Discursive stowaways and human logic (dangling participles part 2)

Dangling participles are less ambiguous than style manuals would have you believe.  They are subject to the same basic rule that governs all modifiers – namely, that human readers with functioning representations of the real world will give them the most plausible interpretations and move on.  At worst, they are just like a lot of adverbials or adjective clauses in that they could conceivably refer to multiple parts of the sentence.  More often, danging participles in common use are essentially idioms with set meanings, whether or not they share a subject with the main clause.  These are the ones you hear on the evening news – keep an ear out and you’ll catch quite a few.

I put together another survey after the last one to further investigate what may make a dangling participle seem more comprehensible or clear besides having the subject of the main clause as its subject.  Specifically, I was interested in a few things that seemed to be the most common implied subjects, and whether using these reliably made a dangling participle more comprehensible than other implied subjects.  My conclusion was not what I had expected.

If that hasn’t already put you to sleep, read on.

The most commonly used implied subjects for dangling modifiers are things I have elected to call discursive stowaways, referents “for free”.  I call them stowaways because they are always salient, if invisible, factors in any situation where language is used and therefore can be referred to without explicitly naming them.  My hypothesized stowaways are:

  1. The speaker, or S
  2. The discourse, or D
  3. The content of another clause, or C

As support for the constant availability of these for reference, I refer to some other grammar structures as examples:

  1. For S, sentence adverbs (“Sadly, he can’t do it”), which invoke the speaker’s perspective without naming him/her.  (The maligned “hopefully” is an example of one whose transition from regular adverb to sentence adverb is still in question by some.)
  2. For D, other adverbials (“On the other hand”, “In the same vein”) that refer to the topic or flow of conversation without needing to name it.
  3. For C, adjective clauses (“I lost my keys, which sucks”) that refer to the entire meaning of another clause without it necessarily being made into a gerund or being a noun clause (which might look like this: “I hate losing my keys, which sucks” or “I hate that I lost my keys, which sucks”).

To study these, I have deliberately avoided using well-worn phrases like “Having said that” or “Moving on”, since one of the clearest findings from the last survey was that these are almost always acceptable.

As I wrote last time,

Hypothesis #1: Dangling participles are less likely to trigger alarms when they are part of a well-worn and familiar phrase like “having said that”, regardless of the contents of the main clause.  Rephrasings of those idiomatic expressions will be judged as less acceptable even when their literal meanings are similar.  This is for well-known principles of idiomaticity, the same reasons we don’t demand normal grammaticality of phrases like “be that as it may”.

Hypothesis #2: Dangling participles seem less abrupt when their implied subjects are either the speaker OR the discourse/conversation/train of thought itself, even when these are nowhere to be found in the main clause.  This came from a bit of reflection that A) the speaker and the discourse are always potential, ambient topics – discursive stowaways if you will – and B) I know many transitional phrases and discourse markers in Japanese work like this (although of course nobody considers this ungrammatical because not only participial phrases but all kinds of phrases in Japanese often lack subjects).

Basically, the survey that forms the basis for this newest post is a deeper exploration of Hypothesis #2 above, with an additional proposed stowaway in C (the contents of another clause).

As it turns out though, my discursive stowaways are not to be the most decisive factor in comprehensibility of dangling participles.  The most powerful deciding factor, a human-like understanding of the world, subsumes my stowaways and numerous other factors, including the prescriptivist notion that participial phrases must always share a subject with their main clauses (see the very bottom of this post).  The ability to make a plausible interpretation from the information given and from unconscious but very human assumptions counts more than grammatical structure.  I was able to find this out by comparing the answers to questions about the sentences I wrote, and realizing that the neither the data nor my understanding of language supported a conclusion that structure itself, including references to stowaways hidden from surface structure, was more important in determining the comprehensibility of participial phrases.

Anyway, this is what I did and how I reached my conclusions.

The survey

I wrote eight sentences with participial phrases, including two each for S, D, and C as well as two other sentences intended to act as a ceiling and a floor on scores.  The fact that I had these last two sentences was actually crucial in my reaching the conclusion that I did, because as we shall see, the floor didn’t actually function as one, and for interesting reasons.

My ceiling was a participle that didn’t dangle; i.e. its subject was the subject of the main clause.  Some participants noticed this.

My floor was a participle that dangled in addition to not having one of my discursive stowaways as its referent.  If my theory about discursive stowaways lessening the awkwardness of dangling participles had been correct, this sentence would have had the lowest comprehensibility and clarity.

Each sentence was followed by 3 questions, intended to suss out comprehensibility (the possibility of at least one interpretation) and clarity (the ease with which a most likely interpretation is determined), as well as for surveyees to give comments on what factored into their decisions.  The first two questions were Likert-style (1-5), and the last was text input.  I had a total of 29 participants this time, mostly gathered from LinkedIn forums and Twitter.

The precise wording of the questions was:

It is easy to take at least one plausible meaning from this sentence.

(Strongly disagree – Disagree – Neither agree nor disagree – Agree – Strongly agree)

It is clear which meaning the author intends for this sentence.

(Strongly disagree – Disagree – Neither agree nor disagree – Agree – Strongly agree)

What factors affected your answers for this sentence?

This was probably a weakness of the survey.  At least one participant wasn’t sure what the first question meant:

Your first question is misleading. “At least one plausible meaning” – there’s a big difference between one plausible meaning and a two plausible meanings

If I do a survey like this again, I’ll have to think of other ways to suss out comprehensibility and clarity (as independent from “acceptability”, a concept which seems to trigger prescriptivism).

Anyway, the sentences (with discursive stowaway in parentheses) were:

  • Approaching the end of the chapter, the final topic is information technology. (Discourse)
  • Costing millions of dollars, the workers accidentally struck a water pipe while digging. (Clause)
  • Crashing into a tree, the airbag deployed automatically. (Floor)
  • Thinking of my job, there was one thing that clearly needed to be done. (Speaker)
  • Opening up the word processor, the student began to work. (Ceiling)
  • Causing the students to break out in cheers, the due date was delayed for a week. (C)
  • Handing in my resignation, there was nothing left to do but clean out my office. (S)
  • Returning to the main point, it is clear that further progress can be expected. (D)

Next, I’ll review the scores that these sentences received, categorized by type of stowaway.  First, the two sentences that were supposed to act as lower and upper bounds.

Floor and ceiling

Screen Shot 2017-06-24 at 23.37.23.png
The floor is the left bar, the ceiling the right.  Both bars have comprehensibility on the left and clarity on the right, which is why they seem to slope at the top.  Comprehensibility and clarity scores were 3.83 and 3.51 respectively for the floor sentence, and 4.34 and 4.31 for the ceiling sentence.  Once again, 1 = Strongly disagree and 5 = Strongly agree.

Here is the first wrench thrown into the works: my floor didn’t function as a floor at all.  In fact, other sentences scored even lower than this sentence that I had intended to be a stereotypically bad dangling participle.  As you can see, the participial phrase in “Crashing into the tree, the airbag deployed automatically” has as its implied subject either the car or the driver, neither of which is present in the main clause and neither of which is one of my proposed discursive stowaways.  Nonetheless it scored well above “Neither agree nor disagree” on average for both comprehensibility and clarity!  Some factors that participants gave as reasons for their judgments on this sentence were:

Takes inference get to get a complete picture of what happened.

Airbags don’t crash. [low scorer]

There is no other noun mentioned, and we infer that it is a car, but could an airbag maybe crash by itself if involved in some kind of experiment?

We know the airbag can’t crash into a tree. A car crashes into a tree and the airbag is then deployed. [high scorer]

Temporal circumstantial element for the development of the situation in the main clause. Circumstantial concatenation.

I could definitely have driven the score lower by intentionally introducing humorous ambiguity into the sentence, a la “Plunging down the rock face, we observed the stunning Victoria Falls”, but I didn’t want to test whether deliberately misleading dangling participles are difficult to comprehend, just if mismatched or missing subjects alone threatened the comprehensibility and clarity of participial phrases.  It appears that they don’t.

As for the ceiling (“Opening up the word processor, the student began to work”), it got the high scores most people would predict that it would.  These were indeed the highest scores among all the sentences presented, meaning that it can function as a ceiling for comprehensibility and clarity of participial phrases in this survey.  Some of the stated factors for its scores were:

In this case the sentence can be correct, with the student opening up the word processor and beginning to work. Unless there is clearly another agent opening the word processor – like the computer, or the processor is opening itself up, this is a correct sentence – if the context does not lead to confusion.

Following the order of what happened (intention to do so)

Is this a dangling modifier? I don’t think it is.

The student opening the word processor is an aspect of them having already commenced to work.

So the non-dangling participle did indeed score the highest, but the clearly dangling participle didn’t score as low as expected.  Both of these sentences presented logically explicable scenarios, which seems to have been a more significant factor in participants’ judgments than the lack of clear subject.  With the (non-functioning) floor and ceiling in mind, let’s look at the other sentences, separated into discursive stowaway categories.

S, the speaker

Screen Shot 2017-06-24 at 23.36.23.png
Again, both bars have comprehensibility on the left side and clarity on the right.  Comprehensibility and clarity for the left bar were 3.83 and 3.86, and for the right bar were 4.03 and 4.04.  The comprehensibility scores for both sentences were correlated with each other at 0.71, and the clarity scores at 0.65, showing that participants who accepted one tended to accept the other.

All the scores for these sentences were above the “Neither agree nor disagree” 3.00, which shouldn’t surprise anyone.  What is surprising, to me at least, is that the first sentence with the speaker as its implied subject only does slightly better than my floor sentence.  That is, “Thinking of my job, there was one thing that clearly needed to be done”, was rated only slightly clearer than “Crashing into the tree, the airbag deployed automatically”.  Some stated factors were:

Clearly, I am thinking.

Could be thought of many ways [middling scorer]

The possessive states a person who is the carrier of the participle and there is no other possibility since it is the only person who can think there.

Not as much context within sentence [low scorer]

The second sentence scored a bit higher, perhaps because it mentions the implied subject as a possessive pronoun in both the participial phrase and the main clause.  “Handing in my resignation, there was nothing left to do but clean out my office” got these comments as factors for its scores:

The first part implies the second

Directly apposed to the person behind “my”.

I would probably be the only person handing in my resignation and cleaning out my office. But here are many other possibilities. Maybe I have a diagnosis of a fatal disease, and my husband or best friend hands in my resignation and cleans my office. [high comprehensibility, low clarity]

General knowledge. Knowing what happens next woensdag you do this. [autocorrected into Dutch?]

Notably, many of the comments have explicitly referred to logic and context as reasons for scoring – the second sentence, which named its subject (in possessive pronouns) in both the main clause and the participial phrase as well as featuring an easy-to-picture physical action, was considered more comprehensible and more clear than the first, with its hard-to-visualize main clause which doesn’t mention the speaker.  This implies that my hypothesis – that the speaker (S) is always available for reference “for free” and makes otherwise dangling participles more comprehensible – is false; the speaker being available for reference depends on the logic of the speaker being a referent given the semantic content of both the main clause and the participial phrase.  I imagine that “Thinking of the children, the spaceship entered the wormhole” probably wouldn’t pass muster for people who aren’t watching Interstellar at that very moment, but not for reasons of grammatical structure.

Let’s see if my other hypothesized stowaways did any better.

D, the discourse

Screen Shot 2017-06-24 at 23.33.19.png
3.86 and 4.00 for the left bar, 3.54 and 3.71 for the right.  Correlations 0.67 for both comprehensibility and clarity.

The first sentence, “Approaching the end of the chapter, the final topic is information technology”, resembles the sentence from Mike Long’s Second Language Acquisition and Task-based Language Teaching that set off this quest in the first place – a dangling participle meant to move the discourse along, with the discourse as its implied subject.  As you can see, it got higher, but not much higher, scores than my floor sentence.  Some reasons given for this were:

The use of words “end” and “final” clarified what to expect.

Understanding if what the author meant; however, I don’t like it.

Seems to be similar meaning as “as we approach”

It obviously is a temporal element for the reader of the chapter. Not dangling but apposed to the reader’s situation

(Another discursive stowaway – the reader/listener – is also possible, and is the only logical referent in an instruction manual sentence like “Holding Piece B steady, insert Leg A into Socket C”.  This doesn’t change my conclusion that the solution to the dangling participle problem is human readers making the most likely interpretations of potentially ambiguous utterances.)

The big news here is that the second sentence, “Returning to the main point, it is clear that further progress can be expected”, has a comprehensibility score a full 0.29 lower than the true dangling participle that I designated my floor.  That is, this sentence was judged less comprehensible than the floor by half the difference between my floor sentence and my ceiling sentence (0.51). Some reasons that participants gave were:

The passive voice and little indication of the subject returning to the main point make this sentence much too vague – aside from guesses, there is definitely not enough supporting evidence for a safe guess.


Because we discussed the previous form of this response [this respondent pilot tested the questions for me.  I probably shouldn’t have included his responses, but this is just a blog after all]

Obvious circumstantial apposition to the main people behind the situation

This sentence is too uninformative. I guess we can expect further progress in something that we talked about previously. [low scorer.  Note that the problem is with the main clause, not the participial phrase]

0.29 is a sizable drop from the sentence that I had specifically designed to be the lowest scorer, and doesn’t support my hypothesis that participles that refer to the discourse itself are always more comprehensible that true dangling participles.  Given that these two sentences both seem to refer to ongoing threads of discourse that readers of just these sentences in isolation aren’t a part of, it is natural rather that they would deem them less comprehensible and less clear because I’ve stymied their ability to form a cohesive understanding of what is happening and what is being described.  That is, again, a general ability to understand the kinds of things that people talk about is the main factor in determining comprehensibility and clarity in a dangling modifier, not adherence to a particular grammatical rule.

C, the contents of another clause

Screen Shot 2017-06-24 at 23.35.42.png
Left bar: Comprehensibility 3.24, clarity 3.21.  Right bar: Comprehensibility 3.92, clarity 3.83.  Correlations 0.54 for comprehensibility and 0.44 for clarity.

The first sentence’s scores reveal something of great interest to my hypothesis – basically disproving it.  I meant this sentence (“Costing millions of dollars, the workers accidentally struck a water pipe while digging.”) to be a clear reference to the action of the workers, not the workers themselves, as would have been the case if this sentence followed the usual subject-matching rule that supposedly governs participial phrases.  However, many participants saw this reference as ambiguous, possibly referring to the wages of the workers, and cited this as a reason for giving it a relatively low score.

This one was entirely indecipherable. Depending upon the number of workers, they could be what costs millions of dollars, though on reflection I doubt it. This is an example of where a dangling modifier leads to a real lack of clarity. [low scorer]

The logic obviously underlying the sentence… But it would have been natural if the main and subclause had changed places…

Made me think that the workers cost money [low scorer]

just move the first four words to the end of the sentence, and it’s clear

Evidently – and this is a rule that makes sense to me although I hadn’t heard it put into words before – participial phrases are free to dangle (or at least refer to an entire clause rather than its subject) when they come at the end of the sentence.  That is, “Resolving the problem of how to split the inheritance, the deceased’s desk was found to contain a will in its bottom drawer” is felt to be more ambiguous than “The deceased’s desk was found to contain a will in its bottom drawer, resolving the problem of how to split the inheritance” (Note that neither the participle-adjacent desk nor drawer is the subject of the participial phrase).

Anyway, what is of interest for this survey is that it was the logical possibility of workers also costing millions of dollars that seems to have caused this sentence’s scores to fall a full 0.59 below those of my floor sentence, not the presence or absence of a discursive stowaway or the participial phrase’s status as dangling (remember, the floor sentence’s dangles even more flagrantly).  Again dangling-ness and reference to things hidden from surface structure but nonetheless present matter less than human logic.

The second sentence doesn’t suffer as much from ambiguity, although its modifier is technically as dangling as the first sentence.  The participial phrase in “Causing the students to break out in cheers, the due date was delayed for a week” pretty unambiguously refers to the act of delaying the due date rather than the due date itself.  Most participants had little doubt where this was concerned, which probably explains its higher scores.

Inferencing the setting and sequence of events

It still sounds awkward, but the agent of the action is not too hard to understand.

Circumstantial concatenation expressing what happened after the announcement

I can make some assumptions, but they are only assumptions. I can assume that the due date is a date when an assignment is due in class. I can assume that learning that the due date has been postpone causes the students to cheer. But there are other possibilities. Maybe someone is having baby, and these are medical students, and they all know that the extra weeks of pregnancy will be beneficial to the mother and child. Or maybe the due date is a library due date, so the students are happy that they have more time to spend with a reading assignment. [middling scorer]

So again we see that trying to match the semantic information to what the reader understands about the world is more important to understanding than strict adherence of the rule that participial phrases must have the same subjects as the main clause, and also more important than reference to my discursive stowaways.  I feel like I’ve typed sentences similar to that at least 5 times too many in this post.  Let’s see if I can avoid doing so again in the conclusions.

In sum

As factors that affect comprehensibility and clarity of dangling participles, my discursive stowaways are not insignificant, but by focusing on them I was missing the bigger picture.  Because the speaker/writer (and the listener/reader), the discourse itself, and the contents of other clauses are always readily available to be used to fill roles in an utterance in order for that utterance to make sense, they are part of a palette of mental resources for making sense of any given sentence.  However, that is because most listeners have the psychological and linguistic resources to see discursive stowaways as part of the representation of the world he/she draws on the make sense of any utterance. The representation of the world also includes things like how cars work, why people usually quit jobs, and what it means to approach the end of a book, that may be more important to how comprehensible and clear a sentence is than my hypothesized stowaways or strict observance of prescriptive rules of grammar.

Also, here are the averaged comprehensibility/clarity scores for each category.

Screen Shot 2017-06-24 at 23.50.06.png
Speaker: 4.00; Discourse: 3.86; Clause: 3.51; Floor: 3.51; Ceiling: 4.31

In closing, here are some final comments from the participants (who often seem more qualified than I am to comment on dangling participles):

Reading student texts gives me more understanding of poor modifiers.

Interesting thought experiment

Maybe some call them dangling but they are clear circumstantial, temporal mostly but not only, apposition to the situation, to the main subject or to the only person directly concern,ed in the utterance.

Though not a purist, I still believe in insisting on proper use of participial phrases since dangling, or misplaced phrases indicate laziness and poor thinking habits, and a lack of respect for the reader (expected to reconstruct the meaning).

To me the examples seem fairly literal translations of a Romance (Portuguese?) sentence (with gerund/’gerúndio’).

I think some dangling-clause phrases will become (have become to some extent) expressions, almost like prepositional phrases — Like “judging from”, which is quite OK if written as “on the basis of …”; the underlying idea is quite clear and the phrase is that frequent that it will become like a prep. phrase…

A ‘dangling participle’ may not earn the beauty conteSt, but the conteXt usually makes the meaning crystal-clear.

I can’t believe I just spent 10 minutes answering a survey about dangling participles. But oh well, I did.

Addendum: Modifiers are usually held to logical and not grammatical standards, and why we should care about prescriptive grammar

Contrary to what your inner prescriptivist may have been led to believe, participial phrases are not alone in having ambiguous referents.  There is no reason to hold participial phrases to this standard when not doing so with other adverbial and adjectival modifiers.  When you deliberately shut off the part of your brain that tries to look for the most natural interpretation of modifiers, you can come to some bizarre interpretations with nothing to argue against them besides being “unnatural”.

From Newsweek (the autoplay video, not the news article):

A new controversial bill regarding birth control has been advanced in Arizona, which would allow employers to fire women who take the pill to prevent pregnancy.

How do we know it’s the bill, not Arizona, that would allow employers to fire women who take the pill?  Would we make the same assumption based on structure alone if the phrase were “A new bill has been advanced in Arizona, which has had its governor threaten to veto all new legislation”?

From CNN:

While it’s likely to change before lawmakers vote on it — possibly late next week — it’s already clear who will benefit and who will lose under the Senate plan.

How do we know that “under the Senate plan” applies to both lose and benefit?  Would we make the same assumption on structure alone if the phrase were “who will spend and who will save under the mattress”?

My point in raising these unintuitive examples is just that we already rely on intuition and logic rather than just structure to put pieces of a sentence together in a way that makes sense.  In a gratuitous appeal to authority, I will also say that the primacy of logic over form is the same conclusion Steven Pinker came to, as I verified by revisiting my copy of The Sense of Style last night.


As a teacher, I have to be aware that next semester, someone else will be grading my current students’ papers.  This person is, even in ESL, likely to be a prescriptivist with a keen radar for “errors”.  The fact that these particular grammatical structures are so salient to prescriptivists and those who listen to them also adds an element of self-fulfilling prophecy to acceptability of participial phrases: Because prescriptivists consider them wrong, and because so many (over-) educated consumers of language are prescriptivists, using them in a “dangling” fashion becomes a sign of bad style much in the same way as split infinitives or “ain’t”.

I mostly appealed to human logic and intuition as an explanation for why dangling participles are usually passable.  Still, part of the meaning of language in human society is basic agreement on ways of using it that display class and intelligence.  I can’t claim that basic facts about human psychology saving dangling participles from “error” status allows us to pass them on as healthy language use when another fact of human psychology requires them to be shibboleths to separate elevated language from casual speech.  There is nothing inherently wrong with dangling participles, but the fact that people consider them wrong, in a way, makes them wrong and worth avoiding, at least when writing for a formal audience (which, ironically, ought to know better).  After all, not everyone will read this blog post and become enlightened about them.  Unless you tell them about it.

2 thoughts on “Discursive stowaways and human logic (dangling participles part 2)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s