On Mon, Apr 5, 2010 at 2:11 PM, Jim Henry <[log in to unmask]> wrote:
> Syntax has remained more in sync between written and spoken languages
> than have phonology or morphology, but if you're going for a NLF2DWS,
> wouldn't the optimal syntax be very different from that of any linear
> mode such as speech?

It would; I was unclear there. I consider 2D syntax to be completely
different from 1d syntax, enough that they are basically completely
unrelated features.

So to be more precise, I think that 1D syntax is not a mode-relevant
spec feature for any medium (though it's certainly relevant at higher
level), and that a NLWS does not *have* 1D syntax (whereas all
temporal modes do). A NLWS does have 2D syntax, and *that* would
certainly be very highly constrained by fidelity to the medium.

As an edge case, my guess is that a 4DWS (a NLWS in which time is a
factor?) would not be *medium*-constrained on the 1D part, in that I
expect that any constraints on 1D syntax will be cognitive (e.g. most
important and/or new and/or independent information goes first, except
if getting someone's attention, in which case low importance / high
salience stuff should go first while they're only half-attentive).

Totally a guess though; I haven't thought this through well.

> speech, a modifier can only come before or after its head; in a 2D
> writing system, it might could come in any of four or six positions --

Or indeed, any arbitrarily large number of positions (I presume we're
not talking about a grid-based writing system), some of which might
not have the same properties (e.g. some might be with respect to
elements of the symbol itself, like within it for a container
relationship, whereas others might be directional like you suggest).

> Arguably (and often arguedly) natlang syntaxes are isomorphic to a
> particular type of 2D system, namely a tree -- but I thought, from our
> previous discussions of some years ago, that you wanted a NLF2DWS to
> be more complex and expressive than a tree.

I do. Graphs can be forced into isomorphy with trees, but IMO it's a
horrible kludge to do so. In any case, isomorphism in the logical
sense is not really the criterion, ne? We want things to be
*expressed* in a native way. Isomorphism is what you get when you
*ignore* how things are expressed.

>> 3. Each grapheme should degrade gracefully on lower resolution, such
>> that you loose the least amount of overall information possible (when
>> including effects of context).
>
> Indeed; and "lower resolution" I'd suggest means display on small and
> low-resolution screens, but also paper, whiteboard etc. on which
> someone is writing in a hasty scrawl.

Sorry if I wasn't clear; this is definitely one of the cases I meant
to be included.

And yes, your topology constraint would definitely help with this. In
the case of alphabets we know, you can see it as a matter of
word-shape. I'm not sure how to formalize the impact of context on
letter design as a property of letters themselves, but basically you
want letters in a 1DWS to be such that they make word outlines
maximally entropic. Something equivalent would hold for a NLWS, though
I wouldn't presume the existence of "words" in the usual sense.


Eldin: ground stomping is actually something deaf people do, like
flicking lights, to get someone's attention. ;-)

Re. "toothbrush speech", I occasionally have a serious inability to
articulate (neurological issue) that's at about the same level of
impact. It's fairly impressive how much one can get across just
through tone patterns - even though English is supposedly not a tonal
language. ;-)

I don't think SCUBA signals, military hand signals, semaphore, or Pig
Latin deserve to be called "language". They are useful communication
techniques that are either closed class or pure codes, thus failing
the usual tests for what's a language.


Brett: You should talk to John Quijada. He used to joke about a
language that has phonemic wardrobing. ;-)


On Mon, Apr 5, 2010 at 6:29 PM, Nika Mirovak <[log in to unmask]> wrote:
> Sai, the reason I advise against solid areas of raised paper

That's not what you said, though:

> be wise to stay with raised dots and not try to use solid raised lines or
> entire sections of raised "areas" on the page.

Raised areas are, of course, not discernable as areas. They would just
be unremarkable surfaces which have edges that are directional lines
(yet another kind!). I'd presume that for a tactile language, the only
way to describe features is in terms of *local* variation; it requires
sight to see an entire raised area at once. (Of course, you can feel
around the edges to figure it out, but then you're talking about some
kind of explicit diagram where the shape is directly semantic, rather
than merely being a "phonological" feature.)

What I was asking was specifically why not have solid *lines*,
contrastive with dotted (or otherwise textured) ones?


On Tue, Apr 6, 2010 at 7:54 AM, Matthew Turnbull <[log in to unmask]> wrote:
> Also something that the clan books bring to mind, what do one handed
> signer's do about thier predicament?

I've never met one, so I'm not sure. However, I have met signers with
significant lateralized disability (e.g. a weak/trembly arm) or who
are temporarily impaired (e.g. by carrying a cup of coffee).

There are relatively obvious ways to degrade signs to be 1.5-handed or
1-handed - you loose the nondominant hand's handshape (no *big* loss
really, it's usually redundant), some or all of its movement (this can
be confusing as signs like WHICH are recognized more by the global
asymmetric movement than by one hand of it), or the like. It's like
someone's mumbling - annoying but understandable.

My guess is that one-handed signers are significantly better at this, though.

- Sai