Tuesday, June 05, 2007

Me Tarzan. You don't understand.

Over on the wall of the Truth Cave I recently etched a link to Geoffrey K. Pullum's Language Log post on apes and language. Pullum's voice is keen as ever and his insights are always worth noting. His post focuses on the silly credulousness in John Berman's piece for ABCNews. Read it. It's an excellent post.

Pullum's point (making it clear that apes can't understand English) is most strikingly made when he parses one researcher's claim: "Qualitatively, there is no difference between Kanzi's language and my language." He writes

Let's look at that first sentence Fields uttered. It has a preposed adverb in adjunct function (qualitatively) at the beginning of an existential clause with a singular postcopular noun phrase with a negative determiner (the determinative no) and a complement preposition phrase headed by between. The preposition between has as its complement a noun-phrase coordination in which the two noun phrases have contrasting genitive noun phrase determiners (Kanzi's and my, respectively). Bill Fields knows how to construct English clauses with typical syntactic sophistication.

Kanzi, on the other hand, sometimes presses the EGG lexigram on hearing a human say "Egg".

Some who would defend Kanzi's ability to understand language would be hard pressed to show that the ape is able to understand how any of those words are functioning. And I'm not talking about analysis. I'm talking about simple competence.

One commenter on the ABCNews site tries hard (with several comment posts) to assure readers that there is good evidence of the ape's English competence. Posting under the name sarunfeldt the commenter says of Kanzi:
He clearly understands spoken English sentences where syntax is critical. ... Apes, and particularly bonobos, are clearly able to communicate symbolically and to use a limited form of ordered syntax.
And in another comment
Kanzi readily distinguishes between the sentences - Kanzi, did you eat the benana?, Kanzi, eat the banana. Kanzi, give me the banana., Kanzi, better not eat that banana. and Kanzi, do you want to eat the banana?
At least sarunfeldt admits that intonation probably has a lot to do with the apes ability to "distinguish between the sentences."

Distinguishing is vastly different from understanding. And it's not a simple "matter of degree." An ape that distinguishes demonstrates little more than that he knows when to choose an appropriate response. Let's look back at the claim that Kanzi shows syntactic competence. The ability to recognize a few (or even many) frozen forms and respond differently to each one does not indicate syntactic competence. One important test is productivity and the ability to understand how a syntactic alternation changes a statement.

Let's look at a very simple example. I noticed this form in a typo that I committed. Imagine your dog is talking with words. Consider two sentences it might say, each missing a word.

1. (*)I have chewed bone.
2. (*)I have chewed bone.

Both sentences look the same. Both are clearly wrong (unless we use "bone" as a mass noun -- that'd be odd) tho the meanings are easy to work out. When assured that they are different and the missing word is the article 'a' the two meanings are then clear to the hearer. Let's assign them correct forms thus

3. I have chewed a bone.
4. I have a chewed bone.

I'll agree that both facts can be "known" or even "understood" by a dog. All it requires is a tiny memory. But could any non-human animal comprehend the critical role of syntax? A simple movement of the article makes the meaning unambiguous by relying on the simple rule that a verb following a determiner must be a participle (a converse of the rule that keeps the participle from preceding the article as adjectives must come after the article). Hence the following obviously contrasted phrases

running the machine
the running machine

lit a match
a lit match

eating the shark
the eating shark

giving the tree
the giving tree


And another simple little difference between sentences 3 and 4 is the verb have. In sentence 3 have is an auxiliary verb which must then take a main verb "chewed" as a complement. And in 4 it's a verb meaning possession and must be, since an article cannot follow an AUX verb. Only a main verb can. In the decontextualized fragment "...have chewed..." the first word is ambiguous: it could be an AUX or main verb, but in "...have a..." it can only be a main verb.

These are rules of English that you understood before reading this elementary analysis. This is what grammatical competence allows you know without ever having heard a sentence.

Let's move beyond the claims that Kanzi shows an ability to distinguish between some simple syntactic alternations. Has Kanzi ever shown the ability to produce any syntactically determined differentiation? And nothing like YOU.BYEBYE or ME.EAT.BANANA (which I doubt he would differentiate from BANANA.EAT.ME).

Let's see Kanzi express the difference between I have related a story and I have a related story using the rules of English he has allegedly acquired.


  1. I don't know this for a fact, but I've heard Russian doesn't have articles (or at least doesn't have the subtle distinctions with articles that English does), which, I've heard, makes translating Chekhov a little hard. I presume they have another way of differentiating between a chewed bone and chewed a bone--in Spanish it would be this:

    "He masticado un hueso"--I have chewed (the verb masticar conjugated in the present participle[?]) a bone (article, noun).
    "Tengo un hueso masticado"--I have (the verb tener conjugated to first person) a bone chewed.

    I guess the adjective following the noun, and having a specific word for the participle (or whatever) conjugation that doesn't stand alone like the English "have" simplifies things for them.

  2. The auxiliary verb--that's what I meant, I think. I'm rusty on the nomenclature.

  3. You remember correctly. Russian has no article. Languages without articles are common. And yet something as simple as an article can confuse translation.

    Confusion can be a sign of greater understanding, especially when the confusion is due to recognized ambiguity.

    Do these people who argue that apes can talk really believe that an ape is aware enough to be confused by an article?

    An ironic argument might be this: The fact that an ape will just as easily learn to understand simple Russian as simple English shows that the ape is not learning the languages -- the ape is merely learning symbolic extensions.


Thanks for reaching out.

You can also contact me at wishydig[at]gmail[d0t]com.