Recently, several social psychologists have posited a "Whodunit" system in the brain that's always looking to assign authorship -- either our own or somebody else's -- to actions. Most of the time, it's pretty easy to tell when we've done something, because we have all sorts of signals coming from the body, along with the brain's awareness of the signal's it's sending. But in some cases, particularly when bodily signals are ambiguous or absent, the "Whodunit" system can be tricked into thinking that someone else caused an action that was really of our own doing, or that we caused an action when someone (or something) else did.
Daniel Wegner, a social psychologist at Haawvad, has spent the last several years studying how we can screw with people's sense of authorship, and even wrote a book about it, titled The Illusion of Conscious Will. Wegner argues that attributions of authorship are inferences, and has shown over and over again that if you screw with the information that the whodunit system uses to make those inferences, you can create some pretty powerful illusions of authorship. His most recent demonstration is in a paper in the January, 2008 issue of the Journal of Experimental Social Psychology(1). In the paper, Dijksterhuis et al. argue that the whodunit system (they don't call it that, but I think it's a good name for it) is constantly keeping track of potential authors. If that's the case, then making them aware of potential authors could trick the system into attributing authorship to them.
So test the idea that the whodunit system is keeping track of potential authors, they conducted three experiments in which subjects were faced with assigning authorship to an action for which the authorship was ambiguous (i.e., it wasn't clear whether they had done it or another agent had). Prior to making the authorship judgments, the participants were primed either with "I" and "me," which should make them attend more to themselves, and thus decide that they had caused the action, or with the name of another agent, which should cause them attend more to that agent and, as a result, decide that the other agent had caused the action.
In the first experiment, they just used the "I" and "me" primes (actually, "Ik" and "mij," as the participants were Dutch). The participants were given a lexical decision task, which involves presenting them with letter-strings and asking them to indicate whether they're words or non-words. Half of the time, the strings were words, and half the time they were non-words (e.g., "gewws"). For half of the 72 trials, the words "Ik" and "mij" preceded the presentation of the letter strings for 17ms (not enough time to consciously notice it, but enough time to process it unconsciously), and for the other half, the word "de" (Dutch for "the") preceded the strings.
These sorts of tasks are used in priming research all the time, but in this case, Dijksterhuis et al. added a twist. They told participants that when they indicated whether a string was a word or non-word, the string would be removed from the screen and the next string would appear, which is how these tasks usually work. But they also told them that sometimes the computer would remove the word and put up the next one before they had pressed anything. The idea, then, is to work faster than the computer. However, Dijksterhuis et al. varied the amount of time that the computer waited before removing the word itself (it waited between 450 and 700 ms; participants' average response time was 487ms), so that it would be difficult for the participant to know whether they'd caused the word to be removed of the computer had.
After each trial, participants had to indicate whether it was they who'd caused the word to disappear or the computer, using a 6-point scale to indicate how sure they were of their authorship judgment. As they expected, they found that on trials in which the participants had been primed with "Ik" and "mej," their self-authorship ratings were slightly higher than the "de" trials. The effect is pretty small (3.83 vs. 3.64), but statistically significant, and given the fact that the prime was subliminal and presumably unrelated to the authorship judgment, that's somewhat impressive.
Their second study used the same lexical decision task paradigm, but this time, English-speaking participants were primed either with the word "computer" (half of the trials) or the word "broccoli" (the other half). The word computer should prime participants to think of the other potential causal agent, and thus bias them towards believing that the computer removed the letter string. And that's what they found: on trials in which they'd been primed with the word "computer," their own sense of authorship was reduced (3.87) compared to trials when they were primed with "broccoli" (4.05).
It's the third experiment, though, that makes the paper cool. This time, half of the trials used the agent-neutral "de" prime again (they're back with the Dutch now), and half of them used "God" as the prime. To analyze the data from this experiment, they divided participants into two groups: believers and non-believers. For non-believers, there was no difference between the two conditions: their own sense of authorship was the same regardless of the prime. For the believers, however, priming them with God significantly reduced (3.05) their sense of authorship compared to the "de" trials (3.63). Interestingly, though the difference was not statistically significant the non-believers' sense of authorship was actually stronger (4.01) in the "God" trials than in the "de" trials (3.72). It's as though they were rebelling against the very idea (unconscious though it may be) that God might have caused the letter strings to disappear. OK, that's a stretch, but this is a blog post, right?
Again, this effect is pretty small, but given the fact that the prime is subliminal, and consciously, the participants know that either they or the computer made the letter strings disappear (and besides, who thinks that God's in the business of screwing with social psychology experiments, other than grad students' whose experiments don't work, I mean?), the fact that they got an effect at all is surprising. Somehow, priming them with "God," who believers think might be messing with lab computers, is enough to trick their whodunit system just a little into believing that maybe, just maybe, neither the computer nor they removed the letter string, but that instead they were witnessing little computer-screen miracles. That's pretty strong evidence, I'd say, that you can screw with the authorship system by suggesting alternative causal agents.
1Dijksterhuisa, A., Preston, J., Wegner, D.M., & Aarts, H. (2008). Effects of subliminal priming of self and God on self-attribution of authorship for events. Journal of Experimental Social Psychology, 44(1), 2-9.
Thanks for the great explanation. What do you make of Dijksterhuis' other work on unconscious deliberation?
Very cool, but I don't really get how this relates to one's sense of will, though. It seems that the judgement is between two competing causal hypotheses regarding non-cognitive physical processes -- whether the key was pressed before or after the switch was determined, did the key-pressing cause the switch -- and not regarding the relation between reason and action or anything of that sort which would be more directly pertinent to will. Am I just off on cog-sci terminology or am I missing something?
Peli, I've actually had that discussion before, with Brandon of Siris I think, and he (assuming that's who it was) convinced me that it doesn't really have much to do with free will. It does, however, have something to do with our own experience of willing, which I suppose could be important for certain discussions of free will.
Jonah, do you mean the deliberation-without-attention stuff? If so, I think that's really interesting, but that there's a lot of work to be done. It's not easy to simulate major decisions in the lab, both because people know they're in the lab, and because the dimensions of complex decisions are... complex. So it's going to take a while before the "dwa" hypothesis has sufficient empirical support, if it's right.
Yeah, I was referring to the deliberation without attention research. (His 2006 Science paper in particular.) I also find it really interesting, but I get the sense that a lot of scientists share your skepticism that complex decisions in the real world can't really be simulated either in the lab or by surveying people in an Ikea.
I do think I remember that discussion; I don't remember convincing you, though! There are positions on free will that are based on something like the experience of authorship, so experiments like this would certainly be relevant to those particular positions.
This is a rather cool experiment. I'd be curious how things would go in cases where the task was cooperative (or where it was not but participants were uncertain whether they were being assisted), and the participants were primed with the name of the (human) cooperator. That is, I wonder if even where authorship is pretty clear the notion that someone else might be cooperating with us might 'dampen' our sense of being the authors of the action by forcing us to spread the possibility of authorship around among potential cooperators.
I think the significance of the experience will-wise strongly depends on how one distinguishes actions from consequences, no? If we take the experiment to have wrecked havoc merely in the attribution of a consequence to an action that it's not very relevant to issues of will, but if we take it to wreck havoc in the attribution of action it's clearly germane.
I think that experimental philosophers who want to challenge incombatibilists should take note of Wegner's findings.
Chris you say:
"Somehow, priming them with "God," who believers think might be messing with lab computers, is enough to trick their whodunit system just a little into believing that maybe, just maybe, neither the computer nor they removed the letter string, but that instead they were witnessing little computer-screen miracles."
This makes it sound as if the researchers asked the believers if they did it or did *God* do it. However, if I am not mistaken, the researchers still kept to the question: did you do it or did the *computer* do it?
Cool study. The numbers are...interesting! They are pretty consistent through the first two experiments, in the 3.8-4.0 range of the scale. And there are small but significant changes (as you make clear) from one condition to the comparison of about 5% -- an amount common in these kinds of priming tasks. From the post, for example, there is a .19 reduction from about 3.8 in the first experiment, and about a .2 reduction from about 4.0 in the second; both 5% decrements.
Then Bang! comes experiment 3. For the believers the reductions after priming with God are huge - almost 20%, and way down to the lowest levels in all the experiments (I am assuming the same scale is used...).
Believers seem to be really searching for a way to unburden themselves...