"Believe (that) PNP"

« previous post | next post »

Following up on yesterday's "'Guess that'", this morning I looked at whether "speakers use the unstressed optional complementiser that to maximise rhythmic alternation of weak and strong syllables" in the case of complements following the verb believe. I again used data from Shuang Li's INTERVIEW: NPR Media Dialog Transcripts dataset.

In order to match more closely the stimuli used in Lee & Gibbons (2007), I limited the pattern to cases of "I believe (that) PROPER_NOUN". And I'll note in passing that just like "I guess", that-less "I believe" is often used as a sort of adverbial, e.g.

Other contenders in the field – I believe Biden and Bennet, for example – they say that goes too far.
The only defendant in the case that has a significant criminal conviction is I believe Frankie Brewster that was convicted of a voluntary manslaughter back in the mid-90s and spent some time in West Virginia Penitentiary as a result of that.
You know, Ann Landers grew up in I believe Sioux City, Iowa, and she and her sister, Dear Abby, you know, they gained a reputation as being very no-nonsense, very practical and having a lot of common sense.

And therefore there are also that-less cases where it's not clear whether "I believe" is taking a complement or not, e.g.

We have testimony saying that the helicopter, the Coast Guard flew over, saw the breach in the morning, confirmed it in the evening, took the pictures which I believe Senator Dayton showed you.

That issue aside, the prediction is that the presence or absence of that should depend on (or at least be influenced by) whether the first syllable of the following word is stressed or not:

I believe that Trump is …
I believe Obama is …

It's not clear what the prediction is for words like Illinois, whose first syllable is a secondary stress — my choice was to put them in the unstressed class. I did the same thing with cases like New York, where there's a word-internal space. The number of those cases was not very great.

As before, I coded a sample of 100 cases. The results:

stressed not_stressed
that 34 8
no_that 41 17

This trends in the predicted direction, but not strongly enough to reject the null hypothesis:

> Bsq$observed
           stress nostress
  that       34        8
  nothat     41       17
> Bsq$expected
          stress nostress
  that     31.5     10.5
  nothat   43.5     14.5
> Bsq$p.value
[1] 0.3493677



  1. Haamu said,

    February 21, 2021 @ 1:40 pm

    To limit any confusion introduced by the adverbial, wouldn’t it be fairly straightforward to filter out instances beginning with “I”?

  2. Jerry Friedman said,

    February 22, 2021 @ 11:06 am

    Is it interesting that there was no convenient corpus with the accented syllables tagged? The information is available in dictionaries. I imagine using it to tag a corpus would face various difficulties, such as words that are pronounced differently depending on context or dialect or whatever, but tagging a corpus for parts of speech also faced difficulties.

    I tried a Very Short Breakfast Experiment at the Spoken section of COCA with "insist" and "insists" versus "insisting" and "insisted" to see whether the stress level of the verb's last syllable made a difference. However, I didn't have time to look at every word, so I just wrote down the nouns that had more than one hit. The results probably don't mean anything, but there might be a tiny little tendency for nouns with unstressed first syllables not to be preceded by "that".

    (One reason the results are so worthless is that COCA has duplicates. There are two hits for "insisted that Mrs." because the same sentence from NPR's Morning Edition, about Mrs. Kennedy being brought to a cabin, is in there twice. But I did it, so here it is.)

    insist|insists NOUN: 78 things, Sotomayors, security, Democrats, police, talks

    insist|insists that NOUN: 74 people, president, Congress, Mr.

    insisting|insisted NOUN: 52 Mr., Republicans, Democrats, president

    insisting|insisted that NOUN: 54 people, Mrs., president, Congress, things, race

    [(myl) The main problems for the cases that I looked at is not determining the stress pattern, which indeed would be easy enough to determine automatically with reasonable accuracy, but rather whether the following material is a (tensed) complement clause or something else.]

  3. Jerry Friedman said,

    February 25, 2021 @ 2:51 pm

    Thanks, that makes sense.

RSS feed for comments on this post