Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.
Which means "Bison from Buffalo (that other) bison from Buffalo intimidate (themselves) intimidate (other) bison from Buffalo."
When you understand the trick of reading the "horse" sentence, it isn't too hard to make sense of it (hint: add a that after horse). But the Buffalo sentence remains incredibly hard to make sense of. For another example of a tough, if not impossible, to follow sentence, see this post at Language Log.
A few days ago I had an e-mail exchange with someone on a mailing list about one of the second type of these sentences. The sentence at first bothered me, and then interested me. Unfortunately, my interlocutor wasn't as interested in getting to the bottom of things (indeed, seemed downright surly about the possibility that he might be wrong). So I dropped the issue on the list, but I'd still like to pursue it a little further.
The sentence is based on the triple homophone bare/bear (n.)/bear (v.), and it's supposed to allow infinite recursion. The sentence is:
Bare bear bear bare bear bear bare bear...
And then it is expanded with additional "bear bare bear" ad inifinitum
That is, "naked bears give birth to naked bears who give birth to naked bears", etc.
My initial instinct was that there are two problems with this sentence. The first involves the subject-verb agreement. For me, the noun "bear" doesn't admit a zero plural in this context.
In other words, in this sentence bare bear is a plural noun phrase. It's a zero plural because instead of adding an -s, we add nothing. A handful of English nouns always take the zero plural (e.g., sheep, deer), and some can go either way (e.g., fish), while others always use the ordinary plural (dog, bird, etc.). There are some nouns that some speakers use with the zero plural, but overall the use is rare, and I think bear falls into this category. Although I don't use it this way (and so my internal grammar sense wants to asterisk such sentences), others do. See this site about Michigan bears for an example.
As an aside, the Comprehensive Grammar of the English Language mentions that "[z]ero tends to be used partly by people who are especially concerned with the animals, partly when the animals are referred to in the mass as game" (p. 307), which may account for the usage on this web page, but I wonder if there's more to it than that. My own grammatically judgments are not all or nothing here. I have different reactions depending on whether bear is an actor in the sentence or the theme. The latter type of sentence I find better than the former.
So, (both examples taken from the Michigan bear site):
Black bear are considered opportunistic feeders, taking advantage of many seasonally available foods.
I find stilted, but OK. On the other hand
Bear eat succulent, new green vegetation in the spring after they leave their dens.
definitely grates on my ear. In googling for different instances, almost all the examples of zero-plural bear that I found were of the former type. That may suggest that even people who use zero-plural bear tend to avoid it when the subject is performing the action. (Or my googling skills may just suck.)
All in all, I can accept that "Bare bear bear bare bear" works as an English sentence, although it's a very marked construction. But what about the recursion claim?
Bare bear bear bare bear bear bare bear...
For this to work, we have to assume a null relative pronoun, of the sort that we find in a sentence like
The bear I saw in the woods seemed unafraid of humans.
Here, that is optional after bear. I won't go into a complete analysis of relative clauses, since I don't want to take the time to explain the difference between the subordinator that and the relative pronouns, but notice that in English, that is not omissible when the relative pronoun is the subject of the relative clause. In the example above, it's the direct object (cf. "I saw [the bear] in the woods."). But in a sentence like
The bear that raided our camp ate most of our food.
You cannot drop that. (cf. "[The bear] raided our camp.")
Yet this is exactly the construction we have with the putative recursion. Filling in a relative pronoun:
Bare bear bear bare bear who bear bare bear...
Notice that who is the subject of the relative clause. So it really can't be omitted without loss of sense. It's much easier to see the problem by changing the sentence so we can more easily make sense of it:
Jenny bore a child bore a child bore a child.
Notice that this does not mean the same thing as
Jenny bore a child who bore a child who bore a child.
At best, it's a repetition of the predicate, as in a song, and you really need to stick commas after the first two childs.
What raises the recursion claim above the level of a mere mistake for me is that the person who mentioned this sentence claimed to have learned it from Sidney Greenbaum, one of the authors of the Comprehensive Grammar. So I ask myself, what was Greenbaum, obviously someone who knows the grammar of English very well, thinking when he claimed that this sentence could take a recursive relative clause? Could it be that with sentences that are this hard to parse, our brains turn off and accept, or at least don't bother to challenge, the claim that the sentence is grammatical? I suspect something like that had to be the case here. When we're confronted with a sentence as tough to process as this one, perhaps we don't really bother to process it at all, and we simply look for a vague meaning. That's the only way I can see that Greenbaum would have made this goof.