Response

In their article, Thunström et al. present the results of a trust experiment that compares Christians, Muslims, and nonbelievers in a variety of ways. The study is innovative, provides some useful insights, and certainly has merit––but also has several issues that should be scrutinized. To summarize briefly, Thunström et al. used quota sampling to select a nonrandom online panel of participants for three religious/belief groups (roughly 200 in each of the three Proposer groups––Christians, Muslims, and nonbelievers; and 200 in each of the Responder groups for just under 1,200 participants total). Proposers were given an amount of money that they could send to a Responder (whose religiosity/nonbelief is made clear). Any money that was sent to the Responder was automatically tripled, and the Responder could decide how much money would be split (i.e., reciprocated) with the Proposer (whose religiosity/nonbelief is also made clear). Both Proposers and Responders were financially incentivized by being allowed to keep the monies. The “best” strategy is for the Proposer to send all their money as this maximizes payoff––but this strategy is only viable if the Proposer trusts the Responder. Thunström et al. employed dozens of uncorrected comparisons and ultimately noted that nonbeliever Responders sent back 2–9 percent less than other Responders. From this, Thunström et al.’s key finding is that nonbelievers are less trustworthy than are Christians or Muslims. In this critique, we will comment on issues in ascending order of importance, as the smaller objections lay the foundation for the larger ones.

Our first issue is that atheism and agnosticism are not religions! Atheism and agnosticism are views toward the existence of god/gods/higher powers. Perhaps, an analogy will help: Atheism is to theism as nonreligion is to religion. Social scientists need to stop conflating positions toward deities with religious affiliations. Thunström et al. compared Christians and Muslims to nonbelievers––which is a beef stew to carrots comparison (i.e., Christianity and Islam include a lot of “ingredients”; atheism is just one). The flawed understanding of what atheism and agnosticism are also introduced a serious problem for Thunström et al.’s entire study. Because people can vary in how religious they are when they have a religious affiliation, Thunström et al. attempted to capture that. But they also seemed to think that atheism and agnosticism were comparable levels of nonreligiosity, which they are not (Cragun 2016). As a result, Thunström et al. presented their participants with more information about the religious Responders and Proposers than they did about the nonbelieving ones––they indicated how devoted they are to their religious identity. Comparing highly religious Christians to agnostics is like comparing a gourmet, 5‐star beef stew to… carrots. To what extent this misalignment of information calls into question the findings in Thunström et al.’s study is not clear, but it raises serious concerns in our minds.

Our second issue is the ambiguity in what the study actually captured. Thunström et al. note that atheists are aware that they are mistrusted and are discriminated against. The authors detail how they “hid” the religious nature of the experiment by including filler items. However, the compensating effect of this is immediately undone as Thunström et al. primed all Proposers and Responders with religious information about their unknown partner for the monetary exchange. As it stands, Thunström et al. only demonstrate that nonbelieving respondents will be less trusting of strangers in some circumstances after being primed about religion. Frustratingly, this could have easily been remedied within the study design by simply adding a control condition where Proposers and Responders were primed on nonreligious information. We would then have a much clearer idea of whether nonbelievers are less likely to reciprocate at the same rate as religious groups, or whether this effect was produced by the religion prime.

Our third issue is the inability of Thunström et al. to generalize beyond their sample. Their data are not remotely representative of Christians, Muslims, or nonbelievers, and the artificial nature of the experiment has no real‐world analogues (i.e., it lacks ecological validity). Moreover, the difference observed in the “key finding,” while statistically significant, has just a small effect size that is exaggerated by expanding the y‐axes in their figures. We are certainly willing to grant that the authors may have found a small difference in reciprocity in this nonrandom, artificial study, but there are no grounds to suggest that this has any real‐world application. Had Thunström et al. limited their inference to describe a small difference in how much money nonbeliever Responders returned to the Proposers in their study, we would have been unlikely to write this response. However, Thunström et al. seem to suggest that their narrow findings are somehow broadly true of nonbelievers generally.

Our final and most important issue is the implications of Thunström et al.’s study. Their conclusion that they have evidence that atheists are less trustworthy—paired with the already widespread prejudice and discrimination against nonbelievers in the United States (Edgell, Gerteis, and Hartmann 2006; Gervais 2014)—is akin to kicking a hornet's nest. Basically, Thunström et al.’s work could easily be used to justify that atheists behave in an untrustworthy manner, thus anti‐atheist prejudice and discrimination are warranted. In a real‐world parallel, Regnerus's (2012) findings were immediately used in court cases to marginalize gay and lesbian individuals (Boccuzzi et al. 2013; Cheng and Powell 2015). To clarify, if robust methods replicated Thunström et al., we would be surprised by the finding, but would recognize the value in exploring and understanding the mechanism behind the result. However, given the list of issues we have presented, we believe that caution should be exercised, as the finding could be used to justify the existing and widespread discrimination against nonbelievers in the United States (Cragun et al. 2012; Hammer et al. 2012).

One of the frustrating aspects of reading Thunström et al.’s article were the missed opportunities with their results. For instance, Thunström et al. found that nonbelieving Proposers were more trusting than the Christian Proposers in their study. Nonbelievers sent comparable levels of money to Christians, Muslims, and nonbelievers; essentially, nonbelievers did not demonstrate an in‐group bias like the Christians did. Additionally, Table 1 in their paper indicates that nonbelieving and Muslim Responders were less likely to receive any money from Proposers. In other words, when people were presented with information that the Responder was either Muslim or a nonbeliever, the Proposer refused to send any money. Thunström et al. excluded Responders who received $.00 from Proposers, and the numbers of religious groups become noticeably lopsided in the Responder condition (N_Christian = 228 vs. N_Muslim = 163 vs. N_Nonreligious = 170). Surprisingly, despite these two findings running contrary to their key‐finding (atheists are less trustworthy), these do not receive meaningful attention in their article. While none of these findings are generalizable, the decision to emphasize a perceived negative characteristic of nonbelievers while ignoring positive characteristics drew our attention.

With the growth of the nonreligious, scholars most assuredly should begin including nonreligious and nonbelieving individuals in their studies (Smith and Cragun 2019). However, including the nonreligious in studies should be done appropriately, reflecting a clear grasp of the relevant concepts (Cragun 2016). Otherwise, scholars will continue to make flawed comparisons. While we applaud Thunström et al. for their efforts to advance our understanding of trust and religion, we urge caution in considering their findings. Given the methods, Thunström et al.’s study is merely preliminary and future efforts in this area need to address the serious shortcomings.

中文翻译：

---

我们不认为这意味着他们的想法：对Thunström等人的回应

回复

在他们的文章中，Thunström等人。展示了一项信任实验的结果，该信任实验以多种方式比较了基督徒，穆斯林和非信徒。这项研究具有创新性，提供了一些有用的见识，并且当然具有优点-但也有一些问题需要仔细研究。简而言之，Thunström等人。使用配额抽样为三个宗教/信仰团体（参加三个提议者组中的每个团体-基督徒，穆斯林和非信徒大约200个；每个响应者组中的大约200个；每个响应者组中的200个，总计不到1200名参与者）选择了一个非随机的在线参与者小组）。向提案人提供了可以发送给响应者的金额（已明确其宗教信仰/不信奉）。发送给响应者的所有资金都会自动增加三倍，响应者可以决定分配多少资金（即，与提议者进行互惠互利（也明确表示了自己的宗教信仰/不信教）。提议者和响应者都被允许保留款项，从而在财务上得到了激励。“最佳”策略是使投标人将其所有资金汇出，以最大程度地获得收益–但是，只有在投标人信任响应者的情况下，该策略才可行。Thunström等。进行了数十次未经更正的比较，最终注意到，非信徒的回复者回传了2–9％少于其他响应者。由此看来，Thunström等人的主要发现是，非信徒比基督徒或穆斯林不值得信赖。在此批评中，我们将按重要性从高到低的顺序对问题进行评论，因为较小的异议为较大的异议奠定了基础。

我们的第一个问题是无神论和不可知论者不是宗教！无神论和不可知论是对神/神/更高能力存在的看法。也许，类比会有所帮助：无神论对有神论就像无宗教对宗教一样。社会科学家需要停止将对神的立场与宗教信仰混为一谈。Thunström等。将基督徒和穆斯林与非信徒进行了比较-这是炖牛肉与胡萝卜的比较（即基督教和伊斯兰教包括很多“成分”；无神论只是其中之一）。对什么无神论和不可知论的误解也为Thunström等人的整个研究提出了一个严重的问题。Thunström等人认为，当人们有宗教信仰时，他们的宗教信仰会有所不同。试图抓住这一点。2016）。结果，Thunström等人。向他们的参与者提供了有关宗教响应者和提议者的更多信息，而不是他们向非信奉者和提议者提供的信息–他们指出了他们对宗教身份的投入。将信奉宗教的基督徒与不可知论者进行比较，就像将美食家的五星级炖牛肉与…胡萝卜进行比较。信息的这种错位在多大程度上使人们怀疑Thunström等人的研究结果，但是却引起了我们的严重关注。

我们的第二个问题是研究实际捕获的内容不明确。Thunström等。请注意，无神论者意识到他们是不信任和受到歧视的。作者详细介绍了如何通过添加填充项来“隐藏”实验的宗教性质。然而，这种补偿作用随Thunström等人的研究而被撤销。向所有提议者和响应者提供有关其未知伙伴进行货币兑换的宗教信息。就目前而言，Thunström等人。仅表明不信的受访者在某些情况下对陌生人的信任会降低关于宗教的环境。令人沮丧的是，在研究设计中，只需添加一个控制条件即可轻松解决这一问题，在该控制条件中，对提议者和响应者的非宗教信息进行了准备。这样，我们就可以更清楚地了解非信徒是否不太可能以与宗教团体相同的速度来往，或者这种影响是否是由宗教总理产生的。

我们的第三个问题是Thunström等人的无能。推广到他们的样本之外。他们的数据不能代表基督徒，穆斯林或非信徒，而且实验的人为性质没有现实世界的类似物（即缺乏生态有效性）。此外，在“关键发现”中观察到的差异虽然具有统计学意义，但其影响范围很小，而通过扩大图中的y轴会扩大该影响范围。我们当然愿意承认作者在这项非随机的，人工的研究中可能发现了互惠性的细微差别，但是没有理由表明它可以在现实世界中使用。有Thunström等。限制他们的推论，以描述非信徒回应者在研究中退还给提案人的金额之间的微小差异，我们将不太可能写下此回复。然而，Thunström等。似乎暗示他们的狭窄发现在某种程度上对于非信徒总体上是正确的。

我们最后也是最重要的问题是Thunström等人的研究的意义。他们得出的结论是，他们有证据表明无神论者不那么值得信赖，再加上在美国已经普遍存在的对非信徒的偏见和歧视（Edgell，Gerteis和Hartmann 2006； Gervais 2014），类似于踢大黄蜂的巢穴。基本上，Thunström等人的工作可以很容易地用来证明无神论者的举止是不可信任的，因此，反无神论者的偏见和歧视是必要的。在现实世界中，Regnerus（2012）的发现被立即用于法庭案件中以将同性恋者边缘化（Boccuzzi等，2013； Cheng和Powell2015）。要澄清的是，如果鲁棒的方法重复了Thunström等人的研究，我们会对该发现感到惊讶，但会认识到探索和理解结果背后机制的价值。但是，考虑到我们提出的问题清单，我们认为应谨慎行事，因为该发现可用于证明美国对非信徒的现有和普遍歧视是正当的（Cragun等，2012； Hammer等，2012）。）。

阅读Thunström等人的文章令人沮丧的方面之一是其结果错失了机会。例如，Thunström等。发现在研究中，不信教的提议者比基督徒的提议者更信任。非信徒向基督徒，穆斯林和非信徒寄出了相当数量的钱；从本质上讲，非信徒没有表现出像基督徒那样的群体内偏见。此外，表1中的文件指出，nonbelieving和穆斯林响应者较少接受任何来自提议者钱。换句话说，当向人们提供响应者是穆斯林还是非信徒的信息时，提议者拒绝汇款。Thunström等。谁从提议者收到$ .00排斥反应者，和宗教团体的数目在响应状态明显变得渐行渐远（ñ_基督教= 228与ň_穆斯林= 163个与ň_不信教= 170）。出乎意料的是，尽管这两个发现与他们的发现背道而驰（无神论者的信任度较低），但它们在本文中并未得到有意义的关注。尽管这些发现都没有一个可以推广的结论，但强调非信徒感知到的负面特征却忽略其正面特征的决定引起了我们的注意。

随着非宗教人士的增长，学者们最确定的应该开始将非宗教人士和非信仰人士纳入研究范围（Smith and Cragun 2019）。但是，应该适当地将非宗教信仰纳入研究范围，这反映了对相关概念的清晰理解（Cragun 2016）。否则，学者们将继续进行有缺陷的比较。尽管我们为Thunström等人鼓掌。对于他们为增进我们对信任和宗教的理解所做的努力，我们敦促谨慎考虑他们的发现。有了这些方法，Thunström等人的研究仅仅是初步的，未来在这一领域的工作需要解决严重的缺陷。