Language and writing system

Aleme's picture

Hello every one , Today I had intense talk about the connection between writing systems and languages . My view is in most cases it has no connection. My question is how important is this connection to type designers ? Is this connection between language and writing systems documented some where ? I know for sure nations who uses ideographic writing system the symbols has no connection to the spoken language. Writing systems like Urdu or Persian are forced in to the Spoken language (which Is Arabic writing system) in short my final question is how much is linguistics important to type designers ?
Thanks a lot and Happy Thanks Giving,m

quadibloc's picture

It is not clear that type designers - unless they're effectively moonlighting by engaging in alphabetic reform - are affected by where a script comes from. Whether it represents the sounds of a language well or poorly does not matter to what the shapes should be.

But Chinese script is connected to the underlying language. Most characters are built from a radical and a phonetic, so that after memorizing a limited number (about 800-1000) of basic characters, a Chinese person can read any character. (The proper "spelling" of uncommon characters, which of several choices of phonetic to use, still has to be memorized, but English has a similar problem.)

It's only because individual syllables are more likely to carry meaning in the Chinese languages that they could get away with a character-based script; Japanese had to add a phonetic script for its inflections.

hrant's picture

{To Follow}

sevag's picture

Hi Aleme,
You wrote ‘My view is in most cases it has no connection.’ – could you elaborate a little bit more on that? Thanks.

John Hudson's picture

These are my current thoughts on this topic:

No particular script is linked in any limiting way to a particular language. This is clearly demonstrated by the ease with which scripts are adapted to multiple languages. Indeed, there are relatively few scripts anywhere in the world that are used to write only a single language: everywhere the tendency is for scripts to be adopted or adapted to multiple tongues.

Yet a writing system is a tool for capturing language, so connections do exist even though they are not limiting. Most of the time, I use the terms writing system and script as synonyms (generally preferring the former term, as script has several other meanings), but there is a more precise use of the term writing system that applies at the level of connection to particular languages or groups of languages. Aleme has already referred to the adaptation of the Arabic script to write Persian and Urdu. What is the nature of that adaptation vis à vis the system of the script? It is what I would characterise as an extension: the nature of the system, the mechanisms it uses to capture language, are essentially the same, only the set of signs is extended. Contrast this with the adaptation of the Arabic script to write Kurdish, though, in which the actual system is altered. To someone who can only read one or neither of the Arabic and Kurdish languages, pages of text in the two might look very similar indeed, but one writing system is an abjad and the other is an alphabet. One script: two writing systems.

I do think it is useful and important for type designers to understand how writing systems work as systems. One of the workshop I run periodically for the MA programme at Reading is focused on getting the students to think in these terms, beginning with a lecture using a diagrammatic approach to the structural representation of phonetic information in various systems.

Thinking about a writing system as a system means understanding how the script captures language, understanding it structurally and not just visually. How important is this? Well, visual understanding -- understanding the set of signs as graphic objects, what they look like individually and how they appear in combination -- can get you a very long way, especially for scripts with fairly simple structures, and there are doubtless good type designs that have been achieved with little more than this. But structural understanding provides insights that affect decisions about readability: if you understand the kinds of information that the reader is looking for in the text and needs to get out of the system, then you can make better decisions about how to present that information within the glyph shapes and their interactions.

There are degrees to this understanding, and at some point the truly useful and the personally fascinating are probably distinguishable. Most people understand that the European alphabetic writing systems all capture both consonant and vowel sounds as in-line letter signs, even if they've never articulated the understanding in such terms. Many, but fewer, also understand that many of these writing systems also employ digraphs, trigraphs, etc. to capture sounds not represented by individual signs. Many understand that secondary marks applied to the basic signs may represent distinct sounds, but fewer appreciate that these may be considered letters in the application of a system to one language but as diacritics in another. Some of this understanding is doubtless important to the type designer, but is all of it? What about an understanding of Turkic vowel harmony rules as reflected in the structure of the Uyghur writing system (something that Tom Milo and I spent an evening examining a few years ago)?

The great majority of linguists simply ignore writing as secondary (and mostly inaccurate) notation systems for spoken language. This is why phonetic transcription systems like the IPA alphabet exist: to make up for the shortcomings of natural orthographies in capturing speech, enabling the linguists to ignore writing. One of the few exceptions to this is Florian Coulmas, and his book Writing Systems: an introduction to their linguistic analysis is worth at least dipping into.

enne_son's picture

[John Hudson] “Most people understand that the European alphabetic writing systems all capture both consonant and vowel sounds as in-line letter signs, even if they've never articulated the understanding in such terms. Many, but fewer, also understand that many of these writing systems also employ digraphs, trigraphs, etc. to capture sounds not represented by individual sounds.”

The last word here (sounds) looks like a typo. Shouldn’t it be signs?

[John Hudson] “Many understand that secondary marks applied to the basic signs may represent distinct sounds, but fewer appreciate that these may be considered letters in the application a system to one language but as diacritics in another.”

Probably you meant to write: “may be considered letters in the application of a system to one language but as diacritics in [the application to] another.”

Can you provide an example of this?

R.'s picture

enne_son: Can you provide an example of this?

Maybe like this: The diaeresis on ‘o’ as in English ‘coöperate’ or Dutch ‘coöperatie’ is a diacritic that merely makes morphological segmentation easier. In Finnish, German or Hungarian, ‘ö’ and ‘o’ are letters that do not represent the same sounds.

enne_son's picture

In the writing of the French, é and è represent distinct sounds too, but don’t constitute distinct letters.
In orthography and collation, a letter modified by a diacritic may be treated either as a new, distinct letter or as a letter–diacritic combination. This varies from language to language, and may vary from case to case within a language.

I guess I’m wondering why this is, or how it came about. It doesn't seem to be a matter of relationships between how the sounds are made (articulation).

Nick Shinn's picture

Unicode will have the final say on this, with /é and /è entering the French alphabet.

Michel Boyer's picture

According to the French wiki Alphabet français, there are 42 letters in the French alphabet, including æ and œ and letters to which are added diacritics (the article does not consider the cedilla as a diacritic; it considers the letter 'ç' as a whole; I would like to know why).

John Hudson's picture

The convention whether to regard a sign as a letter or as a diacritic is practically related to sorting: e.g. does ö sort with o, or by itself. I don't think there is any rule or even general principle by which this develops: it is conventional and language-specific. There may even be debate among the users of a language over the question, as there is over the Dutch IJ digraph.

Nick, I don't understand your comment. Unicode doesn't have any say on what characters are treated as letters in a given orthography and which are not. Unicode provides a generic sorting algorithm, but expressly notes that this is insufficient for language processing.

John Hudson's picture

PS. Thanks for noting the typos, Peter. I have edited these.

Nick Shinn's picture

John, just as Unicode has been instrumental in establishing the distinction between characters and glyphs, I imagine that it will influence which characters are considered letters in particular orthographies, with the trend being towards diacritics becoming letters.

But perhaps French is not a good example for change, as even /œ is not considered a member of the French alphabet, despite its orthographic status.

John Hudson's picture

But Unicode had a specific technical motivation for distinguishing characters and glyphs: it allowed them to define an encoding principle independent of text shaping and to clearly define areas of responsibility between different levels and kinds of software. There simply isn't any parallel motivation in encouraging diacritics to be treated as letters in orthographies. Indeed, insofar as Unicode has anything to say about this at all, it is explicit that the kinds of operations that rely on the definition of a letter within an orthography -- of which collation is the principal example -- are the responsibility of higher level protocols, which generally have no difficulty capturing the vagaries of different orthographic sorting conventions.

Unicode Technical Report #10 defines the Default Unicode Collation Element Table (DUCET), which provides a generic sorting and string comparing method, but the whole point of DUCET is that it is customisable for individual languages. The default collation is only used when individual language information is not available. As soon as one is working with a specific language, the orthographic conventions of that language, including the the distinction of letter and diacritic, become available data. Unicode further provides the Locale Data Markup Language (LDML) to capture that data and make it available to software in a consistent and standard interchange format.

So I'd say that all of Unicode's efforts in this area are towards enabling language-specific conventions, not to influencing a trend towards treating diacritics as letters.

Aleme's picture

Hello every one ,
@Sans ,
When I say language and writing systems has very little connections .I mean speech in a given language and its written representation . For example in different part of China same writing system is read differently .Even in case of English and French where the language has changed since the writing system has been in use. The writing has become less reliable representing the spoken languages.
For type designers this is not a decisive factor but it is additional knowledge which leads to what John Hudson is talking . "thinking about and understanding how writing system works . The workshop John is giving is very ,very important especially for designers who work on different languages.
@John,I am very interested by what you have said"understanding how the script captures the language" do you mean the rule of the writing system?
Or the nature of the writing system. How it came about ?You also said
No particular script is linked in any limiting way to a particular language. I could be wrong but few scripts are created for specific languages Armenian and Georgian comes to mind ( if Hart is around he can tell us more )as far us I know this writing systems are not in use by any other nations .
Thanks a lot

John Hudson's picture

I really do encourage you to read Florian Coulmas' book. As he points out in it, the relationship of writing to language is complex, and it is also subject to the fact that spoken language changes more quickly -- not only over time, but also regionally -- than written language, which has a tendency to standardisation. What this means is that an orthography that starts off being largely phonetically based ends up, like English spelling, being etymologically based, i.e. what it captures about the language is information about roots and relationships, rather than about strictly how words are pronounced.

When I talk about understanding how a script captures a language, I mean understanding what kinds of linguistic information are present in the writing. Generally, there is a mix of information. As you've noted, Chinese script mostly captures semantic content, which means that the same text can be read by people who actually speak quite different languages. Alphabetic scripts mostly capture phonetic content, but because of the ways in which pronunciation changes this is seldom entirely accurate for long, and hence the writing systems end up capturing historical content (this is particularly the case in English due to the great vowel shift).

Elsewhere in the thread today we were talking about French diacritics. Here's another entertaining example of a writing system capturing something historical: the circumflex accent in French in many situations has no readily discernible effect on pronunciation, but rather marks where the spelling of the word used to contain a letter s. Hence, the circumflex in the French word fenêtre indicates that the word used to be spelled fenestre. The accent is maintained as a marker that captures the root relationship to related words, e.g. the verb défenestrer.

These are the sorts of things I mean by understanding how a writing system captures languages. Fundamentally, I think a basic understanding must be had of the core kinds of linguistic information present in the writing and how it is present. Hence my diagram, above, which shows the kinds of ways consonant (blue square) and vowel (red circle) information is present in Indian writing systems.

When I wrote that no particular script is linked in any limiting way to a particular language, I meant that even when a strong association links a particular script with a single language, there is nothing inherent in the script that limits it to this unique use. The strong association is produced by history and culture, not determined by the script, which can always be adopted or adapted to writing another language. And if you look carefully enough, you will almost always find it to be the case that this did, in fact, happen. Even in the case of Armenian, which is now associated solely with the Armenian language and culture, one finds that during the Ottoman period the script was actually used to write -- and indeed print -- both Ottoman Turkish and other Turkic languages. Tom Milo presented an example of Armenian-scripted Turkish orthography during his presentation at this year's ISType conference, making the case that the structure of this system provided the model for the eventual Latin-script orthography introduced in the 1920s.

Likewise, the Georgian script has been used to write, at least, Mingrelian and Svan, and probably other languages of the region.

Aleme's picture

Thanks a lot John.This is fascinating and eye opener .

BTW I had e mail exchanges with Gerry Leonidas.
He has given me valuable information .
Again thanks a lot.

Bendy's picture

Certain writing systems correlate written and spoken forms more closely, where the shapes of characters correspond to the phonological articulation of the sounds they represent. I'm thinking of things like the Inuktitut syllabary, shorthand or the Visible Speech notation.

quadibloc's picture

Of course, although æ and œ are not considered letters of the English alphabet, it used to be that printed works used them frequently for the correct spelling of words of Latin or Greek origin, such as fœtus or encyclopædeia.

John Hudson's picture

the shapes of characters correspond to the phonological articulation of the sounds they represent

Korean Hangul is usually cited as the best example of this, in which the shapes of the jamo components actually correspond to the shape of the vocal system during articulation. But, of course, such a system is as vulnerable as any other to normal variation and change in pronunciation.

To my knowledge, the Inuktitut syllabary is not based on such a correlation. Different phonemes are captured by rotation of a smallish set of signs, which seem to have been devised by Evans and Peck based on efficiency of typefounding rather than a relationship to articulation.

Bendy's picture

Thanks John, I was sloppy in thinking through what I meant. What I should have said was that in Inuktitut, each orientation consistently represents a phoneme, rather than letterforms corresponding to the phonetics of the sound's articulation.

Aleme's picture

What would be ideal writing system? Does it exist ?

John Hudson's picture

I think in order to have an ideal writing system you would need to have an ideal as to the information that it carries and how how that information is carried. Hangul seems an ideal writing system if the information you think a writing system should carry is strong representation of the articulation of spoken language. But writing systems exist to be read, so I would say that any ideal needs to be reckoned in terms of readability. That implies comparing writing systems as read by native speakers in terms of quantifiable measures such as speed, comprehension and fatigue. The trouble with such comparisons is that because different writing systems capture different kinds of linguistic information, it is difficult to isolate information type from representation type, i.e. you can't easily tell whether a measurable difference between writing systems is due to what they capture of how they capture it. So, for instance, might Chinese script be either better or worse than a Latin alphabet in reading performance tests because it captures semantics instead of phonetic and etymological information, or because of the shapes of the signs themselves?

enne_son's picture

On the question of the ideality of writing systems, see what Mark Seidenberg and what David Share have to say on optimality in their commentaries on Ram Frost’s recent discussion paper: ”Towards a universal model of reading,“ available at:
Mark Seidenberg’s commentary (“Writing systems: Not optimal, but good
enough”) starts in page 43, David Share’s (“Frost and fogs, or sunny skies? Orthography, reading, and misplaced optimalism”) starts on page 45.

I especially like the David Share suggestion that we think of writing sysrems as “dual-purpose devices that must provide decipherability for novice readers and automatizability for the expert.”

quadibloc's picture

I was somewhat amused by the claim that "position-invariant letter identification ia a key component of any universal model of reading", as a counterexample came immediately to mind.

However, I knew enough about that "counterexample" to know that it wasn't really one:

the Chinese do indeed identify 人 in 亽 and 亾 and 珍 and 琌... even if it isn't a "letter" in any sense.

Nick Shinn's picture

So I'd say that all of Unicode's efforts in this area are towards enabling language-specific conventions, not to influencing a trend towards treating diacritics as letters.

You’re right.
I had it backwards.
According to my logic, foundry type, in which the accent and letter were cast as a single item, would have been responsible for encouraging people to think of accented letters as distinct alphabetical entities.

The most persuasive concept is probably that if letter and accent overlap to form a contiguous whole, with no space between, then they constitute a distinct letter; as in Polish. Can you incorporate this principle into your boxes and ovals diagram (if you think it makes sense)?

John Hudson's picture

Nick, I think that principle may apply in some orthographies, but it certainly isn't a general one. So, for example, ö is a letter in some alphabets and a diacritic in others.

There was a related discussion about the Dutch IJ on the ATypI list recently, and one of the things I noted was that there seems a tendency for digraphs that cannot be confused in pronunciation (e.g. English VV, Dutch IJ) to become letters (W, IJ), while digraphs that can be confused in pronunciation (English CH, TH) do not become letters.

[The box and oval diagram above is specific to Indic scripts, and shows basic types of consonant and vowel information relationships. Not shown is consonant+vowel ligation that happens in some scripts. But this is independent of the question of what constitutes a letter or not in an alphabetic script.]

Syndicate content Syndicate content