Exploring linguistic correlates of social anxiety in romantic stories

Abstract

The current study used computerized linguistic analysis of stories about either going on a date or taking a walk down a street to examine linguistic correlates of social anxiety in a sample of undergraduate students. In general, linguistic analysis revealed associations of social anxiety with several linguistic variables, including negative emotion, affect, and anxiety words. Participants higher in social anxiety wrote fewer affect words. The relationship between social anxiety and anxiety words depended on gender, whereas the relationship between social anxiety and negative emotion words depended on both gender and the nature of primes (supraliminal vs. subliminal) received. Overall, our findings highlight the potential utility and benefits of using linguistic analysis as another source of information about how individuals higher in social anxiety process romantic stimuli.

Keywords:

• Linguistic analysis
• LIWC
• social anxiety
• romantic
• subliminal priming

Notes

1. Though the subliminal primes used in the current study were not specifically sexual, as defined in Gillath, Mikulincer, Birnbaum, and Shaver (2007, 2008) studies, they were based on physical attractiveness, a romantic construct related to sexuality. We chose attractive versus less attractive faces as our subliminal prime for the following reasons: First, existing research has linked social anxiety to differences in mate selection based on physical attractiveness in photographs (Wenzel & Emerson, 2009), such that individuals higher in social anxiety reported being less likely to initiate relationship and sexual behaviors with physically attractive individuals (and more likely to initiate these behaviors with physically unattractive people). Second, the effects of attractive versus less attractive faces as subliminal primes have been documented in recent studies (McDonald, Slater, & Longmore, 2008). A total of 20 subliminal images were presented to each participant.

2. At the time the study was conducted, our IRB believed such a question was invasive and would not allow the question to be asked of the participants participating in studies for credit. Since the time of this study, they have changed their protocol. One way in which we attempted to conduct the experiment to be consistent with the participant’s sexual orientation was to offer participants the option of notifying the experimenter if they preferred to go on a date with an individual of the same sex; however, we acknowledge that this approach is flawed. For example, participants may have indeed preferred to go on a date with an individual of the same sex but felt uncomfortable sharing his or her preference with the experimenter. The presence of some small number of exclusively homosexual participants seems likely, and would add a small amount of error variance to the study that we cannot quantify without further information.

3. Due to clerical error, the first 59 participants were assigned to condition pseudo-randomly (i.e. participants were assigned to condition in order—first participant to Condition 1, second participant to Condition 2, etc. and then repeated), and the remaining participants were assigned randomly by block. No moderating effects for pseudo-random vs. random were found (ps > .095), suggesting no differences created by this error.

4. Development of the MacBrain Face Stimulus Set was overseen by Nim Tottenham and supported by the John D. and Catherine T. MacArthur Foundation Research Network on Early Experience and Brain Development. Please contact Nim Tottenham at [email protected] for more information concerning the stimulus set.

5. Supplementary analyses suggested that the social anxiety composite results might be more dependent on the SIAS than the SPS.

6. For several reasons, we determined a correction for multiple tests was not called for in the current analyses. However, should we have used one, a Bonferroni correction, our new p-value to determine significance would be p < .00714. According to this new cut-off, social anxiety would no longer significantly predict affect words, and there would be no significant effects for anxiety or negative emotion words. Thus, the only LIWC categories with significant results would be affect and anger, and only anger words would be predicted by social anxiety (in the context of an interaction).