For all the lively discussion set off by his forthcoming article in Science ("Numerical Cognition Without Words: Evidence from Amazonia", published online 19 August 2004), Peter Gordon deserves the thanks of everyone interested in human language, thought and culture. A bit of the discussion took place here on Language Log -- I suggested an analogy to a different skill, and Dan Everett sent a fascinating note reflecting on the issues from the perspective of his 27 years of working with the Pirahãs.
Earlier today, Peter sent a response, which I'm happy to be able to present below.
Don't believe everything you read in the press. If you read the Daily Telegraph you will learn that I have been married to Dan Everett's wife Keren for the last 20 years! Also, the glib headlines: "Whorf was right!!!!!" screaming out are also less subtle than the primary source. In the Science article, I first ask the question of whether there are concepts that you cannot entertain as a consequence of the language that you speak. I then allude to Whorfian theory at this point --sorry Sapir, sorry Boas. This is actually the only place where I mention Whorf in the paper, and I do not finish with some final crescendo that "Whorf was right!" So, I distinguish between "weak determinism" and "strong determinism", which is basically a distinction that derives from John Lucy and that he essentially gets from Brown and Lenneberg. The strong determinism question is 1. whether languages can be incommensurable (i.e., possess concepts that are not intertranslatable), and 2.whether such incommensurability can actually prevent you from entertaining such concepts. B&L suggested that the latter cases would not exist because all languages can express the same range of concepts only some do it more efficiently. Apparently even Whorf believed this, so perhaps my results are not supportive of Whorf in the end.
Mark Liberman asks basically whether it is language or practice that is at stake here with his imagined example of the non-throwing culture. Well, this question always comes up in some form or another --usually just "how do you know that it isn't just because they don't engage in counting that leaves them without number concepts?" And here's how I think about this issue. First, as I say in the article, one has to get a handle on what counts as an interesting case. For example, the fact that the Piraha have no concept of quark or molecule is not going to be an interesting case of determinism. How do we draw the line? Well, basically, if someone didn't know what a quark was, we would not question their command of English, just their scientific knowledge. On the other hand, if you ask some to give you 4 sticks, and they say "Uh what does "four" mean?" then you would have some serious misgivings about their command of English or the intactness of their parietal lobes.
So, let's imagine another Libermanesque culture (invent your own name). It turns out that they make no distinction in their language between definite and indefinite reference. So, we do a bunch of experiments that show that they cannot get their minds around this distinction either. The skeptic then replies: "Well, maybe it has nothing to do with not having words like "the" and "a", but that they just don't engage in making distinctions between definite and indefinite reference, and that is where the causal structure lies, not in the failure of the language to engage in such distinctions." It seems to me that this is a pretty dumb argument because distinctions between definite and indefinite reference are inextricably entwined with language, and so to attempt to separate language and use is pretty meaningless. No one claims that it is just the sounds of the words that give you conceptual distinctions, it is their meanings and how those meanings fit into culturally defined conceptiual systems of interconnected knowledge.
Where does number fit into the continuum between definiteness and quarks? I think it is closer to definiteness, because the practice of counting is inextricably entwined with the words (or signs) we use for number. Research in the development of number ability suggests that we are born with the ability to exactly perceive and represent 1 to 3 elements in memory without counting, and that we can approximate larger numbers. This is precisely what you see in the Piraha (sorry I can't generate a tilde on this crappy computer in this trading post in the wilds of Maine where I am right now). The thing that bootstraps you beyond the small-number exact enumeration, into the realm of 4, 5 and to infinity and beyond, is the language of number. There is no way to do this (at least within the natural bounds of human experience) that does not involve some symbolic representation of exact quantities.
If we now take Liberman's example of "throwing", the parallels break down. Sure, you might question someone's knowledge of English if they didn't know the the word "throw", but I must confess, that I do not know the technical difference between a "lob" and a "toss" and a "hurl" -- maybe the latter is a bit faster? It's a bit like knowing that Elms are trees, but I would not bet more than 10 cents on my abiltity to identify one. It seems to me that you could develop a very cognitively complex representation of throwing distinctions by engaging in this act without using language. For example, baseball batters develop the ability to predict how a pitch is going to come by the configuration of the pitcher as the ball leaves his hand. There is no vocabulary for this, but is something that baseball batters develop. It's also possible that having words for different kinds of throws could (contrary to my own experience) engender some categorical perception for different kinds of throws (move the arm below 20 degrees and it's a "lob", above 20 degrees and it's a "toss"). So, in this case, the language might be crucial. But this is all an empirical question --it's why we do experiments.
My claim then is that because language is so intimately tied to counting, it basically makes no sense to ask whether it's language or counting that is important in acquiring exact numerical abilities. Personally, I think that Whorf was wrong about many things he said. I also think that the Piraha number case is just an existence proof for incommensurability and, in the absence of further empirical inquiry, should not be generalized beyond this case.
[email from Peter Gordon to Mark Liberman, 8/27/2004, for posting on Language Log]
Posted by Mark Liberman at August 27, 2004 01:32 PM