Friday, November 28, 2003

More Than Words For Snow

Ever hear that Eskimos have scores of words for snow? Apparently it's a myth [1], as their language has around the same number of words for snow as English does (snow, sleet, powder, slush, etc).

Whether language shapes and influences perception and thought is a hotly debated issue, and while I do think it does, I don't think it does so in the same ways as it is typically considered by people on either side of this argument.

The Eskimo words for snow are often cited as evidence of language as a shaper of perception and thought, and revealing it as a myth is often used as evidence that language does not shape thought. Another example I've heard concerns the fact that one language may have a term for a particular concept, while there may not be a way to directly express that concept in other languages.

Even if any of these were true, I think they would at most provide pretty weak evidence. And I'd say the same for the other 'evidence' I've heard -- I don't think they really capture the ways that language really does influence perception and thought. I'm not saying that no one has any strong examples, just that I haven't seen any -- in what I've read on the web, in magazines and in books [2] -- and certainly not in any of the popular discussions of this issue.

I want to present an example that illustrates a way I think language does influence thought, and fairly significantly influence it. The setting is as follows. There's a research group in a university computing department who have a novel idea for a piece of computer security software. The idea itself is quite general, and could be applied in a number of situations, and what they're currently looking to do is find some commercial interest that would allow them to pilot the system in a more concrete setting.

They are in a design meeting where they are trying to put together requirements for the system, as a starting point for any potential commercialisation. One member of the team is at the whiteboard, where he's written up two colomns, one saying 'needs to have' and the other 'wants to have'. These coloumns are referring to features the software could have -- and how critical the team thinks they are.

After these colomns have been drawn up on the board, one member of the team interjects that perhaps there could be multiple sets of 'needs to have' and multiple sets of 'wants to have'. The person at the whiteboard turns to the person and with a look of incredulity asks how there could be multiple set of 'needs to have'.

He goes on to say: if the software needs to have a feature, then it goes under 'needs to have' and if it doesn't need to have the feature, then the feature has to be something that we want the system to have or something we don't want the system to have -- there's no other possibility!

And regardless of how watertight that argument may sound, it's wrong. There can be multiple, different sets of 'need to have' because what features are critical (need to have) depends on the context the system will be applied in. As was mentioned earlier, the idea behind the software is quite general, and could be applied in a number of situations. The particular situation may influence which features the system needs to have, and which are merely desirable.

So if the argument was wrong, how can it sound convincing? Well, it's basically because the argument was based on the meaning of the words 'need to have' rather than on a consideration of the situation that those words were being applied to. If you just consider those words you have to conclude that you either need to have something or you don't need it, and that there's no two ways about it.

Essentially, the mistaken argument arised from putting words before reality. That is, thinking of the words first, then trying to think of the reality in terms of those words. We can see the truth by considering the reality first, then considering the words we're concerned with in terms of that reality. This means realising that we're considering features of the system, then thinking that the possible set of features will depend on what setting the ideas are applied to, and that thus 'need to know' depends on the setting.

I'll stop there. Perhaps I haven't made the strongest case for why langauge strongly influences thought in this fashion. Though I've seen a number of other examples of this nature in the past, other examples aren't things I can come up with from memory, so I'll be on the look out for more examples in the future - watch this space. Another thing I'd like to get into someday, hopefully soon, is why I think people tend to put words first. I think there is a very general mechanism at work in our minds that gives rise to this way of thinking -- and much else. Anyway, I've had enough for tonight -- I'm tired and my lower back is killing me sitting here on the chair. Mumble mumble, grumble grumble, bah humbug :-).

[1] See, for example, here and here.
[2] I'm afraid that I haven't kept notes on what I've read, but I can recall Stephen Pinker's The Language Instinct and that there were some articles in New Scientist.

No comments:

Post a Comment