A guide to writing systems

[bradrn]

Interesting, thanks! Do you know if there are any old borrowings which still lack ⟨h⟩?

Yeah, there were transliterations like Corinto for Corinth. More interesting are ones that survived, like purpura (source of 'purple'), calx, Punicus, which by classical standards should have been porphyra, chalix, phoenicus. (The latter was of course borrowed as is, which gives us the doublet Punic/Phoenician.)

Indeed, I was asking about the ones that survived. Note that purple also has a doublet porphyry, still a standard geological term.

4. You could add <g> as a marker of palatalization in Italian: bagno, degli.

EDIT: actually, on second thoughts, where did this originate from? Wikipedia has no information.

This is based on some cursory googling, but note that Italian <gn> can come from 1) <gn> as in agnellus 'little lamb' > agnello, 2) /nj/ as in vinea > vigna, balneum > bagno. Allen tells us that <gn> was pronounced [ŋn]. If that went to [nʲ] by sound change, then <gn> could be used for other instances of nʲ, then applied to lʲ by analogy.

Hmm, your comment about /nj/ sounds familiar for some reason… I’m sure I read something about this the other day. Maybe I’ll be able to find it again.

[Darren]

Allen tells us that <gn> was pronounced [ŋg].

I think that should be [ŋn].

[zompist]

Allen tells us that <gn> was pronounced [ŋg].

I think that should be [ŋn].

Yes, you're right.

[anteallach]

That sounds reasonable to me. (In fact, I don’t think I mentioned vowel-modifying ⟨h⟩ at all.)

Of course it also occurs in German, where in some words it derives from a real consonant but it has been generalised to words where it doesn't.

The most straightforward violation of the definition occurs when single graphemes denote phoneme sequences. This seems to occur most commonly in Greek and derived alphabets. The Ancient Greek alphabet includes ⟨ζ ψ ξ⟩ /zd~dz ps ks/, though in Modern Greek ⟨ζ⟩ has been simplified to /z/. Coptic has inherited both ⟨ⲯ ⲝ⟩ /ps ks/, as well as deriving ⟨ⲋ⟩ /dz/ from an ⟨στ⟩ ligature, and ⟨Ϯ⟩ /ti/ from Demotic Egyptian. Latin inherited ⟨x⟩ /ks/ from a local Greek variant, but lost ⟨ξ⟩. Additionally, the unrelated Armenian alphabet has ⟨և⟩ /(j)ɛv/, derived from a ligature of ⟨եւ⟩.

I suspect that, regardless of its later value, in the earliest versions of the Greek alphabet ⟨ζ⟩ represented an affricate best regarded as a single phoneme. There is also a Z series in Linear B, but there doesn't seem to be anything like the other two.

[WeepingElf]

Never forget that writing systems are usually not designed by phonologists. People sometimes have the oddest notions of what represents a "sound unit" or whatever worthy of a grapheme of its own.

[masako]

Never forget that writing systems are usually not designed by phonologists. People sometimes have the oddest notions of what represents a "sound unit" or whatever worthy of a grapheme of its own.

I think that when being taught "to read", many folks just integrate that a certain symbol can only be used in the representation of one (or maybe a few) sound(s), and that anything outside those boundaries just feels wrong.

"T" doesn't inherently mean anything, most groups that use it do use it for something similar, but you take something like "X", "E", or even "J", and folks will argue for days.

[bradrn]

That sounds reasonable to me. (In fact, I don’t think I mentioned vowel-modifying ⟨h⟩ at all.)

Of course it also occurs in German, where in some words it derives from a real consonant but it has been generalised to words where it doesn't.

I know nothing about German; could you elaborate please?

The most straightforward violation of the definition occurs when single graphemes denote phoneme sequences. This seems to occur most commonly in Greek and derived alphabets. The Ancient Greek alphabet includes ⟨ζ ψ ξ⟩ /zd~dz ps ks/, though in Modern Greek ⟨ζ⟩ has been simplified to /z/. Coptic has inherited both ⟨ⲯ ⲝ⟩ /ps ks/, as well as deriving ⟨ⲋ⟩ /dz/ from an ⟨στ⟩ ligature, and ⟨Ϯ⟩ /ti/ from Demotic Egyptian. Latin inherited ⟨x⟩ /ks/ from a local Greek variant, but lost ⟨ξ⟩. Additionally, the unrelated Armenian alphabet has ⟨և⟩ /(j)ɛv/, derived from a ligature of ⟨եւ⟩.

I suspect that, regardless of its later value, in the earliest versions of the Greek alphabet ⟨ζ⟩ represented an affricate best regarded as a single phoneme.

Oh, probably, but it seems likely that it represented two phonemes at some stage.

There is also a Z series in Linear B, but there doesn't seem to be anything like the other two.

But Linear B is a syllabary, so we would expect a single grapheme to denote multiple segments in any case.

Never forget that writing systems are usually not designed by phonologists. People sometimes have the oddest notions of what represents a "sound unit" or whatever worthy of a grapheme of its own.

Of course… but at the same time, they rarely seem to diverge too far from the ideal phonemic analysis.

[zompist]

I suspect that, regardless of its later value, in the earliest versions of the Greek alphabet ⟨ζ⟩ represented an affricate best regarded as a single phoneme. There is also a Z series in Linear B, but there doesn't seem to be anything like the other two.

I'm afraid I don't think this is a principle at all. Early scripts are notorious for being terrible representations of the phonology. (Examples: not writing vowels; not indicating length; not indicating syllable finals; indicating only certain finals; merging places of articulation.) Allen makes the case for ζ being zd and it's a good one, but it's a very odd choice (on the part of the Greeks) when there was no symbol for st.

[masako]

Allen makes the case for ζ being zd and it's a good one, but it's a very odd choice (on the part of the Greeks) when there was no symbol for st.

I mean, they're Greek, kinda their thing.

Seriously though, it speaks to the arbitrary nature of what writing can be used to represent. While we have categories and classifications of systems, writing is just as idiosyncratic as any other aspect of language. While there are trends and patterns, there are also exceptions. Usually quite a few.

[xxx]

just look at the correspondence between the roman alphabet in the languages that use it and the attached sounds...

we could use the IPA,
but we'd need different scripts in the same country using the same language...
the opposite of the logographies used by different countries...

as we've seen, reading is logographic, even for alphabets; sticking to pronunciation would paralyze reading...

from there to think of a naturalness, if not a superiority, of logographies over blind scripts, which run after phonetic transcription without ever reaching it...

[WeepingElf]

Some writing systems make interesting choices what to assign graphemes to. The two Japanese syllabaries write morae. The Paleo-Hispanic scripts assign letters to what has duration: there are individual letters for vowels and for sibilants, nasals and liquids, but not for stops alone, only for stop+vowel combinations.

[bradrn]

Some writing systems make interesting choices what to assign graphemes to. The two Japanese syllabaries write morae. The Paleo-Hispanic scripts assign letters to what has duration: there are individual letters for vowels and for sibilants, nasals and liquids, but not for stops alone, only for stop+vowel combinations.

Rest assured, in time I will cover all this and more!

[anteallach]

That sounds reasonable to me. (In fact, I don’t think I mentioned vowel-modifying ⟨h⟩ at all.)

Of course it also occurs in German, where in some words it derives from a real consonant but it has been generalised to words where it doesn't.

I know nothing about German; could you elaborate please?

German has vowel+h digraphs as a regular way of writing long vowels, as in Zahn (tooth) and zehn (ten). As I understand it, the <h> in the latter derives from an actual historic /h/, but in Zahn it doesn't, and just indicates the long vowel.

[bradrn]

Typical and atypical alphabets

Beyond the definition given in the last post, most alphabets share a number of other properties. In my opinion, the following ones are particularly prominent:

1. Graphemes are arranged in a single line in one direction
2. Consonants and vowels are graphically equal
3. Each phoneme corresponds to one grapheme

These properties do seem to be widely and strongly associated with alphabets in particular. For instance, WWS has no qualms with calling Meroitic ‘essentially alphabetic’ , even though it’s quite straightforwardly an abugida — presumably because it satisfies both (1) and (2) above. Similarly, Wikipedia calls 'Phags-pa an alphabet without qualification, despite its obviously abugidic nature. (And for that matter, Pollard — which I’ll describe below — gets called an abugida even though it’s an alphabet! Specifically, an alphabet which violates both (2) and (1).)

On the other hand, there exist alphabets which do not satisfy all of these properties. Such alphabets differ interestingly from the norm, so it’s worth having a closer look at them.

(Given previous discussion, I should probably make myself completely clear at this point: I claim no particular psychological effects to points (1), (2) or (3). They’re interesting purely because they show the range of diversity which gets subsumed within the term ‘alphabet’.)

Let’s start with (1): the property of linearity. When using an alphabet, this has an obvious functional motivation — after all, if you’re writing out every phoneme separately, it makes sense to write them out in order.

The most famous example of a nonlinear script is of course Hangeul, used for Korean. This is indisputably an alphabet, and a very good one at that: the correspondence between graphemes and phonemes is near-perfect. However, the letters (in Korean, jamo) are not written out linearly, but collected into distinct syllable blocks. For instance, the jamo ⟨ㅎㅏㄴ⟩, spelling out han, are written in a single block as ⟨한⟩. In general, the first consonant (or ⟨ㅇ⟩ for a null onset) is placed at the top-left, followed by the vowel to the right and/or below. A final consonant goes at the bottom if present.

Hangeul’s nonlinearity is exacerbated by the fact that its writing direction has switched from top-to-bottom to left-to-right. (Thanks to Richard W for this observation!) Hangeul syllable blocks are essentially constructed vertically: a word such as ⟨국문⟩ gungmun would be completely linear if the syllables are written top-to-bottom, as was previously the case. Nonetheless, horizontally-formed syllables as in ⟨사라미⟩ sarami are also common, as are syllables with a mix of directions. Furthermore — excepting the work of typographers such as Ahn Sang Soo — jamo are invariably stretched and squeezed to form square blocks, which further deemphasises any linearity which might be present.

Next, (2). This one also makes obvious sense in an alphabet: if all phonemes are regularly written as separate graphemes, there’s not much motivation to treat consonants and vowels differently within the script.

The most striking counterexample to this I know is Thaana, the writing system of the Dhivehi language (the Indo-Aryan language of the Maldives). In Thaana, only consonants are written as full letters. Meanwhile, vowels are written as diacritics on the previous consonant. Every consonant must have a diacritic: those which have no following vowel receive the sukun diacritic, a small ring. Similarly, a vowel without a preceding consonant can take the alifu ⟨އ⟩ as base. The alifu+sukun combination ⟨އް⟩ is used to mark gemination of the following consonant (and reportedly glottal stop as well, though I see no evidence of this).

The net impression of this writing system is a line of consonants with another line of vowels above (plus a few below). Using the UDHR as an example (source: Wikipedia):

ހުރިހާ އިންސާނުން ވެސް އުފަންވަނީ، ދަރަޖަ އާއި ޙައްޤު ތަކުގައި މިނިވަންކަމާއި ހަމަހަމަކަން ލިބިގެންވާ ބައެއްގެ ގޮތުގައެވެ.

ހު

hu

ރި

ri

ހާ

haː

 

އި

i

ން

n

ސާ

saː

ނު

nu

ން

n

 

ވެ

ve

ސް

s

 

އު

u

ފަ

fa

ން

n

ވަ

va

ނީ،

niː

 

ދަ

da

ރަ

ra

ޖަ

dʒa

 

އާ

aː

އި

i

 

ޙަ

ħa

އް

ː

ޤު

qu

 

ތަ

ta

ކު

ku

ގަ

ɡa

އި

i

 

މި

mi

ނި

ni

ވަ

va

ން

n

ކަ

ka

މާ

maː

އި

i

 

ހަ

ha

މަ

ma

ހަ

ha

މަ

ma

ކަ

ka

ން

n

 

ލި

li

ބި

bi

ގެ

ɡe

ން

n

ވާ

va

 

ބަ

ba

އެ

e

އް

ː

ގެ

ɡe

 

ގޮ

ɡo

ތު

tu

ގަ

ɡa

އެ

e

ވެ.

ve

Transcription: hurihā insānun ves ufanvanī, darajaʾāi ḥaqqu takugai minivankamāi hamahamakan libigenvā baʾegge gotugaʾeve.

(Incidentally, I can’t resist sharing Viethaana, a thoroughly creative Thaana font which recasts it as Latin letters. I have no idea how this compares to the experience of actually reading Thaana, though. Probably poorly!)

Because its letters are not arranged linearly, Thaana violates property (1) in addition to (2). However, it is possible for a writing system to have (2) only: that is, it may treat consonants and vowels unequally, but still write them in a single line.

One near-example of such a script is the Pollard script, invented to write A-Hmao (a.k.a. ‘Large Flowery Miao’), a member of the Hmong–Mien family spoken in China. In Pollard, consonants are large and take up the full height of a line, whereas the syllable rime is written using clearly smaller graphemes (occasionally merged into a single letterform). Possibly uniquely amongst scripts, tone is represented by positioning these letters in different positions relative to the baseline. But most of the time the graphemes within a word remain almost linearly arranged — e.g. ⟨𖼊𖽺𖾐 𖼣𖽔 𖼊𖽱𖾐⟩. Alas, Pollard is not a perfect example: high tone is represented by writing the vowel above the consonant, as in ⟨𖼽𖽔𖾑⟩, and apparently in some usages there’s also a fifth tone written below the consonant.

Finally, there’s property (3): that each phoneme is written with one grapheme, i.e. that the script is phonemic. This is the converse of the definition I gave for alphabets, namely that each grapheme transcribes one phoneme. I won’t go into much detail about violations of this property, for the simple reason that I’m writing in English — if you can read this post, I assume you’re already well-acquainted with how convoluted a non-phonemic alphabet can get! But I’ll probably write a bit about non-phonemic orthographies at some stage.

[masako]

The most famous example of a nonlinear script is of course Hangeul, used for Korean.

How is Hangul not linear? You cannot read a syllable in any way but TBLR. Or are you equating "linear" with only in a ""straight line""?

This is indisputably an alphabet, and a very good one at that: the correspondence between graphemes and phonemes is near-perfect.

When you say "near-perfect", are you including how Hangul is read across syllables? I would hardly call that "correspondence between graphemes and phonemes" "near-perfect", but ok.

However, the letters (in Korean, jamo) are not written out linearly, but collected into distinct syllable blocks.

And those blocks are indeed read linearly, unless you know something new.

Hangeul’s nonlinearity is exacerbated by the fact that its writing direction has switched from top-to-bottom to left-to-right.

Both ways of reading/writing are still in use.

Once again, folks out here playing-up Hangul like it's a magical unique thing. It's a beautiful system, sure, but c'mon folks, it's just an alphabet.

It also helps to use terminology that is contextual to orthography.

[bradrn]

The most famous example of a nonlinear script is of course Hangeul, used for Korean.

How is Hangul not linear? You cannot read a syllable in any way but TBLR. Or are you equating "linear" with only in a ""straight line""?

Yes, by ‘linear’ I mean only ‘in a straight line’. It seems that a lot of people dislike it when I do this, so please feel free to suggest a better term if you have one!

This is indisputably an alphabet, and a very good one at that: the correspondence between graphemes and phonemes is near-perfect.

When you say "near-perfect", are you including how Hangul is read across syllables? I would hardly call that "correspondence between graphemes and phonemes" "near-perfect", but ok.

Yes, which is why it’s ‘near-perfect’ rather than ‘perfect’. There are very few perfect alphabets in existence. (Possibly Finnish? I don’t actually know.)

Hangeul’s nonlinearity is exacerbated by the fact that its writing direction has switched from top-to-bottom to left-to-right.

Both ways of reading/writing are still in use.

I don’t actually know anything about this, beyond the fact that I most commonly see Hangeul written left-to-right. (Unlike Chinese, where I do now and again see long-form text written vertically.) Moose-tache may be more qualified than I am to comment about this.

It also helps to use terminology that is contextual to orthography.

Hmm, that’s a different use of the word ‘linearity’. I can see how that could cause confusion — but again if you have a better suggestion I’m happy to use it.

[Richard W]

This is indisputably an alphabet, and a very good one at that: the correspondence between graphemes and phonemes is near-perfect.

When you say "near-perfect", are you including how Hangul is read across syllables? I would hardly call that "correspondence between graphemes and phonemes" "near-perfect", but ok.

Yes, which is why it’s ‘near-perfect’ rather than ‘perfect’. There are very few perfect alphabets in existence. (Possibly Finnish? I don’t actually know.)

I think there is a natural tendency to split words up into the contiguous parts of morphemes and write the parts. There may be an issue with phonetic allomorphs, and it does seem natural to write suppletive allomorphs differently. It has been suggested that the English past tense forms 'learnt' and 'learned' have slightly different meanings, directly related to mild phonetic symbolism. The battle to respect phonetic syllable boundaries in Hangeul seems to have been lost, and I think lost in the past century.

Hangeul’s nonlinearity is exacerbated by the fact that its writing direction has switched from top-to-bottom to left-to-right.

Both ways of reading/writing are still in use.

It also helps to use terminology that is contextual to orthography.

Hmm, that’s a different use of the word ‘linearity’. I can see how that could cause confusion — but again if you have a better suggestion I’m happy to use it.

In a sense, Hangeul is linear like proteins are mostly linear - it just folds up on itself locally.

[masako]

It seems that a lot of people dislike it when I do this, so please feel free to suggest a better term if you have one!

My preferences are really irrelevant. What I'd suggest though, is that perhaps use a conventual mode of discussing how systems are written/read, as in; Left-to-right (LR); top-to-bottom (TB); boustrophedon, etc. Most of the time - especially when talking about Hangul for some ridiculous reason - folks try to reinvent the orthographic vocabulary instead of just using what's been established for quite a while.

There are very few perfect alphabets in existence.

If by "perfect alphabet" you mean phonemic, the answer is zero. Always has been, always will be. A 1-to-1 phonemic correspondence from phoneme to grapheme is a fever-dream that some Lojbanist had and it's spread like the Bubonic plague. It simply does not exist. Let it go. Zero.

if you have a better suggestion I’m happy to use it

My suggestion is very simple, in fact. Just do loads more reading before trying to rewrite the books that have been out there for decades. Really, you've got a lot of good stuff, but you're repaving the roads to Rome.

EDIT:

In a sense, Hangeul is linear like proteins are mostly linear - it just folds up on itself locally.

That is a vivid way to describe it, but less poetically, you could, just, y'know, call it what it is; an alphabet (with a few abugidic elements) that is mostly LR but can be TB.

[bradrn]

In a sense, Hangeul is linear like proteins are mostly linear - it just folds up on itself locally.

I’m not sure this is the best comparison — proteins do not just fold up locally (‘secondary structure’), but also those structures combine to create the larger-scale 3D structure (‘tertiary structure’) which directly allows proteins to do what they do.

It seems that a lot of people dislike it when I do this, so please feel free to suggest a better term if you have one!

My preferences are really irrelevant. What I'd suggest though, is that perhaps use a conventual mode of discussing how systems are written/read, as in; Left-to-right (LR); top-to-bottom (TB); boustrophedon, etc.

Except those terms are conventionally used to describe arrangement at a higher level than graphemes, which is not what the previous post was about. (I plan to post more about writing direction later.)

And, on that note:

you could, just, y'know, call it what it is; an alphabet (with a few abugidic elements) that is mostly LR but can be TB.

As those terms are usually used, an example of such would be Latin, which is mostly LR but can be TB (e.g. on book spines or street signs). The previous post mostly discussed the arrangements of graphemes to form syllable blocks, as in Hangeul and Thaana and Pollard (and Lao and Zoulai, let me add).

Besides, even if I grant for the sake of argument your usage of ‘LR’ and ’TB’, this still misses something… it’s not just that Hangeul can be TB, it’s that graphemes are forced to be written horizontally or vertically in specific circumstances. Not only that, often both directions are used at once, as in ⟨한⟩. Note that I don’t consider Hangeul particularly special in this regard: similar situations are found in Lao and the other scripts I mentioned. But it’s a clearly different story to scripts like Latin, which pick one direction and stick to it for the rest of the text.

if you have a better suggestion I’m happy to use it

My suggestion is very simple, in fact. Just do loads more reading before trying to rewrite the books that have been out there for decades. Really, you've got a lot of good stuff, but you're repaving the roads to Rome.

It feels like you’re being weirdly hostile here. I’m not trying to rewrite any books, just to be slightly more comprehensive than the various other resources online which describe the ‘alphabet/abjad/syllabary/abugida’ classification and leave it at that.

[xxx]

repaving the roads to Rome...
it is the best way to appreciate the twists and turns and adapt a new pattern...
and it's a typically conlanger way of thinking...
every new language is a layer on the world...
which goes well with mine of reinventing the wheel with each utterance...

as for hangeul, it is linear when analysed syllabically,
and just as alphabetic writing is read logographically,
I don't think Korean writing, which uses spaces between words, is read any differently...