Gen. sg. *dhegwh-s. This would have been parallel to, e.g., the inflexion of the word for 'night', PIE Nom. sg. *n6kwt-s (> Ved. not), Acc. sg. *nokwt-m (> Lat. noct-em), Gen. sg. *nekwt-s (> Hitt. nekuz), cf. Schindler 1972. Of course, it is quite possible that in PCelt. *degwi- the e-grade was introduced analogically to the verb (PIE *dhegwh-o-> Skt. ddhati).
The evolution of meaning from 'flame, fire' to 'hearth' might appear problematic, but these meanings are quite often related. It is true that the attested developments are usually in the reverse direction, from 'hearth' to 'fire', as in the history of Lat. focus in the Romance languages, where the original Lat. ignis was generally replaced by the reflexes of focus in the meaning 'fire' (e.g. It. fuoco, Sp. fuego, Fr. feu). Similarly, Croatian vatra (of unknown origin, but certainly connected with Alb. vater, voter 'fireplace, hearth') changed the meaning from 'hearth' to 'fire' in the contemporary language, replacing the inherited reflex of Proto-Slavic *ognb (Russ. ogon', etc.). However, the extension of meaning from 'fire' to 'fireplace, hearth' is attested in Celtic; the older meaning of PCelt. *tefnet- is found in W tan 'fire', while OIr. teine (Gen. sg. teined) has both meanings, 'fire' and 'fireplace, hearth' (DIL 583).
Our second example involves Lat. hortus 'garden', which has the voiceless -t- rather than the expected *b < *dh, because it developed by thematicization of a root noun, PIE *gh6rdh-s, Acc. *ghordh-m, Gen. *ghrdh-os 'enclosure',2 cf. OCS grad'b 'town', Lith. gardas 'pen', Skt. grhd- 'house' (with the zerograde of the root from the oblique cases, cf. Matasovic 2004: 99). Here too we have to start from a PIE root-noun in which the voiced or aspirated root-final consonant became voiceless
2 For the ablaut patterns of PIE root nouns see especially Schindler1972.
before the word-final *-s.3 The voiceless root-final consonant was subsequently generalized throughout the paradigm. Of course, this etymology has long been recognized (IEW 444, LIV 176), but the true cause of the root-final alternation of *-dh and *-t has never been clearly recognized.
Our third example involves PCelt. *druko- 'bad' (OIr. droch-, MW drwg-, etc., cf. Matasovic 2009: 105). This PCelt. etymon represents a thematicization of the stem *druk-, which is the regular outcome of the PIE root-noun *dhr6wgh-s 'lie, deceit' (cf. Skt. druhyate 'deceives', druh- 'deceit', OHG gi-trog 'deceit', IEW 276, LIV 137) before word-final *-s (with *ghs > *ks). The zero-grade of the root was generalized, apparently, from the oblique cases (Gen. sg. *dhrugh-os, etc.).
The same process will account for the next set of cognates, including OIr. droch 'wheel' (< PCelt. *droko-, Matasovic 2009: 105, LIV 135, IEW 273), Gr. rpoxo? 'id.', Goth. -dragan 'carry', Arm. durgn 'potter's wheel'. The voiceless velar in Celtic may have been generalized from the Nom. sg. of the root-noun (PIE *dr6gh-), and the original lengthened grade, expected in the Nom. sg., may be preserved in Arm. durgn (with -u- regularly from *-6). The same root is attested in the e-grade in the verb *dhregh-o- (Gr. Opexw, etc.). This verb probably formed the s-aorist in PIE (cf. Gr. Hom. eOpeXa 'I ran'), so that another, less probable analogical source of *-k- in Celtic is the 3rd person singular aorist (PIE *dhregh-s-t, which was probably pronounced as [dhrekst].
Next, both PIE *lewbh- and *lewp- 'peel, rind' (IEW 690, LIV 377) may represent the same root with the basic form *lewbh-, if we start from a root noun *l6wbh-s 'peel, rind'. The lengthened grade in the nominative may be preserved in Lith. luobas 'peel', Latv. luobs 'id. '. The full grade is attested in Alb. labe 'rind' and OE leaf 'leaf' (from the accusative singular *lowbh-m), while the zero-grade, from the oblique cases (Gen. sg. *lubh-os, etc.) is attested in Lith. lubena 'peel', luba 'plank', Russ. lub 'bast' and Lat. liber 'bark, book' (if from unattested
3 In this case the generalization of root-final *t may have been Italo-Celtic, cf. also OIr. gort field' < *"enclosed field" and W garth 'pen, enclosure' < PCelt. *gorto- (Matasovic 2009: 164-5).
*luber, cf. de Vaan 2008: 337). The same root is attested with final *-p in Russ. lupa 'peel', Croat. lupez 'thief' and the (presumably denominative) Lith. lupu, lupti 'to peel', and Russ. lupit' 'peel, bark, thrash' (apparently a causative, parallel to Lith. laupyti 'tear off' < *lowp-eye-). Russ. luskd 'skin, peel, chaff', Croat. ljuska 'shell, pod, scale' may be from *lewp-ska (Derksen 2008: 291). The Balto-Slavic forms with the unexpected root-final -p may be explained as analogically extended from the nominative singular of the original root noun (*l6wbhs).
Our next example is less certain, because of the vocalism of Lat. falx 'curved blade, hook, scythe' (IEW 247). Its final voiceless stop may have been generalized from the Nom. sg. of a root-noun *dh6lg-s 'sting', from which we also have OIc. ddlkr 'pin, dagger' and Lith. dilge 'nettle' (with the expected metatony in an -iya stem), dilgus 'stinging, smarting' and OIr. delg 'pin, needle', MW dala 'sting'. The ablaut pattern may have been Nom. sg. *dh6lg-s (> OIc. ddlkr), Gen. sg. *dhlg-os (> Lith. dilge), Acc. sg. *dholg-m > Lat. falx, under the assumption that here *o changed into a after a labial, as in *mori > mare 'sea' (OCS more, OIr. muir, etc.). However, this is doubtful, since this change is regular only in open syllables. The e-vocalism in Celtic may be secondary.
Another possible example involves OIr. liach 'unhappy', Alb. lig 'bad', Lith. liga 'illness' and Gr. loigos 'ruin, death, plague' (IEW 667). The voiceless velar in Celtic may be explained by devoicing in the Nom. sg. of the PIE root noun (*leyg-s, or perhaps *l6yg-s, in which case the OIr. vocalism is secondary), while Albanian and Lithuanian point to the zerograde, generalized from the oblique cases (Gen. sg. *lig-os). Greek may preserve the o-grade of the Acc. sg. (*loyg-m), or it is analogical to other o-stems with the o-grade throughout the paradigm. The meaning of the PIE root noun may have been 'illness, weakness', or something similar.
In our last example, the root is clearly verbal, and there is no evidence for the existence of a root-noun in PIE, so it is possible that the variant with the voiceless root-final stop is due to the generalization of the root-form in the 3rd person singular of the root-aorist. PCelt. *sterka 'love' (Matasovic 2009: 355, cf. OIr. serc, W serch) is derived from PIE *sterg-
'take heed, care for' (Gr. uTepyw 'to love', CTTopyh 'love', OCS stregq, stresti 'to guard', Russ. storoz 'guard'). This root formed the s-aorist in PIE, as evidenced by Gr. CTTepXai (LIV 544), and the de-voiced root-final *-k- in the 3rd person singular (PIE *sterg-s-t) may have been generalized in Celtic.
Finally, there are a few less probable, but still possible cases. PCelt. *meko- 'sack, bushel (of corn)' (OIr. miach) can be plausibly connected to OPr. moasis 'barley', Latv. miezi (pl.), Lith. mieziai (pl.) 'barley', if one starts from PIE root-noun *meyg'hs [meyks] from which the root form with *k could have been extended analogically (Matasovic 2009: 162). However, in the absence of cognates from other languages4, this last example is clearly less persuasive than the others. Similarly, positing a PIE root noun *bh6g-s /*bhog-os may help us to relate etymologically OCS bokb 'hip', Russ. bok, Croat. bok and OE baec 'back', OHG bah both from PGerm. *bakan; for the semantic connection cf. OHG bahho 'side of bacon', MHG bache 'ham, bacon, Germ. Hinter-backe 'buttock'. Slavic would have generalized the voiceless velar from the Nom. sg. while the reflex of the original *-g- would be preserved in Germanic. Likewise, a PIE root-noun *(s)m6wg-s / *(s)mug-os 'smoke' (IEW 971) might account for the correspondence between Arm. mux, mxoy 'smoke', OIr. much, W mwg (< Nom. sg. *m6wg-s) and OE smeocan 'to smoke', Lith. smdugti, Russ. smuglyj 'dark brown' (< Acc. sg. *smowg-). And, finally, positing PIE *w6lg-s / *wlg-os 'wetness' (IEW 1145, LIV 617, 621) can help us explain the relationship between OCS vlaga 'wetness', Russ. volglyj 'wet', Croat. vlaga, and Lith. vdlgyti 'make wet', welc 'wet' (with the expected reflex of PIE voiced *g) and Latv. valks 'wet', OIr. folcaid 'washes' OHG welh, OIr. fliuch 'wet' < PCelt. *wlik-wo-). Again the unexpected reflexes with the voiceless velar may be results of analogical extension of the devoiced stop from the nominative singular.
It is worth mentioning that the same process of the generalization of devoiced root-final stops in root-nouns can be observed in the history of Latin. Lat. pax, pacis 'peace' owes
4 The comparison with Khotanese massa 'field' in Mallory and Adams (eds.) 1997: 51 is extremely doubtful.
its root-final -c- to the generalization of the root form which is expected before the final -s. The comparison with pango 'to fix', Gr. piriynu|ai 'to attach', Skt. pajrâ- 'solid, firm' shows that the PIE root was *peh2g'-, and that the voiceless stop was generalized in Latin on the analogy with the Nom. sg., where *g was regularly devoiced before word-final *-s.
To conclude, then, many apparent root-final alternations of voiced (aspirated) and voiceless stops in PIE can be explained by assuming that the voiced (or voiced aspirated) stop is primary, while the final voiceless stop is the result of the analogical extension of the devoiced allophone, which is expected in the nominative singular of a root noun (before word-final *-s). This insight can help us recognize some new etymologies (as in the case with Lat. focus, which may be derived from PIE *dhegwh- 'burn') and better understand some apparently irregular equations (e.g. between Proto-Slavic *lupiti and *lub^). In most cases we were able to adduce independent evidence for the existence of root nouns in PIE, while in other, less reliable cases, this assumption must remain conjectural.
References
Derksen, R. Etymological Dictionary of the Slavic Inherited Lexicon.
Leiden: Brill, 2008. de Vaan, M. Etymological Dictionary of Latin and the other Italic Languages. Leiden: Brill, 2008.
DIL = Dictionary of the Irish Language. Dublin: Royal Irish Academy, 1990.
Ernout, A. & Meillet, A. Dictionnaire étymologique de la langue latine. Paris: Klinksieck, 1979.
Hamp, E. Focus // Glotta, l992. Bd. 70. S. 82.
IEW = J. Pokorny, Indogermanisches etymologisches Wörterbuch.
Bern: Francke, 1959.
LIV = H. Rix et alii, Lexikon der indogermanischen Verben.
Wiesbaden: Harrassowitz, 2000.
Mallory, J. P. and Adams, D. Q. (eds.) Encyclopedia of Indo-European Culture. Chicago: Fitzroy Dearborn, 1997.
Matasovic, R. Gender in Indo-European. Heidelberg: Winter, 2004. Matasovic, R. Etymological Dictionary of Proto-Celtic. Leiden: Brill, 2009.
Nagy, G. Six studies of sacral vocabulary relating to the fireplace // Harvard studies in Classical Philology 78/1974: 71-106.
Schindler, J. L'apophonie des noms-racines indo-européens // Bulletin de la Société de linguistique de Paris, 1972. T. 67. P. 3138.
Walde, A. & Hoffmann, J. B. Lateinisches etymologisches Wörterbuch. Heidelberg: Winter, 1930-1950.
Summary
Many apparent root-final alternations of voiced (aspirated) and voiceless stops in PIE can be explained by assuming that the voiced (or voiced aspirated) stop is primary, while the final voiceless stop is the result of the analogical extension of the devoiced allophone, which is expected in the nominative singular of a root noun (before word-final *-s). This insight can help us recognize some new etymologies (as in the case with Lat. focus, which may be derived from PIE *dhegwh-'burn') and better understand some apparently irregular equations (e.g. between Proto-Slavic *lupiti and *lubi).
