Almost 2 Weeks Ago I began talking about some research I've done with Seth Finn and Sungkyoung Lee. In that post, I told you that I would return to the topic the next day. Like I say, it's been 2 weeks!
Well, here we go...
My research focuses on how structural features of audio messages can cause listeners to have automatic attention responses to the message. So, for example, some of my earlier work shows that when one speaker's voice is replaced by another's, listeners automatically allocate more attention to that change in the auditory stream. This is a quick enhancement of processing, you understand...only lasting a few seconds...in which the contentof the message would need to be interesting in order for the listener to continuepaying controlled attention to it.
A couple of years ago, one of my graduate students, Sungkyoung, became interested in investigating whether merely including an emotional word in an audio presentation will cause listeners to automatically allocate more attention to the presentation for a few seconds.
As part of doing so, she collected physiology from research subjects while they listened to radio ads containing nothing but spoken words (i.e., no music, no sound effects, etc.).
Then, along came Seth Finn who has long been interested in how unpredictability of words used in written texts lead to more favorable responses in readers, like self-reported enjoyment. He has used a data-collection technique called the cloze procedure--which quantifies the predictability of each word in a text. However, while most cloze analyses quantify the mean predictability of a text OVERALL, Seth has analyzed content words and structure words separately. As an example, consider the sentence: I want to eat cherry pie and vanilla ice cream for my birthday. Content words would include: I, want, eat, cherry, pie, vanilla, ice, cream, my & birthday. "To", "And" & "For" are structure words.
Seth has tested and confirmed hypotheses that effective writing contains highly predictable structure words and highly unpredictable content words. In other words, when a content word contains more INFORMATION VALUE, it leads to more enjoyment in readers. He has also shown that the relative information value of content words was related to self-reported arousal:
- Donohew, L., Finn, S., & Christ, W. (1988). The nature of news revisited: The roles of affect, schemas and cognition. In H. E. Sypher, L. Donohew, & E. T. Higgins (Eds.), Communication, social cognition, and affect (pp. 195-218).
: Sage. Newbury Park, CA
So, as I mentioned earlier, Seth and I got together several years ago and began thinking that if high-information content words are associated with higher enjoyment...and this enjoyment is due (in part) to readers being more aroused by the unpredictability...could the same higher arousal be exhibited in listeners when exposed to audio information that varies in it's predictability?
To test this question, here's what we did:
We took skin conductance data from four of the radio messages that Sungkyoung had played for her subjects. As is usually done, we measured skin conductance on the palms of the hands, where the sweat glands respond not to how hot the body is, but rather how aroused the person is. The four messages we used were the ones which showed the most variation in skin conductance across the entire 60 second ad--because remember we were looking for variance which we could correlate with the amount of information value. We then calculated how the skin conductance changed at each second (t) from each second just before it (t - 1).
Then, Seth created cloze procedure instruments for each of the adds. Here's a partial example:
Each script had eight different cloze scripts: where every 8th word was replaced by a blank, the intial replaced word was rotated across the 8 versions, and 18 subjects completed each version.
Then, an information-value score was created for every content word and every function word. We then calculated how the listeners' arousal (skin conductance change scores) correlated with the information-value scores.
As we suspected, there was no correlation at all between function words (i.e., to, and, and for) and information value. However, when we correlated the arousal with the information of the content words (eat, vanilla, and birthday) we found a positive correlation. You can see it for one of the messages below:
Now, the other three messages showed similar positive correlations. As the information-value of each individual word increased, so did the physiological arousal reaction. This--once again--shows that words are important and shouldn't be chosen lightly in audio production.
More information about this research study will be presented at the Society for Psychophysiological Research conference in October.