Hostname: page-component-74656f9884-npj9w Total loading time: 0 Render date: 2026-11-06T18:42:03.375Z Has data issue: false hasContentIssue false
  • English
  • Français

The broken plural in Soqotri

Published online by Cambridge University Press:  06 February 2020

Leonid Kogan
Leonid Kogan*
Affiliation:
Russian State University for the Humanities / National Research University – Higher School of Economics, Moscow
*
lkog@rggu.ru
Rights & Permissions [Opens in a new window]

Abstract

The article investigates the broken pural in the Modern South Arabian language Soqotri (island of Soqotra, Gulf of Aden, Yemen). It is based on extensive field research and rich collections of lexical evidence. Primarily synchronic in its approach, the article pays attention to historical problems of Modern South Arabian and Semitic phonology and morphology whenever appropriate.

Keywords

SemiticModern South ArabianSoqotriNominal morphologyPlural formation

Information

Type
Article
Information
Bulletin of SOAS , Volume 83 , Issue 1 , February 2020 , pp. 1 - 23
Copyright
Copyright © SOAS University of London, 2020

I. Introduction

Like many other aspects of Soqotri morphology, the present subject has practically no history of research: since Bittner's pioneering remarks (Reference Bittner1918: 62–6), the Soqotri broken plural has not been discussed in the scholarly literature, even within the summary descriptions of Modern South Arabian as a group.Footnote 2 We are lucky, however, to have a clearly elaborated theoretical and comparative background for our study, namely, Robert Ratcliffe's groundbreaking studies on the broken plural in Semitic (Reference Ratcliffe1998a, Reference Ratcliffe1998b). Of particular value for our investigation are Ratcliffe's instructive overviews of the comparable systems in Soqotri's sister tongues, Harsusi (Reference Ratcliffe1998a: 189–95; Reference Ratcliffe1998b: 93–5) and especially Jibbali (Reference Ratcliffe1998a: 196–201; Reference Ratcliffe1998b: 95–7).

This article is based on an exhaustive perusal of the glossaries of the two volumes of Kogan and Naumkin, Corpus of Soqotri Oral Literature (CSOL I, Reference Naumkin and Kogan2014, and CSOL II, Reference Naumkin and Kogan2018). For most patterns, exhaustive – or nearly exhaustive – lists of examples have been compiled, which makes these findings relevant also from a quantitative point of view: they show how frequent (or, for that matter, regular or productive)Footnote 3 is this or that pattern (or, better, this or that singular/plural correlation).

The phonological system of Soqotri vowels, of crucial relevance for any study in the field of Soqotri morphology, follows Kogan and Naumkin Reference Kogan, Naumkin, Elmaz and Watson2014, where only five vocalic elements are considered phonemic: a, e, o, i, u (the latter of rather limited distinctive load). The distinсtion between a and ε, usually conditioned by the surrounding consonants and, therefore, essentially non-phonemic, will nevertheless be maintained in the transcription.

The main goal of this article is to produce an accurate description of the synchronic facts based on solid lexical evidence. No fully-fledged diachronic assessment of the Soqotri system of broken plural has been attempted. Still, a few basic historical observations will be given where appropriate, both in the main body of the article and its concluding section.

Before proceeding to the evidence, I would like to offer a word of gratitude to our beloved Soqotri informants, ‘Isa Gum‘an al-Da‘rhi and Ahmad ‘Isa al-Da‘rhi, whose extraordinary linguistic skills and exemplary devotion have made this study possible. The article is dedicated to the memory of Robert Ratcliffe, whose revolutionary ideas on the broken plural problem have influenced me so profoundly from my earliest student days up to now.

II. The evidence

a. Singular nouns with e or i in the last syllable (mostly, but not exclusively, quadriconsonantal) typically produce plurals with a-replacementFootnote 4 (o < *ā or *ā́)Footnote 5

There are several groups of examples displaying more complex developments certainly or very probably along the same lines.

  • • Examples where the emergence of o in the second syllable is simultaneous with a vocalic shift in the first syllable (most probably a kind of vocalic accommodation): míṣ̌her, pl. máḳhor “goat pen”, míˀšer, pl. médkor “billy goat”,Footnote 13 ḥ ímhim, pl. ḥ émhom “charcoal”, ḷ éḳham, pl. ḷ áḳhom “kind of tree (Commiphora parvifolia)”Footnote 14, móṣfin, pl. méṣfon “something plaited”, ṣ éľheľ, pl. ṣ óḷhoḷ “a little wadi”, kémbeb, pl. kómbob “remains of dry leaves on a palm trunk”, ḳ eḥ íľh in, pl. ḳ aḥ éḷhon “egg”. No clear-cut rules can be posited for this limited and rather heterogeneous evidence, but the main trend is clear: e and especially i tend to be avoided in the plural form, being replaced by a/ε or o.

  • • Examples where the vowel of the second syllable is a/ε rather than o: f érhim, pl. f έrh εm “girl”, míṭh εn, pl. máṭ an “a shrub”, ˁázgeheľ, pl. ˁázg εh εḷ “penis”. While the attribution of these examples to the present class appears tempting,Footnote 15 the unexpected behaviour of the vowel in the second syllable of the plural form is difficult to explain.

  • • Varia: k εn, pl. k óhon “colour, shape”, méraḥt, pl. méroḥt “eating mat”, šεm, pl. šóhom “name”, tε, pl. t óho “meat”, t εyš, pl. toš “he-goat used as a sire”.Footnote 16

The historical background of this type of broken plural formation has been discussed at some length in Kogan Reference Kogan2015: 476–7, where it is considered one of the hallmarks of the linguistic identity of MSA.Footnote 17 For the comparative evidence from continental MSA see Bittner Reference Bittner1909: 70–71, Rubin Reference Rubin2010: 66–7, Watson Reference Watson2012: 58 (Mehri); Ratcliffe Reference Ratcliffe1998a: 193 (Harsusi); Bittner Reference Bittner1915: 55–6, Ratcliffe Reference Ratcliffe1998a: 200, Rubin Reference Rubin2014: 82–3 (Jibbali).Footnote 18

b. Plurals of quadriconsonantal nouns with a-insertion (the “maktab- > makātib-” type)Footnote 19

  • • The diachronically expected variety of this pattern (examples [1]–[29]) displays o (< *ā or *ā́) breaking the C2C3 consonantal cluster of the singular and has i in the last syllable. Many examples involve mV-prefixed nouns, particularly the me-C1C2iC3-o type. Only 7 out of 30 examples from this subgroup do not display the feminine marker in the singular.

  • • Another prominent type is C1eC1eC2eC3 type (examples [30]–[41]), whose diachronical background remains to be established.Footnote 26 None of the twelve examples involved has the feminine marker.

  • • The remaining examples ([42]–[59]) do not easily lend themselves to a diachronic analysis. Most of them must belong to one of the two types discussed above, but the picture has been obscured by the influence of gutturals and emphatics.

For this type of plural formation as a prominent feature of the “South Semitic” area – comprising Arabic,Footnote 27 Ethiopian SemiticFootnote 28 and MSA (in all probability, also Epigraphic South Arabian)Footnote 29 – see Kogan Reference Kogan2015: 167 and 476; and Bulakh and Kogan Reference Bulakh and Kogan2011: 23–6. For parallel phenomena in continental MSA see Ratcliffe Reference Ratcliffe1998a: 193, (Harsusi); Bittner Reference Bittner1909: 66–9, Jahn Reference Jahn1905: 45– 9 (Mehri); Bittner Reference Bittner1915: 53–5, Ratcliffe Reference Ratcliffe1998a: 199–200, Rubin Reference Rubin2014: 83 (Jibbali).Footnote 30

c. Plurals of the nomen unitatis/nomen collectivum type

  • • The high percentage of examples with (presumably, prosthetic) initial vowel ([1]–[15]) is noteworthy.

As far as one can judge from Ratcliffe Reference Ratcliffe1998a: 193 and 199, this pluralization strategy is uncommon in continental MSA,Footnote 32 or at least much less prominent than in Soqotri. In a broader “South Semitic” context, it is well attested in Arabic (Ratcliffe Reference Ratcliffe1998a: 94; Fischer Reference Fischer1987: 49–50) and Tigre (Ratcliffe Reference Ratcliffe1998a: 180–1; Bulakh and Kogan Reference Bulakh and Kogan2011: 8–9, 15–6).Footnote 33 However, in none of these languages has the singulative/collective opposition been grammaticalized into a fully-fledged means of pluralization: in both Arabic and Tigre, the collective forms exist side-by-side with “true” broken plurals, to which they can be opposed functionally/semantically.Footnote 34 In other words, the Soqotri situation at this point appears to be virtually unique in a comparative (South) Semitic perspective.

d. Patterns with prefixed *ˀV-Footnote 35

  • • Examples [1]–[8] display the pattern eC1C2oC3, which can be confidently traced back to *ˀV-C1C2āC3-, reliably attested in Arabic,Footnote 39 Ethiopian SemiticFootnote 40 and, within MSA, in Jibbali.Footnote 41 Nearly all of them represent geminated consonantal roots (C1C2C2);Footnote 42 the only example from a geminated root that does not take this pattern is [11] ḥor, pl. áˀḥrεr “goat pen”.

  • • Examples [9]–[14] and [15]–[19] belong to the patterns eC1C2εC3 and eC1C2eC3 respectively, both diachronically uncertain.Footnote 43

  • • Examples [20]–[28] cannot be reliably attributed to any of the above groups because the vocalism of the second syllable is obscured by the influence of the gutturals and/or emphatics.

  • • Examples with the feminine ending (Bittner Reference Bittner1918: 64) comprise one group with ε-vocalism ([29]–[34]) and another with e-vocalism ([35]–[37]). In the former the feminine ending is -o, while in the latter it is -e. Footnote 44

  • • In examples [38]–[44] the original vocalism is obscured by the neighbouring gutturals and emphatics; the feminine ending is uniformly -o.

  • • In examples [45]–[52] the vowel of the second syllable has become o because of the vocalic harmony around the last radical guttural.

  • • Examples [53]–[58] are derived from roots with the third weak radical and do not follow any systematic pattern.

As one learns from Ratcliffe Reference Ratcliffe1998a: 196–7, the Soqotri inventory of *ˀV-prefixed broken plurals comes very close to the Jibbali system,Footnote 45 where the patterns VC1C2ɔC3, VC1C2εC3 and VC1C2eC3-t are all common, and contrasts with the Harsusi system, where *ˀV-prefixed patterns do not seem to play any significant role (cf. the summary of the Harsusi system in Ratcliffe Reference Ratcliffe1998a: 195). The same is true of the Mehri system as described in Rubin Reference Rubin2010: 66–8 (but note Bittner Reference Bittner1909: 63–4).

e. Plurals in -ihin, -hin, -in

As argued in Kogan Reference Kogan2015: 474–5, the traditional perception of such forms – both in Soqotri and throughout MSA – as reflecting the “sound masculine plural”Footnote 48 can scarcely be correct. There are two main reasons behind this conclusion.

On the one hand, nouns using this way of pluralization in MSA do not belong to morphological classes which normally use external masculine plural in languages where broken plurals are available – namely, “derived nouns such as participles and verbal nouns” (Ratcliffe Reference Ratcliffe1998a: 219).Footnote 49 Indeed, as the chart above plainly shows, the relevant Soqotri lexemes belong to simple morphological structures (most of them can be reasonably qualified as “primary nouns”) which, both formally and functionally, are totally different from participles of derived stems or nomina agentis of Arabic and Geez, typically pluralized with the external masculine markers -ūna/-īna and -ān respectively.

On the other hand, practically all the Soqotri forms involved display o (presumably < *ā) in the second syllable of the singularFootnote 50 – a feature known to be associated with the C1iC2C3ān- pattern in Arabic (Ratcliffe Reference Ratcliffe1998a: 86).Footnote 51

f. Plurals in -yhon Footnote 52

This small group comprises exclusively kinship terms, probably brought together via semantic attraction.Footnote 53 It is difficult to say why these nouns opted for an n-extended broken plural. Since most of the singulars are (synchronically) biconsonantal, Arabic examples like nār- “fire”, pl. nīrān-, extensively discussed in Ratcliffe Reference Ratcliffe1998a: 85–6, readily come to mind as a possible analogy.Footnote 54

g. The С1éС2(h)oC3 pattern

h. The С1íС2(h)oC3 pattern

The patterns С1éС2(h)oC3 and С1íС2(h)oC3, dealt with in this and the preceding section, are almost certainly traceable to a common prototype.Footnote 57 Since o in the second syllable may go back to either long or stressed short *a, both *C1iC2áC3- and *C1iC2āC3- are theoretically possible. There are two reasons to prefer the second solution, although neither of the two is truly compelling.

  • • The pattern *C1iC2aC3- is likely behind С1i/eС2(h)εC3 in group l below, where it is diachronically expected as it rather regularly corresponds to C1eC2áC3-e/ε structures in the singular. It would be a priori unusual to see the same *C1iC2aC3- prototype reflected as С1é/íС2(h)oC3 when applied to nouns with no feminine ending and as С1i/eС2εC3 when this ending is present: an Arabic-like distribution *C1VC2C3- > *C1iC2āC3- (Ratcliffe Reference Ratcliffe1998a: 77–81) vs. *C1i/uC2C3-at- > *C1i/uC2aC3- (ibid. 88–91) would appear more natural.Footnote 58 This is not necessarily so, however, in view of the Geez correlations such as ˀəgr “foot”, pl. ˀəgar (Ratcliffe Reference Ratcliffe1998a: 166–8).Footnote 59

  • • The “parasitic” h, often featuring before o, is usually considered a safe witness of the etymological length of the vowel following it (LS 22–3; Rhodokanakis Reference Rhodokanakis1915: 16 and elsewhere). Now about two-thirds of the 31 plural forms recorded in groups g and h display a “parasitic” h. These rather indecisive statistics are further marred by the fact that about a half of the С1i/eС2εC3 plurals in group l also reflect this feature.

Within MSA, the *С1VС2oC3 pattern is fairly common also in the continental languages, see Ratcliffe Reference Ratcliffe1998a: 189–90 for Harsusi; Bittner Reference Bittner1909: 55–6, Rubin Reference Rubin2010: 67 for Mehri; Bittner Reference Bittner1915: 50, Ratcliffe Reference Ratcliffe1998a: 196, Rubin Reference Rubin2014: 83 for Jibbali. Their diachronic evaluation is faced with the same difficulties as in Soqotri.Footnote 60

i. The С1éС2oyC3 pattern

The pattern С1áС2oyC3 is an exact formal parallel to the Arabic С1aС2āˀiC3- pattern (Bittner Reference Bittner1918: 65),Footnote 64 and the nature of the singular/plural correlation is exactly the same in both languages: the pattern is favoured by nouns with a long vowel in the second syllable, followed by the feminine marker (Ratcliffe Reference Ratcliffe1998a: 97–9). A comparable phenomenon in Geez is somewhat more remote insofar as the “default consonant” in the second syllable of the plural form is usually w rather than y (Ratcliffe Reference Ratcliffe1998a: 170).Footnote 65 The Epigraphic South Arabian evidence, difficult to interpret precisely because of the purely consonantal script, is discussed in Ratcliffe Reference Ratcliffe1998a: 174–5 and Stein Reference Stein2003: 76–7. Within MSA, a few comparable examples have been registered in Harsusi (Ratcliffe Reference Ratcliffe1998a: 194) and Mehri (Bittner Reference Bittner1909: 65), but, apparently, not in Jibbali.Footnote 66

j. The С1aС2yoC3 pattern

This rather marginal pattern can be plausibly connected with the Mehri pattern C1eC2yōC3, extensively dealt with in Diem Reference Diem1979: 64–75 (see also Bittner Reference Bittner1909: 57–58 and Kogan Reference Kogan2015: 469).

k. The С1éС2eC3 pattern

This rather prominent plural pattern is among the most difficult to evaluate diachronically.

Its main structural feature – breaking the C2C3 consonantal cluster of the singular – brings it close to the well-known *C1iC2āC3- and *C1iC2aC3- structures repeatedly discussed throughout this article, but the origin of the e-vowel in both syllables is still difficult to explain. It is easy to observe that the С1éС2eC3 structure of the plural pattern is exactly the same as that of the 3 sg. m. of the perfect of the inactive type in the basic stem. Now the latter can be safely traced to *C1iC2aC3 in view of the corresponding feminine form C1iC2έC3-o (Naumkin et al. Reference Naumkin, Bulakh, Cherkashin, Kogan, Issa and Gumaan2014: 27–8). One could surmise that С1éС2eC3 is indeed a regular outcome of *C1iC2aC3, presumably stressed on the first syllable and leaving unstressed the a-vowel of the second syllable. Within this approach, the Soqotri С1eС2eC3 and С1éС2(h)oC3/С1íС2(h)oC3 patterns could be safely traced back to *C1iC2aC3- and *C1iC2āC3- respectively. We are left, however, with the С1i/eС2(h)εC3 pattern discussed in the following section l: for all practical purposes, *C1iC2aC3 is the most expected and legitimate prototype of this sequence, and it is hard to see why it should have given two competing outcomes – С1éС2eC3 and С1í/éС2(h)εC3 – within the system of the Soqotri broken plural.

l. The С1í/éС2(h)εC3 pattern

  • • In nine out of 16 examples, the correlation C1éC2C3-e/ε Footnote 68 > С1í/éС2(h)εC3 is observed, clearly identical to Arabic C1i/uC2C3-at- > C1i/uC2aC3- (Ratcliffe Reference Ratcliffe1998a: 88–90). This way of plural formation is well attested also in continental MSA: Harsusi (Ratcliffe Reference Ratcliffe1998a: 191–2), Mehri (Bittner Reference Bittner1909: 52–3, Rubin Reference Rubin2010: 67–8), Jibbali (Bittner Reference Bittner1915: 47–9, Ratcliffe Reference Ratcliffe1998a: 199, Rubin Reference Rubin2014: 83–4). Furthermore, in all continental languages a special “mixed” type of plural is common, consisting of the a-inserted base and the external feminine plural marker.Footnote 69 Outside Arabic and MSA, this correlation is very prominent in Tigre (Ratcliffe Reference Ratcliffe1998a: 180–2), where, however, Arabic influence cannot be ruled out (Bulakh and Kogan Reference Bulakh and Kogan2011: 15–21, Reference Bulakh and Kogan2013: 83). For the limited, but relatively certain ESA evidence see Ratcliffe Reference Ratcliffe1998a: 173–4, Stein Reference Stein2003: 75.

  • • In two cases ([3] ḥiṣáľe and [15] žirέme), the singular displays a vowel after the second radical, which does not fit the above correlation, where ε in the plural breaks the C2C3 cluster in the singular. One wonders, therefore, whether these two examples might be alternatively ascribed to group c (the nomen collectivum/nomen unitatis type).

  • • The remaining five cases involve nouns with no feminine marker. Their presence in this group is, at present, difficult to explain.

m. The С1íy(h)εC3 pattern

This group is a transparent sub-type of the correlation described under l, with -y- featuring as the default second consonants for biconsonantal singulars with i < *ī. Footnote 70 Similarly structured pairs are common elsewhere in languages where the *C1VC2C3-at- > *C1VC2aC3- correlation is attested (see Ratcliffe Reference Ratcliffe1998a: 88 for Arabic, p. 180 for Tigre).

n. The С1íy(h)oC3 patternFootnote 71

С1íy(h)oC3 is a subtype of the С1éС2(h)oC3/С1íС2(h)oC3 pattern (sections g and h), with y functioning as default second consonant for biconsonantal singulars. Comparison with the similarly structured correlation in the preceding section m (also six examples) is instructive: five out of six examples in m have an overt i in the singular, but none in n, where e predominates; all examples in m, but only a half in n, display the feminine marker.

o. The “mixed type”

Soqotri does not productively use the “mixed” type of plural in which internal modifications of the stem are accompanied by the addition of the ending *-tVn, otherwise characteristic of external feminine plural. This way of plural formation, very common in Jibbali (Bittner Reference Bittner1915: 52–3, Ratcliffe Reference Ratcliffe1998a: 198–9),Footnote 72 is, however, not totally alien to Soqotri, as seen already by Bittner (Reference Bittner1918: 64) and confirmed by modern research. Such examples, with the stem structure C1eC2oC3-, are limited to a handful of – apparently archaic – examples listed above.Footnote 73

p. A few unique, atypical broken plural formations are summarized here

III. The broken plural of Soqotri in a diachronic perspective

In his discussion of the relevance of the broken plural problem for genealogical classification of Semitic (Kogan Reference Kogan2015: 166–71), the present author has limited the scope of his analysis to Arabic and Geez, usually excluding the MSA facts from consideration because of the heavily understudied historical phonology and morphology of these languages. The Soqotri evidence analysed in this article, aiming to develop and supplement Ratcliffe's pioneering insights into the Harsusi and Jibbali systems of broken plural, allows a preliminary assessment of the place of the Soqotri system both within and outside MSA, thus bringing somewhat greater clarity into the thorny field of Modern South Arabian linguistic diachrony.

In Ratcliffe Reference Ratcliffe1998a: 206, four fundamental features shared by most languages of the “South Semitic” areaFootnote 79 in the domain of the broken plural formation are listed. Soqotri participates prominently in each of them.

  • • The *C1VC2C3-at- > *C1VC2aC3- correlationFootnote 80 = groups l and m.

  • • Plurals with *ˀa- = group d.

  • • The pattern *C1aC2āC3iC4- for quadriconsonantal nouns = group b.Footnote 81

  • • The pattern *C1aC2āy/wiC3- for triconsonantal nouns with a long vowel in the second syllable = group i.

Further commonalities between Soqotri and the languages of the “South Semitic” area are in evidence.

  • • The singulative-collective opposition = group c.Footnote 82

  • • The *C1VC2āC3- > *C1iC2C3ān- correlation = group e. Ratcliffe, adhering to the traditional interpretation of the MSA n-plurals as external masculine, was unable to observe their similarity to the Arabic facts and listed the latter among “features of the Arabic system which do not find parallels in other southern languages” (Reference Ratcliffe1998a: 210). The new evaluation of the MSA correlation allows us to broaden seriously the diachronic importance of this feature, thus counterbalancing the rather indecisive Ethiopian evidence.Footnote 83

  • • The pattern *C1iC2āC3- = groups g, h, n, is very prominent in Arabic (Ratcliffe Reference Ratcliffe1998a: 77), but not attested in Geez. Its presence in Tigre may be attributed to Arabic influence (Bulakh and Kogan Reference Bulakh and Kogan2011: 21–3, Reference Bulakh and Kogan2013: 83–4).

  • • Soqotri shares with other MSA languages one of the two singular-plural correlations which Ratcliffe (Reference Ratcliffe1998a: 210) defines as specific innovations of this subgroup, namely the correlation in which “an (etymologically) long high vowel … in the last syllable of the singular alternates with a long non-high vowel … in the last syllable of the plural” = group a. It does not actively participate, however, in the second of these features, where essentially the same shift affects biconsonantal structures as well.Footnote 84

  • • By far the most remarkable feature of broken plural formation which Soqotri does not obviously share with most other languages of the “South Semitic” group is the *u/ū insertion (Ratcliffe Reference Ratcliffe1998a: 207), prominently attested in Arabic (ibid. 77, 105) and on good grounds reconstructible for Geez (ibid. 167–8) and Ethiopian Semitic in general. There is no need to stress that the validity of this observation is seriously relativized by our inability safely to establish the Soqotri reflexes of proto-Semitic u-vowels.Footnote 85 In such a context, a closer look at the continental MSA evidence is mandatory.

The clearest possible parallel to the Arabic C1uC2ūC3- pattern is C1eC2ewweC3 in Harsusi, which is quite a prominent means of plural formation in this language (Ratcliffe Reference Ratcliffe1998a: 190). The possibility of Arabic influence, apparently hinted at by Ratcliffe (ibid.), can hardly be discarded in this case, particularly since there is no trace of such formation in Jibbali, by far the most conservative among the continental MSA, and even in Mehri.Footnote 86

In Ratcliffe Reference Ratcliffe1998a: 197, the author somewhat obliquely identifies the widespread Jibbali pattern εC1C2eC3(V)t with *ˀa-C1C2uC3-at-, most probably underlying the *ˀaC1C23C3-t pattern in Geez. As we have seen above in connection with group d, this may well be the case also for similarly structured Soqotri pattern(s), but there can hardly be any positive proof for such a conclusion before the fate of *u in both languages is elucidated.Footnote 87

All in all, Ratcliffe's conclusion (Reference Ratcliffe1998a: 207), according to which “both MSA languages have plurals that could plausibly be interpreted as reflecting CuCuuC and ˀaCCuuC” appears rather premature. This leaves enough room for an alternative hypothesis, according to which the u-isogloss has never reached the proto-MSA system of broken plural.

Footnotes

To the memory of Robert Ratcliffe

The article has been written in the framework of the project 34.5109.217 supported by the Ministry of Science and Education of Russia. My sincere gratitude goes to Dr Maria Bulakh for her valuable remarks on this article and, more broadly, for numerous discussions on Soqotri grammar and lexicon.

2 Such as Johnstone Reference Johnstone1975 and Simeone-Senelle Reference Simeone-Senelle and Hetzron1997; Reference Simeone-Senelle and Weninger2011. Recent surveys on comparable phenomena in continental MSA can be found in Rubin Reference Rubin2010: 66–8, Watson Reference Watson2012: 58–60 (Mehri) and Rubin Reference Rubin2014: 82–4 (Jibbali), none aiming at exhaustiveness. A more detailed – even if, in many respects, undoubtedly outdated – treatment of the continental evidence can be found in Bittner Reference Bittner1909: 50–76 (Mehri) and Reference Bittner1915: 47–56 (Jibbali). For Mehri see further Jahn Reference Jahn1905: 35–69.

3 For these notions as applied to our problem see Ratcliffe's important observations (Reference Ratcliffe1998a: 68).

4 This essential concept goes back to Greenberg's classic article (Reference Greenberg and Lukas1955).

5 Bittner Reference Bittner1918: 66. For the time being, we will take for granted the widespread – and, in essence, undoubtedly correct – assumption that long *ā and stressed short *a yield o in Soqotri and MSA in general (Ratcliffe Reference Ratcliffe1998a: 189, 196 and elsewhere). Further remarks on this shift will be made throughout this article, but a final assessment cannot be possible without an eventual comprehensive inquiry into the historical phonology of MSA.

6 Here and below, scientific names of plants are taken from Miller and Morris Reference Miller and Morris2004.

7 On the two l-phonemes of Soqotri and their alternations see, provisionally, CSOL I, pp. 12–13, Naumkin et al. Reference Naumkin, Bulakh, Cherkashin, Kogan, Issa and Gumaan2014: 29.

8 In this example, as well as in [14] and [40], a instead of o in the second syllable is due to the secondary influence of the neighbouring consonants.

9 Or rather “the hard disc at the flower end of a date”. Here and elsewhere below, alternative definitions proposed by Miranda Morris are adduced in footnotes. For these friendly suggestions I am deeply grateful to Dr Morris, a brilliant expert on Soqotri realia (and a competent native speaker of English). The background of the – mostly subtle – discrepancies between these definitions and those I was able to obtain from my Soqotri informants remains to be investigated.

10 “A skin or hide which has had the skin removed but has not yet been tanned, cut and stitched to make a bag or a container.”

11 “Scoop, ladle.”

12 “Any piece of any material used by someone as a wrap or covering.”

13 For palatalizing alternations k > š and ḳ > ṣ̌ in Soqotri nouns v. CSOL I, pp. 16–17.

14 Miller and Morris Reference Miller and Morris2004: 468.

15 As seen already by Bittner (Reference Bittner1918: 66) in connection with férhim “girl”, pl. fέrhεm.

16 Cf. Bittner Reference Bittner1918: 64.

17 See also Johnstone Reference Johnstone1975: 21, Ratcliffe Reference Ratcliffe1998a: 210 and, in nuce, Bittner Reference Bittner1918: 66.

18 In Mehri, the vowel of the second syllable is long in both singular and plural, but at least in some cases – such as kǝbkēb “star”, pl. kǝbkōb (ML 201) – the length is obviously secondary (< PS *kabkab-). It is, therefore, very difficult to say whether o in Soqotri goes back to a long or short a in this case. Total absence of “parasitic h” from the relevant examples would speak for a short vowel in the prototype, but this criterion seems to be less reliable than commonly believed (see Section h below).

19 Bittner Reference Bittner1918: 65.

20 “Pouch made by looping the material over and above the belt of the loincloth (used for carrying smaller items).”

21 “Roofing extended beyond the walls of a building to provide shelter.”

22 This example is unique insofar as there is no consonantal cluster in the singular.

23 “Grazing reserve (especially for goats); wild area which only goats spend time in.”

24 “Noose placed on the ground to catch the foot of an animal.”

25 “Ledge on a mountainside under which is an inhabitable cave and above which are further caves; stepped terrace-like terrain.”

26 Cf. Bittner Reference Bittner1909: 65: “Das â dieser formen erscheint im Soqoṭri als â, häufiger als ô … aber auch als ê”.

27 Ratcliffe Reference Ratcliffe1998a: 95–6.

28 Ratcliffe Reference Ratcliffe1998a: 169–70 (Geez), 182–3 (Tigre).

29 Stein Reference Stein2003: 78–9.

30 A characteristic feature of this pattern of pluralization in Jibbali is the very frequent appearance of b < *w before the third consonant (Rubin Reference Rubin2014: 83). The historical background of this phenomenon, also attested in Mehri (Rubin Reference Rubin2010: 67), remains to be elucidated.

31 “New shoot on a branch; side stem, side branch.”

32 A few remarkable examples from Jibbali can be found in Bittner Reference Bittner1915: 49. Cf. also Ratcliffe Reference Ratcliffe1998a: 193 for Harsusi.

33 But not in Geez (Ratcliffe Reference Ratcliffe1998a: 169). As pointed out by Ratcliffe (Reference Ratcliffe1998a: 181), the actual number of clear-cut examples of the singulative/collective opposition in Tigre is quite low (only eight examples in Palmer Reference Palmer1962: 52). See further fn. 82.

34 Ratcliffe Reference Ratcliffe1998a: 94 (Arabic) and 182 (Tigre).

35 For a penetrating early overview see Bittner Reference Bittner1918: 63.

36 “Overhang, shallow cave, often running along a ledge, not high in a mountainside but near flat ground.”

37 “Area of rolling hills, plain foothills which are bare of vegetation.”

38 Wranik Reference Wranik2003: 506.

39 Ratcliffe Reference Ratcliffe1998a: 77–81.

40 Ratcliffe Reference Ratcliffe1998a: 166–9 (Geez), Reference Ratcliffe1998a: 177 (Tigre). See further Bulakh and Kogan Reference Bulakh and Kogan2013: 78–9.

41 Ratcliffe Reference Ratcliffe1998a: 196–7.

42 A few non-geminated Jibbali examples are listed in Ratcliffe Reference Ratcliffe1998a: 196–7. However, the prominent position of geminated roots in the cognate Jibbali plural pattern has been noted by Ratcliffe (Reference Ratcliffe1998a: 233), who explains it by reduplication of the second root consonant in (originally) biconsonantal nouns. One is tempted to ascertain whether some true geminated roots are also involved, but, at least for Soqotri, etymology is of little help here, as nearly all relevant lexemes are diachronically opaque.

43 In a comparative perspective, *ˀV-C1C2āC3- and *ˀV-C1C2uC3- are the only options available. Within the first possibility, one has to postulate a (hitherto unexplained) split of *ˀV-C1C2āC3- into eC1C2oC3 vs. eC1C2εC3/eC1C2eC3, with a further, similarly unexplained, difference between ε- and e-examples. The second alternative implies that *u can be reflected as ε and/or e, which remains to be co-ordinated with further data on the development of *u in Soqotri.

44 The vocalic difference in the second syllable is perplexing, but the different nature of the feminine endings – a stressed *-át > -o and an unstressed *-at > -e – readily suggests itself as a possible solution, to be developed within a more comprehensive inquiry into the history of MSA vocalism. From the comparative point of view, there are only two plural patterns with prefixed *ˀa- and a feminine ending, viz. *ˀa-C1C2iC3-at- and *ˀa-C1C2uC3-at-. As rightly seen by Ratcliffe (Reference Ratcliffe1998b: 90), the former option is scarcely feasible here: the Arabic pattern ˀa-C1C2iC3-at- is a minor feature strictly associated with bisyllabic singular bases with a long ā in the second syllable. We are thus left with *ˀa-C1C2uC3-at-, presumably underlying the ˀa-C1C2ǝC3-t in Ethiopian (Ratcliffe Reference Ratcliffe1998a: 166–8; Bulakh and Kogan Reference Bulakh and Kogan2013: 78–9), which of necessity presupposes that *u can yield ε and/or e in Soqotri (cf. fn. 85).

45 So already Bittner Reference Bittner1918: 63. A good collection of Jibbali examples can be found in Bittner Reference Bittner1915: 50–52.

46 “Plaited cow-leather support used when climbing date palms.”

47 Miller and Morris Reference Miller and Morris2003: 687.

48 Bittner Reference Bittner1909: 46–9, Reference Bittner1915: 41, Ratcliffe Reference Ratcliffe1998a: 195, 203, Rubin Reference Rubin2010: 64, Reference Rubin2014: 79–80.

49 Cf. Bittner Reference Bittner1909: 46.

50 Exactly the same as in Mehri (Bittner Reference Bittner1909: 47, Rubin Reference Rubin2010: 64) and Jibbali (Bittner Reference Bittner1915: 41, Rubin Reference Rubin2014: 79–80).

51 Cf. γazāl- > γizlān- “gazelle”, γulām- > γilmān- “boy” and many similar examples. One has to admit that the i-vowel in the Soqotri plural forms does not match, at first sight, the expected *ā in the protoform. It should be kept in mind, however, that our knowledge of Soqotri historical phonology is too limited for such a negative conclusion. Compare, for example, such verbal nouns as ˁaríbhin “knowledge” or nikírhin “nostalgia”, where -hin is very likely going back to *-ān, or the primary noun líšin “tongue”, almost inevitably traceable to *lišān-.

52 There remains, admittedly, a small group of plurals with the -hVn ending which cannot be attributed to any of the two classes e and f: ḥánžεhεr “knife” > ḥanžérhon, ˁésrε “thicket, grove” > ˁisárhen, ˁábre “generation” > ˁibárhen. Their historical background remains to be elucidated.

53 Note the reduplication in example [5].

54 Cf. Ratcliffe Reference Ratcliffe1998a: 240: “We choose to interpret them [the C1īC2ān- plurals] as innovations particular to Arabic, although the limited productivity of the forms would admittedly be more consistent with an interpretation of them as old developments”. Mehri examples like gōr “slave”, pl. gīrōn and ẑayf “guest”, pl. ẑīfōn (Rubin Reference Rubin2010: 65) may also be relevant for comparison.

55 For the phone ö and its (almost completely) allophonic status with respect to e, see CSOL I, pp. 17–8.

56 Miller and Morris Reference Miller and Morris2004: 671.

57 I am unable, at present, to explain the vocalic discrepancy in the first syllable. Different consonantal surrounding may have played an important role (thus, there are eight examples beginning with ḥ in h, but none in g, as against seven examples beginning with velars in g, and none in h), but no clear-cut distributional rules are in evidence.

58 Thus, it is not without significance that almost no examples in groups g and h display a feminine ending. The exceptions are [2] bíľe and [16] ṭéľfε in group h and [13] ŝenége and [14] tenéže in group g. This is compatible with Ratcliffe's observation about С1iС2āC3- as a “minor” option for C1aC2C3-at- and C1aC2aC3-at- singulars in Arabic (Reference Ratcliffe1998a: 90). Furthermore, as far as the latter two examples are concerned, an alternative attribution to group c (singulative/collective) is also to be considered.

59 Vestigially, also in Tigre (Ratcliffe Reference Ratcliffe1998a: 177; Bulakh and Kogan Reference Bulakh and Kogan2013: 77–8).

60 Bittner (Reference Bittner1918: 63–4) opts for *С1iС2áC3- in Soqotri. Ratcliffe (Reference Ratcliffe1998a: 190) directly compares the Harsusi forms with Arabic *С1iС2āC3-, but is more cautious in his description of the Jibbali facts (ibid. 196), admitting a short stressed a as an alternative possibility.

61 “A ‘Muslim’ spirit that lives underground: harmless, and indeed often helpful and protective”.

62 “Large hoop ear-ring worn in mid earlobe”.

63 Miller and Morris Reference Miller and Morris2004: 567.

64 The synchronic presence of ˀ (rather than y) in Arabic is secondary from the historical point of view (Diem Reference Diem1980: 93).

65 As opposed to Tigre, which displays -y- as in Arabic and Soqotri (Ratcliffe Reference Ratcliffe1998a: 184). See further Bulakh and Kogan Reference Bulakh and Kogan2013: 80–81.

66 As cautiously observed already by Bittner (Reference Bittner1915: 53). In Ratcliffe Reference Ratcliffe1998a: 236 this type is explicitly qualified as “lost” in Jibbali. Ratcliffe (Reference Ratcliffe1998b: 97) explains this by the loss of contrastive vowel length in Jibbali, but this argument would not work for Soqotri where vocalic length is also non-phonemic. Also in Mehri the functional load of the vocalic length appears to be extremely low (J. Watson in Sima Reference Sima2009: 22).

67 “Flesh of the upper ribs and flank (in a carcass)”.

68 With e > a in the neighbourhood of gutturals and emphatics. The apparently unmotivated distribution of -e and -ε as allomorphs of the feminine marker remains unexplained.

69 Less prominent, but also present in Soqotri (see Section o below).

70 The only exception where the i-vowel does not surface in the singular is [6] sóˁo.

71 The common word ˁag(g) “man, husband” (< *ˁayg) belongs here with its plural ˁeyyúg, in spite of the atypical stress on the last syllable.

72 But, it seems, rare in Mehri (see Rubin's remarks in Reference Rubin2010: 65): ˀayn “eye”, pl. ˀāyε̄ntən, ḥəyḏēn “ear”, pl. ḥəyḏantən, kəwbēt “bitch”, pl. kəlabtən, ẑəfīr “plait of hair”, pl. ẑəfartən. For more examples see Bittner Reference Bittner1909: 60–1.

73 Curiously enough, nouns with n as the last consonant are involved in each case. The o vowel in these examples is phonetically realized as semi-open [ɔ].

74 Probably belongs to С1eС2eC3 (type k), the vowel of the first syllable being syncopated between two voiceless consonants.

75 Bittner Reference Bittner1918: 66 (“eigentümlich”).

76 The peculiar nature of this example has been noted already by Bittner (Reference Bittner1918: 62).

77 In essence, this is certainly a representative of type b. The presence of ou (= u) in the first syllable is very unusual for a primary noun in Soqotri, but an eventual -w- in the prototype is likely (*sawˁid-at-). If so, the present example comes close to such correlations as Geez kokab “star”, pl. kawākəbt (Ratcliffe Reference Ratcliffe1998a: 170).

78 “The flesh of the lower ribs, the flap of flesh at the base of the ribcage (in a carcass)”.

79 That is, Arabic, Epigraphic South Arabian, Ethiopian Semitic, and Modern South Arabian. The diachronic background of this unity – genealogical relationship or language contact – cannot be discussed in the framework of this contribution. For the time being, see Kogan Reference Kogan2015: 599.

80 Ratcliffe is correct to state that the importance of this feature goes beyond the “South Semitic” area and affects West Semitic as a whole. What he means are the North-West Semitic examples like Hebrew malkā “queen”, pl. məlākōt (Reference Ratcliffe1998a: 154). He is also right to observe that this feature is missing from Geez (on p. 236 of his study he plausibly explains it as a secondary loss due to generalization of the feminine marker -t at the expense of -at). Its widespread presence in Tigre is, therefore, contradictory from a diachronic point of view (Bulakh and Kogan Reference Bulakh and Kogan2011: 15–21, Reference Bulakh and Kogan2013: 83): either a deeply rooted archaism, or a result of a very close and prolonged interaction with Arabic. For important remarks on this correlation see further Ratcliffe Reference Ratcliffe1998a: 207–8.

81 In both Arabic and Geez, the “basic” *C1aC2āC3iC4- shape of this pattern can be modified. In Arabic, the vocalic length attraction is at work: when the singular has a long vowel in the second syllable, the pattern of the plural is *C1aC2āC3īC4- (Ratcliffe Reference Ratcliffe1998a: 95). In Geez we are faced with the polarity rule: “singulars without -t usually have a -t suffix in the plural; singulars with -t generally lose it in the plural” (Ratcliffe Reference Ratcliffe1998a: 169, Bulakh and Kogan Reference Bulakh and Kogan2013: 80–81). A Geez-like opposition must have existed also in Epigraphic South Arabian (Ratcliffe Reference Ratcliffe1998a: 174, Stein Reference Stein2003: 78). Neither of the two phenomena is attested in Soqotri, but one may wonder whether the unexplainable difference between the two common allomorphs C1eC2oC3iC4 and C1eC2eC3eC4 may have something to do with the presence vs. absence of the feminine marker in the singular. One can surmise, for example, that the former pattern, primarily associated with tri-syllabic nouns with the feminine marker in the singular, goes back to a longer prototype *C1aC2āC3īC4-, whereas the latter, exclusively associated with bi-syllabic nouns with no feminine marker, goes back to a shorter prototype *C1aC2āC3iC4-.

82 In Ratcliffe's view (Reference Ratcliffe1998a: 207), “the collective type with singulative formed by -at affixation derives from Proto-Semitic. It is attested in Arabic, possibly in the MSA languages, but also in Akkadian. In NW Semitic the singulative with -at has become a feminine singular and the collective acquires a pleonastic masculine plural suffix”. Ratcliffe's claim for the original presence of this opposition in Hebrew and Aramaic (see Reference Ratcliffe1998a: 157–8 for more details) is quite sound. Of particular interest here are examples – not mentioned by Ratcliffe – such as Hebrew boṭn-īm “terebinths”, ṣäˀä̆l-īm “lotuses”, šiḳm-īm “sycamores”, where the (obviously, pleonastic) external plural ending -īm is attached directly to the stem without a-intercalation normal for the plural forms of Hebrew nomina segolata. In most such cases the singulatives *boṭn-ā, ṣäˀä̆l-ā, šiḳm-ā are, unfortunately, not preserved in the Bible, but note a directly attested opposition in pišt-ā, pl. pišt-īm “flax, linen” (BDB 833–4). For the problem see further Blau Reference Blau1998: 191 and Steiner Reference Steiner2003: 63, neither of whom mention the singulative/collective opposition. The Akkadian evidence is much more opaque, however. On p. 153 of his study, Ratcliffe states that “Akkadian has, like Arabic … a singulative derived from a collective by attachment of -t suffix”. He does not provide any Akkadian examples and refers to GAG § 60a, where we learn that “nomina unitatis … sind selten”. Indeed, none of the examples provided by von Soden look particularly impressive. A special study pending, it seems wise to restrict this isogloss to West Semitic languages.

83 Geez has no means of broken plural formation for triconsonantal nouns with a long ā in the second syllable, which normally take the external feminine marker -āt (Ratcliffe Reference Ratcliffe1998a: 171). At this point, it differs sharply from Tigre, which often uses an Arabic-like patter ˀa-C1əC2C3-at in such cases (Ratcliffe Reference Ratcliffe1998b: 93). See further Bulakh and Kogan Reference Bulakh and Kogan2011: 11–5, Reference Bulakh and Kogan2013: 82–3, particularly on possible Arabic influence on Tigre in this case.

84 What is meant are examples such as Mehri ḥə-d īd “paternal uncle”, pl. ḥə-d ōd (Bittner Reference Bittner1909: 72, Rubin Reference Rubin2010: 67). Rubin restricts this plural formation to “several masculine kinship terms”, but this is not quite so, cf. ḥə-n īd, pl. ḥə-n ōd “water-skin” (ML 306). The Jibbali picture appears more complicated: while in some examples the *i > *a apophony is clear (nid > nud “water-skin”), other cases (such as but > bεt “house” or ḳud > ḳεd “rope”, and more examples in Ratcliffe Reference Ratcliffe1998a: 196) may involve quite different processes. As far as I can see, the only residual example of this type of plural formation in Soqotri is the, now obsolete, word ḥ el'- (pl. ḥ oḷ-) “maternal uncle” (LS 166, only remotely known to our informants). Bittner (Reference Bittner1918: 66) may be correct to attribute to this type also dídhe (pl. dódhi) “demon(s)”, presumably “my paternal uncle(s)” (LS 124).

85 The synchronically attested u-vowel in Soqotri is a rare phoneme whose attestations, apart from a few scattered nominal and adjectival lexemes, are mostly restricted to two morphological positions: the diminutive pattern C1uC2aC3 and -C1uC2aC3- as the base of the imperfect passive of the basic and causative stems (Kogan and Naumkin Reference Kogan, Naumkin, Elmaz and Watson2014: 70–71). In both cases, there are good reasons to consider u to be etymological (see, provisionally, Kogan Reference Kogan2015: 54–7 and 477–8 respectively), but this does not mean for certain that all etymolgical u-vowels are reflected as u in Soqotri.

86 Thus, practically all hypothetical *C1uC2ūC3- examples listed in Bittner Reference Bittner1909: 54 involve transparent Arabic loanwords, evidently borrowed together with their plurals.

87 Note, furthermore, that on pp. 208 and 236 of his study Ratcliffe apparently identifies εC1C2eC3(V)t with *ˀa-C1C2iC3-at- in Arabic and Tigre (cf. fn. 44).

References

BDB = Brown, F., Driver, S.R. and Briggs, Ch.A.. 1906. A Hebrew and Aramaic Lexicon of the Old Testament. Oxford.Google Scholar
Bittner, Maximilian. 1909. Studien zur Laut- und Formenlehre der Mehri-Sprache in Südarabien, I. Vienna: Hölder.Google Scholar
Bittner, Maximilian. 1915. Studien zur Šḫauri-Sprache. I. Vienna: Hölder.Google Scholar
Bittner, Maximilian. 1918. “Charakteristik der Sprache der Insel Soqoṭra”, Anzeiger der pilosophisch-historischen Klasse der kais. Akademie der Wissenschaften in Wien 55, 48–83.Google Scholar
Blau, Joshua. 1998. Topics in Hebrew and Semitic Linguistics. Jerusalem.Google Scholar
Bulakh, Maria and Kogan, Leonid. 2011. “Arabic influences on Tigre: a preliminary evaluation”, BSOAS 74/1, 1‒39.10.1017/S0041977X10000698CrossRefGoogle Scholar
Bulakh, Marie and Kogan, Leonid. 2013. “Эфиосемитские языки”, Языки мира. Семитские языки. Эфиосемитские языки. Moscow, pp. 13–141.Google Scholar
СSOL I: = Naumkin, Vitaly, Kogan, Leonid et al. 2014. Corpus of Soqoti Oral Literature. Volume One. Leiden: Brill.Google Scholar
СSOL II = Naumkin, Vitaly, Kogan, Leonid et al. 2018. Corpus of Soqoti Oral Literature. Volume Two. Leiden: Brill.10.1163/9789004376205CrossRefGoogle Scholar
Diem, Werner. 1979. “Studien zur Frage des Substrats im Arabischen”, Der Islam 56, 12‒80.10.1515/islm.1979.56.1.12CrossRefGoogle Scholar
Diem, Werner. 1980. “Untersuchungen zur frühen Geschichte der arabischen Orthographie. II. Die Schreibung der Konsonanten”, Or 49, 67‒106.Google Scholar
Fischer, Wolfdietrich. 1987. Grammatik des klassischen Arabisch. Wiesbaden: Harrassowitz.Google Scholar
GAG = von Soden, W.. 1995. Grundriss der akkadischen Grammatik. Rome: EPIB.Google Scholar
Greenberg, Joseph H. 1955. “Internal a-plurals in Afroasiatic (Hamito-Semitic)”, in Lukas, Johannes (ed.), Afrikanistische Studien. Berlin: Akademie Verlag, 198–204.Google Scholar
Jahn, Alfred. 1905. Grammatik der Mehri-Sprache in Südarabien. Vienna: Hoelder.Google Scholar
Johnstone, T.M. 1975. “The Modern South Arabian languages”, Afroasiatic Linguistics 1/5, 93‒121.Google Scholar
Kogan, Leonid. 2015. Genealogical Classification of Semitic. The Lexical Isoglosses. Berlin: De Gruyter.10.1515/9781614515494CrossRefGoogle Scholar
Kogan, Leonid and Naumkin, Vitaly. 2014. “The vowels of Soqotri as a phonemic system”, in Elmaz, Orhan and Watson, Janet C.E. (eds), Languages of Southern Arabia. Oxford: Archaeopress, pp. 57–80.Google Scholar
LS = Leslau, Wolf. 1938. Lexique Soqoṭri (Sudarabique moderne) avec comparaisons et explications etymologiques. Paris: Librairie C. Klincksieck.Google Scholar
Miller, Anthony G. and Morris, Miranda. 2004. Ethnoflora of the Soqotra Archipelago. Edinburgh: Edinburgh Royal Botanic Garden.Google Scholar
ML = Johnstone, T.M.. 1987. Mehri Lexicon and English–Mehri Word-List. London: SOAS.Google Scholar
Naumkin, Vitaly, Bulakh, Maria, Cherkashin, D., Kogan, L., Issa, A. and Gumaan, I.. 2014. “Studies in the verbal morphology of Soqotri I/1. Strong triconsonantal roots in the basic stem (the analysis)”, Zeitschrift für Arabische Linguistik 59, 25–56.Google Scholar
Palmer, F.R. 1962. The Morphology of the Tigre Noun. London: Routledge.Google Scholar
Ratcliffe, R. 1998a. The “Broken” Plural Problem in Arabic and Comparative Semitic. Amsterdam and Philadelphia: Benjamins.10.1075/cilt.168CrossRefGoogle Scholar
Ratcliffe, R. 1998b. “Defining morphological isoglosses: the ‘broken’ plural and Semitic subclassification”, Journal of Near Eastern Studies 57, 81‒123.10.1086/468623CrossRefGoogle Scholar
Rhodokanakis, N. 1915. Der zweigipflige Akzent in Minao-Sabaischen. Vienna.Google Scholar
Rubin, Aaron. 2010. The Mehri Language of Oman. Leiden: Brill.10.1163/9789004187627CrossRefGoogle Scholar
Rubin, Aaron. 2014. The Jibbali (Shaḥri) Language of Oman. Leiden: Brill.10.1163/9789004262850CrossRefGoogle Scholar
Sima, A. 2009. Mehri-Texte aus der jemenitischen Šarqīyah. Wiesbaden: Harrassowitz.Google Scholar
Simeone-Senelle, Marie-Claude. 1997. “The Modern South Arabian languages”, in Hetzron, Robert (ed.), The Semitic Languages. London and New York: Routledge, pp. 378‒423.Google Scholar
Simeone-Senelle, Marie-Claude. 2011. “Modern South Arabian”, in Weninger, Stefan (ed.), The Semitic Languages: An International Handbook. Berlin: de Gruyter, pp. 1073–1113.Google Scholar
Stein, Peter. 2003. Untersuchungen zur Phonologie und Morphologie des Sabaischen. Rahden: Leidorf.Google Scholar
Steiner, Richard C. 2003. Stockmen from Tekoa, Sycomores from Sheba: A Study of Amos’ Occupations. Washington, DC: The Catholic Biblical Association.Google Scholar
Watson, Janet C.E. 2012. The Structure of Mehri. Wiesbaden: Harrassowitz.Google Scholar
Wranik, W. 2003. Fauna of the Socotran Archipelago. Field Guide. Rostock: University of Rostock.Google Scholar