Note: I was a reviewer on the Stage 1 manuscript.

This is well written and the methods follow what I was expecting from the Stage 1 report.

I have two main comments. The first is of more importance.

1.  The manipulations had little (if any) affect on people's responses to the manipulation check (or the memory distrust scale, but I'll focus here on the manipulation check). This was surprising, particularly for the manipulation check as those questions seem to be asking about the exact psychological constructs that the two non-control conditions are meant to affect. On lines about 411 of the Stage 1 manuscript it states:

Since the manipulation needs to reach a certain level of strength, only if the lower bound of the 80% CI on the effect size is above the minimal effect of interest (raw score difference of 1.6 with a SD of 2), will we consider the manipulation adequate. If the 80% CI is within the equivalence bounds [-1.6, 1.6], we will conclude that the manipulation did not reach an adequate strength.

A summary of the finding in the Stage 2 manuscript and the decision to still go ahead were:

We therefore concluded that the manipulation did not reach an adequate strength in manipulating state memory distrust.  Despite the manipulation not producing the intended effects, we proceeded with the main analyses to examine potential effects that may still inform our research question.

I had to go back and re-read the relevant section of the Stage 1 manuscript. I had interpreted the phrase above as meaning if they found, according the criteria they state above, that the manipulation had not worked, they would work on making a manipulation that, according to their criteria, did work. I understand that having the data they went onto further analyses, but it seems this is now different. Of course it may be that the manipulation checks and the memory scale do not work as they were expected to (e.g., people aren't aware of the influences of the manipulation), but that seems a question worth both addressing and trying to see if there is a manipulation check that does. The alternative is that the manipulation checks do work but the manipulation did not. If the manipulation had no (or little) effect, then the significant findings appear suspect. I went back to my original review to see if I raised issues with the manipulation check. I did:

"My concern at this point is whether the manipulation will have the desired effect on memory distrust that the authors believe. If I read their power analysis correctly they believe it will almost completely account for memory distrust because they base their analysis on the memory distrust to response bias effect, if the causal chain above how they believe that this works. This makes the manipulation check critical. 

As such, details of what counts for the manipulation check working is important for this. "

While the authors did state what they meant by failing the manipulation check, they did not examine, presumably through further data collection, why this occurred. I had assumed they would, but as a Stage 1 reviewer, I guess I should have asked them explicitly what they would do or requested a smaller manipulation check study before okaying the Stage 1 report.  I am relatively new (as I guess we all are!) to reviewing these. My apologies for not requesting these. I had assumed failing the manipulation check would have prompted actions other than just continue, and that is my fault making that assumption. The editor may have views about what failing a manipulation check means for the rest of these analyses.

On a minor note, in the stage 1 review they describe why it might be of interest to compare when the manipulation check was administered. In their code they list:

MC_Distrust_Commission <- 

   afex::aov_car(Manipulation_Check_1~ Condition*Manipulation_order + Error(participantId),data = dat_S2)
This should be reported.  I did find myself when reading the results going back to the code. The code page has a lot more analyses than appear reported.

One other note. In the original footnote 3 (but not included in the new version for some reason) discusses some issues with the manipulation. Thus, the authors were aware that this could be an issue.

2). If it is assumed that the manipulation worked (and therefore there was some unknown problem with the manipulation check), then the findings work as expected. I have two points. First, many of the outcome measures are dependent. For example, the c and beta. Similarly, when mixed models are done (and they aren't all linear, they are all generalized linear) statistics like d' and C are just coefficients in the probit regression (and really close in logit regression) if care is taken in coding the variables. Thus, these do not provide independent evidence and I assume the editor would prefer p-values only for one of these.

The second point is that the author note that the scores on the first recognition test (before the manipulation) could be used to predict the scores during recognition 2, and then measuring the additional impact of the manipulations would likely be more powerful. I skimmed the code file, but this is a long file. Were these done? This seems a better use of additional predictors than, for example, predicting the old/new response from belief ratings and other variables.

 ***************

In summary, the authors appear to have done what they said that they were going to do, just I had assumed (wrongly) what would occur if they failed their manipulation check. The topic they purport to be examining has changed. Their original title was:

The effects of memory distrust toward commission and omission on recollection-belief correspondence and memory errors

But since their manipulation did not seem to affect distrust as expected, the new title focuses on the manipulation. This seems a largely substantial shift between a Stage 1 and Stage 2 document, but I am new to these registered reports. Maybe this is common to change what the focus is on.

With that, and recognizing I am part to blame for not asking the authors to be explicit in what they would do if the manipulation check failed, I recommend acceptance with three caveats. First, I think it is important, assuming that this is accepted, that this aspect is made clear to readers at the start (e.g., in the abstract say the manipulation check failed, and that possible reasons for this are discussed). Second, I believe the paper would be much stronger if the authors address this issue (likely with new data) as part of this paper prior to publishing this manuscript. Third, given their original intent (agreed upon from the title in the Stage 1 manuscript) was exploring the causal chain: manipulation -> distrust -> memory errors, they should stress that this was not shown.

The impression of the great competence of the Authors which I gained after reading Stage 1, only deepened after reading Stage 2. I do not see any weak points in the analyses (in fact, in several places they went beyond my own analytical competence).

The manipulation check did not show the effectiveness of the manipulation. Despite this, the Authors made the difficult decision to continue the analyses. I agree with this decision. Nonsigificant p-values do not imply nonexistence of the effect. It is therefore possible that in reality the manipulation worked, but it was not possible to show this in the check. Still, the effects of the manipulation can be seen in the main analyses. After all, the method used for the manipulation check - individual sentences assessed on a Likert scale - is rather weak, and much weaker than the tests used in the main analyses. The Authors commented in the Discussion that it may be that “state memory distrust measure did not adequately capture the change in state memory distrust levels” (p. 31, l. 4-5; also p. 32, l. 20ff., and ‘Limitations’). It may be added that a single question is a less accurately measure than a questionnaire. This does not mean that other interpretations of the lack of an effect in the manipulation check are invalid. But the explanation may also be the low reliability of the measurement in the manipulation check. Currently, the Authors mainly discuss the issue of the validity of the manipulation check questions. It may also be worth mentioning their possible low reliability. After all, the p-values ​​were not far from the significance level.

Minor points

I suggest not reporting confidence intervals as ‘.00’. Actually, if it was indeed exactly zero, it would mean that the effect is significant. Perhaps increase the decimal places to three in such cases. Writing ‘< .01’ would be problematic unfortunately, as this does not exclude the the lower bound is negative, which was not the case.

As this is a stage 2 review, I have reviewed this manuscript according to the stage 2 crtiera, and categorised my comments within each of the criteria requirements.

Summary:

This study looks at whether giving false feedback about commission or omission erros influences participants' tendancy to respond either more, or less, librally to a memory test.  The intended manipulation of state mistrust did not appear to be effectives, although changes in the criterion were detected across feedback conditions anyway.  Overall, the paper is well constructed and well written, and I have no concerns relating to the stage 2 criteria.

1) Can the data test the hypothesis, by passing the outcome neutral criteria?

Yes,  The authors had previously shown a very thorough approach to their planning, and their approach to the data has been carried out, with all criteria being accounted for.  The approach to excluding participants was well justified, and sampling sizes were achieved.

2) Are the introduction, rationale and hypotheses the same as the stage 1 manuscript?

Yes, I could see no substantial changes from the stage 1 manuscript.

3) Were the registered study procedures adhered to?

Yes, the stage 1 study procedures were adhered to.

4) Where there any significant deviations from the analytical approach set out in stage 1?

The initial plan was followed, but additional exploratory analyses were conducted to investigate the failure of the manipulation to change experimental coniditon state mistrust ratings when compared to control state mistrust ratings.  Personally I found the additonal analyses to be well-justified and appropriate to investigating the potential mechanisms at work, so in terms of a stage 2 review, I have no changes to suggest.

5) Are the conclusions justified on the basis of the data collected and the analysis used?

I believe they are.  I agree with the authors that, on the basis of their analyses, the issue here may be that the state mistrust ratings are either not measuring what the authors' wanted them to measure, or else the criterion shifts did not occur because of changes in explicit state mistrust.  However, overall, I believe that this manuscript fulfills the stage 2 criteria.