ABSTRACT
A previously underappreciated factor, the specific letters used to label the groups, was found to influence the magnitude of the well-established illusory correlation (IC) effect . The typical IC effect of an association between the minority group and the rarer (negative) behavior was strong when the minority group was labeled with an infrequent letter (e.g. X, Z) and the majority group was labeled with a frequent letter (e.g. S, T), but the effect was eliminated (or reduced) with the reverse pairing of the majority group with an infrequent letter. The letter label effect was also found with the A and B labels most commonly used in this paradigm. The results were consistent with an explanation based on the affect associated with the letters due to the mere exposure effect. The findings reveal a previously unexplored way that the names for groups may influence stereotype formation, contribute to the debate on the mechanism underlying IC, and illustrate how arbitrarily chosen labels for groups and other objects in social research may bias processing in unexpected ways.
Acknowledgments
The author declares that there are no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. The author received no financial support for the research, authorship, and/or publication of this article. This article adheres to the ethical guidelines specified in the APA Code of Conduct. All studies reported received ethics approval by the Hofstra University IRB. The materials for the studies are available at: doi.org/10.17605/OSF.IO/7PF3C The data that support the findings of the studies are available at: doi.org/10.17605/OSF.IO/3NVQ4
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability
The data will be made available upon request by contacting the corresponding author.
Open scholarship
This article has earned the Center for Open Science badges for Open Data and Open Materials through Open Practices Disclosure. The data and materials are openly accessible at doi.org/10.17605/OSF.IO/7PF3C; doi.org/10.17605/OSF.IO/3NVQ4
Notes
1. The male names included with each sentence were all very common American male names (e.g., John, Joe, Bob, Steve, Mike). Each name was linked to the same behavior in each condition and thus any associations with a given name would be consistent across conditions.
2. In many past works on IC, researchers have transformed the phi coefficients into Fisher z scores. However, following the recommendation from Haslam and McGarty (Citation1994), I ran the analyses on the raw phi coefficients. The following is the formula for the calculation of the phi coefficient: phi = (AD – BC)/ √(A+B)(C+D)(A+C)(B+D) where A is the value for the majority group positive behaviors, B is the value for the minority group positive behaviors, C is the value for the majority group negative behaviors, and D is the value for the minority group negative behaviors.
3. The original Mullen and Johnson (Citation1990) meta-analysis did not include evaluation measures so I did not include the effect size comparisons for those.
4. The list of articles uncovered in this review can be obtained from the author.
5. The phi coefficients for assignment and estimation correlated strongly in Study 1 (r = .523, p < .001). Consequently, to more efficiently present the study in the online format, the assignment measure was dropped from the procedure in Study 2.
6. The sample size for the estimation task analysis (N = 81) was smaller than the sample size for the evaluation task analysis (N = 92). With the estimation task, participants were asked to place their response in an open ended text box. Some participants did not answer with a discrete numerical value and thus the phi coefficient could not be calculated.
7. The words included in the described comparison were those used by Song and Schwarz (Citation2009) in their Studies 1 and 2. They used different stimuli in a third study but in that study the easy-to-pronounce words were shorter in length than the hard-to-pronounce words which adds a confounding factor that makes a straightforward comparison based on letter frequency problematic.
8. The analysis here focused on Bahník and Vranka’s (Citation2017) study 5 because their other studies used Czech or German samples or included features that made the desired comparisons less straightforward.
9. I thank Stepan Bahnik for providing the harmfulness ratings and suggestion to examine the results of his Study 5.
Additional information
Funding
Notes on contributors
Craig Johnson
Craig Johnson, is an Associate Professor in the Department of Psychology at Hofstra University. His research interests include stereotyping, human-animal relations, and other topics in the area of social cognition.