What is the extent of Featural Writing?

Question

In an earlier question I found out that the earliest I can realistically have a featural writing system is the late Neolithic, or anywhere from 1600 - 800 BC. Working on this languages writing system I realized that there are likely much more information I can store in the glyphs than in Hangul.

What is the extent of information you can store in a featural writing system?

Answer 1

It's fairly common for the design of featural conscript glyphs to be at least partly informed by the manner of articulation and/or place of articulation, but you could also consider:

Articulatory phonetic information – glyph-level

• Place in the sonority hierarchy, or, roughly, the amplitude

  (For all examples I will assume a horizontal script comparable to English.)
  Position in the sonority scale could be represented by increasing lengths of ascenders or descenders, potentially giving every written word a silhouette that follows a low-high-low, low-high, high-low, or high-low-high pattern.

• Length or gemination

  A doubled or tripled ligature, an elongated horizontal stroke, or tick diacritics, could indicate when a consonant should be geminated; the number or length could indicate how long it should be held, if multiple lengths are possible

Pronunciation cues – syllable-/intersyllable- or word-level

• Tone

  In tone-marking scripts (Thai, Chinese Zhuyin and Pinyin, others), a diacritic indicates whether it's a falling, high, mid, rising, or low tone. In some scripts no diacritic indicates no tone, and in others it means a default tone. Tone changes the pronunciation and meaning of a syllable/word, as seen in this example:

• Emphasis or lexical stress

  In accordance with your conlang's stress patterns, a stressed syllable could be indicated by an alternate form of the nucleus (generally a vowel), or a diacritic applied to the vowel or the entire syllable. Changing the stress can, as in the English examples of conduct and conduct, differentiate a noun from a verb, or highlight the important part of a sentence (as in "are we going?" versus "are we going?" versus "are we going?").

• Hiatus or diaeresis of vowels

  Hiatus is a slight pause that occurs when a coda-less syllable abuts the nucleus of the next (as in co-operate); this would only need to be indicated if your languages uses diphthongs (two or more vowel glyphs written in sequence), which isn't exactly featural (though Hangul does it). It might also be useful if your script does not use spaces (as in many eastern Asian scripts) or linkages (as in Devanagari) to define a word space – hiatus would indicate the start of a new word when the preceding word ends with the same vowel and would produce an unintended third word if the pause is omitted.

• Intervocalic transformation of consonants

  Indicating whether or not an intervocalic consonant should undergo lenition could indicate the pronunciation pattern a word should follow. For example, intervocalic t → d (bottle → "boddle") may not change the meaning of the word, but could convey the social context or significance of the speaker or subject (e.g. "boddle" might be slang and carry additional connotations).

• If your language uses vowel harmony or metaphony, diacritics could indicate the scheme to be applied to given vowels

  If your language classifies vowels into classes (as Hangul does) there may be rules that determine which vowels can go together in a word, or in which order they must go. If you have well-established patterns vowels follow, a diacritic may indicate following vowels should be pronounced a particular way (the equivalent of other vowel glyphs, but their use conveys a particular meaning), or that a vowel is exempt from these rules (as it is a loanword, for example).

Structure or parsing – word- or sentence-level

• Prosodic stress

  Prosody can be similar to stress as mentioned above, but operates at a multi-word level to indicate things like sarcasm, focus, type, etc. Specific glyphs such as a mark between words or under/over-lines, could indicate where entire words or clauses should be stressed, how long to pause between them, if cadence should be applied, or if a particular phrase should be understood as an idiom. In English, prosody tells you if a spoken sentence is a question, and the question mark represents the sound change in the final syllable.

Answer 2

There are a lot of options for Informational symbols.

• Sound.
  • Phonetic Information.
    • Sounds, this is an obvious one, but the sounds are important here. You can even go further and make the individual vowel and consonants represent the way your tongue, teeth and mouth interact with each other.
    • Voicing, similar to sound, most conlangs include this without knowing it. For those who do not know, voicing is the act of vibrating the vocal cords while making the sounds; like the difference between F and V.
  • Volume.
    • Loudness, the volume of the voice, pretty simple here.
    • Stress, the emphasis of specific consonants related to others.
• Grammatical Information.
  • Word types, these can be helpful in determining the kind of word; for example, nouns and verbs.
    • Tense, an addition to the verb word type. Adding a symbol for the tense of the verb, ei; past, present and future. You can go a step further and add time independence, something English lacks.
  • Capitalization, while upper and lower case is obvious, the upper case is technically two case; sentence starters and proper nouns.
  • Breaking, there are three types of breaks; sentence breaks, word breaks and syllable breaks. In English, the former is done with periods, the middle with spaces and the latter with common sense. You can technically add symbols to your symbols to represent the former two and since English lacks a good way of telling the latter, all the freedom to you.
• Miscellaneous
  • Emotion, the feeling behind the words.

Answer 3

Another factor to add to your own answer is the concept of bits of information per character, also known as the entropy of information.

To illustrate what that means, let's imagine slowing down a play-back of a person speaking to an extreme level. Suppose the first thing we hear is the person making a /k/ sound. What word is the person starting to say? There are thousands of possibilities, too many to even guess.

As the playback continues, we next hear a short /a/ sound. Perhaps we can now make a guess. Kayak? Cab? Kalamazoo? Cassowary? Castinet? It is not certain.

Next, we hear /t/. Ok. Cat. That thing that's furry, thinks it's cute, says meow a lot. Ah, but it could be the beginning of catacomb. Or catalyst. So it is still not 100%.

The playback continues. No further sounds are heard. Finally, we have total certainty that the person has said the word cat. We know what idea the person is attempting to communicate.

This process is happening inside your brain at a much faster pace every time to read or listen. For each letter you read and each sound you hear, your brain has to choose between multiple possibilities of what might be coming, and can only make its final choice when it has sufficient information. The fact that there are differences between words like Cab, Kalamazoo, cat, and so on, that can only be discriminated when a sufficient number of letters have been read or heard, is referred to as the entropy of the language. The ability to decide between one word and another, like the 0 and 1 in a computer, gives you one bit of information.

It is estimated (as in the linked article) that English text has between 0.6 and 1.3 bits of entropy for each character of message.

In a graphical language, there is going to be many more bits of information per "character." As a slightly tongue-in-cheek example, if I see a road sign with a stick figure with a bucket on its head, there are many things it might be warning me about, so there is still a fair amount of entropy in this one symbol. However, it is clearly not warning me to beware of tigers, of ice on the road, or an exit ahead. So there are many bits of information in that symbol, possibly several hundred.

(I'm not sure anyone has ever studied the information entropy of typical road signs. Anyone want to help me write up the grant proposal?)

Finally, in a featural language, there will be fewer bits of entropy per character on average. In alphabetical language like English, each letter is approximately equal to one phoneme. In featural languages where each character can be less than a full phoneme, there is necessarily less information.

At this point, I can only guess, as I have no knowledge of any examples. Whether fictional Tengwar or real Hangul- I'm not even sure what the differences between the phonemes and features of the word "cat" are. My guess is that the order of magnitude is in the tenths of bits per character.