Implicit and Explicit Process in Social Judgment: Deep and High
Marilynn B. Brewer
The combined themes of this volume – explicit and implicit processing, and social judgment and decision making—are readily juxtaposed, as is well illustrated by the contents of the preceding chapters. Given the pervasiveness of judgment and decision making events in everyday life, it is not surprising to find that these processes play out at different levels of awareness and conscious effort. If many routine or recurring assessments and decisions in the social domain were not relegated to low-effort processing, social exchange would be hopelessly mired. Thus, understanding implicit, automatic, heuristic judgment processes as well as explicit, deliberative, elaborated processes, and their interplay, is essential to a full theory of social judgment and decision making.
Collectively, the chapters in this volume take on this task of explicating the roles of implicit and explicit processes across various domains of human judgment, with at least two major integrative themes that cut across the various contributions. Each of these themes takes up a particular challenge to prevailing notions in the study of human judgment and decision making. One recurring theme is to challenge to the normative view of judgment and decision making, in which heuristic strategies are labeled as “errors” and “biases,” in favor of a functional, adaptive perspective in which such heuristics are viewed as goal-serving. The second challenge is that issued by Kruglanskiand his colleagues (Kruglanski, Thompson, & Spiegel, 1999; Kruglanski et al., this volume) to prevailing dual process theories of social judgment. I will use these two themes to organize my thoughts and comments on the major points represented in preceding chapters.
Social Judgment: Error-prone or Functionally Adaptive?
The debate over the epistemological status of judgment heuristics is represented in the exchange between Kahneman and Tversky (1996) and Gigerenzer (1996), revolving around Gigerenzer’s claim that it is inappropriate to judge decision making against normative statistical models. Instead, Gigerenzer (2000) has argued that the appropriate question to ask about human reasoning is what are the organism’s goals, particularly in naturalistic decision making contexts. According to this goal-serving perspective, judgmental heuristics are not simply functional as mental “shortcuts,” but adaptively valid decision rules.
Several chapters in this volume take up this debate explicitly, applying an adaptive view to specific judgment phenomena. Haselton and Buss (this volume), for instance, argue from a signal detection framework that specific social judgment biases represent necessary trade-offs between the costs associated with false-positive and false-negative types of error. From a cost-analysis perspective, what appear to be judgmental “flaws” may represent rational design, biased toward committing the less costly error in terms of ultimate reproductive success. A similar adaptive argument is made by Fiedler and Freytag (this volume) in their analysis of judgments based on “pseudo-contingencies.” Both Haselton and Buss and Fiedler and Freytag suggest that specific judgmental heuristics may come “built-in” to human decision making, as a function of their long-run adaptive value. Consistent with this adaptive approach is Funder’s (this volume) call for a renewed focus on accuracy in social perception and personality judgments. He argues that an extraordinary level of accuracy is an achievement of person perception processes, and that inaccuracies may be more a function of the information environment than of perceiver biases or inadequacies.
On the other hand, the “error” view of some aspects of human judgment is at least implicit in other chapters in this volume. For instance, Forgas (this volume) speaks of positive mood effects as responsible for a variety of “errors and distortions” in performance assessments and false memories (p. xx). Errors of judgment are also implicated in treatments of attitude-biased information processing (von Hippel et al., this volume), and stereotype-biased attributions (Johnston & Miles, this volume).
From the prevailing “cognitive miser” perspective on social judgment, many implicit processes are viewed as “short-cuts,” or products of mental “laziness.” From this perspective, heuristics function as time-and-effort-saving mechanisms, not goal-serving devices. But further analysis may reveal an underlying adaptive logic, particularly if we shift attention to serving social (rather than cognitive) goals. In this vein, Shaver and Mikulincer (this volume) argue that attachment strategies may underlie certain negative mood congruence effects, and Williams, Case, and Govan (this volume) suggest that implicit prejudice may serve social inclusion needs. These analyses illustrate how a functional, social psychological perspective may illuminate the adaptive significance of many implicit judgment and decision making phenomena.
Social Judgment: Unimodal or Distinct Processes?
Parallel in many ways to the implicit-explicit distinction in the social judgment literature are the myriad “dual process” models that postulate separate, qualitatively distinct routes to social judgments and decisions (cf. Kruglanski et al., this volume). Although the various dual process models characterize this distinction in somewhat different ways, in most cases, one processing mode is represented as quick, low-effort, relatively superficial, and heuristic-based, whereas the alternative processing mode is more deliberate, thoughtful, slow, and involves effortful processing of more information. Dual process theories hold that these processing styles are fundamentally different in that they utilize different information and engage different underlying processing mechanisms to arrive at a judgment or decision.
Against this backdrop, Kruglanski (Kruglanski et al., 1999; this volume) has issued a challenge to the validity of most dual process claims and poses an alternative “unimodel” theory of human judgment in which a set of judgmental parameters capture the variations in judgment styles that have been attributed to different processes. In this multidimensional (but unimodal) model, modes of judgment differ in degree rather than in kind, and most postulated dual modes can be reconceptualized in terms of the intersection of two continua of motivation and task difficulty. As Kruglanski et al. (this volume) put it, “…the fact that some inferences may occur very quickly, outside the individual’s conscious awareness and with only minimal dependence on cognitive resources…does not imply a qualitatively separate cognitive process.” (p. x, footnote 3).
Kruglanski and his colleagues provide a convincing critique of many of the research paradigms and findings that have been garnered in support of qualitatively distinct judgment processes. Nonetheless, I believe that at least some of the phenemona that have been examined in the present volume do meet the definition of “content-free” dual processes that Kruglanski challenges us to defend. The question is, what criteria can and should be proposed to assess whether processes are qualitatively different or not? I suggest that some useful answers can come from considering an analogy to the distinct subsystems that constitute human sensory processing. Although the different modalities can be viewed as parts of a single sensory system, few would deny that vision, audition, olfaction, etc., represent qualitatively distinct forms of sensory processing. Among other things, they differ in the kind (not simply amount) of information extracted from environmental stimuli, rely on different brain pathways, and invoke different signal transducers. Analogously, we may be able to identify distinct cognitive and affective processing subsystems that differ in parallel ways.
In Table1, I have made an attempt to identify criteria for distinguishing two fundamentally different modes of information processing that may underlie social judgments and decision making. For want of a better set of terms, I have labeled these two modes “deep” and “high,” respectively.[1] In making this particular process differentiation, I am drawing heavily on the ideas presented by Lieberman (this volume; see also Lieberman, Gaunt, Gilbert, & Trope, 2002) and his distinction between “reflexive” and “reflective” judgment processes and their neurological substrates.
Deep processes reflect activity primarily at the subcortical level of the brain, perhaps involving the neural structures associated with Lieberman’s “X-system” (composed of the amygdala, basal ganglia, and lateral temporal cortex). Activation of this deep system, and the environmental cues to which it responds, is either “pre-wired” (a product of our biological evolution) or set by early experience and emotional conditioning. As a repository of long-term stable knowledge, it has the properties of a “slow-learning” memory system, reflecting a large body of experiences acquired slowly and incrementally (including evolutionary time) (McClelland et al., 1995; Sherry & Schacter, 1987). As a consequence, this system operates largely by default, in a reflexive, rigid, and mechanistic manner. Deep processing is rapid, associative (Sloman, 1996; Smith & DeCosta, 2000), probably domain-specific, and takes place automatically and unintentionally.
By contrast to deep processing, the high processing mode involves activation of the prefrontal cortex and reflects explicit learning of symbolic rules, principles, procedures, and cultural norms (Smith & DeCoster, 2000). Processing based on this system is reflective, deliberative, and constructive, rather than mechanistic. It is rule-based and abstract, hence domain-general. High processing is also intentional and capable of overriding (or disrupting) deep processes (see Lieberman, this volume). Although the distinction between deep and high processing modes is not equivalent to the distinction between automatic and controlled processing, there is a contingent association in that controlled, deliberate judgments always (necessarily) involve the high processing system, but automatized, low effort, or subconscious judgments do not always or necessarily implicate the deep processing system.
By far the greatest portion of research on social judgment and decision making has, at least historically, been devoted to studying judgments that rely on the high processing mode. Of special interest to social psychologists more recently is new evidence of how much social information processing takes place in the deep mode (see Lieberman, this volume; also, Haselton & Buss; Shaver & Mikulincer). Amygdala activation, in particular, has been implicated in social categorization and automatic stereotyping, as well as detection of threat cues such as fear expressions. Recently, Lieberman et al. (2002) have suggested that the automatic, behavior identification stage of attribution takes place in the lateral temporal cortex, along a pathway activated by cues associated with behavior intentionality. In addition, short-exposure (30-ms) affective priming (Stapel, this volume) involves extraction of diffuse affective content prior to cognitive appraisal and probably reflects deep level processing of social information. Consistent with this idea, recent experiments using event related brain potential (ERP) measurement (N. K. Smith, Cacioppo, Larsen, & Chartrand, 2002) indicates that differential allocation of attention to negative, potentially threatening stimuli takes place well before information reaches the visual cortex for more elaborate processing. Thus, many of the fundamental components of social judgment appear to arise from processes at subcortical levels of the neurological system.
Cognitive and Motivational Outputs of the Two Systems
Although various brain imaging techniques may eventually provide direct evidence of the activation of different modes of processing, much of our knowledge of these systems will come more indirectly, from the different types of output or products of the processing modes. Thus, an important direction for future research will be identifying what types of judgments, evaluations, and motivational states arise from deep processing versus high processing. Reviewing the wide range of judgmental phenomena covered in the present volume, one realizes that we are a long way from being able to classify or explain specific effects in terms of underlying processing mode distinctions. Assimilation and contrast effects, decision heuristics, mood congruence effects, causal attributions, perceptual and conceptual judgments, impression formation, stereotyping, social comparison, goal activation – all may implicate deep or high processing systems in some way.
Daunting though the task may be to match specific judgmental outcomes to processing modes, I have made an attempt to classify some of the phenomena that have been examined in this volume in terms of the distinction between deep and high processes. Table 2 presents a rough classification of the types of judgments that are generated by the two processing modes, along with some specific examples. The reader will notice immediately that the table has three columns rather than two. In only some cases have I been willing to go out on a limb and classify specific judgment outputs as unequivocally representative of deep or high processing. All of the others fall into an ambiguous intermediate, or “mixed” category, which contains most of what has been studied in the domain of “implicit” judgments, evaluations, and inferences.
The most fundamental output of the deep system is object identification and classification based on pattern matching, which is appropriately characterized by connectionist models (Smolensky, 1988). Pattern matching takes place in the prefrontal cortex as well (Zarate & Stoever, this volume), but Lieberman et al. (2002) summarize evidence for automatic categorical pattern matching involving activation of a pathway along the inferotemporal cortex, prior to any conscious recognition or symbolic representation. In terms of identification, there is little or no difference between judgments involving social and nonsocial objects or movements, but other primitive categorizations generated by the deep processing system are particularly relevant to social stimuli and events. These include evaluative categorization (diffuse affect, relevant to affective priming and social emotions), behavioral intentionality categorization (relevant to attribution judgments), me-not me categorization (relevant to attachment and belonging, ingroup-outgroup discimination), and dominance (relevant to status and power judgments). In addition, some domain-specific decision heuristics (cf. Haselton & Buss, this volume) may reflect deep-level processing, along with automatic assessments of self-relevance, and threat versus security.
Most of the judgment outcomes generated by the deep processing system have parallels in judgments and decisions resulting from high system processing. Associative, pattern-matching identification in the deep system is complemented by rule-based classification, explicit comparison, and reasoning by analogy in the high processing system. Automatic evaluative categorization is matched by explicit attitudes and preference judgments in the high system. Behavioral intentionality assessments feed into explicit attributions and personality judgments; automatic decision heuristics have their parallel in explicit “rules-of-thumb;” and security-threat assessments are represented in the high processing system as attachment schemas (cf Shaver & Mikulincer, this volume) and explicit social motives and goals.
Implicit Processes: Deep or High?
In most of the chapters in the present volume, high processing system outputs are discussed in comparison to related implicit, unintentional, or automatic judgments. Some of these so-called implicit phenomena may arise directly from deep system processing, as indicated above. But the concept of implicit processing is used rather broadly in social cognition to refer to judgments that appear to occur outside of awareness and are unaffected by cognitive load (McClure, Sutton, & Hilton, this volume). Many of the social judgments and motives designated as implicit cannot be readily attributed to deep system processing since they implicate symbolic and semantic representations as well as first-order categorizations and evaluative associations. It is my hope that some heuristic value of this chapter will be derived from questions to be answered about the phenomena in this mixed category. More specifically, can implicit processes be differentiated such that some are direct products of deep processing, some are derivatives of high processing judgments that have become automatized, and some reflect more complex interactions of the outputs of both high and deep processing modes?
One example of the juxtaposition of deep and high system processing is represented in Stapel’s (this volume) analysis of affective priming effects in judgments. It seems clear that the diffuse affect that gives rise to early (i.e., very short exposure duration) affective judgments is a product of deep system processing. But this is quickly supplanted by later (moderately short duration) distinct affect judgments that implicate higher level cognitive appraisal processes. Interestingly, the consequences of affective priming (assimilation or contrast) differ dramatically depending on which system is involved.
By contrast to Stapel’s dual-system analysis of affective priming effects, Chartrand and Jefferis (this volume) discuss automatic goal activation as an automatized derivative of intentional, high system processes. But their integration of automatic motives and automatic goal pursuit strategies, such as mimicry, may implicate deep system processes as well. Mimicry in particular may reflect the activation of inclusion motives derived from deep system assessments of insecurity and/or dominance. Similarly, strategies to enact or to counter prejudice/stereotyping (cf Galinsky & Martorana; Johnston & Miles; this volume), maintain or avoid attachment (Shaver & Mikulincer, this volume), and restore inclusion (Williams et al., this volume.) may be so fundamental to survival that they are activated by assessments generated from deep system processes, rather than merely reflecting automatized high system judgments.
The interplay of implicit and explicit judgments is also represented in the discussion by McClure et al (this volume) of the role of implicit covariation assumptions and explicit rules in goal-based explanations of social behavior, in the analysis by Bless, Schwarz, and Wanke (this volume) of the size of context effects in social judgments, in Forgas’s (this volume) analysis of affective influences in elaborative judgments, in Suls’s (this volume) proxy model of the role of social comparison in self-assessment of abilities and opinions, in the analysis by Galinsky and Martorana (this volume) of the differential effects of implicit processes in determining the outcomes of suppression versus perspective-taking as strategies for controlling stereotype activation, and in the analysis by Johnston and Miles (this volume) of the consequences of dispositional versus situational attributions for stereotype-consistent behaviors. In all of these domains, however, it is not clear whether the implicit aspects involve the activation of deep system processes in any way, or whether they might well be incorporated in a “unimodel” framework as argued by Kruglanski et al. (this volume), or, alternatively, whether they implicate yet a third processing system that is qualitatively distinct from either deep or high systems as they have been defined here. It seems to me that this poses an interesting question for further understanding of the role of implicit judgment processes in social cognition.