A guide to writing systems

[zompist]

Knots… the only knotted system I’m aware of is khipu, and that’s probably not even a full writing system.

Our best information is that it's a good spreadsheet system. That is, it represents numbers very well, plus the metadata that organizes them. E.g. we can identify raw data, subtotals, and totals. That in fact is a good reason to trust our decipherment of the numerals, because you wouldn't get neat sets of subtotals if (e.g.) we were erroneously reading lingustic text as numbers.

There are some knot combinations that aren't numbers, and there is non-knot information like color, and the structure of the strings. I don't think we'll ever find that it's a writing system for Quechua, but there are probably the beginnings of one, in e.g. using particular knot codes for place names.

Also, what are the four ways to use clay? Cuneiform is one, obviously, and presumably using a stylus is another. What are the others?

1. Dragging a pointer through the clay to draw images.
2. Pressing the stylus into the clay (cuneiform proper).
3. Pressing an object into the clay (e.g. signature seals).
4. Sculpting small objects (tokens— look up Denise Schmandt-Besserat)

1/2 were used as full writing systems. 3 was more specialized but note the resemblance to printing. 4 is a precursor, not unlike khipu in fact.

Precursors are important because writing systems, unlike languages, are things that develop in stages. So e.g. it's not naturalistic to posit a con-culture inventing an alphabet from scratch (rather than borrowing the idea from a neighboring culture).

[bradrn]

Well, that's what definitions usually do, right? When people bring up definitions, I always have to think of zompist's old Never define rant: http://www.zompist.com/rants05.html (It's the entry for April14th.)

Of course, but I’m using these mostly as jumping-off points for further discussion.

[*] Abugidas, in which consonantal graphemes can be used on their own or with a vocalic modifier to form syllabic blocks; and
...
Next up: abjads and alphabets, most probably. (Unless I change my mind, of course.)

You've fallen into the 'old men and women' ambiguity. Is a system which uses consonant graphemes with a vocalic modifier to form syllabic blocks and uses consonant graphemes on their own an abugida? I'm thinking of English and modern Lao.

It is ambiguous! There’s good arguments either way. (At least for Lao; I can’t see how English falls into this category.)

Knots… the only knotted system I’m aware of is khipu, and that’s probably not even a full writing system.

Our best information is that it's a good spreadsheet system.

This is a way of putting it I hadn’t seen before. I like this perspective!

Also, what are the four ways to use clay? Cuneiform is one, obviously, and presumably using a stylus is another. What are the others?

1. Dragging a pointer through the clay to draw images.
2. Pressing the stylus into the clay (cuneiform proper).
3. Pressing an object into the clay (e.g. signature seals).
4. Sculpting small objects (tokens— look up Denise Schmandt-Besserat)

1/2 were used as full writing systems. 3 was more specialized but note the resemblance to printing. 4 is a precursor, not unlike khipu in fact.

Ah, I thought you were talking only about full writing systems. My mistake!

Precursors are important because writing systems, unlike languages, are things that develop in stages. So e.g. it's not naturalistic to posit a con-culture inventing an alphabet from scratch (rather than borrowing the idea from a neighboring culture).

Of course. But on the other hand, I can just refer people to the ALC for that! In fact, I’m forced to do, since I don’t actually know very much about the origins and development of logographies. My personal interests lie more in the development and use of phonemic writing systems (is that a term?), so the scope of this thread will be mostly constrained to that area.

[Richard W]

[*] Abugidas, in which consonantal graphemes can be used on their own or with a vocalic modifier to form syllabic blocks; and
...
Next up: abjads and alphabets, most probably. (Unless I change my mind, of course.)

It is ambiguous! There’s good arguments either way. (At least for Lao; I can’t see how English falls into this category.)

E-n-g-li-s-h - 2 (C)V blocks, and C on its own for the codas and pre-initials.

Now, Lao is an alphasyllabary, though the subordinate status is not so certain for some of the vowel symbols. The Phags-pa and Semitic Mongolian writing systems are pretty much alphabets, despite having developed from an abugida and an abjad respectively. (Yiddish and Kurdish Semitic scripts preserve marks of their abjad ancestry, despite meeting the definition of an alphabet.)

[bradrn]

It is ambiguous! There’s good arguments either way. (At least for Lao; I can’t see how English falls into this category.)

E-n-g-li-s-h - 2 (C)V blocks, and C on its own for the codas and pre-initials.

But to what degree are these ‘blocks’ really features of the writing system itself? In Lao, the notion of a syllable block is strongly embedded in the writing system: C graphemes are central, and V graphemes are written around them, oftentimes out of linear order. By contrast, English is most straightforwardly analysed as sequences of letters written in the same linear order as sounds: there is nothing about, say, the sequence ⟨li⟩ which justifies treating

Now, Lao is an alphasyllabary, though the subordinate status is not so certain for some of the vowel symbols. The Phags-pa and Semitic Mongolian writing systems are pretty much alphabets, despite having developed from an abugida and an abjad respectively. (Yiddish and Kurdish Semitic scripts preserve marks of their abjad ancestry, despite meeting the definition of an alphabet.)

I plan to cover all of these, in time!

[Moose-tache]

[*] Abugidas, in which consonantal graphemes can be used on their own or with a vocalic modifier to form syllabic blocks; and
...
Next up: abjads and alphabets, most probably. (Unless I change my mind, of course.)

You've fallen into the 'old men and women' ambiguity. Is a system which uses consonant graphemes with a vocalic modifier to form syllabic blocks and uses consonant graphemes on their own an abugida? I'm thinking of English and modern Lao.

I'm slightly confused. Are you suggesting that an alphasyllabary is just another type of abjad? Or that abugidas, as defined, include alphabets?

[bradrn]

[*] Abugidas, in which consonantal graphemes can be used on their own or with a vocalic modifier to form syllabic blocks; and
...
Next up: abjads and alphabets, most probably. (Unless I change my mind, of course.)

You've fallen into the 'old men and women' ambiguity. Is a system which uses consonant graphemes with a vocalic modifier to form syllabic blocks and uses consonant graphemes on their own an abugida? I'm thinking of English and modern Lao.

I'm slightly confused. Are you suggesting that an alphasyllabary is just another type of abjad? Or that abugidas, as defined, include alphabets?

He’s talking about the many ambiguous cases lying between prototypical ‘abugidas’ and ‘alphabets’. In modern Lao, consonants have no inherent vowel: instead, all vowels are written. However, they’re written above and around the consonant letters, as in standard abugidas. I’m not sure what he’s getting at for English, though (for reasons detailed in my last post here).

[Moose-tache]

I think if vowel marks are obligatory, to me that just sounds like another kind of abjad.

I don't think we can say that langages that always pair C and V glyphs because of the language's phonology are alphasyllabaries or syllabaries or anything like that. At some point the usage of the script as defined by the language have to be separate from the rules of the script if we're going to study the script as its own entity. English also doesn't have word-initial ng, but that's not a feature of the English language version of the Latin alphabet.

[Richard W]

By contrast, English is most straightforwardly analysed as sequences of letters written in the same linear order as sounds: there is nothing about, say, the sequence ⟨li⟩ which justifies treating

So do you regard Meroitic as an abugidaʔ I would say that it wasn't an alphasyllabary. I think Kirsty Rowan's description on p30 of her Ph. D. thesis (https://eprints.soas.ac.uk/29209/1/10731304.pdf) is worth bearing mind:

In particular, the inclusion of a distinct set of separate vowel signs in the Meroitic script, and the small number of signs has probably caused the most confusion in its classification, as these give the ‘alphabetic’ appearance of the script and mask its syllabic based principle. Typologically, the Meroitic script is quite rare in its system of organisation, where the Old Persian Cuneiform script is perhaps its most typologically closest equivalent.

I don't think I'd come across the comparison with Old Persian cuneiform before, though I have read a paper of two of hers on Meroitic.

I'll post now as the thesis is a long read.

[bradrn]

I think if vowel marks are obligatory, to me that just sounds like another kind of abjad.

Except it would be a weird abjad, because (a) abjads are defined by not writing the vowels obligatorily, and (b) the vowel marks in Lao are often written in line with the rest of the text, bipartitely in many cases. For instance, /eː/ is ເ◌, /a/ is ◌ະ, /e/ is ເ◌ະ.

I don't think we can say that langages that always pair C and V glyphs because of the language's phonology are alphasyllabaries or syllabaries or anything like that. At some point the usage of the script as defined by the language have to be separate from the rules of the script if we're going to study the script as its own entity. English also doesn't have word-initial ng, but that's not a feature of the English language version of the Latin alphabet.

I’m not going off the phonology. I’m saying that because of all the vowels which literally get written as diacritics above or below their onset consonant, as well as the remaining vowels which get written before or after the consonant (or both). I don’t see any way to analyse this situation which is not in terms of CV blocks.

By contrast, English is most straightforwardly analysed as sequences of letters written in the same linear order as sounds: there is nothing about, say, the sequence ⟨li⟩ which justifies treating

So do you regard Meroitic as an abugidaʔ

I regard Meroitic as yet another in-between system. In fact, it’s almost the reverse of the Lao case: modern Lao is nonlinear but writes every vowel, yet Meroitic is linear while having an inherent vowel. Neither can be unambiguously assigned to one of the ‘traditional’ categories, though like I said I’m inclined to define ‘semi-syllabary’ as a new basic category. (Neither quite fits the ‘abugida’ case, since the prototypical abugida has inherent vowels and writes the remaining vowels nonlinearly.)

(Incidentally, this conversation is giving me some interesting ideas about classification, so thanks!)

[Richard W]

[*] Abugidas, in which consonantal graphemes can be used on their own or with a vocalic modifier to form syllabic blocks; and
...
Next up: abjads and alphabets, most probably. (Unless I change my mind, of course.)

You've fallen into the 'old men and women' ambiguity. Is a system which uses consonant graphemes with a vocalic modifier to form syllabic blocks and uses consonant graphemes on their own an abugida? I'm thinking of English and modern Lao.

I'm slightly confused. Are you suggesting that an alphasyllabary is just another type of abjad? Or that abugidas, as defined, include alphabets?

I'm suggesting that the second interpretation, corresponding to "(old men) and women", does include alphabets, so I think it shouldn't be what is meant. One problem with the definitions involving syllables is that they only really address languages with just the syllable type CV, and leave lots of issues when dealing with a syllable structure CCVC, which gets addressed in many ways.

I wasn't saying that an alphasyllabary is just another type of abjad. It depends whether Hebrew written with vowel points is to be described as being written with an 'abjad'. Daniels refused to address the typological classification of vocalised abjads and other 'impure abjads'. I do happen to think that compulsorily vocalised abjads, with the vowel marks subordinate to the consonant, are a type of alphasyllabary. The Mongolian script with script code Mong is an example of where using matres lectionis has turned an abjad into an alphabet. Curiously, the Mongols learn it as a syllabary, even though it represents consonant clusters the way a typical Greek-derived alphabet does.

[bradrn]

I’ve just realised that we do in fact have another extant script similar to modern Lao: namely, Hangeul! Both are nonlinear yet write all consonants and vowels, though of course Hangeul takes the nonlinearity further than Lao does.

You've fallen into the 'old men and women' ambiguity. Is a system which uses consonant graphemes with a vocalic modifier to form syllabic blocks and uses consonant graphemes on their own an abugida? I'm thinking of English and modern Lao.

I'm slightly confused. Are you suggesting that an alphasyllabary is just another type of abjad? Or that abugidas, as defined, include alphabets?

I'm suggesting that the second interpretation, corresponding to "(old men) and women", does include alphabets, so I think it shouldn't be what is meant.

OK, I didn’t realise this was what you meant. But I think the mention of ‘syllabic blocks’ disqualifies most alphabets, barring edge cases like Mongolian.

(I should note that I’m adopting an emic approach here, where I define these notions script-internally. As you noted, it’s possible to define a notion of ‘syllabic blocks’ for English, but that’s completely useless for analysing how English works, so I feel comfortable saying that alphabets like English don’t really have syllable blocks.)

[Richard W]

I think if vowel marks are obligatory, to me that just sounds like another kind of abjad.

Except it would be a weird abjad, because (a) abjads are defined by not writing the vowels obligatorily, and (b) the vowel marks in Lao are often written in line with the rest of the text, bipartitely in many cases. For instance, /eː/ is ເ◌, /a/ is ◌ະ, /e/ is ເ◌ະ.

Nitpick: ເ◌ະ is /eʔ/. Otherwise, /e/ is ເ◌ັ◌, where the dotted circles represent mandatory consonants. A graphically clustered consonant is possible for the first slot, e.g. ເຫງັນ 'civet'.

By contrast, English is most straightforwardly analysed as sequences of letters written in the same linear order as sounds: there is nothing about, say, the sequence ⟨li⟩ which justifies treating

So do you regard Meroitic as an abugidaʔ

I regard Meroitic as yet another in-between system. In fact, it’s almost the reverse of the Lao case: modern Lao is nonlinear but writes every vowel, yet Meroitic is linear while having an inherent vowel. Neither can be unambiguously assigned to one of the ‘traditional’ categories, though like I said I’m inclined to define ‘semi-syllabary’ as a new basic category. (Neither quite fits the ‘abugida’ case, since the prototypical abugida has inherent vowels and writes the remaining vowels nonlinearly.)

'Semisyllabary' means a system with a mix of phonetic syllables and alphabetic symbols. I think the Iberian alphabets are the prototypical examples. Not all abugidas have higgledy-piggledy vowel marks. It's largely a feature inherited from the early Indic alphabets. Canadian Syllabics are orderly, though I don't like calling them an abugida. Meroitic seems to be designed much the same ways as Brahmi or Kharoshthi - strike out the commonest vowel symbol as unnecessary, though for Meroitic it seems to give only a 25% saving.

[Richard W]

I’ve just realised that we do in fact have another extant script similar to modern Lao: namely, Hangeul! Both are nonlinear yet write all consonants and vowels, though of course Hangeul takes the nonlinearity further than Lao does.

Closer to home, we have the Latin script, e.g. in 17th century English with bars through the ascenders and descenders of 'b', 'd' and 'p' to abbreviate 'er' or 'ar'. and a similar device for 'ro' through the letter 'p', and the origin of the combining tilde as an abbreviation for a following 'm' or 'n'. And some Germanic languages had vertically stacked vowels.

I'm not sure why you think Hangeul takes non-linearity further. It's block by block, top left, right, below, further below - just a series of semi-circles. Lao has tone marks above, whereas the vowel may be below, not just to the left and/or above and/or right. And Hangeul is perfectly linear in NFC backing store.

[bradrn]

I think if vowel marks are obligatory, to me that just sounds like another kind of abjad.

Except it would be a weird abjad, because (a) abjads are defined by not writing the vowels obligatorily, and (b) the vowel marks in Lao are often written in line with the rest of the text, bipartitely in many cases. For instance, /eː/ is ເ◌, /a/ is ◌ະ, /e/ is ເ◌ະ.

Nitpick: ເ◌ະ is /eʔ/. Otherwise, /e/ is ເ◌ັ◌, where the dotted circles represent mandatory consonants.

Ah, thanks! (I know I can count on you to correct me on Indic scripts )

A graphically clustered consonant is possible for the first slot, e.g. ເຫງັນ 'civet'.

I’m aware.

So do you regard Meroitic as an abugidaʔ

I regard Meroitic as yet another in-between system. In fact, it’s almost the reverse of the Lao case: modern Lao is nonlinear but writes every vowel, yet Meroitic is linear while having an inherent vowel. Neither can be unambiguously assigned to one of the ‘traditional’ categories, though like I said I’m inclined to define ‘semi-syllabary’ as a new basic category. (Neither quite fits the ‘abugida’ case, since the prototypical abugida has inherent vowels and writes the remaining vowels nonlinearly.)

'Semisyllabary' means a system with a mix of phonetic syllables and alphabetic symbols. I think the Iberian alphabets are the prototypical examples.

On reflection, I got confused by the fact that Tartessian uses CV-V combinations similar to Old Persian and Meroitic, and I incorrectly assumed that’s what was meant by a ‘semi-syllabary’. But I see now that the other Paleohispanic scripts are genuine syllabary–alphabet combinations.

Not all abugidas have higgledy-piggledy vowel marks. It's largely a feature inherited from the early Indic alphabets. Canadian Syllabics are orderly, though I don't like calling them an abugida. Meroitic seems to be designed much the same ways as Brahmi or Kharoshthi - strike out the commonest vowel symbol as unnecessary, though for Meroitic it seems to give only a 25% saving.

(Did you perhaps mean ‘early Indic abugidas’?)

You can add Ethiopic to the list — if anything, its vowels are even more higgledy-piggledy than Indic ones. But yes, if you relax the condition that vowels should be diacritical (to form graphical syllabic blocks), then Meroitic certainly qualifies as an abugida.

Ultimately, what this comes down to is that the usual classification just isn’t very good — it ignores some distinctions while leaving other systems unclassified or ambiguous. But I’m already getting some ideas around what a better typology might look like, and I’ll probably post it here once I have a firmer idea about the details.

Closer to home, we have the Latin script, e.g. in 17th century English with bars through the ascenders and descenders of 'b', 'd' and 'p' to abbreviate 'er' or 'ar'. and a similar device for 'ro' through the letter 'p', and the origin of the combining tilde as an abbreviation for a following 'm' or 'n'. And some Germanic languages had vertically stacked vowels.

Looking even earlier, medieval script was chock-full of scribal abbreviations, which is why record type exists. But this feels to me like a slightly different flavour of nonlinearity, if that makes any sense.

I'm not sure why you think Hangeul takes non-linearity further. It's block by block, top left, right, below, further below - just a series of semi-circles. Lao has tone marks above, whereas the vowel may be below, not just to the left and/or above and/or right.

My reasoning was that, in Lao, the whole text is still effectively a single line — some marks are positioned relatively above, below or to the side of the consonantal letters, but when you read it, you’re still scanning it from left to right with only brief deviations. By contrast, a Hangeul text requires every single syllable block to be read semicircularly, going from top to bottom and left to right to left, before jumping back to the top of the next block.

But of course this is subjective, and something about which reasonable people can disagree.

And Hangeul is perfectly linear in NFC backing store.

For my purposes I’ll be largely ignoring computer encoding in this thread, for all that it’s a fascinating and intricate topic. (Though I might write a post about it at some point.)

[zompist]

My reasoning was that, in Lao, the whole text is still effectively a single line — some marks are positioned relatively above, below or to the side of the consonantal letters, but when you read it, you’re still scanning it from left to right with only brief deviations. By contrast, a Hangeul text requires every single syllable block to be read semicircularly, going from top to bottom and left to right to left, before jumping back to the top of the next block.

This strikes me as a weird description based on outsider assumptions. Surely you could use an eye-tracker or something to see how Koreans actually read; do you think they are moving their eyes in a circular motion? I'd guess the eye moves in a straight line, reading syllables mostly as a gestalt... as in fact English readers do. Besides...

w h a t

i f

w e

w r o t e

l i k e

t h i s?

That's also a linear order, with a sublinear order for each syllable. Granted that it's new and unfamiliar, it doesn't strike me as hard to read, and if that was how your writing system works you'd be nonplussed by a description of how you have to switch between left-right and top-down orientation for every word. That might describe the logical structure, but the brain might be doing things quite differently. Very often learning something means getting most of the details into parts of the brain we don't consciously access.

[bradrn]

My reasoning was that, in Lao, the whole text is still effectively a single line — some marks are positioned relatively above, below or to the side of the consonantal letters, but when you read it, you’re still scanning it from left to right with only brief deviations. By contrast, a Hangeul text requires every single syllable block to be read semicircularly, going from top to bottom and left to right to left, before jumping back to the top of the next block.

This strikes me as a weird description based on outsider assumptions. Surely you could use an eye-tracker or something to see how Koreans actually read; do you think they are moving their eyes in a circular motion? I'd guess the eye moves in a straight line, reading syllables mostly as a gestalt... as in fact English readers do.

OK, so perhaps I was wrong about the biomechanics of it. All the same, purely considering how the letters are organised with respect to each other, my point is still correct.

Perhaps it would be clearer if I illustrated it graphically:



I hope it should be obvious here that Hangeul letters are arranged less linearly, in that they take full advantage of the space provided by the syllable block, whereas the Lao text is mostly just letters in sequence, even if there’s little bumps and loopbacks here and there. (It helps that only two vowel components can go left of the consonant, something I didn’t fully notice before.)

[zompist]

I hope it should be obvious here that Hangeul letters are arranged less linearly, in that they take full advantage of the space provided by the syllable block, whereas the Lao text is mostly just letters in sequence, even if there’s little bumps and loopbacks here and there. (It helps that only two vowel components can go left of the consonant, something I didn’t fully notice before.)

I got your point, though I do thank you for the nice illustration from Lao, which I don't know. (I'm familiar with vowel signs appearing before or on both sides of the consonant from devanagari, though.)

As I said, structurally you're right, but putting elegance of structure above psychology is how you get Chomskyan UG.

Maybe English examples would be clearer. Reading $100, do you skip over the out-of-place $ (brain: uninterpretable!), then read the number, then move back to the $? I'm guessing you don't, and an eye-tracker would confirm this. You read "dollar-100" and say "100 dollars" without even thinking about it.

Your red lines also assume that (say) a German speaker reading "München" reads the u, then bops up to the umlaut, then bops down to continue rightward. But maybe they start with the umlaut and read the u next. Or maybe they absorb ü as a single unit. Or "München" as a single unit!

This isn't a huge point, but if you're trying to come up with a new analysis to replace Daniels, I think it shouldn't be based on impressionistic ideas of linearity.

(For that matter, some early writing systems, especially Sumerian, were pretty cavalier about linearity, and if you've watched little kids write, they can be too. There are probably good reasons writing systems end up mostly linear! But we should be wary of over-emphasizing a feature that happens to be particularly characteristic of our own alphabet.)

[bradrn]

Your red lines also assume that (say) a German speaker reading "München" reads the u, then bops up to the umlaut, then bops down to continue rightward. But maybe they start with the umlaut and read the u next. Or maybe they absorb ü as a single unit. Or "München" as a single unit!

Ah, I think I see the confusion. My red lines were not aimed at demonstrating any kind of psychological phenomenon. They’re merely showing the mapping from the positions of the glyphs to the linear sequence of spoken phonemes. I’m not trying to ‘put elegance of structure above psychology’ at all: rather, I’m approaching this from a purely descriptive perspective, which is how I tend to approach linguistics generally.

Maybe English examples would be clearer. Reading $100, do you skip over the out-of-place $ (brain: uninterpretable!), then read the number, then move back to the $? I'm guessing you don't, and an eye-tracker would confirm this. You read "dollar-100" and say "100 dollars" without even thinking about it.

This is of course correct. Yet, let’s see how Wikipedia describes its use:

In [many countries], the sign is written before the number ("$5"), even though the word is written or spoken after it ("five dollars", "cinco pesos").

This is pretty much all I’m trying to give: a description of how the glyphs in the writing system correspond to spoken language.

[zompist]

In [many countries], the sign is written before the number ("$5"), even though the word is written or spoken after it ("five dollars", "cinco pesos").

This is pretty much all I’m trying to give: a description of how the glyphs in the writing system correspond to spoken language.

Here at least Wikipedia does describe things neutrally. When you use words like "bumps and loopbacks" it does sound like you're describing movement... this is a very common metaphor ("the woods run along the river") but I'd just try to avoid it as it sounds like it's making claims you don't want to make.

[bradrn]

In [many countries], the sign is written before the number ("$5"), even though the word is written or spoken after it ("five dollars", "cinco pesos").

This is pretty much all I’m trying to give: a description of how the glyphs in the writing system correspond to spoken language.

Here at least Wikipedia does describe things neutrally. When you use words like "bumps and loopbacks" it does sound like you're describing movement... this is a very common metaphor ("the woods run along the river") but I'd just try to avoid it as it sounds like it's making claims you don't want to make.

Fair enough, thanks for the suggestion!