Saturday, July 30, 2005

Fear, Race, and the Unconscious

The amygdala is an often misunderstood brain region. It's closely associated with emotion in general, as well as many emotionally-mediated cognitive and perceptual processes, including face recognition and even affect-mediated memory recall. Despite its many functions, it's often treated in casual discussions and the popular press as the source of one particular emotion: fear. For example, in the recent discussion of the neuroimagining studies of males and females during orgasm, the fact that deactivation was observed in the amygdala was discussed almost exclusively as a decreased fear response, even though there's no other evidence (e.g., in the form of skin conductance responses) that fear responses were decreased. It's dangerous, then, to talk about fear and the amygdala in relation to certain controversial topics. But dangerous is my middle name (seriously, it's on my birth certificate), so I'm going to chance it anyway. Readers should keep in mind the difficulty in interpreting results associated with neuroimagining data indicating increased (or decreased) activation in areas with multiple functions.

While the amygdala is associated with many emotions, there is growing evidence that it is closely associated with fear and fear-related conditioning in nonhuman animals and humans. First, while bilateral damage to the amygdala produce general emotional deficits in humans1, bilateral amygdala lesions are specifically associated with deficits in fear conditioning and produce severe deficits in the recognition of fearful faces2 and vocal intonation3 associated with fear relative to deficits in recognizing those displaying or expressing other emotions (e.g., happiness). Several imaging studies have also observed increased amygdala activation during fear conditioning4, as well as the recognition of fearful faces5, and the visual or cognitive6 representation of fearful stimuli. It's pretty clear, then, that the amygdala is a central part of the system that produces fear recognition and conditioning in humans.

Another line of research, born of the controversy surrounding the claim that the Implicit Association Test (IAT) may reveal implicit racial attitudes, has investigated the neural activity underlying such attitudes. In one study7, researchers presented participants with photos of unfamiliar white and black faces while they were in an fMRI machine. They also gave the participants the IAT, as well as another implicit measure (startle eyeblink), and asked them to describe their attitudes toward African Americans. They found that activation in the amygdala increased for black faces, and was correlated with implicit attitudes (more activation in the amygdala corresponded to more negative associations with African Americans), but not with explicitly expressed attitudes8.

A second study presented white and black participants with photos of white and black faces while they were in an fMRI, and measured amygdalar activation over time9. For ingroup faces (white faces for white participants, black faces for black participants), amygdala activation decreased over time as several ingroup faces were viewed. For outgroup faces, no such decrease in activation was observed. Thus, after the presentation of several ingroup and outgroup faces, amygdala activation was low for ingroup faces but high for outgroup faces.

In a third imaging study, participants were again presented with white and black faces, this time for two different durations, short (30ms) and long (525 ms)10. This study also showed higher amygdala activation for black faces than white faces for white participants, but only for the short durations. At longer durations, there was no difference in amygdala activation between white and black faces, while there was increased activation for black faces in cortical areas associated with conscious control. The researchers argue that this explains the findings from the first study, claiming that initially, implicit attitudes are activated, but longer exposure allows conscious, or explicit attitudes to be activated and to inhibit the implicit associations.

OK, so I told you all of that to tell you about this. These three studies indicate that the processing of faces from members of racial outgroups results in activation in the amygdala. This may indicate that fear responses and racial attitudes share neural systems. There is in fact some behavioral evidence for this. According to terror management theory, which has produced a ton of supporting data over the last decade or so, fear (especially fear of death), produces exaggerated ingroup-outgroup responses11. In particular, negative evaluations of outgroup members are heightened, as are positive evaluations of ingroup members. These findings motivated a just-published study on the relationship between fear conditioning and racial attitudes12. The researchers presented white and black participants with white and black faces, as in the imaging studies, but added a twist. During the presentation of either the black or white faces (varying which face across participants), the participants also received a mild electric shock. After a few trials, they exhibited a conditioned response (measured by skin conductance) to the faces associated with the shock, but not to the faces that were not associated with a shock. After the conditioning phase, the participants were again presented with the faces, but without any shock. The result of interest was the extinction rates for ingroup vs outgroup faces. For some stimuli (the researchers mention snakes and spiders), fear responses seem to elicit prepared conditioning responses, so that when those stimuli are paired with an aversive experience (like electric shocks), extinction takes longer than it does for stimuli that aren't preconditioned to be associated with fear (they use butterflies and birds). If the extinction of conditioned fear responses take longer for racial outgroup faces than for ingroup faces, it would indicate that there is a preconditioned association with fear for those faces. This is in fact what they found: for white participants, the fear responses took significantly longer to extinguish for black faces than white faces, and the opposite was observed for black participants. The only participants who didn't, on average, show this effect were those who had a lot of interracial dating experience (i.e., positive experience with racial outgroup members).

All of this information taken together presents a fairly clear picture: fear and negative racial attitudes are closely associated. However, the authors of the last study make it clear that they don't believe their data implies an innate fear of racial outgroup members, as some have argued such prepared responses to snakes, spiders, and other fear-related stimuli indicate. Here is how they put it:
In other words, because of its relatively recent emergence as an important dimension in human social interaction, race inherently cannot be the basis of the outgroup preparedness result. Instead, it is likely that sociocultural learning about the identity and qualities of outgroups is what provides the basis for the greater persistence of fear conditioning involving members of another group. Most notably, individuals acquire negative beliefs about outgroups according to
their local cultures, and few reach adulthood without considerable knowledge of these prejudices and stereotypes. It is plausible that repeated exposure to information about outgroups might prepare individuals to fear newly encountered outgroup members.
Studies like these will can help people in combating negative racial attitudes. The empirical evidence I've discussed here shows that conscious attitudes can inhibit implicit (likely unconscious) negative racial attitudes, and that positive exposure to members of racial outgroups (as in those participants who had interracial dating experience in the last study) can create positive implicit attitudes towards the members of other races. This means that people are less likely to act on implicit negative attitudes towards members of other races when they have the time and cognitive resources to allow their explicit attitudes to take control. It also means that negative associations can be combated through positive exposure to members of other races. In other words, diversity is important, because it allows for positive exposure to the members of other races. Hopefully future research will find other ways to diminish the fear-related emotional responses to other races as well.

1 Aggleton, J.P. (1992). The functional effects of amygdala lesions in humans: A comparison with findings from monkeys. In J.P. Aggleton (ed.), The Amygdala: Neurobiological Aspects of Emotion, Memory, and Mental Dysfunction. New York: Wilely-Liss, 485-503.
2 Calder, A.J., Young, A.W., Rowland, D., Perrett, D.I., Hodges, J.R., Etcoff, N.L. (1996). Facial emotion recognition after bilateral amygdala damage: differentially severe impairment of fear. Cognitive Neuropsychology, 13, 699-745.
3 Scott, S.K., Young, A.W., Calder, A.J., Hellawell, D.J., Aggleton, J.P., Johnson, M. (1997). Impaired auditory recognition of fear and anger following bilateral amygdala lesions. Nature, 385(6613), 254-257.
4 Quirk G.J., Armony, J.L. & LeDoux, J.E. (1997). Fear conditioning enhances different temporal components of tone-evoked spike trains in auditory cortex and lateral amygdala. Neuron, 19, 613-624; LaBar, K.S. Gatenby, J.C., Gore, J.C., LeDoux, J.E., & Phelps, E.A. (1998). Human amygdala activation during conditioned fear acquisition and extinction: a mixed trial fMRI study. Neuron 20, 937-945.
5 Morris, J.S., Frith, C.D., Perrett, D.I., Rowland, D., Young, A.W., Calder, A.J., Dolan, R.J. (1996). A differential neural response in the human amygdala to fearful and happy facial expressions.
6 Phelps, E.A., O'Connor, K.J., Gatenby, C., Gore, J.C., Grillon, C., & Davis, M. (2001). Activation of the left amygdala to a cognitive representation of fear. Nature Neuroscience, 4, 437-441.
7 Phelps, E.A., O'Connor, K.J., Cunningham, W.A., & Funayama, E.S. (2000). Performance on indirect measures of race evaluation predicts amygdala activation. Journal of Cognitive Neuroscience, 12(5), 729-738.
8 Interestingly, in their second experiment, they used familiar (famous) and positively regarded black faces. In this case, they didn't observe increased amygdalar activation to black faces. This confirms the "safe black person" phenomenon that is so common these days.
9 Hart, A.J., Whalen, P.J., Shin, L.M., McInerney, S.C., Fischer, H., & Rauch, S.L. (2000). Differential response in the human amygdala to racial outgroup vs ingroup face stimuli. NeuroReport, 11(11).
10 Cunningham, W.A., Johnson, M.K., Raye, C.L., Gatenby, J.C., Gore, J.C., & Banaji, M.R. (2004). Separable neural components in the processing of black and white faces. Psychological Science, 15(12), 806-813.
11 Solomon, S., Greenberg, J., Schimel, J., Arndt, J., & Pyszczynski, T. (in press). Human awareness of death and the evolution of culture. In M. Schaller and C. Crandal (eds.), The psychological foundations of culture. Erlbaum.
12 Olsson, A., Ebert, J.P., Banaji, M.R., & Phelps, E.A. (2005). The role of social groups in the persistence of learned fear. Science, 309, 785-787.

Tuesday, July 26, 2005

Moral Psychology I: Where Is Morality in the Brain?

Long ago, in response to which post I do not remember, Brandon requested posts on topics related to the philosophy of David Hume, including causal reasoning and moral psychology. I responded with a post on causal reasoning, and left it at that. The primary reason I left it at that is that while there is a ton of research on causal reasoning (it's one of the big areas in cognitive psychology), and it would be impossible to summarize it on a blog, the research at least fairly straightforward, and I've actually done research in the area, so I feel pretty comfortable with the literature. The cognitive scientific literature on moral psychology, on the other hand, is hopelessly muddled. This is due in part to the fact that most of the research on the topic has taken place outside of the core cognitive sciences, in fields like developmental, social, and clinical psychology. It's also a product of the fact that much of the empirical and theoretical work on moral psychology within the cognitive sciences is dominated by conflicts of ideology. You see, cognitive science, while it does have a dominant paradigm (the computationalist paradigm -- though its dominance is diminishing fairly rapidly), is a fairly fractured discipline. You've got the old computer metaphor people, the connectionists (who sometimes, but not always, mingle with the computer folks), the dynamic systems people (who sometimes mingle with the connectionists), the embodied cognition folk (who sometimes hang out with the dynamic systems or connectionist people), and then within each of these, a host of subgroups, all of whom have pictures of the mind that may vary a little or a great deal. Thus there is a growing debate between "rationalists," who are less wedded to old school rationalism than they are to the symbolist (computer) view of mind, and the "intuitionists" who are less anti-rationalist than they are pro-connectionist or anti-symbolist. It's a big mess, full of speculation and less-than-empirically-based conclusions, and as a result, it's hard for little old me (someone with symbolist and embodied cog. tendencies) to sort things out.

(Un)Fortunately for you, I've decided to give it a go anyway. I'm motivated in part by some posts on related topics that you probably didn't read, but which raised interesting questions. For example, there was this post on the work of Jonathan Haidt, which you will soon find out is the leading edge of the "intuitionist" picture of moral judgment. There was also this post which made some pretty strong claims about the ways in which philosophers should use moral psychology to constrain ethical theories. And there was this post, which brings to mind a question that I tend to ask about a lot of things: what makes a person an expert? In this case, what makes a person a moral expert? Hopefully, by the time I'm done, you will have some idea of what the intuitionist view of moral judgment is, in what ways moral psychology and moral philosophy should interact, and who, if anyone, might be a moral expert. There are a bunch of other issues that I'll try to touch on as well. Is morality a natural kind in the brain, or to use a stranger label, a cognitive kind? How much influence does conscious reasoning have on our moral judgments and behavior? How does communication affect moral judgment? These and other difficult questions will be answered definitively in these posts. OK, so maybe not definitively, or at all, but I'm at least going to touch on them.

As I've done in the past, I'm going to start with the neuroscience. I usually do this for two reasons. First, cognition happens in the brain. This may come as a shock to some of you, especially if you're still clinging to some sort of 16th century mental dualism (a note just for you folks: there is no evidence that the pineal gland is involved in moral judgment), but it's true. Second, while neuroscience may present the best and hardest evidence in some of the lower-level areas of cognitive science, like perception, when we get into higher-level cognition, it often gives us the weakest, or at least the most equivocal evidence we have. This is especially true for areas that are wrapped up in social contexts, as moral judgment certainly is. It would be a shame to end on a weak note, so instead I'll start with the neuroscience and build from there.

The Neuroscience

The first line of neuroscientific work comes from studies of patients with damage to the ventromedial prefrontal cortex (as did Phineas Gage, whose skull and death mask are on the left). Much of this work has been done by Antonio Damasio and his colleagues in the development and testing of the somatic marker hypothesis. This hypothesis states that bodily reactions to stimuli create affective markers that influence our decision making1. Thus we would expect that individuals with lesions to certain areas associated with somatic markers will display less optimal decision making behavior, because the influence of affect on their decisions will be diminished, or less immediate. The ventromedial prefrontal cortex is thought to be one of the key areas associated with somatic markers.

The evidence for the somatic marker hypothesis comes primarily from the Iowa Gambling Task (see this paper for a description of the initial experiments, and the IGT itself). This task involves four decks of cards (A, B, C, and D) that each give an immediate reward when flipped over. Participants are given a "loan" to start, and told to earn as much money as possible by flipping over cards from one of the four decks. The A and B decks give a large immediate reward ($100), while the C and D decks give a smaller immediate reward ($50). By reward alone, then, we should pick cards from A and B. However, in each deck there are some cards that also result in a penalty. The size of the penalty in decks A and B is such that if you pick cards from those decks, you will end up with a net loss, while the penalty in decks C and D allows you a net gain. Thus, the optimal strategy is to pick from C and D. Bachara et al. (the paper linked above) found that normal participants (those without brain damage) quickly learn this, even though they are not consciously aware of the knowledge. In fact, when Bachara et al. measured skin conductance responses (which signal certain affective responses), normal participants at first had them after selecting cards, with different SCRs for rewards and punishments. After a short time, though, they began to have "anticipatory" SCRs, which resembled punishment responses when picking from A or B, or reward responses when picking from C or D. These anticipatory responses appeared before they expressed knowledge that A and B were bad decks.

Bachara et al. also tested patients with damage to the prefrontal cortex, specifically the ventromedial prefrontal cortex. These patients were horrible at the task, tending to pick from A and B very often, and thus ending with net losses. This was the case even when they expressed knowledge that A and B were bad choices. Furthermore, these patients never displayed anticipatory skin responses, indicating that the reason for their poor choices was the lack of an affective reaction influencing their behavior. From this, Damasio and others have concluded that this affective reaction is necessary for making rational choices. It's important to note, as well, that these patients tend to have a difficult time behaving appropriately in social situations, as well as understanding the behavior of others, even though they retain normal IQs and can verbalize the social knowledge they seem unable to act upon2.

What does this have to do with morality? Well, patients like those studied by Damasio developed brain damage in adulthood, and thus while they make bad decisions and often act confused in social situations, they tend to behave morally. This appears to be due to the fact that they had developed behavioral patterns and associations over the course of childhood that allow them to continue to behave morally3. What happens if the damage occurs early in development, and thus before moral knowledge is learned? To address this, Steven Anderson and his colleagues (who included Damasio, who is the last author on pretty much every study anyone conducted on any topic from 1994-2005) studied two patients whose prefrontal cortex damage had occurred prior to 16 months of age4. These patients showed many of the decision-making deficits that characterize prefrontal damage in adults, but they also showed much, much more. They were unable to learn social conventions and moral rules, and showed poor moral reasoning, and were just all around bad people. They lied, cheated, stole, were terrible parents, and for all of this, they showed no guilt or regret. They were so bad that Anderson et al. put it in these strong terms:
Thus early-onset prefrontal damage resulted in a syndrome resembling psychopathy.
Psychopathy! In fact, studies of the emotional responses of psychopaths have shown that they, like patients with prefrontal cortex damage, show a lack of emotional response to material to which normal participants generally show strong emotional reactions5. Thus, it appears that the prefrontal cortex, and especially the right ventromedial prefrontal cortex. In fact, a recent study of patients who developed lesions of the right ventromedial prefrontal cortex in adulthood showed that, like psychopaths, they show impaired empathic responses, or the lack of empathic responses altogether6.

So, there is good evidence from lesion/brain damage studies that the prefrontal cortex, and the right ventromedial prefrontal cortex specifically, is involved in the emotional and empathic responses that are associated with moral judgment. Moll et al. 7 presented participants with moral and nonmoral photos while they were in an fMRI machine. In this and similar studies (some of which involved moral and nonmoral sentences instead of photos), they found that the moral stimuli caused activation in several areas, including the ventral and medial prefrontal cortex (VMPC), as in the brain damage studies, later superior temporal cortex (STC), right posterior superior temporal sulcus (STS), frontal pole, medial frontal gyrus, right cerebellum, left orbifrontal cortex, left precuneus, and posterior globus pallidus8. These areas are associated with social reasoning and theory of mind (VMPC, medial frontal gyrus, left precuneus), social perception (STC), general social cognition (STS), as well as affect, reward, emotional memory, and several other social and emotional functions.

From the study of brain damaged patients by Damasio and others, and the imaging studies of Moll, it's quite clear that areas associated with social reasoning, behavior, and memory, and areas associated with affect (both positive and negative) are important for moral reasoning. It's not clear, however, that areas of the frontal cortex associated with deliberative, conscious reasoning are involved.

There is another study, though, in which activation in such areas was observed during the performance of moral judgments. Greene et al. (notice these are all et al., because it takes 500 people to conduct a neuroscience experiment) used methods similar to those in the Moll experiments, in which people were exposed to moral and nonmoral stimuli while in an fMRI machine8. However, they added a twist. Some of the moral stimuli involved personal involvement, while others did not. They called them "personal" and "impersonal," respectively. The example Greene uses to illustrate the difference is the classic Trolley Problem. In one version of the trolley problem, a run-away train (or trolley) is hurtling towards five people who are unknowingly in its path, and who will certainly be killed unless the trolley is diverted. You can flip a switch and cause the trolley to change tracks. On the other set of tracks, there is one person who will be killed if you flip the switch. Should you flip the switch? In another version, the trolley is heading towards the same five people, but instead of flipping a switch, the only way to divert the trolley is to through a large man onto the tracks in front of it. Should you do so? Most people answer yes to the first dilemma, and no to the second. The difference between these two problems, according to Greene et al., is that the first problem is impersonal, and the second personal. Here is Greene's more formal definition of the two different kinds of problems:
A moral violation is personal if it is: (i) likely to cause serious bodily harm, (ii) to a particular person, (iii) in such a way that the harm does not result from the deflection of an existing threat onto a different party. A moral violation is impersonal if it fails to meet these criteria. (Greene & Haidt, p. 519)
Consistent with their distinction, they found two distinct activation patterns for personal and impersonal moral stimuli. The personal moral stimuli, like the second version of the trolley problem, caused activation in several of the areas that were active in Moll's studies, including the medial frontal gyrus, posterior cingulate gyrus, angular gyrus, and the superior temporal sulcus, areas associated with affect. The impersonal moral stimuli, however, showed activation in areas that are active during cognitive tasks, particularly those involving working memory, but less active during emotional responses, including the middle frontal gyrus, dorsolateral prefrontal cortex, and parietal lobe. These areas were also active in the processing of nonmoral social stimuli. Thus, it appears that the processing of personal moral stimuli takes place largely in the prefrontal and precortical social and affective areas of the brain, while nonmoral social and impersonal moral stimuli utilize cognitive areas of the brain.

The take home message of this is hard to pin down. It appears that areas of the brain associated with social reasoning and affect are heavily involved in moral reasoning, at least when we are personally invested in the situation, while cognitive processing, or what we would usually call moral reasoning, is less involved in these contexts, but does the bulk of the work when we're making moral judgments in situations in which we are not invested. It is also quite clear that moral judgment does not take place in a single area or group of highly interconnected areas, but is instead spread throughout much of the brain. Each of these brains is associated with a variety of functions that are related to moral judgment, such as social reasoning. We can reasonably conclude, then, that there is no single moral center in the brain, and that moral judgment is not a "cognitive kind," but is instead the product of multiple affective and social reasoning abilities that are activated in particular social contexts. We should be careful, though, in interpreting these results any further. Looking at photos or reading sentences with moral content while in a cramped fMRI machine is much different from participating in real world social interactions in which moral decisions upon which we must act are necessary. It may be that the imaging studies conducted so far present only a fraction of the neural picture of moral judgment.

In the next post, I'll look at the behavioral data, and get into the rationalist-intuitionist debate that the neuroscientific data has, in large part, sparked. After that, I'll start to try to answer some of the questions, looking specifically at communication, expertise, and the role of moral psychology in moral philosophy.

The second installment is here.

1 Damasio, A. (1994). Descarte's Error: Emotion, Rationality and the Human Brain, Grosset Books: New York, Putnam.
2 Ibid.
3 Greene, J.&d Haidt, J. (2002) How (and where) does moral judgment work? Trends in Cognitive Sciences, 6(12), 517-523.
4 Anderson, S.W., Bechara, A., Damasio, H., Tranel, D., Damasio, A.R. (1999). Impairment of social and moral behavior related to early damage in human prefrontal cortex. Nature Neuroscience, 2(11), 1032-1037.
5 Greene & Haidt, 2002.
6 Shamay-Tsoory, S.G., Tomer, R., Berger, B.D., & Aharon-Peretz, J. (2003). Characterization of empathy deficits following prefrontal brain damage: The role of the right ventromedial prefrontal cortex. Journal of Cognitive Neuroscience, 15(3), 324-337.
7 Moll, J., Oliveira-Souza, R., Eslinger, P.J., Bramati, I.E., Mourão-Miranda, J., Andreuiolo, P.A., & Pessoa, L. (2002). The neural correlates of moral sensitivity: A functional magnetic resonance imaging investigation of basic and moral emotions. The Journal of Neuroscience, 22(7), 2730-2736.
8 Greene, J.D., Sommerville, R.B., Nystrom, L.E., Darley, J.M., & Cohen, J.D. (2001). An fMRI investigation of emotional engagement in morajudgmentnt. Science, 293, 2105-2108.

Monday, July 25, 2005

Reading Group Update

Today I sent out an email to everyone who expressed interest in participating in the reading group. If you gave me your email address, but didn't receive an email, please let me know. Also, if you're interested but haven't yet said so, let me know, and I'll forward the email to you. The email contains some questions designed to help us iron out the details of the group so that we can get started. Some people have already replied, and one of the really good suggestions was that I start a yahoo or google group. So I just created a yahoo group. I've never done yahoo groups before, but it seems easy enough. The group is called Cognitive Science Blog Reading Group, and the following are the email addresses that yahoo gave me when I created it:
Post message:
If you want to go ahead and subscribe, feel free. You can subscribe even if you don't want to participate, though I think anyone can read the postings even without subscribing.

Someone also suggested a cog sci wiki, which I think would be an excellent idea if we could get other cognitive scientists to participate, but I don't know if it's very practical right now. However, I do highly recommend the MIT Encyclopedia of the Cognitive Sciences. All of the entries were written by experts, and contain great references. It used to be available for free on the internet, but when they finally got around to publishing it in book form, they made the online version subscription only. The book is not cheap, either, at $68. So you may want to just ask me, or someone else in the field, if you have questions about specific topics or concepts. If I don't feel qualified to answer, then I'll know someone who is.

Friday, July 22, 2005

The Mind Is Stranger Than Fiction

A warning: I have no idea how to explain the research results that I'm about to describe, and neither John Bargh nor Mark Chen or Lara Burrows can offer a compelling explanation either. It may very well be that the results are due to the entirely mundane effects of increased and decreased arousal that have more to do with the nature of the tasks than the content of the primes, but it's impossible to tell. The "experiments" don't really test any explanations, they just demonstrate effects, as is the social psychologist's way. Still, the "experiments"are pretty cool, and the results are really weird, and sexy (they've made Bargh a minor celebrity), so I'm going to talk about them anyway.

We've known for some time that attitudes, including those we don't really want to have (e.g., racist, sexist, or homophobic attitudes), can be activated automatically (i.e., without intention), and often unconsciously, and that they can influence our subsequent thoughts and behavior. John Bargh and his colleagues have argued that this is only part of the story. They believe that specific "social-behavioral responses" can be automatically (and unconsciously) activated by situational factors in the way that attitudes can. In their 1996 paper, they refer to James' "principle of ideomotor action" (by the way, if all philosophy is merely a footnote to Plato, then all psychology is merely a footnote to James), and give the following quote from The Principles of Psychology:
We may lay it down for certain that every representation of a movement awakens in some degree the actual movement which is its object.
In other words, the mental representation of a movement increases the likelihood of the performing of that movement. Since movements, or behaviors, involve mental representations just as attitudes do, then it stands to reason that representing a behavior increases the likelihood of performing it. The thought primes the action. Furthermore, since thoughts can be primed by external information, behaviors can too. The most common examples of such an automatic stimulus-behavior link, these days, refer to aggression (though people sometimes draw really strange conclusions from that research), but you get the idea. In fact, because attitudes and other conceptual representations are closely associated with behavior representations, activating one can activate the other, and thus prime behaviors.

Thus, Bargh, Chen, and Burrows set out to demonstrate the existence of "automatic social behavior." They conducted three experiments, each targeting different behaviors. In the first experiment, they first gave participants a scrambled sentence test, which involves presenting five scrambled words and asking the participant to form a grammatically correct sentence out of four of them as quickly as possible. They developed three different lists of scrambled words, one of which primed the concept RUDE, another that primed POLITE, and a third that was neutral with respect to rudeness/politeness. The primes in these lists included adjectives, adverbs, or verbs that were associated with the concepts (e.g., brazen, aggressively, or disturb for RUDE, and considerate, patiently, and respect for POLITE). While the participant was completing the scrambled sentence test, the experimenter left and began talking to a confederate (an experimenter posing as another participant). When the participant finished, he or she came out of the room to look for the experimenter to receive instructions for the next task (as the experimenter had instructed). However, the participant always found the experimenter talking to the confederate. Bargh et al. then measured the time it took for the participant to interrupt the conversation between the experimenter and the confederate.

Guess what they found. Of the participants who did the RUDE version of the sentence test, more than 60% interrupted in under ten minutes (they cut it off at ten minutes -- can you imagine how frustrated some of those participants were after standing there for ten minutes?), whereas fewer than 20% of the POLITE-primed participants interrupted in that time. The neutral list participants were in between at around 40%. The RUDE participants also interrupted a full 3 minutes sooner than neutral participants, and almost 4 minutes sooner than the POLITE participants.

But it gets weirder. In experiment two, they looked at the behavioral effects of a particular stereotype, the "elderly stereotype." Once again, they gave participants the scrambled sentence test, but this time, one version containing words associated with the elderly (e.g., worried, old, lonely, and no I'm not making this up, Florida), and the other was age-neutral. The participant was then "partially debriefed" (i.e., lied to), and then told that he or she could go. The experimenter then timed the participants walk to the elevator down the hall, and wouldn't ya know it, the people who had been primed with the elderly words, thus activating the elderly stereotype (which apparently includes slowness) walked to the elevator more slowly. It took them a full second longer (8.28 seconds vs. 7.30 seconds for those who took the neutral test), on average, to get there.

Recognizing that there might be at least one alternative explanation for their findings (other than the automatic activation of a stereotype which then influences behavior), they conducted a second version of this experiment to make sure that the elderly word list wasn't depressing the participants, and that sadness was responsible for their slowness. They gave each participant a version of the Affect-Arousal Scale, and found that the elderly-primed participants were no more depressed than those in the neutral condition. Thus, sadness can't explain their results. Who knows what can?

But it gets even weirder. Brace yourselves for this one. In the third and final experiment, they presented participants with a "boring and tedious" task that took forever (each participant completed 130 trials). Before each trial, a picture of an African American male face, or a "young Caucasian male face" popped up on the screen subliminally. This was designed to activate racial stereotypes without the participants being aware of it. After the 130th trial, a message came up on the screen saying that there had been a problem in saving their data, and they would have to complete the experiment again. They videotaped the participants' reactions to this horrible news (if you've ever participated in one of those tedious psychology experiments, you will understand just how terrible that news is), while the experimenter rated the hostility of the reaction in first person. The video tapes were used so that two other independent raters could rate the hostility later on. The inter-rater reliability wasn't really very high, with the two independent raters agreeing about 64% of the time, but since they were using a 5-point scale, that's not surprising. They just averaged the ratings that they couldn't agree on after discussion (if one had given a 2, and another a 4, then they used 3). Once again, the stereotype seemed to automatically affect the participants behavior. Those who had seen the African American faces were rated as more than half a rating-point more hostile than those who had seen the white faces.

I wish I had something really smart to say to wrap this post up, summarize the findings, and at least attempt to explain them, but I don't. If these results show what Bargh and his colleagues think they do, then it is both very interesting, and very scary. The unconscious and automatic influence of pernicious racial, gender, or even age stereotypes on our behavior is not something pleasant to think about, and could ultimately have major social and political implications. However, no one really knows what's driving these experimental results, and to date, no one's really attempted to tease out the causes experimentally. All we have is these really freaky results from a 1996 paper, along with the other, often equally bizarre effects that Bargh and others have since demonstrated since then. We're no closer to knowing what's really going on than we were 9 years ago. I do know one thing, though. Reading entire social psychology papers makes me write more speculatively and less analytically. Is that yet another case of stereotypes automatically influencing behavior? (OK, that was a bit of low blow, but it was funny, right?)

POST SCRIPT: I'm not really trying to be mean to social psychologists. I just wrote a post the other day that favorbably cited the research of several social psychologists. As with all science, each piece of social psychological research should be evaluated individually. It should not be assumed that because some social psychologists don't actually test explanations, and just look for really sexy results intead, that all social psychologists do that.

POST POST SCRIPT: And no, I'm not just making these jabs because I'm anonymous. I have been saying these things for years, often quite loudly and within earshot of social psychologists. In fact, in the graduate course I took on social psychology, I spent an entire semester saying these things to a whole room of social psychologists. It's all in good fun, people, all in good fun.

Thursday, July 21, 2005

Spelke on Sex Differences in Math

In a comment on this post at Majikthise, commenter Nancy mentioned this great paper by Elizabeth Spelke on sex differences in mathematics, of which I was not previously aware. Some of you may have heard of Dr. Spelke because she recently debated Stephen Pinker. You may not be aware, though, that much of her recent work has been on the development of the different cognitive mechanisms underlying our mathematical abilities. Her work, along with that of Fei Xu, Karen Wynn, and Susan Carey, has shaped the debate in the literature on infant numerical cognition over the last several years. In other words, unlike Stephen Pinker (do you know what Pinker actually studies? I recommend this paper to get an idea) and some of the others who have commented on sex differences in mathematical ability, she knows what she's talking about. She's an expert on the cognitive capacities that underlie mathematical ability, and she's also an expert on the literature (and the author of much of it) within which evidence for sex differences, and innate sex differences in particular, would be found.

The paper, which I highly recommend you read if you've been at all swayed by the arguments hinted at by Summers, and later made by Pinker and many others, about the potential role of innate sex differences in cognitive abilities in creating the gender disparity in math and science departments, discusses the cognitive abilities that underlie mathematical cognition, and shows that in development, few, if any sex differences in math-related ability have been observed. In fact, where they have been observed they actually favor female children. She discusses research demonstrating that rather than differences in ability, differences in problem-solving strategies actually underlie the observed gender differences in math performance. Along the way, she debunks the oft-cited study showing that male infants prefer objects, while female objects prefer people; she points out that there is absolutely no evidence for the innateness of sex differences underlying the observed differences in math test scores (and argues that innateness is irrelevant to the actual debate over gender disparities among faculties in certain fields); she discusses the inadequacy of the SAT-M, the test on which most of the sex differences have been observed; and she discusses the actual performance of men and women at the top end of the curve. The last two allow her to debunk the oft-cited claim that men outperform women at the top end of the curve (and in the process, debunk the 12:1 ratio that is also frequently cited). In the end she concludes, correctly I think, that there is no evidence for sex differences in math-related cognitive abilities that can account for the gender disparity in math and science departments, while there is a wealth of evidence for gender equality in both primary and secondary mathematical abilities.

The paper has yet to be published, or even submitted, which means that it has yet to be peer reviewed. This may give some people pause, but as someone who knows the literature that she discusses fairly well, I can assure you that her factual claims are all true. You can evaluate the arguments she develops from these facts for yourself.

After you read that one, you might also want to read this very good book chapter by Spelke and one of my favorite cognitive scientists, Marc Hauser, on the evolution of mathematical ability. It would make a great addition to my list of writings on evolutionary psychology (as opposed to Evolutionary Psychology), as it explicitly avoids the primary sin discussed in the paper to which Lindsay linked in her post.

Tuesday, July 19, 2005

A Post That Serves As a Demonstration of Its Point

One of the things that I've lamented over the last few years is the loss of what might be called innocence in observing and interpreting people's behavior, a certain connection to the world out there. When I first started studying cognition, I looked at how people behaved, and used that to inspire my search for research and theories that might explain that behavior. There were problems with this approach. I was not yet sufficiently familiar with theory to interpret people's behavior in deep and systematic ways, and was thus limited to surface features. This sometimes made it difficult to find the relevant research, but it also made me evaluate theories and experiments based on my real-world experience with the behaviors that they were trying to explain.

By now, that innocence is long gone. It is difficult for me not to interpret people's behavior using my knowledge of research and theory. In other words, my approach to people's behavior is the opposite of it once was. Instead of going from behavior to theory, I go from theory to behavior. This has its advantages. I am able to interpret behavior at a much deeper, more systematic level. However, it also has its disadvantages. One of the most common criticisms of cognitive psychological research is that it is not connected with the real world. We cognitive psychologists lock ourselves in our labs and conduct strange experiments, in strange contexts, on university undergrads. Few of us perform experiments out in the real world, because we're worried about control. Before, I frequently noticed the disconnect between cognitive scientific explanations and real world behaviors. Now it's more difficult for me to notice them.

Interestingly, this is exactly what you would expect based on the research on the development of expertise. The primary difference between novices and experts concerns depth of processing. Novices tend to lack the relevant theoretical knowledge that would allow them to notice complex relational structures, and are thus limited to surface features when processing information. This is how novices solve physics problems, play chess, and perceive the information in x-ray photos, for instance. Experts, on the other hand, reason about information in their area(s) of expertise at a deeper, more relational level. Their theoretical knowledge allows them to immediately recognize deep relations between information that are not apparent on the surface, and thus to reason about that information at a deeper level. The development of expertise follows a pattern that moves from using surface features to understand theory to using theory to understand observed surface features based on underlying relations and principles. This is exactly what has happened to me.

Sometimes I wish that weren't the case. I was able to come up with some very interesting and novel ideas when I first started grad school, because I was coming at cognitive scientific problems from a naive perspective, and using my real-world experience to develop ideas about them. It's virtually impossible for me to do that now. Notice how I can't even write a blog post about it without invoking the theoretical explanation. I'm afraid that my ideas have suffered in some ways as a result.

Gender, Math, Stereotype Threat, and Testosterone

The recent debate of the remarks of a certain university president about gender differences in mathematical ability has focused on the role of innate factors in those differences. I've argued previously that there is no direct evidence for innate differences, though innate factors cannot be ruled out. There is, however, a great deal of evidence for social/environmental factors that influence the observed gender differences in performance on standardized math tests. Much of the recent research on social factors has focused on Claude Steele's concept of stereotype threat1. According to Steele (1999), stereotype threat involves "the threat of being viewed through the lens of a negative stereotype or the fear of doing something that would inadvertently confirm that stereotype" (p. 46). More technically, in contexts in which particular stereotypes are active, individuals who are members of the negatively stereotyped groups will be conscious of the content of those stereotypes, and this may negatively affect their performance. In the context of performance on standardized math tests, women who are currently aware of gender stereotypes related to math ability may experience anxiety related to the confirmation of those stereotypes, and as a result, their performance on the math tests will suffer.

There is a lot of evidence that stereotype threat can affect intellectual performance. Steele's early work on the topic looked at the affect of racial stereotypes on the intellectual performance of African Americans. However, he's found that even the performance of white males can be affected by stereotype threat if task-related stereotypes are active. Specifically, when the stereotype that Asian males have better math skills than white males is active, white men perform worse on standardized math tests than when the stereotype is not activated2. More recently, researchers (at Harvard, even) have shown that female undergrads in male-dominated majors (math, science, and engineering majors) experience high levels of stereotype threat compared to women in majors that are not male-dominated, to the extent that they are much more likely than males to consider changing majors3. If women experience stereotype threat in the context of math education, and given that stereotype threat has been shown to affect intellectual performance, it would not be surprising to find that stereotype threat affects women's performance on standardized math tests, and thus plays a role in the much-discussed gender differences in math ability.

The only study on the role of stereotype threat in gender differences in math ability that I remember seeing cited during the debate over the Summers remarks was one by Jason Osborne. Using data from more than 15,000 individuals drawn from a national study of high school seniors, Osborne found that anxiety (not necessarily stereotype-related anxiety) mediated the gender differences in scores on standardized math tests, though the effect of anxiety was fairly small. In other words, anxiety played a role in the gender differences Osborne observed in math performance, but the role was a small one. This studies strength is that it used an extremely large and diverse sample. It's weakness, however, is that it did not look at the role of stereotypes specifically, and that its only measure of anxiety was post-test self-report.

Other studies, which as far as I can tell have been neglected in the discussion of Summers' remarks, have looked at the role of stereotype threat more directly. Spencer et al., for example, found that for a sample of men and women who were highly and equally qualified, gender differences in mathematical performance could be eliminated by reducing stereotype threat4. After showing that their math test produced the frequently observed gender differences, they had participants perform the same test after being told either that the test does not produce gender differences, or that it does produce gender differences. Women who were told that it does not produce gender differences performed as well as men, while women who were told that it does produce them performed significantly worse than men.

Brown and Josephs conducted a similar study, with similar results5. In their first experiment, they either told participants that the math test they were about to take would show whether they were weak in math, or whether they were strong in math. Consistent with the hypothesis that stereotype threat would affect women's performance, female participants who had been told that the test would determine whether they were weak at math (the stereotype-consistent test description) performed worse than female participants who had been told that it twould test whether they were strong in math. Men showed the opposite pattern. They performed better when they were told that the test measured math strength than when they were told that it measured math weakeness. This indicates that the effects of gender-related math stereotypes can be positive or negative. If you are a member of a group that is negatively stereotyped, then stereotype threat will hurt your performance, whereas the performance of members of the group that is positively stereotyped will perform better when the stereotype is active.

Past research (see the paper for citations) has shown that the presence of an "external handicap," i.e., some external factor that may hurt performance, can alleviate stereotype threat. In experiments 2 and 3, Brown and Josephs allowed participants to practice math problems prior to completing the math test. However, for some participants, they produced an external handicap by having the computer crash at the beginning of the practice session. If stereotype threat is hurting women's performance on math tests, then alleviating that threat by providing an external handicap should reduce the effect of the stereotype, and increase performance. Consistent with this prediction, they found that emale participants who experienced the external handicap and were told that the test measured math weakness performed significantly better than participants in the "weakness" condition than women who did not experience the external handicap. Thus, experiments two and three provide another piece of evidence that gender-related math stereotypes are affecting women's math performance.

Another factor that influences the effect of stereotypes on performance is group identification. Individuals who highly identify with a stereotyped group will be more susceptible to the negative effects of that group's stereotypes. Thus, we would expect that women who place more importance on their gender identity will perform worse on math tests when math-related gender stereotypes are activated than women who place less importance on their gender identity. To test this, Schmader conducted a study using participants with either high or low gender identification, and found that women for whom their gender identity was important performed worse on a math test when they were told that the test produced gender differences than men, while women who placed little importance on their gender identity performed as well as men on the same test6.

In another set of studies, Josephs et al. looked at the relationship between stereotype threat and social status7. They hypothesized that for individuals who view math ability as important, high levels of social status concern will increase the efffects of stereotype threat. Josephs and others (see paper for citations) have shown that for both men and women, high levels of testosterone (high-T) are correlated with high levels of status concern. Thus, they predicted that high-T women (women with high levels of testosterone compared to other women),would be more affected by the stereotype, and thus their performance would suffer more than women who had low levels of testosterone (low-T). To test this prediction, they first took saliva samples from participants who rated math ability as important, to measure their current testosterone levels, and then had them complete a short questionaire. Half of the participants completed a questionaire that had previously been shown to prime math-related gender stereotypes, while the other half completed a control questionaire. The participants then completed 20 questions from the quantiative section of the GRE. In their fist study, they found that high-T women performed worse when the stereotype was activated (when they had completed the stereotype-activating questionaire) than when it was not, whereas low-T women performed equally well in both conditions. In fact, high-T women in the stereotype-prime condition performed worse than low-T women in both the stereotype-prime and control conditions, indicating that the effect of the stereotype was quite large for high-T women. Interestingly, in their second experiment, they showed that, consistent with the findings of Brown and Josephs, the effects of testosterone on men are reversed. High-T men actually performed better when gender stereotypes were activated than when they weren't, and better than low-T men in either condition, while low-T men's performance did not differ in the two conditions. High-T men who have had the gender stereotype primed apparently view the test as a way to show off their mathematical abilities, and thus increase their social status.

Finally, researchers have begun to look at the ways in which stereotype threat might affect women's performance on math tests. In one study, Johannes Keller used methods similar to those of Steele et al. (2002), in which participants complete a math test either after having the gender stereotype primed or without having it primed8. Consistent with the other research, he found that females (he used high school students) performed worse than men when the stereotype was active, but as well as men when it was not. He also found that the decrease in performance was largely due to self-handicapping, which is common in the face of stereotype threat. Self-handicapping involves things like decreased effort and attention, procrastination, and similar performance-reducing behaviors. In Keller's study, female participants form who the stereotype was primed were much more likely to perform self-handicapping behaviors.

In another study, Quinn and Spencer primed math-related gender stereotypes, and again showed that women in the primed condition performed worse than women for whom the stereotypes were not active, and that for the latter group, performance was equal to that of men9. In addition to completing the math test, participants in their study also described their problem-solving strategies. Quinn and Spencer then coded these strategies, and found that women in the stereotype-primed condition, women produced fewer and less effective problem-solving strategies. This indicates that stereotype-threat made it more difficult for women to produce problem-solving strategies, and that this reduced their performance on the test. Another interesting result from their study that is not directly related to the role of stereotype threat was that in their first experiment, in which stereotypes were not activated, women performed as well as men on numerical problems, but worse on word problems.

The point of all of this is that several studies have shown that stereotype threat influences women's performance on math tests, and thus is likely responsible for at least part of the observed gender differences in math ability. In fact, given that in most of the studies, gender differences were eliminated when stereotype-threat was absent or reduced, either through indicating that the test does not produce gender differences or producing an external handicap, or in participants for whom stereotypes are not as relevant due to the low importance of gender identity or low levels of social status concern, the role of stereotype-threat in gender differences may be quite large. The fact that Osborne found only a small, though statistically significant effect of anxiety is interesting, but it doesn't speak directly to the role of stereotype-threat, and is overshadowed by the long list of studies that directly demonstrated large effects of stereotype-threat. It would be difficult for innate factors, even those relating to the influence of spatial reasoning ability, to account for much of the data on gender differences in math ability. As of yet, no one has a theory explaining how innate differences account for the fact that gender differences disappear in untimed tests, in numerical problems vs. word problems, and when stereotype threat is alleviated. This doesn't mean that there aren't innate differences, but it does mean that for now, the best evidence we have indicates that social factors play a strong role in gender differences in math, and it would be a mistake to overlook them, particularly in the search for innate differences that cannot explain the data.

1 See e.g., Steele, C. M. (1997). A threat in the air: How stereotypes shape intellectual identity and Performance. American Psychologist, 52, 613-629, or Steele, C.M. (1999, August). Thin ice: “Stereotype threat” and Black college students. Atlantic Monthly, 44-54.
2 Aronson, J., Lustina, M., Good, C., Keough, K., Steele, C., & Brown, J. (1999). When white men can't do math: Necessary factors in stereotype threat. Journal of Experimental Social Psychology, 35, 29-46.
3 Steele, J., James, J.B., & Barnett, R.C. (2002). Learning in a man’s world: Examining the perceptions of undergraduate women in male-dominated academic areas. Psychology of Women Quarterly, 26, 46-50.
4 Spencer, S.J., Steele, C.M., & Quinn, D.M. (1999). Stereotype threat and women's math performance. Journal of Experimental Social Psychology, 35(1), 4-28. See also Walsh, M., Crystal, H., & Duffy, J. (1999). Influence of item content and stereotype situation on gender differences in mathematical problem solving. Sex Roles, 41(3-4), 219-240, for a study using similar methods and producing similar results in a sample of middle school-aged participants.
5 Brown, R.B., & Josephs, R.A. (1999). A burden of proof: Stereotype relevance and gender differences in math performance. Journal of Personality and Social Psychology, 76(2), 246-257.
6 Schmader, T. (2002). Gender identification moderates stereotype threat effects on women's math performance. Journal of Experimental Social Psychology. 38(2), 194-201.
7 Josephs, R.A., Newman, M.L, Brown, R.P., & Beer, J.M. (2003). Status, testosterone, and human intellectual performance: Stereotype threat as status concern. Psychological Science, 14, 158-163.
8 Keller, J. (2002). Blatant stereotype threat and women's math performance: Self-handicapping as a strategic means to cope with obtrusive negative performance expectations. Sex Roles, 47(3-4), 193-198.
9 Quinn, D.M., & Spencer, S.J. (2001). The interference of stereotype threat with women's generation of mathematical problem-solving strategies. Journal of Social Issues, 57(1), 55-71.

Reading Cognitive Science

One of the purposes of this blog is to promote the field that is near and dear to me, cognitive science. Posts on what I think are interesting or socially/politically-relevant cognitive science topics is one way to do that. Another way would be to engage people more directly. I get emails fairly frequently from students who are interested in grad school in cognitive science, or people who are in other disciplines but want to learn more about cognitive science. A good way for students to prepare, or people outside of cog sci but interested to learn more, is to read books on the topic. I thought it might be interesting, then, to have a sort of blogosphere reading group for people interested in cognitive science (I owe the idea of a blog reading group to Clark). We could read the books and discuss them on our individual blogs, in comments, or even through emails. It would only take a few interested individuals for me to participate, though naturally, the more the better. No special expertise would be required, of course, though anyone out there in cognitive science or related disciplines would be welcome to participate. You don't have to have a blog, either. You can participate in the discussion through comments or email.

I have a few ideas for books that might be interesting to read, but we can work out the specifics of what to read and how we'll go about talking about it when we've got a list of people who are interested, and what cog sci topics they might be interested in. So, if you think this is something you'd like to do, leave a comment or email me (if you leave a comment, be sure to leave an email address).

UPDATE: It looks like there's definitely enough interest to do this. I'm going to wait until Monday for people to join in (which is not to say that you can't join in after Monday). Then I'll send an email to everyone who's expressed an interest in participating, and we can work out the specifics.

Monday, July 18, 2005

One More Quick Note On Lakoff In the NYTs

I wanted to say something about this passage from the article (see the previous post for a link):
According to Lakoff, Democrats have been wrong to assume that people are rational actors who make their decisions based on facts; in reality, he says, cognitive science has proved that all of us are programmed to respond to the frames that have been embedded deep in our unconscious minds, and if the facts don't fit the frame, our brains simply reject them. Lakoff explained to me that the frames in our brains can be ''activated'' by the right combination of words and imagery, and only then, once the brain has been unlocked, can we process the facts being thrown at us.
I would remove some of the language in the paragraph. "Programmed," for instance, is a nasty word outside of the context of the computer metaphor for mind, and immediately invokes the thought of brainwashing that show up in the article. That's not what I want to talk about, though. What I want to talk about is in this sentence:
Cognitive science has proved that all of us are programmed to respond to the frames that have been embedded deep in our unconscious minds, and if the facts don't fit the frame, our brains simply reject them.
If we ignore more poor wording ("proved," in the context of science, and in particular, a science as young as cognitive science? I mean, come on! Popper and Quine are both rolling over in their graves), this says something that is both true and not quite true. Our schemas (or frames) tune our perceptual and conceptual mechanisms, and thus, information that is not relevant to those schemas (I call that information schema-irrelevant) tends to get ignored. I emphasize "tends," because sometimes that information does get in.

But what Lakoff's short description doesn't mention, and thus what makes it not quite true, is that information that does not "fit the frame," but which is not irrelevant to it (which, for instance, contradicts some of the information contained in the frame) will, in fact, be more readily noticed (I call this schema-inconsistent information). If cognitive science has "proved" anything about schemas, it is not that we reject irrelevant information but that, if information is relevant but inconsistent (i.e., it's schema-inconsistent), we are actually more likely to notice and remember it! Information that is consistent with our schemas or frames (you can probably guess that I call it schema-consistent information) is easily missed, because we're simply expecting it. We're less likely to remember the particulars of that information, and once our schemas are activated, we're more likely to remember that schema-consistent information was there even if it wasn't. Schema-inconsistent information is surprising, and we're built to notice surprising information. If instead what Lakoff said were true, then framing simply would not work, because if our schemas simply reject information that is inconsistent with them, then they would reject all new frames, no matter how often we repeated them. We'd also be pretty maladapted creatures, running around never noticing any new or surprising information after we've developed a lot of schemas.

I imagine Lakoff is aware of this, but if we read only that statement, we'd miss one of the most important aspects of framing. The best way to counter a frame is to provide another that presents the reader/listener with a bunch of relevant but frame/schema-inconsistent information. This new information is in fact required to get people to think about the issues from a new perspective, using new base domains, or new schemas. When we are coming up with new frames, then, we have to carefully consider what information we want people to notice, and present it in such a way that it is immediately relevant and at the same time inconsistent with the entrenched frame. Doing so will highlight that information while simultaneously highlighting the problems with the opposition's way of approaching an issue. I'm pretty sure that's what Lakoff wants Democrats to do.

Once again we see that if we listen to Lakoff's version of framing, we won't be able to do what Lakoff wants.

Lots to Say About Lakoff

Lakoff talk is back, this time in the New York Times (via Language Log). The article, titled "The Framing Wars," is long, and full of good, bad, and just plain false points about Lakoff, framing, and the uses of framing. The best parts of the article are the discussion of some successful real-world applications of framing by Democrats, in the social security and filibuster debates, and the point that one of the most important aspects of the Democrats embracing framing is their newfound discipline in staying on message and using the frames that their research has shown will work. While the article points out that much of the discussion of framing in the media, among politicians, and I should add, on blogs, has been overly simplistic, the fact that Democrats are sticking to their frames shows that they've grasped at least one of the more complex points about framing, namely that in order for it to work, it has to be consistent and coherent enough to change the way people actually think about an issue. If you switch back and forth between multiple frames, or alternate between your frames and your opposition's frames, you will have a hard altering anyone's schemas.

These two good aspects of the article are overshadowed, however, by much that isn't so good. The popular press simply can't get it right when it comes to science, even when that science is as shitty as Lakoff's. I'm glad that framing itself is still getting attention, even in a non-election year, but I hope that more liberal cognitive scientists who study concepts, metaphor and analogy, rhetoric/communication, and related topics will start to speak up within earshot of politicians and the press (i.e., not on podunk blogs like this one). For now, I'm going to speak up on a podunk blog like this one (identical with it, in fact) about the NYT article.

The article doesn't get to Lakoff until the third page (of the online version), and when it does, it starts off on the wrong foot, calling Lakoff "the father of framing." This will come as a shock to anyone who read Erving Goffman's work on framing in the mid-1970s, or to anyone who wrote on framing between the publication of Goffman's book and Lakoff's work on the topic in the 1990s (or even the 1980s, if we are generous and describe Lakoff's early work on metaphor as work on framing). Lakoff has popularized framing, however, and for that he does deserve recognition. I can't help but think that Matt Bai, the author of the New York Times piece, has been buying into Lakoff's framing of himself, though. He follows the bogus "father of framing" metaphor with this:
A year ago, Lakoff was an obscure linguistics professor at Berkeley, renowned as one of the great, if controversial, minds in cognitive science but largely unknown outside of it.
As far as linguistic professors go, Lakoff hasn't been obscure for a long time. After Noam Chomsky, he has probably been the most famous linguistics professor in the U.S. for at least two decades. Metaphors We Live By has been one of the most popular cognitive science books among non-cognitive scientists for 20 years, and I'd wager that Women, Fire, and Dangerous Things has sold more copies that virtually any other cog sci book of similar length as well. This is not because cognitive scientists are reading it (most of the ones I know haven't). Furthermore, among cognitive scientists, Lakoff is certainly not considered one of the great minds of our discipline. He was controversial at one time, though now he's more recalcitrant in the face of overwhelming data against his view than he is controversial. But this article appears to be a sort of myth-building exercise, and "largely ignored by cognitive scientists, Lakoff has begun to gain a great deal of influence in politics" doesn't seem like a good start to a myth.

To Bai's credit, he does try to hint at some of the deeper aspects of framing. He writes:
According to Lakoff, Democrats have been wrong to assume that people are rational actors who make their decisions based on facts; in reality, he says, cognitive science has proved that all of us are programmed to respond to the frames that have been embedded deep in our unconscious minds, and if the facts don't fit the frame, our brains simply reject them. Lakoff explained to me that the frames in our brains can be ''activated'' by the right combination of words and imagery, and only then, once the brain has been unlocked, can we process the facts being thrown at us.
This could be an OK short description of how framing works. Framing is about changing and utilizing largely unconscious representations. (A terminological note: I call these representations schemas, even though Lakoff calls them frames. I think it's important to call the representations "schemas" because if we call both the language and the conceptual representations that framing analysis deals with "frames," then we might be tempted to treat "framing" as a strictly linguistic phenomenon, which is what many, including most of the politicians the article discusses, appear to do.) The article at least attempts to make it clear that this is about changing how people think about issues, not just changing how we talk about them, and that's a good thing. Still, as Bai himself shows, the use of the word "unconscious" immediately activates common misconceptions of the cognitive unconscious. For example, Bai writes:
This notion of ''activating'' unconscious thought sounded like something out of ''The Manchurian Candidate'' (''Raymond, why don't you pass the time by playing a little solitaire?''), and I asked Lakoff if he was suggesting that Americans voted for conservatives because they had been brainwashed.
Lakoff quickly points out how this is wrong, saying:
That's true, but that's different from brainwashing, and it's a very important thing. Brainwashing has to do with physical control, capturing people and giving them messages over and over under conditions of physical deprivation or torture. What conservatives have done is not brainwashing in this way. They've done something that's perfectly legal. What they've done is find ways to set their frames into words over many years and have them repeated over and over again and have everybody say it the same way and get their journalists to repeat them, until they became part of normal English.
What Lakoff doesn't really make clear is that framing in politics is not about changing people's entire world-views by talking about things differently. Instead, it's about changing the way we talk about things so that people will represent them within particular schemas that they already have. So, for instance, we might want to reframe abortion so that, instead of being about life and death, it is represented utilizing the schemas we use to represent freedom and choice. This is nothing like brainwashing (it's not, as Lakoff suggests, different simply because it's legal). It involves showing people new relations between domains, and then utilizing the schemas from one domain (e.g., freedom) to represent another domain (abortion). If we want to use framing successfully, we have to realize this. It's important that, instead of going out and conducting surveys and focus groups to figure out which frames work best for each individual issue, we go out and look for the base domains (the schemas that people already have) that work the best, and then develop analogies (Lakoff would call them metaphors, but in order to avoid the appearance of endorsing his bogus cognitive theories, I'm going to call them analogies) from those. Sometimes this will require figuring out which base domains can best be used to represent individual domains, but it will also require developing large-scale narratives (or systems of analogies) from a set of base domains, which is what Lakoff argues Republicans have done so well.

Finally, the article includes the gratuitous "other perspective" (you know, to make it "objective"). There are two versions of this, one from Republicans and one from anti-framing Democrats. The Republican critique comes from Frank Luntz. Here's what Bai writes of his objections:
The problem with Lakoff, Luntz said, is that the professor's ideology seemed to be driving his science. Luntz, after all, has never made for a terribly convincing conservative ideologue. (During our conversation, he volunteered that the man he admired most was the actor Peter Sellers, for his ability to disappear into whatever role he was given.) Luntz sees Lakoff, by contrast, as a doctrinaire liberal who believes viscerally that if Democrats are losing, it has to be because of the words they use rather than the substance of the argument they make. What Lakoff didn't realize, Luntz said, was that poll-tested phrases like ''tax relief'' were successful only because they reflected the values of voters to begin with; no one could sell ideas like higher taxes and more government to the American voter, no matter how they were framed. To prove it, Luntz, as part of his recent polling for the U.S. Chamber of Commerce, specifically tested some of Lakoff's proposed language on taxation. He said he found that even when voters were reminded of the government's need to invest in education, health care, national security and retirement security, 66 percent of them said the United States overtaxed its citizenry and only 14 percent said we were undertaxed. Luntz presented this data to chamber officials on a slide with the headline ''George Lakoff Is Wrong!!''
I'll be damned if I know what Luntz' admiration of Peter Sellers has to do with any of this, but I can say for sure that Luntz' doesn't understand Lakoff's work. First of all, while it is highly likely that Lakoff's politics are guiding his representation of conservatives and liberals (through his descriptions of the STRICT FATHER and NURTURANT PARENT metaphors), that's not science. That's Lakoff's own idiosyncratic informal cognitive linguistic analysis of political discourse. The science, which I will admit isn't very good science, but which makes many of the same points that good science does, was around long before Lakoff got political. In addition, Luntz' data doesn't show that Lakoff's view of framing (or framing analysis itself) is wrong. It does show that Lakoff might be wrong about which frames to use for taxation, but even that's not clear. Lakoff himself notes that simply presenting frames once or twice over a short period of time is not sufficient to change people's deeply-entrenched and largely unconscious (and therefore difficult to consciously change) representations. It takes the extended and consistent use of frames to create large-scale representational change. As Bai puts it, Lakoff "does not suggest in his writing that a few catchy slogans can turn the political order on its head by the next election."

The Democratic criticisms are slightly better, but completely misdirected. Here is how Bai describes it:
Lakoff's detractors say that it is he who resembles the traveling elixir salesman, peddling comforting answers at a time when desperate Democrats should be admitting some hard truths about their failure to generate new ideas. ''Every election defeat has a charlatan, some guy who shows up and says, 'Hey, I marketed the lava lamp, and I can market Democratic politics,''' says Kenneth Baer, a former White House speechwriter who wrote an early article attacking Lakoff's ideas in The Washington Monthly. ''At its most basic, it represents the Democratic desire to find a messiah.''
This is simply unfair to Lakoff. He is not a charlatan, even if he's not a good cognitive scientist. His central points, about the importance of morality in political discourse, and the use of what we know about how people think to help communicate our ideas better while counteracting the deceptive use of language by Republicans, are dead on. Yet, it is obviously true that many Democrats have latched onto Lakoff without fully understanding what framing entails. It does not mean that we can call taxation "dues for services rendered," and think that people will immediately vote for our candidates.

It's a shame that the article included either of these criticisms, at least as directed toward Lakoff (the latter is pretty dead on when directed toward Democrats who don't understand framing). It's also a shame that the article doesn't do more to correct the misrepresentations that it itself laments, including the ones that it tries to perpetuate (e.g., with talk of brainwashing). It would be really nice if the article actually mentioned the criticisms of cognitive scientists (you know, the people who actually know what they're talking about), too. But all of that is probably too much to expect from an article in the New York Times, even if it is 12 frickin' pages long.

Thursday, July 14, 2005


At some point over the last couple days -- I'm not sure when, and I'm too lazy to go back and look -- I hit 40,000 "unique" visitors. I'll be damned if I know how I've gotten that many. Thanks, though, for stopping by, and I hope some of you have developed an interest in cog sci, or had your interest in it strengthened.

Feel free to drop me a comment or an email if there's anything you'd like to see posted or changed, or if you just want to say hi. Thanks again.

Tuesday, July 12, 2005

Male Privelege, Female Privelege, and Why More Men Should Take Feminism Seriously

These two posts by Lindsay of Majikthise got me thinking. I’m a bit of a geek, and more than a bit of a social misfit (I should have been a pair of ragged claws, and all that), and as a result, I’ve been pretty terrified of women in social contexts since the onset of puberty. Being a social misfit, particularly one who’s terrified of women, means you end up being around other men who are social misfits and terrified of women, so I've had a lot of experience with men like me. While I’ve never suffered from nice-guy syndrome (I’ve always recognized my own assholeness), which involves the belief that (attractive) women always end up with jerks, and then complain about men being assholes, while nice guys (which pretty much means misfits in this context) get the shaft, I have known plenty of afflicted men. These men genuinely believe that if only the woman would take the time to get to know them, and look past the surface (i.e. physical attractiveness), they would fall for them instead of the attractive jerk. It’s a strange belief that underscores the points about “privilege” that Lindsay and some of the commenters made in her first post. It basically faults women for being superficial by choosing the attractive guy, who ends up being a jerk, rather than the nice guy who’s less attractive but will treat her right. At the same time, it allows the self-described nice guy to be superficial, by choosing a woman for her looks (remember, if she hasn’t taken the time to get to know him, then he hasn’t had the time to know her). In other words, the nice guy is faulting the woman for doing exactly what he’s doing, and this sort of reasoning is pretty widely accepted by men everywhere. If that’s not a double standard of the sort that defines male privilege, I don’t know what would be.

Of course, not all the fear of women held by us misfits is totally unjustified. The same system that produces male privilege also produces something that might be called female privilege. It’s the privilege that allows women to reject men with little regard for their feelings, or require things (drinks, gifts, or whatever) in return for the honor of spending time with them. Even attractive men will be rejected more often than not, and often quite harshly, and it takes a pretty thick skin to take all of the inevitable rejection without developing some fear of women. The privilege arises out of the part of the social system that requires men to approach women (and in many ways frowns on women who approach men), creating the belief in many women that men should earn or deserve their attention, affection, or just the chance to sit next to them in a bar.

The point I’m trying to make here is that the male-dominated, and largely male-created social system tends to harm about as many men as it does women. Too many of us are on the outside looking in because we fit into propertly into the gender roles that the system (the patriarchy, as some like to call it) itself has created. Both men and women are harmed by both the male and female-oriented double standards of that system, and often we’re too blinded by those same double standards (as in nice guy syndrome, or harsh rejection disorder) to see that. It’s for that reason that I think more men should take feminism seriously. Feminism’s primary purpose is to expose the problems in the system, showing where it has created double standards, along with what the causes and effects of those double standards are. If those effects are just as often bad for men as they are for women, then we’d do well to take notice of them and, if possible, work to change them, at least in ourselves.

Yet, as the comments on Lindsay’s post show all too well, many of the men who are getting the short end of the patriarchal stick are a long way from seeing this. For them, feminism is just women bitching, blaming men for everthing, or making whacky claims about how fluid mechanics are neglected in physics because the fluid is associated with the feminine. Instead of thinking about what feminists actually say, these men are doing exactly what the double standard requires: blaming women. It’s women’s fault that I’m alone, bitter, depressed, afraid, or had to settle for a woman I didn’t really want, because (attractive) women exercise poor judgment when it comes to men, picking, as they do, the attractive over the nice. Or so the reasoning goes. But that reasoning has never gotten anyone anywhere, and it never will. It will just reaffirm the system that causes so much trouble for all of us. And trust me, that's just what the men (and women) who are genuinely benefitting from the system, at the expsense of the women (and men) who aren't, want.

Post Script: Yes, as the title of this blog, and the allusion in the second of this post indicate, I read too much Eliot. What of it?

Monday, July 11, 2005

Leiter on EP II

After writing a short post on Leiter's relatively short critique of Evoluionary Psychology, I started to feel bad. I said that his arguments were basically just Gould's (with a little Lloyd thrown in for good measure), and that they weren't very good arguments, and just referenced a book as support for that claim. In other words, I behaved a lot like Leiter. So this post is really just my attempt at peace of mind. It's probably also me being overly kind to Leiter, as we can be reasonably certain that he would never do this for me. For these reasons, and countless others, you'd probably do well to just ignore it.

Here is Leiter's (or Gould's) argument against EP (from Leiter's post):
[I]t is extremely reasonable, given what we know, to express doubts about evolutionary psychology and its selectionist hypotheses about differences between the sexes, since none of these hypotheses (as in none) have been confirmed by standards that approach those in biology [This is the Lloyd part. - Chris]. The fundamental difficulty is that there exist important non-selectionist evolutionary mechanisms (for example, genetic hitch-hiking or genetic drift), so that one can not, as evolutionary psychologists do, treat the selectionist explanation as the default one [This is the Gould part. - Chris]. This is just bad science. This point is also the stuff of baby biology textbooks [This is the Leiter-style gratuitous jab. - Chris] ; herewith Stearns & Hoekstra (OUP, 2000), p. 8:: “much of the variation in DNA sequences [over time] is neutral with respect to selection.” The challenge for evolutionary biologists studying, e.g., sex differences, is to figure out what role selection, if any, is really playing. Evolutionary psychology is silent on this problem. (There is a separate problem, of course, pertaining to the role of non-biological factors in observed sex differences.)
He then goes on to give examples of some traits in nonhuman animals that did not evolve as selection-induced adaptations. The problem with this critique, of course, is that it doesn't say anything about any of the claims, arguments, or fundamental tenets of EP. It's not true, in practice, that Evolutionary Psychology assumes that "the selection explanation" is the "default one." This is how Buller correctly summarizes the picture of EP that Leiter and Gould are operating with:
This pattern of reasoning begins with the observation of a purported adaptation and then attempts to reconstruct its evolutionary history. (p. 90)
But, as Buller notes (also correctly), this is not what Evolutionary Psychologists do.
I can't think of any instance of an Evolutionary Psychologist using this line of reasoning, and I don't think we can expect Leiter to give us one (Gould never does). Instead, what Evolutionary Psychologists do is the following (also from Buller, and on the same page):
If early humans faced such-and such an adaptive problem in the EEA, then our species should have evolved this or that proximate behavior-control mechanism to solve that problem; so, if modern humans posses such a proximate mechanism, it is an adaptation.
In other words, Evolutionary Psychologist hypothesize (i.e., they don't assume) that if certain adaptive conditions existed in the environment of evolutionary adaptiveness (EEA), then certain "proximate behavior-control mechanisms" should have evolved to deal with those problems. The existence of the proximate mechanisms are predictions derived from the hypotheses about the adaptive problems in the EEA. They then reason about the modern-day cognitive/behavioral manifestations of those proximate mechanisms, and go out and collect data to look for those cognitive/behavioral manifestations. If they find data that supports the existence of those manifestations, then their predictions are confirmed, and their hypotheses supported.

Now there are all sorts of problems with this line of argument. For one, it involves what is little more than speculation about the adaptive conditions of the EEA, along with the counter-to-fact belief that those adaptive conditions were static over an extended period of time in human evolution (and not, in fact, influenced by our adapting to them!). It also requires some pretty speculative reasoning about what sort of proximate mechanism we should have evolved to deal with those statically-conceived adaptive problems. In addition, in practice, it always involves some pretty sketchy reasoning about the modern-day manifestations of some pretty abstractly described proximate mechanisms. What this leads to is bad reasoning leading to speculative hypothes, resulting in poor empirical investigations that yield data that don't support those hypotheses. It's not surprising, then, that for virtually every EP empirical investigation, there are either many equally plausible alternative interpretations of the data, or there is in fact a wealth of data that is inconsistent with the EP interpretation (see Buller's chapters on mating, marriage, and parenthood for examples on these topics, and this post and the references for a discussion of the empirical evidence for social exchange theory).

What a critique of EP requires, then, and what Leiter (echoing Gould) does not provide, is a discussion of the problems with the reasoning by which Evolutionary Psychologists come to the conclusion that a certain proximate mechanism, if it exists, is an adaptation, and more importantly, a case-by-case look at the empirical support for the various EP hypotheses.