dark magic?

A recent study just published in PNAS, hit the mainstream news and attendant blogsphere circuit yesterday.  In the study, as most of you must have heard by now, the authors (Eugene Caruso, Chicago Booth School of Business; Nicole L. Mead, Tilburg University Marketing and Emily Balcetis, NYU Psychology) show that respondents who are shown a picture of a biracial political candidate tend to choose doctored photos of that candidate—some “ligthened” and other one’s “darkened”—as “most representative” of that candidate depending on how close they are to that candidate’s political views. When there is a match, the candidate is seen as better represented by his lighter self, and when there is a mismatch, the candidate is seen as better represented by his darker alter-ego (of course this is an experiment so exposure to stimuli is randomized).

The study is in the press because the authors of the study were smart enough to not only use a nondescript biracial candidate (they did that in study 1, the (poor?) guy’s name is Jarome Iginla, “a 32-year old biracial male whose father was a Black Nigerian and whose mother was a  White American.”), but then (in study’s 2 and 3) they selected the most famous biracial candidate in the history of modern democratic politics.  They also planned well, since study 2 was conducted just before the 2008 election and study 3 was conducted right after.  What they find is that the “photoshop” effect was replicated with Obama and that the extent to which participants judged that a lightened photograph better represented Obama’s “true essence”* predicted both voting intentions (study 2, before the election) and restrospective reports of actual voting behavior (study 3, post-election), controlling for both standard lib-con placement and (implicit and explicit) racial attitudes (the last of which were surprisingly impotent in predicting anything in this study; sorry Larry).

This is a very cool study, the authors have pretty big substantive (not just statistically) significant effects and this could probably be easily replicated.  The effect is real and (given the news attention) spectacular.  The issue is of course interpretation.  I believe that the main interpretation that’s circulating in the news report (and which the authors don’t appear to be doing much to combat) is a (slight) misinterpretation.  Most news reports are blabbing about how political views alter or skews the perception of the photograph. In fact the study is literally called “Political Partisanship Influences Perception of Biracial Candidates Skin Tone.”  But while this title was obviously designed with the press-release in mind, this is not what the study is actually about.  In fact, Marc Ambider’s summary at the Atlantic is actually a more accurate characterization of the effect: “Lighter Skin, More Like Me.”

For the key issue in this study is not “partisanship” but the extent to which you rate the candidates view as similar to your own. The confusion stems from the fact that this was explicit in Study 1 (where candidates where informed about the unknown biracial candiates view and then had to judge how representative these views where of their own) but became implicit in studies 2 and 3, since there all that you needed to know was the respondent’s lib-con placement (Obama’s was presumed to be known of course).

But most importantly (and I’m sure the Psychologist in the study knows this very well) this was not a perception task (e.g. in the psycho-physical sense, “estimate how dark or light this photograph is…”).  It was a judgment task (“how well the photographs represent…”), so “partisanship is not skewing perception in the raw sense.  In judgment (like judging distance) perception is involved but it is more complicated than partisanship making you “see” a photograph as lighter or darker. In fact, the best explanation for this effect in my view, is Dan Kahneman’s “attribute substitution” story.  All of the elements for an attribute substitution effect are there: (1) an ill-defined, loosely bounded judgment task (like judging distance), (2) a readily available (cognitively and affectively accessible) cue that can be used as a heuristic (the relative distance between the person and the candidate in terms of political views) in order to produce a judgment.

So I propose that the effect occurs in two steps:

1) First, we substitute the (impossible to define) “true essence” of the candidate with the more accessible “how close is this candidate from my values?” heuristic:  close to me=good essence, far from me=bad essence. This simplifies the problem to one of being asked to judge which picture best represents a bad essence (if candidate disagrees with my views) or a good essence (if candidate agrees).

2) We then answer this (more answerable) question with a second substitution, using a simple (and culturally entrenched heuristic [e.g. see Gandolf’s robe]) dark=bad and light=good); both steps occur at an implicit level of course.  So there is not effect of partisanship on perception.  The partisanship effect is cognitive/affective, and occurs during the first substitution (when judging “distance from self”).  The second effect is “cultural” and (you don’t have to see the scene in Spike Lee’s Malcolm X when the title character goes through a dictionary, or read Durkheim and Mauss’ Primitive Classification to get this), thoroughly intuitive (e.g. Dark Magic, etc.).

*The actual prompt is as follows:

…participants read instructions describing that photographs can differ in how well they ‘‘represent a politician’’ and capture his or her ‘‘true essence.’’ They then rated how much each of three photographs—one lightened, one unaltered, and one darkened—represented the candidate  on  scales  ranging  from  1  (not  at  all)  to  7  (a  great  deal).

Written by Omar

November 25, 2009 at 1:45 pm

13 Responses

Subscribe to comments with RSS.

  1. I haven’t read the paper, so maybe they mention this, but Jarome Iginla’s an NHL all-star.



    November 25, 2009 at 2:23 pm

  2. Hi Omar
    If the heuristic really is close to me=good = lighter, far from me=bad = darker, then it is striking that the IAT (which tries to measure implicit white=good, black=bad associations) neither affects the results nor has reported significant effects on its own.



    November 25, 2009 at 2:35 pm

  3. Omar, thanks for linking this study. I think you have an interesting insight on what these results mean, but I wonder a little about the second part of the model. If people define candidates as good or bad based on how close the political beliefs are (which is very reasonable), could we see something similar with the issue of skin tone? In other words, is whiteness always good, or do people want the candidates who share their values to share their skin tone as well, and it just so happens that the majority of experiment participants were white so they prefer lighter skin? I think there’s theoretical ambiguity here, and room for future studies.

    Simon: They don’t say Iginla an all-star, but they imply that he’s good. (They also say no participant knew who he was from the photo.)



    November 25, 2009 at 6:38 pm

  4. I know nothing about Hockey, but if the guy is an NHL player that would make the study even cooler.

    brubineau , I think that the fact that the usual measures of prejudice don’t matter here is actually a good indication that this is not (primarily) a “racial prejudice” effect, but requires the postulation of a more general mechanism to be accounted for. All that the IAT picks up is automatic associations between perceptions of black (or white) people and and concepts that carry a negative or stereotypical association with race (e.g. violence). I think that given the task at hand here things are a bit different. Persons are “choosing” among a “dark” and a “light” representation of a racially ambiguous subject as corresponding more closely to what they are perceiving to be the “true essence” of that person. So my sense is that it is going to be general connotations associated with light and dark (of which the color of the relative shade in which the person’s skin appears is going to be a salient but not necessarily the only component).

    For instance, in the “lighter” Obama picture, the effect was made to appear natural by showing him as “bathed” with some sort of natural light, while the in the “darker” picture, he appears in a more “shadowy” cast ( I mean cinematographers (not to mention Classical Renaissance painters) have known of this psychological “effect” of dark and light portrayal of persons for a long time (persons shot in shadows appear sinister; persons shown in light appear morally superior, etc.), and I think that interpreting the influence of the light/dark portrayals narrowly as a pure “skin color” effect discounts the fact that the pictures are perceived, and (implicitly) evaluated as a gestalt (in that respect the various jokes that Obama was a lot like Jesus for liberals might had a grain of truth to them!). Once again the fact that prejudice does not matter militates against the narrow skin color interpretation of the findings.

    Noah, I think that what you propose is intuitive, but I think that the association of dark with bad and light with good is such a general, entrenched feature of the culture that I would be shocked if it wasn’t generally shared across racial and ethnic groups (just like a IAT results show that blacks can be as “implicitly prejudiced” against their “own” as anybody else). But possible socio-demographic sources of heterogeneity in the effect in question here is certainly something that should be considered.



    November 25, 2009 at 7:28 pm

  5. @Noah

    This video, a segment of an 8 minute teen directed documentary, shows the results of a replication of Clark’s “doll test.” Of 21 black children, 15 preferred a white baby doll over a black baby doll, and associated the white baby doll with words like “nice” and “good” while the black baby doll was associated with words like “bad.” While we don’t know how the children were sampled or much about their socioeconomic status or overall environment, the results are still powerful.

    The clip of the doll test can be found here:

    You can watch the whole documentary here:


    Alotta Errata

    November 25, 2009 at 8:43 pm

  6. I have a difficult time with some of the brief conclusions made by the authors. They seem to forget the history of race relations in the US when they term the biracial political candidates to be “racially ambiguous.” Many people still hold firm to the “one drop rule” in which simply having a black ancestor somewhere in your line of ancestors automatically makes you “black.” Moreover, if you follow the conception of “color” in the Latin American sense to contain phenotype, facial and hair features, education, and culture, then it isn’t so hard to see the IAT came up not significant. I generally feel like they could have elaborated on their findings a bit more instead of keeping things so brief to lead to misinterpretation.



    November 26, 2009 at 12:44 am

  7. Hillbilly: I don’t understand your point about the one drop rule. Are you saying the authors should have described the target’s race in the paper in accordance with the one drop rule?
    Also, are you suggesting that the subjects in the study may have shared the Latin American sense of color? Why do you think that?


    Thorstein Veblen

    November 26, 2009 at 6:28 am

  8. Thorstein, perhaps I should have stuck to the original point I was trying to make. I think that describing the individuals in the pictures as racially ambiguous misses the point that, in fact, in Western society they are not. Their use of “true essence” is a bit troubling as well as historically the word “essence” was used in a more racial manner to delineate who was “really white” and who was not along with the world “blood” in the sense that black and white individuals have “black blood” and “white blood.”

    I’m not arguing that their findings are wrong, they’re not. I think that the trouble I’m having with the article is how much is not written that I feel would clear things up for not only me, but others and add more weight to their findings. I did like their idea of media consumption as an alternative explanation, but perhaps there could have been a few other explanations they could have (may have for another article?) tested.



    November 26, 2009 at 12:51 pm

  9. I don’t mind their use of the word “racially ambiguous.” If they’d said the targets were all black it would like they were endorsing the one drop rule.

    The sort of discussion where one “problematizes” key concepts and results, a style often favored in sociology, isn’t really possible at PNAS because of serious space limitations. I’m not saying the authors would have delivered the discussion you would have liked, but for this journal it would have been impossible.


    Thorstein Veblen

    November 26, 2009 at 6:14 pm

  10. I agree that PNAS does not allow in-depth discussions, Thorstein. I guess my next question would be why they chose to pursue PNAS as a venue for their article and not another journal (not saying PNAS is a fourth-tier journal or anything like that).



    November 26, 2009 at 10:14 pm

  11. Omar – two questions, since I can’t access the article through my University’s library:

    1. Did the authors actually use that IAT in the study?

    2. What were the demographics of the participants?

    I’m curious to those two answers, as they’ll end up skewing my comments about the article…



    November 27, 2009 at 5:17 am

  12. Hi Annie,

    They used both the IAT and an “explicit” racial attitude scale (in studies 2 and 3). These came in as controls in the models in which they were trying to predict future voting intentions and reports of voting behavior (light-dark judgment as their main predictor and lib-con placement as the other control variable). They had nothing to do with the main effect that was reported in the study, which was simply an association between “light-dark representativeness” and the extent to which the candidate was close or far from the subject’s own political ideology (that’s just a simple chi-square test not a multiple regression).

    The demographics of the participants as reported in the study were standard psychological experiment demographics: “In exchange for entry into a $50 lottery, 108 people from a Chicago-based participant pool completed an online study…” (study 1); “In exchange for course credit, 221 undergraduates at Arizona State University participated in an online study…” (study 2); “In exchange for entry into a $50 raffle, 53 undergraduates from Florida State University participated in the study.” (study 3).

    In terms of race of the participants, this is what the authors had to say:

    Given the small percentage of Black participants in Studies 2 (4%) and 3 (10%), we did not have enough Black participants to test reliably for differences between Black and White participants. Because we base our predictions on the participants’ political group membership (and not their race), we have not excluded any participants based on race in the results we report here. None of the results meaningfully changes when Black participants are excluded from the analyses.



    November 27, 2009 at 2:00 pm

  13. […] skin tone either lighter or darker than it was in the original photograph.  This is, as Omar mentioned, a very cool […]


Comments are closed.

%d bloggers like this: