Aside from reviewing published work from Project Implicit researchers or highlighting relevant articles in implicit cognition, we’ll also use this space to discuss some of the smaller studies being run on Project Implicit.
Researchers using the IAT often worry about what category labels to use. A common concern is that, when choosing between similar category labels, even subtle differences between labels can alter IAT performance. One example comes from IATs looking to assess gender-career associations, meaning how strongly the concepts of male and female are associated with the concepts of career and family. In this IAT, participants categorize typically male or female names as well as words related to family or career. Results typically show stronger associations between male with career and female with family.
However, it is possible that there is an “alliteration effect” with participants completing an IAT that uses both “Family” and “Female” as labels. That is, during those parts of the IAT when “Family” words and “Female” words are categorized using the same key, participants may find that easier because both labels start with the letter “F”, compared to if the IAT used similar labels that did not share the same starting letter, such as “Family” and “Woman”. By potentially making responses easier (and therefore faster) when “Family” and “Female” words are paired together, this alliteration effect could contribute to (or even create) IAT scores suggesting a stronger association between the concepts of female and family (as well as between male and career).
To address this possibility, Project Implicit researchers and University of Florida professors Kate Ratliff and Colin Smith ran a study where over 500 participants were randomly assigned to complete one of two versions of an IAT. Both IATs used Career and Family as attributes, but one used Male/Female as the category labels and the other used Man/Woman. In both versions of the IAT, participants viewed the same typically male or female names for each category-- all that differed between the two IATs was whether the category labels in the IAT were male vs. female or men vs. women.
The data showed no evidence of an alliteration effect, as there were no differences in overall performance on the two IATs. Although both IATs indicated a strong association between male/man and career, and between female/woman and family, the overall IAT scores were very comparable between participants completing the man/woman version (Average IAT D score = 0.35, Standard Deviation = 0.37, Sample Size = 246) and those completing the male/female version (Average IAT D score = 0.33, Standard Deviation = 0.36, Sample Size = 255). These data suggest that, for measuring gender-career associations at least, using Male/Female or Man/Woman as category names does not meaningfully impact performance on the IAT. As such, researchers can choose which of these labels they feel more comfortable with, perhaps based on which words fit more appropriately into the type of self-report items that may be included in the study design.
These results harken back to earlier work on the IAT investigating the impact of category labels versus the stimuli used to represent those categories (interested readers can look at De Houwer, 2001 or Bluemke & Friese, 2006 for more on this topic). Of note, the current data clearly illustrates the importance of testing our assumptions. From our perspective, the alliteration effect makes perfect sense. Further, it is a mechanism for IAT effects that is commonly-raised by participants at Project Implicit and by attendees of our talks. However, the current results do not suggest any evidence of an “alliteration effect” when completing the IAT.