Filled Pause
Research Center

Filled Pause
Research Center

Filled Pause
Research Center

Investigating 'um' and 'uh' and other hesitation phenomena

Investigating 'um' and 'uh' and other hesitation phenomena

Investigating 'um' and 'uh' and other hesitation phenomena

October 3rd, 2015

Wait, what? A written text can be disfluent, too?

I recently read an article entitled, "Disfluency disrupts the confirmation bias" (Hernandez, I., & Preston J. L. (2013). Disfluency disrupts the confirmation bias. Journal of Experimental Social Psychology. 49(1), p. 178-182.) which is a rather interesting look at how confirmation bias—the tendency to give greater credibility to evidence that comports with one's already held beliefs—can be dampened by presenting information to people in a disfluent manner. I was interested in the article for obvious reasons (yay, disfluency!) and quickly added it to my Kindle for reading during my next walk to and from campus. However, I got about one or two pages in before I realized something: The article is not about speech disfluency, but rather text disfluency.

Wait, what? I returned to the first page and started re-reading and discovered a couple of the phrases I had glossed over quickly as, I thought, metaphorical (e.g., "visual clarity", "disfluent font") were intended literally and that the researchers were going to describe an experiment in reading a disfluent text rather than listening to disfluent speech. As someone who has studied speech fluency for nearly two decades, I became rather indignant at the blatant "borrowing" of a term that is central to the study of speech. I was reminded of a physics professor in my undergrad days who complained about sociologists absconding the term "Heisenberg principle" to refer to the "observer's paradox" in their sociological studies: "Show me an electron and then I'll show you what Heisenberg uncertainty really is!" (my paraphrase)

Anyway, I put my bias aside and read the rest of the research article, and it's very interesting. They observe that participants' pre-existing political confirmation biases as well as experimentally-assigned confirmation biases can be dampened by presenting written texts in a "disfluent" manner (i.e., light-gray bold italicized Haettenschweiler font as in Figure B) versus a fluent manner (i.e., Times New Roman as in Figure A).

Heisenberg principles printed in a fluent presentation

Figure A

Heisenberg principles printed in a disfluent presentation, ver. 1

Figure B

Hernandez and Preston argue that the disfluency causes people to attend much more carefully to the content and hence logical reasoning of the text than in the fluent condition where readers may rapidly skim for a basic comprehension. It's very interesting work and if the hypothesis really is true, then it certainly could suggest a useful techniques to prevent confirmation bias in, say, courtroom proceedings, classroom lectures, and so on. Personally, I'm not totally convinced by their results (and the authors don't overstate their conclusions, either) and would like to see confirming evidence in comparable experimental manipulations. Nevertheless, it was very interesting.

After reading the article, I thought more about their use of the term fluency in this context and while I believe their use of it muddles the meaning somewhat, I think a partial analogy can be made. Fluency, and thus, disfluency, has long been thought of and studied as a property of speech production rather than perception (though more on the perception side below). If we say John delivered a fluent speech in Spanish, we understand that John's production of the speech was fluent according to some objectively-held standard. Since writing is also a productive process, it could make sense to talk about writing fluency. And in fact, there is already quite a large literature studying this (Google Scholar: "writing fluency" where it means something like the timely production of a written text by an individual. In this respect, writing disfluency would seem to refer to the long pauses that occur while writing (already an area of study, in fact, or even the self-repairs that may occur during the writing process.

But Hernandez and Preston are not looking at the writing productive process, but rather the perception of writing: that is, reading. In speech fluency research, Norman Segalowitz (Segalowitz, N. (2010). Cognitive Bases of Second Language Fluency. Routledge: London. has distinguished between the fluency of speech production first at the cognitive level (cognitive fluency) as well as at the articulatory level (utterance fluency) on the one hand and the perception of the speaker's fluency by the hearer on the other hand. This, he refers to as perceptual fluency. It seems to me that it is perceptual fluency which is the closer analog to what Hernandez and Preston are studying: How the presentational issues of the text influence the reader's cognitive orientation toward the content.

I'm still not totally drawn in by this analogy, however, because in their study, they have put the entire text into a uniformly 'disfluent' form. But speech disfluency and even writing disfluency are temporary phenomena: That is, we can usually identify a start and end point to any particularly disfluent interval. We do not usually refer to a speaker as uniformly and continuously disfluent from the start to finish of their speech. Thus, it seems that what Hernandez and Preston have done is analogous to presenting a spoken text to a listener in white noise or with multi-speaker babble, or even reverb. I would therefore imagine a disfluent text to be one that has, say, repeated or missing words, typographical errors, grammatical errors, and so on, of the type that one might make when typing out a text rapidly, as in Figure C. But I'm not quite sure if this sort of text would yield the same results as Figure B does.

Heisenberg principles printed in a disfluent presentation, ver. 2

Figure C

In short, I guess I still have to say that I'm not quite resolved to their characterization of texts as fluent or disfluent. While researchers are free to use terminology as they see fit as long as they define terms clearly and use them in an internally-consistent manner within their writings, I think their use of "disfluency" to describe a text that presents uniformly challenging perceptual difficulty is ... well ... disfluent. Further, this disfluency has not dampened my terminological bias, contra their hypothesis. Perhaps age is a mediating factor? (Maybe not...)