I appreciate the opportunity to be a part of the review process for this replication. Below are some small suggestions for improvement: 

- on p. 19, when listing total participants, was this after exclusions? Please list the number of exclusions, and the reasons for excluding. 

- I noticed in the Methods and Results the authors refer to Studies 2 and 3. This can be a bit confusing, as it is easy to interpret this as the authors running 3 studies. It would be clearer (albeit take more space) to write "replication of Study 2" and "replication of Study 3". 

- On p. 31, under the heading "Replication: Extension Analysis", second line from the heading, I believe the authors meant to write "gratitude but not indebtedness" rather than "gratitude but no indebtedness".

- I noticed in the Methods in p. 32 (under the "Order Effects" heading), the authors used a mix of past, present, and future tense verbs. It would be clearer if the verb tenses are consistent, such as all past tense. Do check before p. 32, as it may have also occurred before then, but it stood out the most to me on this page. 

- On p. 52, the authors suggest that their results are counter to previous research linking gratitude to prosociality. I think this is too big of a jump. The authors do not measure prosocial behaviors, or even prosocial behavioral intentions. Additionally, Ma et al.'s 2017 meta-analysis linking gratitude to prosociality suggests there generally is a relationship, and the relationship is not necessarily countered by only one study. If the authors feel strongly about leaving this point in, I suggest making the point more tentatively. This same issue appears again on p. 54 where the authors suggest that De Steno's work was not replicated. Again, the authors did not measure dependent measures that De Steno did, and there is a large literature supporting the relationship between gratitude and prosociality. 

Signed, 

Jo-Ann Tsang

Overall, I believe the authors did an excellent job in implementing the study. They adhered closely to the stage 1 registration to conduct the study, test the hypotheses, and report the results. Very glad to see that this important pioneering work by Tsang (2006), which I cited a lot in my own manuscripts, has been successfully replicated. Congratulations to the authors, and thank you for all the efforts in working this out.

Here are some of my concerns and suggestions to improve the manuscript.

For Study 3, H7, the authors proposed a different hypothesis on gratitude regarding the ambiguous condition (benevolent>ambiguous>selfish) compared to Tsang (benevolent>selfish>ambiguous). This hypothesis is not well justified, nor are the results (and its comparison with Tsang) adequately discussed. Additionally, I would expect the authors to report the t-test comparing the benevolent vs. selfish conditions on gratitude, which is crucial but currently missing. Specifically, the effect size seems much larger in this study (6.59 vs. 5.48) compared to the original one (6.76 vs. 6.49). These findings are important and could be discussed.

The authors did several two-way ANOVAs to examine the interaction effects. What these results mean and its implications are somewhat unclear and unexplained.

On page 40 par.2, under manipulation check, it states, “However, we found no support for rated helper intention in the benevolent condition as different from that in the ambiguous condition”. This reporting appears incorrect, as this paragraph is about the rated magnitude of favor.

I did not see the report related to the order effect. I understand that the authors registered to do this only if the data failed to support the hypothesis. But as the authors indicated, “one disadvantage is that answers to one scenario may bias participants”. Therefore, I believe it’s still important for the readers to understand whether, or to what extent, combining two studies into one may have generated bias.

The results for the extension part are a bit confusing to me. The authors concluded that gratitude and indebtedness are both not associated with reciprocity tendency, but no statistical analysis was presented to directly support this statement. Moreover, based on this conclusion, the authors discussed that this is a conceptual replication of Bartlett & DeSteno (2006) and Peng et al., (2020), which I disagree with. In these two articles, the IV is Help vs. No Help, and the DV is actual helping behavior rather than reciprocity tendency, which are quite different from the current experiment setting.

Best of luck with your future research!

Cong Peng

Dear authors/recommender, 

It was a pleasure to see this article in its final form, after having reviewed this as a stage 1 protocol. I thought the protocol was solid, and the completed study reinforces that. It has been carefully and thoroughly conducted (which of course is partly due to the cooperation of the original authors), and written up with plenty of detail which allow the reader to evaluate its quality. As I believe I mentioned in my original review, I think it could be enhanced by the addition of Bayesian statistics, but that isn't a deal-breaker, especially as the p-values do not really come into the ambiguous territory (i.e., most of them are quite small and would be likely to easily recalculate to pro-alternative BFs). 

All-in-all, congratulations to the authors on a very well-done study! I have no qualms in seeing this article in print! 

Sarahanne M. Field