The Neglect of the Environment by Cognitive Psychology

In 1955, Egon Brunswik presented a paper in which he argued that neglect of the environment and over emphasis of the organism was the major downfall of cognitive psychology. His critiques have largely been ignored and research is discussed that demonstrates the same organismic-asymmetry Brunswik detailed in 1955. This research is discussed in attribution terms since experimental psychologists make behavioral attributions. This organismicasymmetry has resulted in a body of research that is guilty of the fundamental attribution error. Brunswik’s theory of representative design, proposed to address organismicasymmetry, is discussed and contrasted with calls for ecological validity. Although calls for ecological validity are well intentioned, they lack any systematic theory of the environment and fall significantly short of Brunswik’s ideal. In 1955, Kenneth Hammond presented Brunswik’s paper, “Scope and aspects of the cognitive problem,” at one of the first ever conferences on cognition. Hammond (2001) writes that “in this paper he [Brunswik] made a strong effort to make the whole of his life’s work intelligible to his peers” (p. 298). The theme of this paper, if not the language, is simple; psychology has neglected the environment. Brunswik (1957/2001) writes “if there is anything that still ails psychology in general, and the psychology of cognition specifically, it is the neglect of the investigation of environmental or ecological texture in favor of that of the texture of organismic structures and processes.” Brunswik’s argument still carries weight today and psychology in general, and cognitive psychology specifically, has still not dealt with the criticisms levied against it by Brunswik 50 years ago. Brunswik began that 1955 paper with the following sentences, “One of the broadest and most universally accepted definitions of psychology conceives of psychology as being concerned with the interrelationships between organism and environment. In this definition both organism and environment appear as equal partners” (p. 300, 1957/ \\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 2 21-MAR-07 8:13 140 Journal of Theoretical and Philosophical Psy. Vol. 26, 2006 2001). At face value, there seems little to object to in this definition. However, a look at the definition of psychology in most introductory textbooks will reveal that the environment is not mentioned. A review of most of the major standard-format introductory textbooks (modular style texts not included) revealed that of 36 textbooks, the environment was only mentioned in the definition of psychology in one of them while “mental”, “mind”, “cognitive”, or “experience” was mentioned in 34 of the 36 definitions. At face value we can take this as support that psychology today is still more focused on “organismic structures and processes” than on the environment in which behavior takes place. For Brunswik, this asymmetric focus of psychology on the organism was the central problem of experimental psychology. Brunswik strongly advocated that the organism-environment interaction should be the central focus of psychology and that a cumulative science would only be had with this systemic focus. Both Hammond (2001) and Brunswik (1957/2001) point out that while theories of organismic behavior abound in psychology, there have only been two explicit theories of the environment, those of Brunswik and Gibson. Without explicit theories of the environment, organismic attributions of behavior will change with experimental paradigms without a clear linkage between contexts. Linkages between the organism, behavior, and environment will be post-hoc and remain limited to haphazard changes in experimental paradigms so long as psychology maintains this asymmetric focus on the organism. Research in cognitive psychology, including social cognition, typically demonstrates this asymmetric-organismic focus. Sternberg (2003), in his textbook on cognitive psychology, defines cognitive psychology as “the study of how people perceive, learn, remember, and think about information” (p. 529). The environment is not mentioned in this definition and there are almost no theories linking cognition to the environment in an explicit a-priori way (exceptions include Gibsononian and Brunswikian inspired theories). While Brunswik explicitly emphasized the achievement of the organism within an environment, much of cognitive psychology has instead focused on errors in cognition with the assumption that learning about these malfunctions can teach us about how normal cognition operates. Terms like “bounded rationality” and “biases” are often used to describe these limits. In the abstract, the paradigm is to construct an experiment that demonstrates these flawed cognitive processes at work and attribute them to limits in our information processing abilities. Little attention is initially given to the effects of the experimental context on behavior since it is implicitly assumed that we are learning about cognition, which resides within the organism. Later, when a different experimental procedure yields different results, theories of cognition \\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 3 21-MAR-07 8:13 Organismic-assymetry 141 are modified. What is missing from the start is the systemic focus on organism-environment achievement that Brunswik advocated 50 years ago. Since experimental psychologists are in the business of making attributions about the causes of behavior, perhaps this asymmetric-organismic focus can be better understood through the lens of behavioral attributions. The fundamental attribution error (FAE), or correspondence bias, refers to the tendency for people to make dispositional attributions over situational attributions when establishing the cause of another person’s behavior. Gilbert (1995) and Gilbert, Pelham, and Krull (1988) developed a two-stage model of attribution that describes the possible mechanism that produces this causal attribution bias. The theory posits that when we first observe a behavior, we make an automatic personal attribution (the person is the cause of the behavior). This personal attribution is automatic, triggered by an observed behavior, requiring minimal cognitive resources, and runs until completion. The second stage however is not automatic. Rather, it is an effortful contextual adjustment (taking the power of the situation into account when making an attribution). For example, you may be more patient when dealing with a friend or loved one if you know they are experiencing an unusually high load of personal stress. Consideration of how the context may have caused the behavior requires conscious effort according to this theory of behavioral attribution. Since the contextual attribution requires attentional resources, one can make the prediction that if people are cognitively taxed, they will not have the resources available to make the effortful contextual adjustment. Instead, they will over-attribute the cause of an observed behavior to factors residing in the person. An example will help make the phenomenon more concrete. In one study participants were asked to watch: seven silent clips from a videotape of a female target having a discussion with a stranger. In five of the seven clips, the target appeared extremely anxious. Half the subjects learned that in these five clips the target had been discussing anxiety-inducing topics (e.g., her sexual fantasies). The remaining subjects learned that in all seven clips the target had been discussing relaxation-inducing topics (e.g., world travel). Half of the subjects in each of these conditions were required to perform a cognitive rehearsal task (i.e., remembering the discussion topics in their proper sequence) while viewing the tape, and the remaining subjects were not. After viewing the tape, subjects rated the target’s trait anxiety, predicted the target’s future state anxiety, and attempted to recall the discussion topics. (Gilbert et al., 1988, p. 734) \\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 4 21-MAR-07 8:13 142 Journal of Theoretical and Philosophical Psy. Vol. 26, 2006 Gilbert et al. hypothesized that the added cognitive load of memorizing the discussion topics would inhibit this group from making the effortful contextual adjustment. Ironically, they predicted that those subjects who were memorizing the contextual variables were less likely to use the contextual variables when making a causal attribution. This is precisely what they found. On average, the target discussing the relaxing topics was judged as less trait-anxious by participants who had the memory task than those without the memory task. Likewise, the target discussing the anxious topics was judged to be more trait-anxious by participants who had the memory task than those without the memory task. This finding is ironic because it shows that participants who memorize the contextual variables are less likely to use those same variables in subsequent judgments! Cognitive psychologists, like the participants in Gilbert et al.’s study, are in the business of making causal attributions about behavior. Cognitive psychologists may also have a tendency to attribute behaviors to personal factors rather than situational factors. This, in and of itself, is nothing shocking because it simply suggests that researchers fall prey to some of the same biases that other people do. However, this becomes more surprising when you consider that the researchers create the situations, usually with the intent of showing the behavior. When the behavior is observed, it is attributed to some biased mechanism of the cognitive system, and the situation that was designed to show the behavior is under-weighted. Consider how similar the preceding argument is to the participants in Gilbert et al.’s study who, upon memorizing all the situational variables, attributed the cause of the behavior to dispositional factors. Below I discuss examples of organismic-asymmetry in psychology and argue that the conclusions are examples of misattribution. Specific examples of how changes in the experimental situations moderate the observed effects are discussed. All of the examples demonstrate a lack of systemic organism-environment focus as advocated by Brunswik. It is important to note that I am not arguing against any of the findings. Rather my goal is to show that Brunswik’s central critique of cognitive psychology 50 years ago has still not been heeded and that a more explicit organism-environment focus would advance the science of psychology. In short, our bias toward personal attributions limits our understanding of situational factors, and ultimately, human behavior. These examples represent a convenience sample who’s basis for inclusion was the initial personal attribution made by psychologists that later had to be reconsidered when the experimental context was altered. No claim is made as to the representativeness of this sample. Rather, they should be viewed as hand-picked examples for the purpose of making a point. \\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 5 21-MAR-07 8:13 Organismic-assymetry 143 The Negative Effects of Introspection on Judgment Wilson and colleagues have several studies showing the detrimental effects of introspection on decision making (Hodges & Wilson, 1993; Wilson, Kraft, & Dunn, 1989; Wilson & LaFleur, 1995; Wilson & Schooler, 1991). The title of one paper, “Thinking too much: Introspection can reduce the quality of preferences and decisions” (Wilson & Schooler, 1991), accurately describes some of their findings. Wilson, Dunn, Bybee, Hyman, and Rotondo (1984) asked some participants to “list all the reasons you can think of why your relationship with your dating partner is going the way it is“ (p. 11). Half the participants were asked to perform this reasons analysis task while the other half were not. Both participants were then asked to fill out a dyadic adjustment scale assessing their relationship. Participants were then contacted 32-41 weeks later and asked if they were still with their romantic partner. Participants who had performed reasons analysis prior to making their relationship satisfaction judgment showed a significantly lower correlation (0.08) between their responses and the state of their relationship several weeks later than participants who had not performed reason analysis (0.56). Introspecting on the reasons why they felt the way they did toward their partner resulted in a less accurate judgment about the relationship. The judgment tasks chosen by Wilson and colleagues are typically preference judgments (Hodges & Wilson, 1993; Wilson et al., 1989; Wilson & LaFleur, 1995; Wilson & Schooler, 1991). Preference judgments are ill-defined and ambiguous and therefore not likely to benefit from a rational analysis (Hammond, 1996). So perhaps it is not as bad as ‘Introspection reduces the quality of decisions’ but more like ‘Specific types of introspection mixed with specific types of judgment tasks can reduce the quality of decisions.’ This last view paints the findings in a more positive light while still acknowledging the effect. This last view is the approach that was taken by Millar and Tesser to explain such findings. Millar and Tesser (1986, 1989, 1992) have developed a mismatch model to explain the low attitude-behavior relationships found after reasons analysis by Wilson and colleagues. Millar and Tesser’s theory assumes that attitudes have both an affective and cognitive component. When asking people to analyze the reasons why they do something, the cognitive component of the attitude is made salient. When asking people to focus on how they feel about something, the affective component of the attitude is made salient. Attitude-behavior consistency will be strong when the attitude component made salient (cognitive versus affective) matches the primary drive of behavior. Consumatory behaviors, those done for their own purpose, are considered to be affectively based. Instrumental behaviors, those done to aid in goal attainment, are considered to be cognitively based. Any given behavior can con\\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 6 21-MAR-07 8:13 144 Journal of Theoretical and Philosophical Psy. Vol. 26, 2006 ceivably be performed for instrumental or consumatory purposes. The mismatch model predicts that when reasons analysis is performed prior to making a global evaluation, the cognitive component is activated. Since preference judgments are affectively based, the low attitudebehavior relation found in the work of Wilson and colleagues should be expected. This example demonstrates that introspection does not necessitate poor judgment. Such a statement is an attribution bias in that it fails to consider the wider context of the introspection instructions. Poor judgment following introspection is a result of a multitude of factors that include the context that induces the introspection. The effects of introspection on judgment are not limited to factors residing in the person (lack of introspective access) but include situational factors set up by the researchers including the type of introspection induced and the type of judgment task selected. The Effects of Affect Many researchers have found that negative moods (sadness, anxiety) increase self-focused attention and happy moods decrease self-focus (Wood, Saltzberg, & Goldsamt, 1990; Carver & Scheier, 1986). Such a view emphasizes how negative moods can lead to increased self-focus, rumination, and consequently depression. In short, this work often views negatively valenced moods as having negative consequences. This conclusion is also an attribution bias as little attention is given to the environmental context. Green and Sedikides (1999) argue that “affect orientation [self or environment] exerts effects on self-focused attention that are above and beyond those exerted by affect valence.” They argue that emotions often serve to prepare an organism for action or inaction and are adaptive responses to basic environmental demands like mate selection, food gathering, and fighting predators. Based on this psychoevolutionary perspective, Green and Sedikides hypothesized that self-focus would be a function of the orientation of the affect rather than the valence. They induced either sad or content emotions in participants and found equal levels of self-focus even though these states are oppositely valenced. Similarly, they found no difference between thrilled and angry affective states in terms of self-focus. However, the thrilled and angry participants were significantly less self-focused than the sad and content participants. These results require an explicit consideration of the situation and show that a predominantly personal attribution is insufficient to explain the relationship between self-focus and affect. Negative emotions provide feedback that may indicate goal-fulfillment (watching a sad movie may be the goal) and goal-progress (no pain, no gain). In short, moods, whether positive or negative, provide \\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 7 21-MAR-07 8:13 Organismic-assymetry 145 feedback about the environment (and our reaction to the environment) that aid in adaptive functioning (Martin, 1999). The simple view that negative moods lead to rumination and depression is a FAE because personal attributions are insufficient in that they do not account for the situational factors that are necessary to understand the effects of affect. The Irrational Processor There are many examples of irrational processing in social cognition, cognitive psychology, and the joint literature on judgment and decision making. One popular example of the irrational processor is the phenomenon of base rate neglect. In studies of base rate neglect, people are given specific information about a case, plus information about the population distribution from which the case was drawn (or base rate information) and asked to make a judgment about the likelihood of the case. Participants often underweight the base rate information when judging the likelihood of the event. For example, the following is a commonly used base rate problem (Bar-Hillel, 1980; Kahneman & Tversky, 1972; Tversky & Kahneman, 1980): A cab was involved in a hit and run accident at night. Two cab companies, the Green and the Blue, operate in the city. You are given the following data: 85% of the cabs in the city are Green and 15% are Blue. A witness identified the cab as Blue. The court tested the reliability of the witness under the same circumstances that existed on the night of the accident and concluded that the witness correctly identified each one of the two colors 80% of the time and failed 20% of the time. What is the probability that the cab involved in the accident was Blue rather than Green? (Tversky & Kahneman, 1982, p. 157) Bayes’ Theorem is the normative rule usually used to obtain the correct answer. Using the odds form of Bayes’ the cab is more likely to be Green (.59) than Blue (.41). Despite this fact, “the median and modal answer is typically .80, a value which coincides with the credibility of the witness and is apparently unaffected by the relative frequency of blue and green cabs” (Tversky & Kahneman, 1982, p. 157). There are many other examples in the literature of participants ignoring or under-weighting base rates (see Kahneman, Slovic, & Tversky, 1982; Koehler, 1996). The finding is so common that Bar-Hillel (1980) stated, “The genuineness, the robustness, and the generality of the base-rate fallacy are matters of established fact” (p. 215). Similarly, Kahneman and Tversky (1972) stated, “man is apparently not a conservative Bayesian: he is not Bayesian at all” (p. 450). \\server05\productn\T\THE\26-1-2\THE1205.txt unknown Seq: 8 21-MAR-07 8:13 146 Journal of Theoretical and Philosophical Psy. Vol. 26, 2006 Such work on base rate neglect implies that we are inherently doomed to be irrational processors of information. Our irrational processing is attributed to limitations in our information processing systems (Gigerenzer & Hoffrage, 1995; Lichtenstein, Fischhoff, & Phillips, 1982). Research by Gigerenzer and Hoffrage (1995) demonstrates that such a simple personal attribution is an error. Gigerenzer and Hoffrage (1995) argue that the representation format used in classic base rate neglect studies may not be in line with how our cognitive systems have evolved. Classic base rate studies present information as percentages. Such a format is a recent development and not necessarily the format our cognitive systems are tuned for. Gigerenzer and Hoffrage demonstrate that by changing a standard base rate problem from a percentage to a frequency format, which they argue is more natural, the proportion of participants who use a Bayesian algorithm and get the correct answer jumps from 16% to 46%. Arguing that base rate neglect is inherent to our information processing system is a misattribution if changes to the surface characteristics of the task drastically increase Bayesian reasoning. The most natural format to present base rate information is to let participants experience base rates in a series of unique events rather than as summary statistics (Gigerenzer & Hoffrage, 1995). Under such condition of direct experience Dunwoody, Goodie and Mahan (2005) found that participants differentially utilize base rates as a function of their environmental predictiveness. Participants used base rate information more when it was consistent than when it was inconsistent. In addition, when participants had a choice between information that was consistent but less useful versus an inconsistent but more useful alternative, they choose the more useful piece of information regardless of whether it was the base rate information. This finding was particularly surprising considering that the predictiveness of the information varied by a mere 10%. Base rate usage was found to be a function of the statistical characteristics of the environment and not a fixed bias inherent in the participants.