F1000Research F1000Research 2046-1402 F1000 Research Limited London, UK 10.12688/f1000research.25573.2 Research Article Articles Stage 2 Registered Report: How subtle linguistic cues prevent unethical behaviors

[version 2; peer review: 2 approved]

 Guo Wen Conceptualization Data Curation Formal Analysis Investigation Methodology Writing – Original Draft Preparation Writing – Review & Editing https://orcid.org/0000-0002-5544-0033 1 Liu Huanxu Conceptualization Data Curation Formal Analysis Investigation Methodology Writing – Original Draft Preparation Writing – Review & Editing https://orcid.org/0000-0001-7078-5956 1 Yang Jingwen Conceptualization Data Curation Formal Analysis Investigation Methodology Writing – Original Draft Preparation https://orcid.org/0000-0002-6050-656X 1 Mo Yuqi Conceptualization Data Curation Formal Analysis Investigation Methodology Writing – Original Draft Preparation 1 Zhong Can Conceptualization Data Curation Formal Analysis Investigation Methodology Writing – Original Draft Preparation 1 Yamada Yuki Conceptualization Funding Acquisition Investigation Methodology Project Administration Resources Supervision Validation Writing – Original Draft Preparation Writing – Review & Editing https://orcid.org/0000-0003-1431-568X a 2 Graduate School of Human-Environment Studies, Kyushu University, Fukuoka, Japan Faculty of Arts and Science, Kyushu University, Fukuoka, Japan yamadayuk@gmail.com

No competing interests were disclosed.

 16 11 2020 2020 9 996 3 11 2020 Copyright: © 2020 Guo W et al. 2020 This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Background: Differences in descriptions can influence people’s evaluations and behaviors. A previous study by Bryan and colleagues suggested that subtle linguistic differences in ethical reminders can differentially prevent readers’ unethical behavior. The present study tried to replicate the previous finding in the Japanese context (Experiment 1); additionally, we explored the influence of unfamiliar Japanese instruction words that captured participants’ attention (Experiment 2).

Methods: In two online experiments, participants were asked to make 10 coin-tosses and report the number of “heads” results, which would indicate the amount of money that they could earn. In Experiment 1, we analyzed the difference in the number of “heads” results as reported by 768 participants under three conditions with different instructions (“Don’t cheat” vs. “Don’t be a cheater” vs. baseline as a control). In Experiment 2, we conducted an extended experiment with an additional task in which more attention was directed toward the text.

Results: In Experiment 1, we successfully replicated the results of the original experiment. The results of Experiment 2 showed no evidence that the results in Experiment 1 were influenced by attentional factors.

Conclusions: In conclusion, the results of the present study supported the hypothesis that self-identity-related words of moral reminder curb unethical behaviors more effectively.

Stage 1 report: https://doi.org/10.12688/f1000research.20183.4

cheating persuasion attention moral self construal labeling Japan Society for the Promotion of Science 15H05709 16H01866 17H00875 18H04199 18K12015 20H04581 This research was supported by JSPS KAKENHI (15H05709, 16H01866, 17H00875, 18H04199, and 18K12015). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Amendments from Version 1

Based on the reviewers' comments, some minor changes were made to improve clarity and accuracy. 1. At the reviewer’s suggestion, we have revised the position of a sentence in the introduction section that is part of Stage 1, and we are therefore stating it here. Such modifications do not change the initial logic or structure of Stage 1. Nor do they affect the hypotheses or conclusions of this study at all. (the 6th paragraph of Introduction) 2. The “Outlier extraction” section in Stage 1 has been replaced with “Exclusion criteria” in Stage 2. We set the “Exclusion Criteria” section before the “Data analyses” section. These changes in expression and layout do not affect the hypothesis and conclusion of our study. (the Exclusion criteria section of Experiment 1 Methods)

 Introduction

When people behave dishonestly, they usually downplay the seriousness of the dishonest act (e.g., Monin & Jordan, 2009; Steele, 1988), weakening the link between the dishonesty and one’s self-identity (e.g., Bandura, 1999) to avoid the correspondent inference ( Jones & Nisbett, 1972; Ross, 1977) that one is the kind of person who behaves dishonestly. According to self-concept maintenance theory, individuals in general strive to create and maintain an image of themselves as good and ethical people ( Markus & Wurf, 1987; Mazar et al., 2008).

According to Blasi (1984), a moral person is one for whom moral categories and moral notions are central, essential, and important to self-understanding. Morals cut deeply to the core of what and who such people are as individuals. However, one study revealed that highly constructed self-identities are associated with more unethical behaviors ( Cojuharenco et al., 2012).

Regarding ethical behavior, a moral-character model has been proposed, where moral character consists of motivation, ability, and identity elements ( Cohen & Morse, 2014). Moral identity here refers to being disposed toward valuing morality and wanting to view oneself as a moral person. This disposition should be considered when attempting to understand why people who behave unethically tend to apply a variety of strategies to weaken the behavior–identity link ( Bandura, 1999). The use of “euphemistic labeling” to describe one’s attributes and weaken the link regarding language should also be included in this disposition.

Different ways of description can easily influence people’s evaluation and judgment about something, even if they have a wealth of previously established knowledge ( Fausey & Boroditsky, 2010). For instance, using a transitive verb (agentive description, e.g., “Timberlake ripped the costume”) to describe an accident makes participants significantly more likely to blame the actor compared to the same description with the words changed to an intransitive verb (nonagentive description, e.g., “The costume ripped”). Another study found that, for children aged 5–7 years old, when a noun label was employed to describe a character (e.g., “She is a carrot-eater”) rather than a verbal predicate (e.g., “She eats carrots whenever she can”), their judgment about those characteristics would be more stable over time ( Gelman & Heyman, 1999). The same phenomenon has been demonstrated regarding self-perception ( Walton & Banaji, 2004). It is possible that language has some effect in this category ( Gelman et al., 2000) because when nouns are used to refer to something, one may have a deeper understanding of it, which is noted to “enable inductive inferences” ( Gelman & O’Reilly, 1988).

Once the subtle description is used to refer to oneself, a noun label may have a stronger effect. Bryan et al. (2011) found that more people would choose to vote if they heard the words “be a voter” rather than “to vote” on the day before election day. Additionally, research showed that, compared to “helping,” “being a helper” encouraged more children to conduct kind behaviors toward others ( Bryan et al., 2014). However, subsequent research found that although “being a helper” can lead to more kind behaviors initially, once there is a setback, the backlash may also be stronger accordingly ( Foster-Hanson et al., 2020). The reason underlying this phenomenon is as follows: as category labels, nouns bear a strong link to identity and may lead to self-doubt once one fails.

According to Bryan et al. (2011), the effect of noun expression comes from a motivation-driven process. When a noun is involved with a positive identity such as “voter” and “helper,” people simply see themselves as voters or helpers and they produce more correlated behaviors; When the noun is involved with undesirable (negative) identities, however, these kind of words should cause people to avoid correlated behaviors. Highlighting a self-identity word will prevent unethical behaviors to some degree ( Bryan et al., 2013).

In social psychology, experiments of priming of unethical behaviors and its subsequent prevention typically involve money or time ( Gino & Mogilner, 2014; Gino & Pierce, 2009; Mogilner & Asker, 2009; Vohs et al., 2006). A mere exposure to money is associated with unethical outcomes ( Kouchaki et al., 2013). In Gino et al.’s experiment (2014), participants were asked to complete a scrambled-sentences task using some money-related words or time-related words; results showed that priming time (rather than money) makes people behave more ethically.

In contrast, another experiment by Bryan et al. (2013) allowed experimenters to prevent unethical behaviors through semantic priming. They manipulated the task’s instructions by changing the use of verbs (“Don’t cheat”) to noun labels (“Don’t be a cheater”) to inhibit participants from engaging in unethical behaviors. The self-identity related group (“don’t be a cheater”) had significantly lower proportion of unethical cheating behaviors.

In the present study, we aim to replicate Experiment 3 of Bryan et al. (2013), for the following reasons:

First, the participants in Experiment 1 in Bryan et al. (2013) were asked to think of a number from 1 to 10. If the number was even, they were paid $5; if it was odd, there was no reward. Bryan et al. (2013) paid for even numbers because it has been reported that participants typically show a strong bias toward odd numbers in a random number generation task ( Kubovy & Psotka, 1976), but this oddness bias had not been confirmed for betting behaviors. Furthermore, an even or odd number participants think of is just imaginary, occurring in one’s inside world, not an external real event; hence, it is difficult to use it as an index of falsification. An index used for cheating should emphasize that participants’ reports can differ from the fact. Thus, we abandoned the method of Bryan et al. (2013) Experiment 1. In their Experiments 2 and 3, they used a coin-tossing task: participants were asked to toss a coin and receive a reward corresponding to the result of their coin flips. We choose this method for our experiment because tossing a coin induces a real external event, which is more objective and operable, and hence it is better than thinking of a number to measure cheating behavior. In addition, compared with Experiment 2 in the original study, which just used two conditions, “cheater” and “cheating”, a baseline group was included in Experiment 3, which made Experiment 3 more complete in its design—an approach we followed also.

Moreover, we found that the effect size in Experiment 3 was small ( f = 0.302 in G*Power (significance level α = 0.05, power level 1-β = 0.95), meaning that Experiment 3 required at least 174 participants; in fact, only 99 people joined the original research. From this, we suppose that the effect size in Experiment 3 was overvalued.

According to the above review, high levels of self-identity and the willingness of individuals to maintain a positive self-view should prevent unethical behaviors. We predict that the self-relevant noun “cheater” will curb cheating behaviors more significantly than the verb “cheating” and the baseline condition (in which there is no reminder in the instruction).

 Methods Experiment 1

Our experiment was conducted online in a private and impersonal way, which means that participants did not meet or be expected to meet the experimenters. We aimed to replicate Experiment 3 of Bryan et al. (2013), in which there were three conditions: “cheater,” “cheating,” and “baseline”; in the baseline condition, a reminder about cheating was not mentioned.

Participants. Participants were users of the Yahoo! Crowdsourcing Service in Japan. Participants were required to meet the a priori criterion that they were native Japanese speakers. We conducted a pilot test to determine the shortest time in which one could reasonably participate in the experiment in good faith. This pilot test is detailed in a later section ( Outlier extraction). Participants were excluded if they completed the experiment faster than the pilot test time. Repeat participation was prevented.

Procedure. At the beginning of the experiment, demographic information of participants’ age and gender was collected. Online instructions indicated that a recent controversial article has claimed to report the first scientific evidence for paranormal phenomena ( Bem, 2011); this was the same cover story used in the original study.

We asked the participants to find a coin at home and, while trying to influence the outcome of each toss with their minds, flip the coin 10 times, making it land on “heads” as often as possible. They were asked to ensure that they were “properly motivated.” For every toss landing on “heads,” they received 10 T-points (≈ 10 yen) as a reward. The instructions indicated that participants would receive a reward corresponding to the result of their coin flips. Noting that the average Yahoo! Crowdsourcing user receives two T-points for one task, the reward we set was five times that amount. Even those participants who reported a chance level of five “heads” received 25 times more points than for a normal task. This reward should be large enough for some participants to engage in cheating behavior.

Participants were randomly assigned the three conditions. All three conditions were posted on the crowdsourcing website, and a person who participates in one of the conditions cannot participate in the others. This resulted in passive randomization. In the “cheater” and “cheating” conditions, the manipulation was embedded in the instructions. We referred to the instructions in the original experiment ( Bryan et al., 2013) and translated it into Japanese, as follows:

NOTE: Please don’t [cheat/be a cheater] and report that one or more of your coin flips landed heads when it really landed tails! Even a small [amount of cheating/number of cheaters] would undermine the study, making it appear that psychokinesis is real.

[嘘をつかない/嘘つきにならない]でください。コイントスの結果，実際にはコインの裏が出たのに表が出たと報告しないでください。

たとえわずかでも[嘘があると/嘘つきがいると]，この研究を台無しにし，念力が実在するように見せてしまいます。

The manipulation was also embedded in the instructions on the next page, where participants log the outcomes of their 10 coin-flips. At the top of the page, a message was displayed in large red words: “嘘をつかないでください/嘘つきにならないでください” This means “PLEASE DON’T [CHEAT/BE A CHEATER]” as in the original experiment ( Bryan et al., 2013).

In the baseline condition, the instructions were the same as above, except that the cheating message was not mentioned.

Power analysis and sample size. Because Experiment 3 of Bryan et al. (2013) did not report the effect size, η 2, first, we calculated the effect size of the analysis of variance (ANOVA) result from the F and df values. Bryan et al. (2013) reported the statistics of their one-way ANOVA as F(2, 96) = 4.38, p = .015. Hence, we calculated η 2 based on Cohen’s (1973) method, as η 2 = .0836. Then, we calculated the effect size, f, as follows: f = √(η 2/(1 – η 2) = 0.302. The small sample size may overestimate the effect size so, as a replication convention (e.g., Nitta et al., 2018), we halved the effect size of the original experiment, and used G*Power 3.1.9.3 ( Faul et al., 2009) to conduct a power analysis (i.e., to 0.151). In G*Power, we set the significance level α = 0.05, power level 1-β = 0.95, and effect size f = 0.151. According to the conditions of the original experiment, we divided the participants into three groups. The required total sample size was 681, with 227 participants in each group; therefore, we tried to recruit at least 681 participants, and data collection did not exceed 810 participants. This stopping rule was set because it was difficult for us to limit the number of participants to exactly 681, due to the characteristics of the simultaneous participatory online recruitment system; therefore, we allowed for up to 120% of the required sample size (i.e., 810). If more than 810 people participated in the experiment, we selected the data of the first 810 participants based on the time stamp and used this for the analysis. Also, we set the number of participants (max. 365 males and 445 females) to match the gender distribution of the original study (male: female = .45:.55).

Exclusion criteria. For our online experiment, we established a minimum completion time (MCT) for inclusion in the final sample by asking five colleagues who were unfamiliar with this experiment to complete the experiments as fast as possible, then calculating the mean completion time. Specifically, each colleague performed a coin toss ten times; after each toss they recorded the result on the experiment website. This pilot test did not include the attempt to motivate psychokinesis and measured only the required time of the coin toss and recording. Bryan et al. (2013) also used the MCT as an extraction criterion. We excluded those participants who completed it faster than the MCT, because they might have rushed through the experiment and failed to complete it in good faith.

Data analyses. In this study, the dependent variable was the mean number of “heads” reported. In the original experiment, a one-way ANOVA and t-test were performed. Specifically, the ANOVA was performed for analyzing the main effect of the three groups. A problem in the original study was that the authors did not report adjustments for any significance level in subsequent multiple comparisons. Therefore, in the present study, we used a one-way ANOVA and Tukey’s method for the multiple comparisons. Additionally, in order to check the cheating in each group, the original study performed one-sample t-tests between the mean number of “heads” reported and the chance level (i.e., 50%; 5 times out of 10 flips). These analyses were performed using jamovi (version 1.0.5). The original results are summarized in Table 1.

Results of Experiment 3 of <xref ref-type="bibr" rid="ref-4">Bryan <italic toggle="yes">et al.</italic> (2013)</xref>.
Analysis types Reported p-value Degree of freedom Effect size
Main effect: three groups .015 96 f = 0.302
t-test “cheating” vs “cheater” .013 96 d = 0.71
“cheater” vs baseline .004 96 d = 0.66
“cheating” vs baseline > .80 96 d = 0.05
“cheating” vs “chance” < .0005 36 d = 0.79
baseline vs “chance” < .0005 35 d = 0.78
“cheater” vs “chance” > .30 25 d = 0.19

Moreover, as the dependent variable was based on the counts of “heads” reported and that the 10 coin tosses were nested within each participant, a quasi-Poisson or Poisson regression was used for exploratory analyses. In the (quasi-)Poisson model, the variance was assumed to be the mean multiplied by a dispersion parameter ( Ma et al., 2014). Dispersion parameters with a value greater than one indicate that overdispersion exists; in this case, quasi-Poisson regression will be performed. Thus, which analysis to used depends on the result of variance and the mean of “heads” counts. We first tested the original hypothesis. Then, information of gender and age were added as predictors to establish a regression model.

 Experiment 2

This experiment was employed as an extended, conceptual replication of Experiment 3 in the original study ( Bryan et al., 2013). Our Experiment 2 was only performed when the results of Experiment 1 successfully replicated those of the original experiment. In the original experiment, the numbers of heads claimed in the “cheater” condition was significantly lower than that in the “cheating” and baseline conditions, but no difference was found between the “cheating” and baseline conditions. Here we cannot easily interpret the non-significant results based on self-identity alone. We aimed to test whether lower levels of attention to the instruction in the “cheating” condition reduced the effectiveness of preventing dishonest behaviors in our Experiment 1. Thus, we conducted Experiment 2, adding a “cheating” with attention task condition in which we used tasks concerning an instruction to ensure that participants’ attention was captured (e.g., Folk et al., 1992; Folk et al., 2002). When we translated the instruction into Japanese, we felt the unfamiliarity of a “cheater” condition in a Japanese language situation. Participants in our experiment might find that the reminder “don’t be a cheater” commands extra attention because of this sense of deviation. Therefore, even if the result of the original experiment was completely reproduced in our Experiment 1, it would not fully support the finding of the original experiment, as the reason for the possible different dishonest behavior rates between the “cheating” and “cheater” conditions in our Experiment 1 might be that the participants in the “cheating” group paid relatively less attention to the instruction; for this reason, “cheating” might have worked weakly as a moral reminder in this condition. Because the experiments are conducted online, it was difficult to ensure that the participants have actually seen and understood the instruction; in addition, it was also possible that the participants ignored the instructions of Experiment 1 due to satisficing, (e.g., Chandler et al., 2014; Oppenheimer et al., 2009; Sasaki & Yamada, 2019), further diminishing the effect of the unattended reminder (i.e., “cheating”). In this Experiment 2 we addressed these attention-related effects.

Noticeably, the main difference between our Experiment 1 and the original Experiment 3 lay in the different language used in the instruction. Thus, if our Experiment 1 was a successful replication, we would then choose to focus on the expression used in the Japanese instruction, rather than the English instruction of the original Experiment 3.

To support this approach, we conducted a preliminary experiment, asking participants to evaluate their familiarity with certain expressions in Japanese. The expressions “Don’t cheat” and “Don’t be a cheater” were translated into Japanese, and native speakers evaluated their familiarity with them (1: not familiar to 5: very familiar) via an Internet survey on Yahoo! Crowdsourcing. The protocol of this experiment was registered on the Open Science Framework ( Guo et al., 2020). The results showed that the familiarity rating score in the “cheater” condition was significantly lower than that in the “cheating” condition, t(64) = 6.73, p < .001, Cohen’s d = 0.834. Hence, we conjectured that the anticipated difference in the results between the “cheating” and “cheater” conditions in Experiment 1 may partly occur due to differences in attention paid to the instruction, instead of the preservation of a positive self-image proposed by the original study ( Bryan et al., 2013). This meant that part of the effect of the “cheater” condition was due to the unfamiliar expression, which attracts people’s attention then played a role in preventing them from conducting unethical behavior. See Extended data for details about this experiment.

In our Experiment 2, we manipulated the way in which participants saw the instructions to explore the differences between the “cheating” and baseline conditions. Experiment 2 comprised three conditions: “cheating,” “cheating” with attention task, and baseline. We predicted that the “cheating” with attention task condition would be more effective in curbing unethical behaviors than the “cheating” and baseline conditions, because the task would arouse more attention. While the instruction in the “cheating” condition was in large red capital letters, this should entail no significant difference compared with baseline.

Procedure. The procedure for Experiment 2 was identical to that of Experiment 1, except for important differences in two aspects. In Experiment 2, we focused on whether the participants read the instructions as diligently as we expect. First, we omitted the original “cheater” condition and included instead another “cheating” condition (i.e., “cheating” with attention task condition). Second, in the “cheating” with attention task condition, we added a task page in which participants were asked to choose the exact expression (i.e., “Don’t cheat”) that appeared on the screen from three sample sentences. We reminded participants of this task in advance to ensure they read the instructions carefully.

Power analysis and participants. Because the power analysis of Experiment 2 was the same as in Experiment 1, we intended to recruit participants in the same way as Experiment 1. The minimum completion time was also established for participants to be included in the final sample. This exclusion standard was similar to that in Experiment 1.

Data analyses. In Experiment 2, the dependent variable was the mean number of “heads” reported. We still used a one-way ANOVA and Tukey’s method for the multiple comparisons. To check the cheating rate in each group, a one-sample t-test between the mean number of “heads” reported and the chance level (50%) was analyzed. The data of participants who failed to provide the right answer to the attention task were not used for further analysis. Another analysis by a (quasi-)Poisson regression model was also performed to explore the contribution factors of cheating counts.

 Study timeline

After Stage 1 acceptance, our colleagues were asked to complete the pilot test to calculate the MCT. Then, we posted our experiments on the Yahoo! Crowdsourcing Service to recruit participants. The experiments and subsequent analysis were conducted from March 29, 2020 to July 10, 2020.

Ethical approval and consent to participate

The present study received approval from the psychological research ethics committee of the Faculty of Human-Environment Studies at Kyushu University (approval number: 2019-004). Completion of experiments by participants was regarded as consent to participate; they also had the right to withdraw from the experiment at any time without providing a reason. In addition, we protected participants’ personal information. Because this study was conducted online, even if participants engaged in cheating behaviors, we could not identify them or meet the participants face-to-face.

 Results Experiment 1

We recruited 1298 users of the Yahoo! Crowdsourcing service who clicked on a survey called “Coin Game.” Of these participants, 803 met our minimum completion time (MCT) criterion for good faith during the course of the experiment. Moreover, we deleted the data of participants exceeding the limit we set to match the gender distribution of the original study; finally, we used data from 768 participants (males = 315, females = 445, no response = 8, M age = 45.68) for analyses.

In this study, the dependent variable was the mean number of “heads” reported. We performed a one-way ANOVA and Tukey’s method for the post hoc comparisons. The results of ANOVA showed a significant main effect of condition ( F (2,765) = 5.86, p = .003, f = 0.123). Multiple comparisons showed that participants in the “cheater” condition reported having obtained significantly lesser “heads” than the participants in the “cheating” condition ( t (765) = 2.554, p = .029, Cohen’s d = 0.22) and the baseline condition ( t (765) = 3.24, p = .004, Cohen’s d = 0.29). The “heads” reported in the baseline condition had no significant difference with the “cheating” condition ( t (765) = 0.754, p = .731, Cohen’s d = 0.07). The results of Experiment 1 suggested that participants instructed with “cheater” had a significantly lesser cheating rate than those instructed with “cheating” and those that were uninstructed ( Figure 1A). Hence, Experiment 1 successfully replicated the results in the original study ( Bryan et al., 2013).

The mean number of “heads” in each instruction condition in Experiment 1 (Panel A) and Experiment 2 (Panel B). Error bars indicate 95% confidence intervals of the mean.

Moreover, we compared each condition to a chance level (i.e., five times) to check bias in each group. The “cheating” ( M = 5.22) and baseline ( M = 5.33) groups had a significantly higher rate than five ( t (265) = 2.20, p = .029, Cohen’s d = 0.135 and t (241) = 3.31, p = .001, Cohen’s d = 0.212, respectively). The number of “heads” in the “cheater” ( M = 4.86) group did not significantly differ from five ( t (259) = 1.36, p = 0.176, Cohen’s d = 0.084). In this respect too, we succeeded in replicating the results of the original experiment.

The dependent variable was based on counts of reported “heads” and 10 trials were nested within each participant. A Poisson regression analysis on the number of reported “heads” was performed. We first added “condition” as a factor into the model. The results indicated that condition ( χ 2 (2) = 6.23, p = 0.044) significantly contributed to explaining variance related to the number of “heads.” In contrast, when participants’ gender and age were used as predictors for the regression model, the results indicated that gender ( χ 2 (2) = 1.11, p = .575) and age ( χ2 (58) = 26.40, p = 1.00) did not significantly predict the number of “heads.”

Although we successfully replicated the results in the original study, it was still premature to conclude that the subtle difference in the instruction affected the cheating behavior. In particular, as there is a significantly different degree of familiarity between the words “cheater” and “cheating” in the Japanese instruction (see our preliminary experiment), there was still a possibility that attentional factors might have influenced the cheating. Therefore, we conducted Experiment 2.

 Experiment 2

We recruited 1255 users of the Yahoo! Crowdsourcing service. Among these, 800 (males = 365, females = 425, M age = 43.35) met the MCT criterion for good-faith participation and were used for analyses. The number of participants by gender did not exceed the maximum sample size we had set.

In Experiment 2, three conditions, “cheating,” “cheating” with attention task, and baseline groups, respectively, were employed. The analysis was identical to Experiment 1. The result of the one-way ANOVA did not show a significant main effect of condition [ F (2, 797) = 0.667, p = .514, f = 0.045 ( Figure 1B)]. The cheating behavior in the “cheating” with attention task condition did not significantly decrease more than the baseline and “cheating” conditions ( t (797) = 0.957, p = .604, Cohen’s d = 0.01 and t (797) = -1.034, p = .555, Cohen’s d = 0.09).

Moreover, we also performed a t-test between each condition and a chance level of five. The “cheating” with attention task condition ( M = 5.25) and baseline ( M = 5.23) condition had a significantly larger number than five ( t (242) = 2.380, p = .018, Cohen’s d = 0.153 and t (286) = 2.389, p = .018, Cohen’s d = 0.141, respectively). The number of “heads” in the “cheating” condition had no significant difference with five ( t (269) = 0.937, p = .350, Cohen’s d = 0.057).

A Poisson regression analysis on the number of reported “heads” was also performed. Condition [ χ 2 (2) = 0.628, p = .731] did not significantly contribute to explain variance related to the number of “heads.” Additionally, gender [ χ 2 (2) = 0.728, p = .695] and age [ χ 2 (57) = 27.139, p = 1.00] did not significantly predict the number of “heads.”

These findings do not suggest that the settings made for participants to pay full attention to the reminder instruction significantly reduced the incidences of cheating. Therefore, it is reasonable to assume that the significant effect of “cheater” in Experiment 1 was not due to excessive attention to the instruction.

 Discussion

In this study, we directly replicated Experiment 3 of Bryan et al. (2013) (Experiment 1) and addressed potential attentional effects related to translation (Experiment 2) using a Japanese sample. These registered experiments were perfectly performed without any deviation from the protocol. As a result, in Experiment 1, we successfully replicated the original results of Bryan et al. (2013). We found a significant main effect of condition similar to the original study, and the “Don’t be a cheater” had a minimal number of reported “heads” ( M=4.86), which meant participants in this group had the lowest rate of dishonest behavior. However, we could not conclude that the participants in the “cheater” group have lesser cheating behaviors than those in the “cheating” group. As we predicted in the Stage 1 protocol and confirmed in a preliminary experiment, the unfamiliar expression of “Don’t be a cheater” in the Japanese context would have made participants pay more attention to this sentence; hence, it would have made for a more effective reminder. In other words, “Don’t cheat” in the Japanese context might be ignored by some participants and lose its role as a reminder. Therefore, there were more cheating behaviors in the “cheating” condition.

To examine the attentional factor on cheating behaviors, we added an additional “Don’t cheat” condition with a recognition task for the reminder. We aimed to test whether the lower levels of attention caused by a familiar instruction (“Don’t cheat”) in the “cheating” condition reduced the effectiveness of preventing unethical behavior in our Experiment 1. One of the mechanisms of the theory of self-concept maintenance relies on the attention to standards of one’s conduct ( Mazar et al., 2008). As Mazar et al. (2008) found, when reminded of their standards, people had to face their actions, and therefore, could be more honest. However, in Experiment 2, no significant main effect of conditions was shown, which clarified that the proper attention to the reminder “Don’t cheat” was still ineffectual in curbing the cheating behavior. We did not obtain results consistent with Mazar et al. (2008). Thus, we excluded the role of the attentional factor in verb-based and noun-based reminders in the Japanese context. The results supported the hypothesis proposed by Bryan et al. (2013) that, in such cases, undesired self-relevant nouns can cause people to avoid a particular behavior and such nouns influence behavior by emphasizing its implications for identity ( Bryan et al., 2011).

Although we successfully replicated the original study, some differences were present in terms of results. Note that the effect size of the present study was considerably smaller than the original study ( f = 0.302 for the original study and f = 0.123 for the present study). Similarly, in the multiple comparisons, as per the trend (“cheating” – “cheater” = 0.71 and “cheater” – baseline = 0.66 for the original study vs. “cheating” – “cheater” = 0.22 and “cheater” – baseline = 0.29 for present study), it is likely that there was some inflation of effect size due to the small sample size in previous studies (99 participants in the original study vs. 768 participants in the present study). Another replication of the original study by Savir & Gamliel (2019) also obtained results with a lower effect size. They added positive-valence appeals to check the effectiveness of self-appeals in reducing unethical behavior. In their meta-analysis, “cheater” was actually more effective than “cheating”, whereas they reported a lower effect size (overall estimate of 0.17) as compared with Bryan et al. (2013) (overall estimate of 0.80). In the present study, the effect size was comparable to the previous replication study ( Savir & Gamliel, 2019).

Besides the comparison between conditions, to check bias in each group, we compared the “heads” times in each condition in both experiments to a chance level. Inconsistent with the original experiment and Experiment 1, in Experiment 2, there was no significant difference between the “cheating” condition and the chance level. This inconsistency in the results between the present experiments does not affect the interpretation of the main hypothesis we tried to test here. Yet, whether this “cheating” reminder does or does not affect behavior is inconsistent across studies as well, and hence it will be interesting to further investigate what factors define the difference. Note that the present study showed that the effect size of the reminder itself is very small, so it is clear that no promising hypothesis exists at this point as a crud factor ( Orben & Lakens, 2020).

The present study contributes to research across cultures and to the generalization of the original effect. Previously published studies employing a similar method have studied this effect in US and Israel ( Bryan et al., 2013; Savir & Gamliel, 2019), but to the best of our knowledge, the subtle linguistic effects of moral reminders on cheating have not been investigated in Asia. The results of this study also provide a cross-cultural validity of the inhibitory effects of different lexical-based warnings on unethical behaviors.

In the present study, we selected a sample of Japanese people in Asia (instead of Americans in the original experiments) and set up an authentic online survey environment as in the original study (instead of a laboratory experiment) to ensure that participants behave naturally. Given that in the two language contexts (Japanese and American English) self-identity words curb unethical behaviors more effectively, we expect that our results with users of a crowdsourcing platform can be generalized to other online contexts. However, given that the priming of unethical behavior in the present study (the same as Bryan et al., 2013) was based on the coin flip paradigm, the pattern of results, which might hold only for the instant coin flip paradigm, do not generalize to curbing all kinds of unethical behaviors. It should also be noted that linguistic expressions conveying the same meaning may still vary greatly in two different language families, depending on the cultural habits and the language structure itself (as in our preliminary experiment, “Don’t be a cheater” was not a familiar expression in Japanese, as it is not usually used in that way); therefore, it is still necessary to consider the generality of this effect in countries with a similar linguistic system in future research. Considering the results of the Poisson regression analysis, we have no reason to believe that the results depend on other characteristics of the participants. It is unclear at this time whether other materials and contextual factors could be constraints on generality ( Simons et al., 2017).

A limitation that we want to discuss is the considerably high rate of dropout: 495 people of 1298 (38.1%) and 455 people of 1255 (36.2%) in the two experiments, respectively. In future online surveys, we may need to recruit participants over 150% of the required sample size to ensure an adequate number of participants. There were advantages and disadvantages to conducting the cheating experiment online; because of anonymity, participants were free from social expectations and could cheat more easily ( Grym & Veronica, 2016). However, we did not have a true grasp of whether participants completed the experiment in good faith (e.g., Sasaki & Yamada, 2019).

 Conclusion

In conclusion, this study successfully replicated the results of Experiment 3 of Bryan et al. (2013) as a registered report without any deviation from the protocol, and excluded the possibility of the attentional factor when the reminder was translated to Japanese. We closely followed their experimental design and analytic methods. The results of our study supported the hypothesis that self-identity-related nouns of moral reminder curbed unethical behaviors more effectively than when the same meaning was expressed using verbs; furthermore, they supported the cross-linguistic validity of these findings. Moreover, we found that the attentional factor showed no effect in decreasing unethical behaviors in the Japanese context. Future research should also consider personal traits when focusing on the inhibiting effects of language on individual unethical behaviors.

 Data availability Underlying data

Open Science Framework: How subtle linguistic cues prevent unethical behaviors, https://doi.org/10.17605/OSF.IO/AXDHV ( Guo et al., 2020).

This project contains the following underlying data:

Data collected for Experiment 1 (XLSX format, uploaded 15/07/2020)

Explanation for dataset of Experiment 1 (RTF format, uploaded 13/07/2020)

Data collected for Experiment 2 (XLSX format, uploaded 13/07/2020)

Explanation for dataset of Experiment 2 (RTF format, uploaded 13/07/2020)

 Extended data

Open Science Framework: How subtle linguistic cues prevent unethical behaviors, https://doi.org/10.17605/OSF.IO/AXDHV ( Guo et al., 2020).

This project contains the following extended data:

Protocol for the pilot study conducted for Experiment 2 (PDF format, uploaded 31/07/2019)

Data collected for the pilot study conducted for Experiment 2 (XLSX format, uploaded 23/07/2019)

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).

 Bandura A : Moral disengagement in the perpetration of inhumanities. Pers Soc Psychol Rev. 1999;3(3):193–209. 15661671 10.1207/s15327957pspr0303_3 Bem DJ : Feeling the future: experimental evidence for anomalous retroactive influences on cognition and affect. J Pers Soc Psychol. 2011;100(3):407–25. 21280961 10.1037/a0021524 Blasi A : Moral identity: Its role in moral functioning. Morality, Moral Behavior, and Moral Development. 1984;128–139. Reference Source Bryan CJ Adams GS Monin B : When cheating would make you a cheater: implicating the self prevents unethical behavior. J Exp Psychol Gen. 2013;142(4):1001–5. 23127418 10.1037/a0030655 Bryan CJ Master A Walton GM : "Helping" versus "being a helper": invoking the self to increase helping in young children. Child Dev. 2014;85(5):1836–1842. 24779480 10.1111/cdev.12244 Bryan CJ Walton GM Rogers T : Motivating voter turnout by invoking the self. Proc Natl Acad Sci U S A. 2011;108(31):12653–12656. 21768362 10.1073/pnas.1103343108 3150938 Chandler J Mueller P Paolacci G : Nonnaïveté among Amazon Mechanical Turk workers: consequences and solutions for behavioral researchers. Behav Res Methods. 2014;46(1):112–130. 23835650 10.3758/s13428-013-0365-7 Cohen J : Eta-squared and partial eta-squared in fixed factor ANOVA designs. Educ Psychol Meas. 1973;33(1):107–112. 10.1177/001316447303300111 Cohen TR Morse L : Moral character: What it is and what it does. Res Organ Behav. 2014;34:43–61. 10.1016/j.riob.2014.08.003 Cojuharenco I Shteynberg G Gelfand M : Self-construal and unethical behavior. J Bus Ethics. 2012;109(4):447–461. 10.1007/s10551-011-1139-8 Faul F Erdfelder E Buchner A : Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav Res Methods. 2009;41(4):1149–1160. 19897823 10.3758/BRM.41.4.1149 Fausey CM Boroditsky L : Subtle linguistic cues influence perceived blame and financial liability. Psychon Bull Rev. 2010;17(5):644–650. 21037161 10.3758/PBR.17.5.644 Folk CL Leber AB Egeth HE : Made you blink! Contingent attentional capture produces a spatial blink. Percept Psychophys. 2002;64(5):741–753. 12201333 10.3758/bf03194741 Folk CL Remington RW Johnston JC : Involuntary covert orienting is contingent on attentional control settings. J Exp Psychol Hum Percept Perform. 1992;18(4):1030–44. 1431742 10.1037//0096-1523.18.4.1030 Foster-Hanson E Cimpian A Leshin RA : Asking Children to "Be Helpers" Can Backfire After Setbacks. Child Dev. 2020;91(1):236–248. 30230532 10.1111/cdev.13147 6424657 Gelman SA Heyman GD : Carrot-eaters and creature-believers: The effects of lexicalization on children's inferences about social categories. Psychol Sci. 1999;10(6):489–493. 10.1111/1467-9280.00194 Gelman SA Hollander M Star J : The role of language in the construction of kinds. Psychol Learn Motiv. 2000;39:201–263. 10.1016/S0079-7421(00)80035-3 Gelman SA O'Reilly AW : Children's inductive inferences within superordinate categories: the role of language and category structure. Child Dev. 1988;59(4):876–887. 10.2307/1130255 Gino F Mogilner C : Time, money, and morality. Psychol Sci. 2014;25(2):414–421. 10.1177/0956797613506438 Gino F Pierce L : The abundance effect: Unethical behavior in the presence of wealth. Organ Behav Hum Decis Process. 2009;109(2):142–155. 10.1016/j.obhdp.2009.03.003 Grym J Veronica L : To cheat or not to cheat? The effect of a moral reminder on cheating. Nordic Journal of Business. 2016. Reference Source Guo W Liu H Yang J : How subtle linguistic cues prevent unethical behaviors.2020. http://www.doi.org/10.17605/OSF.IO/AXDHV Jones E Nisbett R : The actor and the observer: Divergent perceptions of the causes of behavior. Attribution: Perceiving the Causes of Behavior. 1972;79–94. Reference Source Kouchaki M Smith-Crowe K Brief A : Seeing green: Mere exposure to money triggers a business decision frame and unethical outcomes. Organ Behav Hum Decis Process. 2013;121(1):53–61. Reference Source Kubovy M Psotka J : The predominance of seven and the apparent spontaneity of numerical choices. J Exp Psychol Hum Percept Perform. 1976;2(2):291–294. Reference Source Ma L Yan X Qiao W : A quasi-Poisson approach on modeling accident hazard index for urban road segments. Discrete Dyn Nat Soc. 2014;2014(3):8. 10.1155/2014/489052 Markus H Wurf E : The dynamic self-concept: A social psychological perspective. Annu Rev Psychol. 1987;38(1):299–337. 10.1146/annurev.ps.38.020187.001503 Mazar N Amir O Ariely D : The dishonesty of honest people: A theory of self-concept maintenance. J Mark Res. 2008;45(6):633–644. 10.1509/jmkr.45.6.633 Mogilner C Aaker J : “The time vs. money effect”: Shifting product attitudes and decisions through personal connection. J Consum Res. 2009;36(2):277–291. 10.1086/597161 Monin B Jordan A : The dynamic moral self: A social psychological perspective. Personality, Identity, and Character: Explorations in Moral Psychology. 2009;7:341–354. 10.1080/03057240.2011.541998 Nitta H Tomita H Zhang Y : Disgust and the rubber hand illusion: a registered replication report of Jalal, Krishnakumar, and Ramachandran (2015). Cogn Res Princ Implic. 2018;3(1):15. 29780877 10.1186/s41235-018-0101-z 5954052 Oppenheimer DM Meyvis T Davidenko N : Instructional manipulation checks: Detecting satisficing to increase statistical power. J Exp Soc Psychol. 2009;45(4):867–872. 10.1016/j.jesp.2009.03.009 Orben A Lakens D : Crud (re)defined. Adv Methods Pract Psychol Sci. 2020;3(2):238–247. 10.1177/2515245920917961 Ross L : The intuitive psychologist and his shortcomings: Distortions in the attribution process.In: Adv Exp Soc Psychol. 1977;10:173–220. 10.1016/S0065-2601(08)60357-3 Sasaki K Yamada Y : Crowdsourcing visual perception experiments: a case of contrast threshold. PeerJ. 2019;7:e8339. 31875164 10.7717/peerj.8339 6927342 Savir T Gamliel E : To be an honest person or not to be a cheater: Replicating the effect of messages relating to the self on unethical behaviour. Int J Psychol. 2019;54(5):650–658. 30079556 10.1002/ijop.12519 Simons DJ Shoda Y Lindsay DS : Constraints on generality (COG): A proposed addition to all empirical papers. Perspect Psychol Sci. 2017;12(6):1123–1128. 28853993 10.1177/1745691617708630 Steele CM : The psychology of self-affirmation: Sustaining the integrity of the self.In: Adv Exp Soc Psychol. 1988;21:261–302. 10.1016/S0065-2601(08)60229-4 Vohs KD Mead NL Goode MR : The psychological consequences of money. Science. 2006;314(5802):1154–1156. 17110581 10.1126/science.1132491 Walton GM Banaji MR : Being what you say: The effect of essentialist linguistic labels on preferences. Soc Cogn. 2004;22(2):193–213. 10.1521/soco.22.2.193.35463 10.5256/f1000research.28223.r69709 Reviewer response for version 1 Asano Michiko 1 Referee https://orcid.org/0000-0001-8362-9311 Department of Psychology, College of Contemporary Psychology, Rikkyo University, Saitama, Japan

Competing interests: No competing interests were disclosed.

 26 10 2020 Copyright: © 2020 Asano M 2020 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve

This study nicely replicates Experiment 3 of Bryan et al. (2013), which showed that a self-relevant noun of moral reminder (i.e., “Please don’t be a cheater”) curbed unethical behaviours more effectively than a verb (i.e., “Please don’t cheat”) did, in an online survey environment with Japanese speakers. The results of Experiment 2 of the current study, which addressed potential attentional effects related to translation, strengthened their claim.

I only have minor suggestions, as described below, and approve this manuscript for publication.

Minor suggestions:

The last sentence of the second paragraph of Introduction “However, one study revealed that highly constructed self-identities are associated with more unethical behaviors (Cojuharenco et al., 2012).”:

It is not very clear to me what point the authors are trying to make by citing this study here.

The fourth paragraph of Discussion: “Inconsistent with the original experiment and Experiment 1, in Experiment 2, there was no significant difference between the “cheating” condition and the chance level. This discrepancy is not important because it is not relevant to the claimed hypothesis... Note that the present study showed that the effect size of the reminder itself is very small, so it is clear that no promising hypothesis exists at this point as a crud factor (Orben & Lakens, 2020).”:

I understand that the lack of a significant difference between the “cheating” condition and the chance level is not relevant to the hypothesis of Experiment 2, and the effect size of the reminder itself is small. However, I still feel the authors’ claim here (“This discrepancy is not important”) is too strong and perhaps needs to be toned down because it is the prerequisite for this line of study that cheating behaviours occur in the “cheating” condition.

Is the work clearly and accurately presented and does it cite the current literature?

Yes

If applicable, is the statistical analysis and its interpretation appropriate?

Yes

Are the data able to test the authors’ proposed hypotheses by satisfying the approved outcome-neutral conditions?

Yes

Did the authors adhere precisely to the registered experimental procedures? If not, has an explanation been provided regarding any change?

Yes

Are all the source data underlying the results available to ensure full reproducibility?

Yes

Is the study design appropriate and is the work technically sound?

Yes

Are the conclusions drawn adequately supported by the results?

Yes

Are the introduction, rationale and stated hypotheses the same as the approved Stage 1 submission?

Yes

Are any unregistered post hoc analyses added by the authors justified, methodologically sound and informative?

Not applicable

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

NA

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Yamada Yuki Kyushu University, Japan

Competing interests: No competing interests were disclosed.

 31 10 2020

Thank you for reviewing our registered report again! We have responded to each of your comments as follows.

Minor suggestions:

1. The last sentence of the second paragraph of Introduction “However, one study revealed that highly constructed self-identities are associated with more unethical behaviors (Cojuharenco et al., 2012).”:

It is not very clear to me what point the authors are trying to make by citing this study here.

Reply: In Cojuharenco et al. (2012), high levels of self-identities construction elicited more unethical behaviors. This finding was inconsistent with the conclusion of the experiment in Bryan et al. (2013) that we intended to replicate, thus here we cited this research with different conclusions in order to place more emphasis on the importance of replicating the results of Bryan's experiment. However, to avoid excessive and unnecessary changes of the introduction (the stage 1 manuscript), we decided only to explain it here, but do not add extra explanations to the manuscript.

2. The fourth paragraph of Discussion: “Inconsistent with the original experiment and Experiment 1, in Experiment 2, there was no significant difference between the “cheating” condition and the chance level. This discrepancy is not important because it is not relevant to the claimed hypothesis... Note that the present study showed that the effect size of the reminder itself is very small, so it is clear that no promising hypothesis exists at this point as a crud factor (Orben & Lakens, 2020).”:

I understand that the lack of a significant difference between the “cheating” condition and the chance level is not relevant to the hypothesis of Experiment 2, and the effect size of the reminder itself is small. However, I still feel the authors’ claim here (“This discrepancy is not important”) is too strong and perhaps needs to be toned down because it is the prerequisite for this line of study that cheating behaviours occur in the “cheating” condition.

Reply: Thank you for your constructive suggestion. As you pointed out, we have to admit that we discussed the “cheating” condition in Experiment 2 with a too strong tone. It is meaningful to explore the effectiveness of “don’t cheat” reminder in different circumstances, thus here we cannot assert that this discrepancy is not important. We toned down the discussion and modified the corresponding part of the manuscript as follows:

This inconsistency in the results between the present experiments does not affect the interpretation of the main hypothesis we tried to test here. Yet, whether this “cheating” reminder does or does not affect behavior is inconsistent across studies as well, and hence it will be interesting to further investigate what factors define the difference.

 10.5256/f1000research.28223.r69708 Reviewer response for version 1 Cervera-Torres Sergio 1 2 Referee https://orcid.org/0000-0003-1689-4654 Leibniz-Institute für Wissensmedien (IWM), Tübingen, Germany Salumedia Labs SLU, Sevilla, Spain

Competing interests: No competing interests were disclosed.

 22 9 2020 Copyright: © 2020 Cervera-Torres S 2020 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve

The study nicely replicates, in a Japanese sample, the findings by Bryan et al. (2013) about how subtle linguistic cues prevent unethical behaviour based on a coin-tossing task. 

As a reviewer of Stage 1 of this study, I must say that all my major concerns have been successfully amended. I only have minor suggestions:

- Introduction: I still think that the storyline could be improved a bit. I find unclear whether the statement "In general, we believe that highlighting a self-identity word..." at the beginning of the second paragraph is an especial belief of the authors or an argument. Maybe it could suit better at the end of the paragraph "According to Bryan et al. (2011)...".

-   Experiment 1. I would use Inclusion (or exclusion) criteria as a heading instead of "outlier extraction" and move the section before the "data analyses" section. 

-   Experiment 2. You use three different expressions to name the new condition: "cheating" with task condition, "cheating with attention task" condition, and "don't cheat" condition with a recognition task. I suggest being consistent through the paper and use, "cheating with attention task". In the procedure, I would use different wording for "..we deleted the original "cheater condition and added..." such as "...we omitted the original cheater condition and included instead..."

- Discussion. In the second paragraph, you state "we aimed to test whether the effect of the reminder was enhanced by participants' excessive attention to the unfamiliar expression [cheater] in the Japanese context". This is not exactly what you did.

As you state in the introduction of Experiment 2  "We aimed to test whether lower levels of attention to the instruction in the “cheating” condition reduced the effectiveness of preventing dishonest behaviours in our Experiment 1".

The difference is subtle but important.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

Yes

Are the data able to test the authors’ proposed hypotheses by satisfying the approved outcome-neutral conditions?

Yes

Did the authors adhere precisely to the registered experimental procedures? If not, has an explanation been provided regarding any change?

Yes

Are all the source data underlying the results available to ensure full reproducibility?

Yes

Is the study design appropriate and is the work technically sound?

Yes

Are the conclusions drawn adequately supported by the results?

Yes

Are the introduction, rationale and stated hypotheses the same as the approved Stage 1 submission?

Yes

Are any unregistered post hoc analyses added by the authors justified, methodologically sound and informative?

Not applicable

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

Cognitive & e-Health Psychology.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Yamada Yuki Kyushu University, Japan

Competing interests: No competing interests were disclosed.

 31 10 2020

Thank you for reviewing our registered report again! We have responded to each of your comments as follows.

1.- Introduction: I still think that the storyline could be improved a bit. I find unclear whether the statement "In general, we believe that highlighting a self-identity word..." at the beginning of the second paragraph is an especial belief of the authors or an argument. Maybe it could suit

better at the end of the paragraph "According to Bryan et al. (2011)...".

Reply: In the second paragraph, we wanted to state how the self-identity influences ethical behaviors through the moral core of a person. However, this topic was also discussed in Bryan et al. (2013) and it makes sense to put it at the end of the sixth paragraph of the introduction as your suggestion. As this is a revision of a part of the manuscript of stage 1, please let us add a footnote before the “Data Availability” part to explain this modification does not change the logic or the hypothesis that we registered in our stage 1 manuscript.

2.- Experiment 1. I would use Inclusion (or exclusion) criteria as a heading instead of "outlier extraction" and move the section before the "data analyses" section.

Reply: Thank you for your suggestion. We have replaced the “Outlier extraction” with “Exclusion criteria” and set the “Exclusion criteria” section before the “Data analyses” section.

3.- Experiment 2. You use three different expressions to name the new condition: "cheating" with task condition, "cheating with attention task" condition, and "don't cheat" condition with a recognition task. I suggest being consistent through the paper and use, "cheating with attention task". In the procedure, I would use different wording for "..we deleted the original "cheater condition and added..." such as "...we omitted the original cheater condition and included instead..."

Reply: According to your suggestion, the expression of the new condition has been unified as “cheating” with attention task. Besides, we also modified the corresponding part in the manuscript.

4.- Discussion. In the second paragraph, you state "we aimed to test whether the effect of the reminder was enhanced by participants' excessive attention to the unfamiliar expression [cheater] in the Japanese context". This is not exactly what you did.

As you state in the introduction of Experiment 2 "We aimed to test whether lower levels of attention to the instruction in the “cheating” condition reduced the effectiveness of preventing dishonest behaviors in our Experiment 1". The difference is subtle but important.

Reply: The purpose of Experiment 2 was to explore the role of the attention factor on unethical behavior in different samples from Experiment 1. As you said, we should revise the expression to describe our work more clearly, so we have rephrased the description of the discussion as follows:

We aimed to test whether the lower levels of attention caused by a familiar instruction (“Don’t cheat”) in the “cheating” condition reduced the effectiveness of preventing unethical behavior in our Experiment 1.