How I (almost) won an argument using corpus linguistics

I’m taking a class called “Morphology and Syntax.” It’s currently the third week of the class and it has been interesting. In the first week, I got into an interesting debate with another student.

The argument was about the usefulness of an intrusion test to determine phrasal boundaries (1). He argued that a test to see where an adverb can be placed within the sentence to determine the boundaries of a phrase is not completely reliable because, while incorrect according to a prescriptive grammar, some discourse strategies allow intrusion in a phrasal verb (2). In principle, I agree with him. I feel strongly that how people actually use language *should supersede any prescriptive grammar. The question then becomes How do people actually use the language? and this is where corpus linguistics comes in.

The examples my classmate used are below:

(A) The cat will eat its, I suspect, lamb chops …
(B) John rang, almost certainly, up his accountant.
(C) John rang, almost certainly his accountant up.

I left (A) alone in the debate that followed.

Then I turned to COCA (Corpus of Contemporary American English) and discovered after an hour of messing around that I had absolutely no idea how to do what I wanted to do. So I did what any rational tweep would do and turned to our friendly neighborhood #TESOLgeek and #corpusexpert, @muranava.

Screen shot 2013-03-14 at 3.51.00 PM

This led to some more research on commands to use in a corpus, but also led to an answer to my question:

Screen shot 2013-03-14 at 4.01.03 PM
Most commonly, a direct object is inserted between ‘rang’ and ‘up’. 9 and 10 look strange. Judging by the context (“her laughter rang shrill up and down the river”) the phrasal verb is not involved for 10. 9 is a mystery: “I rang turn up on the field phone.”

Next I searched for adverbs:

Screen shot 2013-03-14 at 4.17.41 PM

This image shows that adverbs appear before ‘rang up’. This was not surprising. Next I checked to see if people ever use adverbs within the phrasal verb:

Screen shot 2013-03-14 at 4.21.24 PM

Unsurprisingly, I came up empty.

Of course there are problems with my informal research: COCA is limited to American English for one thing, while the phrasal verb ‘rang up’ is perhaps not very common in American English (compared to ‘called up’ with 549 entries, ‘rang up’ has only 46). Incidentally, I did the same research for ‘called up’ and discovered that ‘called back up’ has a couple entries.

A second consideration is that COCA’s results, while across a variety of genres, are nevertheless limited. My classmate suggested that a search of a non-dialect-specific corpus might give better results. Ah, if only the BNC were free! The random 50 entries for ‘rang _ up’ in BNC all insert direct objects and I cannot do a more advanced search from the “simple search” page.

In the end, my classmate conceded that his specific example might not have been the best one, but his point still held that intrusion tests are not sufficient to determine phrasal boundaries – and I agreed.

The argument ended when the professor chimed in to say that wherever this sort of interruption might occur within a phrasal verb, one is also likely to find disfluency features like throat-clearing or false starts. His post made it clear that he personally finds the intrusion plausible.

Now the more I think about it, the more I realize that I was being a bit of a prescriptivist myself. I was trying to use a descriptive resource to prove my prescriptive point. It didn’t work quite the way I hoped it would, but I learned a few things!

From this interesting experience I learned a lot about the power and limitations of corpus linguistics. I see the value in having access to how people actually use English as an alternative to the prescriptive approach (how we *should use it). I also see the limitations since corpora are not exactly a cross-section of English use yet.

I also started to learn how to search a corpus (two, in fact. Thanks to Mura for getting me started!). These are tools I want to use more often.


Have you ever used corpora? How do you use them?
Feedback of all kinds is appreciated. 



Notes on terminology and/ or grammar:
(1) A bit more about phrases and intrusion tests:
Phrases include, for example, noun phrases (“the big red dog”), prepositional phrases (“through the garden”), etc. In order to determine where the boundaries of a phrase are, there are tests. The intrusion test says that an adverb (for instance) can only be inserted at the boundaries of a phrase.
So in the sentence, “The big red dog ran through the garden,” we can say
“Quickly the big red dog ran through the garden”
and “The big red dog quickly ran through the garden”
and “The big red dog ran quickly through the garden”
and “The big red dog ran through the garden quickly.”
But we can’t say “*The quickly big red dog ran through the garden”
or “*The big red dog ran through quickly the garden.”
This shows that “the big red dog” and “through the garden” are phrases.

(2) Verbs often stand alone, except phrasal verbs. Phrasal verbs, like “rang up,” “called up,” “tried out” are verbs that are inseparable from their particles.
The only thing that can be inserted between the verb and particle is a noun phrase containing the direct object, so “rang the doctor up” or “called the speedy taxi driver up” or “tried the new piano out” are all okay.
On the other hand, “*rang quickly the doctor up” and “*called spontaneously the speedy taxi driver up” and “*tried yesterday the new piano out” – not so much. Thus the intrusion test once again shows that these verbs are phrases.


Post a comment or leave a trackback: Trackback URL.


  • eflnotes  On March 15, 2013 at 10:05 am

    fantastic post, don’t know much about linguistic theory so nice to learn about intrusion test, look fwd to more of the same 🙂

    the rang turn up seems to be a typo should be rang him up –

    re BNC you can search BNC on COCA after search click on the side by side drop down.

    recently found this web-based corpus (free registration)


  • breathyvowel  On March 19, 2013 at 8:25 am


    Like all my best linguistic inspirations this comes from listening to Dr. Dre.

    It struck me that it’s quite possible to shove an expletive in the middle of a phrasal verb quite naturally, as in “shut the **** up”. I’m assuming the expletive is an intensifying adverb here, and not a direct object, Anyway, if this is the case it would seem to cast a little doubt on the intrusion test.

    I also think that the intrusion test is very much a product of written grammar, rather than spoken, and that people probably do split phrasal verbs occasionally in speech with no problems.

    • gfkpth  On February 10, 2014 at 10:57 am

      I’m not sure if this test generalises to cases like “take the f*ck out the bins” or “take the f*ck the bins out” (not a native speaker)?
      Also note that these sorts of terms can even enter into what would commonly be described as words, e.g. Ala-f*ckin’-bama.
      But of course it may well be that particle verbs aren’t phrasal in the same sense that, say, a noun phrase is.

      • breathyvowel  On February 16, 2014 at 1:55 pm

        The two examples above feel a bit unnatural to me, but playing with word order “take the bins the fuck out” sounds ok, probably related to where you can put an adverb in a construction like this. But right on, expletives are about the most versatile items in the language.


  • […] final one (recommended by Mura Nava) was Anne Hendler’s post on “How I (almost) won an argument using corpus linguistics“. I really liked this article because it showed some practical uses for corpus and a bit of […]

Leave a Reply

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

You are commenting using your 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 )

Google+ photo

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

Connecting to %s

%d bloggers like this: