Do Alphabets Always = European?

[Xenops]

As far as I can tell, and from my reading in Geoffrey Sampson's book "Writing Systems: A Linguistic Introduction", it appears that we have our first alphabet (having symbols for each individual consonant and vowel) from the ancient Greeks. The Greeks took the Phoenician script (which is an abjad, since it has no special symbols for vowels?) and adjusted it to have symbols for vowels to better fit the Greek language. Now what I am wondering is: are alphabets mostly a European, Western-world phenomena? I ask because I do not want my conlang to be obviously Western in flavor, but I am biased toward alphabet scripts. The only other script I am familiar with are the Japanese syllabaries, which I am told are mora syllabaries, and are rather odd as far as syllabaries are concerned. I am also biased toward alphabets because (my opinion) they can be closer to the actual pronunciation of the word, rather than a syllabary (using katakana to spell the English word cheese, for example).

Thanks in advance for your insight.

[Travis B.]

You are overlooking abugidas, where, in the basic case (not counting things like ligatures to mark consonant clusters and like) each unit represents a consonant, usually with an inherent vowel, and then diacritics attach to it, which modify or remove the vowel in question. These are commonly found in Ethiopia, South Asia, and Southeast Asia, and Canadian Aboriginal Syllabics are also a case of this.

[vokzhen]

I am also biased toward alphabets because (my opinion) they can be closer to the actual pronunciation of the word, rather than a syllabary (using katakana to spell the English word cheese, for example).

That's more to do with katakana than with syllabaries. There's no reason a syllabary can't accommodate a close pronunciation if it needs to.

[zompist]

I'm not quite sure I understand the concern... if you have a conworld, you could have any culture develop an alphabet. If you're setting something on our planet, it's most likely that your culture would borrow from whoever is nearest. But if you want a non-European alphabet, Korean is a good example (and it's covered in Sampson).

[Terra]

As far as I can tell, and from my reading in Geoffrey Sampson's book "Writing Systems: A Linguistic Introduction", it appears that we have our first alphabet (having symbols for each individual consonant and vowel) from the ancient Greeks. The Greeks took the Phoenician script (which is an abjad, since it has no special symbols for vowels?) and adjusted it to have symbols for vowels to better fit the Greek language. Now what I am wondering is: are alphabets mostly a European, Western-world phenomena? I ask because I do not want my conlang to be obviously Western in flavor, but I am biased toward alphabet scripts. The only other script I am familiar with are the Japanese syllabaries, which I am told are mora syllabaries, and are rather odd as far as syllabaries are concerned. I am also biased toward alphabets because (my opinion) they can be closer to the actual pronunciation of the word, rather than a syllabary (using katakana to spell the English word cheese, for example).

Thanks in advance for your insight.

(1) Japanese has a very simple syllable structure: (C)V(N) . Hiragana and katakana are perfectly sufficient to write this limited syllable structure. Perhaps if Japanese had a more complicated syllable structure, they would've developed a way to unambiguously record these syllables. (Note that both hiragana and katakana were created from kanji.)
(2) What's so odd about them?
(3) Alphabets do not always represent the way that a word is spoken. Take English, for instance. If you read stuff from 500-1000 years ago, you'll see all kinds of variation in the spellings of common words. There wasn't a standard, and people (well, the few that were literate) weren't afraid to spell fonettikly. Nowadays, people have accepted that a word has a certain spelling, no matter what dialect it's spoken in. In this way, the word has become a character of its own, in the same way that hanzi/kanji have: it represents a meaning, and the reader just has to know the right pronunciation.

[Richard W]

By and large, alphabets in the strict sense are a Greek invention, but other systems have converged to them. There were two well-nigh unique (i.e. if not unique, might as well have been) inventions leading to alphabets.

The history of the European alphabet as I see it progresses as follows, with some simplification:

Steps 1 to 3 happened in the Egyptian script:
1) Start with a logographic system for an inflected system with many vowel changes.
2) Adapt some logograms to represent the consonants of the inflections.
3) Extend the use of the single-consonant symbols to clarify the meaning of the logograms, which could represent several different words with related meanings.

The next step seems to have occurred with the adaptation of the system to write a Semitic script:
4) The logograms were dropped, leaving just the consonant symbols. This step is close to unique. The resulting system is called an abjad. It was experimented with in Egyptian, but it may have been a copy of the Semitic system, and something similar happened to yield the Meroitic script.
5) Write vowels on an equal footing with consonants. This happened with the adaptation of the Phoenician style to write Greek. This step is close to being unique.

Now, at step 4, semivowels can come to manifest themselves as vowels, and this can lead to some vowels being written using consonant signs - matres lectionis. This can also lead to a system best described as an alphabet, such as the Mongolian script.

Another development from step 4 is to add marks to the consonants to indicate vowels. If the system already has matres lectionis, they may just be an optional extra, as in Arabic and Hebrew. The vowel symbols may include a symbol to indicate the absence of a vowel. One trick to speed up writing is to use the absence of a vowel symbol to indicate the commonest of absence or a particular vowel. If absence of a mark indicates a particular vowel, then one has an abugida.

These vowel marks are mostly placed above, below or within the consonants - they have little or no effect on their spacing. (Well, vowel marks can affect how Arabic letters are joined together, but that is because Arabic letters may, to varying extents, be stacked vertical. However, in Indic scripts, some of the marks are placed before or after the consonants.

Now, technically, if every vowel is indicated by a mark, one has not an abugida, but an alphabet. I don't find this a useful definition, but by this definition, the modern Lao writing system is not an abugida but an alphabet. It's not a useful distinction, because the simple vowels may go before, after, above or below. There are a few Indic scripts in which all the vowel marks follow the consonants, and these are very similar to alphabets, and for some languages, they may actually be alphabets. The one that comes to mind is Phags-Pa. However, Phags-pa was designed for use in the Mongol empire, and was intended to replace the Mongolian script, so the tidy placement of the vowels may have been inspired by the Mongolian script.

An abugida with vowels written after consonants is not very different from an abjad with matres lectionis; indeed, it is not clear to me how one should classify the Orkhon script. However, that script is a cultural artefact indicative of Levantine influence, albeit distant. With 4 vowel symbols, it is only slightly more impoverished than the original Roman alphabet, with 5 vowels. It makes up for the poor vowel system by doubling the number of word initial consonants. So, although the Orkhon script looks vaguely Runic, it owes little if anything to European alphabets.

So, it is not true that alphabets are restricted to 'European' culture, but they are rare beyond its reach.

[Richard W]

Japanese has a very simple syllable structure: (C)V(N) . Hiragana and katakana are perfectly sufficient to write this limited syllable structure. Perhaps if Japanese had a more complicated syllable structure, they would've developed a way to unambiguously record these syllables. (Note that both hiragana and katakana were created from kanji.)

Or perhaps kana wouldn't have developed!

Another possibility is that they could have come up with something like cuneiform, another 'moraic' writing system, which typically writes /CVC/ as <CV-VC> and may write /CV:C/ as <CV-V-VC>. While that handles /(C)V(C)/ fine, I don't get the impression that it handles Hittite so well.

Also, writing systems often deliberately neglect the fine details - the Philippine scripts, which are abugidas, traditionally truncated CVC to CV (and it has taken a lot of effort to get some acceptance of a no-vowel mark), the Cypriot and Cretan syllabaries couldn't write CVC or CCV as such, the Vai syllabary ignores tone, tone marks had a strong tendency to go walkies in Tai scripts, Latin dropped length marks and word dividers, and old Latin and early Brahmi writing ignored consonant gemination. Coming to modern times with widespread education, the Cree seem loath to write final consonants, and reportedly the aspiration distinction is frequently dropped in informal writing in some American Indian languages.

[Zaarin]

Also, even the Semitic scripts of Phoenician, Aramaic, and perhaps Ugaritic used consonantal symbols to indicate long vowels (as already mentioned) before developing niqqud--<w> for /ū ō/, <y> for /ī ē/, <ʕ ʔ> for /ā ē/. By late Punic, these were even being used for short vowels. At which point you basically have an alphabet.

[Xenops]

First of all, thanks to everyone for your input.

You are overlooking abugidas...

Indeed, I failed to mention abugidas, but I have not forgotten them. In fact I attempted to make one based on the Brahmic script, and found that my language is not strictly CV in format, and I am not familiar with languages that have abugidas, so I was not sure how to fit an abugida to fit a word that is CVCCV, for example.

That's more to do with katakana than with syllabaries. There's no reason a syllabary can't accommodate a close pronunciation if it needs to.

Looking at Greek's Linear B, I guess I could have symbols for consonant codas, etc?

I'm not quite sure I understand the concern... if you have a conworld, you could have any culture develop an alphabet. If you're setting something on our planet, it's most likely that your culture would borrow from whoever is nearest. But if you want a non-European alphabet, Korean is a good example (and it's covered in Sampson).

The simplified version is that the language is spoken /written by non-human entities that have close interaction with humans (more like faeries than aliens, I think). And thank you, I need to look at hangul more closely.

(1) Japanese has a very simple syllable structure: (C)V(N) . Hiragana and katakana are perfectly sufficient to write this limited syllable structure. Perhaps if Japanese had a more complicated syllable structure, they would've developed a way to unambiguously record these syllables. (Note that both hiragana and katakana were created from kanji.)
(2) What's so odd about them?

Indeed, like abugidas, the Japanese syllabaries fit languages with C(V) format the best. Just judging from Wikipedia, there are only like five languages that have the mora structure, so I am hesitant to use it because Japanese would the language most familiar to people. And interesting point about English spelling compared to kanji: we do need to memorize every meaning and pronunciation. xD

To Richard W, thank you for the detailed history: it sounds like the European transformation is very linear, while alphabets in Asia have more spontaneous origins, depending on the language? I will have to look up the languages you mentioned.

Also, even the Semitic scripts of Phoenician, Aramaic, and perhaps Ugaritic used consonantal symbols to indicate long vowels (as already mentioned) before developing niqqud--<w> for /ū ō/, <y> for /ī ē/, <ʕ ʔ> for /ā ē/. By late Punic, these were even being used for short vowels. At which point you basically have an alphabet.

That is an interesting point: something else that I need to look at.

[Travis B.]

You are overlooking abugidas...

Indeed, I failed to mention abugidas, but I have not forgotten them. In fact I attempted to make one based on the Brahmic script, and found that my language is not strictly CV in format, and I am not familiar with languages that have abugidas, so I was not sure how to fit an abugida to fit a word that is CVCCV, for example.

There are multiple way to do this. One is to have special diacritics that mark the lack of a vowel. Another way to do it is to use ligatures and like to mark consonant clusters. As for initial vowels, they can be marked with a special base symbol that marks the lack of a consonant.

[Richard W]

In fact I attempted to make one based on the Brahmic script, and found that my language is not strictly CV in format, and I am not familiar with languages that have abugidas, so I was not sure how to fit an abugida to fit a word that is CVCCV, for example.

There are actually many different ways.

As far as I can make out, in most Indian Prakrits, -CC- would usually be a geminate consonant, and the initial mechanism was to ignore the gemination. Later, the trick adopted was to stack the consonants of the -CC- from top to bottom, and this is still the general rule in SE Asia and Central Asia (definitely Tibetan & Phags-pa). Often these consonants would from a ligature that then took on a life of its own. Sometimes one or other consonant would be simplified. In some case, if the first consonant was simplified, the stack would rise so that the second consonant would be at the level of the normal consonants, and the first consonant above it. When this first consonant is <r>, it is called repha, and discussions of Indian typography will have many mentions of it.

In India, the descending CC stacks were a nuisance in printing, and there has been a tendency to make the general arrangement linear instead. You will find discussion of this under the topic of half-forms. Where the second consonant had a part rising to the hanging baseline on the left, this phonetically second consonant may now occur before the phonetically first consonant.

A method of dealing with final consonants is to add a 'no vowel' symbol to the consonant - generally called virama, though the preferred Hindi term is halant. In non-Indic languages (e.g. Malayalam, Tamil, Burmese), CVCCV is frequently broken into 3 orthographic syllables - CV-C-CV. The middle consonant has the 'no vowel' symbol.

ISCII used a character that represented either stacks the adjacent characters or represents the 'no vowel' symbol; Unicode adopted it and called it virama.

Another method of dealing with final consonants was to write them smaller, and they seem then to have drifted under the previous CV. If the vowel symbol is to the right, one gets a VC stack. This is current in the Lanna script, and has largely gone out of fashion in Khmer and Tibetan. If the vowel symbol is not to the right, CVC and CCV may be indistinguishable, though the phonotactics of the language may greatly reduce ambiguity, as in Northern Thai written in the Lana script. In the Lanna script, CVCCV may be written as CVC-CV with stacking or as CV-CCV with stacking, and some words may be written either way according to the author's taste.

Finally, a few scripts neither stack the consonants nor use a 'no-vowel' symbol. The best known example is Thai. If a consonant cluster has a vowel written on the left, the vowel will be written to the left and the reader has to guess where the vowel goes phonetically. This <okrŋ> is /kroːŋ/, but <osḷs> is /soːlos/. (The last /o/ is the implicit vowel.) There seem to be several words where a misreading has become the standard pronunciation!

Looking at Greek's Linear B, I guess I could have symbols for consonant codas, etc?

You could, but that's more Cree or Elamite cuneiform than Linear B. Linear B doesn't have special symbols for codas. Linear B just dropped codas, just as the Philippine abugidas did.

Also, even the Semitic scripts of Phoenician, Aramaic, and perhaps Ugaritic used consonantal symbols to indicate long vowels (as already mentioned) before developing niqqud--<w> for /ū ō/, <y> for /ī ē/, <ʕ ʔ> for /ā ē/. By late Punic, these were even being used for short vowels. At which point you basically have an alphabet.

That is an interesting point: something else that I need to look at.

This is close to the route that the Central Asian scripts took - Orkhon is an early example, and the Mongolian script is a later one. Both seem to derive from the Sogdian script.

[Richard W]

As for initial vowels, they can be marked with a special base symbol that marks the lack of a consonant.

Gurmukhi seems to be the only script that's done that, though it has three variants, rather like the three alephs of the Ugaritic script.

The SE Asian Indic scripts (Burmese, Thai, Khmer etc.) have reinterpreted the independent vowel symbol <a-> as being the consonant <ʔ>, turning it into a consonant. This is often misinterpreted as a base symbol marking the lack of a consonant. The writing system assumes an initial glottal stop for words beginning with a vowel in some of the lesser languages written in the Thai script, e.g Pali and Pattani Malay. The Tibetan script does the same for Sanskrit. At least one Nagari-script Nepali language has gone the other way, and used independent vowels as combined glottal stop and vowel, and represented /ʔw/ by subjoining <v> to the independent vowel.

[din]

As for initial vowels, they can be marked with a special base symbol that marks the lack of a consonant.

Gurmukhi seems to be the only script that's done that, though it has three variants, rather like the three alephs of the Ugaritic script.

Canadian Aboriginal Syllabics do the same, right? They have the ᐃ ᐊ ᐅ ᐁ /i a o e/ series for words (syllables) starting with a vowel.

[Terra]

Japanese has a very simple syllable structure: (C)V(N) . Hiragana and katakana are perfectly sufficient to write this limited syllable structure. Perhaps if Japanese had a more complicated syllable structure, they would've developed a way to unambiguously record these syllables. (Note that both hiragana and katakana were created from kanji.)

Or perhaps kana wouldn't have developed!

Another possibility is that they could have come up with something like cuneiform, another 'moraic' writing system, which typically writes /CVC/ as <CV-VC> and may write /CV:C/ as <CV-V-VC>. While that handles /(C)V(C)/ fine, I don't get the impression that it handles Hittite so well.

Ancient Mayan, a language whose prototypical syllable's shape was CVC (and strangely disallowed CV ones), was written with a moraic system much like Japanese kana, had a different solution:

(*) A CVC syllable was written CV-CV, where the two vowels (V) were the same: yo-po [yop] 'leaf'
(*) A syllable with a long vowel (CVVC) was written CV-Ci, unless the long vowel was , in which case it was written CiCa: ba-ki [baak] 'captive', yi-tzi-na [yihtziin] 'younger brother'
A(*) syllable with a glottalized vowel (CV’C or CV’VC) was written with a final a if the vowel was [e, o, u], or with a final u if the vowel was [a] or : hu-na [hu’n] 'paper', ba-tz’u [ba’ts’] 'howler monkey'.
(*) Preconsonantal [h] is not indicated.

[M Mira]

Tibetan deals with that by using deliberate syllable boundry mark ་ to delimit a "chunk" of consonants that goes with a vowel, and within a "chunk" the vowel always follows the last consonant on the central stack, and there can be up to 6 consonants per vowel arranged in CCCCVCC structure.

It's basically pushing the definition of abugida to the extreme, where 5 out of 6 letters in a syllable are effectively alphabets, carrying no vowel while remaining in unmarked form and not necessarily in the central stack, though it still fits the rule of abugida in that isolated (delimited by ་) letters automatically forms /Ca/, and other vowels (i, u, e, o) are represented by the diacritics  ིེོུ.

Null onset is treated like a consonant, and its letter is ཨ /0a/

An example of a maximum syllable: བསྒྱུརད་ , composed of བ /ba/, ས /sa/, ག /ga/, ཡ /ja/ (in the subscript form ྱ), vowel mark ུ /u/, ར /ɹa/ or /ra/, ད /da/. It's parsed from left to right, top to bottom, vowel after the last on the stack, regardless whether it's marked above, below or unmarked. It's supposed to represent something like /bsgjurd/ back in the days of Old Tibetan.

To avoid ambiguity, Tibetan places those higher on the sonority hierarchy on the stack if before the vowel, and those lower to the left, so among those that can appear before the "core" consonant, /g d b m/ and འ (phonemic value lost) are placed to the left, while /r l s/ are placed on top of the "core", forming a stack. Glides and liquids that can immediate precede the vowel, that is, /j w l r/, are placed below the "core" and also become a part of the stack, and so ambiguities that would otherwise be a problem in 1-dimensional structure are avoided. 

[Richard W]

What's the evidence that it was /bsgjurd/ rather than /bəsgjurd/ when the writing system was created?

The vowelless orthographic prefixes were prefixes grammatically as well, and they usually take a subordinate position in sorting words alphabetically. The <s> at the start of a cluster often counts as a prefix for sorting as well, though at least how it counts is phonetically determined. The number of final clusters not ending in /s/ was rather limited, and I'm not sure how often the final /s/ of final clusters was not an inflection.

[M Mira]

What's the evidence that it was /bsgjurd/ rather than /bəsgjurd/ when the writing system was created?

The vowelless orthographic prefixes were prefixes grammatically as well, and they usually take a subordinate position in sorting words alphabetically. The <s> at the start of a cluster often counts as a prefix for sorting as well, though at least how it counts is phonetically determined. The number of final clusters not ending in /s/ was rather limited, and I'm not sure how often the final /s/ of final clusters was not an inflection.

To demonstrating maximal syllable, it would always be a past tense of a verb with the form of /bCCCVC(s/d)/ the b- and -s/-d are indeed affixes of past conjugation, and -d was likely an allophone of -s after /n r l/, and was elided by Classical Tibetan, so that leaves only /VCs/ as a valid final cluster.

However, the horizontal affixes are not always grammatical, for the left-siding ones, see the /g-, b-, d-/(ག བ ད) on Tibetan numerals, which, if rendered as prefixes, doesn't seem to convey any meaning. For suffixes, there are གངས (gangs) "snow", ཐབས (thabs) "way, method", and ཐུགས (thugs) "heart, spirit", which have -s in final consonant clusters, and are nouns, which don't decline in Tibetan.

Regarding whether there's a schwa in prefixes to the left, I can't tell if there's a phonetic one, but likely not a phonemic one. First, there are no contrasts, the prefixes to the left can't be placed on top of the stack, so either that 1. The prefix arrangement was only following sonority hierarchy or something similar, or 2. Certain consonants, if prefixed, always carry a schwa. Second, no descendants reflect the schwa, but there is [βzɟɯr] in Japhug rGyalrong as a reflex of the word in question, with the entire initial cluster preserved (G. Jacques (2005))

IIRC, Traditional Tibetan dictionaries sort words by their root letter, disregarding the ones on top, bottom, or left of it, though I'm not sure whether this was phonetically motivated or orthographically motivated.

And for purposes of conscripting, there can be some alternation to this syllable-boundary-marking, 2-dimensional system to avoid ambiguity if the Tibetan rule doesn't work. For example, mandating that the prevocalic cluster first grows up to a stack, then grows left in-line, which would mark position of the vowel unambiguously. Another idea is to have an onset stack and an coda stack for closed syllables, and onset stack only for open syllables.

[Curlyjimsam]

It's certainly true that, in the real world, alphabetic writing systems have not arisen independently very many times and many cultures (even very advanced ones) use other types of system. So this is perhaps something to bear in mind, that it may be unrealistic for every (or even most) cultures on a planet to have an alphabet - but this doesn't mean you should never have any.

I doubt many people will see an alphabetic script and think "this is terribly European!"; it doesn't seemed to be one of the stereotyped Europesque features of conlangs, although perhaps it should be ...

[rotting bones]

My conlang currently has a writing system with glyphs representing consonants and (mostly) qualities of phonation like breathy, creaky and glottalized. Words are distinguished by vowels, but not to the same extent as in human languages. There are dozens of letters corresponding to each (represented) phoneme and spelling is used to distinguish words that would otherwise be spelled the same because vowels are not represented. Two of the tones are directly represented, a third can be inferred from a reliable regularity, and there are rules of thumb for another, but some wouldn't be distinguishable at all if not for the nightmarish spelling system.

That's still technically an alphabet, right? (Also, any idea where I should look to see what kind of states an alien glottis might have that might be interestingly different from human ones? And if this is too implausible even for an alien language, that would be good to know.)

[CatDoom]

(Also, any idea where I should look to see what kind of states an alien glottis might have that might be interestingly different from human ones? And if this is too implausible even for an alien language, that would be good to know.)

Glottal states fall along a continuum from fully open to fully closed, so unless the aliens in question have a glottis that does something other than contract and relax, there's not a whole lot of room for "new" glottal states. It might be possible for aliens with more acute hearing or muscle control or something to distinguish more glottal states than any human language does, having sounds that fall at many places along the spectrum.

[Richard W]

I don't think Rotting Bones's writing system is an alphabetic script. To be honest, it seems more to be a logographic system, such as the English writing system is falsely accused of being. Has a history been created for it?

If one ignores the multiplicity of symbols for each consonant, which I don't think one should, one can attempt to classify it amongst the segmental systems. If the tone marks are treated like consonants, one might try arguing that it was alphabetic, though not as we know it, with the roles of tones and vowels the other way round to human writing systems. There are quite a few writing systems which ignore tones; I can't think of any that write tones but not vowels. If the tone marks have a subsidiary role, it's really an abugida in the useful sense of the term.

[Salmoneus]

I think if it doesn't represent vowels, it can't be an alphabet.

Ignoring that, I think there are three possibly interestingly distinct possible systems:
- a single alphabet, but many-one spelling rules, so that words that sound the same (or differ only in features, like tone, that are not shown in the script) look different because they can be spelled in different ways
- a single very large alphabet with many possible glyphs for each phoneme, allowing the same (or similar) sounds to be spelled in many ways
- multiple small alphabets with different words written in different alphabets.

The first is what you might call an anti-phonemic alphabet. English has elements of this, and I've played around with spelling that goes in that direction: intentionally disambiguating homonyms through the use of variant spellings. So just as we have "be" vs "bee" and "right" vs "rite" vs "wright" vs "write", we might also have "can" (modal verb) vs "kan" (fluid container), or "duck" (action) vs "ducque" (bird), "router" (thing that provides a route) vs "roubter" (one who inflicts a rout), etc. At some point this probably does become a logography, but there's also a large area before you get that far where it's just a baroque alphabet.

The third is something I've toyed with for some of my conlangs, but never actually made. I used to call it an "alphalogography". It operates on a similar principle to the ideal chinese system, with one element suggesting meaning and one element suggesting pronunciation - but instead of the phonetic element being a single symbol, it's broken into a series of symbols - an alphabet, abjad, or syllabary - and then instead of the semantic element being another symbol, it's instead the choice of the set of phonetic symbols. To put that into concrete terms: the word for 'duck' would use one set of letters, the bird-letters, when it meant the bird, and another set of letters, the action letters, when it meant the action. Likewise, 'hawk' would use bird-letters when it meant the bird, but action letters, or maybe trade letters, when it meant 'to sell'.

I imagine this system would originate in the simplification process by which initial ideograms or logograms gain phonetic function. But instead of one single logogram being chosen to now represent all instances of /b/, there would be a dozen or so different symbols for /b/, and different ones would be used depending on the meaning - the 'b' in 'beetle' might be a picture of a bee (invertebrate set), whereas the 'b' in 'cabbage' might be a picture of a bean (vegetable set). [obviously, when I say 'picture', I mean 'symbol that originally represented...'] Presumably this would occur when there was some serious ambiguity due to the script's failure to represent some vital element (like vowels, or voicing, etc) - although the whimsy and over-ornamentation (perhaps ritualisation and religious beliefs - like how people used to use CAPITALS for words relating to the LORD) of priests and scribes could also help explain it.

The second possibility listed above is sort of the intermediate condition between the first and the third possibilities.

[Richard W]

I think if it doesn't represent vowels, it can't be an alphabet.

It depends what allowances one should make for it being an alien writing system. I think substituting tones for vowels is valid. After, all writing tones but not vowels is not, I believe, something that has happened in human writing systems. Presumably no language with an abjad has undergone tonogenesis in the East Asian manner, where some consonants turned into tones.

The first is what you might call an anti-phonemic alphabet. English has elements of this, and I've played around with spelling that goes in that direction: intentionally disambiguating homonyms through the use of variant spellings. So just as we have "be" vs "bee" and "right" vs "rite" vs "wright" vs "write",

Except that that isn't how these spellings arose. Deliberate disambiguation such as this is rare.

However, I can see an inspiration from Arabic script writing for non-Semitic languages (e.g Persian) and from Thai. Persian vowels often have to be deduced by recognising the consonant sequence as a particular word. In Persian, you can have as many as 4 different letters for a sound. In Thai, ignoring silent letters, a final /t/ could be written in as many as 6 + 6 + 6 = 18 different ways, though not all of them actually occur. The 6-way split arises because Thai readily makes 5-way distinctions in manner and tone at each place of articulation, and the 6th is just a preservation of its Indic heritage. Syllable-finally, these distinctions but do not apply, but mostly Indic loanwords result in these distinctions being preserved in writing syllable-finally. Finally, the three Indic orders of palatal, retroflex and dental are not distinguished word finally in Thai. In Thai, dents and stretches have been added to letters to indicative pre-glottalisation (now realised as voicing) and fricativisation. Finally, Thai has two old tone marks, which go along way towards covering the existing five tones (or six in the north). If you combine these features and exaggerate the number of consonants in an abjad with say 3 unrecorded vowels (as in Persian), you get something like Rotting Bones's writing system.

However, for a human language, one would be double-counting phonation contrasts - once in the letters, and again in the phonation glyphs. It would be hard to do by dissimilar languages borrowing the writing system, for even with massive borrowing, there would be a tendency to systematically use a few letters for each sound in native words. That's why I favoured the analysis as alphalogographic, though I hadn't come up with the idea of anything quite so elegant with the semantics. I suppose such a system could arise in part from the ligatures of purely phonetic symbols and determinatives. Egyptian Demotic might have given rise to such a system, though tone and phonation markers would have had to be added somehow.

[rotting bones]

Glottal states fall along a continuum from fully open to fully closed, so unless the aliens in question have a glottis that does something other than contract and relax, there's not a whole lot of room for "new" glottal states. It might be possible for aliens with more acute hearing or muscle control or something to distinguish more glottal states than any human language does, having sounds that fall at many places along the spectrum.

I'm looking for information on the range of features of a sound wave that can be affected by glottal states in humans, and how these features are affected by the shape, etc. of the glottis. Eg. What features of the sound wave correspond to the creak in creaky voice and how does the glottis produce it?

Has a history been created for it?

My currently preferred version of the history involves elements being lifted off logograms, simplified and regularized to an extent to form the letters. The elements lifted off the logogram representing word X are often, though not always owing to regularizations, etc., used to spell the word X, hence the multiplicity of letters corresponding to each phoneme. More regularizations tend to occur with each regime change.

(Also, elements from "similar" logograms tend to be used in regularizations and to spell words that never had logograms. "Similarity" can be based on meaning as well as sound.)

(Yes, I know, it's extremely primitive. My explanation is just that the aliens find phonation contrasts more significant than vowel contrasts and they haven't cleaned up their writing system. Although there are considerations like maintaining a similar writing system across a cultural area.

I have an alternate history of borrowing from another sapient species, but I haven't developed it as much, whereas I actually have the logograms the script was supposed to have come from.)

If one ignores the multiplicity of symbols for each consonant,

Not just consonants. Phonations and two of the tones too. (And they are not subsidiary.)

If the tone marks are treated like consonants, one might try arguing that it was alphabetic, though not as we know it, with the roles of tones and vowels the other way round to human writing systems. There are quite a few writing systems which ignore tones; I can't think of any that write tones but not vowels. If the tone marks have a subsidiary role, it's really an abugida in the useful sense of the term.

For two of the tones and a bucket category for all the other tones, each combination of one phonation and one tone is represented by a distinct set of dozens of symbols. (Though not all combinations exist.)

(I can't promise that my subconscious does not intend this as satire. BTW, I should have mentioned before that zompist has already done something similar with Eteodaole distinguishing words by tone rather than vowel IIRC. I came up with this before I knew about Eteodaole, but zompist has probably had the idea long before he published it. I never intend to publish my conworlding, though. This is solely for my own amusement.)

I think if it doesn't represent vowels, it can't be an alphabet.

So if I just let the writing system distinguish between at least two sets of vowels similar to the way I do with tones, then will it become an alphabet? (Wouldn't work with the language as it stands. Because it distinguishes words by only two vowels, if they are distinguished in writing, then all the vowels will be distinguished. Although perhaps they can be imperfectly distinguished? But that probably won't be an alphabet.)

Ignoring that, I think there are three possibly interestingly distinct possible systems:
- a single alphabet, but many-one spelling rules, so that words that sound the same (or differ only in features, like tone, that are not shown in the script) look different because they can be spelled in different ways
- a single very large alphabet with many possible glyphs for each phoneme, allowing the same (or similar) sounds to be spelled in many ways
- multiple small alphabets with different words written in different alphabets.

I noticed that English tends to have the same letter be pronounced in different ways more often than in Bengali, whereas in Bengali, we have three different letters for an s-like sound that are distributed throughout the words on the basis of Sanskrit spelling, even though most of them are simply pronounced [ʃ] in educated speech. (plenty of obvious exceptions exist, like [ɔst̪ʰir])

If you combine these features and exaggerate the number of consonants in an abjad with say 3 unrecorded vowels (as in Persian), you get something like Rotting Bones's writing system.

I haven't studied Thai, but you're right about Persian being one influence.

However, for a human language, one would be double-counting phonation contrasts - once in the letters, and again in the phonation glyphs.

No, it's like the vowels carry tone and phonation like in Jalapa Mazatec: https://en.wikipedia.org/wiki/Jalapa_Mazatec

But the writing system only represents the phonation contrasts and two of the tones, because vowel distinctions aren't as important as they are in human languages in any of the languages used by this species.

[Valdeut]

After, all writing tones but not vowels is not, I believe, something that has happened in human writing systems. Presumably no language with an abjad has undergone tonogenesis in the East Asian manner, where some consonants turned into tones..

Punjabi, perhaps? It lost the Indo-Aryan breathy voiced stops (they became tenuis or voiced) but they did leave behind a tonal distinction.
https://en.wikipedia.org/wiki/Punjabi_language#Tone

Punjabi can be written in an Perso-Arabic abjad (called the Shahmukhī script), although it is also written in an abugida (usually the Gurmukhī script). I might be mistaken because I was able to find much more written about the Gurmukhī script than the Shahmukhī, but I believe it is the case in both scripts, that Punjabi is written as if it still had breathy voiced stops, which means that the choice of consonant letter can indicate tone. So that would mean that the abjad leaves out short vowels while indicating tone, at least to some extent.