Research on political judgment and decision-making has converged with decades of research in clinical and social psy- chology suggesting the ubiquity of emotion-biased motivated reasoning. Motivated reasoning is a form of implicit emotion regulation in which the brain converges on judgments that minimize negative and maximize positive affect states associ- ated with threat to or attainment of motives. To what extent motivated reasoning engages neural circuits involved in ‘‘ cold ’’ reasoning and conscious emotion regulation (e.g., suppres- sion) is, however, unknown. We used functional neuroimag- ing to study the neural responses of 30 committed partisans during the U.S. Presidential election of 2004. We presented subjects with reasoning tasks involving judgments about in-

INTRODUCTION

In political science, cognitive science, economics, law, and business, the predominant models of judgment and decision-making today are ‘‘bounded rationality’’ models (Simon, 1990). These models suggest that people are rational within limits imposed by cognitive shortcuts and heuristics (Westen, Weinberger, & Bradley, in press; Gigerenzer & Selten, 2001; Kahneman & Tversky, 2000). In political science, a long-standing body of research on ‘‘partisan’’ biases in political judgment (e.g., Taber, Lodge, & Glathar, 2001; Campbell & Converse, 1960) points to another set of limits to rational judgment im- posed by emotion-biased or motivated reasoning (i.e., reasoning biased to produce emotionally preferable con- clusions; Kunda, 1990; Lord, Ross, & Lepper, 1979). Motivated reasoning can be viewed as a form of implicit affect regulation in which the brain converges on solu- tions that minimize negative and maximize positive affect states (Westen & Blagov, in press; Westen, 1985, 1994, 1998). Freud (1933) described such processes decades ago, using the term ‘‘defense’’ to denote the processes by which people can adjust their cognitions to

Emory University

formation threatening to their own candidate, the opposing candidate, or neutral control targets. Motivated reasoning was associated with activations of the ventromedial prefrontal cor- tex, anterior cingulate cortex, posterior cingulate cortex, in- sular cortex, and lateral orbital cortex. As predicted, motivated reasoning was not associated with neural activity in regions previously linked to cold reasoning tasks and conscious (explicit) emotion regulation. The findings provide the first neuroimaging evidence for phenomena variously described as motivated reasoning, implicit emotion regulation, and psy- chological defense. They suggest that motivated reasoning is qualitatively distinct from reasoning when people do not have a strong emotional stake in the conclusions reached. &

avoid aversive feelings such as anxiety and guilt. We use the term motivated reasoning here because of its wide- spread use (although, strictly speaking, all reasoning is typically motivated by emotions such as interest, excite- ment, anxiety, etc.; see Marcus, 2002; Westen, 1985).

Neural network models of motivated reasoning sug- gest that in affectively relevant situations, the brain equilibrates to solutions that simultaneously satisfy two sets of constraints: cognitive constraints, which maxi- mize goodness of fit to the data, and emotional con- straints, which maximize positive affect and minimize negative affect (Westen, Feit, Arkowitz, & Blagov, 2005; Thagard, 2003; Westen, 1998). Decision theorists have long argued that people gravitate toward decisions that maximize expected utility (or in emotional terms, that optimize current or anticipated affect; Simon, Krawczyk, & Holyoak, 2004; Mellers, 2000; Westen, 1985). Contem- porary views of motivation similarly emphasize approach and avoidance systems motivated by positive and nega- tive affect (Carver, 2001; Davidson, Jackson, & Kalin, 2000). The same processes of approach and avoidance, motivated by affect or anticipated affect, may apply to motivated reasoning, such that people will implicitly approach and avoid judgments based on their emotional associations.