2. Hypotheses of a Mediated Borrowing

Much (Reference Much1893:34) and Solmsen (Reference Solmsen1895) were the first to propose a Celtic mediation. Their argument is historically quite sound: Celtic tribes are well known to have had intense contact with Rome since at least the 4th century b.c.e. (Untermann Reference Untermann1955:394) as well as, later on, with the Germanic settlers of northern and central Europe; Gothic shows transparent or very likely Celtic loanwords such as reiks ‘king’, siponeis ‘disciple’, kelikn ‘tower, dining hall’, etc. (Lehmann Reference Lehmann1986, s.vv.). Alew would therefore come from a celticized Latin form ⁺olēuo- later undergoing the regular Germanic ⁺o > ⁺a development.

Much’s and Solmsen’s theory, however, soon came under criticism. Zupitza (Reference Zupitza1897) showed it to be untenable on purely linguistic grounds. He rejected the possibility of a Celtic mediation for relying too heavily on reconstruction and not taking into due account the historical Celtic words for ‘oil’. A hypothetical proto-form ⁺olēuo- would yield *⁺oluw in Welsh, for instance, not the attested olew. The latter, together with Irish ola and Breton oleo, points instead to a prehistoric ⁺oleuo-, with a short e. Here the labiovelar glide, Zupitza notes, is not a reflex of Latin v but an indigenous Celtic development: The same is observed in Welsh pydew < Latin puteus ‘well, pit’. The source of the Celtic forms is therefore classical Latin oleum, so the vocalism of the Gothic word remains unexplained.

The “Celtic hypothesis”, though still having proponents in the present day (see, for example, Untermann Reference Untermann1955:394–399, Green Reference Green1998:156–158), subsequently tended to be abandoned in favor of an Illyrian mediation (Kretschmer Reference Kretschmer1948:25). This theory is based on the assumption that the Illyrian language, which is practically unattested if one keeps it distinct from Messapian, was spoken in prehistoric times over an area stretching from the North Sea to Palestine (Krahe Reference Krahe1940). However, more recent research has limited the home of the Illyrians to a relatively small territory roughly corresponding to historical Dalmatia, and their language seems to be represented by onomastic data only (Polomé Reference Polomé, Dietrich and Geckeler1981:509).

The difficulties of the Celtic and Illyrian theories prompted Polomé (Reference Polomé, Ölberg, Schmidt and Bothien1985a:309–311) to propose a Raetic mediation. Polomé excludes a direct borrowing from Latin for historical and geographical reasons: Archaic Latin ⁺olēvom, as noted above, would have disappeared from usage by 150 b.c.e., too early for it to reach the Goths. A language capable of borrowing the word from Latin before that time while having contact with Germanic tribes would be Raetic. Raetic is first attested in the 6th century b.c.e., in inscriptions found in the eastern Alpine region. Many of these inscriptions, however, still need to be conclusively deciphered. The language is thought to be related to Etruscan and, like Etruscan, it lacks an ⁺/o/ phoneme and substitutes a for o in loanwords (Pisani Reference Pisani1964:324–325). Raetic tribes remained active in the eastern Alpine region until they were conquered by Augustus in 15 b.c.e. and gradually Romanized. Polomé postulates that Latin ⁺olēvom was borrowed into Raetic as ⁺alewa-, and that the latter may have been brought north by the defeated Cimbri at the end of the 2nd century b.c.e.

As can be seen, this “Raetic theory”, like the “Celtic” one, relies exclusively on reconstructed forms. As Lehmann (1986, s.v. alew) notes, there is no written trace of a Raetic word for ‘oil’. It should be added that nothing in the attested Raetic corpus leads one to think that vowel length was a distinctive feature in that language and that archaic Latin ē would have been kept as such in it. Moreover, the Romans came into contact with the Raetic tribes during the first half of the 2nd century b.c.e., by which time the Latin word for ‘oil’ is already well attested as oleum in Plautus (for example, Poenulus 1.2) and Cato (for example, De agri coltura 64). Cato, incidentally, is also the first Latin writer to use the adjective raeticus. The parallel form olīvom, modeled on the oblique cases (Ernout & Meillet Reference Ernout and Meillet2001, s.v. oleum) is rare and almost exclusively used in verse. Finally, as already observed, the ⁺o > ⁺a substitution in alew does not strictly require an intermediate language.

In light of these difficulties, it is not perhaps unhelpful to wonder if the hypothesis of a direct borrowing from Latin is really less plausible than that of a mediated borrowing. In what follows, I try to show that there is no cogent case against the former and that it is, in fact, the most parsimonious explanation of the origin of alew.

3. The Case for a Direct Borrowing

The presence of a in the Gothic form, as I already stated, can be explained in two different ways. It could be the result of the Germanic ⁺o > ⁺a shift, which came to completion by the end of the 1st century b.c.e. at the latest (Ringe Reference Ringe2006:145; see also Polomé Reference Polomé and Ureland1985b). This hypothesis is not altogether irreconcilable with what is known about the timing of the Gothic migrations in Europe: The Gutones are found on both banks of the middle Oder in the years 7 b.c.e.–17 c.e., and their territory stretched, in that period, as far south as Lower Silesia. They were certainly known to Roman geographers of the time (Wolfram Reference Wolfram1979:35).

The alternative hypothesis is plain sound substitution: The Germanic merger of Indo-European ⁺o with ⁺a would have virtually eliminated the vowel from the Gothic phonemic inventory. It is entirely plausible that ⁺/o/ would have been substituted by ⁺/a/ in an early borrowing. This explanation would be more parsimonious and more realistic in terms of chronology, allowing the likely date of borrowing to be postponed to the Common Era, the time when Gothic-Roman trade and linguistic contact were surely more intense. Typologically, alew would be consistent with other early borrowings from Latin that pertain to the vocabulary of trade and designate such typically Mediterranean commodities as wine (Gothic wein < vīnum) and vinegar (Gothic aket < acētum), borrowings that go back to the 1st–2nd century c.e. (Corazza Reference Corazza1969:10f.). The hypothesis of sound substitution is also to be preferred for reasons mentioned below.

While substitution of Latin o by Gothic a therefore does not seem to be an insurmountable problem, the same cannot be said of the word-internal e found in the Gothic form. Based on etymology, the Gothic <e> graph is customarily assumed to stand for a long, mid-high front vowel, the reflex of Proto-Germanic ⁺ē₁ and ⁺ē₂, whereas the e in Latin oleum is short, and short e is generally assumed to be a mid-low vowel in Latin. The two sounds thus seem irreconcilably different in quality and quantity. This point deserves some discussion.

As already noted, the attribution of a long quantity to Gothic <e> rests chiefly on comparative evidence. Wulfilian orthography makes no distinction between long and short vowels. It has been argued, therefore, that <e> did not actually stand for a distinctively long vowel in historical Gothic and that the presumed difference in quantity between Latin and Gothic e in alew is a false problem (Szemerényi Reference Szemerényi1989:139). In fact, dissenting voices about the distinctiveness of vowel length in historical Gothic have existed since the beginning of modern comparative linguistics. Rasmus Rask was the first to note that Gothic offered no internal evidence for phonemic vowel length but rather seemed to distinguish vowels by means of qualitative differences only (Streitberg et al. Reference Streitberg, Michels and Jellinek1936:404). Marchand (Reference Marchand1955:79ff.), in one of the lengthiest extant studies on the matter, concluded that it would be arbitrary to state that historical Gothic distinguished vowels by means of quantity. The distinctive value of vowel length in Gothic was also excluded by Hamp (Reference Hamp1958:360) and Bennett (Reference Bennett1959:429). Views to the contrary were held by Wright (Reference Wright1910:25), Mossé (Reference Mossé1956:58), Vennemann (Reference Vennemann1971) and generally by scholars taking a traditionalist comparative approach, while structuralists tend to side with Marchand (Beck Reference Beck1973:115ff.). An extensive overview of the debate can be found in Moulton Reference Moulton, Cardona, Zide and Hoenigswald1987.

The distinctiveness of vowel length in historical Gothic, however, is not so relevant as one may think to the issue at hand. All extant studies into the etymology of alew fail to properly consider the wide time gap between (Pre-)Gothic as it was spoken by the beginning of the Common Era and historical Gothic as one knows it from Wulfila’s work, or, to be more accurate, from the surviving Gothic manuscripts. It cannot be taken for granted that Gothic sounds and phonemes stayed perfectly identical through the 1st, 2nd, 3rd, and 4th centuries. Whether alew is regarded as a direct or indirect borrowing, one should not forget that the form in which the surviving documents show it represents a different stage of linguistic development than that when the word was presumably adopted into Gothic. For this reason, it is not so important to know how the Wulfilian graphemes of the word in its written form were read when the extant Gothic manuscripts were compiled, but rather what Pre-Gothic sounds could come to be designated by each of those graphemes.

It has been claimed (Moulton Reference Moulton1948:80, Marchand Reference Marchand1957:353f.) that Pre-Gothic had a “triangular” short vowel system: ⁺/i/ ⁺/u/ ⁺/a/, due to the coalescence of Proto-Germanic ⁺e and ⁺i, with ⁺e becoming a mere positional variant of ⁺i before r, h, and ƕ. It seems safer to assume, however, that the Pre-Gothic merger of ⁺e and ⁺i only reduced the frequency of ⁺e without obliterating it as a phoneme. The vowel could still be found in the reduplicating syllable of strong verbs, at least before internal open juncture, as in ai-auk ‘I added’ versus bi-uhti ‘custom’ (Cercignani Reference Cercignani1979:274), and where the consonant causing breaking was obscured by assimilation, as in aiþþau ‘or’ < ⁺i-ƕe-þau (Cercignani Reference Cercignani1984). Pre-Gothic ⁺e was a mid-low front vowel, and it had a long counterpart in the ⁺[ɛ:] sound resulting from the lowering of ⁺ē before vowels, as in saian “to sow” < PGmc ⁺sē(j)anan (Bennett Reference Bennett1967:7). The historical reflexes of long and short mid-low front vowels would come to be graphically rendered as <ai> in a very consistent fashion in Wulfilian orthography. Wulfilian orthography also consistently keeps mid-low front vowels distinct from the historical reflex of Pre-Gothic ⁺ē, spelled <e> in 98.28% of cases (Snædal Reference Snædal2013:287). The scholarly consensus regards the latter, as already noted, as a mid-high front vowel (Mossé Reference Mossé1956:58, Marchand Reference Marchand1973:75, Vennemann Reference Vennemann1971:126–130, among others).

No short mid-high front vowel seems to have existed in Pre-Gothic. There is, however, one instance of such a vowel being introduced into the language through an early loanword. Gaulish celicnon ‘tower, building’, which was probably borrowed into Gothic during the Goths’ stay in Central Europe (Green Reference Green1998:158), shows a short mid-high e sound (the vowel is occasionally spelled <i> in inscriptions). In historical Gothic, the word appears consistently spelled kelikn, therefore there is at least one unambiguous instance of the <e> graph standing for the historical reflex of an etymological short vowel. As Eska (Reference Eska1990:67) has proposed, the most likely explanation is that Gaulish e, an alien sound to the native Gothic phonemic inventory, distinguished both from Pre-Gothic ⁺/ɛ/ (spelled <ai> in Gothic) and ⁺/i/, was borrowed as such into Gothic and spelled <e> to keep it distinct from the sounds designated by the graphs <ai> and <i>.Footnote ¹

Another point that has been overlooked by scholars investigating the origin of alew is the Vulgar Latin treatment of vowels in hiatus. It is a fundamental rule of Latin prosody that any given vowel is short when immediately followed by another vowel, and this applies to oleum, of course. Short Latin e is generally reconstructed as a mid-low vowel, on the basis of its Romance reflexes. As Meadows (Reference Meadows1946) notes, however, when originally long Latin vowels are shortened in hiatus they tended to retain their relatively closed pronunciation in everyday speech, while originally short ones tend to be raised. As mentioned above, the e in oleum was originally long. In unstressed position the tendency to a closed pronunciation of e in hiatus shows up quite early: The vowel is occasionally spelled <i> already in Archaic Latin, and such spellings become frequent in informal writing during the Empire (for example, casiu for caseum ‘cheese’ in Pompeiian graffiti; Väänänen Reference Väänänen1981:96).

In stressed syllables, the development happens much later, so much so that it leaves virtually no written trace in antiquity and comes to completion only in Medieval times and with considerable dialectal variation. For example, Old French has a double form for ‘mine’: mieu for Latin meus and moie for Latin meam, the former reflecting a Vulgar Latin form with a mid-low e and the latter reflecting a Vulgar Latin form with a mid-high e.Footnote ² In unstressed syllables, the vowel would eventually develop into a glide and oleum would be reduced to a disyllable, although vulgar forms with e are still attested in late antiquity (for example, oleu in a 6th century Egyptian papyrus, Cavenaile Reference Cavenaile1956–1958: nr. 277). Furthermore, in some areas of the Latin domain, as I show below, the -eu- sequence would occasionally develop into -ewu-, not into -ju-.

The historical Gothic reflex of the e found in oleum thus becomes explainable if one assumes a mid-high quality for it in Latin and for its subsequent spelling <e> in historical Gothic, by the same mechanism observed in the case of Gaulish celicnon ‘tower, building’. This assumption makes the hypothesis of sound substitution (a for o) in the first syllable preferable. The borrowing must have taken place after the Gothic raising of ⁺e, otherwise ⁺aliw would have resulted. The latter change must have taken place after the pan-Germanic ⁺o > ⁺a shift, therefore after the beginning of the Common Era, but before the operation of Gothic Brechung, which it feeds. Brechung must have taken place late enough to affect pre-Wulfilian Latin loanwords (as in Latin urceus > Gothic aurkeis ‘ewer’). This line of reasoning would put the most likely date of borrowing of oleum at the 2nd century c.e., approximately at the time of borrowing of vīnum ‘wine’ and acētum ‘vinegar’. By that time, according to Tacitus, the Gutones were well known to the Romans and the Empire had intense trade relations with the Germanic tribes.

Zupitza’s (1897) claim that the labiovelar glide found in the Celtic words for ‘oil’ is a Celtic development has been challenged by later studies. Jackson (Reference Jackson1953:87) has shown that the insertion of w to fill a hiatus is a feature of British Latin; such forms as puteus ‘well, pit’, oleum ‘oil’ came to be pronounced ⁺putéwus, ⁺oléwum in Latin as spoken in Britannia, and hence Welsh pydew, olew. This, Jackson argues, seems to be a Latin dialectal feature peculiar to the province of Britain: Nothing of the sort can seem to be found in Continental Vulgar Latin. The phenomenon must be regarded as Latin, not Celtic, because it is exclusive to loanwords: Other examples are Latin leo > Welsh llew ‘lion’; Mattheus > Mathew; Iudaeus > Iddew ‘Jew’. This view is now generally accepted (Haarmann Reference Haarmann1970:22).

Some criticism has, however, been raised against Jackson’s view of w-insertion as a purely Britannic phenomenon. Szemerényi (1989:129ff.), in a lengthy treatment of Gothic alew and its etymology, notes that a few continental Latin loans into Germanic and Slavic languages show a similar addition of w in hiatus. Old High German leuuo ‘lion’ is one such case, suggesting derivation from a Vulgar Latin form ⁺lewo, just like Welsh llew. The Old Church Slavonic word for ‘lion’, lьvь, also shows the reflex of a labiovelar glide. The Slavic form cannot be explained as a borrowing from German, as the regular Old Church Slavonic reflex of German e is e, not ь (Kiparsky Reference Kiparsky1934:284). Szemerényi explains the Slavic and German forms as borrowings from a hypothetical Gothic ⁺liwa < Vulgar Latin ⁺lewo, as already suggested by Stender-Petersen (Reference Stender-Petersen1927:362f.).

The main difficulty with this hypothesis, as Szemerényi himself concedes, lies in the fact that a Gothic word for ‘lion’ is not attested. The recently discovered Gothic fragment from Bologna (Gothica Bononiensia), however, seems to yield a belated confirmation of Szemerényi’s thesis, namely, the reading of laiwa for Greek λέων ‘lion’ (Falluomini Reference Falluomini2017:288). Be this a direct or mediated borrowing, it does not stand in isolation. Another instance of w-insertion in hiatus is found in the Old Church Slavonic stem židov- ‘Jew’, a form that, similarly to Welsh Iddew, points to a Vulgar Latin ⁺jūde-w-us.

That the insertion of a hiatus-filling w after front vowels was not exclusive to Britannic Latin is also shown by some continental inscriptions. At Pompeii one finds the vulgar form paeuoniam for paeoniam ‘peony’ (the phenomenon, therefore, was already taking place during the early decades of the Principate): As shown by other inscriptions from Pompeii (for example, in the inverted spelling maeae for meae ‘mine’), the Latin diphthong ae had undergone monophthon-gization to e in 1st-century Vulgar Latin as spoken in Italy. The labiovelar glide found in Gothic alew may therefore be the reflex of the insertion of a hiatus-filling w after post-consonantal e, a development frequent in Britannic Latin and rare, but not absent, in other areas of the Latin-speaking world (Dilts Reference Dilts1977:297). Like Slavic, Gothic seems to have had some contact with a variety of Latin, as suggested by the recently discovered reading of laiwa mentioned above.

One might wonder why the e in leo was not subject to the same treatment as the e in oleum, resulting in Gothic ⁺lewa. This can be explained by bearing in mind that the Vulgar Latin raising of mid vowels in hiatus affected stressed vowels at a much later time than nonstressed ones, as already noted. Leo has a stressed e that should be assumed to have still been mid-low at the time of the borrowing.