Content-coding motive measures can be approcimated with automated word counts
Many researchers are interested in working with the Picture Story Exercise (PSE), the most extensively validated measure of implicit motives, but are deterred by the amount of work necessary in training coders and having them content-code PSE stories. But now there may be a method that may allow researchers to approximate content-coded motive scores through the use of the inexpensive text-analysis software Linguistic Inquiry and Word Count (LIWC) by Pennebaker, Francis, and Booth (2001). In a paper recently published in Frontiers in Psychology, HuMAN-Lab director Oliver C. Schultheiss explores whether LIWC analyses of PSE stories yield motive scores that converge with scores resulting from content-coding of PSE stories.
.jpg)
Example of a content-coded PSE story (from Schultheiss, 2013, Frontiers in Psychology)
“The process of content coding is not something that can be done by a computer at this point. It requires an understanding of the entire sentence and sometimes also the context the sentence is embedded in. Thus, content-coding is based on sophisticated understanding of language. This is something that computers do not yet excel at. So the question was: how close can we get through a dumb and automated process of counting words, which is something that computer do excel at?” explains Schultheiss. To answer this question, Schultheiss used the LIWC software to analyze PSE stories that he had collected in the US and in Germany and compared LIWC-derived counts of specific words with content-coded scores for the same PSE stories. Across studies and samples, content-coded motive scores correlated with specific word categories in a consistent and conceptually meaningful way. For instance, n Power scores were associated with the LIWC word category “anger”, n Achievement with the category “achievement”, and n Affiliation with the category “friends”. Each motive could consistently be linked to several such categories, and once LIWC categories were combined for each motive, the LIWC-based scores correlated with the original content-coding scores between .35 and .52. Comments Schultheiss: “These numbers suggest that LIWC-based motive scores really only approximate content-coded scores. However, I found that even with this rough approximation, you can detect some key validity correlates of implicit motives that you also consistently get with content-coded scores, such as a gender gap in n Affiliation or an association between implicit motives and emotional well-being that depends on the advancement of personal goals.” While some may shrug off the word-count approach implemented with LIWC as a poor researcher’s motive measure, Schultheiss is not so sure that this approach is necessarily inferior to semantic coding: “In all those PSE studies over the years, people have tried to find tell-tale signs of motivational arousal in complex patterns of semantic meaning. But what if there’s also something going on below this level? What if motivational arousal also manifests itself in the increased frequency of some types of words and the decreased frequency of others? No one so far has really given that fair scrutiny.”
Exploiting the full potential of thematic apperception through profile analysis
High progesterone is associated with less coherent brains (August 2012)
What color naming speed reveals about the wisdom of
one's goal choices (December 2010)
Are you high on testosterone and is that a good thing? Listen to Podcast of interview with Dr. Oliver Schultheiss on UM NewsService
New edited book on implicit motives available (October 2009)
What the word "not" may reveal about your ability to handle stress (October 2008)
Estrogen fuels female power (February 2008)
High-testosterone people reinforced by others’ anger, new study finds (February 2007)
Study finds US students more motivated to achieve, less power-hungry than German students (August 2006)
Are all people stressed out by a defeat or does it hurt some more than others? (April 2006)
|
