No CrossRef data available.
Published online by Cambridge University Press: 22 April 2026
This article was suggested by that of Edgerton in Lang. 10.235 ff., in which he shows the insufficiency of the arguments adduced from Sanskrit for the existence of a schwa secundum in Indo-European times. Since the facts alleged in support of the theory from this language may as well be explained otherwise, the question of the existence of the ъ is thrown back upon other languages, notably upon Latin and Greek, the proof from which is supposed to be most conclusive. The writer of this article consequently wishes to point out that the evidence from these also has no compelling power, and that it can be interpreted otherwise more persuasively. He also wishes to show the complete lack of conclusiveness of the a-prioristic generalities used to support the schwa-secundum theory. While he therefore does not take it upon himself to say that he is positive that there was no such thing, he is certain that no decisive evidence for it has as yet been found.
1 The material discussed in this article is drawn from the two principal more recent discussions of the schwa secundum: Hermann Güntert, Indogermanische Ablautsprobleme, Strassburg 1916; Hirt, Indogermanische Grammatik 2.76 ff. These two works are usually referred to merely by the name of the author and the page number.
2 In accordance with the more recent practice I use the sign ъ to designate the schwa secundum under all circumstances, although this deviates partially from Hirt's notation, and completely from that of Güntert.
3 In case of IE changes it is self-evident that such stages cannot be established from actual documents, nor would it help if this could be done, since the existence of ъ would presuppose the persistence of such a stage to a later period.
4 Güntert (128 ff.) makes complicated suggestions, but recognizes that they are no more. The more important of these are discussed below. Others, as the assumption of varying sandhi conditions, or of allegro and lento forms, are so completely incapable of proof that they cannot even be discussed.
5 Cf. Edgerton, Lang. 10.263 f.; Sturtevant, Lang. 11.184.
6 Hirt (194) admits that both zero grade and reduced grade occur before the accent in dissyllabic words, but thinks he can explain the conditions. For longer words he also suggests accentual conditions to account for the divergence. Against such attempts, Güntert 123.
7 Even if we admit this a of many Latin words to have been a weak-grade vowel, this does not mean that it was necessarily a schwa secundum. Since a is the regular Latin representative of IE ∂, it is at least as probable that this was found in the e-series under certain conditions, as that there were two weak-grade vowels.
8 On the representation of ъ by Gr. α adherents of the schwa-secundum theory vary widely. Güntert (43 ff.) thinks α is the spontaneous development of ъ, Osthoff (MU 6.212) thought it appeared only after a liquid or nasal, while Hirt (80) doubts Osthoff, and rejects Güntert completely.
9 Cf. Schwyzer 272. A possible suggestion that *-tur- as well as *k ṷъ- arose in oblique cases, as the gen. pl.
, and made its way from there to πίσυρες, cannot be accepted because πίσυρ- is confined to the nominative and accusative plural, where *kṷét- is certainly original. It is true, however, that no gen. pl. of ‘four’ occurs in Homer, but, if *πισυρῶν had existed, there would have been no reason for avoiding it, whereas τεσσάρων or πεττάρων would not fit into the meter.
10 According to the usual interpretation of Osc. patensíns it would be derived from *patnō or *patenō (so Buck, Gramm. of Osc. and Umbr. 163). It would thus differ in its structure from both Gr. πίτνŋμι and Lat. pateo, although one would naturally suspect a close affinity to the latter. Perhaps Osc. en is here merely a way of spelling ē before s, which became possible because n before s apparently was no longer pronounced. In this case Osc. patensíns was identical with Lat. patērent.
11 Hirt (loc. cit.) admits such possibilities, but claims that not all instances can be thus explained. It is difficult to follow him, since accentual variation in every instance led to the continuance of the parallel strong form to serve as a pattern to cause metathesis in the weak grade and thus to bring it into conformity with itself.
12 Güntert (98) claims for ъ even the case-forms of Skt.
gen. bhruv-áḥ, although OHG brāwa ‘eyebrow’ < *bhrēṷā shows that the weak grade in its first stage must have had a schwa primum, and that ū in
came form ∂u∂.
13 See note 10.
14 Brugmann, Gr. 11.407, thought of the influence of Italic *mais = Osc. mais on Lat. magis, and then on magnus, but even *mais and *megis were much farther apart phonetically than *meg- and the *mak- of *makro-s. Cf. Osthoff MU 6.224, Güntert 52. In this connection it is interesting that Hitt. mekkis ‘great’ (:Gr.
, Goth. mikils, etc.) owes its kk < IE k to contamination with this *mak-, cf. Sturtevant, HG 83. Thus both Lat. magnus and Hitt. mekkis were affected by the same root.
15 It cannot be determined whether these mutual contaminations also affected the quantity of the i, since the two words occur mainly in prose, and inability to prove
or *patīna may be accidental.
16 Hirt of course assumes ъ for frango and fractus as well as for Goth. brukans.
17 Mere mention will suffice for Czech. čtyři and Pol. cztery ‘four’ (Osthoff, MU 6.212), for which derivation from *čъtyr- is very improbable because of the regular OBlg. četyre. Cf. Brugmann, IF 28.370, who explains the Czech and Polish forms by dialectic loss of e. Güntert (47) counters with the surprising and dogmatic statement that the same change could not have taken place independently in the two languages. One need refer only to the independent and yet parallel development of Umlaut in English and German or to the diphthongization of ī and ū in both.
18 Cf. Walde-Hofmann s.v. aser.
19 Only the Oscan word is close enough to trabs semantically to make the comparison cogent. If the others are eliminated trab- is the regular weak grade with ∂ corresponding to the normal *trēb- in trííbarakavúm.
20 This reconstructed form is of course my own, since others insist on original *z-dhi = Av. zdī. To me the fact that the strong grade
- took the place of the original weak grade *σ- also in the dual and plural indicative (excepting the 3. pl.) indicates that in all structurally transparent forms the strong
- was adopted. It is no stranger that *
θι took the place of *z-dhi than that *s-te became
. It is to be observed also that the accent on the penult favors original *es-dhi although this is not decisive, since the tendency of verbs to retract their accent could have been at work here just as well as in ἴσθι: οῑδα.
21 For
Wackernagel, IF 25.329, suggests a phonetic change of ηλι to ιλι, for which he also finds other evidence.
22 It is of no advantage to think of a sonant z in some cases, as did Thurneysen, KZ 30.351 ff. The theory has been generally rejected.
23 I refrain from citing Gr. κρῑθή ‘barley’ because it is etymologically ambiguous. If it is cognate with Lat. hordeum and OHG gersta ‘barley’ it arose somehow from *gherzdā and would belong here. It may, however, go with OEngl. grātan and Engl. groats (Uhlenbeck KZ 40.555), and come from IE *ghrīdh-. In either case Grassmann's Law accounts for Gr. κ- instead of χ-, and there is no reason for definitely adopting the latter alternative, which is less enlightening semantically, and has its own phonetic difficulty in the lack of correspondence of the final consonants in the compared words.
24 Holger Pedersen, IF 2.293, briefly suggested the influence of the ι of ἴστημι, τίθημι, πίμπλημι, etc., on σκίδνημι beside the aorist
.
25 Of the various explanations offered for the much-discussed ι of ιππος, two merit serious consideration, without being so convincing that they can be definitely accepted. Brugmann, IF 22.202, suggested that the wandering ιννος ‘colt of mule’, which is the Lat. hinnus, first infected
to become ἴκκος, and that the ι went from this to *ἴππος (still found in compounds like Ἀρίστ-ιππος). Although
and ἴννος before their contamination were not so much alike as to suggest similarity of sound as a contributing factor, the semantic association was perhaps strong enough to account for the change. The second possibility is that suggested by Meillet, Mem. 9.136 f.:
, and the latter with υ like νύξ. However, the assumed change
is dubious. Of the other alleged examples νύξ (AJP 46.56 f.) is best explained as receiving its υ from a contrasting *λύξ ‘light, daylight’: Lat. lūx, and the υ of ὄρτυξ and ὄνυξ may be due to suffix variation. The apparent advantage of explaining the rough breathing of ἴππος by the older stage *u-, which according to the theory must have existed at the time every initial υ became aspirated, is illusive because compounds like Ἀρίστ-ιππος, just mentioned, show the aspiration of ἴππος to have been much younger than that of initial υ. If we accept Meillet's explanation in spite of these difficulties, as does Güntert (25), there is no advantage in assuming an original
. On the contrary, ἴππος, accented like Skt. áçva-ḥ, must have had a full grade vowel from IE times. Cf. Holger Pedersen, KZ 38.417.
For the rough breathing of ἴππος I suggest the following explanation, which takes for its starting-point the fact that in Homeric times horses were used only war, and that in the warfare of those days horses were not ridden, but attached to chariots (Gr. ἄρματα), so that
alternated with ἀφ‘ ἄρματος and
ἂρματος in the sense ‘from the chariot’ and ‘on the chariot’. Occasionally ἴπποι and ἄρματα were even combined into one phrase, as in Λ 198
θ' ἴπποισι καὶ ἄρμασι. Under these circumstances there is no difficulty in assuming that the rough breathing slipped over to ἴππος from ἄρμα.
26 This statement ignores two words (
) in case of which accentuation of the ι is prevented by the long ultima. These are consequently of no value in determining the original accentuation of syllables with alleged ъ.
I suggest that probably χϑιζός was originally accented χϑίος, and shifted its accent to the ultima through analogy, perhaps after πρωϊζός ‘on the day before yesterday’. Thus the change εe to ι, in as far as it was a phonetic change (it was analogical in the group πίτνημι), was conditioned on the vowel being accented, which is in line with the musical nature of the Greek accent, since ι has a higher pitch than ε.
27 I refer only to ъ as a distinct phoneme in opposition to ∂. That the latter may have been pronounced differently in different surroundings, is self-evident; but, since no diverging development can be proved, there is no evidence for two phonemes instead of one. One may compare the German unaccented e, with varying pronunciation, but phonemically, i.e. psychologically, a unit.