The Seductive Allure of Neuroscience Explanations

Deena Skolnick Weisberg, Frank C. Keil, Joshua Goodstein, Elizabeth Rawson, and Jeremy R. Gray

Abstract & Explanations of psychological phenomena seem to gener- vs. with neuroscience) design. Crucially, the neuroscience in- ate more public interest when they contain neuroscientific formation was irrelevant to the logic of the explanation, as information. Even irrelevant neuroscience information in an confirmed by the expert subjects. Subjects in all three groups explanation of a psychological phenomenon may interfere with judged good explanations as more satisfying than bad ones. people’s abilities to critically consider the underlying logic of But subjects in the two nonexpert groups additionally judged this explanation. We tested this hypothesis by giving naı¨ve that explanations with logically irrelevant neuroscience infor- adults, students in a neuroscience course, and neuroscience ex- mation were more satisfying than explanations without. The perts brief descriptions of psychological phenomena followed neuroscience information had a particularly striking effect on by one of four types of explanation, according to a 2 (good nonexperts’ judgments of bad explanations, masking other- explanation vs. bad explanation) Â 2 (without neuroscience wise salient problems in these explanations. &

INTRODUCTION (Lombrozo & Carey, 2006; Kelemen, 1999). People also Although it is hardly mysterious that members of the tend to rate longer explanations as more similar to ex- public should find psychological research fascinating, perts’ explanations (Kikas, 2003), fail to recognize circu- this fascination seems particularly acute for findings that larity (Rips, 2002), and are quite unaware of the limits were obtained using a neuropsychological measure. In- of their own abilities to explain a variety of phenomena deed, one can hardly open a newspaper’s science sec- (Rozenblit & Keil, 2002). In general, people often believe tion without seeing a report on a neuroscience discovery explanations because they find them intuitively satisfying, or on a new application of neuroscience findings to eco- not because they are accurate (Trout, 2002). nomics, politics, or law. Research on nonneural cogni- In line with this body of research, we propose that tive does not seem to pique the public’s people often find neuroscience information alluring be- interest in the same way, even though the two fields are cause it interferes with their abilities to judge the quality concerned with similar questions. of the psychological explanations that contain this infor- The current study investigates one possible reason why mation. The presence of neuroscience information may members of the public find so be seen as a strong marker of a good explanation, re- particularly alluring. To do so, we rely on one of the gardless of the actual status of that information within functions of neuroscience information in the field of psy- the explanation. That is, something about seeing neu- chology: providing explanations. Because articles in both roscience information may encourage people to believe the popular press and scientific journals often focus on they have received a scientific explanation when they how neuroscientific findings can help to explain human have not. People may therefore uncritically accept any behavior, people’s fascination with cognitive neuroscience explanation containing neuroscience information, even can be redescribed as people’s fascination with explana- in cases when the neuroscience information is irrelevant tions involving a neuropsychological component. to the logic of the explanation. However, previous research has shown that people To test this hypothesis, we examined people’s judg- have difficulty reasoning about explanations (for reviews, ments of explanations that either do or do not contain see Keil, 2006; Lombrozo, 2006). For instance, people can neuroscience information, but that otherwise do not dif- be swayed by teleological explanations when these are not fer in content or logic. All three studies reported here warranted, as in cases where a nonteleological process, used a 2 (explanation type: good vs. bad) Â 2(neurosci- such as natural selection or erosion, is actually implicated ence: without vs. with) design. This allowed us to see both people’s baseline abilities to distinguish good psy- chological explanations from bad psychological explana- Yale University tions as well as any influence of neuroscience information

D 2008 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 20:3, pp. 470–477 on this ability. If logically irrelevant neuroscience infor- circular restatements of the phenomenon, hence, not mation affects people’s judgments of explanations, this explanatory (see Table 1 for a sample item). would suggest that people’s fascination with neuropsy- For the With Neuroscience conditions, we added neu- chological explanations may stem from an inability or roscience information to the good and bad explanations unwillingness to critically consider the role that neurosci- from the Without Neuroscience conditions. The added ence information plays in these explanations. neuroscience information had three important features: (1) It always specified that the area of activation seen in the study was an area already known to be involved in tasks of this type, circumventing the interpretation that EXPERIMENT 1 the neuroscience information added value to the expla- Methods nation by localizing the phenomenon. (2) It was always Subjects identical or nearly identical in the good explanation and the bad explanation for a given phenomenon. Any There were 81 participants in the study (42 women, general effect of neuroscience information on judgment 37 men, 2 unreported; mean age = 20.1 years, SD = should thus be seen equally for good explanations and 4.2 years, range = 18–48 years, based on 71 reported bad explanations. Additionally, any differences that may ages). We randomly assigned 40 subjects to the With- occur between the good explanation and bad explana- out Neuroscience condition and 41 to the With Neuro- tion conditions would be highly unlikely to be due to science condition. Subjects thus saw explanations that any details of the neuroscience information itself. (3) either always did or always did not contain neuroscience Most importantly, in no case did the neuroscience information. We used this between-subjects design to information alter the underlying logic of the explanation prevent subjects from directly comparing explanations itself. This allowed us to test the effect of neuroscience that did and did not contain neuroscience, providing a information on the task of evaluating explanations, stronger test of our hypothesis. independent of any value added by such information.1 Before the study began, three experienced cognitive neuroscientists confirmed that the neuroscience infor- Materials mation did not add value to the explanations. We wrote descriptions of 18 psychological phenomena (e.g., mutual exclusivity, attentional blink) that were meant Procedure to be accessible to a reader untrained in psychology or neuroscience. For each of these items, we created two Subjects were told that they would be rating explana- types of explanations, good and bad, neither of which tions of scientific phenomena, that the studies they contained neuroscience. The good explanations in most would read about were considered solid, replicable re- cases were the genuine explanations that the researchers search, and that the explanations they would read were gave for each phenomenon. The bad explanations were not necessarily the real explanations for the phenomena.

Table 1. Sample Item

Good Explanation Bad Explanation Without Neuroscience The researchers claim that this ‘‘curse’’ happens The researchers claim that this ‘‘curse’’ happens because subjects have trouble switching their because subjects make more mistakes when point of view to consider what someone else they have to judge the knowledge of others. might know, mistakenly projecting their own People are much better at judging what they knowledge onto others. themselves know. With Neuroscience Brain scans indicate that this ‘‘curse’’ happens Brain scans indicate that this ‘‘curse’’ happens because of the frontal lobe brain circuitry because of the frontal lobe brain circuitry known to be involved in self-knowledge. known to be involved in self-knowledge. Subjects have trouble switching their point of Subjects make more mistakes when they have view to consider what someone else might to judge the knowledge of others. People are know, mistakenly projecting their own much better at judging what they themselves knowledge onto others. know.

Researchers created a list of facts that about 50% of people knew. Subjects in this experiment read the list of facts and had to say which ones they knew. They then had to judge what percentage of other people would know those facts. Researchers found that the subjects responded differently about other people’s knowledge of a fact when the subjects themselves knew that fact. If the subjects did know a fact, they said that an inaccurately large percentage of others would know it, too. For example, if a subject already knew that Hartford was the capital of Connecticut, that subject might say that 80% of people would know this, even though the correct answer is 50%. The researchers call this finding ‘‘the curse of knowledge.’’ The neuroscience information is highlighted here, but subjects did not see such marking.

Weisberg et al. 471 For each of the 18 stimuli, subjects read a one-paragraph We also found a significant interaction between expla- description of the phenomenon followed by an expla- nation type and neuroscience information [F(1, 79) = nation of that phenomenon. They rated how satisfying 18.8, p < .01]. Post hoc tests revealed that although the they found the explanation on a 7-point scale from À3 ratings for good explanations were not different without (very unsatisfying)to+3(very satisfying) with 0 as the neuroscience (M = 0.86, SE = 0.11) than with neuro- neutral midpoint. science (M = 0.90, SE = 0.16), ratings for bad expla- nations were significantly lower for explanations without neuroscience (M = À0.73, SE = 0.14) than explanations Results and Discussion with neuroscience (M = 0.16, SE = 0.16). Note that this Preliminary analyses revealed no differences in perfor- difference is not due to a ceiling effect; ratings of good mance based on sex or level of education, so ratings explanations are still significantly below the top of the were collapsed across these variables for the analyses. scale [t(80) = À21.38, p < .01]. This interaction indi- Additionally, subjects tended to respond similarly to all cates that it is not the mere presence of verbiage about 18 items (Cronbach’s a = .79); the set of items had neuroscience that encourages people to think more acceptable psychometric reliability as a measure of the favorably of an explanation. Rather, neuroscience infor- construct of interest. mation seems to have the specific effect of making bad Our primary goal in this study was to discover what explanations look significantly more satisfying than they effect, if any, the addition of neuroscience information would without neuroscience. would have on subjects’ ratings of how satisfying they This puzzling differential effect of neuroscience in- found good and bad explanations. We analyzed the rat- formation on the bad explanations may occur because ings using a 2 (good explanation vs. bad explanation) Â participants gave the explanations a more generous in- 2 (without neuroscience vs. with neuroscience) repeated terpretation than we had expected. Our instructions measures analysis of variance (ANOVA; see Figure 1). encouraged participants to think of the explanations as There was a significant main effect of explanation type being provided by knowledgeable researchers, so they [F(1, 79) = 144.8, p < .01], showing that good expla- may have considered the explanations less critically than nations (M = 0.88, SE = 0.10) are rated as significantly we would have liked. If participants were using some- more satisfying than bad explanations (M = À0.28, SE = what relaxed standards of judgment, then a group of 0.12). That is, subjects were accurate in their assess- subjects that is specifically trained to be more critical of ments of explanations in general, finding good expla- judging explanations should not fall prey to the effect nations to be better than bad ones. of added neuroscience information, or at least not as There was also a significant main effect of neuroscience much. [F(1, 79) = 6.5, p < .05]. Explanations with neuroscience Experiment 2 addresses this issue by testing a group information (M =0.53,SE = 0.13) were rated as sig- of subjects trained to be critical in their judgments: stu- nificantly more satisfying than explanations that did not dents in an intermediate-level cognitive neuroscience include neuroscience information (M =0.06,SE =0.13). class. These students were receiving instruction on the Adding irrelevant neuroscience information thus some- basic logic of cognitive neuroscience experiments and on how impairs people’s baseline ability to make judgments the types of information that are relevant to drawing about explanations. conclusions from neuroscience studies. We predicted that

Figure 1. Novice group. Mean ratings of how satisfying subjects found the explanations. Error bars indicate standard error of the mean.

472 Journal of Cognitive Neuroscience Volume 20, Number 3 this instruction, together with their classroom experience As with the novices in Experiment 1, we tested whether of carefully analyzing neuroscience experiments, would the addition of neuroscience information affects judg- eliminate or dampen the impact of the extraneous neuro- ments of good and bad explanations. For the students in science information. this study, we additionally tested the effect of training on evaluations of neuroscience explanations. We thus ana- lyzed the students’ ratings of explanatory satisfaction using a 2 (good explanation vs. bad explanation) 2 EXPERIMENT 2 Â (without neuroscience vs. with neuroscience) Â 2 (pre- Methods class test vs. postclass test) repeated measures ANOVA Subjects and Procedure (see Figure 2). We found a significant main effect of explanation type Twenty-two students (10 women; mean age = 20.7 years, [F(1, 21) = 50.9, p < .01], confirming that the students SD = 2.6 years, range = 18–31 years) were recruited from judged good explanations (M = 0.37, SE = 0.14) to an introductory cognitive neuroscience class and received be more satisfying than bad explanations (M = À0.43, no compensation for their participation. They were in- SE = 0.19). formed that although participation was required for the Although Experiment 1 found a strong effect of the course, the results of the experiment would have no im- presence of neuroscience information in explanations, pact on their class performance and would not be known we had hypothesized that students in a neuroscience by their professor until after their grades had been posted. course, who were learning to be critical consumers of They were additionally allowed to choose whether their neuroscience information, would not show this effect. data could be used in the published research study, and However, the data failed to confirm this hypothesis; all students elected to have their data included. there was a significant main effect of neuroscience Subjects were tested both at the beginning of the se- [F(1, 21) = 47.1, p < .01]. Students, like novices, judged mester and at the end of the semester, prior to the final that explanations with neuroscience information (M = exam. 0.43, SE = 0.17) were more satisfying than those without The stimuli and task were identical to Experiment 1, neuroscience information (M = À0.49, SE = 0.16). with one exception: Both main variables of explanation There was additionally an interaction effect between type and presence of neuroscience were within-subject explanation type and presence of neuroscience [F(1, due to the small number of participants. 21) = 8.7, p < .01], as in Experiment 1. Post hoc analy- ses indicate that this interaction happens for the same reason as in Experiment 1: Ratings of bad explanations Results and Discussion increased reliably more with the addition of neurosci- Preliminary analyses showed no differences in perfor- ence than did good explanations. Unlike the novices, the mance based on class year, so this variable is not con- students judged that both good explanations and bad sidered in the main analyses. There was one significant explanations were significantly more satisfying when interaction with sex that is discussed shortly. Responses they contained neuroscience, but the bad explanations to the items were again acceptably consistent (Cronbach’s were judged to have improved more dramatically, based a = .74). on a comparison of the differences in ratings between

Figure 2. Student group. Mean ratings of how satisfying subjects found the explanations. Error bars indicate standard error of the mean.

Weisberg et al. 473 explanations with and without neuroscience [t(21) = in the Without Neuroscience condition and 20 subjects 2.98, p < .01]. Specialized training thus did not discour- in the With Neuroscience condition. age the students from judging that irrelevant neuro- We defined our expert population as individuals who science information somehow contributed positively to are about to pursue, are currently pursuing, or have both types of explanation. completed advanced degrees in cognitive neuroscience, Additionally, our analyses found no main effect of , or strongly related fields. Our par- time, showing that classroom training did not affect ticipant group contained 6 participants who had com- the students’ performance. Ratings before taking the pleted college, 29 who were currently in graduate school, class and after completing the class were not significant- and 13 who had completed graduate school. ly different [F(1, 21) = 0.13, p > .10], and there were The materials and procedure in this experiment were no interactions between time and explanation type [F(1, identical to Experiment 1, with the addition of four de- 21) = 0.75, p > .10] or between time and presence of mographic questions in order to confirm the expertise neuroscience [F(1, 21) = 0.0, p > .10], and there was no of our subjects. We asked whether they had ever partic- three-way interaction among these variables [F(1, 21) = ipated in a neuroscience study, designed a neuroscience 0.31, p > .10]. The only difference between the preclass study, designed a psychological study that did not neces- data and the postclass data was a significant interaction sarily include a neuroscience component, and studied between sex and neuroscience information in the pre- neuroscience formally as part of a course or lab group. class data [F(1, 20) = 8.5, p < .01], such that the dif- The average score on these four items was 2.9 (SD = ference between women’s preclass satisfaction ratings 0.9), indicating a high level of expertise among our for the Without Neuroscience and the With Neurosci- participants. ence conditions was significantly larger than this differ- ence in the men’s ratings. This effect did not hold in the Results and Discussion postclass test, however. These analyses strongly indicate that whatever training subjects received in the neurosci- Preliminary analyses revealed no differences in perfor- ence class did not affect their performance in the task. mance based on sex or level of education, so all subse- These two studies indicate that logically irrelevant neu- quent analyses do not consider these variables. We roscience information has a reliably positive impact on additionally found acceptably consistent responding to both novices’ and students’ ratings of explanations, par- the 18 items (Cronbach’s a = .71). ticularly bad explanations, that contain this information. We analyzed subjects’ ratings of explanatory satisfac- One concern with this conclusion is our assumption that tion in a 2 (good explanation vs. bad explanation) Â 2 the added neuroscience information really was irrele- (without neuroscience vs. with neuroscience) repeated vant to the explanation. Although we had checked our measures ANOVA (see Figure 3). items with cognitive neuroscientists beforehand, it is We found a main effect of explanation type [F(1, still possible that subjects interpreted some aspect of 46) = 54.9, p < .01]. Just like the novices and students, the neuroscience information as logically relevant or the experts rated good explanations (M = 0.19, SE = content-rich, which would justify their giving higher rat- 0.11) as significantly more satisfying than bad ones (M = ings to the items with neuroscience information. The À0.99, SE = 0.14). subjects’ differential performance with good and bad Unlike the data from the other two groups, the ex- explanations speaks against this interpretation, but per- perts’ data showed no main effect of neuroscience, indi- haps something about the neuroscience information cating that subjects rated explanations in the same way genuinely did improve the bad explanations. regardless of the presence of neuroscience information Experiment 3 thus tests experts in neuroscience, who [F(1, 46) = 1.3, p > .10]. would presumably be able to tell if adding neuroscience This lack of a main effect must be interpreted in light information should indeed make these explanations of a significant interaction between explanation type and more satisfying. Are experts immune to the effects of presence of neuroscience [F(1, 46) = 8.9, p < .01]. Post neuroscience information because their expertise makes hoc analyses reveal that this interaction is due to a them more accurate judges? Or are experts also some- differential effect of neuroscience on the good explana- what seduced by the allure of neuroscience information? tions: Good explanations with neuroscience (M = À0.22, SE = 0.21) were rated as significantly less satisfying than good explanations without neuroscience [M =0.41,SE = EXPERIMENT 3 0.13; F(1, 46) = 8.5, p < .01]. There was no change in Methods ratings for the bad explanations (without neuroscience M = À1.07, SE = 0.19; with neuroscience M = À0.87, Subjects and Procedure SE = 0.21). This indicates that experts are so attuned to Forty-eight neuroscience experts participated in the study proper uses of neuroscience that they recognized the in- (29 women, 19 men; mean age = 27.5 years, SD = sufficiency of the neuroscience information in the With 5.3 years, range = 21–45 years). There were 28 subjects Neuroscience condition. This recognition likely led to the

474 Journal of Cognitive Neuroscience Volume 20, Number 3 Figure 3. Expert group. Mean ratings of how satisfying subjects found the explanations. Error bars indicate standard error of the mean.

drop in satisfaction ratings for the good explanations, We analyzed subjects’ ratings of how satisfying they whereas bad explanations could not possibly have been found the explanations in the four conditions. We found improved by what the experts knew to be an improper ap- that subjects in all groups could tell the difference be- plication of neuroscience information. Informal post hoc tween good explanations and bad explanations, regard- questioning of several participants in this study indicated less of the presence of neuroscience. Reasoning about that they were indeed sensitive to the awkwardness these types of explanations thus does not seem to be and irrelevance of the neuroscience information in the difficult in general because even the participants in our explanations. novice group showed a robust ability to differentiate These results from expert subjects confirm that the between good and bad explanations. neuroscience information in the With Neuroscience con- Our most important finding concerns the effect that ditions should not be seen as adding value to the expla- explanatorily irrelevant neuroscience information has on nations. The results from the two nonexpert groups are subject’s judgments of the explanations. For novices and thus due to these subjects’ misinterpretations of the students, the addition of such neuroscience information neuroscience information, not the information itself. encouraged them to judge the explanations more favor- ably, particularly the bad explanations. That is, extrane- ous neuroscience information makes explanations look GENERAL DISCUSSION more satisfying than they actually are, or at least more satisfying than they otherwise would be judged to be. Summary of Results The students in the cognitive neuroscience class showed The three experiments reported here explored the im- no benefit of training, demonstrating that only a semes- pact of adding scientific-sounding but empirically and ter’s worth of instruction is not enough to dispel the conceptually uninformative neuroscience information effect of neuroscience information on judgments of ex- to both good and bad psychological explanations. Three planations. Many people thus systematically misunder- groups of subjects (novices, neuroscience class students, stand the role that neuroscience should and should and neuroscience experts) read brief descriptions of not play in psychological explanations, revealing that psychological phenomena followed by a good or bad logically irrelevant neuroscience information can be explanation that did or did not contain logically irrelevant seductive—it can have much more of an impact on par- neuroscience information. Although real neuropsycho- ticipants’ judgments than it ought to. logical data certainly can provide insight into behavior However, the impact of superfluous neuroscience in- and into psychological mechanisms, we specifically formation is not unlimited. Although novices and stu- wanted to investigate the possible effects of the presence dents rated bad explanations as more satisfying when of neuroscience information, regardless of the role that they contained neuroscience information, experts did this information plays in an explanation. The neurosci- not. In fact, subjects in the expert group tended to rate ence information in the With Neuroscience condition good explanations with neuroscience information as thus did not affect the logic or content of the psycholog- worse than good explanations without neuroscience, ical explanations, allowing us to see whether the mere indicating their understanding that the added neurosci- mention of a neural process can affect subjects’ judg- ence information was inappropriate for the phenome- ments of explanations. non being described. There is thus some noticeable

Weisberg et al. 475 benefit of extended and specific training on the judg- current stature both in psychological research and in the ment of explanations. popular press. However, we believe that our results are not necessarily limited to neuroscience or even to psy- chology. Rather, people may be responding to some more Why are Nonexperts Fooled? general property of the neuroscience information that Nonexperts judge explanations with neuroscience in- encouraged them to find the explanations in the With formation as more satisfying than explanations without Neuroscience condition more satisfying. neuroscience, especially bad explanations. One might be To speculate about the nature of this property, people tempted to conclude from these results that neuroscience seeking explanations may be biased to look for a sim- information in explanations is a powerful clue to the ple reductionist structure. That is, people often hear goodness of explanations; nonexperts who see neurosci- explanations of ‘‘higher-level’’ or macroscopic phenom- ence information automatically judge explanations con- ena that appeal to ‘‘lower-level’’ or microscopic phe- taining it more favorably. This conclusion suggests that nomena. Because the neuroscience explanations in the these two groups of subjects fell prey to a reasoning current study shared this general format of reducing heuristic (e.g., Shafir, Smith, & Osherson, 1990; Tversky psychological phenomena to their lower-level neurosci- & Kahneman, 1974, 1981). A plausible heuristic might entific counterparts, participants may have jumped to state that explanations involving more technical language the conclusion that the neuroscience information pro- are better, perhaps because they look more ‘‘scientific.’’ vided them with a physical explanation for a behavioral The presence of such a heuristic would predict that phenomenon. The mere mention of a lower level of subjects should judge all explanations containing neuro- analysis may have made the bad behavioral explanations science information as more satisfying than all explana- seem connected to a larger explanatory system, and hence tions without neuroscience, because neuroscience is itself more insightful. If this is the case, other types of logically a cue to the goodness of an explanation. irrelevant information that tap into a general reductionist However, this was not the case in our data. Both novices framework could encourage people to judge a wide and students showed a differential impact of neuroscience variety of poor explanations as satisfying. information on their judgments such that the ratings for There are certainly other possible mechanisms by bad explanations increased much more markedly than rat- which neuroscience information may affect judgments of ings for good explanations with the addition of neuro- explanations. For instance, neuroscience may illustrate a science information. This interaction effect suggests that connection between the mind and the brain that people an across-the-board reasoning heuristic is probably not re- implicitly believe not to exist, or not to exist in such a sponsible for the nonexpert subjects’ judgments. strong way (see Bloom, 2004a). Additionally, neuroscience We see a closer affinity between our work and the so- is associated with powerful visual imagery, which may called seductive details effect (Harp & Mayer, 1998; Garner, merely attract attention to neuroscience studies but which Alexander, Gillingham, Kulikowich, & Brown, 1991; Garner, is also known to interfere with subjects’ abilities to explain Gillingham, & White, 1989). Seductive details, related but the workings of physical systems (Hayes, Huleatt, & Keil, logically irrelevant details presented as part of an argument, in preparation) and to render scientific claims more con- tend to make it more difficult for subjects to encode and vincing (McCabe & Castel, in press). Indeed, it is possible later recall the main argument of a text. Subjects’ attention that ‘‘pictures of blobs on brains seduce one into thinking is diverted from important generalizations in the text to- that we can now directly observe psychological processes’’ ward these interesting but irrelevant details, such that they (Henson, 2005, p. 228). However, the mechanism by perform worse on a memory test and have a harder time which irrelevant neuroscience information affects judg- extracting the most important points in the text. ment may also be far simpler: Any meaningless terminol- Despite the strength of this seductive details effect in ogy, not necessarily scientific jargon, can change behavior. this previous work and in our current work, it is not im- Previous studies have found that providing subjects with mediately clear why nonexpert participants in our study ‘‘placebic’’ information (e.g., ‘‘May I use the Xerox ma- judged that seductive details, in the form of neuroscience chine; I have to make copies?’’) increases compliance with information, made the explanations we presented more a request over and above a condition in which the re- satisfying. Future investigations into this effect could an- searcher simply makes the request (e.g., ‘‘May I use the swer this question by including qualitative measures to Xerox machine?’’) (Langer, Blank, & Chanowitz, 1978). determine precisely how subjects view the differences These characteristics of neuroscience information among the explanations. In the absence of such data, may singly or jointly explain why subjects judged ex- we can question whether something about neuroscience planations containing neuroscience information as gen- information in particular did the work of fooling our erally more satisfying than those that did not. But the subjects. We suspect not—other kinds of information be- most important point about the current study is not that sides neuroscience could have similar effects. We focused neuroscience information itself causes subjects to lose the current experiments on neuroscience because it their grip on their normally well-functioning judgment provides a particularly fertile testing ground, due to its processes. Rather, neuroscience information happens to

476 Journal of Cognitive Neuroscience Volume 20, Number 3 represent the intersection of a variety of properties that Without Neuroscience conditions, both the good and the bad can conspire together to impair judgment. Future re- explanations in the With Neuroscience conditions appear less elegant and less parsimonious than their without-neuroscience search should aim to tease apart which properties are counterparts, as can be seen in Table 1. But this design provides most important in this impairment, and indeed, we are an especially stringent test of our hypothesis: We expect that planning to follow up on the current study by examining explanations with neuroscience will be judged as more satisfying comparable effects in other special sciences. We predict than explanations without, despite cosmetic and logical flaws. that any of these properties alone would be sufficient for our effect, but that they are more powerful in combina- tion, hence especially powerful for the case of neuro- REFERENCES science, which represents the intersection of all four. Bloom, P. (2004a). Descartes’ baby. New York: Basic Books. Regardless of the breadth of our effect or the mech- Bloom, P. (2004b). The duel between body and soul. The anism by which it occurs, the mere fact that irrelevant New York Times, A25. information can interfere with people’s judgments of Farah, M. J. (2005). Neuroethics: The practical and the explanations has implications for how neuroscience philosophical. Trends in Cognitive Sciences, 9, 34–40. Feigenson, N. (2006). Brain imaging and courtroom information in particular, and scientific information in evidence: On the admissibility and persuasiveness of fMRI. general, is viewed and used outside of the laboratory. International Journal of Law in Context, 2, 233–255. Neuroscience research has the potential to change our Garner, R., Alexander, P. A., Gillingham, M. G., Kulikowich, views of personal responsibility, legal regulation, educa- J. M., & Brown, R. (1991). Interest and learning from text. tion, and even the nature of the self (Farah, 2005; American Educational Research Journal, 28, 643–659. Garner, R., Gillingham, M. G., & White, C. S. (1989). Bloom, 2004b). To take a recent example, some legal Effects of ‘‘seductive details’’ on macroprocessing and scholars have suggested that neuroimaging technology microprocessing in adults and children. Cognition and could be used in jury selection, to ensure that jurors are Instruction, 6, 41–57. free of bias, or in questioning suspects, to ensure that Harp, S. F., & Mayer, R. E. (1998). How seductive details they are not lying (Rosen, 2007). Given the results re- do their damage: A theory of cognitive interest in science learning. Journal of Educational Psychology, 90, 414–434. ported here, such evidence presented in a courtroom, a Hayes, B. K., Huleatt, L. A., & Keil, F. (in preparation). classroom, or a political debate, regardless of the scien- Mechanisms underlying the illusion of explanatory depth. tific status or relevance of this evidence, could strongly Henson, R. (2005). What can functional neuroimaging tell sway opinion, beyond what the evidence can support the experimental psychologist? Quarterly Journal of (see Feigenson, 2006). We have shown that people seem , 58A, 193–233. Keil, F. C. (2006). Explanation and understanding. Annual all too ready to accept explanations that allude to Review of Psychology, 51, 227–254. neuroscience, even if they are not accurate reflections Kelemen, D. (1999). Function, goals, and intention: Children’s of the scientific data, and even if they would other- teleological reasoning about objects. Trends in Cognitive wise be seen as far less satisfying. Because it is unlikely Sciences, 3, 461–468. that the popularity of neuroscience findings in the pub- Kikas, E. (2003). University students’ conceptions of different physical phenomena. Journal of Adult Development, 10, lic sphere will wane any time soon, we see in the cur- 139–150. rent results more reasons for caution when applying Langer, E., Blank, A., & Chanowitz, B. (1978). The mindlessness neuroscientific findings to social issues. Even if expert of ostensibly thoughtful action: The role of ‘‘placebic’’ practitioners can easily distinguish good neuroscience information in interpersonal interaction. Journal of explanations from bad, they must not assume that those Personality and Social Psychology, 36, 635–642. Lombrozo, T. (2006). The structure and function of outside the discipline will be as discriminating. explanations. Trends in Cognitive Sciences, 10, 464–470. Lombrozo, T., & Carey, S. (2006). Functional explanation and the function of explanation. Cognition, 99, 167–204. Acknowledgments McCabe, D. P., & Castel, A. D. (in press). Seeing is believing: The effect of brain images as judgments of scientific We thank Paul Bloom, Martha Farah, Michael Weisberg, two reasoning. Cognition. anonymous reviewers, and all the members of the Cognition Rips, L. J. (2002). Circular reasoning. , 26, and Development Lab for their conversations about this work. 767–795. Special thanks is also due to Marvin Chun, Marcia Johnson, Rosen, J. (2007, March 11). The brain on the stand. The New Christy Marshuetz, Carol Raye, and all the members of their York Times Magazine, 49. labs for their assistance with our neuroscience items. We ac- Rozenblit, L., & Keil, F. (2002). The misunderstood limits of knowledge support from NIH Grant R-37-HD023922 to F. C. K. folk science: An illusion of explanatory depth. Cognitive Reprint requests should be sent to Deena Skolnick Weisberg, De- Science, 92, 1–42. partment of Psychology, Yale University, P. O. Box 208205, New Shafir, E. B., Smith, E. E., & Osherson, D. N. (1990). Typicality Haven, CT 06520-8205, or via e-mail: [email protected]. and reasoning fallacies. Memory & Cognition, 18, 229–239. Trout, J. D. (2002). Scientific explanation and the of understanding. Philosophy of Science, 69, 212–233. Note Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185, 1124–1131. 1. Because we constructed the stimuli in the With Neuro- Tversky, A., & Kahneman, D. (1981). The framing of decisions science conditions by modifying the explanations from the and the psychology of choice. Science, 211, 453–458.

Weisberg et al. 477 This article has been cited by:

1. Nancy N Potter. 2010. Recent developments in philosophy of mind and psychopathology. Current Opinion in Psychiatry 23:6, 542-545. [CrossRef] 2. Torsten Porsch, Rainer Bromme. 2010. Effects of epistemological sensitization on source choices. Instructional Science . [CrossRef] 3. Jan Slaby. 2010. Steps towards a Critical Neuroscience. Phenomenology and the Cognitive Sciences 9:3, 397-416. [CrossRef] 4. J. Nicole Shelton, Jan Marie Alegre, Deborah Son. 2010. Social Stigma and Disadvantage: Current Themes and Future Prospects. Journal of Social Issues 66:3, 618-633. [CrossRef] 5. Sonia K. Kang, Michael Inzlicht, Belle Derks. 2010. Social Neuroscience and Public Policy on Intergroup Relations: A Hegelian Analysis. Journal of Social Issues 66:3, 585-601. [CrossRef] 6. Frank C. Keil. 2010. The Feasibility of Folk Science. Cognitive Science 34:5, 826-862. [CrossRef] 7. Kurt W. Fischer, Usha Goswami, John Geake. 2010. The Future of Educational Neuroscience. Mind, Brain, and Education 4:2, 68-80. [CrossRef] 8. Bette Anton. 2010. CQ Sources/Bibliography. Cambridge Quarterly of Healthcare Ethics 19:02, 230. [CrossRef] 9. Andrea Lavazza, Mario De Caro. 2010. Not so Fast. On Some Bold Neuroscientific Claims Concerning Human Agency. Neuroethics 3:1, 23-41. [CrossRef] 10. Virginia Hughes. 2010. Science in court: Head case. Nature 464:7287, 340-342. [CrossRef] 11. Lesley J. Sylvan, Joanna A. Christodoulou. 2010. Understanding the Role of Neuroscience in Brain Based Products: A Guide for Educators and Consumers. Mind, Brain, and Education 4:1, 1-7. [CrossRef] 12. Celine van Golde, Stefanie J. Sharman, Ingrid Candel. 2010. High prevalence information from different sources affects the development of false beliefs. Applied Cognitive Psychology 24:2, 152-163. [CrossRef] 13. Judy Illes, Mary Anne Moser, Jennifer B. McCormick, Eric Racine, Sandra Blakeslee, Arthur Caplan, Erika Check Hayden, Jay Ingram, Tiffany Lohwater, Peter McKnight, Christie Nicholson, Anthony Phillips, Kevin D. Sauvé, Elaine Snell, Samuel Weiss. 2010. Neurotalk: improving the communication of neuroscience research. Nature Reviews Neuroscience 11:1, 61-69. [CrossRef] 14. Nicole A Vincent. 2010. On the Relevance of Neuroscience to Criminal Responsibility. Criminal Law and Philosophy 4:1, 77-98. [CrossRef] 15. Lutz Jäncke. 2010. Hirnforschung: sinnvolle Ergänzung oder überflüssiges Anhängsel der Psychologie?. Psychologische Rundschau 61:4, 191-198. [CrossRef] 16. Lutz Jäncke, Franz Petermann. 2010. Zum Verhältnis von Biologie und Psychologie. Psychologische Rundschau 61:4, 175-179. [CrossRef] 17. Rainer Mausfeld. 2010. Psychologie, Biologie, kognitive Neurowissenschaften. Psychologische Rundschau 61:4, 180-190. [CrossRef] 18. Ilina Singh, Nikolas Rose. 2009. Biomarkers in psychiatry. Nature 460:7252, 202-207. [CrossRef] 19. Chris Kaposy. 2009. Will Neuroscientific Discoveries about Free Will and Selfhood Change our Ethical Practices?. Neuroethics 2:1, 51-59. [CrossRef] 20. Bruce Luber, Carl Fisher, Paul S. Appelbaum, Marcus Ploesser, Sarah H. Lisanby. 2009. Non-invasive brain stimulation in the detection of deception: Scientific challenges and ethical consequences. Behavioral Sciences & the Law 27:2, 191-208. [CrossRef] 21. Daniel A. Martell. 2009. Neuroscience and the law: Philosophical differences and practical constraints. Behavioral Sciences & the Law 27:2, 123-136. [CrossRef] 22. Boba M. Samuels. 2009. Can the Differences Between Education and Neuroscience be Overcome by Mind, Brain, and Education?. Mind, Brain, and Education 3:1, 45-55. [CrossRef] 23. Suparna Choudhury, Saskia Kathi Nagel, Jan Slaby. 2009. Critical Neuroscience: Linking Neuroscience and Society through Critical Practice. BioSocieties 4:1, 61-77. [CrossRef] 24. Donna Coch, Stephen A. Michlovitz, Daniel Ansari, Abigail Baird. 2009. Building Mind, Brain, and Education Connections: The View From the Upper Valley. Mind, Brain, and Education 3:1, 27-33. [CrossRef] 25. 2009. Connecting the dots. Nature Neuroscience 12:2, 99-99. [CrossRef] 26. Martha J. Farah, M. Elizabeth Smith, Cyrena Gawuga, Dennis Lindsell, Dean Foster. 2009. Brain Imaging and Brain Privacy: A Realistic Concern?Brain Imaging and Brain Privacy: A Realistic Concern?. Journal of Cognitive Neuroscience 21:1, 119-127. [Abstract] [Full Text] [PDF] [PDF Plus] 27. USHA GOSWAMI. 2008. Principles of Learning, Implications for Teaching: A Cognitive Neuroscience Perspective. Journal of Philosophy of Education 42:3-4, 381-399. [CrossRef] 28. Tyler K. Perrachione, John R. Perrachione. 2008. Brains and brands: developing mutually informative research in neuroscience and marketing. Journal of Consumer Behaviour 7:4-5, 303-318. [CrossRef] 29. Emily R. Murphy, Judy Illes, Peter B. Reiner. 2008. Neuroethics of neuromarketing. Journal of Consumer Behaviour 7:4-5, 293-302. [CrossRef] 30. Gary Greenberg, Ty PartridgeBiology, Evolution, and Psychological Development . [CrossRef]