To appear in Gideon Keren, ed., Perspectives on Framing. New York: Psychology Press / Taylor & Francis.
*I thank Denis Hilton and Gideon Keren for extremely helpful comments, and Samara Klar and Thomas Leeper for research assistance.
How do people form preferences? This question is fundamental for social scientists across disciplines. Psychologists seek to understand how people think, feel, and act, and preferences often reflect or determine these activities. Sociologists explore how preferences stem from and impact social interactions. Economists, particularly in light of the trend towards behavioral economics, often study the causes and consequences of preferences that deviate from well-defined, self-interested motives. Political scientists, for whom citizens’ preferences serve as the basis for democratic governance, investigate the roots of political preferences as well as the extent to which governing elites respond to and influence these preferences.
In some ways, there is fruitful inter-disciplinary collaboration on understanding the causes and consequences of preferences; in other ways, cross-discipline communication is lacking. Both these perspectives are apparent when one considers the idea of “framing.” Framing receives substantial attention across the social sciences—for many, it plays an important role in explaining the origins and nature of preferences. Yet, “framing” continues to be used in different and sometimes inconsistent ways across (and even within) disciplines. For example, some reserve it to refer to semantically distinct but logically equivalent portrayals, such as 95% unemployment versus 5% employment, while others employ a relaxed definition that includes emphasis on any alternative consideration (e.g., economic concerns versus humanitarian concerns when thinking about welfare). In short, although more than a decade old, Nelson, Oxley, and Clawson’s (1997: 222) claim that the “heightened interest in frames... conceals a lack of conceptual clarity and consistency about what exactly frames are…” still seems accurate (also see Fagley and Miller 1997: 357, Kühberger 1998, Levin, Schneider, and Gaeth 1998: 151, Druckman 2001a: 226-231, 2004, McCombs 2004: 89, Sniderman and Theriault 2004).
In this chapter, I attempt to reduce this conceptual ambiguity. I begin by offering a simple model of preference formation that makes clear exactly what frames are and how they might work. This enables me to draw a distinction between prominent usages of the framing concept. I then focus on a particularly relevant conceptualization used in political science. I review work that shows how political elites (e.g., politicians, the media) engage in framing, and how these frames influence political opinion formation. A brief summary concludes.
What Is A Frame?
To explain what framing is, I begin with the variable of ultimate interest: an individual’s preference. A preference, in essence, consists of a rank ordering of a set of objects or alternative actions. For example, an individual might prefer the socialist party to the environmental party to the conservative party, the immediate withdrawal of foreign troops in Iraq to piecemeal withdrawal, a defined benefit retirement program to a defined contribution one, or chocolate ice cream to vanilla to strawberry. In some definitions, particularly those used by economists, the rank orderings must possess specific properties including transitivity (e.g., if one prefers chocolate to vanilla, and vanilla to strawberry, then he/she must prefer chocolate to strawberry too) and invariance where different representations of the same choice problem should yield the same preference (e.g., a person’s preference should not change if asked whether he/she “prefers chocolate to vanilla” as compared to being asked if he/she prefers “vanilla to chocolate”) (Tversky and Kahneman 1987).
Preferences over objects derive from comparative evaluations of those objects (Hsee 1996); for example, an individual prefers the socialist party to the conservative party if he/she holds a relatively favorable evaluation of the socialists (Druckman and Lupia 2000). Social psychologists call these comparative evaluations attitudes, which is “a person’s general evaluation of an object (where ‘object’ is understood in a broad sense, as encompassing persons, events, products, policies, institutions, and so on)” (O’Keefe 2002: 6). It is these evaluations (i.e., attitudes) that underlie preferences.
A common portrayal of an attitude is the expectancy value model (e.g., Ajzen and Fishbein 1980, Nelson, Oxley, and Clawson 1997) where an attitude toward an object consists of the weighted sum of a series of evaluative beliefs about that object (this portrayal is akin to utility theory). Specifically, Attitude = ∑vi*wi, where vi stands for the evaluation of the object on attribute i and wi stands for the salience weight (∑wi = 1) associated with that attribute. For example, one’s overall attitude, A, toward a new housing development might consist of a combination of negative and positive evaluations, vi, of the project on different dimensions i. An individual may believe that the project will favor the economy (i=1) but harm the environment (i=2). Assuming this individual places a positive value on both the economy and the environment, then v1 is positive and v2 is negative, and his attitude toward the project will depend on the relative magnitudes of v1 and v2 discounted by the relative weights (w1and w2) assigned respectively to each attribute (Nelson and Oxley 1999).
The general assumption of the expectancy value model that an individual can place different emphases on various considerations about a subject serves as a useful abstraction for discussing framing. This conceptualization applies to any object of evaluation (and, thus, any set of objects over which individuals have preferences). For instance, a voter’s attitude toward a party may depend on whether the voter favors the party on dimensions such as platform issues and leadership that are of varying importance (e.g., economic issues may be seen as being more important than foreign affairs and leadership experience) (see Enelow and Hinich 1984). The voter might prefer one party (e.g., conservatives) when the evaluations are based on foreign affairs (e.g., foreign affairs receives considerable weight) but another when based on economic considerations (e.g., socialists). As another example, the extent to which an individual assigns blame to a welfare recipient may depend on evaluations of the recipient’s personal efforts to stay off of public assistance (dimension 1) and the situational factors that the recipient has faced (dimension 2) (see Iyengar 1991). Similarly, one’s tolerance for a hate group rally may hinge on the perceived consequences of the rally for free speech, public safety, and other values, with each value receiving a different weight. For these examples, if only one dimension matters, the individual places all of the weight (wi = 1) on that dimension in forming his attitude. Without loss of generality, ican be thought of as a dimension (Riker 1990), a consideration (Zaller 1992), a value (Sniderman 1993) or a belief (Ajzen and Fishbein 1980).
The dimension or dimensions—the “i’s”—that affect an individual’s evaluation constitute an individual’s frame in thought. This is akin to Goffman’s (1974) depiction of how frames organize experiences or Johnson-Larid’s (1983) mental model. If an individual, for example, believes that economic considerations trump all other concerns, he/she would be in an “economic” frame of mind. Or, if free speech dominates all other considerations in deciding a hate group’s right to rally, the individual’s frame would be free speech. If instead, he/she gave consideration to free speech, public safety, and the effect of the rally on the community’s reputation, then his/her frame of mind would consist of this mix of considerations. The examples given thus far constitute what scholars call “emphasis” frames, “issue” frames, or “value” frames. For these cases, the various dimensions of evaluation are substantively distinct—that is, one could reasonably give some weight to each consideration such as free speech and public safety or the economy and foreign affairs. The varying weights placed on the dimensions often play a decisive role in determining overall attitudes and preferences (e.g., more weight to free speech leads to more support for the rally).
Another type of frame is “equivalency” or “valence” frames. In this case, the dimensions of evaluation are identical; this typically involves casting the same information in either a positive or negative light (Levin, Schneider, and Gaeth 1998: 150). The most famous example is Tversky and Kahneman’s (1981) Asian disease problem. This problem—which is described in detail in the introductory chapter—shows that individuals’ preferences shift depending on whether equivalent outcomes are described in terms of the number of lives saved out of 600 (e.g., 200 are saved) as opposed to the number of lives lost (e.g., 400 are lost). Analogous examples include more favorable evaluations of an economic program when the frame (dimension) is the employment rate rather than the unemployment rate, a food product when the frame is the percentage fat free rather than the percentage of fat, and a crime prevention programs when the frame is the percentage not committing crime instead of the percentage of criminals (e.g., Levin, Schneider, and Gaeth 1998, Piñon and Gambara 2005, O’Keefe and Jensen 2006). Related to equivalency or valence framing effects are question wording effects in surveys (see Druckman 2001a, Bartels 2003 for discussion).
Unlike emphasis or value frames, the dimensions in equivalency frames are not substantively distinct and are in fact logically equivalent. Thus, one’s evaluation should not inherently (or ideally) differ based on the dimension of evaluation (e.g., one should not change his/her evaluation of an economic program when he/she thinks about it terms of 95% employment instead of 5% unemployment). The fact that preferences tend to differ reflects cognitive biases that also violate the aforementioned invariance axiom of preference formation.
In sum, a frame in thought can be construed as consisting of the dimensions on which one bases his/her evaluation of an object. These dimensions involve either substantively distinct considerations (i.e., emphasis frames) or logically equivalent ones (i.e., equivalency frames). In both cases, the frame leads to alternative representations of the problem and can result in distinct evaluations and preferences.
Frames in Communication
The frame that one adopts in his/her mind (e.g., the dimensions on which evaluations are based)—and that, consequently, can shape preferences—stems from various factors including prior experiences, ongoing world events, and so on. Of particular relevance is the impact of communications from others including friends and family, and in the case of politics, politicians and the media. In presenting information, speakers often emphasize one dimension or another; in so doing, they offer alternative frames in communication. For example, if a speaker, such as news outlet, states that a hate group’s planned rally is “a free speech issue,” then the speaker invokes a “free speech” frame (emphasis frames). Alternatively, in describing an economic program, one can emphasize its consequence for employment or unemployment (equivalency frames).
Frames in communication and frames in thought are similar in that they both concern variations in emphasis or salience (see Druckman 2001b). However, they differ with the former usage focusing on what a speaker says (e.g., the aspects of an issue emphasized in elite discourse), while the latter usage focusing on what an individual thinks (e.g., the aspects of an issue a citizen thinks are most important) (also see Entman 1993). In this sense, the term frame refers to two distinct, albeit related entities; as Kinder and Sanders (1996: 164) explain, “frames lead a double life… frames are interpretative structures embedded in political discourse. In this use, frames are rhetorical weapons… At the same time, frames also live inside the mind; they are cognitive structures that help individual citizens make sense of the issues…” When a frame in communication affects an individual’s frame in thought, it is called a framing effect.
When it comes to studying frames in communication and concomitant framing effects, a few clarifications are in order. First, it makes sense to define a frame in communication as a verbal or non-verbal statement that places clear emphasis on particular considerations (on non-verbal frames, see Grabe and Bucy 2009, Iyengar 2010). Other types of communications that do not explicitly highlight a consideration (e.g., a factual statement such as “a hate group has requested a permit to rally”) may still affect individuals’ frames in thought, but such an effect does not make the statement a frame in communication (i.e., the speech act should not be defined based on its effect) (see Slothuus 2008 for a more general discussion). Frames in communication sometimes will and sometimes will not influence individuals’ frames in thought. For example, a free speech activist or a journalist who possesses strong beliefs in free speech are unlikely to be influenced by a public safety frame when it comes to a hate group rally—in other words, such individuals have clearly defined prior beliefs that prevent a frame in communication from exerting an effect (also see, e.g., Furnham 1982 on how values condition attributions).
Second, many scholars employ the concept of frames in communication to analyze trends in elite discourse. For example, Gamson and Modigliani (1987) show that, over time, opponents of affirmative action shifted from using an undeserved advantage frame to a reverse discrimination frame. That is, the discourse changed from questions such as “have African Americans earned or do they deserve special rights?” to the question of “is it fair to sacrifice the rights of whites to advance the well-being of African Americans?” Analogous examples include studies on support for war (e.g., Dimitrova, Kaid, Williams, and Trammell 2005), stem cell research (Nisbet, Brossard, and Kroepsch 2003: 48), cynicism toward government (Brewer and Sigelman 2002), and the obesity epidemic (Lawrence 2004). These analyses provide insight into cultural shifts (Schudson 1995, Richardson and Lancendorfer 2004: 75), relative media biases (Tankard 2001), and public understanding (Berinsky and Kinder 2006). For now, however, I turn to a discussion of how frames in thought exert their effects on attitudes, as this provides insight into when a frame in communication will influence one’s preference. (In what follows, I do not regularly distinguish frame “in thought” from those “in communication” as it should be clear from the context to which I am referring.)
The conceptualization of frames in thought as constituting the dimensions on which one bases his/her attitudes leads straightforwardly to a psychological model of framing (Chong and Druckman 2007a). The starting point is that individuals typically base their evaluations on a subset of dimensions, rather than on the universe of possible considerations (e.g., Ajzen and Sexton 1999). At the extreme, they focus on a single dimension such as foreign policy or economic affairs in evaluating a party, free speech or public safety when considering a hate group rally request, or lives saved or lives lost in assessing medical programs. Even when they incorporate more than one dimension, there exists a limit such that individuals rarely bring in more than a few considerations (e.g., Simon 1955). The dimensions used are available, accessible, and applicable or appropriate (Price and Tewksbury 1997, Althaus and Kim 2006; also see Higgins 1996, Chen and Chaiken 1999). (Recall I construe frames in thought as consisting of the dimensions of evaluation.)
A consideration must be stored in memory to be available for retrieval and use in constructing an attitude (e.g., Higgins 1996). For instance, an individual needs to understand how a hate group rally might threaten public safety, or how the First Amendment pertains to unpopular political speech for these considerations to matter. Similarly, the individual must understand how the unemployment (or employment) rate connects to a given economic program. A consideration is available only when an individual comprehends its meaning and significance.
Accessibility refers to the likelihood that an available consideration exceeds an activation threshold to be used in an evaluation (e.g., Higgins, Rholes, and Jones 1977). Put another way, the available consideration stored in long-term memory enters an individual’s mind when forming an evaluation. Increases in accessibility occur through “passive, unconscious processes that occur automatically and are uncontrolled” (Higgins and King, 1981: 74).1
Accessibility increases with chronic or frequent use of a consideration over time or from temporary contextual cues—including communications (e.g., frames in communication)—that regularly or recently bring the consideration to mind (Bargh, Bond, Lombardi, and Tota 1986, Bargh, Lombardi, and Higgins 1988). Repeated exposure to a frame, such as frequently hearing someone emphasize free speech or lives lost, induces frequent processing, which in turn increases the accessibility of the frame.2
The impact of an accessible consideration also can depend on its applicability or appropriateness to the object being evaluated (e.g., Strack, Martin, and Schwarz 1988). For instance, concern that a rally will tie up traffic may be available and accessible, but the individual may view it as irrelevant and give it no weight. The likelihood that a consideration raised by a frame will be judged applicable and shape an individual’s opinion increases with conscious perceptions of its strength or relevance (Eagly and Chaiken 1993, Nelson, Clawson, and Oxley 1997).
Individuals do not, however, always engage in applicability evaluations (as it requires conscious processing); doing so depends on personal and contextual factors (Higgins 1996, Druckman 2004). Individuals motivated to form an accurate attitude will likely deliberately assess the appropriateness of a consideration (Fazio 1995, Ford and Kruglanski 1995, Stapel, Koomen, and Zeelenberg 1998).3 The information context also matters, as the introduction of conflicting or competitive information (e.g., multiple, alternative frames) can stimulate even less personally motivated individuals to engage in conscious, deliberate assessments of the appropriateness of competing considerations (Lombardi, Higgins, and Bargh 1987, Strack, Martin, and Schwarz 1988, Martin and Achee 1992, Jou, Shanteau, and Harris 1996). On the other hand, individuals lacking personal motivation or the stimulus of competition likely rely uncritically on the considerations made accessible through exposure to a message—applicability or appropriateness is a non-factor in these cases. They base their preferences on whichever frames happen to be accessible.
When conscious processing occurs, the perceived applicability or strength of a frame depends on two factors. Strong frames emphasize available considerations; a frame focused on unavailable considerations cannot have an effect (i.e., it is inherently weak). The other factor is the judged persuasiveness or effectiveness of the frame. This latter factor is akin to what Pan and Kosicki (2001: 49) call “framing potency” (also see McCombs 2004: 91-97 on “compelling arguments”).
Empirically, frame strength is established by asking individuals (e.g., in a pre-test) to rate the effectiveness or persuasiveness of various frames in communication, on a particular issue. For example, study participants may view a hate group rally frame emphasizing free speech as effective and one highlighting traffic problems as less compelling. If so, then, when individuals are motivated (by individual interest or the context) to engage in applicability evaluations, only the free speech frame should have an impact since they will assess and follow strong frames only. If, however, motivation to evaluate applicability is absent, then either frame might matter since individuals rely on any accessible frame. While this approach to operationalizing frame strength is empirically practical—since it allows a researcher to isolate strong as opposed to weak frames—it leaves open the important question of why a particular frame is seen as strong. This is a topic to which I will later return.
Equivalency Frames Versus Emphasis Frames
The portrayal of an individual’s frame as depending on the availability, accessibility, and, at times, applicability of distinct dimensions applies to emphasis (i.e., issue or value frames) and equivalency (i.e., valence) frames. It also accentuates the difference between them. Equivalency frames have their differential effects when an individual bases the evaluation on whatever dimension—such as lives saved or employment—happens to be accessible (see Jou, Shanteau, and Harris 1996: 9, Levin, Schneider, and Gaeth 1998: 164-166, Druckman 2004). Accessibility can increase due to the description of a problem (which is akin to a frame in communication), as when an individual learns of medical programs described in terms of lives lost. If an individual engages in applicability evaluations due to motivation or contextual conditions, the differential impact of the logically equivalent frames should dissipate. This occurs because the individual will consciously recognize that deaths can be thought of as lives saved or employment equals unemployment, and thus, will not focus on gains or losses (e.g., he/she will recognize the equivalency). This renders the framing effect mute. The individual will not be in losses frame of mind, but rather, will consider losses and gains.4
Druckman (2004) offers evidence along these lines. Specifically, he replicates traditional framing effects—using four distinct problems5—finding, for example, that those exposed to a negative frame (e.g., money lost) exhibit distinct preferences from those receiving an equivalent positive frame (e.g., money gained) (also see Hsee 1996). The effects disappear, however, among participants who receive multiple competing frames (e.g., both the gain and loss frames). The competitive information context presumably stimulates applicability evaluations leading participants to recognize the equivalency of the frames, making them ineffective. Other work shows that, even in non-competitive environments, motivated individuals—such as those with high cognitive ability (Stanovich and West 1998) or strongly held attitudes (Levin Schneider, and Gaeth 1998: 160)—exhibit substantially less susceptibility to equivalency framing effects (also see Fagley and Miller 1987, Miller and Fagley 1991, Takemura 1994, Sieck and Yates 1997). This further supports the idea that when stimulated to assess applicability, individuals recognize alternative ways of viewing the problem—they appreciate that one can construe the problem as lives lost or lives saved—and the equivalency framing effects vitiate. In other words, motivation leads one to recognize the equivalent ways of viewing the problem.
The effect of applicability evaluations differs when it comes to emphasis framing, where individuals consider substantively distinct dimensions (e.g., free speech and public safety, or foreign affairs and the economy). Conscious recognition and evaluations of these dimensions will not lead individuals to view them as identical (as with equivalency frames); instead, individuals will evaluate the dimensions’ strengths. As explained, strength involves availability, but perhaps more importantly, persuasiveness—which dimension is most compelling? In their study of opinions about limiting urban sprawl, Chong and Druckman (2007b) exposed some participants to a communication using both a pro community frame (e.g., limiting urban sprawl creates dense, stronger communities) and a con economic costs frame (e.g., limiting sprawl will increase housing prices). As with equivalency framing effects, exposure to multiple competing frames likely stimulated applicability evaluations.6 However, unlike the equivalency framing case, with emphasis framing the evaluations do not mute the effects. This is the case because thinking about alternative frames does not lead one to conclude they are logically equivalent (since they are not); instead, individuals evaluate the substantive strength of the alternative dimensions.
Chong and Druckman had previously identified the economic frame as being strong and the community frame as being weak (based on the previously discussed pre-test approach where individuals rate the effectiveness of various frames). As expected, then, only the economic costs frame influenced opinions; competition did not cancel out competing frames but rather led to the strong frame winning. Unfortunately, Chong and Druckman (2007b), like most others, offer little insight on what factors lie behind relative strength.
The effect of individual motivation similarly differs in its effect for the distinct types of framing effects. Unlike with equivalency frames, emphasis frames often have larger effects on motivated individuals. These individuals have the ability to connect distinct considerations to their opinions (i.e., they have a broader range of available considerations) (Chong and Druckman 2007a: 110-111). In the urban sprawl study, Chong and Druckman (2007b) find that a single exposure to the open space frame only affected knowledgeable participants because open space was relatively more available. Chong and Druckman (2007b: 647) explain that “less knowledgeable individuals require greater exposure to the open space… frame before their opinion shifts…. Knowledgeable individuals may be quicker to recognize the significance of a frame.”
Many communication scholars distinguish framing effects from priming, agenda setting, and persuasion (e.g., Scheufele 2000). The utility of these distinctions remains unclear (Chong and Druckman 2007c). The term “priming” entered the field of communication when Iyengar and Kinder (1987: 63) defined it as: “By calling attention to some matters while ignoring others, television news influences the standards by which governments, presidents, policies, and candidates for public office are judged. Priming refers to changes in the standards that people use to make political evaluations” (also see Iyengar, Kinder, Peters, and Krosnick 1984). For example, individuals exposed to news stories about defense policy tend to base their overall approval of the president (or some other political candidate) on their assessment of the president’s performance on defense. Thus, if these individuals believe the president does an excellent (or poor) job on defense, they will display high (or low) levels of overall approval. If, in contrast, these individuals watch stories about energy policy, their overall evaluations of the president’s performance will tend to be based on his handling of energy policy.
Importantly, this connotation of priming, in the political communication literature, differs from how most psychologists use the concept. For instance, Sherman, Mackie, and Driscoll (1990: 405) state that “priming may be thought of as a procedure that increases the accessibility of some category or construct in memory.” The typical “procedure” for increasing accessibility is not the same as exposing individuals to continual media emphasis of an issue. Yet, Iyengar and Kinder and many others assume, to the contrary, that media emphasis on an issue passively increases the accessibility of that issue. Miller and Krosnick (2000) present evidence to the contrary in claiming that the effects of media emphasis on an issue do not work through accessibility and thus is not akin to priming as defined by psychologists (for discussion, see Druckman, Kuklinski, and Sigelman n.d.).
In my view, the psychological model of framing presented above can be generalized to political communication priming by assuming that each consideration constitutes a separate issue dimension or image (Druckman and Holmes 2004) on which the politician is assessed. When a mass communication places attention on an issue, that issue will receive greater weight via changes in its accessibility and applicability. If this is correct, then framing effects and what communication scholars have called priming effects share common processes and the two terms can be used interchangeably (also see Chong and Druckman 2007a: 115). (Again, this argument does not apply to how psychologist employ the term priming.)
A similar argument applies to agenda setting, which occurs when a speaker’s (e.g., a news outlet or politician) emphasis on an issue or problem leads its audience to view the issue or problem as relatively important (e.g., McCombs 2004). For example, when a news outlet’s campaign coverage focuses on the economy, viewers come to believe the economy is the most important campaign issue. This concept straightforwardly fits the above psychological model, with the focus (i.e., dependent variable) lying on assessments of the salience component of the attitude (rather than the overall evaluation of the object). The aforementioned example can be construed as the news outlet framing the campaign in terms of the economy, and the researcher simply gauging the specific salience weights (wi) as the dependent variable.
A final conceptual distinction concerns framing and persuasion. Nelson and Oxley (1999) differentiate framing from persuasion by referring to the former as a change in the weight component, wi, of an attitude in response to a communication, and the latter as a change in the evaluation component, vi(also see Johnston, Blais, Brady, and Crete 1992: 212, Miller and Krosnick 1996: 81, Wood 2000). For example, in assessing a new housing project, framing takes place if a communication causes economic considerations to become more important relative to environmental considerations. Persuasion occurs if the communication alters one’s evaluation of the proposal on one of those dimensions (e.g., by modifying one’s beliefs about the project’s economic consequences). This distinction stems from the focus in most persuasion research on the evaluation components of an attitude. The key, yet to be answered questions, is whether the processes (and mediators and moderators) underlying changes in the weight and evaluations components differ; if they do not, then perhaps the concepts should be studied in concert. These are not easy issues to address, however, since piecing apart the specific process by which a communication influences overall attitudes is not straightforward.
My argument that these various concepts, unless evidence can be put forth to establish meditational and/or moderator distinctions, can all be enveloped under a single rubric (and psychological model) should facilitate further theoretical development (also see Iyengar 2010). It will enable scholars studying priming, agenda setting, and framing to avoid redundancy and focus more on pressing unanswered questions.