Theiling Online    Sitemap    Conlang Mailing List HQ   

Re: Multimodal language (was: Wordless language (was: NonVerbal Conlang?))

From:Patrick Littell <puchitao@...>
Date:Monday, July 3, 2006, 20:44
On 7/3/06, Kalle Bergman <seppu_kong@...> wrote:
> > > If I have understood things correctly, this could be a > problem, since sign languages don't rely exclusively > on a sequential delivery of phonemes, as spoken > languages do.
Actually, you've slightly misunderstood my aims, although understandably inasmuch as I didn't provide any implementation details. I'm not suggesting that we take, say, a spoken phonology and come up with a manual analogue of it. (And the grammar of Western European-style spoken language and come up with a manual version of that, too.) Like below, you noted that things like spelling out a spoken phonology and doing English in signs are too slow and cumbersome, and would put deaf users at a disadvantage. This is entirely true, but I wasn't suggesting anything so naive.
> Whereas spoken language is > one-dimensional (phonemes are spoken one at a time in > a long "row"), sign language uses all three spatial > dimensions in addition to the time dimension used by > spoken language.
Treating spoken language as one dimensional isn't wholly adequate, actually. We started out working with words as strings of phonemes, but the last 50 years have broadened our understanding of their nonlinear and suprasegmental properties. (Note: this doesn't mean that some language X can't be described in a wholly linear phonology, but rather that spoken languages have more options than just put this after this after this. English doesn't make much use of them, but they're there.) So even though sign phonology is indeed multidimensional, so is oral phonology. (This is not to say that the dimensions of sound space could make as many precise distinctions as body space does, just that a mapping between them need not completely impoverish sign's wealth of dimensional distinctions.) So let's take a practical example, dealing with one of the big stumbling points this project might come across: simultaneity. Spoken language does, as you say, mostly work by putting one thing after each other. But take a sign for "close it (the window) repeatedly". Quite a few things might be going on at once: the motion for closing, the classifier appropriate for windows, the movement to indicate habitual aspect. This superimposition is going to be one of the trickiest things to get right for a project like this. I don't suggest we leave things like this out entirely; the result would be, as you say, a very *unnatural* sign language. But on the other hand, working out a spoken language in which these simultaneously-occuring sounds are represented suprasegmentally will not lead us to an unnatural spoken language. It will lead us to a language very unlike English, yes, and possibly to a typologically improbable language, but not to something unspeakable. (As a side note, we often forget the wealth of features that might resonably be suprasegmental in a spoken language. The first thing that comes to mind for this project is tone, of course, but vowel quality features belong to more than one segment in languages with vowel harmony, consonant POA features in languages with consonant harmony, other features like voice or laryngealization... even *nasality* can be suprasegmental... for example, take a look at languages in the Yanomami or Macro-Je families.) ------------------------------- Anyway, let's play around with this. I'll take David's analogy (his section on the similarities and differences between spoken and signed phonology is very good if you haven't yet read it) and go from there. Here's my specification of the spoken language: Verbs work by the sort of root-and-pattern morphology we find in Semitic languages, in which the root meaning of the verb is indicated by consonants and some of the vowels, and other vowels are left unspecified, to be filled in as part of aspectual inflection. The root for "close" is t*t*, where * is a vowel slot. The inflection for completive aspect involves filling it in with [a]s, whereas habitual aspect involves [ai]s instead, and progressive is [u], etc. Furthermore, the language exhibits classification-by-verbs, aka classificatory incorporation, aka type IV noun incorporation, although instead of doing this by affixation or compounding this is realized suprasegmentally by tone patterns. Say, a high-high pattern for flat things, a high-rising pattern for cylindrical things, a low-low pattern for roundish things, etc. In this case, "close it (the window) repeatedly" comes out as "taitai" with two high tones. (Note: although this language is absolutely nothing like English, this sort of game is not "unnatural" for speech. All of these are perfectly reasonable things for a spoken language to do.) Now, take something like David's mappings of consonants to discrete signs, vowels to movements, and handshape to tone. If the phoneme realized orally as [tj] is realized manually as lateral hand contact, and [ai] as a vertical circular movement, and the high-high tone corresponds to a flat handshape... then we get a case where a natural sign and a natural spoken word correspond. The sign in which one puts two flat hands in contact while moving in a vertical circle is just spoken as [tjaitjai] with HH tones. And negation might be realized manually as a head-shake, but orally as a [+NASAL] suprasegment over the entire word, giving us [njainjai]. (Sign language still wins against spoken when it comes to the number of things that can be happening at once, but spoken language does have usually-unused resources that at least help it catch up.) -------------------------------- For instance; signers often place reocurring
> concepts in their dialogue at a certain position in > the space around them, and when referring to that > concept, they indicate the position in question with > their hands (this is a kind of pronouns).
This is so, and I'm not 100% sure we could find a way to implement this with sounds. It may just be a thing that gets left out in the end, just like subject agreement might be left out in order to allow classifiers instead. It wouldn't necessarily lead to an *unnatural* signed language, though, just as leaving out subject agreement or plural inflection doesn't lead to an impoverished spoken language. (On the other hand, I think that leaving out classification and aspect-as-movement might well lead to a unnatural sign language, so I think they should be on the short list of things-to-keep.)
> I have the feeling that this property of sign language > would make it problematic to create a language whose > underlying representation can be realized as either > speech or signs - at least, and this is, granted, a > big "at least", if one wishes the language to behave > as a natural sign language. > > Of course, one can always create a inventory of signs > which corresponds to an inventory of spoken phonemes, > and which are suppposed to be used in a strict > sequential manner like the phonemes of spoken > language. This would then be reminiscent of "signed > english" - that is, the convention of assigning a hand > gesture to each letter in the english alphabet, used > to spell out english words in sign language. It is > illustrative, however, that signed english isn't used > as a native language by any community of deaf people. > The problem with signed english and similar > conventions is that they're simply too slow; one can > never achieve the same speed when signing letters as > when speaking, which everyone can easily convince > themselves of by looking up signed english on the web > and attempting to sign a few easy words as quick as > they can - and then comparing with the speed of saying > the words in question. Natural sign languages, on the > other hand, which use four dimensions instead of just > one, can easily compress their information rate to a > level equalling spoken language. > > If one aims to construct an auxlang which can be used > equally well by hearing and deaf, then this is the > wrong way to go; the hearing people would get an > obvious advantage, because the language would not > behave as a natural sign language.
Entirely true. As I mentioned above, though, a sign language rigidly constrained to behave like a spoken language wasn't what I was going for. The game isn't to try to force one mode of communication into the mold of another, but to try to find those similarities that allow for natural communication within both. Not easy, definitely, but I do believe possible. -- Pat


Kalle Bergman <seppu_kong@...>SV: Re: Multimodal language (was: Wordless language (was: NonVerbal Conlang?))