1. The qameṣ

One of the peculiarities of the Tiberian tradition of Biblical Hebrew (henceforth: T) is the use of the one sign, <אָ> (qameṣ),Footnote ¹to denote the reflex of two distinct historical vowels. Although qameṣ supposedly designates a unitary vowel quality [ɔ],Footnote ²it can reflect both a historically short */u/ in closed, unstressed syllables, and a historically secondarily lengthened */aː/,Footnote ³principally in open or stressed syllables. Thus, in a word such as קָרְבָּן [qɔɾˈbɔːn],Footnote ⁴“sacrifice”, the first qameṣ reflects a historical */u/, whereas the second one reflects a historical */aː/: */quɾˈbaːn/ > [qɔɾˈbɔːn].Footnote ⁵

While the dating of the */u/ > [ɔ] shift in closed, unstressed syllables is considered late, given its absence in the closely related Babylonian tradition (B),Footnote ⁶the dating of the former shift was a major subject for the past century. Several different approaches were used to tackle it in scholarship, including internal analyses of vowel qualities within T (Khan Reference Khan1987: 44–5; Blau Reference Blau1998: 247–50), the shift’s relation to similar shifts in light of historical linguistics (Kutscher Reference Kutscher1969: 84–91, esp. 90–1; Suchard Reference Suchard and van Beek2025: 233–4),Footnote ⁷a comparison with other traditions of Hebrew (Morag Reference Morag1963: 102–4),Footnote ⁸and comparisons of T qameṣ to scriptio plena <ו> in the Dead Sea Scrolls.Footnote ⁹

To the various studies cited above, I wish to offer a renewed analysis of data from a source hitherto almost unmentioned: the Greek transcriptions in Origen’s Secunda. Origen, a church father and biblical scholar (c. 185, Alexandria–253, Caesarea), compiled a multi-columnar synoptic edition of the Bible known as the Hexapla (ἑξαπλᾶ, “sixfold”), whose second column (the Secunda) contained a transcription of Hebrew into Greek letters. Since Greek uses letters rather than diacritics to express vocalic quality, the Secunda exhibits one of the most important textual sources for appreciating the vocalic qualities of Biblical Hebrew, qualities which derive from a period that significantly pre-dates the evidence of T and of other Hebrew vocalization systems by several centuries.

I know of only one study that has attempted to cite evidence from the Secunda for the */aː/ > [ɔː] shift. In his doctoral dissertation, Benjamin Kantor (Reference Kantor2017: 274–9) noted the Greek transcription of Ps. 7: 8: ουαλεα (corresponding to T: וְעָלֶיהָ [vɔ̆ʕɔːˈlɛːhɔː], “and upon her”), in which <ε> represents an [ɛ] quality rather than [e]. Since in this environment [ɛː] is the result of vowel harmony with a qameṣ following a guttural (*/eː/ > [ɛː]/_Gɔ; see Blau Reference Blau1996: 21–3; Reference Blau2010: 181), the appearance of an [ɛː] implicitly assumes that the final vowel, represented by <α>, had already shifted to [ɔː], placing this shift sometime before Origen’s time. While Kantor deals at length with the plausibility of [ɔː] being represented by Greek <α>, he nevertheless acknowledges the indirectness of the proof and the need to base it on more detailed textual criticism of the relevant sources.

To Kantor’s example of indirect evidence for the shift, I wish to suggest the possibility of direct evidence for the shift. If this possibility is deemed true, it may offer us hints about the development and chronology of the shift.

2. Correspondences with Secunda <ο> and <ω>

The text of the Secunda survives for the most part in a single major copy of Psalms, the Mercati palimpsest,Footnote ¹⁰and a detailed linguistic study of it was published by Einar Brønno (Reference Brønno1943). Listing parallels between T vowels and their Greek correspondences in the Secunda, Brønno (Reference Brønno1943: 354) cited four lexemes in which qameṣ is equivalent to a Greek /o/-grade vowel (either <ο> or <ω>):Footnote ¹¹

1. 1. Ps. 18: 39: εμωσημ (T: אֶמְחָצֵם [ʔɛmħɔːˈs^ʕeːm], “I [shall] smite them”).

2. 2. Ps. 18: 43: ουεσοκημ (T: וְאֶשְׁחָקֵם [vɛ̆ʔɛʃħɔːˈqeːm], “and I [shall] grind them”).

3. 3. Ps. 89: 35: σφωθαι (T: שְׂפָתַי [săfɔːˈθaj], “my lips”).

4. 4. Ps. 49: 2, 89: 48: ολδ (T: חָלֶד [ˈħɔːlɛd], “world”).

Morphologically speaking, lexemes 1–2 are both prefix-form 1sg G-stem verbs, with a 3pl.m pronominal suffix. Lexeme 3 is a plural noun with a 1sg possessive suffix, and lexeme 4 (two identical tokens) is a pausal “segolate” nominal form.

Scholars seem to have interpreted these primarily as morphological variants. For the two verbal forms, Brønno (Reference Brønno1943: 34), following Speiser (Reference Speiser1933: 38), assumed that the /o/-grade vowels represented a different Hebrew pattern to T, yiqtol rather than yiqtal. The same opinion is taken up by Janssens (Reference Janssens1982: 167), Yuditsky (Reference Yuditsky2017: 119) and Maurizio (Reference Maurizio2023: 130–1), who all understand these forms to be yiqtol.Footnote ¹²It is worth mentioning that as a morphological discrepancy, it would be discrepant not only from T but also from B (Yeivin Reference Yeivin1985: 1.469).

As for σφωθαι, scholars agree that the vowel differs from T. But while Brønno (Reference Brønno1943: 100) and Janssens (Reference Janssens1982: 139) explained this as a different form,Footnote ¹³Yuditsky (Reference Yuditsky2017: 36, n. 24; Reference Yuditsky, Joosten and Rey2008: 234–5) suggested that there was a phonetic change in the glide, where /śVp.woː.taj/ had shifted to /śVp.ʔoː.taj/. This theory can also be supported by the form בשפאותיכה, with an <א>, /ʔ/, in the Dead Sea Scrolls.Footnote ¹⁴While entertaining Yuditsky’s explanation, Kantor (Reference Kantor2017: 231–2) adds the possibility that <ω> could be a scribal designation of [w]. Another possibility is that the historical */aː/ vowel was rounded in assimilation to the preceding labial <φ> (/p^h/),Footnote ¹⁵or a scribal error, where the circularity of the nearby <φ> and <θ> may have resulted in intermediate <ω>.Footnote ¹⁶

And finally, ολδ is explained as a different segolate form.Footnote ¹⁷Brønno (Reference Brønno1943: 136–7) notes that if it were a qatl form, as in T, the expected transcription would be *ελδ, citing Arabic خُلْدٌ /ˈḫuldun/, “paradise”, as support for a qutl form. Yuditsky (Reference Yuditsky2017: 188) concurs, noting the discrepancy with the other vocalized traditions, T and B,Footnote ¹⁸and citing evidence for the same form in the Dead Sea Scrolls and the church fathers.

Our case in point pertains to the forms εμωσημ and ουεσοκημ, which have been noted as prefix-form 1sg G-stem verbs with pronominal suffixes. The G-stem in Hebrew exhibits two different patterns in the prefix-form, distinguished by their “theme vowel”: yiqtol, featuring historically in fientive (or dynamic) verbs, and yiqtal, featuring historically in stative verbs.Footnote ¹⁹Since both “smite” and “grind” are fientive verbs, this concurs prima facie with the verbs having an /o/-grade realization before their third radical. However, as is often the case in other West Semitic languages as well, in roots whose second- or third-radical consonant is a guttural, the final vowel was lowered to [a], resulting in a secondary yiqtal form in fientive verbs as well.Footnote ²⁰Since both verbs listed here have a guttural consonant for the second radical, the expected form would be yiqtal, as is indeed the case in T.

The discrepancy between the different forms has enjoyed two explanations. The first one, seemingly implied by Brønno (Reference Brønno1943: 34), suggests that vowel-lowering never took place in the tradition of the Secunda. This follows from the general characteristics of the gutturals in Hebrew, which have a greater effect on preceding vowels than on succeeding ones.Footnote ²¹Thus, for verbs whose third radical is a guttural, we find a preceding [a] (or [ɔ]) vowel across the paradigm – whether by lowering the original vowel or adding a secondary [a] vowelFootnote ²²– but for verbs whose second radical is a guttural, the following vowel, which is in the same syntagmatic position, is generally retained. We find a lowered vowel [a] only in the prefix form of such verbs, but not in the closely related imperative or infinitive (Bergsträßer Reference Bergsträßer1965: 2.116 [§22e]; Kautzsch Reference Kautzsch1976: 169 [§64b]; Joüon and Muraoka Reference Joüon and Muraoka2011: 169 [§69a]). The discrepancy between T yiqtal and Secunda yiqtol can be thus explained: since the effect of a second-radical guttural consonant on subsequent vowels is relatively weak, it could be that in Secunda Hebrew this process, which is partial in T to begin with, might never have taken place.

Another suggestion was raised by Yuditsky (Reference Yuditsky2017: 121), whereby some yiqtal forms changed back into yiqtol. Since in the Secunda this occurs only in forms with a pronominal suffix, Yuditsky suggested this was done by means of analogy to the suffixed infinitive form liqtol-, a process which Qimron (Reference Qimron2018: 197–8) first identified in Qumran Hebrew, using it to explain the existence of yVqotl- forms (that is, prefix-form verbs in the G stem with an /o/-grade vowel before the second radical rather than after it).Footnote ²³Since yVqotl- forms are unique to Qumran Hebrew, by connecting them with similar forms in the Secunda, Yuditsky was able to effectively support both sides of the equation, showing that these are not two sets of anomalies but a single process common to both traditions.

The difficulty with Brønno’s suggestion is that it does not explain why yiqtol features specifically in forms with pronominal suffixes. Yuditsky’s explanation, like Qimron’s, fully accounts for that, wherein lies its strength. And yet methodical difficulties have recently been noted in it (Stadel Reference Stadel2024: 210–1),Footnote ²⁴in turn raising a new suggestion for the Qumran forms: namely, that the unstressed vowel was elided, and a subsequent epenthetic vowel was inserted in the resulting consonantal cluster: yVqtoleni > *yVqtleni > yVqotleni (Stadel Reference Stadel2024: 214–6). Without attempting to pass a verdict on the matter, suffice it to say that if such a position is adopted, this reopens the question of the Secunda forms. My suggestion below is therefore contingent upon the postulate that yVqotVl- forms in Qumran may not be the result of an analogy to the infinitive (while not denying that they may well be so).

3. Vowel-rounding in open syllables

Any explanation of the Secunda’s forms needs to account for the conditioning where apparent yiqtol occurs with pronominal suffixes but yiqtal occurs without it. Unlike Brønno, I believe these forms are better accounted for as genuine yiqtal forms; and unlike Yuditsky, I think a phonological explanation may be provided, which in turn would (albeit distantly) tie this feature to B and T, rather than to the Qumran forms. The addition of a pronominal suffix means that the final consonant of the root changed from being the coda of the second syllable to the onset of the third syllable, in turn positioning the /a/ theme-vowel of the yiqtal form in an open syllable. We know from both T and B that the qameṣ shift */a/ > [ɔ] occurred (among other cases) in open syllables. If such a process had occurred in the Hebrew of the Secunda as well, the rounding of the */a/ vowel may have resulted in a phone that was perceptually mapped to */o/, subsequently being rendered as <ο> (or <ω>) in Greek transcription.

If we look at all yiqtal prefix-forms of the G stem with a second-radical guttural, the process will adequately explain the forms.

Although these are only six tokens, they seem to hint at a conditioned distribution, where an */a/ theme vowel is rendered as <α> in closed syllables but as <ο> or <ω> in open ones. This corresponds to T, which displays the theme vowel as pataḥ ([aː]) in closed syllables but as qameṣ ([ɔː]) in open ones. The one exception marring this distribution is ὀυαθθεμας which, following the correspondence with T, we would expect to see with an /o/-grade vowel in the tonic. However, the T qameṣ here represents a pausal form. It could be construed that the Secunda tradition did not read a pausal form here, or alternatively that the */aː/ > [ɔː] shift, which occurs in here in the environment of closed, stressed syllables, had not yet begun to be operative (as it is in T). Since no other pausal forms of *yiqtal are attested in the Secunda, neither explanation can be properly confirmed or refuted based on the extant evidence.Footnote ²⁷

A small “control group” of forms that have a third guttural radical all show that vowel-lowering was maintained before the guttural:

• • Ps. 30: 9: εκρα (T: אֶקְרָא), 46: 2: νεμσα (T: נִמְצָא)Footnote ²⁸and 89: 27: ικραηνι (T: יִקְרָאֵנִי).

These are all inflections of the verbal roots with a third guttural radical, /ʔ/, classified as yiqtal forms (Yuditsky Reference Yuditsky2017: 118). But this classification is synchronic, resulting from the vowel-lowering that precedes the guttural (which survived quite late; Suchard Reference Suchard2020: 80–81).

While evidently not the only explanation, the suggestion that */aː/ > [ɔː] occurred already in the Secunda in open syllables of (secondary) yiqtal forms of the G stem is supported from the three instances above with pronominal suffixes, against the three cases without pronominal suffixes. That these are indeed secondary yiqtal forms with vowel-lowering is confirmed from the three instances of the “control group” with a third radical /ʔ/.

4. Dating */aː/ > [ɔː] in the Secunda and the Masoretic traditions

If early traces of the */aː/ > [ɔː] shift are indeed recognizable in the Secunda, these need to be corroborated with the chronology of the shift in other Hebrew traditions, namely T and B.Footnote ²⁹The first immediate material evidence of rounding of a historical */aː/ vowel is in Hebrew citations in Aramaic incantation bowls, which can be dated to the late sixth and early seventh centuries, and are assumed to correspond to B.Footnote ³⁰Indirect evidence has also been drawn from relative chronology of phonological processes. With regard to T, Geoffrey Khan (Reference Khan1987: 44–5) has shown that the */aː/ > [ɔː] shift had to precede the secondary lengthening of */a/ vowels (long pataḥ), a process which postdates the Secunda but is attested in the Karaite transcriptions of T into Arabic characters. As for B, Joshua Blau (Reference Blau2013: 30–31) has credited pausal forms like יָדַךָ [jɔːˈðaːxɔː] “your hand” to the result of an */ɛ/ > [a] shift particular to B, which must have postdated the */aː/ > [ɔː] shift in open syllables, as evident from pausal forms like שָׁמָר [ʃɔːˈmɔːɾ] (< */ʃaːˈmaːɾ/).Footnote ³¹In other words, both T and B, as evident from vocalized manuscripts of the late first millennium, exhibit the results of shifts particular to them, which postdate the */aː/ > [ɔː] shift. It also seems reasonable to consider */aː/ > [ɔː] to have occurred in the common ancestor to B and T, termed by Kantor (Reference Kantor2023: 77) “Proto-Masoretic”, rather than a parallel innovation. Of course, we do not know precisely at what point the two traditions diverged. Khan has suggested that their common origin can be traced to the Second Temple period, as late as the first century ce, but I believe that a significantly later date can also be justified.Footnote ³²

Another suggestion was raised by Kantor. Noting that T and B both feature an unusual qameṣ before <ו> in words such as מָוֶת “death” [ˈmɔːvɛθ] and קָו [qɔv] “line”,Footnote ³³this shift would have had to occur when the consonant was still labio-velar ([w]) in Palestine rather than labio-dental ([v]). By examining the epigraphic evidence of Greek and Latin transcriptions, Khan and Kantor were able to date this consonantal shift “by the Byzantine period at the latest”, i.e. third–fourth centuries in Palestine, which means that the */aː/ > [ɔː] shift would have pre-dated it.Footnote ³⁴But this evidence is problematic: while in words such as מָוֶת and קָו the [ɔ] is indeed assumed to be the result of rounding before a labial ([w], [m]),Footnote ³⁵the shift is particular to such an environment. It is thus unable to explain the broader shift attested in open syllables or in closed, stressed ones. Moreover, such rounding would have had to pre-date the more regular */aː/ > [ɔː] shift, since the assimilated vowels would have necessarily been unrounded, i.e. pre-shifted, for them to undergo assimilation.

Unlike assimilation to labials, nuclei of open syllables in the prefix-form of the G stem are not bound by the same environmental conditioning, being wholly determined by syllabic structure. But while the Secunda evidence does not indicate an environmental restriction on the */aː/ > [ɔː] shift, it does confine it to a given morphological category. We would thus need to explain how a very limited “shift” in the Secunda is reflected much more extensively in the later-attested traditions of B and T.

5. Lexical diffusion

According to the standard neogrammarian model, sound shifts are regular, applying uniformly across the language system, and are principally unconditioned (Murray Reference Murray, Honeybone and Salmons2015: 22). However, it has been shown that some sound shifts follow a different mechanism termed “lexical diffusion”, in which a change begins at a certain area of the language, gradually spreading across its breadth (“lexicon”). Diffusion of sound changes has been argued to be primarily a result of language (or dialect) acquisition by adults (Labov Reference Labov2007), and the trajectory of such shifts has often been explained by token frequency (Phillips Reference Phillips, Honeybone and Salmons2015). Still, for some sound shifts token frequency has been shown not to be a factor, or at least not a sole factor, in the gradient of the shift (Trapateau Reference Trapateau2020).Footnote ³⁶In fact, although not formally named so, a form of lexical diffusion was used to explain the prehistoric shift of stress in Hebrew from penultimate to ultimate: at first, final short vowels were elided in some nominal and verbal forms, then other forms aligned stress-wise with the former ones by means of analogy (Blau Reference Blau2010: 144–50). Stress shift in prehistoric Hebrew was thus a gradual process which diffused, albeit incompletely, across the “lexicon”.

The fact that the supposed */aː/ > [ɔː] shift is attested in the Secunda in a specific morphological pattern, whereas the /ɔː/ vowel (as a reflex of secondarily lengthened /aː/) in B and T occurs much more extensively, could be accounted for by means of lexical diffusion. Whether or not token frequency could be accurately established for the kind of Hebrew recited in the synagogue, in turn associating the shift with frequency-based diffusion, is a matter for a separate study. The point is that since the readings of the Secunda have a terminus ante quem of the middle of the third century, though they probably pre-date it by some time,Footnote ³⁷and given that the shared stem of B and T may be later than that,Footnote ³⁸it could be the case that as far as this specific parameter is concerned, the Secunda in fact attests a feature whose kernel precedes the shared node of B and T, thus presenting an earlier stage of the shift, before its extensive diffusion.Footnote ³⁹

The difficulty still remains that while lexical diffusion can account for the geographical and linguistic gap between the Secunda (second or third century Palestine) and the incantation bowls (sixth or seventh century Babylonia), this point is moot so long as evidence has not been presented to show the trajectory of */aː/ > [ɔː] between a single verbal inflection and a systemwide shift. Ex silentio, this would be impossible to prove or to refute, but just because evidence had not been presented does not preclude it from existing. It may be worth revisiting the question of Jerome’s Latin transcriptions of Hebrew from the fourth century, in which Carl Siegfried noted places where T “A-Laute” (in essence: /ɔ/) was rendered by a Latin <o>. Examples include בָּשָׂר = <bosor>, זָכָר = <sochor> and רָקָב = <recob> (Siegfried Reference Siegfried1884: 75). The evidence was met with suspicion, not only because Jerome rendered [ɔ]-reflexes of historical */u/ with <a> (Blau Reference Blau1998: 247), or because Siegfried followed T to the extent that he sometimes silently emended Jerome’s texts (Sperber Reference Sperber1938: 105), but mainly due to inconsistent and conflicting data. For vowels equivalent to T /ɔ/, we also find <basar> (alongside <bosor>! Siegfried Reference Siegfried1884: 74) for בָּשָׂר and <soor> for צָחַר (Siegfried Reference Siegfried1884: 75).Footnote ⁴⁰With the provision of pending textual criticism, I subscribe to James Barr’s understanding of these inconsistencies as evidence of Jerome’s difficulty in phonemizing the Hebrew that he heard and in aptly rendering the articulated phonemes in Latin script (Barr Reference Barr1967: 30–1).Footnote ⁴¹Still, if the alternation between <o> and <a> forms (as representations of historical secondarily-lengthened */aː/) in Jerome’s writings indeed consists of a greater number of types and environments than <ο> and <ω> (as equivalent representations of the same phoneme) in the Secunda, this could provide valuable evidence for tracing the diffusion of [ɔː] across the Hebrew lexicon in between the evidence of the Secunda (third century ce or earlier) and that of the earliest B-like incantation bowls (sixth–seventh centuries).

6. Conclusion

Previous scholars explained the odd Secunda forms <εμωσημ>, <ουεσοκημ>, and <θεσοδηνι> as yiqtol forms, whether by allomorphy with T (and B) yiqtal (Brønno) or by analogy to liqtol- infinitives with pronominal suffixes (Yuditsky). My proposal is to interpret these forms as early examples of the */aː/ > [ɔː] shift, whose results are more prominently seen in the incantation bowls, T and B. The benefit of this interpretation is that it does not rely on a specific explanation drawn from a number of eccentric Qumran forms, but follows a well-attested shift, whose gradual spread is explained by means of lexical diffusion. It is also suggested that in Jerome’s transcriptions of Hebrew there could be evidence of an intermediate stage of this shift.