Many of you are probably familiar with the story of Phineas Gage. He was widely regarded as a kind and generous man, but he suffered from a freak accident during his work on a railroad, and the result was that a railroad spike ended up entering his brain. After the accident was over, the person who remained was not a kind or generous man. He was impulsive, callous, and clearly lacked all of the moral virtues that Phineas had previously shown.
Now, let's consider this case as a problem of personal identity. In particular, let's ask yourself whether the following sentence is correct:
- The original man named Phineas does not exist anymore; the man after the accident is a different person.
Many people have the intuition that this sentence is correct. It might seem, then, that our intuitions conform to an approach to personal identity that emphasizes psychological similarity. Since the man after the accident is not sufficiently similar to the original Phineas, we conclude that they are not the same person.
In a new paper in Analysis, Kevin Tobia make an incisive criticism of this interpretation. As he points out, it is indeed the case that the man after the accident is dissimilar in certain respects from the original Phineas, but there is also another quite salient fact about him. Specifically, he is morally worse than the original Phineas. That is, it is not just that he differs psychologically in some way; he specifically differs by lacking some of the original Phineas's moral virtues. Might that be the explanation of our intuitions here?
To find out, Tobia conducted a simple and elegant experiment. Some participants received the actual historical story of Phineas Gage. Then other participants received a version that was tweaked in one important respect. In particular, they received a 'reverse Phineas Gage case.' That is, these participants were told that Phineas started out as a person who was callous and impulsive, then suffered from an accident, after which there remained a person who was kind and generous. In other words, all participants received a case in which the people before and after the accident are highly dissimilar, but some received a case in which the person after the accident is morally worse while others received a case in which the person after the accident is morally better.
All of these participants then received the question posed above. They were asked whether it was right to say that the original man named Phineas doesn't exist anymore, and the person after the accident is actually a different person.
Confirming numerous claims from the existing literature, participants who received the story of the actual historical Phineas Gage tended to say that the person after the accident was a different person. But those participants who got the reverse Phineas Gage case said just the opposite! They tended to say that the kind and generous person after the accident truly was still Phineas Gage. As Tobia points out, this result suggests that people's intuitions about personal identity are determined not only by the degree of psychological dissimilarity but by a moral question, namely, whether it is a case in which moral virtues are gained or lost.
The full text of Tobia's forthcoming paper is available online, but please do feel free to comment even if you haven't read the paper. I'm very curious to hear any thoughts people might have about how to explain these results or what implications they might have for larger philosophical problems.
It is strange, I would have answered that it is still the same person, in both cases !
I wonder if there may be a principle of charity at work here. When a person acts badly or has a bad character, perhaps we’re more likely to look for reasons to let them off the hook, whereas, when they act well or have a good character, we’re more likely to unreflectively attribute those things to the person.
So, in the Phineas Gage case, participants are likely to see the injury as a reason to think that Gage’s moral character just isn’t attributable to him and, so, to say things like, “He’s just not himself anymore.” On the other hand, in the reverse Phineas Gage case, participants may be more likely to assume that he is still himself and to instead say things like, “His injury has really changed his outlook on life,” or something to that effect.
I’ve not read the full text, so perhaps Tobia has already addressed this point.
Question on the experimental design. Tobia says that “the magnitude of dissimilarity is constant between conditions” but in the vignettes, it isn’t clear to me that there are any magnitudes mentioned. And each subject sees only one condition, right? So, what reason do we have to think that subjects think the dissimilarities are the same in the improvement and deterioration cases? And how could we control for this? Perhaps when people anchor on kind, they adjust less when told that the post-trauma person is cruel. And when people anchor on cruel, they adjust more when told the post-trauma person is kind. That might just reflect optimism in the participants. And it would mean that the answers reflect an asymmetry in the way the story is interpreted, not an asymmetry about personal identity judgments.
Also, the effect here is small or smallish-medium. (About d=0.4 in each study.) I would like to see a bunch of replication before I take it too seriously.
In any event, I think the more interesting thing in the study is that people tend to reject the claim that there is a different person, regardless of whether the person improves or deteriorates. I wonder if there is any vignette that gets people to strongly endorse the claim that on each side of a trauma (like in the Gage example) or on each side of development (like in the Russian example) we find a different person.
With respect to the moral improvement, I wonder what we would see if we gave people a case with improvement or deterioration along a different dimension. I would have liked to see some studies based on the Flowers for Algernon example (pages 8-9). Supposing Charlie doesn’t change morally but does change with respect to intelligence, does improvement/deterioration predict how people will think about personal identity? What if we ran a study using a fictional character like Captain America, where the change is entirely physical? We’d have to be careful in making the stories really symmetric, but I would be interested to see how it comes out.
Nathan,
Nice point. I think there is definitely something to that. However, if you just say that people think that there is some deeper good within the bad agent, it doesn’t seem that you will be able to explain the asymmetry. (After all, in the deterioration case, you should say that the person after the accident is actually better than he appears, hence more similar to the original Phineas.) The hypothesis might work, however, if you say that the idea is that people’s good qualities are somehow more revealing of their ‘essence.’
Cyril and Jonathan,
Looking at the original historical Phineas case for myself, I am inclined to say something like: ‘There is a certain sense in which he is clearly Phineas, but there is also a sense in which he is not truly Phineas anymore.’ If I were forced to make a single judgment as to whether is Phineas or not, I might choose a point at around the midpoint of the scale, but that response would be a little bit misleading. It is not really that I am halfway in agreement with the claim that he is Phineas; rather, my sense is that there is one sense in which completely correct to say that he is Phineas and one sense in which it is completely incorrect. So I wonder if participants might give far higher ratings if one replaced the statement with something like: ‘There is a sense in which the person after the accident is not Phineas at all.’
Jonathan,
Always helpful to conduct further replication studies, but I was just curious to hear what you think might be most helpful here. Is it that you think these exact results might be due to type 1 error, meaning that we need to just run these exact studies again? Or is it that you think the results might be due to the peculiarities of these particular methods, meaning that we should try running further studies with different vignettes?
With respect to replication, I would like to see both. I would like to see the studies repeated as they stand but with more like 500 or 1000 participants per condition. My main worry here is one expressed very nicely by Andrew Gelman in relation to a silly ESP experiment a few years back (http://andrewgelman.com/2011/01/11/one_more_time_o/). He writes, “When you’re studying small effects and you use statistical significance as a filter and don’t do any partial pooling, whatever you have that’s left standing that survives the filtering process will overestimate the true effect.” So my guess is that if there is a real effect here at all, it is quite small. And that makes it very hard to detect again with similarly-powered studies.
I would also like to see the same vignettes with a different response set — your suggestion looks interesting here, and you could do it with a multiple choice. It might be worth running the vignette with a free response just to see what comes out.
But more importantly, I would like to see a larger range of vignettes, some vignettes intended to falsify the view on the assumption that it is false, and (if possible) some alternatives to vignette-plus-question methods, like false-recall tasks or belief-updating tasks. And it would be nice to see something within subjects.
Incidentally, in your write-up, you say, “Confirming numerous claims from the existing literature, participants who received the story of the actual historical Phineas Gage tended to say that the person after the accident was a different person.” But that isn’t what the paper reports. In the improvement condition for the straight Gage probe, the mean was 2.61, and in the deterioration condition, the mean was 3.26. Both of those are on the “same person” side of the midpoint of 4. It would be interesting to see the histograms.
Could this result also signal an expectation of normal moral development? We expect people to become kind and generous, which is partly why we blame failures to do so.
In the Phineas case, some might interpret the accident as a disruption of normal moral development. As a disruption, Phineas would not be blameworthy for his resulting character. One way to suspend blame is to attribute the bad character to a different person.
In the reverse Phineas case, some might interpret the accident as enhancing or accelerating normal moral development. Since he would/should have arrived at that character eventually anyway, he might be praiseworthy for his resulting character. He deserves praise only if he is substantially the same person.
Michael,
That’s a really nice point. Just as you say, it might be that people have a certain conception of normal development (a human being’s natural teleology, as it were), and their personal identity judgments are based in part on the degree to which a person’s development diverges from this normal trajectory.
Actually, this hypothesis makes for an interesting contrast with the one I discussed based on Nathan’s comment above. One possibility would be that this effect reflects an intuition that human beings have a good essence, while another would be that this effect reflects an intuition that people have a normal developmental trajectory that moves toward the good.
Jonathan,
Thanks for the correction of my original write-up (definitely very much appreciated), and also for the points about replication. Tobia actually has some further studies on this phenomenon, which I assume will appear in subsequent publications, but either way, you are completely right to be focusing on this issue. It is, or at least should be, absolutely central to contemporary work in experimental philosophy.
Hi all, thanks so much for the comments.
Jonathan – thanks very much for all these extremely helpful points. I’m very grateful for these suggestions, and I share Josh’s sentiments about the value of replication in xphi/cogsci.
For what it’s worth, I just ran an exact replication of study 1 (Phineas Gage vignette). The original finding was replicated: improvement ratings (M = 2.93, SD = 1.79) were lower than deterioration ratings (M = 3.69, SD = 1.93), t(140) = 2.43, p = .016, d = .41. Do you still think it would be valuable to run a replication with 500 or 1000 participants per condition (1000 or 2000 total)? As a graduate student, that’s ever so slightly over my budget, but I’d love to see it, too!
The other thing worth mentioning is that I have some other unpublished studies showing an improvement/deterioration effect in this same direction for different cases (including brain transplant cases in which the mean is closer to the “non-identical” side of the scale).
More generally, I would be happy to help anyone who is interested in any further replication tests and/or extensions. Someone could also pursue this on his/her own; I believe all the material is in the paper (except the prompt, which is: “Do you agree more with Art or Bart?”).
There’s one other related thought about your suggestion about using vignettes “intended to falsify the view.” That also sounds great, but I want to make sure I’m not interpreted as endorsing a claim too broad. I certainly do not think (or predict) that in EVERY case, direction of change would affect identity intuitions. Personal identity thought experiments (even before contemporary experimental philosophy came on the scene) are notorious for eliciting varying intuitions (e.g. some elicit intuitions indicating that bodily continuity is crucial, others elicit intuitions indicating bodily continuity does not matter very much at all). My guess is that the same would be true for direction of change. One really interesting question (to me, at least) is whether direction of change is actually relevant to personal identity. Our answer to this question may be informed by learning in which cases direction of change affects identity intuitions – and in which cases it does not.
I’m very intrigued by your other suggestion about improvement/deterioration along different dimensions (e.g. intelligence, body, etc.). This relates to Michael’s nice point about an expectation of normal moral development. One possible start of an explanation of the moral improvement/deterioration effect might be that we expect people to become kind and generous and see them as persisting when they change in these directions. But we might also expect people to grow in other (non-moral) ways. For instance, I think there is nothing inherently moral about language ability, but perhaps we would judge the persistence of someone whose language ability gradually improves differently from that of someone whose language ability begins to deteriorate. I haven’t investigated this experimentally, but I wonder if others share these intuitions about non-moral improvements and deteriorations?
Kevin,
Since you actually see the same thing in your replication, it isn’t really necessary to do a larger study. The main reason to do a larger study is if you are worried that the initial study over-estimates the size of the effect. If, for example, the measured effect in your first study was 0.4 but the *true* effect (insofar as there is one) was 0.2, then attempts to replicate that assumed an effect size of 0.4 would be unlikely to actually detect the effect. When I suggested a very large sample size, I was supposing that the true effect size was much smaller than 0.4, and I was setting a high bar for the power of the test (0.9 instead of the usual 0.8).
I think it would be interesting to run a replication where you also ask participants the following questions.
1. Before the accident, how cruel [kind] was Gage? You could put this on a large scale, say 0 to 100 or -100 to 100.
2. After the accident, how cruel [kind] was Gage? Same scale.
If you find that the before and after ratings are mirror-images (and the effect persists), that would address my first worry that the dissimilarities might not be the same in the improvement and deterioration cases. It would also let you do some cool modeling of the size of the effect given direct measures of degree of improvement or deterioration. (And then you could try to manipulate the degree of improvement or deterioration to see what happens.)
With respect to falsifying, I agree that you are not on the hook for *every* case. What I had in mind was two kinds of study: one where you say to yourself, “If I don’t see the effect I’m expecting here, I have to be wrong,” and one where you say, “If I see an effect here, I have to be wrong.”
Kevin,
Thanks for this work. Very interesting.
Have there been any considerations to run the experiment solely on Buddhists? Doing so would at least have the potential shed light on whether direction of change is actually relevant to ascriptions of personal identity or whether something else is being captured by the broader, common category of “same person.”
So if for example Buddhists demonstrate statistically smaller variance with respect to direction of change, that would give a reason to believe that ascriptions of “same person” that are asymmetrical are tracking personal identity. If, on the other hand, Buddhists demonstrate a similar variance with respect to direction of change, that would give a reason to believe that ascriptions of “same person” that are asymmetrical are tracking something else (e.g. appropriateness of punishing/rewarding for past behavior, for example).
Regardless, really interesting stuff. Thanks for sharing
Jonathan – thanks so much for the follow-up. That sounds like a very cool study. I like how it might offer the ability to manipulate the degree of reported improvement or deterioration.
Regarding the falsification point, I think there are a few different claims to consider. One is very simple, something like an existence claim that direction of change at least *sometimes* affects personal identity judgments. I suppose that could be challenged by evidence or argument that what I’ve found is not really an effect of direction of change but is instead something else.
Other interesting things to investigate (falsify, verify, support, etc.) are various plausible competing hypotheses about what theory explains the finding. This strikes me as a job with both empirical and non-empirical parts: what theories plausibly explain why direction of change affects personal identity intuitions – or the identity relation itself; and what are the philosophical costs and benefits of these? [There’s a bigger philosophical question in the background, which is: regardless of the findings about personal identity *intuitions*, is direction of change relevant to the actual personal identity relation? If so, what implications does this have for personal identity (as well as moral/legal philosophy)? On the other hand, if we hold direction of change is irrelevant to personal identity, should we discount thought-experiment results about Phineas Gage and the Russian Nobleman in light of this finding?]
I’m still thinking a lot more about this, but two of the main theoretical contenders have already been suggested here. One is the group of views that takes good parts of a person to be more essential than bad parts. Nathan’s early post seems like one articulation of this view: we attribute good character/properties to a person more readily than we attribute bad character/properties. Michael’s suggestion about “normal moral development” strikes me as an example of another class of views: personal identity is affected by expectations about normal development.
Marcus – thanks very much! That’s a really intriguing idea. If I understand correctly, the thought is that assuming the popular view that Buddhists deny the existence of a persisting self, any kind of connection posited between earlier and later agents would have to be evidence of some other relation besides personal identity – or of something else entirely (e.g. as you suggest, appropriateness of punishing/rewarding)? I think the study you suggest would be really interesting and worthwhile.
I believe Shaun Nichols, Jay Garfield, and Nina Strohminger are currently conducting a large research project, “Death and the Self,” involving cross-cultural research on questions of personal identity among Hindus and Buddhists in India and Christians in the United States. I don’t know if any of their findings are published or available yet, but I think those results will definitely be of great interest to those pursuing these topics.
Chiming in very late and have not had a chance to read the paper, but–and this is sort of along the lines of Nathan’s point and Michael’s point–couldn’t many people believe, correctly or incorrectly, that there’s something about near-death experiences that, normally, prompts people to transcend their self-centered perspective? If so, then it might seem normal for someone that has had a near-death experience and survived to be grateful to be alive and to be, in this ‘elevated’ mood, more kind and generous; by contrast, it might seem like we need a special story for someone who, after a near death experience, becomes cruel.