Readers and listeners can find it very difficult to process sentences that embed one clause in the middle of a 2nd clause and then embed that 2nd clause in the middle of a 3rd clause. In this post, I review an example that embeds one relative clause inside another.
Sentence with no embedding
Let’s start with a simple sentence with no relative clauses (or other sentences) embedded, as in example 1.
(1) The dog was yelping.
The structure of example 1 is simple and easy for readers and listeners to understand. Figure 1 shows one way to depict that structure. A noun phrase (‘the dog’) combines with a verb phrase (‘was yelping’) to form a complete clause or sentence.

The figures in this post treat all clauses and all sentences as containing a complementizer (C), within a complementizer phrase (CP). The complementizer is sometimes explicit (such as that in examples 4 and 5) but sometimes not expressed overtly (as in examples 1-3).
The term subordinator is sometimes used instead of complementizer. What is a preposition? – Language Miscellany
One embedded relative clause
If we embed a relative clause in example 1, we create example 2.
(2) The dog the cat was scratching was yelping.
Although this sentence is understandable, it’s structure makes it harder to process than example 1. As figure 2 shows, this is because the embedded clause the cat was scratching interrupts the outer (main) clause ‘the dog was yelping’. This structure forces the reader (or listener) to suspend processing of the outer (main) clause in order to process the embedded clause.

Centre embedding within centre embedding
If we go further and embed another relative clause inside example 2’s clause the cat was scratching, we produce example 3.
(3) The dog the cat the fox was chasing was scratching was yelping.
Example 3 is much harder to process than example 2. Indeed, processing it is so hard that many readers may completely fail to understand it at all.
To show more clearly what example 3 is saying, I repeat it as example 3a. This marks the most deeply embedded clause with italic type and the 2 components of the intermediate clause with bold type
(3a) The dog the cat the fox was chasing was scratching was yelping.
Figure 3 shows why example 3 is so hard to process.

The reader (or listener) must interrupt their processing of the outer sentence (‘the dog was yelping’) in order to begin processing the 1st embedded sentence (the cat the fox was chasing was scratching). But then, before finishing the processing of the 1st embedded sentence, the reader (or listener) encounters the 2nd embedded sentence (the fox was chasing) and must suspend the processing of the 1st embedded sentence.
Clause embedded at edge of another clause
Example 3 is so hard to process and understand because it embeds an (inner) sentence within the middle of a 2nd (intermediate) sentence and then embeds the intermediate 2nd sentence (including the embedded (inner) sentence within the middle of a 3rd outer sentence. Example 4 and figure 4 show one way to eliminate that feature of example 3.
(4) The fox was chasing the cat that was scratching the dog that was yelping.

In example 4, the reader (or listener) can process the whole of the top layer of the main sentence (the fox was chasing the cat), before starting to process the intermediate layer (the cat was scratching the dog). The reader then begins processing the inner layer (the dog was yelping).
That in relative clauses
English relative clauses can often start with an optional complementizer that. For instance, example 2 could instead read: ‘The dog that the cat was scratching was yelping.’ Yet including that in example 3 would make it sound even odder than it does without that. On the other hand, example 4 would sound odd without that.
Moving an embedded clause to the end of a higher clause
Unfortunately, example 4 isn’t an exact paraphrase of sentence 3 That is because sentence 3 is about the dog but sentence 4 is about the fox.
Example 5 shows one way to eliminate the centre-embedding from example 3 while focusing on the dog, rather than on the fox. Example 5 replaces:
- the 2 active verbs scratching and chasing with passive verbs; and
- the intransitive verb yelping with an existential phrase there was yelping.
(5) There was yelping by the dog that was being scratched by the cat that was being chased by the fox.

As figure 5 shows, these changes move the subjects of the 3 clauses in example 3 to the end of those clauses in example 5. AS a result, the following relative clause then follows that noun directly. Thus, in example 5:
- nothing interrupts the dog that was being scratched; and
- nothing interrupts the phrase the cat that was being chased.
It is widely held that it takes more effort to process passive sentences than to process active sentences. And, indeed, sentence 5 is harder to process than sentence 4. But sentence 5 is much easier to process than sentence 3.
I present example 5 only to help show where the problems lie in example 3. I’m not saying that example 5 is the best possible way to convey the meaning intended by example 3—only that example 5 is much less bad than example 3.
Conclusion
Example 2 is a little harder to process than example 1. But example 3 is dramatically harder than example 2, and is also much harder than example 4 and 5.
To put this in more general terms:
- Embedding one single clause in the middle of another clause produces a structure that is somewhat more difficult to process than a clause with no other clauses embedded in it. It is better to avoid these structures if a clearer structure is available.
- Embedding one clause in the middle of a 2nd clause and then embedding the 2nd clause (containing the 1st clause) within a 3rd clause is never a good idea. Processing that structure is exceptionally difficult, and often impossible.
Further reading
Examples 2, 3 and 4 are from Bever (1970)
Fodor (2013) discusses what makes sentences like example 3 so difficult to process. She suggests that the main explanation may lie in difficulties in packaging such sentences up into prosodic phrases.
Language Down the Garden Path: The Cognitive and Biological Basis for Linguistic Structures, edited by Montserrat Sanz, Itziar Laka and Michael K Tanenhaus (2013)
The Cognitive Basis for Linguistic Structures, by Thomas G Bever (1970), reprinted in Sanz, Laka and Tanenhaus (eds, 2013)
Pronouncing and comprehending center-embedded sentences, by Janet Dean Fodor, in Sanz, Laka and Tanenhaus (eds, 2013)