More people have thought about Escher sentences than I have--so bear with me on this one. What is an Escher sentence? Well, you've just seen one:
More people have thought about Escher sentences than I have.As Geoff Pullum notes, such sentences go down pretty well. So well, in fact, that most people don't even notice that they seem to be the semantic equivalent of hamburger. In this respect, these sentences are like Escher paintings--they have a prima facie coherent compositional structure that crumbles on closer scrutiny.
In a follow up post, David Beaver says, "It's actually clear in broad terms what [is] intended [by such sentences]." Roughly, Beaver claims, the intended meaning of an Escher sentence like the one above is the following:
Others have thought about Escher sentences more (i.e., in greater detail) than I have.Or, perhaps a bit more perspicuously:
Other people have thought more about Escher sentences than I have (thought about Escher sentences).But something along these lines seems quite right to me.
Now the first point I want to make is that Beaver's claim that the intended meaning of the sentence is pretty clear and Pullum's claim that the sentence is semantic hamburger are at odds with one another. If the meaning of the sentence is pretty clear, then it isn't semantic hamburger; and if it is semantic hamburger, then its meaning isn't all that clear. At the same time, however, both Pullum and Beaver seem to be right. What gives?
You might think that the resolution of this (second order) puzzle is straightforward. Pullum is claiming that, as far as the compositional semantics goes, the sentence is semantically infelicitous; Beaver, on the other hand, is simply making a point about what we can conversationally infer the speaker to mean by his utterance. Unfortunately, this simple solution is problematic. The puzzle of Escher sentences is that we don't notice their semantic infelicity. But one would think that any attempt to account for the intended meaning noted by Beaver in terms of conversational pragmatics would require that the semantic infelicity of the sentence be salient. [You might be able to get around this worry by appeal to some sort of interpretive laziness principle according to which audiences don't really bother first figuring out what a speaker is actually saying, but simply make an educated guess at what it would be reasonable for the speaker to be saying in context given some partial semantic analysis of her utterance.]
So if conversational pragmatics isn't the way out, what is? Well, it seems to me that Escher sentences raise an important question about the nature of semantic compositionality. They are puzzling because their apparent meanings cannot be derived from the apparent meanings of their parts (plus their mode of combination). Think of it this way. When you consider an Escher sentence holistically, it has a relatively clear and determinate meaning. However, when you consider it atomistically and try to build-up the meaning on the basis of compositional principles, everything goes to hell.
If that is right, it suggests that our interpretive capacities take into account holistic informational characteristics of linguistic constructions and don't simply generate meanings by way of "bottom up" recursion principles. In particular, I take Escher sentences to instantiate a quasi-idiomatic linguistic construction (call it the Escher construction):
More X's have [transitive verb]-ed NP than I have.The claim is that the Escher construction is a productive linguistic construction that has a determinate meaning in English. The determinate meaning is:
Other X's have [transitive verb]-ed NP more than I have [transitive verb]-ed NP.So the illusion of semantic felicity is not an illusion after all. Once we understand semantic compositionality right, Escher sentences are fully compositionally coherent (and, so, not Escherlike after all!).