On succinctness
Posted: Sat Oct 29, 2011 11:21 am
Recently it struck me that natural languages tend to be more succinct than conglangs, so I started thinking about how they manage it.
First, what exactly do I mean by 'succinctness'? Not just that the language produces shorter sentences, on average (that would be 'brevity'), but that the same meaning is conveyed in a shorter sentence, on average. Shorter how?
- Fewer words for the same meaning? No.
- Fewer characters for the same meaning? No.
- Fewer syllables for the same meaning? Maybe.
- Fewer phonemes for the same meaning? Close.
- Least time to speak/read the same content? Yes!
Since using a stopwatch to test how long it takes to read a sentence is annoying, counting the number of phonemes is a useful first approximation.
What strategies can make a conlang succinct?
#1: Increase the number of symbols
- Make a large number of phonetic distinctions (the Speedtalk/Ithkuil strategy).
- Allow a large number of consonant clusters and polyphthongs (and thus syllable types).
- Give every short, legal sound sequence a meaning.
Increasing the number of phonemes beyond a certain point actually slows down the language, as speakers are forced to lengthen or repeat each phoneme to make certain the listener heard it correctly. Similarly, if every legal one-syllable word has a meaning, speakers have to be able to distinguish between every legal syllable in every context. In short, you can only take this strategy so far.
#2: Morpheme length inversely proportional to frequency
This seems to be the biggest advantage of natural languages. It's a lot of work to design a conlang this way, because you generally have to use the language before you can figure out which morphemes tend to show up most frequently, and changing the length of a morpheme sometimes affects how often you use it, leading to a lengthy cycle of frequent changes to the vocabulary.
#3: Fusion replaces long multi-morpheme constructs with a single (hopefully shorter) morpheme, at the cost of increasing the number of morphemes occupying the symbol-space.
- Give common compound words a single-root synonym.
- Give common affix combinations a single-affix synonym.
- Give common root+affix combos a single-root synonym.
- Give common phrases a single compound-word synonym.
Note that when you replace a combo with a single morpheme, the frequencies of the components of the combo go down, meaning that the components may have to have their lengths increased to match their new, lower frequency.
#4a: Don't repeat yourself
- Use pronouns or other proforms, or definite members of general categories, whenever you need to repeat something.
- Avoid repeating something at all by merging shared components.
(Basic example: "John reads and Bill reads" vs "John reads and Bill does too" vs "John and Bill read".)
- Use as little agreement as possible, as this is essentially repeating the same information in two places. (E.g., marking adjectives with the number and case of their head nouns, like Esperanto.)
- Avoid obligatory markings, since this forces you to repeat information that could have been stated only once.
#4b: Omit anything you can infer from context
- Make as few things mandatory as possible, since that forces you to state unnecessary information
- No dummy arguments. Just "Raining." instead of "It's raining."
- No dummy verbs. Just "The cup red." instead of "The cup is red." (Yes, a bit glib here. I hope you understand what I mean.)
- No obligatory markings. E.g., allow tense, aspect, and number to be inferred from context.
- Allow gaps in relative clauses instead of relative pronouns.
#5: Homonyms (context changes the meaning)
- Roots with different meanings when used with verbal vs nominal affixes.
- Affixes with different meanings when used on verbal vs nominal roots.
- Verbs provide context for nouns. E.g., bat (animate) vs bat (inanimate).
- Nouns provide context for verbs. E.g., run (a race) vs run (a company).
I suspect this is one of the least exploited strategies in conlangs, especially engineered conlangs with their emphasis on one-word-equals-one-meaning. There are lots more ways to do this. E.g., Lojban has an interesting variant where the same syllable has a different meaning when used as a morpheme in a compound word or as a stand-alone word.
I've considered embedding several domain-specific mini-languages into my main language, each one reusing the same vocabulary. The idea is to create a system of homonyms by finding a collection of morphemes that usually appear together and rarely apart, and then finding several such collections that rarely overlap with each other. In the main language, there would be a way of switching contexts, e.g., a way to say "I'm about to talk about economics, so interpret everything that follows as belonging to the economics mini-language.", and a set of words whose meanings depend entirely on which context is in force. Possible mini-language topics include numeric expressions (e.g., numbers, mathematical operators, and a system of units of measurement), particular semantic domains, like kinship terms, tense/aspect and temporal/spatial relationships, or vocabulary to allow one to answer with more precision the question "How do you feel?", and jargon for each profession (e.g., a "battle-language" for soldiers, with specialized jargon such as 'flanking', 'enfilade fire', etc., and short expressions for important situations like "I'm low on ammunition!" or "Someone just took a shot at me!").
Anyway, does anyone know any other ways to make your language succinct? Even if they're only narrowly applicable tricks or tips (like using gaps instead of relative pronouns). Also, feel free to post specific examples of succinctness, either in your conlang or some natlang, or give examples of verbosity, where you wish there were a way to state things more compactly in your conlang.
First, what exactly do I mean by 'succinctness'? Not just that the language produces shorter sentences, on average (that would be 'brevity'), but that the same meaning is conveyed in a shorter sentence, on average. Shorter how?
- Fewer words for the same meaning? No.
- Fewer characters for the same meaning? No.
- Fewer syllables for the same meaning? Maybe.
- Fewer phonemes for the same meaning? Close.
- Least time to speak/read the same content? Yes!
Since using a stopwatch to test how long it takes to read a sentence is annoying, counting the number of phonemes is a useful first approximation.
What strategies can make a conlang succinct?
#1: Increase the number of symbols
- Make a large number of phonetic distinctions (the Speedtalk/Ithkuil strategy).
- Allow a large number of consonant clusters and polyphthongs (and thus syllable types).
- Give every short, legal sound sequence a meaning.
Increasing the number of phonemes beyond a certain point actually slows down the language, as speakers are forced to lengthen or repeat each phoneme to make certain the listener heard it correctly. Similarly, if every legal one-syllable word has a meaning, speakers have to be able to distinguish between every legal syllable in every context. In short, you can only take this strategy so far.
#2: Morpheme length inversely proportional to frequency
This seems to be the biggest advantage of natural languages. It's a lot of work to design a conlang this way, because you generally have to use the language before you can figure out which morphemes tend to show up most frequently, and changing the length of a morpheme sometimes affects how often you use it, leading to a lengthy cycle of frequent changes to the vocabulary.
#3: Fusion replaces long multi-morpheme constructs with a single (hopefully shorter) morpheme, at the cost of increasing the number of morphemes occupying the symbol-space.
- Give common compound words a single-root synonym.
- Give common affix combinations a single-affix synonym.
- Give common root+affix combos a single-root synonym.
- Give common phrases a single compound-word synonym.
Note that when you replace a combo with a single morpheme, the frequencies of the components of the combo go down, meaning that the components may have to have their lengths increased to match their new, lower frequency.
#4a: Don't repeat yourself
- Use pronouns or other proforms, or definite members of general categories, whenever you need to repeat something.
- Avoid repeating something at all by merging shared components.
(Basic example: "John reads and Bill reads" vs "John reads and Bill does too" vs "John and Bill read".)
- Use as little agreement as possible, as this is essentially repeating the same information in two places. (E.g., marking adjectives with the number and case of their head nouns, like Esperanto.)
- Avoid obligatory markings, since this forces you to repeat information that could have been stated only once.
#4b: Omit anything you can infer from context
- Make as few things mandatory as possible, since that forces you to state unnecessary information
- No dummy arguments. Just "Raining." instead of "It's raining."
- No dummy verbs. Just "The cup red." instead of "The cup is red." (Yes, a bit glib here. I hope you understand what I mean.)
- No obligatory markings. E.g., allow tense, aspect, and number to be inferred from context.
- Allow gaps in relative clauses instead of relative pronouns.
#5: Homonyms (context changes the meaning)
- Roots with different meanings when used with verbal vs nominal affixes.
- Affixes with different meanings when used on verbal vs nominal roots.
- Verbs provide context for nouns. E.g., bat (animate) vs bat (inanimate).
- Nouns provide context for verbs. E.g., run (a race) vs run (a company).
I suspect this is one of the least exploited strategies in conlangs, especially engineered conlangs with their emphasis on one-word-equals-one-meaning. There are lots more ways to do this. E.g., Lojban has an interesting variant where the same syllable has a different meaning when used as a morpheme in a compound word or as a stand-alone word.
I've considered embedding several domain-specific mini-languages into my main language, each one reusing the same vocabulary. The idea is to create a system of homonyms by finding a collection of morphemes that usually appear together and rarely apart, and then finding several such collections that rarely overlap with each other. In the main language, there would be a way of switching contexts, e.g., a way to say "I'm about to talk about economics, so interpret everything that follows as belonging to the economics mini-language.", and a set of words whose meanings depend entirely on which context is in force. Possible mini-language topics include numeric expressions (e.g., numbers, mathematical operators, and a system of units of measurement), particular semantic domains, like kinship terms, tense/aspect and temporal/spatial relationships, or vocabulary to allow one to answer with more precision the question "How do you feel?", and jargon for each profession (e.g., a "battle-language" for soldiers, with specialized jargon such as 'flanking', 'enfilade fire', etc., and short expressions for important situations like "I'm low on ammunition!" or "Someone just took a shot at me!").
Anyway, does anyone know any other ways to make your language succinct? Even if they're only narrowly applicable tricks or tips (like using gaps instead of relative pronouns). Also, feel free to post specific examples of succinctness, either in your conlang or some natlang, or give examples of verbosity, where you wish there were a way to state things more compactly in your conlang.