Motivated Reasoning II: Are Political Partisans Irrational?

Your response, I imagine, is "duh." Partisans are emotional; stop the presses, get me rewrite. Perhaps. But I find the graphic clarity of colorful brain scans to be sobering. It'’s one thing to know that some people get obnoxious during political arguments; it's another thing to see that 30 adult men who read candidates' quotes while strapped down in MRI machines didn't even fire up the thinking parts of their brains. - Dick Meyer, in his column on the Westen et al. study

"Didn't even fire up the thinking parts of their brains." Let me quote that one more time. "Didn't even fire up the thinking parts of their brains." When I read something that stupid in an article in the mainstream media, an article linked and quoted by many bloggers, I am at a loss for words. OK, maybe not a loss; an excess, really, if you consider how long the last post was, and that it was only the first in a two-part series. But I really am blown away. You don't even have to have read anything about the study to know that Meyers must be wrong. That's just not how the brain works. But since he's spread that nonsense all over the web, I feel it's important to explain exactly what the study does show. So, after giving the background of the study in the last post, I'm ready to talk about the methods, results, and conclusions of the study in this one. If you'd like to read the unpublished write-up of the study yourself, you can request a copy from Westen here¹.

The study begins from the following position (p. 3):

Neural network models of motivated reasoning suggest that in affectively relevant situations, the brain equilibrates to solutions that simultaneously satisfy two sets of constraints: cognitive constraints, which maximize goodness of fit to the data, and emotional constraints, which maximize positive affect and minimize negative affect.

In other words, as I described in the post yesterday, there are two systems at work in motivated reasoning: a cold system, that works to come to an accurate conclusion given the input from the environment and memory, and a hot system, driven by emotion, that seeks to come to a desired conclusion (and thus "maximize positive affect and minimize negative affect") by biasing the input and interpretation of information from the environment and memory with which the cold system can work.

The study is concerned specifically with how political partisans, people who are committed to one political party, and specifically to one candidate, utilize the "hot" and "cold" systems in reasoning when presented with "threatening" information about that candidate. Utilizing imaging work like that of Goel and Dolan (discussed in yesterday's post), Westen et al. predict that when partisans are presented with threatening information about their candidate, they will see increased activation in areas associated with "hot" cognition, specifically the ventromedial prefrontal cortex and the anterior cingulate cortex (an area associatwiththe emotion and reward), along with suppression of activity in areas associated with "cold" cognition, such as the lateral prefrontal cortex.

During the 2004 presidential campaign, Westen et al. recruited 28 right-handed men between the ages of 22 and 55 who were self-reported "committed Republicans or Democrats." They presented each participant with a series of statements either about their favored candidate (same party condition), about the opposing candidate (opposing party condition), or about a "politically neutral" individual (Tom Hanks, Hank Aaron, or William Styron; neutral party condition). I'll let Westen et al. describe the statements (p. 5):

Each statement set consisted of seven slides presenting verbal material, designed to present a clear contradiction between the target person'’s words and actions and then to resolve that contradiction... Slide 1 presented an initial statement, usually a quote from the target individual. Slide 2 presented a contradictory statement suggesting that the target'’s words and actions were inconsistent. Slide 3 asked subjects to consider whether the target's “statements and actions are inconsistent with each other,” and Slide 4 asked them to rate the extent to which they agreed that the target'’s words and deeds were contradictory, from 1 (strongly disagree) to 4 (strongly agree) using a four-button pad. Slide 5 presented an exculpatory statement that explained away the inconsistency. Slide 6 then asked subjects to consider whether the target'’s “statements and actions are not quite as inconsistent as they first appeared.” The final slide asked them once again to rate the extent to which they agreed with this statement, using the same 4- point scale.

The key slides for the data analysis are those that present the contradiction (Slide 2) and those that present an exculpatory statement (Slide 5). Both slides present what are, in fact, contradictions to earlier information. The difference between the two is that, iin the same party condition, the contradiction slide gets inachievingof acheiving a desired conclusion, and thus should elicit motivated reasoning by triggering a negative emotional response, while the exculpatory statement allows the participant to arrive at his desired conclusion using cold cognition exclusively. Thus, based on the predictions described above, we would expect there to be more activity in the "hot" system areas while reading the contradiction slide than while reading the exculpatory slide. We would also expect their to be more activity in the "hot" system after reading the contradiction slide in the same party condition than in the neutral condition.

To test these predictions, Westen et al. performed four contrasts using the subtraction method. The first contrast involved subtracting the brain activity observed during the contradiction slide in the neutral condition from the activity observed during the contradiction slide in the same party condition. The results of this contrast can be seen in this figure (Figure 3 from Westen et al.):

As the figure indicates (in case you couldn't tell), this is what they found (p. 7, all emphasis mine):

[P]rocessing emotionally threatening information about one's preferred candidate relative to a neutral target activated distributed sites in medial prefrontal cortex, including particularly the ventral ("affective"”) subdivision of the [anterior cingulate cortex] but also the more rostral ("“cognitive"”) subdivision. Also activated were a small superior medial prefrontal region and a larger ventromedial region of [the prefrontal cortex] associated with affective processing. The other notable finding was a large area of activation in the posterior cingulate cortex (along with coextensive regions of the precuneus and inferior parietal cortex), associated in prior studies with neural information processing related to social emotions, moral evaluations, and judgments of forgivability.

To summarize the findings of the first contrast: relative to the neutral condition (let me repeat that again, for Meyer and anyone who read him, relative to the neutral condition), they found increased activation in areas associated with emotion and affect while reading the contradiction slide in the same party condition. Note, however, that they also found some increased activity in areas associated with cognitive processing, as well.

The second contrast tested the prediction that there would be a difference between the contradiction and exculpatory slides within the same party condition. Thus, the activity observed while reading the exculpatory slides was subtracted from the activity observed while reading the contradiction slide, all within the same party condition. This is what they found (p. 8, the figure is Figure 5 from Westen et al., all emphasis mine):

The contrast analysis showed activations in the left lateral inferior frontal cortex and left insula (not shown: maximum at -36, -18, 18), both consistent with processing of negative affect. Also seen were activations in the inferior orbitofrontal cortex (gyrus rectus) bilaterally, indicative of emotion processing as well as the precuneus (suggesting evaluative judgments, as above). The only other prominent activations were bilateral activations in the parahippocampal gyrus and extending to the hippocampus, perhaps indicative of efforts to generate solutions (rationalizations) based on memory retrieval. We again observed no differential activation of [dorsal lateral prefrontal cortex], suggesting that motivated reasoning did not engage regions previously linked with conscious attempts to reason, suppress information, or regulate affect.

To summarize, in the same party condition they found increased activation in response to the contradiction, relative to the exculpatory statement, in areas consistent with negative emotional experiences, evaluation, and according to their speculative interpretation, "efforts to generate solutions (rationalizations) based on memory retrieval). Both of these contrasts are thus consistent with the hypothesis that partisans are engaged in motivated reasoning, i.e., biased interpretations, evaluations, and memory searches, in response to threatening information about their favored political candidate (I won't get into the third contrast, because it's a bit complex, but its results are consistent with the first two).

The fourth contrast, which many reporters and bloggers seem to have found incredibly interesting, was, the author admitted, more exploratory than the first three. It was designed to test the hypothesis that the motivated reasoning that occurred while reading the contradictory slide would reduce negative affect. They thus contrasted the activity observed while reading the slide (slide 3) that was shown after the contradictory slide (which was slide 2). If motivated reasoning serves to decrease negative emotional responses to the threatening material, then we would expect a decrease in the activation in areas associated with negative emotional responses (e.g., the left insular cortex). And that is, in fact, what they found: relative to the contradictory slide, activation in areas associated with negative emotions, such as the insular cortex and the lateral orbital frontal cortex decreased significantly. Activation was also seen in the anterior cingulate cortex (suggesting the processing of emotion, though not necessarily negative emotion), and in the left inferior regions of the parietal lobe (suggesting "effortful processing," or as Westen et al. speculate, "rationalization). Finally (and this is what the reporters and bloggers found so sexy), they also found increased activation in the ventral striatum, an area associated with the brain's reward system.

Summarizing all that jargon: when looking at the slide that follows the activation of motivated reasoning, they found decreased activation in areas associated with negative emotions, but continued processing of emotional information, along with areas potentially associated with rationalization (attempting to come up with a post hoc explanation for arriving at one's desired conclusion), and activation indicating that overcoming the threatening information and arriving at the desired conclusion (and the participants did arrive at the desired conclusion, as indicated by their ratings of the candidates) is rewarding. How was this reported by Meyer and others? Overcoming the threat activates the same areas of the brain that are activated by drug use and sex! Well, duh. So is any other rewarding experience (like, say, correctly classifying a chair as a chair).

OK, now we're in a position to see where Meyer went astray. First, he has no idea what the photos of brain activity data mean. If you look at the photos above you will see, as Meyer's said, little or no activity in the parts of the brain associated strictly with the "cold" system, or as Meyer's nonsensically put it, the "thinking parts of their brains." You know why you don't see that activity? BECAUSE THE PHOTOS ARE OF THE ACTIVITY EXCLUSIVE TO THE CONTRADICTION CONDITION! It is the photo of the activity observed after subtracting the activity from the neutral or exculpatory conditions. You don't see activity in the "thinking parts of their brains" because that activity was subtracted out for the analysis! Anyone, and I mean anyone, who knows anything about neuroscience would have recognized that immediately. But Meyer writes an article read by thousands, and picked up by who knows how many bloggers, who were then read by who knows how many thousands more (Leiter's blog alone gets thousands of readers per day), in which he tells everyone that partisans aren't using the "thinking parts of their brains." When Westen et al. use words like "absence," they don't mean that the "thinking parts" of the brain are shut off, they mean that motivated reasoning itself doesn't engage those parts any more than they are already engaged during cold cognition (e.g., in the neutral conditions). But trust me, the thinking parts of the brain are still working just fine, and they can even override the emotional areas activated by motivated reasoning, if the facts are overwhelming. Partisans, then, are not completely irrational. They are perfectly capable of objectively reasoning about even their own candidates, if given the right incentives and information. But as happens with all of us, when their "hot" systems were activated, it biased their own reasoning processes, without them even knowing it.

Meyer argues in his article that these findings may make political persuasion impossible. It should be obvious, by now, that he's full of shit. Nothing about the Westen et al. results implies, in the least, that persuasion is impossible, even in the case of steadfast political partisans. It does imply that when strong emotions are associated with social concepts, in this case political candidates, persuasion will be more difficult. You will have to help the cold cognition system to overwhelm the hot cognition system by presenting partisans with facts that they can't rationalize. That's easier said than done, of course, but the fact that the U.S. population, once staunchly in favor of both Bush and the Iraq War, have begun to change their opinions of both in the face of overwhelming evidence, is proof positive that it is not impossible. One thing is for certain, though: it's impossible to get people who have no clue what they're talking about to stop writing about science.

One more thing. I think it's important, whenever we talk about cognitive neuroscientific studies, to note their limits. The subtraction method is full of problems, not the least of which is that just because an area doesn't increase or decrease its activity in a targeted condition relative to the condition that is subtracted from it, does not mean that those areas are not performing vital functions exclusive to the processes activated by the targeted condition. You should always take cognitive neuroscience with a salt truck-full of grains of salt. Think of the studies, and their interpretations, as interesting speculations that may be borne out by future research. But don't, I repeat do not, treat them as scientific gospel.

¹Westen, D., Blagov, P.S., Harenski, K., Kilts, C., & Hamann, S. (Unpublished Manuscript). An fMRI study of motivated reasoning: Partisan political reasoning in the U.S. Presidential election.