A fresh approach to English dangling modifiers

Recently defended:

Control in Free Adjuncts: The “Dangling Modifier” in English by James Donaldson. Ph.D. dissertation, Univ. of Edinburgh, 2020 (supervisors: Geoffrey K. Pullum and Nikolas Gisborn).

Donaldson presents a fresh approach to the topic, uniting a huge body of commentary on observed examples by reference to sentence processing in real time.

Below, the dissertation abstract and a “lay summary”. These are not (yet) for quotation: this is not yet the final form of the dissertation — as is common in academia, there will be some editorial revisions before final submission, though it’s to be expected that there will be no substantive changes. I provide the abstract and lay summary here because I think the leading ideas deserve to be heard and appreciated.


In this dissertation, I present an account of the control of free adjuncts that relies on incremental processing. While many free adjuncts are controlled by the subject of the matrix clause (1a), this is not always the case. Some seem to be controlled by non-subject elements within the matrix clause (1b), others are apparently controlled by the discourse topic (1c), and still others involve the perceiver of the matrix clause in logophoric control (1d). These control patterns have raised the ire of many grammatical prescriptivists, who often label such constructions as ‘danglers’.


  1. Turning the corner on his motorcycle, he saw a church.
  2. Turning the corner on his motorcycle, his grip began to slip.
  3. While under development, they put all other projects on hold.
  4. Turning the corner on my motorcycle, a church came into view.

There have been several explanations of these patterns. Many researchers see free adjuncts as obligatorily controlled by the subject (1a) except where this is not possible, in which case logophoric control arises (1b,d). But such approaches cannot account for (1c), in which the controller is inanimate and thus incapable of perceiving anything. Other researchers regard non-subject control as the result of either an attempt to establish semantic coherence between two apparently unrelated clauses or an exhaustive search for alternative controllers based on a complex set of syntactic, semantic, and pragmatic clues. These approaches predict processing difficulty whenever a mismatch occurs, but most language users process sentences like (1b-d) fairly easily.

My central claim is that the patterns found in adjunct control arise because the estab- lishment of control continues throughout the process of understanding a given sentence. The language user, on encountering a free adjunct, guesses at a suitable controller. Disruption occurs when another potential controller arrives that is at least as adequate as the current guess. I support this claim through analysis of an extensive collection of attested examples, taking care to cover the relevant syntactic, semantic, pragmatic, and processing facts. I also emphasise how important it is for theoretical and descriptive studies to make specific predictions that could in principle be vindicated or falsified by future work in historical syntax or experimental psycholinguistics.

Lay summary

Complaints about the use of English often appear in newspaper opinion columns and letters to the editor, but these grievances can only rarely be justified. There are some common targets, including split infinitives, stranded prepositions, and the use of less with countable nouns, but none of these is inherently bad; purging them might make your writing more formal, but it won’t make it any clearer or more effective.

The dangling modifier is different. Danglers can make the reader stumble and come to unexpected conclusions about who is involved in what sort of situation. Take this blurb for a mystery novel: “Violent, clever and funny, I loved it!” The person who wrote the blurb probably intended to describe the book as “violent, clever and funny”, but it sounds as though he is talking about himself. I have read this sentence many times and know what he wants to say, but that knowledge does not seem to make the incorrect reading go away.

Teachers and editors circle danglers with red ink and claim that they should always be understood with reference to the subject of the main part of the sentence (in this example, “I”). The problem with this advice is that even the most skilful authors do not follow it. While danglers can be difficult to interpret, they are naturally produced by nearly everyone who speaks or writes in English. It is puzzling that there should be such a mismatch between the way we understand language and the way we produce it. Are all danglers equally bad? If not, then why are some better than others?

There have been a few attempts to address these questions, but their answers have been incomplete. In some cases, not all of the naturally-occurring data can be accounted for, while in others, elaborate searches are required that seem to predict that these sentences should cause considerable difficulty for the listener. On the contrary, some danglers are so easy to process that they are nearly undetectable. Each of the prior approaches brings some insight, but none provides a comprehensive approach.

In this dissertation, I have attempted to tie the insights of these studies together using a single mechanism: the gradual processing of language. We do not wait until the end of a sentence to try to understand it. Instead, we jump right in from the start and do our best with the information we have. Every new bit of language we hear or read helps us to know more about what the other person is trying to communicate. We can better understand why a given dangler is good or bad by looking at the moment-to-moment experience of the person trying to make sense of what is being said. It turns out that some danglers can even serve an expressive purpose, making our writing better. They are not all bad.

Leave a Reply

%d bloggers like this: