Context effects -- 1
The Size of Context Effects in Social Judgment
Herbert Bless
University of Mannheim, Germany
Norbert Schwarz
University of Michigan
Michaela Wänke
University of Erfurt, Germany
Running head: Size of context effects
Draft for the 5th Sydney Symposium on Social Psychology, March 2002
The reported research was supported by grants Bl 289/5 from the Deutsche Forschungsgemeinschaft to H. Bless, N. Schwarz, & M. Wänke. Correspondence should be addressed to Herbert Bless, Fakultät für Sozialwissenschaften, Universität Mannheim, D-68131 Mannheim, Germany.
Context effects -- 1
Introduction
All of us have first-hand experience with the context-dependency of social judgment. In one situation we consider ourselves assertive, a politician trustworthy, and the stock market full of potential. At other moments our assertiveness seems less compelling, the politician less trustworthy and the stock market less enticing. In many cases the variation of judgments is due to change of judgmental context. This context influences which information comes to mind – or how the accessible information is used. So simply by changing the judgmental context, we may come up with out different evaluations about our assertiveness, the politician or the stock market.
The context dependency of social judgment is perhaps one of the most fascinating phenomena in social psychology. Countless findings from experimental studies and natural observations indicate that the evaluation of a target stimulus may be either assimilated or contrasted to the context in which the stimulus is presented. We refer to assimilation effects whenever the judgment reflects a positive relation between the implications of some piece of information and the judgment, and refer to contrast effects whenever the judgment reflects a negative (inverse) relationship of the judgment and the implications of some piece of information.
Given the interest in the context dependency of social judgment it is no surprise that a wide range of conceptual approaches have been offered. In general, it seems that there is agreement among researchers that assimilation and contrast effects may reflect the operation of a number of different processes, which have often been conceptualized in independent theories. One group of models focused on the distribution of the contextual stimuli, assuming that they influence the adaptation level (Helson, 1964), standard of comparison (e.g., Thibaut & Kelley, 1959), or scale anchor (Ostrom & Upshaw, 1968, Parducici, 1965) used in making a judgment. Another group of models focused on categorization processes (e.g. Herr, Sherman & Fazio, 1983, Martin, Seta, & Crelia, 1990; Tajfel, 1981, Turner, 1987). According to these models, assimilation effects are likely to emerge when the target stimulus and the context stimuli are assigned to the same category, whereas contrast effects may emerge when they are assigned to different categories.
Emphasizing the role of information accessibility and categorization processes we have proposed the inclusion/ exclusion model (Schwarz & Bless, 1992a) as a general framework for conceptualizing assimilation and contrast effects in social judgment. Building on theorizing by Barsalou (1987), Herr et al. (1983), Kahneman & Miller (1986) and Martin (1986), the model offers predictions about the direction and the size of context effects. In the present chapter we focus on the latter, often neglected, aspect: What determines the size of assimilation and contrast effects? We first review the core assumptions of the inclusion/exclusion model. Next, we elaborate on the model's predictions pertaining to the size of context effects and review supporting evidence. Finally, we highlight the simultaneous operation of direct and indirect context effects, which presents a rarely addressed complication in predicting the size of contextual influences.
The Inclusion/Exclusion Model
The inclusion/exclusion model (Schwarz & Bless, 1992a, see also Bless & Schwarz, 1998; Schwarz, Bless & Wänke, in press) assumes that the influence of a given piece of information depends on how it is used. Evaluative judgments that are based on features of the target (rather than on the perceiver's affective response; see Schwarz & Clore, 1996) require two mental representations, namely a representation of the target and a representation of a standard against which the target is evaluated. Both representations are formed on the spot, drawing on information that is chronically or temporarily accessible. Information that is used in forming a representation of the target results in assimilation effects; that is, the inclusion of positive (negative) information results in a more positive (negative) judgment. Conversely, information that is used in forming a representation results in a contrast effect; that is, more positive (negative) information results in a more positive (negative) standard, against which the target is evaluated less (more) favorably. Hence, the same piece of accessible information can have opposite effects, depending on how it is used. The variables that influence information use can be organized by assuming that perceivers tacitly ask themselves three questions, which serve as filters that channel information use.
Why Does It Come to Mind?
The first filter is: "Am I only thinking of this information because it was brought to mind due to some irrelevant influence?" If so, the accessible information is not used in forming a representation of the target. Accordingly, awareness of the priming episode, for example, undermines use of the primed information, resulting in contrast effects (e.g., Lombardi, Higgins, & Bargh, 1987; Martin, 1996; Strack et al., 1993).
Does it Bear on the Target?
When the information passes this first test, the second filter is: "Does this information represent a feature of the target?" This decision is driven by the numerous variables known to influence the categorization of information, including the information's extremity and typicality (e.g., Bless & Wänke, 2000; Bless, Schwarz, Bodenhausen, & Thiel, 2001), as well as the presentation format and related context variables (for reviews see Schwarz & Bless, 1992a; Martin, Strack, & Stapel, 2001).
The studies reviewed in the present chapter take advantage of a particularly robust determinant of categorization, namely the categorical relationship between context information and the target of judgment (e.g., Bless, Igou, Schwarz & Wänke, 2000; Schwarz & Bless, 1992b; Stapel & Schwarz, 1998). Suppose, for example, that a preceding question brings a particularly untrustworthy politician to mind (say, Richard Nixon) and participants are subsequently asked to judge the trustworthiness of "American politicians" in general. Nixon is a member of the superordinate target category "American politicians." He is therefore included in the representation formed of that category, resulting in an assimilation effect in the form of lower trustworthiness judgments. Suppose, however, that participants are instead asked to rate the trustworthiness of another specific politician (say, Newt Gingrich). In this case, Nixon cannot be included in the representation formed of Gingrich because lateral categories (like exemplars) are mutually exclusive -- Gingrich is, after all, not Nixon. Nevertheless, Nixon may influence the judgment because he can be used in forming a representation of a standard, against which Gingrich is evaluated. This results in a contrast effect, making Gingrich look more trustworthy by comparison (e.g. Schwarz & Bless, 1992b). Hence, the same context information (Nixon) can, paradoxically, decrease judgments of the trustworthiness of the group (superordinate category), but increase judgments of the trustworthiness of every individual group member (lateral categories) assessed. Our subsequent discussion of the size of context effects will draw on this robust phenomenon, which highlights the role of categorization processes in the emergence of assimilation and contrast effects.
Conversational Norms
The third and final filter pertains to the norms of conversational conduct that govern information use in conversations: "Is it conversationally appropriate to use this information?" Conversational norms prohibit redundancy and invite speakers to provide information that is new to the recipient, rather than information that the recipient already has (for a review see Schwarz, 1996). Hence, highly accessible information is not used when it violates this conversational norm, again resulting in contrast effects (e.g., Schwarz, Strack, & Mai, 1991; Strack, Martin, & Schwarz, 1988).
Information that passes all three tests is included in the representation formed of the target and results in assimilation effects. Information that fails any one of these tests is excluded from the representation formed of the target, but may be used in forming a representation of the standard, resulting in contrast effects. In addition to specifying the direction of context effects, the model offers predictions about the size of context effects. In the remainder of this chapter we will focus on this aspect and will discuss the direction of context effects only as far as necessary (for overviews see Schwarz & Bless, 1992a; Schwarz et al, in press).
Set-Size Effects in Mental Construal
Determinants of the size of assimilation and contrast effects have often been neglected in psychological research. One of the more obvious variables is the extremity of the accessible context information: The more extreme the information is that is included in the representation of the target, or the representation of the standard, the larger is the resulting assimilation or contrast effect (Schwarz & Bless, 1992a). Hence, a major political scandal would influence judgments of politicians' trustworthiness more than a minor scandal, for example. While this logic seems straightforward, a potential caveat needs to be noted. Although more extreme exemplars should be more influential, extreme exemplars are also more likely to be perceived as atypical and may consequently trigger exclusion from, rather than inclusion in, the representation formed of the superordinate category (see Bless & Wänke, 2000; Schwarz & Bless, 1992a). This, in turn, results in contrast effects (e.g., Herr, Sherman, & Fazio, 1983), rather than more pronounced assimilation effects. Moreover, it is worth noting that it is not the extremity of the context information per se but its relative difference to otherwise accessible information that influences the size of an effect: If the chronically accessible information is as extreme as the context information, including the context information in the temporary representation formed of the target has little additional impact.
Theoretically more interesting is a variable that may increase as well as decrease the size of context effects, namely the amount of information used in forming a representation of the target or the standard. As many models of social judgment assume (e.g., Anderson, 1981; Wyer & Srull, 1989), the impact of a given piece of information decreases as the overall amount of information considered increases. We assume that this general principle also holds when individuals construct mental representations of the target and the standard of comparison.
Returning to our earlier example, suppose that an individual constructs a mental representation of a social category ("American politicians") and an extreme exemplar (Nixon) comes to mind, due to contextual influences. Including this exemplar in the representation formed results in an assimilation effect, as discussed. The size of this effect, however, should depend on the amount and extremity of other information used in forming this representation. This logic is depicted in the left panel of Figure 1, where either three other six exemplars (E) are used in forming a representation of a superordinate target category. Leaving aside potential differences in the weighing of the exemplars, adding Nixon would contribute either one fourth or on seventh to the representation of the social category. In the latter case, the impact of Nixon would hence be smaller than in the former case. These predictions are consistent with numerous studies that investigated information integration in social judgment (for a review see Anderson, 1981). For example, Schwarz, Strack, and Mai (1991) observed in a different content domain that inducing participants to think about the quality of their marriage affected subsequent judgments of general life-satisfaction more when marriage was the only life-domain brought to mind than when marriage was only one of four life-domains addressed in preceding questions. As expected, the accessibility of a larger amount of competing information attenuated the impact of marriage related information.
Conversely, bringing three, rather than one, extreme exemplars to mind would increase the impact of context information, reflecting that more temporarily accessible information is added to the representation of the target.
Figure 1 about here
The inclusion/exclusion model assumes that the same set-size principle applies to the construction of standards of comparison, an issue that has not been addressed in information integration research. If standards of comparison are constructed on the basis of accessible information, the size of contrast effects should parallel the above predictions. First, the impact of a given extreme piece of information should decrease as a function of the amount of other information used in constructing the standard (see right panel of Figure 1), resulting in attenuated contrast effects. Conversely, contrast effects should increase, the more extreme information is included in the representation of the standard.
In summary, the inclusion/exclusion model predicts that the size of context effects is a function of the amount and extremity of the information used in forming temporary representations of the target and the standard. On the one hand, the size of assimilation and contrast effects should increase, the larger the amount and the extremity of the information used in forming the respective representation. This prediction is consistent with numerous models of social judgment and will not be further pursued in this chapter. On the other hand, the impact of extreme context information should be attenuated the more other information with different implications is used in forming the respective representation. Accordingly, including additional exemplars in the representation of the social category should attenuate the resulting assimilation effect on judgments of the social category, whereas including additional exemplars in the representation of the standard should attenuate the resulting contrast effect on judgments of other exemplars.
The following discussion of set size effects will primarily focus on this second set of predictions. We first address the influence of differential amounts of temporarily accessible information and subsequently turn to a discussion of chronically accessible information, which bears on the role of expertise in the emergence of context effects.
Temporary Accessibility:
Reducing Context Effects By Adding Context Information
Building on Schwarz and Bless' (1992b) scandal study, we tested the implications of our set-size assumption by varying the number of accessible exemplars (see Bless, Igou, Schwarz, & Wänke, 2000). We asked German participants to evaluate either the trustworthiness of “German politicians in general“ (superordinate category) or the trustworthiness of specific well-known politicians (lateral categories). Prior to making these judgments, participants received a list of politicians who, either currently or in the recent past, held the office of prime minister in one of the states of the Federal Republic of Germany. Participants were asked to indicate the state each prime minister represented. As a context manipulation, this list either did or did not include a prime minister who was involved in a well-known scandal, namely Max Streibl, the former prime minister of Bavaria. Consistent with our previous findings, based on a different scandal (Schwarz & Bless, 1992b), we again observed that bringing the scandal-ridden politician (Max Streibl) to mind decreased judgments of the trustworthiness of German politicians in general, but increased judgments of the trustworthiness of specific exemplars who were not involved in the scandal, as shown in the left hand panel of Table 1.