The Qurʔānic feminine ending

The feminine ending in the both the orthography of the QCT and of general Classical Arabic is somewhat unusual. I have been putting off this blogpost (presumably my last on the QCT orthography, for now), because I've had trouble deciding what exactly to do with it.

The feminine ending in Classical Arabic: Was it a diptote?

Let's start with the feminine ending in Classical Arabic, and its spelling in Classical Arabic. The feminine ending in Classical Arabic is marked by -at- followed by the case endings. These case endings in Classical Arabic are u/i/a when definite or in contruct, and un/in/an in the indefinite form. This is exactly the same as the masculine nouns. 

The feminine noun, however, unexpectedly is always written with an <h> (or in later orthography with two dots on top, the ta marbuta ة). This is, of course, rather unexpected if one pronounces it like t. This is traditionally understood as representing the pausal pronunciation. In phrase final position the nunation and short vowels would have been lost, and the resulting -at would undergo a shift -at > -ah.

I've expressed my doubts about 'Pausal spelling' as being the origin of some of the spellings of Arabic before, as it fails to explain, for example, why the construct plural -ū/-ī is spelled <-w>, <-y> and not <-wn>/<-yn>. And if we allow words in 'construct' to be written as they are actually pronounced, why is the feminine -at written with <h> while it is in construct in the Classical Orthography? But whether we allow the possibility of it being a 'pausal' spelling or not, we are faced with a problem with the indefinite accusative.

The pausal rule for the indefinite accusative -an in masculine nouns undergoes a loss of the final n and a lengthening of the preceding vowel: -ā. This is reflected in the spelling, e.g. raǧulā <rǧlʔ>'a man (acc.)'. However, if the feminine noun had this same ending -an, we would expect the same development, and expect an indefinite accusative -atā spelled as **<-tʔ>, but instead we find <-h>. This points to an interesting facts about the dialect that lies at the origin of the orthography of Classical Arabic. It seems that the feminine ending was diptotic, and as such did not take nunation, and therfore the accusative was simply -ata > -at > -ah in pause (remember the essentially reversed argument why Ṯamūd was a triptote in the QCT).

Feminine names with the feminine ending -at-, even in Classical Arabic are normally diptotic, but there is no evidence in the grammar of the language that this was the case for all nouns. But there is real evidence that suggests  that at least in some dialects of Arabic, the feminine endings were diptotic. First of all, there are the dialect of Yemen that retain reflexes of nunation in the Tihama and Saʕdah region.

In the Tihama indefinite nouns have a suffix -ū which disappears whenever a noun is definite, or in construct.  So: miftāḥū 'a key', am-miftāḥ 'the key', miftāḥ ... 'the key of ...'. Nouns that are typically diptotes, like the masculine elative pattern ʔakbar-, however, do not have this alternation. This clearly suggests that ū is a reflex of case vowel + nunation (probably the nominative -un > -ũ > -ū). The Saʕdah dialects leave no doubt that this should be attributed to nunation, as the suffix in indefinite nouns is actually -in.

What the Tihama and Saʕdah dialect have in common, however, is that the feminine nouns never have this reflex of nunation, i.e. the feminine nouns are treated as diptotes and are -ah (or -ih) in indefinite form, and -at/-it in the construct form. Somewhat puzzlingly, the -at/-it form also appears in the definite form. Whatever the explanation is for the definite form, it is clear that the indefinite form is not followed by nunation.  So: bagarah 'a cow', ib-bagarit'the cow', bagarit ... 'the cow of ...'.

It is  not  just in recent dialectal forms that show this distribution. Nabataean Aramaic's sizable Arabic element must be understood in the same way. Many names have so called wawation, nouns that have a final <w>. Names that have this never end in the feminine ending <-t> /-at/, noun/adjectival suffix <-n> /-ān/ (normally diptotic) or the ʔakbar- elative pattern. Also some other nouns can be shown to lack wawation according to them being diptotic, such as the famous pair of ʕamr-un, even in Classical Arabic spelling today spelled <ʕmrw> versus the diptotic name of the same root ʕumar<ʕmr>. This distribution clearly suggests that the Arabic dialect of the Nabataeans underwent the same development as Tihama Yemeni Arabic (shared innovation, or individual development?). Feminine names never having the wawation (< nunation) is however not too  surprising, even within the context of Classical Arabic, as Feminine names are typically diptotic in Classical Arabic too, cf. the pair fāṭimatun'a suckling woman' vs Fāṭimatu'Fatimah'.

For the similarity between Tihama Yemeni, Classical Arabic Orthography and Nabataean Arabic to be completely analogous, a non-Name feminine noun should show up in Nabataean Aramaic written with <-t> rather than <-tw>; These can be distinguished from native Aramaic nouns because Aramaic nouns write the feminine singular ending as <-h>, as Aramaic underwent the -at > -ah shift much earlier than Arabic.

Going by Cantineau's glossary of Nabataean there appear to be some genuine nouns with this <-t> ending, confirming the diptotic nature of regular feminine nouns in Nabataean Arabic as well, e.g.

<bgrt> 'navel', cf. Ar. buǧrah'protrusion of the navel' (which may only occur in Personal names, the glossary is a bit ambiguous).

<šqylt> 'money', šuqaylah, also attested as a name, it doesn't actually have a Classical Arabic equivalent, but the diminutive formation is clearly Arabic.

<ḥlt> 'aunt', cf. Ar. xālah 'id.'

<nsḥt> 'copy', cf. Ar. nusxah'id.', this one is important because it stands in a context in the text where this word certainly cannot be in construct, and we are therefore certainly not dealing with a native, or nativized loan. The feminine <-t> can only be understood as coming from a form of Arabic that 1. doesn't have nunation on feminine nouns and 2. has -at for feminine nouns rather than -ah.

    O'Connor does not consider this word, or xālat above, loanwords. Because both roots have a broader Semitic attestation. This, to me, appears to be an obvious mistake. As mentioned -at in the Absolute State does not occur in Aramaic, so finding a <t> in the absolute is definitive proof that it was borrowed from a language that had a feminine ending -at, which in the case of Nabatean Aramaic, is most likely the dialect of Arabic that its writers almost certainly spoke.

    xālat is a little unclear. What I understand from Jongeling & Hoftijzer, the word is only attested with a possessive following it, in which case it is only attested in construct and it is difficult to tell. But for nusxat, there is no doubt it is attested in the Indefinite form.

<šrkt> 'federation', cf. Ar. širkah'id.'

So, while the data is obviously ambiguous (As Nabatean Aramaic isn't Arabic), it certainly leans towards the general pattern found in Yemeni Arabic, where triptotes in the indefinite receive -ū, while diptotes (which include all feminine nouns) do not.

While the language of the Classical Arabic orthography certainly did not have -ū as a suffix, it did have a reflex of an older stage of nunation in the indefinite accusative -ā. The absence of this -ā on indefinite accusative feminines points to a shared isogloss between Tihama Yemeni Arabic, Arabic of the Classical (and QCT!) Orthography and Nabatean Aramaic. The fact that these three dialect groups are in a more or less geographically contiguous region, is probably relevant.

The feminine ending in construct: A similarity between the modern dialects and the QCT

In most modern dialects, the distribution of the feminine ending -ah versus -at is as follows: -ah is found in the definite and indefinite, while -at is found in the construct form (for two dialects that do not follow this pattern see Van Putten forthcoming). This distribution is quite similar to Nabatean Aramaic, but very different from Classical Arabic, where the feminine ending is simply always -at except in sentence-final position, e.g. Benghazi  Arabic:

ubgura'a cow', l-ubgura 'the cow', ubgurut il-binit 'the cow of the girl'

In Classical Orthography, the feminine ending is written with <h> regardless of whether it is in construct or not. In the Qurʔān however, there are many examples of construct feminine being written as -at while the indefinite and definite are written as -ah. This distribution suggests that a language similar to the modern dialects influence the orthography of the QCT in some way.

The unusual thing, however, is that the majority of construct feminines are written in the regular Classical way. From my count there are 47 cases of a <t>-construct. If we look at the actual attestations of the <t>-constructs, there appears a certain pattern:

sunnah'practice/ways' + al-ʔawwalīna 'the former people', allāhi 'God' (5x vs. 8x with <h>):

Q8:38: sunnatu (a)l-ʔawwalīna<snt ʔlʔwlyn>, Q35:43: sunnata (a)l-ʔawwalīna<snt ʔlʔwlyn>

Q40:85: sunnata (a)llāhi<snt ʔllh>, Q35:43: li-sunnati (a)llāhi<lsnt ʔllh> (2x),

niʕmah 'grace' + allāhi 'God' / rabbi-ka 'your lord' (11x vs 12x with <h>):

Q2:231, Q3:103, Q5:11, Q14:28, Q14:34, Q16:83, Q16:114, Q35:3 niʕmata (a)llāhi<nʕmt ʔllh>, Q16:72 wa-bi-niʕmati (a)llāhi<wbnʕmt ʔllh>, Q31:31 bi-niʕmati (a)llāhi <bnʕmt ʔllh>

Q52:29: bi-niʕmati rabbi-ka <bnʕmt rbk>

raḥmah 'mercy' + allāhi 'God' / rabbi-ka 'your lord' (7x vs 12x with <h>):

Q2:218, Q7:56 raḥmata (a)llāhi Q11:73 raḥmatu (a)llāhi Q30:50 raḥmati (a)llāhi<rḥmt ʔllh> 'mercy of god'

Q19:2 raḥmati rabbi-ka Q42:32 raḥmata rabbi-ka <rḥmt rbk>, Q43:32 wa-raḥmatu rabbi-ka <wrḥmt rbk>

laʕnah 'curse' + allāhi 'God' (2x vs 5x with <h>):

Q3:61 laʕnata (a)llāhi Q24:7 laʕnata (a)llāhi <lʕnt ʔllh>

kalimah 'word' + rabbi-ka 'your lord' (5x vs 1x + rabbi-ka 7x + other nouns with <h>)

Q40:6,  Q6:115, Q7:137 Q10:33, Q10:96 kalimatu rabbi-ka <klmt rbk>

imraʔah 'woman, wife' + nūḥin 'Noah', lūṭin 'Lot', firʕawna 'Pharaoh', ʕimrāna 'Imran, the mother of Mary ', al-ʕazīzi 'Al-ʕazīz' (7x vs 0x with <h>)

Q66:10 (i)mraʔata nūḥin <ʔmrʔt nwḥ>

Q66:10 wa-(i)mraʔta lūṭin<wʔmrʔt lwṭ>

Q28:9 (i)mraʔatu firʕawna Q66:11 (i)mraʔata firʕawna <ʔmrʔt frʔwn> 'the wife of the pharaoh'

Q3:35 (i)mraʔatu ʕimrāna<ʔmrʔt ʕmrn>

Q12:30, Q12:51 (i)mraʔatu (a)l-ʕazīzi<ʔmrʔt ʔlʕzyz>

ġayābah 'bottom' + al-ǧubbi 'the well' (2x vs 0x with <h>)

Q12:10, Q12:15 fī ġayābati (a)l-ǧubbi<fy ġybt ʔlǧb> 'into the bottom of the well'

maʕṣiyah 'disobedience' + ar-rasūli 'the messenger' (2x vs 0x with <h>)

Q58:8, Q58:9 wa-maʕṣiyati (a)r-rasūli <wmʕṣydt ʔlrswl> 'disobedience of (=to) the messenger'

6 other examples that do not repeat remain:

Q11:86 baqiyyatu (a)llāhi <bqyt ʔllh> 'the remnant of God'

Q30:30 fiṭrata (a)llāhi <fṭrt ʔllh> 'the nature of Allah'

Q44:43 šaǧarata (a)z-zaqqūmi <šǧrt ʔlzqwm> 'the tree of the Zaqqūm'

Q56:89: wa-ǧannatu naʕīmin<wǧnt nʕym> 'a garden of pleasure'

Q28:9 qurratuʕayni  'a comfort of the eye' (said of a child, usually qurratu ʔaʕyun<qrh ʔʕyn> 'comfort of the eyes').

Q66:12 (i)bnata ʕimrāna <ʔbnt ʕmrn> 'the daughter of ʕimrān'

So the 47 attestations of the t construct consist of a total of 14 different lexical items. What is remarkable about these words is that they are all part of religious phraseology or common religious stories also found in Judaism and/or Christianity so not necessarily Islamic phraseology [Edit: Benjamin Suchard rightfully points out that is not true for the tree of Zaqqūm]. This distribution, to me, suggests that the these spellings were somewhat "fixed" phrases from a Pre-Islamic religious scribal tradition. This scribal tradition presumably had a -ah/-at distribution similar to the modern dialects today (alternatively, one might imagine they were imitating the Aramaic distribution, but that might be a stretch).

This then suggests that the -ah/-at distribution as found in these phrases do not necessarily reflect the language of the QCT.

What was the feminine ending in the language of the QCT?

So if the language of the QCT did not have the 'Neo-Arabic' -ah/-at distribution (unlike the dialect that was the donor of the spelling of the religious phraseology), then what was the way the feminine ending was pronounced?

If we go by the spelling, the feminine ending would be pronounced -ah in all contexts. This is, however, not a very satisfactory suggestion, as there is absolutely no evidence in Old Arabic or the modern dialects that such a dialect ever existed. Instead, I would like to propose that the QCT language had -at in all positions (like earlier Shammari) except perhaps in Pause, where it may have had -ah, (as in Kuwaiti Dōsiri). The scribes must have been aware of the -ah/-at distribution, because of the couple of common religious phrases that they learned to write with such alternation. They however, were not able to 'natively' reproduce this alternation, and therefore settled on writing <h> in every single environment, rather than just in construct, as to these speakers there would have been no discernible difference.

The choice of <h> over the much more obvious <t> may have motivated by it giving the ability to differentiate from the feminine plural suffix -āt, which they consistently spelled <-t>.

This then, leaves us at a situation which is surprisingly close to the Classical Arabic situation, the Feminine ending would have had -at in all environments (except maybe in Pause), but, different from Classical Arabic, must have been diptotic. So, depending on whether the QCT language had short final vowels or not (something which is difficult to decide with the evidence at hand), the paradigm of the feminine ending must have looked as follows:

 -at spelled as <t> Outside of construct

There are two cases of feminine t outside of construct, they can both plausibly be read as actually denoting the sound feminine plural ending -āt, rather than -at. And they are both read as such in a variety of reading traditions.

Q35:40fa-hum ʕalā bayyinatin min-hu<fhm ʕly bynt mnh> 'So they are [standing] on evidence therefrom?'

    In We II 1913 and Tübingen Manuscript.

The old attestation of this spelling suggests that it's old, and not some unusual result. Some of the reading traditions (Ibn ʕāmir and Šuʕbah) read <bynt> as the plural bayyināt, which would be expected to be spelled as such in Qurʔānic orthography (feminine plural is spelled <ʔt> only once). 'evidences' or 'evidence' seems to have little influence on the meaning of this verse, so I am inclined to accept that reading as the original one.

Q77:33ka-ʔannahū ǧimālatun ṣufrun<kʔnh ǧmlt ṣfr> 'As if they were yellowish [black] camels'

This form is attested as such in some Qurʔān documents, but this part of the Qurʔān is simply not attested in the oldest documents.

As with Q35:40, this line is read in a variety of reading traditions as ǧimālātun instead (Nāfiʕ, Ibn Kaṯīr, Abū ʕamr, Ibn ʕāmir and Šuʕbah), which would not alter the meaning greatly, and it would explain the spelling.

[Edit] Ahmad Al-Jallad made the useful observation that I should probably mention that Joshua Blau (2006) already saw the clear similarities between the Nabatean Arabic wawation, and the indefinite -ū in Yemeni Arabic.