On the fitness of abjads

[Kuchigakatai]

tl;dr: believe it or not, I argue that the Arabic script fits Classical Latin a lot better than Classical/Standard Arabic.



I've never quite understood why people here sometimes say abjads fit the Semitic languages well. As if an abjad goes together well with a triconsonantal language or something like that. Shouldn't it be otherwise, since wouldn't a language whose morphology involves vowels a lot more than it's usual instead of using affixes more (which often have extra consonants) then need to express those vowels more clearly than other kinds of languages?

I think I can clarify what I mean by using Standard Arabic as an example. Let's start by having a look at the how the inflection of verbs in the suffix conjugation is spelled. Don't worry, I'll be using a fairly close transliteration since I imagine many of you don't know how to read the Arabic script.

(Note about the transliteration used in this post: 2alif without a hamza is <a>, 2alif with any hamza is <2>, taa2 marbuuTa is <ẗ>.)

## Verb inflection (suffix conjugation) ##
word: kataba 'to write'
Active Voice --- Passive Voice
katabtu <ktbt> --- kutibtu <ktbt>
katabta <ktbt> --- kutibta <ktbt>
katabti <ktbt> --- kutibti <ktbt>
kataba <ktb> --- kutiba <ktb>
katabat <ktbt> --- kutibat <ktbt>
katabtumaa <ktbtma> --- kutibtumaa <ktbtma>
katabaa <ktba> --- kutibaa <ktba>
katabataa <ktbta> --- kutibataa <ktbta>
katabnaa <ktbna> --- kutibnaa <ktbna>
katabtum <ktbtm> --- kutibtum <ktbtm>
katabtunna <ktbtn> -- kutibtunna <ktbtn>
katabuu <ktbwa> --- kutibuu <ktbwa>
katabna <ktbn> --- kutibna <ktbn>

Read the forms vertically and horizontally. All the passive forms are spelled the same as their corresponding active voice forms. No wonder Arabic speakers often, but by no means consistently, add a disambiguating "u" vowel mark on the first consonant, in this case <k>. On the other hand, even if we stay within a single voice (and we shouldn't because they aren't always distinguished in writing), you can see that there are four inflections with potentially a different sound, katabtu, katabta, katabti and katabat, that are all spelled the same, <ktbt>. I'm saying "potentially" because in practice people often pronounce both katabtu and katabta as "katabt" (and kutibtu and kutibta as "kutibt"), but even this would still leave us with three (really six) forms that are regularly written the same, <ktbt>.

It gets much, much worse in the prefix conjugations.

## Verb inflection (prefix conjugations) ##
word: kataba 'to write'
Active indicative --- Passive indicative --- Active subjunctive --- Passive subjunctive --- Active jussive --- Passive jussive
2aktubu <2ktb> --- 2uktibu <2ktb> --- 2aktuba <2ktb> --- 2uktiba <2ktb> --- 2aktub <2ktb> --- 2uktib <2ktb>
taktubu <tktb> --- tuktibu <tktb> --- taktuba <tktb> --- tuktiba <tktb> --- taktub <tktb> --- tuktib <tktb>
taktubiina <tktbyn> --- tuktibiina <tktbyn> --- taktubii <tktby> --- tuktibii <tktby> --- taktubii <tktby> --- tuktibiii <tktby>
yaktubu <yktb> --- yuktibu <yktb> --- yaktuba <yktb> --- yuktiba <yktb> --- yaktub <yktb> --- yuktib <yktb>
taktubaani <tktban> --- tuktibaani <tktban> --- taktubaa <tktba> --- tuktibaa <tktba> --- taktubaa <tktba> --- tuktibaa <tktba>
yaktubaani <yktban> --- yuktibaani <yktban> --- yaktubaa <yktba> --- yuktibaa <yktba> --- yaktubaa <yktba> --- yuktibaa <yktba>
naktubu <nktb> --- nuktibu <nktb> --- naktuba <nktb> --- nuktiba <nktb> --- naktub <nktb> --- nuktib <nktb>
taktubuuna <tktbwn> --- tuktibuuna <tktbwn> --- taktubuu <tktbwa> --- tuktibuu <tktbwa> --- taktubuu <tktbwa> --- tuktibuu <tktbwa>
taktubna <tktbn> --- tuktibna <tktbn> --- taktubna <tktbn> --- tuktibna <tktbn> --- taktubna <tktbn> --- tuktibna <tktbn>
yaktubuuna <yktbwn> --- yuktibuuna <yktbwn> --- yaktubuu <yktbwa> --- yuktibuu <yktbwa> --- yaktubuu <yktbwa> --- yuktibuu <yktbwa>
yaktubna <yktbn> --- yuktibna <yktbn> --- yaktubna <yktbn> --- yuktibna <yktbn> --- yaktubna <yktbn> --- yuktibna <yktbn>

Read the forms horizontally. As you can see in the first row, every one of the six forms has a different sound, but they're all spelled the same, <2ktb>. As with the suffix conjugation, Arabic speakers often, but by no means consistently, add a "u" vowel mark on top of the first consonant to indicate it's passive, reducing the ambiguity. But it is not normal to distinguish the three moods.

Nominal inflection isn't that much better.

## Noun inflection (triptote declension) ##
word: wajh 'face (singular)'
Indefinite --------- Definite --------- Construct
wajhun <wajh> --- al-wajhu <alwjh> --- wajhu <wjh>
wajhan <wajha> --- al-wajha <alwjh> --- wajha <wjh>
wajhin <wajh> --- al-wajhi <alwjh> --- wajhi <wjh>

The triptote declension is the most common declension for nouns in Arabic, and none of the cases are distinguished in writing in this declension except for the indefinite accusative wajhan <wjha>! Neither are the indefinite and construct distinguished most of the time. In practice, this doesn't bother modern Arabic speakers that much because they don't make case distinctions anymore anyway, nor the indefinite-construct distinction except for a few (common) endings, but if the Arabic script was designed for Classical Arabic, how can we really say the script "fits" it when all these forms have different sounds but the same spelling?

I'm not going to talk about derivational morphology in Standard Arabic because that does usually involve consonantal affixes, often with a long vowel, so there's not much a problem there. It is regrettable that the derivational prefixes mi- (which derives tools), ma- (which derives places) and mu- (which derives agent nouns and participles) are spelled the same though, <m> before the triconsonantal root. This gives results such as maktab 'desk', mukattib 'write.CAUS.AGT.PTCP' and mukattab 'write.CAUS.PASS.PTCP' being spelled the same, <mktb>.




Now I was thinking what would be a language that the Arabic script would fit better, hypothetically speaking. Probably a language that uses consonants often even for word inflection, and otherwise often long vowels so that they could be written out with a mater lectionis consonant. Latin. I think Latin might do.

## Verb inflection (-āre conjugation) ##
word: amāre <2mar> 'to love'
Present indicative --- Present subjunctive
amō <2mw> --- amem <2mm>
amās <2mas> --- amēs <2mys>
amat <2mt> --- amet <2mt>
amāmus <2mams> --- amēmus <2myms>
amātis <2mats> --- amētis <2myts>
amant <2mnt> --- ament <2mnt>

Here I'm comparing two of the more similar tenses of the most common conjugation or verb group. We do get a merge in 3rd person forms across mood (amat and amet become <2mt>), but this *is* doing a better job than the verb inflection in Arabic so far.

Latin does have something like the similar-sounding prefix conjugations of Arabic, and they're called the imperfect subjunctive, the future perfect and the perfect subjunctive put together.

Imperfect subj. (active) --- Future perfect --- Perfect subj.
amārem <2marm> --- amā(ve)rō <2ma(w)rw> --- amā(ve)rim <2ma(w)rm>
amārēs <2marys> --- amā(ve)ris <2ma(w)rs> --- amā(ve)rīs <2ma(w)rys>
amāret <2mart> --- amā(ve)rit <2ma(w)rt> --- amā(ve)rit <2ma(w)rt>
amārēmus <2maryms> --- amā(ve)rimus <2ma(w)rms> --- amā(ve)rīmus <2ma(w)ryms>
amārētis <2maryts> --- amā(ve)ritis <2ma(w)rts> --- amā(ve)rītis <2ma(w)ryts>
amārent <2marnt> --- amā(ve)runt <2ma(w)rnt> --- amā(ve)rint <2ma(w)rnt>

The imperfect subjunctive has passive forms, but they're distinct enough thanks to their consonantal endings that I decided not to include them. The future perfect and perfect subjunctive have no pesky passive forms.

Latin at its worst gives us a much better situation than what we get in Arabic at its worst! There is some merging of moods as well, especially noticeable in the third person short forms, amāret amārit amārit <2mart>, and amārent amārunt amārint <2marnt>, but overall this is far milder.

Now let's look at the most essential of Latin's famous nominal declensions.

## Noun inflection (1st declension) ##
word: lingua 'tongue'
Singular
lingua <lngwẗ> --- linguae <lngwy>
linguae <lngwy> --- linguārum <lngwarm>
linguae <lngwy> --- linguīs <lngwys>
linguam <lngm> --- linguās <lngwas>
linguā <lngwa> --- linguīs <lngwys>

## Noun inflection (2nd declension) ##
word: digitus 'finger'
Singular --------- Plural
digitus <dgts> --- digitī <dgty>
digitī <dgty> --- digitōrum <dgtwrm>
digitō <dgtw> --- digitīs <dgtys>
digitum <dgtm> --- digitōs <dgtws>
digitō <dgtw> --- digitīs <dgtys>
digite <dgt> --- digitī <dgty>

## Noun inflection (3rd declension) ##
word: rex 'king'
Singular --------- Plural
rēx <ryks> --- rēgēs <rygys>
rēgis <rygs> --- rēgum <rygm>
rēgī <rygy> --- rēgibus <rygbs>
rēgem <rygm> --- rēgēs <rygys>
rēge <ryg> --- rēgibus <rygbs>

This is amazingly fitting, actually! The one thing that gets lost is -em vs. -um in the 3rd declension. When applying this in practice we would have some merges across the declensions (not within them), notably the loss of gender distinction in amīcum amīcam <2mykm> and amīcī and amīcae <2myky>, but I think a naturalistic application of the Arabic script would probably maintain the distinction by spelling the two feminine forms with an <a>: amīcam <2mykam>, amīcae <2mykay>. The latter would even be naturally justifiable as influence of the poetic amīcāī!

Interestingly, we would gain distinctions that are not made in Classical Latin's own native script. For example, on the side of verbs the future perfect and perfect subjunctive canonical forms would be more clearly distinguished. On the side of nouns, there would be a distinction between e.g. palus <pls> 'stick' and palūs <plws> 'marsh', both of which were spelled PALVS in the ancient tongue of the Romans. This would especially affect the less common 4th declension. Let's look at it.

## Noun inflection (4th declension) ##
word: manus 'hand'
Singular --------- Plural
manus <mns> --- manūs <mnws>
manūs <manws> --- manuum <mn2m>
manuī <man2y> --- manibus <mnbs>
manum <mnm> --- manūs <mnws>
manū <mnw> --- manibus <mnbs>

Although it's true that on the other hand, the distinction between 3rd declension artibus 'for the arts' and 4th declension artubus 'with the limbs' would be lost, <2rtbs>.

Also, we could consider grabbing <p> from the Persians. That'd be a good idea.



I hope I have entertained you today. I also hope at least one of you feels like writing a little Latin in the Arabic script now.

Here's a couple sentences in Latin. Let's see if any of you who know a non-trivial amount of Latin can have a crack at what they say.

سكيتسن، پإري، كوس فإرت تتس پمپونإس أتكس؟ كرنيلإس نپوس ديكت تم بن لكوريتر إت جريكي إت لتيني أت، كم پإيمات لجرت، سپرا نهل پست أدي

<skytsn, p2ry, kws f2rt tts pmpwn2s 2tks? krnyl2s npws dykt tm bn lkwrytr 2t gryky 2t ltyny 2t, km p2ymat lgrt, spra nhl pst 2dy.>

[mèþru]

The problem is that in languages without consonant roots too many roots could have identical spellings.

[Pabappa]

Persian does it and it is IE. Turkish was once written in Arabic. However just because a natlang does it doesn't mean it's good.... I've heard Persian script described as the worst of all natlang scripts. The turks ablolished theirs.
Yiddish seems to be written with vowels as full letters. Some Muslims in China also use Arabic script with vowels (but not tones) given their own glyphs. All in all I'd say your idea is workable, but only a small amount of ambiguity is tolerable... if it gets to the point that you can't tell a bee from a boy then you might have to start letting vowels have their own letters.

[mèþru]

You can't say an abjad is suitable for a language when the rules are modified to create vowel glyphs because then it isn't an abjad anymore.

[Curlyjimsam]

The problem is that in languages without consonant roots too many roots could have identical spellings.

I am not sure how big of a problem this would be in practice, particularly in a language like Latin whose vocabulary is mostly derived from a consonantal root language.

[KathTheDragon]

Calling PIE a consonantal root language doesn't make any sense, since that implies a strong resemblance to Semitic, which just doesn't exist. Ablaut is PIE is extremely limited and unproductive, being limited in which vowels can appear (though subject to lengthening in specific morphological contexts) and in exactly one ablaut slot in most cases, though some roots that make specific stems can shift that ablaut slot exactly one consonant to the right. Meanwhile in Semitic, any vowel, short and long, can appear between any two consonants. The only quality they share is "vowel apophony" and even that isn't exclusive.

[mèþru]

Exactly. The vowel changes in consonant roots are way more involved than anything I've seen in IE linguistics.

[Zaarin]

Persian does it and it is IE. Turkish was once written in Arabic. However just because a natlang does it doesn't mean it's good.... I've heard Persian script described as the worst of all natlang scripts. The turks ablolished theirs.

What's more fascinating is Middle Persian using Aramaic as a logographic script. I'd argue Arabic is a better fit than that.

[masako]

The only thing I would be able to add to this already in-depth discussion is that the Arabic abjad has been adapted for and is currently used by around 4 dozen non-consonantal root languages, so...yeah. Not really noteworthy that it could be easily tweaked to fit Latin.

[Tropylium]

Ablaut is PIE is extremely limited and unproductive

Depends on who you ask; at minimum Roland Pooth would disagree.

[mèþru]

It definitely is very limited compared to any Semitic language or to zompist's Skourene
And unproductive in daughter languages

[Curlyjimsam]

Whether or not PIE fits the definition of a consonant-root language is quibbling over terminology which is irrelevant to the point; the degree of variation permitted in the vowels has no bearing on the argument made. In PIE, as in Arabic, the meaning of roots themselves is basically carried on the consonants alone. Thus, an abjad for this language would not introduce ambiguities between roots which are not homophonous in speech anyway. The same is likely to be largely true of PIE-derived vocabulary in Latin.

[mèþru]

...I strongly doubt that that is true of Latin or most other PIE derivatives, regardless of its truth for PIE.

[zompist]

...I strongly doubt that that is true of Latin or most other PIE derivatives, regardless of its truth for PIE.

Yet it's been used for PIE derivatives, cf. Persian and Urdu (and relatives).

People will use writing systems with astonishing deficiencies— e.g. Linear B is a terrible system for writing Greek. But then, once you know the language and the system well, they're a) just good enough, and b) a bear to change.

[Vijay]

Arabic script has been used for quite a wide variety of languages, for whatever it's worth. It's even used for Mandarin Chinese - and has been used for Spanish, Polish, Albanian, Belarusian, Bosnian, Greek, and French.

[mèþru]

I was remarking about suitability for those languages as an abjad. I'm well aware of its actual use in other languages as well as of other Semitic scripts.

[Moose-tache]

I've often had the same thoughts about Arabic "not needing" vowels. In my experience, native speakers seem to mostly treat it as a mnemonic device rather than a precise script. I've never seen single words written down except where context makes it obvious (i.e. stop signs and the like). You need the vowels, plain and simple. But I never noticed how well an abjad could suffice for a language like Latin. I think all the nitpicking about PIE roots being "consonantal" or not (as if that were meaningful to begin with) are missing the point of the experiment. Imagine if you were told your whole life, without proof, that camels are really appropriate transportation in the desert, and then one day you discover that camels are actually fully aquatic mammals that only thrive at the bottom of freshwater lakes. That's the feeling you should be getting from all this.

[WeepingElf]

People tend to change only what doesn't feel like working well enough. Underspecifiying scripts (and that's what an abjad is) are not the exception but the norm. Also, I doubt that any Greek ever thought, "Hmm, that's nice, But what about the vowels?" and went forth inventing vowel letters. Rather, the Greeks understood words beginning with /j/, /w/ or laryngeals to begin with vowels, and used those letters for vowels when they started writing their own languages in Phoenician letters, and thus it happened.

Likewise, the invention of the abugida by the Indians (and, probably independently, the Ethiopians) may also not have been a conscious attempt to "improve" the script, They essentially just made vowel diacritics, which were already a thing at that time, mandatory.

[Salmoneus]

I think Ser's missing the point about what writing is for. The reader's knowledge of their language's grammar can fill in all the grammatical stuff - stuff like gender agreement can be deduced from context. Even part of speech. The problem is writing isn't to make sure you can work out that "never kid a gangster" has 'kid' the verb, and "never eat a kid" has 'kid' the noun - grammar and semantics tell you that. Nor is it to tell that the verb in "the man kids the gangster" is marked to agree with a third-person singular subject. Again, the reader can work that out really quickly.

What the reader can't immediately work out without help from the spelling is whether the sign says "never eat a kid" or "never eat a cod". Whether the young woman in the story is asking for a pat, or a pet, or a pot. (or a pit or a put or a putt or a poot or a peat or a part or a port...). What's vitally important, because you can't deduce it from the grammar, and you often have difficulty deducing it from context, is not the grammar but the root.

That's why people say abjads are well-suited to semitic languages: because losing the vowels very rarely misleads you as to the root. And the other information is almost all redundant anyway.


That's also why it's not appropriate for Latin. Because someone reading Latin doesn't really have to worry whether the sentence says king-abl or king-dat - at least, much less than they have to worry whether it says "king" or "grave" (rogus). Sure, it's nice to be able to distinguish "He loves the girl" from "he would love the girl" - but the really pressing concern is instead to distinguish "he loves the girl" (amat) from "he purchases the girl" (emit). Whether he removes the girl (demit) or tames or breaks the girl (domat). Whether I rule the girl (rego), or ask the girl (rogo), or just moisten her (rigo). And that's before you come to ambiguities between parts of speech - whether RGS RGM LGS means "you moisten the grave of the law!" or "I would oversee the laws of the kings", or "the kings select the grave" or "I bandage the king of graves", and so on and so forth. Ambiguities in the area of root meaning can lead to many awkward circumstances!

[mèþru]

Thanks Salmoneus. This is exactly what I was trying to write but couldn't.