Tuesday, April 16, 2013

What is language?

Daniel Everett recently gave a talk on his language-as-a-cultural-tool idea at our UC Irvine Center for Language Science.  (Here's a decent review of his recent book.)  It was an entertaining talk but misses the point, in my view, regarding what language really is.

Today I'm reading a very nice review paper by Agnes Roby-Brami and colleagues on the relation between language and praxis (suggested by a sage reviewer of a submitted paper of mine).  It's worth a look.  Very informative and scholarly, but that's not what I want to talk about here.  Instead I want to focus on their definition of language:
Language refers to a system of signs (indices, icons, symbols) used to encode and decode information so that the pairing of a specific sign with an intended meaning is established through social conventions.
This definition is exactly the kind of conception about language that lends itself to Everettesque monologues about language as culture.  Yes, yes culture is reflected through language, but it doesn't mean that language IS culture.

But back to signs: how does the notion of language as a system of signs tilt the playing field toward a language as culture viewpoint?  It does so by focusing on the arbitrary aspects of language, those aspects that most closely related to "cultural conventions"; things like words.

It's not the signs that define language.  To see this, think about Roby-Brami and colleague's definition in visual terms:
Vision refers to a system of objects (forms, movements, and so on) that contain information so that the pairing of a specific object with its associated meaning is established through social conventions.
What?! Vision isn't a social convention, you say.  It's a neural system that analyzes visual form, motion, and location to transform physical information into conceptual representations or into motor patterns for interacting with those objects.  Vision is NOT in the objects themselves!

Of course you are right.  But we can easily think of vision as social if we misconstrue it.  Take any number of your modern objects and show them to a member of a hunter-gatherer society.  Take your iPhone, you coffee maker, your bicycle, your eyeglasses, your credit card, your zipper, your whatever.  They will have no idea what those objects are.  These "object-signs" have no meaning to them. We know what they are only because we have learned the object-signs through cultural experience.  It's as if we are "speaking" different visual languages!  Therefore,  we should conclude, vision is a cultural system.

We can see the flaw in *that* argument, of course.  What vision takes as input and what previous associates we have with particular objects and their uses doesn't define the perceptual/perceptual-motor system.  What defines vision are the computations that are used to extract form, motion, and location and associates these with higher-order conceptual systems, motor systems, and so on.

Why can't we see that the same flaw holds in the language argument?  The "word-signs" with their culturally agreed on meanings don't define language.  What defines language are the computations that are used to extract sound patterns from acoustic babble, to segment them, to combine them, and ultimately link them to complex conceptual representations or to motor speech gestures.


16 comments:

Bastien Boutonnet - Photo Diary said...

Ok. Clear definitions are a good things to have. However, defining language in such a way may push people to think that that's it. Language has repercussions on other cognitive processes and looking at language in that way allows for intellectually interesting avenues of research, which if carried out well are better than clear cut theoretical definitions.

Here are some examples...
http://www.pnas.org/content/106/11/4567

http://www.plosone.org/article/info:doi/10.1371/journal.pone.0008559

http://sapir.psych.wisc.edu/papers/lupyan_spivey_2008.pdf

Dan Mirman said...

It seems like a big part of your argument rests on whether semantics is considered part of the system. Some people consider object semantics part of the visual system, but many think of vision as all the pre-semantic stuff. On the other hand, I think most people consider lexical semantics part of language, so arguing that language IS speech perception is just as limited as arguing that language IS semantics. In any case, the cognitive and neural boundaries between perception (visual or speech) and semantics are probably not as sharp as the boundaries between academic sub-domains.

Greg Hickok said...

Hi Bastien,
Are you suggesting that my framing of the definition of language somehow precludes research of the sort you refer to? I don't see why that should be the case. The same sorts of studies could be carried out (and probably have been) in the visual domain.

Greg Hickok said...

You raise and interesting question Dan and I agree completely that the computational boundaries are not as sharp as the academic boundaries. You mention semantics, but I think the very same point can be made regarding phonology. The study of the "sound pattern" of language (phonology) is typically conceptualized as an amodal domain quite outside of the auditory system. In vision, however, abstract object representations (the "visual patterns" of objects) is never discussed as being outside of vision. Why the difference? Because phonology crosses the boundary between perception and production, one might argue. But this can't be the reason because object shape also crosses the perception-production boundary.

Re: your first statement. I don't understand why my argument rests on whether semantics is considered part of the system. I don't think it matters what you call it or where you draw boundaries (and I don't care much); language is that system that performs the transformations from acoustics to meaning or acoustics to motor gestures, etc.

Dan Mirman said...

I just meant that since semantics is the culturally-constructed part, how much semantics is in the system determines how much of that system is considered "cultural".

Another piece here, I think, is that everyone tends to overestimate the centrality of their research area. I (mostly) study semantics, so to me it seems like language is (mostly) about semantics. Phonology people think it is about phonology, syntax people (Chomsky?) think it is about syntax, etc. A while ago I was chatting with some socio-linguists and it was hard to convince them that language was anything other than a way of communicating race, class, age, gender, etc.

Greg Hickok said...

That's an interesting point and probably close to true for many language folks. Do you think the same is true in vision? Do people studying high-level vision (say face perception) think vision is mostly about face perception? Or would they acknowledge that their focus is just one part of a bigger picture that includes lower level feature extraction and so on? I think the latter is closer to the truth. We could ask the same of a vision scientist work on rods and cones in the retina. Do they believe vision is mostly about rods and cones, or would they acknowledge higher-level processes. Again, I would guess they are happy to view their work in the context of a larger computational system.

Why are language people different? What is it about language that makes us think about it so oddly?

Anonymous said...

“What?! Vision isn't a social convention, you say. It's a neural system that analyzes visual form, motion, and location to transform physical information into conceptual representations or into motor patterns for interacting with those objects. Vision is NOT in the objects themselves!”

“What defines vision are the computations that are used to extract form, motion, and location and associates these with higher-order conceptual systems, motor systems, and so on.”


To start with vision, the key words, for me, are “transform” and “associates.” Transform or associate here, especially transform into “conceptual representations” but also into more basic motor behaviors, has to do with the kind of creature one is. I often turn to the thought of a sighted, anthropomorphosized electron or photon. The line of object-medium-internal representation cannot come apart, or does so very difficultly, perhaps only in some more comprehensive scientific analysis. But the way we cut the world up, but also in general who we are and what we want (medium-sized-objects), means we normally “see” the “world” a certain way. And, thus vision or “our vision” is mediated and determined by our organism and our culture and the world we happen to find.

Can we talk merely about the eyes and brain processes, sure, but some important factors of our seeing will be understood only within the limits of more complex conceptual structures that organize such light into conceptual schemes or that we objectify, which is structured by evolution and then culture and personal experience. To understand “vision” I would argue requires the latter analysis, which is going to require the circumstance of environment, culture, personal history, and conceptual/scientific schema that allows us to see the “iPhone,” as you say. So, to a certain (and important) extent vision is a social convention and is personalized and species-directed. The important factors of the “vision” of any single creature will require us to explain, and will be inseparable from, the contingency of that specific creature. That may just mean we cannot fully understand vision without understanding more complex "representational" brain processes that vision is parcel to and partially constituted by.

One more example, we can explain the “vision system” of the crow, many of the mechanisms both of the eye and brain processing, but when we go to explain the “vision” of the crow AS IT sees and processes a plastic cap on a food product that it has learned to peck off, there now comes a need to explain “vision” in a way that surpasses some non-cultural, non-environmental rendering of vision. We have to explain the relation between basic sight mechanisms and how that fits into learning, conceptualizing, and higher representations that play back onto how sight will “order” the world, specifically one where this crow now sees a plastic cap and how to peck if off, or our human sees the “exit key” of his iPhone. Leaving out the latter capacity of classification and behavior would mean we have not explained visual processes.

Just my thoughts there . . . and that’s not to bemoan any specific scientific work on vision structures of the brain . . . I am only pondering broader definition and conceptual issues.

As to language . . .

Lyndon

Greg Hickok said...

Whoa there, Lyndon. Yours thoughts are plentiful and span quite a theoretical range from protons to culture. Let's rein them in a bit.

So it seems you're a Gibsonian. That's fine. Gibson was correct, in my humble opinion, to emphasize that perception occurs in the context of environments and the biology of the organism. I don't think anyone would doubt that today. I certainly don't. If you've read this blog for any length of time, you've heard me harp on 'ecological validity' -- a Gibsonian notion. And I would not argue with the view that the environment shapes perception in important ways (the system self-tunes to relevant features or reorganizes to some extent).

But! This doesn't mean that you can push all of perception out into the environment or even more so into cultural convention. Does culture determine the distribution of rods and cones on the retina? The layering pattern in the LGN? Axonal targets from LGN cells to the calcarine sulcus? Retinotopy? Yeah, we have to refer to culture and technology to explain why some of us hold a black rectangular object to the side of our face and then talk to no one present, while others couldn't imagine doing such, but we have to look inside the head to understand how these same two people can segregate figure from ground, process form and motion, and represent the rectangular black object in the first place such that it can be grasped or recalled from memory at a later time.

I'm not questioning that understanding a system from the widest possible perspective requires making sense of the biology and computational networks that provided by evolution and also how that system is shaped by personal experience and culture. What I find curious is that in the case of language many people find the latter the only relevant topic of scientific investigation and ignore the biology. We are far less likely to make this mistake for vision. So again I ask, why is this the case?

VilemKodytek said...

A reason may be that, unlike philology, there is no ophtology.

Marc Ettlinger said...

This reminds me of the of the time I was on some panel with a rhetoric PhD who studied language, or language and I thought it'd be great, we'd have so much to discuss. Needless to say, we didn't. I care about the cognitive and neural instantiation of language in the human mind and brain. He was interested something else I didn't quite understand. But it made me want to understand more.

Fast forward a few years and many pages of Bourdieu and Saussure, and I get the general gist. Language is a thing to be studied in and of itself. Indeed, prior to Chomsky, this was the dominant way of thinking about language - as a thing to study and analyze. And Chomsky's view, at least the first few decades, was to study the system of language itself with the idea that that will provide insight into the neural mechanisms involved and not vice versa.
Personally, I've come to believe that everyone studying language should at least have to read some Saussure, even if they are only interested in aphasia, for example.

The cognitive computations we are so interested in are things that shape and influence language, in this view. But so are a lot of other things, like population dynamics, sociology, society as a whole, economics and so on. So, instead of neurolinguistics being the sine qua non of language research, it's just one of many ways of understanding why languages look the way they do.

I now don't see either approach as superior to the other. I have my personal preference, but studying language qua language is definitely a worthy endeavor that has provided tremendous amounts of insight.

Anonymous said...

Hi Greg,

I was trying to work within your definition of vision, and I think it is telling that your definition did not point to only the narrow aspects and processes of rods and cones or the LGN, for example. If we fully understand the exact process of light hitting the eye and the immediate processes that happen, that is of course important, but it will not constitute “vision” as most people understand it and as even you define it. A similar thing holds for what many people want from a theory of language.

From your short blog post I am not entirely sure what you represent others’ language positions to be and what your exact problem with them is, but for me, the capacity to cognize or conceptualize and put into linguistic use those cognitions and conceptualizations is infinite, and is capable of being immensely changed by the most radical and strangest environment we could think of. The basic structures that are postulated to derive from the brain, which certainly there are important structures worth analyzing there, are still always mediated by the environment those beings find. There are limits to survivability and to conceptualization and thus to linguistic representation by humans, but language capacities and structures cannot be reduced to and explained only by the structures of brain systems (except perhaps in a far more complicated way) or by only analyzing the languages we find in the world.

We can imagine, say artificially creating, a bizarre environment and linguistic system, raising individuals in such, and I would claim these beings would still conceptualize and behave in robust ways. Will some of the cognitive processing and abilities be limited and controlled by important structures inherent to all homo sapiens brains? Yes, but we can imagine enough radical differences to understand that language and the conceptualization of phenomena can come apart from the limited brain/mind structures of any specific brain we find today and that some claim as universally grounding certain aspects of human language. Which again, some universal structures of brain limitations are present, but what we are trying to explain and do with language, just as what we want from a theory of vision, is going to surpass those narrow claims.

To go further still, there is an infinite many conceptualizations and possible worlds out there, and there will be significant relational differences between different conceived or experienced worlds. There is good reason to believe that we have an infinite capacity to cognize and represent, and thus to signify and to create different relational aspects of those language processes.

Lyndon

Anonymous said...

Unless you think there are basic anatomical differences in the brains of human beings from culture to culture, I suggest that before you dismiss the idea that language/culture shape cognition it would be good for you to consider the work of Lera Boraditsky http://psych.stanford.edu/~lera/papers/

Regards,

Larry Irons

Greg Hickok said...

Hi Marc,
It's not a question of superiority. It's a question of what research question you want to answer. If you are interested in the neural computations that transform thought into speech and back again, then cultural variation isn't that helpful in answering the question. If you are interested in sociology, then you can find reflections of it in language for sure and therefore language will be a good (but far from the only) data source. But you have to be clear about what you are studying. I'm questioning folks who claim to be interested in the neural computation question but then define language in social terms. It doesn't make sense.

Greg Hickok said...

Larry,
These sorts of language/thinking relations are on the very fringe of the problem that neurolinguists are typically interested in solving. So what if one culture uses cardinal directions (north, south, east, west) and another uses relative terms (left of, right of) to describe spatial relations? This doesn't explain how the sound pattern corresponding to these terms is analyzed acoustically, how it activates some form of a lexical representation, how it is combined with other words in the sentence and ultimately how some conceptual representation is activated in the first place; or vise versa going from thought to spoken language output. To say that culture shapes language (and it is not at all clear it does in this spatial example) just misses the point.

Marc Ettlinger said...

"But you have to be clear about what you are studying. I'm questioning folks who claim to be interested in the neural computation question but then define language in social terms."

I would certainly agree with that, but I'm not seeing where sociolinguists or cultural linguists are claiming to be interested in neural computation.

The only example of culture+language+neuron research that I can think of off the top of my head is the color terms imaging study from Rich Ivry's lab, which was a really great paper.

http://linguistics.berkeley.edu/~regier/papers/lateral-whorf.pdf

Greg Hickok said...

I agree generally but have a look at Daniel Everett's work. He is using cultural arguments to make computational claims.