The feminine ending: A "hypercorrection" in orthographic innovation?

Motivated by the posting of a series of religious early Islamic inscriptions  on Twitter written by women, I'd like to discuss once again the spelling of the feminine ending in Arabic.

One of the striking things about the Feminine ending, which in Classical Arabic is -at- followed by case endings and pronounced -ah in pause, is that it is always spelled with a final ه <h>, even in construct.

This is normally explained as "pausal spelling", that is, words in Classical Arabic are spelled in the way that they are pronounced in phrase-final position. While this explains the spelling of the feminine ending as ه in most positions, it fails to explain its form in Construct. A construct cannot appear in phrase-final position, and therefore does not have a pausal form. As such, we would expect construct feminines to simply be written with a ـت. This is comparable to other nouns that have a different form in construct than in pause, where also the orthographic distinction is made.

The construct masculine plural, for example is Nom. -ū Gen./Acc. -ū/-ī, while the non-construct form is -ūna, -īna (/-ūn/ and /-īn/ in pause). But the construct masculine plural is never spelled ـون <-wn> or ـين <-yn>, which would be their "pausal" pronunciation. The same could be said for ʔab 'father', whose construct form is ʔabū/ʔabā/ʔabī, which is simply just always spelled ابو <ʔbw>, ابا <ʔbʔ> and ابى <ʔby>.

The spelling of the feminine singular in construct with ه <h>, then, is an actual exception to how Classical Orthography is "supposed" to work.

If we look at the pre-Classical orthography of the feminine ending, in the Quran, early Islamic inscriptions and papyri, we find something interesting. Feminine construct are much more commonly spelled phonetically with ـت, examples of which are مريم ابنت عمرن <mrym ʔbnt ʕmrn> /maryam ibnat ʕimrān/ 'Mary the daughter of ʕimrān'. So far, I have not encountered a single example of this noun being spelled as anything other than in its phonetic spelling with ـت <t>. (It is the spelling of this word found in the inscriptions linked above that motivated me to write this blogpost)

Almost a quarter of all the feminine construct phrases present in the Quran are spelled phonetically, and many other such phrases are commonly, or even regularly spelled with ـت <t> in the early islamic inscriptions, e.g. رحمت الله <rḥmt ʔllh> /raḥmat aḷḷāh/ 'the mercy of God'.

While the construct spelling with ـت <t> is fairly common in the early Islamic period, the innovative spelling with ـه <h> is well-attested and more common from the very earliest period. This leads one to wonder: where does this spelling convention come from?

A recent article by Laïla Nehmé, who looked at the pre-Islamic treatment in Nabataean-Aramaic and Nabataean-Arabic inscriptions failed to give a very satisfying answer (due to no fault of her own). Aramaic, the language in which Nabataean inscriptions were mostly written, and from which Arabic derives its orthographic practices (and script), has a very similar distribution from that of the "anomalous" spellings found in Arabic. Indefinite feminine nouns ended either in /-ah/  or /-ā/ and were spelled with a final <-h>, whereas construct feminine which ended in /-at/ were spelled with final <-t>. One might imagine that overlap in this orthography may have lead Arabic scribes to take over this practice. And as a result, you might already expect the gradual replacement of construct feminines -- originally written with <-t> -- to undergo replacement by <-h> in construct position. This is a very reasonable hypothesis, which is conclusively disproven by Nehmé, who shows that as late as the 5th century there is absolutely no sign of this (and the 6th century does not yet yield any useful data).

As far as is currently possible to be seen, using ـه <-h> regardless of context for the feminine ending is a uniquely Arabic feature of the early Islamic period. That is not to say that it is an innovation that only happened after the Islamic period, but due to a lack of the vast majority of written material before Islam (no papyri have been found yet from the late 6th/7th century, nor ones written in Arabic), it is a little difficult to see when it happened, but whatever the case, it appears to have happened fairly late.

So then the question becomes: How did this happen? I have an idea, but we will have to draw a little bit of the context of early Islamic orthography.

There is a reason why I call early Islamic orthography that. As of yet, there is no written evidence for early islamic orthography before Islam. While more and more pre-Islamic Arabic inscriptions that are recognisably writting in the Arabic script are becoming known to us, pre-Islamic Arabic writing has not undergone certain innovations that are typical of the early Islamic orthography:

1. Before Islam, many names are still written with the otiose wawation, a final wāw at the end of --mostly triptotic-- nouns. This has reasonably been taken as a vestige of the nominative ending /-u/. In the modern orthography the only name that still has this is عمرو ʕamr-un where it seems to be retained because it allowed it to be easily distinguished from the diptotic name which would have otherwise been a homograph: عمر ʕumar-u. An innovation of early Islamic orthography is getting rid of this no-longer meaningful final ـو.
2. Before Islam, the ا <ʔ> sign was used to write the glottal stop (or Hamz as it is called by Arabists), a name like ḏiʔb would be spelled دابو <dʔbw>, and not like ذيب <ḏyb> as we find it in the early Islamic period. It has been convincingly argued (especially by Werner Diem, but also others) that the dialect on which the early Islamic orthography was based had lost the glottal stop. What is important here is that the orthography was innovated accordingly. This need not be the case. English still writes knight, right and night with an otiose gh that has been lost hundreds of years ago (cf. Dutch knecht, recht and nacht -- where it is actually pronounced).
3. Word-internal long ā, which could not be expressed in Pre-Islamic Arabic, came to be (optionally) designated by ا <ʔ> even in places where it did not etymologically  have a glottal stop. In Classical Arabic orthography, the spelling becomes mandatory, but that certainly isn't the case in, e.g. the Quran.
4. Before Islam, the feminine ending was spelled ـت <t>, regardless of context. Presumably because the earliest form of Nabatean Arabic still pronounced this feminine ending as /-at/ in all contexts. Later, Nabataean Arabic (and other forms of Arabic that had taken on the Nabataean Arabic orthography) undergoes the shift from *-at > -ah.

In other words, early Islamic orthography clearly represents a explicit effort to represent Arabic more closely to its actual pronunciation, rather than the archaizing orthographic practice that came along with the scribal practices associated with the Arabic script. The early Islamic orthography is the result of a spelling reform. When exactly this spelling reform took place, is not exactly clear, but it seems that it was some time before the advent of Islam and some time after the beginning of the 6th century. But if we are correct that this spelling reform took place in or around Mecca & Medina, this dating should in fact be taken with a grain of salt. We do not yet have an extensive and detailed view of Pre-Islamic writing form this region. It may therefore have happened earlier.

Something that speaks in favour of the spelling reform taking place closer to the 7th century rather than farther away from it, is that early Islamic Arabic still seems to clearly represent a transitional phase. Early Islamic Papyri like PERF 558 still attests a name written with the Archaic wawation on a name other than عمرو namely, حديدو <ḥdydw> /ḥudayd/ or /ḥadīd/, and the new use of ا <ʔ> for writing /ā/ is still far from a well-established practice.

I think in light of the idea of a conscious (although not necessarily very centralized) spelling reform, I think we might be able understand the unusual spelling of the construct feminine ending being spelled as ه <-h>.

As I alluded to earlier, the feminine ending /-at/ of Nabataean Arabic appears to have shifted to /-ah/ fairly early. Despite this, it continued to be conventional to spell this ending with a final <t>. One of the evidences that hints at an early shift of this is the occasional spelling of feminine names with a final <h>. e.g. <ʔmt> occasionally shows up spelled as <ʔmh> suggesting a pronunciation /ʔamah/ rather than /ʔamat/. However, there are plenty of attestations of the spelling <ʔmt> that postdate the attested inscription that has <ʔmh>, suggesting that the conventional spelling was with a <t>, despite being pronounced as /-ah/. This is corroborated by Greek transcriptions of Nabataean names, which as early as the 1st century BCE transcribe the feminine ending without a sign of the final t, as Ahmad Al-Jallad shows in a recent article (pg. 157).

If we assume this conventionalized spelling persisted in the writing schools of Hijaz right up until the spelling reform of the early Islamic orthography (which, I stress again, might not have actually been in the early Islamic period), we might understand where Arabic's unusual treatment of feminine constructs comes from.

First, scribes would have been used to conventionally write ـت <-t> for the feminine ending, regardless of context. This feminine ending, however, would have been most commonly pronounced /-ah/, and only in construct be pronounced /-at/.

Then, when the spelling reform was enacted a rule was establish: "The feminine ending you conventionally write ـت <-t> will from now on be spelled ـه <-h> in line with its pronunciation, that is /-ah/." This rule is incorrect for creating a phonetic orthography, but it is a rule that would be easily formulated. When giving a list of words that used to be written with ـت <-at> and are now to be written ـه <-h>, such a list would of course never contain examples of the feminine noun in construct. As such, this reform principle could easily have been applied hypercorrectly to every position where the feminine ending was spelled ـت <-t>.

This idea of a rather "mechanical" substituation of the ـت <-t> ending with ـه <-h> is, I think, not altogether unlikely. Pre-reform written Arabic would have been very far removed from the spoken language, and scribes would have been used to principles of "mechanical substitution", where they would have to convert any word that sounded like it ended with /-ah/ to be spelled as if it was pronounced /-at/. This mismatch and mechanical substitution that was part of the scribal tradition could have given rise to the hypercorrect mechanical substituation in the other direction.

The not at all uncommon spellings where the construct feminine is spelled with ـت <-t> in the early Islamic writings must then simply be considered accurate phonetic transcriptions, and occasionally archaic spellings in fixed phrases. It does not seem coincidental to me that such spellings are especially common with a fixed phrase like رحمت الله 'The mercy of god', and phraseology that would almost always have a construct position such as ابنت [personal name] 'daughter of PN' and امرت [personal name] 'wife of PN' (which is regularly spelled as such in the Quran as well without exception).

I am happy with this explanation, but it is admittedly still quite speculative, due to a lack of late 6th century written material from the critical area that we are discussing. So here are some ways further evidence could falsify or corroborate this hypothesis:

If we would find that in the Hijaz writing the feminine ending conventionally as ـت <-t> had fallen out of use well before the 6th century, but we nevertheless still find it written as such consistently in construct with no sign of the 'new' spelling with ـه <-h>, this would clearly falsify this hypothesis.

Finding the use of ـه <-h> in construct in a pre-Islamic text, would confirm my suspicion that the spelling reform is not a purely Islamic-period innovation. We are lucky that the Arabic word for 'year' is a feminine noun (sanah) and that if we find a dated text, we are likely also find examples of this spelling!

A final but completely unrealistic way of proving it, would be the finding of an actual prescriptive text laying out the new spelling conventions. This is unlikely to happen.

Any thoughts or comments on this topic will be greatly appreciated!