Since other replies are drifting in this direction, I’ll reply to my own post with a comment about Heinlein’s fictional conlang Speedtalk, to which Ithkuil has been compared. Like a lot of people, it was one of the ideas that got me interested in conlangs. But after a bit of research I concluded that it wasn’t a fruitful direction to head in. I ran into some research in which the rate of information transmission of various natural languages was compared. It turns out that in languages that are spoken faster, as measured in phonemes per second, the information carrying content, measured in bits per phoneme, is smaller. The result is that you really don’t seem to get a lot of bang for your buck by monkeying around with your language design to try to increase the rate of information transmission. The bottleneck at the high end seems to be in the processing capacity of the brain, not the structure of the language.
I think a core feature of a new conlang should be that it has a systematized way to express concepts. After making my first draft of a language I decided to dig deeper into modern applied ontology. I would recommend Barry Smith’s and Katherine Munn’s Applied Ontology: An Introduction.
The better you understand the ontological structure of the world the better you will be able to design a language that can precisely describe the ontological structure of the world.
The result is that you really don’t seem to get a lot of bang for your buck by monkeying around with your language design to try to increase the rate of information transmission.
If you have a sentence like “Four plus five is nine” that’s very hard to transmit in a language like Pirahã that doesn’t have a word for four but has only “one, two and many”. It might be able to say two-two for four and two-two-one for five but it’s really hard to express the sentence.
I think English does have cases where it’s a bit like Pirahã. We have four cardinal directions and if we want to go in a 45° angle we say north-east.
Base 10 is integrated into our language in a way that we can’t simply switch to base 12 or base 16 when we want to do so on a whim.
Another quite horrible case is the English word of “feel”. It mixes so many different cases together. You don’t have a similar distinction as between “see” and “look” for feel. Feel get’s used both to thing inside your own body and outside.
Láadan has:
loláad = to perceive internally, to feel a mental state or emotion, perceive with the heart (metaphorically). Passive internal feeling. “I feel sad.” lowitheláad = to feel, as if directly, another’s feelings (pain/joy/ anger/grief/surprise/ etc.); to be empathetic, without the separation implied in empathy dama = to touch, to feel with the skin. Active touching, feeling. “I am feeling the texture of the yarn.” (see láad oyanan, passive touch) náril = to feel internally, to fix your internal attention actively upon something, to-continue-to-present-time. Active internal feeling. “I am feeling angry.” (see loláad, passive internal feeling) láad oyanan = to perceive with the skin, to feel something on your skin, touch. Passive feeling. “The silk feels good to me.” (see dama, active touch)
When you say “Alice is afraid of Bob”, the sentence doesn’t transmit whether the author is speaking about a feeling or an emotion. Modern psychology has distinct notions of feeling and emotion but it’s hard to express it, because the language hasn’t really caught up with it. That’s again a situation where it’s important to familiarize yourself with the modern ontology instead of simply copying the ontology of the existing language.
Having a richer language that can express more shades of meaning is a reason for people write literature in the language instead of writing in English. Literature that also can’t be easily translated back into English.
Having different words for different concepts is great, but creating a new vocabulary will not be enough. There needs to be some training to make people use the words correctly. Otherwise they will just use the new words incorrectly.
This usually happens when people learn a foreign language that maps one word from their native language to two or more words in the foreign language. For example native English speakers often have problems differentiating “ser”, “estar” and “hay” in Spanish, which all get translated as “to be” in English.
Getting enough training will be especially important for a conlang, because you can’t just expect it to happen “naturally”, if all speakers of the language will keep making the same mistakes.
Having different words for different concepts is great, but creating a new vocabulary will not be enough. There needs to be some training to make people use the words correctly. Otherwise they will just use the new words incorrectly.
Yes, a language is more than just vocabulary.
Getting enough training will be especially important for a conlang
Duolingo style training could file that role well.
Apart from that the quality of the textbook for the a conlang is vitally important for it. It has to showcase the features of the new language.
Applied Ontology: An Introduction. The better you understand the ontological structure of the word the better you will be able to design a language that can precisely describe the ontological structure of the world.
Understanding the true ontological structure of the world is very non trivial, and you might want an improved language to do it in before get finished.
Understanding the true ontological structure of the world is very non trivial, and you might want an improved language to do it in before get finished.
I don’t think that you need to know the ultimate truth to learn useful things from applied ontology to design a better language than you would design if you ignorant of applied ontology.
I think this is one of those cases where philosophy is helpful and it makes sense to read people like Barry Smith.
If you want to speak about obligations (may/should/must) it makes sense to not simply copy the existing words of the English language but first read serious philosophy on what kind of categories of obligations exist. Yes, the resulting language won’t be perfect but it will be better than the language that you will be building when you simply copy English.
I’ve looked into the subject of ontologies (I did research on knowledge base design years ago). The problem wasn’t finding ontologies, but finding non-arbitrary ontologies. That is, no matter how one ontology categorized entities, you could always find another that categorized them differently, and no non-arbitrary reason to select one over the other. And I didn’t want to give in to the temptation to just choose one and use it regardless. I finally gave up and decided that treating each concept in isolation (for the purpose of dictionary building) was better than using an ontology that some users might find highly counter-intuitive.
I finally gave up and decided that treating each concept in isolation (for the purpose of dictionary building)
What do you mean with that sentence? That you want to use the ontology of naive English?
If we would have a name for 75 that’s isolated from the name for other numbers it would be quite hard to do math.
Ordering enities into categories provides the possibility to systematize them instead of making everything a special case.
If you look at Lojban’s place system is a huge mess because it has specific rules for the places of every single gismu.
When it comes to feelings, I think the distinction of feelings/emotions/moods and physical sensations (pain/warmth etc) is highly useful.
It makes a language more difficult to learn to have more distinctions but it makes the language more functional. A person gains something when they learn it.
Agreed. This was one of my more painful realizations, that I might have to do more than one iteration of the conlang before developing a finished product, because there will be no way to understand the flaws in the first version well enough to correct them until after learning to speak it fluently.
Sorry, I don’t have a link for it. The result is just something that I remember reading about many years ago. I looked at the link that redding posted and while it probably isn’t the same paper (I think I read about this before 2011) the result seems to match what I remember. There’s a possibility that if the linked paper could be retrieved, then whatever I read may be in the bibliography, although I don’t know if I’d recognize it as such.
I ran into some research in which the rate of information transmission of various natural languages was compared. It turns out that in languages that are spoken faster, as measured in phonemes per second, the information carrying content, measured in bits per phoneme, is smaller.
This doesn’t seem likely to generalize reliably to specialist constructed languages. I believe that it is true it when it comes down to average people talking about everyday things, but specialist subjects use specialized jargon and a shared bank of knowledge to communicate very complicated ideas very quickly. As a simple example, words like xor and nand, once they are fully understood and become automatic, do increase the rate of information transmission; likewise introducing the concepts behind ‘bacteria’, ‘molecule’, and ‘atom’ results in much quicker communication about a certain aspect of the world.
If you are constructing a language to hold rational debate, it does make sense to increase information transfer by expecting, and teaching, the language learners to match complex concepts with simple words. This should mean, in practice, more information per phoneme per second in the targeted areas.
While there will be a trade-off between time spent learning the framework and ability to communicate quickly, most rationalists are happy to spend time learning useful frameworks.
Since other replies are drifting in this direction, I’ll reply to my own post with a comment about Heinlein’s fictional conlang Speedtalk, to which Ithkuil has been compared. Like a lot of people, it was one of the ideas that got me interested in conlangs. But after a bit of research I concluded that it wasn’t a fruitful direction to head in. I ran into some research in which the rate of information transmission of various natural languages was compared. It turns out that in languages that are spoken faster, as measured in phonemes per second, the information carrying content, measured in bits per phoneme, is smaller. The result is that you really don’t seem to get a lot of bang for your buck by monkeying around with your language design to try to increase the rate of information transmission. The bottleneck at the high end seems to be in the processing capacity of the brain, not the structure of the language.
I think a core feature of a new conlang should be that it has a systematized way to express concepts. After making my first draft of a language I decided to dig deeper into modern applied ontology. I would recommend Barry Smith’s and Katherine Munn’s
Applied Ontology: An Introduction. The better you understand the ontological structure of the world the better you will be able to design a language that can precisely describe the ontological structure of the world.
If you have a sentence like “Four plus five is nine” that’s very hard to transmit in a language like Pirahã that doesn’t have a word for four but has only “one, two and many”. It might be able to say two-two for four and two-two-one for five but it’s really hard to express the sentence.
I think English does have cases where it’s a bit like Pirahã. We have four cardinal directions and if we want to go in a 45° angle we say north-east.
Base 10 is integrated into our language in a way that we can’t simply switch to base 12 or base 16 when we want to do so on a whim.
Another quite horrible case is the English word of “feel”. It mixes so many different cases together. You don’t have a similar distinction as between “see” and “look” for feel. Feel get’s used both to thing inside your own body and outside. Láadan has:
loláad = to perceive internally, to feel a mental state or emotion, perceive with the heart (metaphorically). Passive internal feeling. “I feel sad.”
lowitheláad = to feel, as if directly, another’s feelings (pain/joy/ anger/grief/surprise/ etc.); to be empathetic, without the separation implied in empathy
dama = to touch, to feel with the skin. Active touching, feeling. “I am feeling the texture of the yarn.” (see láad oyanan, passive touch)
náril = to feel internally, to fix your internal attention actively upon something, to-continue-to-present-time. Active internal feeling. “I am feeling angry.” (see loláad, passive internal feeling)
láad oyanan = to perceive with the skin, to feel something on your skin, touch. Passive feeling. “The silk feels good to me.” (see dama, active touch)
When you say “Alice is afraid of Bob”, the sentence doesn’t transmit whether the author is speaking about a feeling or an emotion. Modern psychology has distinct notions of feeling and emotion but it’s hard to express it, because the language hasn’t really caught up with it. That’s again a situation where it’s important to familiarize yourself with the modern ontology instead of simply copying the ontology of the existing language. Having a richer language that can express more shades of meaning is a reason for people write literature in the language instead of writing in English. Literature that also can’t be easily translated back into English.
Having different words for different concepts is great, but creating a new vocabulary will not be enough. There needs to be some training to make people use the words correctly. Otherwise they will just use the new words incorrectly.
This usually happens when people learn a foreign language that maps one word from their native language to two or more words in the foreign language. For example native English speakers often have problems differentiating “ser”, “estar” and “hay” in Spanish, which all get translated as “to be” in English.
Getting enough training will be especially important for a conlang, because you can’t just expect it to happen “naturally”, if all speakers of the language will keep making the same mistakes.
Yes, a language is more than just vocabulary.
Duolingo style training could file that role well.
Apart from that the quality of the textbook for the a conlang is vitally important for it. It has to showcase the features of the new language.
Understanding the true ontological structure of the world is very non trivial, and you might want an improved language to do it in before get finished.
I don’t think that you need to know the ultimate truth to learn useful things from applied ontology to design a better language than you would design if you ignorant of applied ontology.
I think this is one of those cases where philosophy is helpful and it makes sense to read people like Barry Smith. If you want to speak about obligations (may/should/must) it makes sense to not simply copy the existing words of the English language but first read serious philosophy on what kind of categories of obligations exist. Yes, the resulting language won’t be perfect but it will be better than the language that you will be building when you simply copy English.
I’ve looked into the subject of ontologies (I did research on knowledge base design years ago). The problem wasn’t finding ontologies, but finding non-arbitrary ontologies. That is, no matter how one ontology categorized entities, you could always find another that categorized them differently, and no non-arbitrary reason to select one over the other. And I didn’t want to give in to the temptation to just choose one and use it regardless. I finally gave up and decided that treating each concept in isolation (for the purpose of dictionary building) was better than using an ontology that some users might find highly counter-intuitive.
What do you mean with that sentence? That you want to use the ontology of naive English?
If we would have a name for 75 that’s isolated from the name for other numbers it would be quite hard to do math.
Ordering enities into categories provides the possibility to systematize them instead of making everything a special case. If you look at Lojban’s place system is a huge mess because it has specific rules for the places of every single gismu.
When it comes to feelings, I think the distinction of feelings/emotions/moods and physical sensations (pain/warmth etc) is highly useful. It makes a language more difficult to learn to have more distinctions but it makes the language more functional. A person gains something when they learn it.
Agreed. This was one of my more painful realizations, that I might have to do more than one iteration of the conlang before developing a finished product, because there will be no way to understand the flaws in the first version well enough to correct them until after learning to speak it fluently.
I’m interested into that research. Can you link it?
Not sure if this is what KevinGrant was referring to, but this article discusses the same phenomenon
http://rosettaproject.org/blog/02012/mar/1/language-speed-vs-density/
Sorry, I don’t have a link for it. The result is just something that I remember reading about many years ago. I looked at the link that redding posted and while it probably isn’t the same paper (I think I read about this before 2011) the result seems to match what I remember. There’s a possibility that if the linked paper could be retrieved, then whatever I read may be in the bibliography, although I don’t know if I’d recognize it as such.
This doesn’t seem likely to generalize reliably to specialist constructed languages. I believe that it is true it when it comes down to average people talking about everyday things, but specialist subjects use specialized jargon and a shared bank of knowledge to communicate very complicated ideas very quickly. As a simple example, words like xor and nand, once they are fully understood and become automatic, do increase the rate of information transmission; likewise introducing the concepts behind ‘bacteria’, ‘molecule’, and ‘atom’ results in much quicker communication about a certain aspect of the world.
If you are constructing a language to hold rational debate, it does make sense to increase information transfer by expecting, and teaching, the language learners to match complex concepts with simple words. This should mean, in practice, more information per phoneme per second in the targeted areas.
While there will be a trade-off between time spent learning the framework and ability to communicate quickly, most rationalists are happy to spend time learning useful frameworks.