A guide to writing systems

[xxx]

the advantage of constructed language is that you can blur the boundaries...

in 3SDeductiveLanguage(1Sense=1Sign=1Sound), writing is first, based on pictograms very simplified,

they are composable in a purely logographic framework,

by making sounds correspond, and we obtain a syllabic system,

by means of certain modes we can write it by distinguishing the semantic signs attached to consonants from those attached to vowels to make an alphabet of it...

which makes it at once a pictography, a logography, a syllabic and alphabetic system...

to this we must add that, in the absence of a constituted lexicon other than the hundred or so primitives,
self-defined compositions, made on the fly, without segregation into words,
can only with difficulty be read as blind characters, in gestalt form,
and must be semantically deciphered as we go along...

[Richard W]

Of course, the Hangul isn't being written as it was designed to be. It's overall direction used not to be left to right, but top to bottom. It's not certain that Lao ແ should be seen as two components, rather than one. Incidentally, except when used as the sole vowel component, ເ can be taken as a register shift, changing the interpretation of the rest of the vowel part of the syllable.

Finally, the Lao layout in your example seems designed for easy mechanical or computerised layout. In particular, smart font technology has not been used to position the tone marks closer to the consonants when there is no vowel above. The suboptimal layout led to your missing one tone mark on the third line. Marks seem to normally be applied to the syllable as a whole, rather than the last consonant of the onset, with the result that they normally appear further to the right, especially in Lao. This helps bind words/syllables together - the Thai and Lao languages use spaces to separate phrases and clauses, or at least mature writers do. (Young Thai children also use white space to separate syllables.) The four <i>-type vowels may be an exception - in Thai they get written immediately after the consonant, and one can often see ligatures where the pen has not been completely lifted off the paper.

For fluent writers, the sequence of actually writing Thai characters other than marks above and below is the left to right order. I haven't watched Lao being written by hand.

[Moose-tache]

I understand that bradrn isn't making some weird claim about eye-tracking or the psychology of jamo-stacking, but I'm not sure what the point is.

It's true that Korean, at a purely theoretical level, makes use of vertical space by stacking meaningful elements, but what does that entail? From a naive perspective we could say that any writing system other than Morse code makes use of vertical space. How "meaningful" the pieces are that we're stacking is more of a spectrum than a neat binary. For example, the decenders on y and especially j could be pedanticly treated as consonantizers.

Either way, readers engage with text on a much less granular level than any writing system's internal orgnanization scheme, so how can we be sure the internal organization scheme isn't just a soft play area for bored linguists? As Zompist hinted with his comparison to Chompsky (ptew), linguistic analysis that doesn't measure itself against usage quickly becomes circular.

tl;dr: I guess what I'm getting at is that if we're trying to come up with some way of classifying writing systems, philology needs to referee. We need to come up with a classification that has implications for how people read and write.

[bradrn]

Of course, the Hangul isn't being written as it was designed to be. It's overall direction used not to be left to right, but top to bottom.

This is actually really interesting — I hadn’t realised how much more ‘linear’ Hangeul becomes when written vertically. All the consonants are on the left, and all the vowels are written to the right. Admittedly the consonants still get squished and expanded to fill up as much space as possible, but it becomes much more comparable with systems such as Lao.

Finally, the Lao layout in your example seems designed for easy mechanical or computerised layout. In particular, smart font technology has not been used to position the tone marks closer to the consonants when there is no vowel above.

Indeed, this is a pretty terrible font! The underlying image of Lao text is simply taken from Omniglot — I don’t know which font they used.

(I did notice the missed tone mark, but only after I posted the image.)

Marks seem to normally be applied to the syllable as a whole, rather than the last consonant of the onset, with the result that they normally appear further to the right, especially in Lao. This helps bind words/syllables together - the Thai and Lao languages use spaces to separate phrases and clauses, or at least mature writers do. (Young Thai children also use white space to separate syllables.) The four <i>-type vowels may be an exception - in Thai they get written immediately after the consonant, and one can often see ligatures where the pen has not been completely lifted off the paper.

And this is really interesting too! Do you happen to have any images you could post to demonstrate these points?

For fluent writers, the sequence of actually writing Thai characters other than marks above and below is the left to right order. I haven't watched Lao being written by hand.

Yes, I suspected this might be the case. But on the other hand, in English I also write the dollar sign before the number. It doesn’t affect the mapping from graphemes to phonemes, which is what I was talking about.

tl;dr: I guess what I'm getting at is that if we're trying to come up with some way of classifying writing systems, philology needs to referee. We need to come up with a classification that has implications for how people read and write.

Of course, which is what I’m trying to do. It should make more sense once I finish properly thinking it through, and post it as a coherent hole. (If I ever come to any firm conclusions, of course.)

(But I don’t see what ‘philology’ has to do with it — I typically associate that term with the diachronics of classical languages.)

[masako]

tl;dr: I guess what I'm getting at is that if we're trying to come up with some way of classifying writing systems, philology needs to referee. We need to come up with a classification that has implications for how people read and write.

Do we though?

I understand the impulse, but what might be more effective is to describe how people CAN (potentially) read. Much like the apparent purpose of this thread, once we define something as open as "writing", we tend to run into divergent and wandering ideas. "How people read" is likely gonna be even more open because unlike the physical medium of writing, "how" people do something that occurs within the brain leads to far more open and wandering questions.

I was working on a paper about "What is writing?" with the main thesis being: The application of force to a surface to generate markings indicative of sound and/or meaning intended to travel space and time.

I've made further notes, but I've also run into having to define multiple other aspects of language, orthography, philology, etc. It's a worthy endeavorsure, but I don't know how much we 'need' it.

[Moose-tache]

(But I don’t see what ‘philology’ has to do with it — I typically associate that term with the diachronics of classical languages.)

Minor but interesting point. Could be my age showing, but in older literature "philology" was basically short for "knowing languages in their own terms," i.e. speaking languages as opposed to studying linguistics. Back in the day a good education started with a conversational understanding of Greek and Latin, so it's easy to see how the shift in usage happened.

[sasasha]

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 can see this has been discussed already. I just want to try to agree with Moose-tache that writing is a psychological behaviour, and if there is no psychological phenomenon of non-linearity, then there is no non-linearity. (Linearity as a term may be at fault for misrepresenting itself here, given that a robust sequence need not be a line.)

To follow any sequence we need rules. These are never going to be completely consistent. We have plenty of ‘loopback’ elements in our own writing system, and plenty in our speech, too. (These parentheses signify an aside the way Lao ⟨e⟩ signifies a syllable with /e/ coda, and it doesn't mean that we are speaking backwards just because we use a pronoun to refer to something we named earlier.)

Classifying writing systems based on how linear they are thus isn't sitting well for me. As zomp suggested, we are in danger of bringing bias to that. I think the psychological experience of reading and writing needs to be explored to understand sequence in writing systems. (For instance, there's probably a beautiful straight line connecting the typical eye-gaze points of both your Hangeul and Lao examples. We might even find metrics by which we could consider Hangeul ‘more linear’: the eye-gaze points might be more regularly spaced, for instance.)

Anyway, this has been an interesting discussion to read and I'm looking forward the next one.

[bradrn]

I can see this has been discussed already. I just want to try to agree with Moose-tache that writing is a psychological behaviour, and if there is no psychological phenomenon of non-linearity, then there is no non-linearity. (Linearity as a term may be at fault for misrepresenting itself here, given that a robust sequence need not be a line.)

Is there no value at all in relating the sequence of graphemes to the sequence of phonemes, irrespective of psychology? Quite the contrary: I’d say that’s the key thing to do when discussing a writing system.

In any case, all this debate about psychology feels a little disingenuous to me. When we’re talking about syntax, no-one says ‘we shouldn’t classify languages by the order of the adjective and noun, because we haven’t proved that there is a psychological phenomenon of adjective-ordered-ness’. We just describe the language given what we observe, and let the psychology come afterwards when possible.

To follow any sequence we need rules. These are never going to be completely consistent.

It’s not a matter of consistency at all. It’s purely a matter of describing what those rules are and how they work. In Latin, the rule is, ‘speak each letter from left to right’. In Lao, it’s ‘speak each consonant+vowel from left to right, unless you see one of ⟨ເ ແ ໂ⟩’. And, for that matter, in Pahawh Hmong, it’s ‘take each VC block and pronounce the phonemes in reverse order’.

Anyway, this has been an interesting discussion to read and I'm looking forward the next one.

Thanks! (And apologies I haven’t written the next one yet; there’s a few other things I’m working on at the moment.)

[xxx]

It’s purely a matter of describing what those rules are and how they work. In Latin, the rule is, ‘speak each letter from left to right’. In Lao, it’s ‘speak each consonant+vowel from left to right, unless you see one of ⟨ເ ແ ໂ⟩’. And, for that matter, in Pahawh Hmong, it’s ‘take each VC block and pronounce the phonemes in reverse order’.

I think that if we stick to instructions alone, purely phonemic scripts (at least those that separate words, as the others seem to retain oral pre-eminence...) don't apply the instruction but decipher words globally, as patterns, as with logographic scripts...

this is because, unlike logographic scripts, the less phonics it includes, the better it stays in phase with its instructions...
phonemic scripts quickly no longer encode the reality of speech and, beacause they encode no meaning, become blind, losing all their supposed advantage over logographic scripts, which avoided the need to learn a large number of written words...

the invention of syllabaries and the alphabet was a last resort, making it possible to adapt the writing perfectly suited to one language to another that was too different to use it...
contrary to the occientalo-centric vision of writing, logographic scripts have the advantage, and it's not an evolution but a regression that leads from them to alphabets...

thus, languages with a logographic script don't change to a phonic script without external constraints, which enables them to endure culturally;
but logographic usage is a necessity for languages with a phonic script, to limit their rapid obsolescence...

[sasasha]

When we’re talking about syntax, no-one says ‘we shouldn’t classify languages by the order of the adjective and noun, because we haven’t proved that there is a psychological phenomenon of adjective-ordered-ness’. We just describe the language given what we observe, and let the psychology come afterwards when possible.

Fair point, which I concede!

[zompist]

In any case, all this debate about psychology feels a little disingenuous to me. When we’re talking about syntax, no-one says ‘we shouldn’t classify languages by the order of the adjective and noun, because we haven’t proved that there is a psychological phenomenon of adjective-ordered-ness’. We just describe the language given what we observe, and let the psychology come afterwards when possible.

We do? Depends on who you ask, and if it's Chomsky, when you ask them.

The difference between Chomskyan and cognitive linguistics is precisely in when, if ever, the psychology is allowed to come in. For the Chomskyan, some time in the far future when "performance" is tackled. (And yet the entire system is supposed to live in the genome.) For the cognitivists, right now: plausible theories about how the brain works are more important than structural notions.

Order is obviously a key topic in syntax. OTOH, there are also interesting examples of phonological simultaneity, and a few languages that are unusually uninterested in order (look up Jiwarli in my book).

it’s the key thing to do when discussing a writing system

I think departures from linearity are interesting, but not "key." I'm not convinced that it's more important than the scope of graphemes (features / phonemes / syllables). And as I said, I think you're imposing an unwarranted psychological interpretation in some of these cases. E.g. ü is not necessarily an example of "verticality"; it could also be an example of "simultaneity." After all, if the overall flow is left to right, then vertical arrangement could simply indicate identical placement in time— that is, no ordering at all.

I'm not sure why people came up with कि ki vs. की kī. All I can glean from WWS is that this doesn't go back to Brahmi. I wonder if one reason is that premodern people seem very reluctant to invent entirely new symbols, so a weird use of an existing one is preferred to creating a new one.

[Moose-tache]

I appreciate that bradrn is trying to observe and classify objective phenomena, and leave the psychology of speakers as a separate problem. My concern is that it is slightly illusory to call our observations and classifications "objective" when we can't quantify things like "verticalness" or formulate our claims to make them falsifiable. Once we concede that speaker psychology is not part of our task, we run the risk of building sandcastles that exist only in the minds of one person. What insight does this provide, if it cannot lead to a productive conversation about how language works? I don't think bradrn is guilty of Chomspkyian "under the surface, every clause is secretly a bashful weevil named Steve" linguistics, but I do think this approach leads ultimately to the same place.

[bradrn]

The difference between Chomskyan and cognitive linguistics is precisely in when, if ever, the psychology is allowed to come in. For the Chomskyan, some time in the far future when "performance" is tackled. (And yet the entire system is supposed to live in the genome.) For the cognitivists, right now: plausible theories about how the brain works are more important than structural notions.

But there’s a third tradition in modern linguistics: what Dixon and Dryer call Basic Linguistic Theory. This is the purely descriptive perspective from which most reference grammars are written, and is where I tend to come from.

Order is obviously a key topic in syntax. OTOH, there are also interesting examples of phonological simultaneity, and a few languages that are unusually uninterested in order (look up Jiwarli in my book).

Most Australian languages fit into this category, actually. (I found a very interesting paper on the subject just last month, but alas, I seem to have lost the reference.)

it’s the key thing to do when discussing a writing system

I think departures from linearity are interesting, but not "key." I'm not convinced that it's more important than the scope of graphemes (features / phonemes / syllables).

Fair enough; I claimed too much there.

And as I said, I think you're imposing an unwarranted psychological interpretation in some of these cases. E.g. ü is not necessarily an example of "verticality"; it could also be an example of "simultaneity." After all, if the overall flow is left to right, then vertical arrangement could simply indicate identical placement in time— that is, no ordering at all.

I don’t recall ever having mentioned terms such as ‘verticality’ or ‘simultaneity’. Neither did I make any claim about ⟨ü⟩ — largely because it represents one phoneme, so phoneme sequencing doesn’t play any role at all.

(For the record, I consider most Latin-script diacritics an attempt to superimpose a featural system on an existing script, so the letter+diacritic combination generally acts as a single unit.)

I'm not sure why people came up with कि ki vs. की kī. All I can glean from WWS is that this doesn't go back to Brahmi. I wonder if one reason is that premodern people seem very reluctant to invent entirely new symbols, so a weird use of an existing one is preferred to creating a new one.

Judging by the tables on Wikipedia, this appears to have resulted from gradual evolution. In Early Brahmi, both -i and -ī appear at the top-right of the syllable block; by Late Brahmi (a.k.a. Gupta), -ī had become visually distinguished by moving more to the top-left. By the time of Siddhaṃ, both have developed long tails which drop down beside the letter, leading to the current situation in Devanāgarī.

(Also, what’s WWS?)

I appreciate that bradrn is trying to observe and classify objective phenomena, and leave the psychology of speakers as a separate problem. My concern is that it is slightly illusory to call our observations and classifications "objective" when we can't quantify things like "verticalness" or formulate our claims to make them falsifiable. Once we concede that speaker psychology is not part of our task, we run the risk of building sandcastles that exist only in the minds of one person.

Of course, but then again I never claimed to be ‘objective’. (And your last sentence could apply to basically all of descriptive linguistics, which seems to happily run along without considering psychology at all.)

[zompist]

Judging by the tables on Wikipedia, this appears to have resulted from gradual evolution. In Early Brahmi, both -i and -ī appear at the top-right of the syllable block; by Late Brahmi (a.k.a. Gupta), -ī had become visually distinguished by moving more to the top-left. By the time of Siddhaṃ, both have developed long tails which drop down beside the letter, leading to the current situation in Devanāgarī.

Ah, neat!

(Also, what’s WWS?)

The World's Writing Systems, edited by Daniels and Bright. A fantastic reference.

[bradrn]

(Also, what’s WWS?)

The World's Writing Systems, edited by Daniels and Bright. A fantastic reference.

Ah, thanks! I’ll have to see if I can get a copy.

EDIT: Just looked at the contents, and… wow, this is astonishingly comprehensive. But their aim seems different (and neatly complementary) to mine: they focus on detailed descriptions of individual scripts, whereas I aim to write more about typology and evolution.

[Richard W]

I'm not sure why people came up with कि ki vs. की kī. All I can glean from WWS is that this doesn't go back to Brahmi. I wonder if one reason is that premodern people seem very reluctant to invent entirely new symbols, so a weird use of an existing one is preferred to creating a new one.

Judging by the tables on Wikipedia, this appears to have resulted from gradual evolution. In Early Brahmi, both -i and -ī appear at the top-right of the syllable block; by Late Brahmi (a.k.a. Gupta), -ī had become visually distinguished by moving more to the top-left. By the time of Siddhaṃ, both have developed long tails which drop down beside the letter, leading to the current situation in Devanāgarī.

And to compensate -e (and so -o, which has long been notated as a combination of -e and -ā) ceased to be (partly) before the consonant. It's still before the consonant in the closely related Eastern Nagari script (i.e. Bengali etc.). And indeed, it can be before the consonant in some allegedly Devanagari styles - Unicode was persuaded to add a separate character for the preceding -e, namely DEVANAGARI VOWEL SIGN PRISHTHAMATRA E, which people often forget. The other two vowels affected by that style, -ai and -o, have to be written as a combination of the new (but archaic) -e and another vowel.

[bradrn]

Alphabets: basics

In the last post, I defined alphabets as writing systems where one grapheme corresponds to one phoneme. This gives them a particularly simple structure: ideally, it should be possible to read out a sentence simply by reading each grapheme in turn. An example should illustrate the principle well (using the first sentence of the UHDR in Georgian):

ყველა ადამიანი იბადება თავისუფალი და თანასწორი თავისი ღირსებითა და უფლებებით.

ყ

qʼ

ვ

v

ე

e

ლ

l

ა

a

 

ა

a

დ

d

ა

a

მ

m

ი

i

ა

a

ნ

n

ი

i

 

ი

i

ბ

b

ა

a

დ

d

ე

e

ბ

b

ა

a

 

თ

t̪ʰ

ა

a

ვ

v

ი

i

ს

s

უ

u

ფ

pʰ

ა

a

ლ

l

ი

i

 

დ

d

ა

a

 

თ

t̪ʰ

ა

a

ნ

n

ა

a

ს

s

წ

tsʼ

ო

o

რ

r

ი

i

 

თ

t̪ʰ

ა

a

ვ

v

ი

i

ს

s

ი

i

 

ღ

ɣ

ი

i

რ

r

ს

s

ე

e

ბ

b

ი

i

თ

t̪ʰ

ა

a

 

დ

d

ა

a

 

უ

u

ფ

pʰ

ლ

l

ე

e

ბ

b

ე

e

ბ

b

ი

i

თ

t̪ʰ

Romanisation: q'vela adamiani ibadeba tavisupali da tanasts'ori tavisi ghirsebita da uplebebit.

Other examples of modern-day alphabets include Latin, Greek, Cyrillic, Armenian, Mongolian, N'ko, Adlam and Osage. Ancient examples include Coptic, Avestan, Futhark and Ogham.

Of course, not everything that is called an ‘alphabet’ satisfies the definition as strictly as Georgian. Instead, most display cases in which graphemes do not correspond to single phonemes. This can occur in several ways.

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⟩.

(Note that in all these cases, only a few graphemes represent phoneme sequences. If this occurs systematically for a large number of graphemes, then the writing system is better analysed as an abugida, syllabary or semi-syllabary, rather than an alphabet.)

A related phenomenon is observed in Cyrillic, in which many graphemes add palatalisation to the preceding consonant. This is the case for vowel graphemes ⟨Е Ё Ю Я⟩,which palatalise the previous consonant if there is one, and represent the phoneme sequences /je ju ja/ otherwise. Historically, ⟨Е Ю Я⟩ derive from ligatures ⟨Ѥ Ю Ꙗ⟩, though only ⟨Ю⟩ has retained the ligated shape. There is also the soft sign ⟨Ь⟩, which palatalises the previous consonant while representing no phoneme of its own, and similarly the hard sign ⟨ъ⟩ which depalatises the previous phoneme.

(Again, Cyrillic can still be called an alphabet because this only affects a small number of graphemes. A writing system where many graphemes are used only to mark subphonemic features might be better analysed as a featural system.)

A similar case occurs in Irish, where many vowel graphemes serve only to mark the secondary articulation of an adjacent consonant. All vowel graphemes are considered either caol ‘slender’ ⟨e é i í⟩, indicating palatalisation, or leathan ‘broad’ ⟨a á o ó u ú⟩, indicating velarisation. Thus, for instance, ⟨dáil⟩ ‘assembly’ is /d̪ˠaːlʲ/: the ⟨i⟩ only marks the palatalisation of the following consonant. Presumably this system originated by transcribing the offglides of consonants.

Many alphabets also contain multigraphs: grapheme sequences which are treated and read as a single unit. (Two graphemes form a digraph, three a trigraph, and so on.) Examples include ⟨ch sh th ng⟩ /tʃ ʃ θ~ð ŋ(ɡ)/ in English, ⟨ch cz sz rz⟩ /x t͡ʂ ʂ ʐ/ in Polish, ⟨ch dd ff ll ng ph rh si th⟩ /χ ð f ɬ ŋ(ɡ) f r̥ ʃ θ/ in Welsh, and ⟨αι ει οι ου υι γγ τσ τζ γκ μπ ντ⟩ /e̞ i i u i (ŋ)ɡ ts dz ɡ b d/ in Greek. Particularly close-knit digraphs sometimes tend to ligate into a single grapheme, as seen e.g. with ⟨IJ⟩ in Dutch.

Interestingly, it seems that writing systems often have one or two preferred strategies they use to form multigraphs. This can be seen in the examples: English adds ⟨h⟩, Polish adds ⟨z⟩, Welsh adds ⟨h⟩ or doubles the letter, and Greek adds ⟨ɩ⟩ or combines a voiced continuant with a voiceless stop. Why this happens, I’m not entirely sure, but we can speculate on some historical drivers:

• Shared inheritance: for instance Ancient Greek ⟨φ θ χ⟩ /pʰ tʰ kʰ/ were represented in Latin as ⟨ph th ch⟩, before the Greek letters shifted to /f θ χ/, forcing the Latin digraphs to be treated as single units. Similarly Greek ⟨αι ει οι ου υι⟩ became single units via sound change.
• Analogy: given an existing digraph, forming new digraphs using the same strategy is intuitively reasonable.
• Disambiguation: more of a functional motivation, but it seems plausible that multigraphs might become easier to distinguish from ordinary letters when given some regular marker.

Finally, most (if not all) writing systems display some degree of irregularity. In an alphabet, this will of course obscure the grapheme–phoneme correspondence. In extreme cases, such as English, I’ve seen people seriously argue that this approaches a logography — the spelling of homophones such as ⟨by⟩/⟨bye⟩/⟨buy⟩ is related to semantics, rather than phonology. But I’d suggest such a system is still primarily alphabetic, since the spelling still corresponds systematically to phonology. (Of course, it might be a different story if we were to spell /ba͡i/ as ⟨xaz⟩.)

Next up: beyond the definition, what properties are common in alphabetic scripts? And what happens when those are violated?

[zompist]

A few additions:

1. Another source of digraphs is sound change:
* that's really what's going on with the vowels in Greek.
* Romance languages fricativized c/g before front vowels. That in turn allowed French to use <ge> for ʒ, as in plongea 'dived'; cf. also Italian <gi> as in giorno /dʒorno/.
* In French, <qu> went from /kw/ to /k/, and got a second life as a digraph for /k/ before front vowels.

Related: "silent" e in French and English was of course once pronounced.

2. Allen says the use of <h> for aspiration in Latin goes back to the 2nd century BCE (before that the Romans simply ignored aspiration). Of course, sound change turned this from a representation of aspiration to one of fricativization.

3. I may be wrong, but I think the use of h to modify vowels in English is an imitation of transliterations of Hebrew, as in words like Sarah, Micah, then applied to interjectons (oh! ah!), and to foreign /a/ as in Burmah.

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

[bradrn]

1. Another source of digraphs is sound change:
* that's really what's going on with the vowels in Greek.
* Romance languages fricativized c/g before front vowels. That in turn allowed French to use <ge> for ʒ, as in plongea 'dived'; cf. also Italian <gi> as in giorno /dʒorno/.
* In French, <qu> went from /kw/ to /k/, and got a second life as a digraph for /k/ before front vowels.

Related: "silent" e in French and English was of course once pronounced.

That’s what I was getting at by ‘shared inheritance’. I even mentioned sound change for Greek! The extra examples are great, though.

2. Allen says the use of <h> for aspiration in Latin goes back to the 2nd century BCE (before that the Romans simply ignored aspiration). Of course, sound change turned this from a representation of aspiration to one of fricativization.

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

3. I may be wrong, but I think the use of h to modify vowels in English is an imitation of transliterations of Hebrew, as in words like Sarah, Micah, then applied to interjectons (oh! ah!), and to foreign /a/ as in Burmah.

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

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

Ooh, that’s a great one! Let me edit it in.

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

[zompist]

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.)

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.

(In Spanish, there was another source, <nn> or <nm>, as in annus > año. The tilde is just a tiny <n>.)