**Original citation.** van Dijk, E., van Kleef, G.A., Steinel, W., & van Beest, I. (2008). A social functional approach to emotions in bargaining: When communicating anger pays and when it backfires. *Journal of Personality and Social Psychology, 94*(4), 600-614. **Target of replication.** In this experiment, the original authors manipulated both the consequences of rejection and the communicated emotions of the recipient to “see whether, if the consequences of rejection were low, participants would make lower offers to the angry recipient than to the happy recipient” (van Dijk et al., 2008). We particularly focus the original authors' finding that participants offered significantly fewer chips to angry recipients than happy recipients in low consequence situations, F(1,99) = 16.62, p < 0.0001. **A priori replication criteria.** We planned to conduct a 2 (recipients emotion: happy or angry) x 2 (consequences of rejection: no money or 10% reduction) between subjects ANOVA on the perceived anger ratings. We expected that participants in the angry recipient condition would perceive the recipient to be more angry than those participants in the happy recipient condition. A second manipulation check was planned to be conducted on perceived happiness ratings and a final manipulation check was planned to be performed on the consequences of rejection. Materials are available upon request. ## Testing Sites ## Tech: ML5.gameswithwords.org **Boston College** Team: Joshua Hartshorne, Lauren Skorb, Lily Feinberg, Jon Ravid Protocols: Protocol 1 and 2 IRB: Boston College Target number of participants: 120 (60 for each protocol) Compensation: $8/participant Language: English Notes: Protocols will be followed exactly **Wharton School** Team: Gideon Nave, Dylan Manfredi Protocols: Protocol 1 IRB: Wharton Target number of participants: 120 (60 for each protocol) Compensation: $8/participant Language: English Notes: **Universidad de los Andes** Team: William Jimenez-Leal, Andres Montealegre Protocols: Protocol 1 and 2 IRB: Local Target number of participants: 100 (50 for each protocol) Compensation: 14,000 pesos/participant Language: Spanish Notes: **Georg-August-Universität Göettingen** Team: Thomas Schultze, Stefan Schulz-Hardt, Katrin Rentzsch Protocols: Protocol 1 and 2 IRB: Local Target number of participants: 120 (60 for each protocol) Compensation: 7 euro/participant Language: German Notes: **Eotvos Lorand University** Team: Balazs Aczel, János Salamon, Barnabas Szaszi, Bence Bakos, Martin Kovacs, Peter Szecsi Protocols: Protocol 1 IRB: Boston College Target number of participants: 60 Compensation: $8/participant Language: Hungarian Notes: **Milestone Institute** Team: Anna Fedor Protocols: Online (Protocol 3) IRB: Boston College Target number of participants: 200 Compensation: $3/participant Language: English Notes: Collected through Crowdflower **FernUniversitat Hagen** Team: Mathias Kauff, Oliver Christ Protocols: Online (Protocol 3) IRB: Boston College Target number of participants: 128 Participant payment: 7 euro/participant Language: German Notes: Potentially very large sample **Maria Curie-Sklodowska University** Team: Rafał Muda, Ewa Hałasa, Karolina Krasuska, Katarzyna Kuchno, Emilian Pękala, Damian Pieńkosz, Barbara Sioma Protocols: Protocol 1 and 2 IRB: Local Target number of participants: 120 (60 for each protocol) Participant payment: $8/participant Language: Polish Notes: planning to recruit from Faculties of Mathematics, Law, and Administration ## The Manuscript ## A reproducible manuscript is available in Replications->Manuscript. To run the R Markdown file, you will need to have downloaded the entire Manuscript folder as well as the Replications->Data folder. They must be in the same root directory. ### Notes ### The main analyses were pre-registered, and the manuscript went through data-blind review. Both are available as registrations within OSF (see registrations affiliated with this project). In principle, nothing should change after these registrations. In practice, we found that we discovered unforeseen issues and outright errors that had to be corrected. These are noted as appropriate in the manuscript. We highlight them below here as well: 1. The text of the results-blind paper assumed that all regression analyses would report a single regression coefficient for each factor of interest. However, one of the factors was protocol, which has three levels. We chose to analyze significance using model-comparison, which is standard practice in our laboratory. This is now clarified in the Method. 2. Note that in one case, the dummy-coded regression coefficients *are* directly necessary for understanding the model, and so in that case they are also included (see Table 4 and related text). Also, we did not conduct model comparison for some of the supplementary analyses, mostly convenience of exposition. 3. Throughout, the results-blind manuscript neglected to specify how we were going to calculate p-values. We intended to use Wald’s z, which is standard practice in our lab is to use Wald’s z, so this is what we reported. No other method was tried or considered. This is now clarified in the Method. 4. The results-blind manuscript stated that we would use step-wise regression to determine the structure of the main regression model. It also stated that the main regression model would include the interaction of protocol and emotion. However, this interaction was not justified by the step-wise regression. We additionally included a model with that interaction and clarified this issue in the text. We also made a few minor additions/changes that I fully expect to be non-controversial: 5. “Follow-up analyses” appeared in both the main text and the supplementary analyses. We deleted it from the supplementary analyses. 6. In consultation with editor, added to “Disclosures” that authors were not blind to results in writing. 7. Data from six subjects was unusuable: three because no other subjects showed up for the testing time, and three because the software crashed. This possibility was not mentioned in the results-blind manuscript (and, indeed, I had forgotten about it until I triple-checked the R Markdown code before submission). 8. I corrected some spelling and grammar errors. These corrections do not change the meaning of the text. Finally, this is not strictly about the result-blind manuscript, but: 9. In our pre-registered analyses, there is a typo in a number of places, where in lmer notation it says (1|Subject) and clearly is intended to be (1|TestingSite). (The former makes no sense in a study with only one response per subject.)