I've been looking at Classical Hebrew a lot for inspiration on some features it has. Turns out, it wasn't what I expected (such as that its not actually a VSO language, it just looks like that due to it being pro-drop and the bible really liking to link sentences together with conjunctions). However, in the process I learned some things about its abjad.

Of course, abjads normally get stereotyped as impractical and antiquated. However, in Semitic languages, they're not to impractical as they may look. Due to numerable reasons, figuring out the vowels on your own isn't actually that hard. Sure, its not 100% reliable, but let's be real - it's no worse than English's infamously inconsistent orthography (how many sounds can the gh alone digraph represent?) Basically, an abjad is more akin to a compression algorithm. You just write what's needed, because the rest can be deduced from that. This actually interests me; it's like a syllabary in reducing the number of characters needed to write a word, but it doesn't require nearly as many letters, and syllable structure is also far more free since you don't need a symbol for every possible syllable (languages that do use them tend to have a very small number of syllables, thus increasing word length simply because fewer short words are possible).

Also, I thought it may be an interesting creative challenge. Make a conlang that works perfectly fine for an abjad. However, this is proving harder than I expected.

For instance, you can't make as much use out of stem changes. Think of the English words speak and spoke. If you wrote them with just consonants, then both would be spk. Clearly the language can't have something like this. In Hebrew, just about every possible 'vowel template' also involves the addition of consonant somewhere, meaning the extra consonant essentially shows what vowels are used. Thus any changes in vowels must be accompanied by some consonant. I guess part of speech would work fine though, however you'd obviously need to make it easy to determine part of speech. English does this fine, and it even makes liberal use of zero derivation. On the topic of English, y cn ndrstnd nglsh vn wtht th vwls, a rather old trick. Then there's Yiddish, which is a dialect of German written in the Hebrew script, and it makes use of stem changes arguably more so than English does! If they can get away with it, surely a language specifically made for an abjad could do so.

Another problem is it sorta reduces the number of possible words. Let's say you had a normal 5-vowel inventory. In the case of a CVCV word, this means any one pair of consonants could have up to 25 variants. This would result in a lot of ambiguity of course if the vowels weren't written. Thus, at most only a few of these could be used. This obviously would put quite a limit on the number of possible words. I could fix that by just making an abnormally large consonant inventory. Speaking both English and German, I can easily handle quite a few, but that would have a minor drawback in that it could make an English transliteration hard (how do you write three different r sounds?)

Before people say it, my language could fit an abjad quite well actually. Its phonotactics for example would make it pretty easy to guess where vowels have to go. In fact, for all of the possible combinations of 3 consonant words, there's only 1 or 2 possible arrangements of vowels. This doesn't give you the exact vowels of course, but that too could easily be guessed. Part of speech is one (all verbs in Hebrew only use one template for their infinite forms, regardless of which of the language's 7 conjugations it belongs to). Gender/noun classes could be used to help diversify vowel templates for nouns. You'd have to remember which gender each noun was, yes, but it's not like a lot of natlangs don't require that.

I'm still finding it limiting though not being able to use simple stem changes. I mean, Kebreni, a conlang that also uses this tri-consonantal roots thing obviously couldn't be written with an abjad. It relies solely on changing vowels to indicate the benefactive and anti-benefactive! Its a strategy is something I really like, since it prevents you having to lengthen words. Here though, I could not do such. I have thought about doing the matres lectionis, such as for vowel affixes, but I would like to make this actually a pure abjad if I could. Adding an extra symbol is fine I guess; no worse than adding a consonant symbol for any other affix. It could do cause some ambiguities; is that final w a syllable that begins with 'w' or is it an u suffix?

Clearly I need more techniques for morphology that would be compatible with such a system. Does anyone have any recommendations?

3 Answers 3


If the abjad has a rule where all words or syllables must start with a vowel, you might want to use a zero consonant in your conlang. However, if your conlang doesn't have that rule, it may not be crucial for your conlang.

EDIT: I used to have one on my conlang but it swiftly went in jeopardy.

  • I was already thinking that, yes. I was also thinking of using the symbol to indicate where vowels go in case its ambiguous. It could also be used to mark diphthongs, though both the latter uses could cause confusion. I was also thinking of using it just to indicate the number of syllables. Most words would have a number of syllables equal to the number of consonants minus 1. This however could also cause a fake out, because there would be exceptions (such as if all three consonants were a plosives). Not so sure about that. Point is, I was already thinking a lot about using such a letter.
    – user6046
    Jan 16 at 15:00

For instance, you can't make as much use out of stem changes. Think of the English words 'speak' and 'spoke'. If you wrote them with just consonants, then both would be 'spk'. Clearly the language can't have something like this.

To offer a counterexample, many conjugations in Ancient Egyptian worked like this, and thus were simply not differentiated in writing: sḏm can be infinitive sāḏam, subjunctive saḏma, participle sāḏim, the stative or imperative that I can't find vocalizations for…

These just had to be distinguished by context. The writing system used determinatives to distinguish between different lemmas with the same consonant, but made no attempt to distinguish different forms. And it worked well enough for thousands of years.

(For another counterexample, English-speakers can figure out if "read" is /rid/ or /rɛd/ by context.)

  • Regarding 'read', its normal convention to write the past tense using the perfect form. Thus instead of saying' 'I read' when you mean past tense, you say 'I have read'. Also, this isn't an ambiguity with a third person. I.e. 'He reads' vs 'He read'.
    – user6046
    Jan 3 at 0:16
  • @user8600 Some people might use a convention like that, but it's not necessary to be understood: sentences like "I read that already" are plenty common.
    – Draconis
    Jan 3 at 0:57
  • Still, I do know that this system doesn't fit the Semitic languages 100%. Arabic still has to indicate where long vowels go, and Hebrew does write some long vowels. Both also have optional markers for all vowels. My project is sorta a challenge; make a conlang where a pure abjad does work 100% reliably. That's obviously hard to do, especially since I clearly have no examples. I'm probably just going to have to keep morphology to a minimum (to minimize morphology that only involves vowels, minus the addition of V syllables of course).
    – user6046
    Jan 3 at 1:24
  • 1
    @user8600 Egyptian is an example of a "pure" abjad, insofar as it never indicates vowels in native words.
    – Draconis
    Jan 3 at 1:27
  • 1
    @user8600 Only to distinguish lemmas, though, not forms of a word. All forms of sḏm "hear" have an ear determinative, with no attempt made to indicate which conjugation of "hear" it is.
    – Draconis
    Jan 3 at 2:15

You have already explored some of the difficulties of matching a language with an abjad. It is of course not overly difficult to design a semitic-like conlang by studying Semitic languages and how they build up a huge number of forms from triconsonatal roots. Note that Semitic languages use more devices from just adding vowels inside the root, such as consonant gemmination and suffixes containing additional consonants. Note also that no vowel at all is an option, too, and that a vowel can be added at the beginning or the end of a word form. Radically deviating from that is not that easy: Two-consonantal roots are probably not enough to build the vocabulary of a full language but maybe still worth a try, for a minimalistic core vocabulary you may want to look at Toki Pona. Four-consonantal root are possible and an interesting field of investigation.

Than, your example of English speak, spoke shows another possibility: Indivisible consonant clusters (ICCs). You will have to fiddle with your abjad to mark those consonant cluster an to distinguish speak from seapk and I can envision a lot of creative possiblities here (adding separate characters for the ICCs, using ligatures, using some special cluster joiner or cluster divider characters like (sp)k vs. s(kp)).

  • I've been having a hard time finding info on Hebrew that isn't meant to teach the language. Most of what I know about this system comes from the conlang Kebreni, which doesn't work quite the same way. It avoids the infamous problem this system creates where you end up with atrocious consonant clusters everywhere. Kebreni mainly relies on messing with the vowels to mark things; only two possible markings actually involve the addition of phonemes. I could lay out how it all works here, but it would take me several comments to do so. It would probably be better just to look it up yourself.
    – user6046
    Jan 2 at 10:40
  • As for root size, I realize I'm going to mainly need 3 consonant roots. Even that though doesn't result in quite enough possible words, thus why I would need an abnormally large consonant inventory. Besides, that would sorta fit this. If there's only 3 vowels but over 30 consonants, why not focus on them? Also, I do know Hebrew uses at least one two-consonant affix. Its feminine marker. Arabic on the other hand apparently just lengthens the final vowel to mark the feminine. if I wanted all my affixes to be a single consonant, I also know that would put a severe limit on their number.
    – user6046
    Jan 2 at 10:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.