Understanding center embedding (3 hours)

↵ Back to module homepage

Now that we understand how relative clauses are built in English, we can try to understand the complicated sentences that you translated at the beginning of this module. First, let's try starting from a simple sentence and then adding relative clauses to make it more and more complicated.

  1. The ice cream melted.

This is a simple sentence. It should be easy to understand. But what if we want to be more specific? What if there is lots of different ice cream, and we need to clarify which ice cream is the ice cream that melted? We can use a relative clause! Let's treat "The ice cream melted" as the matrix sentence, and take "The boy bought the ice cream" as an embedded sentence. We can turn that sentence into a relative clause using our procedure that we figured out in the previous activity, and we end up with "The ice cream that the boy bought melted":

  1. The ice cream melted.
  2. The ice cream that the boy bought melted.

That is a little bit more complicated, but probably still easy to understand. But what if we want to be more specific? What if there are lots of boys around, and we need to clarify which boy bought the ice cream? We can add another relative clause. So instead of our embedded sentence being "The boy bought the ice cream", let's add a relative clause inside here to make it more specific. Let's say it's not just any boy, but it's a specific boy, a boy that the girl liked. So we have a matrix sentence "The boy bought the ice cream", and an embedded sentence "The girl liked the boy", and we use our procedure to combine these into "The boy that the girl liked bought the ice cream."

Now we have a sentence with a relative clause, "The boy that the girl liked bought the ice cream". But wait, we're not done yet! Remember that we're ultimately describing the ice cream; this sentence is just describing who bought the ice cream. So, we're still going to let this be the embedded sentence inside our original matrix sentence ("The ice cream melted"). Let's follow our procedure to create that sentence.

  1. First, we identify the head noun, "the ice cream", which is present in both sentences: "The ice cream melted" (matrix sentence), and "The boy that the girl liked bought the ice cream" (embedded sentence).
  2. Then, we replace "the ice cream" in the embedded sentence with that: "The boy that the girl liked bought the ice cream that."
  3. Next, we move that to the beginning of the embedded sentence: "that the boy that the girl liked bought."
  4. Last, we insert the embedded sentence after the head noun in the matrix sentence: "The ice cream [that the boy that the girl liked bought] melted.

So, to sum up, we have started with a simple sentence, and we have added relative clauses to make it more and more complex:

  1. The ice cream melted.
  2. The ice cream [that the boy bought] melted.
  3. The ice cream {that the boy [that the girl liked] bought} melted.

If you're like me, you might feel that sentences #1 and #2 are pretty easy to understand and sound like pretty good English, but sentence #3 is very hard to understand and sounds very bad.

As we can see, however, the sentence "The ice cream that the boy that the girl liked bought melted" is meaningful; it's not a nonsense sentence. There is some ice cream, and that ice cream melted. The ice cream was bought by a boy. And the boy was liked by a girl. Or, to put it another way: A girl liked a boy. That boy bought ice cream. And that ice cream melted.

Furthermore, the sentence "the ice cream that the boy that the girl bought melted" does not seem to violate any rules of English grammar. We created a procedure (i.e., a rule) that combines two sentences to make one sentence with a relative clause. When we combined simple sentences, that procedure was fine. There's no clear reason why that same procedure shouldn't also be fine when we combine bigger sentences. In other words, the sentence is grammatical: it can be formed following normal rules of English grammar, and it doesn't break any rules (unless we create arbitrary new rules just to explain this).

So we are left with a problem. The sentence makes sense, and it doesn't break any rules. So why is it so hard to understand, and why does it sound so terrible? Below let's look at two possible explanations.

Is there a grammar rule against having too many relative clauses in one sentence?

Like I mentioned above, it seems like the sentence doesn't break a rule, but maybe we just need to make another rule? For example, maybe our relative-clause-making procedure can only use simple sentences; maybe there is a rule that if a sentence already has a relative clause, you can't put it into another sentence as a big relative clause. Will that solve our problem?

Let's try another example to test this. Here are three sentences we can try to combine together using our procedures:

  1. The boy bought the ice cream.
  2. The girl likes the boy.
  3. The girl sang.

Can we combine these using our procedure? First let's combine α ("The boy bought the ice cream") and β ("The girl likes the boy") to create: "The girl likes the boy that bought the ice cream." Next, let's combine that complex sentence with γ ("The girl sang"), following our procedure:

  1. "The girl sang" will be the matrix clause, and "The girl likes the boy that bought the ice cream" will be the embedded clause. "The girl" is the head noun.
  2. In the embedded clause, replace "the girl" with "that" to create: "that the girl likes the boy that bought the ice cream".
  3. Put the modified embedded clause after the head noun of the matrix clause to create: "The girl [that likes the boy that bought the ice cream] sang."

Now we have two complex sentences; each one was created by following the same procedure to combine three simple sentences into one:

  1. The ice cream that the boy that the girl liked bought melted.
  2. The girl that likes the boy that bought the ice cream sang.

Do you feel any difference between these sentences? Does either sentence feel worse (or more difficult to understand?

For me, #1 sounds much worse, and is much harder to understand. #2 is not beautiful English, but it still sounds basically OK.

If you agree with or accept my judgments about these sentences, then it seems there cannot be a grammar rule forbidding us from making a sentence with too many relative clauses. If there were such a rule, sentence #1 and sentence #2 would both be equally bad. But if sentence #1 is worse than sentence #2, the explanation must be something else.

We could keep trying to think of more and more nuanced grammar rules that might explain the difference between sentence #1 and sentence #2. But instead, let's forget about grammar for now, and try another approach instead.

Is there a psychological explanation?

Before we think about this possibility, let me point out an important difference between the two complex sentences we just saw. To see the difference, let me show you the sentence again, but this time I will color-code the parts of the sentence to make it easier to remember which original sentences each part (each relative clause or matrix sentence) came from:

  1. The ice cream{that the boy [that the girl liked] bought} melted.
  2. The girl{that likes the boy [that bought the ice cream] } sang.

A key thing to notice here is that in sentence #1 (the harder, worse-sounding sentence) the innermost relative clause ("that the girl liked") is embedded completely inside another relative clause ("that the boy bought"). It's in the center of another relative clause, in other words, it is center-embedded.

On the other hand, in sentence #2 (the easier, better-sounding sentence), the innermost relative clause ("that bought the ice cream") is not embedded in the center of another relative clause. It's embedded at the end of another relative clause ("that likes the boy"), but not in the center of it.

This difference has some clear consequences. For sentence #1, we have awkward strings of three noun phrases in a row ("the ice cream that the boy that the girl...") and three verbs in a row ("liked bought melted"), which are pretty rare in normal language. For sentence #2, we don't have those.

To think more formally about why this center-embedding is a problem, we can think about psychology. So far we have only been discussing the linguistic or grammatical aspect of how relative clauses work. Grammar is the system of rules that govern how a language works. As you've learned about in the "Intro to linguistics" module, grammar is not necessarily rules from a school or textbook, but rather the informal "rules" that you naturally know as a speaker of the language. Linguistics is the study of how grammars work. So far we've only approached the relative clause problem by thinking about the linguistic issues, but we haven't considered the psychological issues. Psychology is the study of how your mind works, and the constraints on how your mind works.

One important aspect of psychology is memory. To understand a sentence with a relative clause, you need to use a lot of memory. Consider the simple sentence "The ice cream that the boy bought melted." If you are reading or hearing that sentence, you don't process the entire sentence at once; rather, the sentence gradually unfolds as the sounds reach your ears (or as your eye moves across the sentence on the page or computer screen). Therefore, by the time you read or hear melted, if you want to understand the sentence you still have to remember what the subject was ("ice cream"). And the memory issue is even more challenging than that. When you read "The ice cream..." you might assume that it's the subject of the sentence, and you might start trying to figure out what the verb is; in other words, your mind might be thinking, "Ok, what did the ice cream do?". But then you read "The ice cream that the boy...". When you read "...that the boy...", you must realize you're reading a relative clause. Now you must be trying to figure out the verb of the embedded sentence; in other words, your mind might be thinking, "Ok, what did the boy do to the ice cream?" But at the same time you are trying to understand the embedded sentence ("the boy... did something to the ice cream"), you also have to keep remembering what's going on in the matrix sentence: at the same time you're trying to understand "...that the boy bought...", your mind has to also keep remembering, "I'm still waiting to hear what the ice cream did!"

What that means is: understanding a relative clause requires a lot of memory. You have to use memory to understand the relative clause that's currently unfolding, and at the same time you have to keep remembering the matrix clause of the whole sentence, and you have to think about how they go together. And that's just a simple sentence with one relative clause.

When you try to understand a center-embedded sentence, the problem compounds. Reading "The ice cream..." is no problem; you are just trying to understand what the ice cream did. When you read "... that the boy ...", you need more memory: you are trying to understand what the boy did to the ice cream, and remembering that the main sentence is about ice cream, and trying to understand what the ice cream did. When you go on to read "... that the girl ...", it gets even worse: now you are trying to understand what the girl did to the boy, while trying to remember that the first relative clause is about the boy, and trying to figure out what the boy did to the ice cream, and trying to remember that the main sentence is about ice cream, and still trying to figure out what the ice cream did. So there is a period where you are trying to handle three unfinished sentences at once.

With the non-center-embedded sentence (#2) the problem is not quite as serious. No sentence is completely embedded in the center of another one, so you don't have to remember so many things at the same time.

This is a slightly oversimplified explanation of why center-embedding is so difficult. The important point is that the reason center-embedding is difficult to understand, and sounds bad, is not because of linguistic/grammatical rules, but because these sentences overload your memory capacity. In other words, they are not "wrong" or "ungrammatical"; they just require so much memory to understand that they cause your mind to crash. So the underlying reason for the problem is psychological—it's because of the limitations of human memory. But the grammatical rules of the language create the situations where a certain kind of sentence will become center-embedded (and thus difficult to understand) whereas another kind of sentence might not.

If you want to learn more details about the psychology of center-embedded sentences, see the following articles:

With this information about center-embedding, we can go one step further, to consider how psychological and linguistic constraints interact across different languages. But first, let's practice applying these concepts to some more problems.

In the reading you just completed, we talked about two kinds of sentences:

Recall, also, that in the previous taskof this module we also learned about the difference between subject relative clauses and object relative clauses.

What kinds of relative clauses are used in the difficult sentence: object relative clauses, subject relative clauses, or both?

What kind of relative clauses are used in the easier sentence?

Think back to the article you translated at the beginning of this module. Choose any TWO center-embedded sentences from the article, and try to understand what they mean and explain how they were formed.

For each sentence, do the following things:

Consider again the two kinds of sentences we discussed:

Try to translate these into grammatical Chinese. Try to make a literal translation (i.e., saying exactly what the English sentence says and preserving the same relative clause structure, without trying to change things to make it sound more natural or better); only make the minimal changes needed to make it grammatical (i.e., don't just put the words in the exact same order as they were in English, because that won't make a grammatical sentence in Chinese; remember that being grammatical is not the same thing as sounding good). In other words, take the relative clause procedure that makes relative clauses in Chinese, and apply that procedure exactly (without modifying the sentence to sound more acceptable) to get grammatical Chinese relative clause sentences.

(If you don't know any Chinese languages, you can ask a Chinese-speaking friend or classmate from another class to try it—but you might need to explain the sentences, and the idea of relative clauses and center-embedding, to them! Alternatively, you can try translating it into another language you know.)

When you have finished these activities, continue to the next section of the module: "Putting it all together".


by Stephen Politzer-Ahles. Last modified on 2021-05-01. CC-BY-4.0.