While there is no doubt that ẓ̂áˀteʰ ultimately goes back to PS *ŝalāṯ-at- (contra Jahn 1905:74), the exact path of its development is uncertain (Bittner 1913a:85–88), not the least because of the numerous, at times highly peculiar variants attested in both Soqotri and other MSA.

● Western (Qishn) Mehri ŝaγatīt elicited by A. Jahn (1905:74).

● W. Hein’s Western Mehri texts feature a number of other variants (Rubin 2018:296), among which ẑaγtīt is the most remarkable one for the purpose of our survey.

● Eastern (Dhofari) Mehri ŝāṯáyt. Since post-consonantal ˁ is regularly lost in this Mehri variety, the Central (Eastern Yemeni) Mehri ŝaˁṯáyt is to be considered the direct avatar of the Dhofari ŝāṯáyt. Conversely, ŝāṯáyt is unlikely to go back to a prototype with *-l- as this would have yielded -w- in Eastern Mehri (Rubin 2018:296).

● Johnstone’s Jibbali form is ŝɔṯét. Since neither ˁ nor γ are lost in Jibbali, Rubin (2014:36) is prima facie correct to attribute it to the loss of pre-consonantal l (cf. Bittner 1916:50), well attested in Jibbali (but cf. Rubin 2018:296).

● In Morris–Watson 2019:71, the Jibbali forms are given as ŝaˁṯit and ŝaγṯit.

● The Harsusi form elicited by T.M. Johnstone (HL 120) and D. Eades (apud Morris–Watson 2019:35) is ŝāfayt, with an unusual -f-. For Mehri forms with -f- see Bittner 1913a:86.

● A. Nakano’s Hobyot form is ŝaˀṯét (2013:265), but cf. ŝaˁṯēt and ŝāṯēt in Morris–Watson 2019:118.

● Throughout the Vienna coprus, the Soqotri form is consistently spelled with ẓ̂- (as in our corpus), but with -d- instead of -ˀt- (ẓ̂ádhe). The latter feature cannot be regarded as Müller’s idiosyncrasy in view of Wellsted’s thadder and ṭaṭa’ah (Simeone-Senelle 1991:127).

● In Morris–Watson 2019:85, the Eastern Soqotri form is rendered as ŝáˁte, its Western Soqotri parallel, as ŝóγote (cf. also Simeone-Senelle 1997:395).

At least some of the irregularities pertaining to this numeral may be due to the influence of the widely attested proto-MSA root *ṣ̂γṯ ‘to hold, to take a handfuľ, *ṣ̂aγṯ- ‘handfuľ: Mhr. ẑeγāṯ ‘to crumple, to squeeze, to take a grip on’, Jib. ẓ̂aγáṯ ‘to grip, to hold, to take a handful; to squeeze, to crumple’, ẓ̂aγṯ ‘handfuľ. Soq. ẓ̂óˁot ‘to grasp’, ẓ̂aˁt ‘handfuľ. Such a contamination is capable to explain the presence of -γ- as the second radical in some of the MSA forms (*ŝaγṯ-at-, *ṣ̂aγṯ-at-), as well as the shift from ŝ to ẓ̂ in the Soqotri forms recorded by Müller and our team (as well as in some of the Qishn Mehri variants).

As far as the form ẓ̂áˀteʰ, standard in our corpus, of special interest is the diachronic backgound of the combination -ˀt-, in all probability going to a lost geminate (*-tt- < *-γt-?).