Centre embedding

(n.) the insertion of one clause inside another

Our grammar works in such a way that we can insert an entire clause inside another and still have the statement as a whole make sense. A sentence like The boy thinks the girl is mean, for instance, contains a perfectly sound sentence—the girl is mean—within its confines.

Theoretically, there is no limit to the number of clauses a sentence can embed inside one another like this. But practice is considerably different from theory—especially when it comes to a confounding phenomenon called centre embedding.

Centre embedding is the insertion of a clause wholly inside another (known as the superordinate clause), so that unlike in the example above, words from the superordinate lie either side of the embedded clause. For instance, in a sentence like The boy the girl likes thinks she’s mean, the clause The girl likes is embedded entirely inside The boy thinks she’s mean. This gives us a structure something along the lines of y–x–y, where the shorter clause (x) is entirely surrounded by elements of the larger one (y).

Understandably, constructions like this make for fairly complex statements that, stylistically at least, we might prefer to reword. Exactly the same information in The boy the girl likes thinks she’s mean can be communicated like this: The girl likes the boy, but the boy thinks she’s mean. This longer version would likely be preferred in writing, because it presents all the necessary information much more straightforwardly. But in fast-paced, get-as-much-information-out-as-quickly-as-possible spoken language, we might find ourselves inadvertently producing the more succinct, yet trickier to interpret, centrally-embedded version.

That causes problems, however. There might be no theoretical limit to the limits of this kind of embedded sentence, but as soon as a third component is thrown into the mix, our brains tend to start to struggle to keep track of what’s going on, and which nouns ally to which verbs. In fact, some studies have shown that no more than three embeddings are ever used in natural language; no workable four-tier embedded sentence has even been found.

Precisely what causes us to struggle with embedded sentences like this is a matter of ongoing research, but the most obvious theory is that having a multiplicity of layered sentence elements simply saps our short-term memory. And with all our memory space taken up trying to keep track of who or what is doing what to whom, we struggle to parse (i.e. compute) the entire sentence.

The most oft-cited example of this phenomenon is as follows:

The rat the cat the dog chased killed ate the malt

Despite appearances, this is a grammatically perfectly sound sentence. But in practice, it’s all but indecipherable. That string of subjects (the rat, the cat, the dog), followed by their verbs (chased, killed, ate) proves mind-boggling, and we struggle to keep track of precisely who is doing the chasing, the killing, and the eating, to what and to whom.

In fact, it is the dog that chased the cat. It was that cat that killed the rat. And the deceased rat had eaten some malt. Put another way:

The rat that the cat who the dog chased killed had eaten the malt.

Not that that version is all that much clearer, of course, but it is a slight improvement. Again, if we were to need to ever actually communicate this event, we might stylistically prefer to reword it:

The dog chased the cat. The cat caught the rat that had eaten the malt.

But style is not grammar. There is no denying that The rat the cat the dog chased killed ate the malt is a perfectly sound sentence—it’s just not a particularly good one, nor, when it comes to communicating information, a particularly useful one. Grammar isn’t interested in how good or bad a sentence is, but simply whether it obeys the rules or not. The rat the cat the dog chased killed ate the malt obeys them perfectly, just not in a way that we would ever find particularly easy or useful.

The implications of this is that in creating our languages, we humans have inadvertently built a system that we ourselves cannot comprehend. Put another way, we now have a grammar that permits the construction of sentences that defy our comprehension.

That being said, given enough time (and perhaps a pen and paper), sentences like these can be understood—and once you get your head around who is doing the chasing, killing, and eating in the example above, you can probably begin to understand what’s happening when The shirt the man the neighbour saw washed blew away, or when The car the man the woman likes bought broke down. But with no upper limit to how recursive these sentences can be, even once you get the hang of them they can still prove impregnable:

The dog the girl the boy the teacher the mechanic the owner of the dog uses lives with likes knows saw ran away.