This blog post is directed to those who are interested in coordinating and/or participating in a many-analyst project on a large, multi-site study investigating the effects of gendered job occupations and gender roles on the naming of male or female exemplars. The original authors will make ⅔ of the data available to analysis teams, who will then pre-register their hypotheses before the remaining ⅓ will be released (in case power is insufficient for this split then the analysis team can request the full dataset). My job as the editor (Hans Rocha IJzerman) is to connect potentially interested parties to each other, who can then coordinate with each other to submit one single many-analyst manuscript to the International Review of Social Psychology. The manuscript will pass a normal review by an independent editor. Below you will find a results-blind summary of the project, which is nearing its completion.
The many-analyst project can focus on a re-examination of the hypotheses set forth by Brohmer et al. using different analysis approaches and/or analyses of secondary variables not analyzed in the main project to form a more advanced theoretical model.
Summary of the project
In this project (see https://osf.io/fdrn6/), Brohmer and colleagues aimed to investigate whether generic-masculine forms of job occupations and societal roles evoke more masculine exemplars. In languages such as German, French, or Spanish, plural forms of job occupations and societal roles are often in a generic-masculine form. That is, instead of employing gender-neutral forms, people often use the male plural form to address people in general. Several studies have found that using the generic-masculine form (instead of more neutral forms) reduces the cognitive availability of women for specific societal roles and job occupations. Yet, despite the societal importance of such linguistic conventions, rarely any replications exist to demonstrate the robustness of the effect. Brohmer and colleagues picked one prominent German study by Stahlberg, Sczesny and Braun (2001, Experiment 2), which revealed that people indeed came up with more male exemplars when they were asked to “name three politicians”, “athletes”, “tv hosts”, and “singers” in the generic-masculine form (e.g., male politician or “Politiker”), compared to two alternative gender-neutral forms (naming both form: female and male politician or “Politikerinnen und Politiker”; internal-I form: “PolitikerInnen”). As an extension, Brohmer and colleagues included two more forms – a neutralized control form (which does not imply a gender; e.g., “Name three persons from politics”) and a popular and the more inclusive gender-star form (“Politiker*innen”) along with several other control variables. Moreover, they added two more job categories (“writers” and “actors”) to increase precision of the outcome estimates. Data collection for this replication and extension study across eleven labs and N ≥ 2100 participants is nearly finished.
For the main analyses of the project, the authors perform generalized linear multilevel models analyzing how many women were mentioned across the six job categories (with a maximum of three persons, categories being nested in participants). For the confirmatory analysis, participants’ own gender and perceived base rate per category were included as predictors (i.e., they indicated how many men or women are in these jobs).
In addition, Brohmer and colleagues collect data on the following list of variables that are not analyzed as part of the main projects’ confirmatory analysis. A complete list of variables is provided below (* marks variables for the confirmatory analysis).
- participants’ device (computer/laptop, tablet, smartphone)
news consumption
- smartphone usage
- preferred information source
- gender-fair language scale (9 items)
- participants’ women count for the DV
- if applicable: perceived reasons why for not mentioning women or men (4 answer options and open text field)
- age
- political orientation
- social dominance orientation (short scale, 3 items)
- social equality preference scale (short scale, 3 items)
- location of residence
- highest education
- nationality
- distraction during answering questions
- three funnel debriefing questions
The researchers analyzing the dataset are also welcome to add additional variables to the dataset via data-scraping (see e.g., here for the first of a series of posts on data-scraping).
If you are interested in signing up for the project, please go to this form. The editor will then connect the people interested in this many-analyst project. We anticipate that the data will be available in October.
Image is reposted from https://www.bayesianspectacles.org/call-for-crowdsourced-analysis-many-analysts-religion-project-marp/