Joseph P. Forgas
University of New South Wales, Sydney, Australia
How does temporary mood influence the way people think and behave in social situations? Although most people are intuitively aware that feelings can have a profound influence on their thoughts, judgments and behaviors, we do not yet fully understand how and why these influences occur. This paper presents an integrative review of past and present ideas about the role of affect in social behavior, and offers a theoretical explanation of these effects based on the multi-process Affect Infusion Model (AIM; Forgas, 1995a). A series of programmatic experiments looking at affective influences on social thinking, judgments, and behaviors carried out in our laboratory will also be described. It will be argued that the different information-processing strategies people adopt in different situations play a key role in promoting, inhibiting, or even reversing affective influences on cognition and behavior. Indeed, the evidence to be reviewed here will suggest that it is the very complexity and indeterminacy of many social situations that promotes affective influences on interpersonal behaviors. The principle appears to be that the more complex and ambiguous a social situation, the more likely it is that people will need to engage in open, elaborate, and constructive thinking, drawing on their own memory-based ideas in order to produce an appropriate response. A number of theories as well as empirical studies now predict that such open, elaborate processing strategies are especially likely to be influenced by affective states (Fiedler, 1991; Forgas, 1995a).
Since the dawn of civilization, understanding the delicate relationship between affect and cognition, feeling and thinking, has been one of the recurrent puzzles that occupied artists, writers, and philosophers. The basic idea that affect may overwhelm and subvert rational processes has been echoed in many social and psychological theories throughout the ages. For example, the psychoanalytic view of affect suggested that managing affective states required considerable psychological resources. However, important advances in social cognition, neuroanatomy, and psychophysiology during the last decade led to the recognition that affect is not necessarily a disruptive influence on social thinking and behavior (Adolphs & Damasio, 2001; Ito & Cacioppo, 2001). For example, research with brain-damaged patients showed that individuals who suffer lesions that interfere with affective reactions but leave cognitive capacities intact tend to make disastrous social decisions and their social relationships suffer accordingly, even though their intellectual problem-solving ability may be unimpaired (Damasio, 1994).
Recent affect-cognition research also emphasizes the close interdependence between feeling and thinking in human social life (Forgas, 2000a; Zajonc, 2000). Much of the present discussion will focus on the role of moods in interpersonal behavior. Moods, unlike emotions, are relatively low-intensity, diffuse, and enduring affective states that have no salient cause and little cognitive content (such as feeling good or feeling bad, being in a happy or sad mood). Paradoxically, because moods tend to be less subject to conscious monitoring and control, and often elude awareness, their effects on social thinking, judgments, and behavior tend to be potentially more insidious, enduring, and subtle.
Freud's psychodynamic theory played a key role in highlighting the important, if often latent, influence of affect on interpersonal behavior. According to this view, affective impulses are largely located within the id, and in seeking expression, they exert “pressure” against the countervailing forces of rational, controlled ego mechanisms. These ideas rapidly permeated popular thinking about affect. In a typical early study Feshbach and Singer (1957) tested the psychoanalytic prediction that attempts to suppress affect should generate greater “pressure” for affect infusion into unrelated judgments or behaviors. They found that fearful subjects were in fact more likely to “perceive another person as fearful and anxious” (p. 286), and this effect was especially strong when subjects were trying to suppress their fear, because “suppression of fear facilitates the tendency to project fear onto another social object” (p. 286). However, such psychoanalytic explanations soon declined in popularity, because of the absence of any direct evidence for the psychodynamic mechanisms postulated, and the devastating epistemological criticisms leveled at psychoanalysis as fundamentally unscientific by philosophers such as Karl Popper and others.
Although radical behaviorism contributed little to our understanding of affective phenomena, the Watsonian principle (Watson & Rayner, 1920) that affect may influence unrelated thoughts and judgments through conditioned associations was eventually applied to complex social situations by Byrne and Clore (1970; Clore & Byrne, 1974). According to this view, experiencing aversive or pleasant environments (the unconditioned stimuli) can produce an affective reaction (the unconditioned response) that can become associated with an otherwise neutral target (such a person first met in that environment). Thus, Griffitt (1970) found that people who felt bad after exposure to excessive heat and humidity made more negative judgments of a newly encountered person, indicating that “evaluative responses are determined by the positive or negative properties of the total stimulus situation” (p. 240). Other experiments by Gouaux (1971) used a similar paradigm and found that prior exposure to happy or depressing films influenced liking for another person, as if attraction was “a positive function of the subject's affective state” (p. 40). It is interesting that whereas mood congruence in the 1960s and 1970s was largely explained in terms of “blind” conditioning principles, contemporary studies focus on the cognitive, information-processing mechanisms that link feelings and thinking (Berkowitz, Jaffee, Jo, & Troccoli, 2000; Clore, Schwarz, & Conway, 1994). Ultimately, neither psychoanalytic nor conditioning theories could account for the apparent situation- and context-specificity of affect infusion. Recent social cognitive theories focus on the information-processing mechanisms responsible for mediating affective influences on thinking and action.
By the early 1980s, affect was once again to occupy center stage in psychological theorizing (Bower, 1981; Zajonc, 1980). However, it soon transpired that affective influences on cognition are most likely to occur in circumstances that facilitate open, constructive processing and thus promote the incidental use of affectively primed information (Bower & Forgas, 2001; Eich & Macauley, 2000; Fiedler, 1991; Forgas, 1995a). “Constructive” processing may be defined as involving the active elaboration and transformation of the available stimulus information, requiring the activation and use of previous knowledge structures, and resulting in the creation of new knowledge from the combination of stored information and new stimulus details. It now appears that affective influences on social cognition and behavior are largely dependent on what kind of information-processing strategies people employ to deal with a particular task, as also argued in the Affect Infusion Model (Forgas, 1995a).
Two theories that explain affective influences on social judgments and behavior have received empirical support. The affect-as-information model (Schwarz & Clore, 1988) argues that people may directly use their affect as information when inferring a response to social situations. The alternative affect priming theory predicts that affect should influence social thinking and behavior through selectively priming affect-related constructs, facilitating their use when planning and executing social behaviors (Bower, 1981; Bower & Forgas, 2001).
According to this view, when “computing a judgment individuals may ask themselves: ‘How do I feel about it?’, and, in doing so, they may mistake feelings due to a pre-existing state as a reaction to the target” (Schwarz, 1990, p. 529). Thus, affective influences occur because of an inferential error. For example, when people have just seen happy or sad films and are then asked to respond to unexpected questions in a street survey, there is a significant pattern of affect congruence in their reactions (Forgas & Moylan, 1987). Similarly, off-the-cuff responses to an unexpected telephone survey may show similar effects (Schwarz & Clore, 1988), as respondents may simply rely on their mood to infer a positive or negative response. However, most strategic interpersonal behaviors involve more extensive processing, and are unlikely to be influenced in such a simple way by misattributed affect. A complementary theory, the affect priming model, can explain how affect infusion occurs when more elaborate processing strategies are used.
According to the affect priming model (Bower, 1981), affect is an integral aspect of cognitive representations about the social world. Affect thus automatically primes related ideas and memories, facilitating their use in tasks that require constructive thinking. Numerous studies support these predictions (Baron, 1993; Bower, 1981; Clark & Isen, 1982; Forgas & Bower, 1987; Isen, 1984, 1987; Sedikides, 1992, 1995). Current research suggests that affect infusion into strategic behaviors and judgments is most likely when actors face complex and demanding social situations that require the use of open, constructive processing strategies that facilitate the incidental use of affectively primed information (Fiedler, 1991, 2000; Forgas, 1995a; 1999a,b; Sedikides, 1995). Theories such as the Affect Infusion Model (AIM; Forgas, 1995a) specifically argue that the nature and extent of affective influences on social behavior should critically depend on the kind of information-processing strategies people employ.
Affect may not only influence the content of thinking and behavior, but also the process of cognition, that is, how people think about and evaluate social information (Bless, 2000; Clark & Isen, 1982; Fiedler, 2000; Fiedler & Forgas, 1988). Early research suggested that positive affect produces less effortful and more superficial processing strategies; in contrast, negative affect seemed to trigger a more effortful, analytic, and vigilant processing style (Clark & Isen, 1982; Isen, 1984, 1987; Schwarz, 1990). More recent work showed, however, that people in a positive mood are more likely to adopt creative, open, constructive, and inclusive thinking styles, use broader categories, show greater cognitive and behavioral flexibility, and perform well on secondary tasks (Bless, 2000; Fiedler, 2000; Isen, 1987). Based on such evidence, Fiedler and Bless (in press) argued that the processing consequences of affect can best be understood in terms of a fundamental dichotomy between accommodation and assimilation, a distinction originally discussed by Piaget. Accommodation involves focusing on the demands of the external world. Assimilation is a complementary process where the individual relies on well-established internal schemas and routines to respond to a situation. Thus, positive affect should promote a more assimilative, schema-based, top-down processing style, whereas negative affect produces a more accommodative, bottom-up, and externally-focused processing strategy (Bless, 2000; Fiedler, 2000; Higgins, 2001). These processing consequences of affect do have significant implications for how people plan and execute various social tasks, as the research to be reviewed here will suggest.
The Affect Infusion Model was developed as an integrative theory that can explain both the presence and the absence of affective influences on social thinking and behavior in various circumstances. The model maintained that social actors rely not just on one, but on a variety of processing approaches available to them when dealing with social situations. The multi-process Affect Infusion Model incorporates the affect priming principle, but specifically predicts that this mechanism should only lead to mood congruence when constructive, substantive processing is used. The model also predicts that mood congruity is relatively easily eliminated in circumstances when a response can be based on reproducing prior reactions, or when motivational goals come to dominate responding (Berkowitz et al., 2000; Fiedler, 1991; Forgas, 1990, 1991). Thus, the AIM (Forgas, 1995a) predicts that affect infusion should only occur in circumstances that promote an open, constructive processing style, and also specifies the kind of task-, person-, and situation-characteristics that produce such a processing style (Fiedler, 1991; Forgas, 1992b, 1995b).
The model identifies four alternative processing strategies people may use in responding to a social situation: direct access, motivated, heuristic, and substantive processing. These four strategies differ in terms of two basic dimensions: the degree of effort exerted in seeking a solution, and the degree of openness and constructiveness of the information search strategy. The combination of these two processing dimensions, quantity (effort), and quality (openness/constructiveness), produces four distinct processing styles: direct access processing (low effort, closed, not constructive), motivated processing (high effort, closed, not constructive), heuristic processing (low effort, open, constructive), and substantive processing (high effort, open, constructive) (Fiedler, 2001). According to the AIM, affect infusion is most likely when a constructive processing strategy is used, such as substantive or heuristic processing. In contrast, affect should not influence the outcome of closed, merely reconstructive tasks involving motivated or direct access processing (see also Fiedler, 1991, 2001).
The direct access strategy involves the direct retrieval of a pre-existing response, and is most likely when the task is highly familiar and when no strong cognitive, affective, situational, or motivational cues call for more elaborate processing. The motivated processing strategy involves highly selective and targeted thinking that is dominated by a particular motivational objective. This strategy also precludes open information search, and should be impervious to affect infusion (Clark & Isen, 1982). In fact, motivated processing may also produce a reversal of affect infusion effects, for example, when the goal is affect control (Berkowitz et al., 2000; Erber & Erber, 2001; Forgas, 1991; Forgas & Fiedler, 1996).
The two other processing styles, heuristic and substantive processing, do require a degree of constructive and open-ended information search, and thus facilitate affect infusion. Heuristic processing is most likely when the task is simple, familiar, of little personal relevance, cognitive capacity is limited, and there are no motivational or situational pressures for more detailed processing. Heuristic processing can lead to affect infusion as long as people adopt the “how do I feel about it” heuristic to produce a response (Clore et al., 1994; Schwarz & Clore, 1988). Substantive processing requires individuals to select, encode, and interpret novel information and relate this information to their pre-existing memory-based knowledge in order to produce a response. Substantive processing is promoted when the task is unusual, demanding, complex, or personally relevant, there are no direct access responses available, there are no clear motivational goals to guide processing, and there are adequate time and other processing resources available. Substantive processing is an inherently open and constructive strategy, allowing affect to prime access to, and facilitate the use of, related thoughts, ideas, memories, and interpretations. The AIM makes the interesting and counterintuitive prediction that affect infusion should be greater when more extensive processing is required to deal with more complex, demanding, or novel tasks. This nonobvious pattern has now been confirmed in several experiments (Forgas, 1992b; 1993; 1995b; 1998a,b; Forgas & Bower, 1987; Sedikides, 1995).
The AIM is fundamentally a contextual theory of affective influences on social behavior. The model asserts that personal and situational factors combine to recruit one of four distinct kinds of processing strategies that moderate the kind of influence that affective states can have on behavior. The AIM also incorporates the mood-congruence mechanisms proposed by affect-priming and affect-as-information theories, and specifies the circumstances under which they are most likely to be used. The AIM was formulated in such a way that its empirical predictions can be clearly tested. The definition of four distinct processing strategies as moderators opens up whole new avenues of research where analyzing the information-processing strategies of social actors becomes a key aspect of understanding affective influences on their behaviors. The implications of this model have now been tested in a number of experiments, as the next section will summarize.
Skilled social behavior necessarily requires constructive inferential thinking (Heider, 1958). One of the key predictions of the AIM is that affect infusion should be significantly dependent on the processing strategies used. The more people need to draw on their pre-existing, memory-based ideas and interpretations to produce a response, the more likely that affectively primed information will influence the outcome.
This prediction has been tested in a number of experiments where the complexity of the social situation was manipulated to create more or less demand for an open and substantive processing style. For example, if one observes usual or unusual people in a public setting such as a restaurant or a cafe, a well-matched couple is much more “typical” and should require less elaborate and constructive processing than do couples where the partners are obviously mismatched in terms of features such as age or physical attractiveness. Several recent experiments tested this prediction (Forgas, 1993, 1995b). In a controlled replication of the restaurant scenario, participants feeling happy or sad after viewing standard mood induction films were presented with images of well-matched or badly matched couples. Their judgments showed significant mood congruence as happy participants formed more positive impressions of the couples than did sad participants. However, when the couples were atypical and badly matched, mood had a much greater effect on judgments than for couples who were typical and well matched (Forgas, 1993, 1995b). In fact, the size of mood effects on judgments was strongest when the couples were most mismatched, intermediate when they were partly matched, and smallest when they were well matched (Forgas, 1995b).
Similar results were also obtained when we asked participants to respond to people who varied in terms of their prototypicality (Forgas, 1992a). An analysis of processing latency and recall memory data confirmed that forming impressions about more unusual and atypical persons took longer, and there was correspondingly greater affect infusion into these more constructive responses. Surprisingly, the same kinds of effects can also be obtained when people respond to highly realistic social information, such as making judgments about their own intimate relationships (Forgas, 1994). In a counterintuitive pattern, mood effects were consistently greater when more extensive, constructive processing was required to deal with more complex and serious rather than simple, everyday interpersonal issues. Jointly, these series of experiments provide strong evidence for the process sensitivity of affect infusion into cognition and interpersonal behaviors. Similar effects may also influence the way people interpret real-life social behaviors, as the next section will suggest.
As making sense of ambiguous observed behaviors by definition requires some degree of substantive processing, there should also be affect infusion into behavior interpretation. This hypothesis was tested (Forgas, Bower, & Krantz, 1984) by asking happy or sad participants to monitor and rate their own and their partner's behaviors on a videotape recording their social interactions on a previous occasion. As predicted, happy people “saw” significantly more positive, skilled and fewer negative, unskilled behaviors both in themselves and in their partners than did sad subjects. Objective observers who received no mood manipulation displayed no such monitoring biases. In terms of the AIM, these effects occur because affect priming subtly influences the kinds of interpretations, constructs, and associations that become available as people evaluate intrinsically complex and indeterminate social behaviors in the course of substantive, inferential processing. The same smile that is seen as “friendly” in a good mood may be judged as “awkward” when the observer is in a negative mood; discussing the weather could be seen as “poised” in a good mood but “boring” when in a bad mood, and so on.
Later experiments found that these effects also extend to the way people evaluate themselves. People in a negative mood made more critical, self-deprecatory interpretations of their own behaviors, but those in a positive mood selectively looked for and found lenient and optimistic explanations for identical outcomes (Forgas, Bower, & Moylan, 1990). Rather surprisingly, such mood-induced distortions can also influence reactions to highly familiar, intimate events involving close partners (Forgas, 1994). When partners in long-term intimate relationships were asked to evaluate their own and their partner's behaviors in more or less serious interpersonal conflicts, positive mood produced lenient, self-serving explanations. These mood effects were even stronger when the events judged were more complex and serious and thus required more constructive processing.
Affect can also influence the kind of information-processing strategies people adopt. We recently found, for example, that the kind of vigilant, systematic attention to external stimulus details recruited by negative moods tends to reduce or even eliminate such common judgmental mistakes as the fundamental attribution error (FAE; Forgas, 1998c). As positive affect often produces a more schematic, top-down, and heuristic processing style (Bless, 2000), the schematic perception of a “unit relation” between the actor and the act leading to the FAE appears to be promoted by positive affect, and reduced by negative affect. This pattern was also confirmed in unobtrusive field studies. People who had just seen happy or sad films made judgments about the writers of popular and unpopular essays in an ostensible “street survey”. Positive affect again increased and negative affect decreased the FAE. Happy judges mistakenly assumed that the views advocated in a coerced essay were in fact the writer's own, but sad judges realized that the essays were coerced and discounted their relevance. Subsequent mediational analyses (Forgas, 1998c, Exp. 3) confirmed that these attributional biases were due to affect-induced differences in processing strategies: sad judges also thought longer about their responses than did happy judges.
Affective influences on processing strategies also influence eyewitness accounts of observed social events. In a recent investigation (Forgas, 2000b) participants witnessed complex social events such as a wedding or a robbery on a videotape. Some time later, they were exposed to a mood induction, and then received questions about the incident that either included, or did not include, “planted” misleading details. After a further interval, participants’ recognition memory for the incidents was tested. Temporary positive mood at the time when the misleading information was presented significantly increased, and negative mood decreased the mistaken incorporation of planted information into eyewitness memories. These effects were replicated in a field study, where students observed a staged incident during a lecture (Forgas, 2000b). About a week later, they were induced to feel good or bad through watching videotapes, and were then exposed to questions about the incident that included or did not include planted incorrect information. Once again, those feeling good while hearing the planted information were more likely later to incorporate “planted” details heard during the questioning with actually witnessed details. In contrast, negative affect reduced the incidence of such mistakes. These results, together with the evidence for mood effects on inferential mistakes such as the FAE, confirm that mild, transient affective states can have a marked influence on the way people process, interpret, and remember observed social behaviors. Does affect also impact on actual interactive behaviors? This possibility was explored in several recent studies.
As long as an open, constructive strategy is used, positive affect should prime positive information and produce more confident, assertive, optimistic and cooperative “approach” behaviors, whereas negative affect should prime negative memories and produce more avoidant, defensive, or unfriendly behaviors. This prediction was evaluated in a simple experiment we carried out in collaboration with Anoushka Gunawardene (Forgas & Gunawardene, 2000). Female undergraduates were first induced into a positive or negative mood by watching happy or sad videotapes as part of an “unrelated” study. Next, they participated in an interview about student life with a confederate, and their behavior was recorded by a hidden video camera. These videotapes were subsequently observed and rated by trained judges who were blind to the manipulation. Results indicated a significant pattern of mood-congruence: happy participants displayed more smiles, communicated more, disclosed more personal information about themselves, and generally behaved in a more poised, skilled, and rewarding manner. Sad participants were generally rated by observers as significantly less friendly, confident, relaxed, comfortable, active, interested, and competent than were happy participants (Figure 1).
Making a request is a difficult and complex interpersonal task. Requesting involves uncertainty, and requesters must try to maximize the likelihood of compliance (by being direct), yet avoid the danger of giving offence (by not being too direct). In terms of the AIM, happy people should adopt a more confident, direct requesting style, as a result of the greater availability and use of positively valenced thoughts and associations in their minds (Forgas, 1998b, 1999a,b). Further, in terms of the AIM these mood effects should be greater when the request situation is more complex and difficult, and requires more substantive processing. This prediction was tested in several experiments. In one study, mood was induced by asking participants to recall and think about happy or sad autobiographical episodes (Forgas, 1999a, Exp. 1). Next, participants were asked to identify more or less polite request forms they would prefer to use in an easy, routine and a difficult, embarrassing request situation. Happy participants generally preferred more direct, impolite requests, whereas sad persons preferred more cautious, indirect, and polite requests. Further, mood effects on requesting were much stronger when the request situation was demanding and difficult, and required more extensive, substantive processing (Figure 2).
In a follow-up experiment, participants formulated their own open-ended requests, which were subsequently rated for politeness and elaboration by trained judges (Forgas, 1999a, Exp. 2). Positive mood again produced more direct, impolite and less elaborate requests than did negative mood, and these mood effects were greater when the request situation was difficult and problematic. Further studies also showed (Forgas, 1999b, Exp. 1) that affective influences on request preferences seem greatest when people consider using direct, impolite and unconventional requests that most clearly violate cultural conventions, and thus recruited the most substantive, elaborate, processing strategies.
Overall, these results show that affect infusion into requesting is increased or reduced depending on just how much open, constructive processing is required to deal with a more or less difficult interpersonal task. These mood effects also occur in real-life interactions. In an unobtrusive experiment (Forgas, 1999b, Expt. 2), affect was induced by asking participants to view happy or sad films. Next, in an apparently impromptu development, the experimenter casually asked participants to get a file from a neighboring office. Their words in requesting the file were recorded by a concealed tape recorder.
Negative mood resulted in significantly more polite, elaborate, and more hedging requests than did positive mood (Figure 3). Those in a negative mood were also more hesitant, and delayed making their requests for significantly longer. An analysis of recall memory for the exact words they used—an index of elaborate processing—showed that recall accuracy was positively related to the degree of affect infusion, as predicted by the AIM.
Frequently we must respond almost instantaneously to a new social situation. When such events involve uncertainty and require constructive processing, responses should also be subject to affect infusion effects. This prediction was evaluated in a series of recent field experiments (Forgas, 1998b), where we assessed how people respond to being unexpectedly approached by another person in a public place such as a university library. Affect was induced by leaving folders containing pictures (or text) designed to induce positive or negative mood on unoccupied library desks. Students entering the library were surreptitiously observed as they exposed themselves to the mood induction. A few minutes later, they were approached by another student (in fact, a confederate) who made an unexpected polite or impolite request for several sheets of paper needed to complete an essay. Their responses were noted.
A short time after this incident a second confederate approached the participants and explained that the situation was in fact staged, and asked them to complete a brief questionnaire assessing their perception and evaluation of the request and the requester, and their recall of the request. Students who received the negative mood induction were significantly more likely to report a critical, negative evaluation of the request and the requester and were less likely to comply than were positive mood participants. In a particularly interesting result, mood and the politeness of the requester had an interactive effect on responses: mood effects were markedly greater when responding to an impolite and unconventional rather than a polite request. An analysis of later recall memory for the incident confirmed the more substantive processing (and better recall) of impolite requests. Conventional, polite requests, on the other hand, were processed less substantively, were recalled less well, and responses were less influenced by mood. These results confirm that affect infusion into responding to unexpected situations is also significantly moderated by the processing strategy people employ.
If affect infusion is a function of substantive, elaborate processing, we might expect affective states to play a particularly important role in elaborately planned interpersonal encounters such as bargaining and negotiating encounters (Forgas, 1998a). In these studies, mood was induced by giving participants positive, negative, or neutral feedback about their performance on a verbal test. Next, they engaged in highly realistic interpersonal and intergroup negotiation in what they believed was a separate experiment. The question we were interested in was how temporary moods might influence people's goals, plans, and, ultimately, their behaviors in this interaction.
Results showed that happy participants were more confident about the encounter, formed higher expectations about their success, and also planned and used more optimistic, cooperative, and integrative strategies than did control or negative mood participants (Figure 4). Surprisingly, these mood effects on bargaining behavior actually produced significantly better outcomes for happy participants than for those who were feeling bad. These findings clearly suggest that even slight changes in affective state due to an unrelated prior event influenced the goals that people set for themselves, the action plans they formulated, and the way they ultimately behaved and succeeded in strategic interpersonal encounters.
In these studies we also found that high Machiavellians and those high in need for approval were less influenced by mood in formulating their plans and behaviors than were low scorers on these measures. Theoretically, as implied by the AIM, affect infusion should be constrained for individuals who habitually approach interpersonal tasks such as bargaining from a highly motivated perspective. It is almost as if high Machiavellians and those high in need for approval had their minds made up about what to do even before they started, thus limiting the extent of open processing and incidental affect infusion. There is growing evidence that individual differences related to information-processing styles play an important role in mediating affective influences on social cognition and behavior. For example, we also found that people who score low on traits such as Openness to Feelings seem less likely to be influenced by affect in how they respond to social information than are high scorers (Ciarrochi & Forgas, 2000). Trait anxiety can also moderate affect infusion (Ciarocchi & Forgas, 1999). Low anxious people respond in a mood-congruent, negative manner to an out-group when experiencing aversive mood. However, high trait anxious individuals displayed an opposite pattern, consistent with their greater sensitivity to aversive affect, and reliance on motivated processing designed to eliminate negativity.
Amateur persuaders—and that means all of us—must produce their persuasive strategies on-line, and be sensitive to the situation and any feedback they receive. As we have seen, affect can significantly influence information- processing strategies. People in a negative mood often pay greater attention to the demands of the situation (Bless, 2000; Fiedler, 2000; Forgas, 1998a,b). Thus, negative affect may also have a beneficial influence on some social influence strategies such as the production of persuasive messages. Despite much research on affective influences on how people respond to persuasion (Petty, DeSteno, & Rucker, 2001) there has been little work on how affect influences the production of persuasive messages. This question was investigated in several recent unpublished experiments (Forgas, Ciarrochi, & Moylan, 2000a). In the first study, student participants were induced into positive or negative mood by watching short videotapes. After the mood induction, they were asked to produce persuasive arguments either for, or against (1) the proposition that student fees should be increased, and (2) Aboriginal land rights in Australia (most students were against fees, and for land rights). Each participant argued the popular position on one issue and the unpopular position on the other, in a 2 × 2 design, with mood and issue popularity as the independent variables. The arguments produced were rated by two judges blind to the manipulations in terms of their overall quality, persuasiveness, and valence (positivity-negativity). As quality and persuasiveness were strongly correlated (r = .78), these two scales were combined. Mood significantly influenced argument quality: those in a negative mood produced significantly higher-quality and more persuasive arguments than did happy persuaders. Issue popularity had no main or interactive effects on argument quality, indicating that mood influenced argument quality irrespective of the issues argued. An analysis of argument valence showed a trend towards a mood-congruent pattern, as happy persons produced more positive, and sad persons produced more negative, arguments. A second experiment confirmed these findings. This time, happy and sad participants were asked to argue either for or against Australia becoming a republic, and for or against the right-wing One Nation party. Results showed that sad mood resulted in arguments that were of higher quality and more persuasive than by happy persons, with an intermediate performance by the neutral group (Figure 5). This result is consistent with negative mood promoting a processing style that is more attuned to the requirements of a particular situation.
However, happy participants produced more arguments overall than did sad participants. This result is interesting because it contradicts some earlier suggestions that the processing benefits of negative mood are simply due to more effortful thinking. It seems that affective influences on argument quality were not simply caused by more or less effort being extended. Rather, they occurred because of the more or less top-down, internally driven or bottom-up, situationally oriented processing styles promoted by feeling good or feeling bad (Bless, 2000; Fiedler, 2000). The next, third study replicated these effects, using a different, autobiographical mood induction method.
In Experiment 4 persuasive arguments were produced while interacting with a “partner” through a computer keyboard as if exchanging e-mails. In fact, the computer was preprogrammed to “respond” indicating increasing agreement or disagreement. Motivation to be persuasive was also manipulated, by offering some participants a significant reward (the chance to win highly desired movie passes). The task was to persuade another student to volunteer for an unpleasant experiment, an adaptation of the well-known forced-compliance persuasion procedure used in cognitive dissonance research (Festinger, 1957). Responses by the “partner” communicated increasingly positive, accepting (“I am somewhat persuaded ”, “I like what you are saying ”, . . “I agree to do the study”), or increasingly negative and rejecting reactions to the persuasive messages (“I somewhat disagree ”, “I don't like what you are saying ”, “I don't want to do the study”). Participants readily accepted the genuineness of the procedure.
Two raters rated each argument for quality, complexity, persuasiveness, originality, and valence. As ratings of argument quality and originality were correlated, an analysis of covariance was carried out. There was a significant mood main effect, as the negative mood group generated significantly higher quality arguments than the neutral group, who in turn did better than the positive group. In an interesting pattern, this mood effect was further qualified by a significant interaction with the reward condition. Mood had a greater effect on argument quality in the low reward condition than the high reward condition. This finding is consistent with the Affect Infusion Model, and shows that the provision of a reward reduced the size of mood effects on argument quality by imposing a strong motivational influence on how the task was approached. These experiments provide convergent evidence that even slight changes in mood can produce profound differences in the quality and effectiveness of persuasive arguments. Similar results were obtained both in hypothetical situations and in realistic interactions, with a variety of attitude topics, using a range of different mood induction procedures and irrespective of the popularity and social desirability of the position argued. These results make sense in terms of our theoretical predictions, and suggest that negative affect promotes a more externally focused and bottom-up information-processing style that was previously also found to lead to the reduction of some attribution errors and the reduction of eye-witness memory distortions (Forgas, 1998c). This kind of processing also produces discernible behavioral effects, leading to more persuasive interpersonal strategies.
The research reviewed so far shows that affective states have a marked informational and processing influence on the way people perform strategic interpersonal behaviors. In this section we will consider some of the possible extensions and applications of this research to some related areas of inquiry. As positive affect often promotes a schematic, top-down information-processing strategy (Bless, 2000), it also facilitates reliance on group stereotypes as long as there are no other demands for more elaborate processing. Once activated, stereotyped knowledge may guide responses such as the allocation of rewards to members of ingroups and outgroups (Forgas & Fiedler, 1996). Negative affect can also function as a warning signal, indicating the need for a motivated reassessment of potentially prejudiced responses. We found that this “alerting” effect of negative mood is particularly strong for individuals who are habitually anxious and score high on trait anxiety (Ciarrochi & Forgas, 1999).
Affect also plays an important role in mediating the attitude-behavior link, as several recent reviews indicate (Harmon-Jones, 2001; McGuire & McGuire, 2001; Petty et al., 2001). For example, as cognitive dissonance involves feelings of arousal and negative affect, affective states can influence the presence and extent of dissonance reduction strategies (Harmon-Jones, 2001). Several studies suggest that positive affect decreases, and negative affect increases dissonance reduction even if the source of affect is unrelated. And conversely, once consonance is restored, affective state also tends to improve (Harmon-Jones, 2001).
Affect also has an additional important influence on social behavior: positive mood may serve as a resource that allows people to overcome defensiveness and deal more effectively with potentially threatening situations (Trope, Ferguson, & Ragunanthan, 2001). People in a positive mood are more likely to expose themselves to threatening but diagnostic information from others. This mood-as-a-resource hypothesis may well also extend to behavioral effects. The evidence given earlier, demonstrating the superior bargaining performance by people in a good mood, certainly supports such a hypothesis (Forgas, 1998a). People in positive mood also process in greater detail and remember negatively valenced arguments about health risks better (Trope et al., 2001).
The role of affect in health-related behaviors is receiving growing attention (Salovey, Detweiler, Steward, & Bedell, 2001). Positive or negative affective states may promote healthy or unhealthy attitudes and behaviors, and may ultimately also influence physical wellbeing. Numerous studies found a clear correlation between good moods and positive health outcomes. For example, sick students who were suffering from cold or flu reported nearly twice as many aches and pains when feeling sad than those made to feel happy—even though there were no differences between the two groups before the mood induction (Salovey et al., 2001). It is not really surprising that ill-health should be associated with negative moods. The more interesting possibility is that induced mood may have a causal influence on health symptoms. Individuals who experience negative moods report more and more severe physical symptoms, and report more negative attitudes and less confidence in their ability to manage their health. Optimism, for example, can have a direct effect not only on health-related behaviors, but also on immune system functioning (Salovey et al., 2001). Recent studies confirm that affective states can also have a significant influence on many organizational behaviors such as personnel selection, appraisal, consumer reactions, and bargaining strategies (Ciarrochi & Forgas, 2000; Forgas, 1998a). Different information-processing strategies seem to play a critical role in moderating these effects (Forgas & George, in press).
The research reviewed here also implies that substantive processing and motivated processing may function as two countervailing cognitive strategies that jointly constitute a dynamic, self-correcting mood management system (Ciarrochi, Forgas & Mayer, in press). Substantive processing facilitates affect infusion and the maintenance and accentuation of the prevailing affective state; once affect intensity reaches a threshold level, motivated processing may be triggered producing affect-incongruent responses, leading to an attenuation of the affective state. Such an automatic switch from affect infusion to affect control was illustrated in a suggestive study by Sedikides (1994), and was further developed in the mood-management model developed by Forgas (2001; Forgas, Ciarrochi, & Moylan, 2000b). This model predicts that mood should initially lead to affect-infusion until a threshold level of negativity is reached, at which point people should automatically switch to motivated mood control and mood-incongruent responses (Figure 6). Several recent studies evaluated this hypothesis (Forgas & Ciarrochi, in press), and found that when responses are monitored over time, initial mood congruence spontaneously gives way to mood incongruent responses, especially in negative moods. In later experiments, we found that participants who score high on self-esteem are more effective in managing their moods in this way.
The existence of a homeostatic mood-management mechanism involving the spontaneous alternation between substantive processing (affect infusion) and motivated processing (affect control; Figure 6) suggests that in normal subjects at least, extreme mood effects on thinking and judgments are likely to be relatively short-lived and spontaneously reversed. Indeed, many experimental studies suggest that laboratory mood induction typically does not last very long. These mood management strategies are likely to influence not only cognitive responses, but also behavioral choices in complex situations.
This paper argued that mild everyday affective states do have a significant influence on the way people perceive and interpret social behaviors, and the way they plan and execute strategic interactions. Different information-processing strategies seem to play a key role in explaining these effects. Multiprocess theories such as the Affect Infusion Model (Forgas, 1995a) offer a simple and parsimonious explanation of when and how affect infusion into social behaviors occurs. Several experiments found that more extensive, substantive processing enhances mood congruity effects, consistent with the predictions of the AIM (Forgas, 1992b; 1994; 1995b). The paper also reviewed a number of empirical studies demonstrating how such principles can be translated into behavioral research, and how affective states impact on both simple and complex interpersonal behaviors.
These experiments show that affect can influence behavior monitoring and interpretation, as well as the actual performance of interpersonal behaviors, such as the formulation of and responses to requests; the planning and execution of strategic negotiations; and the production of persuasive arguments. In contrast, affect infusion is reduced or absent whenever a social cognitive task could be performed using a simple, well-rehearsed direct access strategy, or a highly motivated strategy. In these conditions, there is little need and little opportunity for incidentally primed mood-congruent information to infuse information processing (Fiedler, 1991; Forgas, 1995a). Several of the field experiments also showed that affect infusion occurs not only in the laboratory, but also in many real-life situations. These findings have many applied implications. Affect is likely to have a significant influence on relationship behaviors, group behaviors, organizational decisions, consumer preferences, and health-related behaviors. The tendency to alternate between substantive and motivated processing strategies, producing affect infusion and affect control respectively, could also be considered as part of an ongoing homeostatic strategy of self-regulatory mood management (Forgas, 2001; Forgas et al., 2000b).
Other studies showed that affective states have an important and asymmetrical processing effect on cognition. It seems that positive mood increases and negative mood decreases judgmental errors in how interpersonal behavior is monitored and explained, such as the fundamental attribution error (Forgas, 1998c), and the accuracy of eyewitness accounts (Forgas, 2000b).
In summary, this paper argued that different information-processing strategies play a key role in explaining how affect influences social cognition and interpersonal behavior. The Affect Infusion Model in particular offers a parsimonious integrative account of the conditions likely to facilitate or inhibit affect infusion processes. Much of the evidence reviewed here suggests that affect infusion is most likely in conditions requiring constructive, substantive processing. Other processing strategies such as direct access or motivated processing result in the absence, or even reversal, of affect infusion. Obviously a great deal more research is needed before we can fully understand the multiple influences that affect has on interpersonal behavior. Hopefully, this summary will help to stimulate further interest in this fascinating and important area of inquiry.
This work was supported by a Special Investigator award from the Australian Research Council, and the Research Prize by the Alexandor von Humboldt Foundation. The contribution of Joseph Ciarrochi, Stephanie Moylan, Patrick Vargas, and Joan Webb to this project is gratefully acknowledged.
Adolphs, R., & Damasio, A. (2001) The interaction of affect and cognition: A neurobiological perspective. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Baron, R.A. (1993) Affect and organizational behavior: When and why feeling good (or bad) matters. In J.K. Murnighan (Ed.), Social psychology in organizations: Advances in theory and research (pp. 63–88). Englewood Cliffs, NJ: Prentice-Hall.
Berkowitz, L., Jaffee, S., Jo, E., & Troccoli, B.T. (2000) On the correction of feeling-induced judgmental biases. In J.P. Forgas (Ed.), Feeling and thinking: The role of affect in social cognition (pp. 131–152). New York: Cambridge University Press.
Bless, H. (2000) The interplay of affect and cognition: The mediating role of general knowledge structures. In J.P. Forgas (Ed.), Feeling and thinking: The role of affect in social cognition (pp. 201–222). New York: Cambridge University Press.
Bower, G.H. (1981) Mood and memory. American Psychologist, 36, 129–148.
Bower, G.H., & Forgas, J.P. (2001) Mood and social memory. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Byrne, D., & Clore, G.L. (1970) A reinforcement model of evaluation responses. Personality: An International Journal, 1, 103–128.
Ciarrochi, J.V., & Forgas, J.P. (1999) On being tense yet tolerant: The paradoxical effects of trait anxiety and aversive mood on intergroup judgments. Group Dynamics: Theory, Research and Practice. 3, 227–238.
Ciarrochi, J.V., & Forgas, J.P. (2000) The pleasure of possessions: Affect and consumer judgments. European Journal of Social Psychology, 30, 631–649.
Ciarrochi, J.V., Forgas, J.P., & Mayer, J. (Eds.). (in press). Emotional intelligence: A scientific approach. Philadelphia, PA: Psychology Press.
Clark, M.S., & Isen, A.M. (1982) Towards understanding the relationship between feeling states and social behavior. In A.H. Hastorf & A.M. Isen (Eds.), Cognitive social psychology (pp. 73–108). New York: Elsevier/North-Holland.
Clore, G.L., & Byrne, D. (1974) The reinforcement affect model of attraction. In T.L. Huston (Ed.), Foundations of interpersonal attraction (pp. 143–170). New York: Academic Press.
Clore, G.L., Schwarz, N., & Conway, M. (1994) Affective causes and consequences of social information processing. In R.S. Wyer & T.K. Srull (Eds.), Handbook of social cognition (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
Damasio, A.R. (1994) Descartes’ error. New York: Grosste/Putnam.
Eich, E., & Macauley, D. (2000) Fundamental factors in mood-dependent memory. In J.P. Forgas (Ed.). Feeling and thinking: the role of affect in social cognition (pp. 109–130). New York: Cambridge University Press.
Erber, M.W., & Erber, R. (2001) The role of motivated social cognition in the regulation of affective states. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Feshbach, S., & Singer, R.D. (1957) The effects of fear arousal and suppression of fear upon social perception. Journal of Abnormal and Social Psychology, 55, 283–288.
Festinger, L. (1957) A theory of cognitive dissonance. Palo Alto, CA: Stanford University Press.
Fiedler, K. (1991) On the task, the measures and the mood in research on affect and social cognition. In J.P. Forgas (Ed.), Emotion and social judgments (pp. 83–104). Oxford: Pergamon Press.
Fiedler, K. (2000) Towards an integrative account of affect and cognition phenomena using the BIAS computer algorithm. In J.P. Forgas (Ed.), Feeling and thinking: The role of affect in social cognition. New York: Cambridge University Press.
Fiedler, K. (2001) Affective influences on social information-processing. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Fiedler, K., & Bless, H. (in press). The formation of beliefs in the interface of affective and cognitive processes. In N. Frijda, A. Manstead, & S. Bem (Eds.), The influence of emotions on beliefs. New York: Cambridge University Press.
Fiedler, K., & Forgas, J.P. (Eds.). (1988) Affect, cognition, and social behavior: New evidence and integrative attempts (pp. 44–62). Toronto: Hogrefe.
Forgas, J.P. (1990) Affective influences on individual and group judgments. European Journal of Social Psychology, 20, 441–453.
Forgas, J.P. (1991) Mood effects on partner choice: Role of affect in social decisions. Journal of Personality and Social Psychology, 61, 708–720.
Forgas, J.P. (1992a) Affect in social judgments and decisions: A multi-process model. In M. Zanna (Ed.), Advances in experimental social psychology, Vol. 25 (pp. 227–275). New York: Academic Press.
Forgas, J.P. (1992b) On bad mood and peculiar people: Affect and person typicality in impression formation. Journal of Personality and Social Psychology, 62, 863–875.
Forgas, J.P. (1993) On making sense of odd couples: Mood effects on the perception of mismatched relationships. Personality and Social Psychology Bulletin, 19, 59–71.
Forgas, J.P. (1994) Sad and guilty? Affective influences on the explanation of conflict episodes. Journal of Personality and Social Psychology, 66, 56–68.
Forgas, J.P. (1995a) Mood and judgment: The affect infusion model (AIM). Psychological Bulletin, 117, 39–66.
Forgas, J.P. (1995b) Strange couples: Mood effects on judgments and memory about prototypical and atypical targets. Personality and Social Psychology Bulletin, 21, 747–765.
Forgas, J.P. (1998a) On feeling good and getting your way: Mood effects on negotiation strategies and outcomes. Journal of Personality and Social Psychology, 74, 565–577.
Forgas, J.P. (1998b) Asking nicely? Mood effects on responding to more or less polite requests. Personality and Social Psychology Bulletin, 24, 173–185.
Forgas, J.P. (1998c) Happy and mistaken? Mood effects on the fundamental attribution error. Journal of Personality and Social Psychology, 75, 318–331.
Forgas, J.P. (1999a) On feeling good and being rude: Affective influences on language use and request formulations. Journal of Personality and Social Psychology, 76, 928–939.
Forgas, J.P. (1999b) Feeling and speaking: Mood effects on verbal communication strategies. Personality and Social Psychology Bulletin, 25, 850–863.
Forgas, J.P. (Ed.) (2000a) Feeling and thinking: The role of affect in social cognition. New York: Cambridge University Press.
Forgas, J.P. (2000b) The effects of mood on the accuracy of eyewitness reports of observed social events. Unpublished manuscript, University of New South Wales, Sydney, Australia.
Forgas, J.P. (Ed.). (2001) The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Forgas, J.P., & Bower, G.H. (1987) Mood effects on person perception judgements. Journal of Personality and Social Psychology, 53, 53–60.
Forgas, J.P., Bower, G.H., & Krantz, S. (1984) The influence of mood on perceptions of social interactions. Journal of Experimental Social Psychology, 20, 497–513.
Forgas, J.P., Bower, G.H., & Moylan, S.J. (1990) Praise or blame? Affective influences on attributions for achievement. Journal of Personality and Social Psychology, 59, 809–818.
Forgas, J.P., & Ciarrochi, J.V. (in press). On managing moods: Evidence for the role of homeostatic cognitive strategies in affect regulation. Personality and Social Psychology Bulletin.
Forgas, J.P., Ciarrochi, J.V., & Moylan, S.J. (2000a) Affective influences on the production of persuasive messages. Unpublished manuscript, University of New South Wales, Sydney, Australia.
Forgas, J.P., Ciarrochi, J.V., & Moylan, S. J. (2000b) Subjective experience and mood regulation: The role of information-processing strategies. In H. Bless & J.P. Forgas (Eds.), The message within: The role of subjective experience in social cognition. Philadelphia, PA: Psychology Press.
Forgas, J.P., & Fiedler, K. (1996) Us and them: Mood effects on intergroup discrimination. Journal of Personality and Social Psychology, 70, 36–52.
Forgas, J.P., & George, J.M. (in press). Affective influences on judgment, decision making, and behavior in organizations: An information-processing perspective. In Organizational Behavior and Human Decision Processes (Special Issue Ed., H. Weiss).
Forgas, J.P., & Gunawardene, A. (2000) Affective influences on spontaneous interpersonal behaviors. Unpublished manuscript, University of New South Wales, Sydney, Australia.
Forgas, J.P., & Moylan, S.J. (1987) After the movies: The effects of transient mood states on social judgments. Personality and Social Psychology Bulletin, 13, 478–489.
Gouaux, C. (1971) Induced affective states and interpersonal attraction. Journal of Personality and Social Psychology, 20, 37–43.
Griffitt, W. (1970) Environmental effects on interpersonal behavior: Ambient effective temperature and attraction. Journal of Personality and Social Psychology, 15, 240–244.
Harmon-Jones, E. (2001) The role of affect in cognitive dissonance processes. In J.P. Forgas (Ed.). The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Heider, F. (1958) The psychology of interpersonal relations. New York: John Wiley.
Higgins, E.T. (2001) Promotion and prevention experiences: Relating emotions to non-emotional motivational states. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Isen, A.M. (1984) Towards understanding the role of affect in cognition. In R. S. Wyer & T. K. Srull (Eds.), Handbook of social cognition, Vol. 3 (pp. 179–236). Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
Isen, A.M. (1987) Positive affect, cognitive processes and social behaviour. In L. Berkowitz (Ed.), Advances in experimental social psychology, Vol. 20 (pp. 203–253). New York: Academic Press.
Ito, T., & Cacioppo, J. (2001) Affect and attitudes: A social neuroscience approach. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum and Associates Inc.
McGuire, W.J., & McGuire, C.V. (2001) Dimensions of the social mind: Size, asymmetries, congruence, and sex differences in thought systems focused on self or other persons. In J.P. Forgas, K.R. Williams, & L. Wheeler (Eds.), The social mind: Cognitive and motivational aspects of interpersonal behavior. New York: Cambridge University Press.
Petty, R.E., DeSteno, D., & Rucker, D. (2001) The role of affect in attitude change. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Salovey, P., Detweiler, J.B., Steward, W.T., & Bedell, B.T. (2001) Affect and health-relevant cognition. In J.P. Forgas (Ed.), The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Schwarz, N. (1990) Feelings as information: Informational and motivational functions of affective states. In E.T. Higgins & R. Sorrentino (Eds.), Handbook of motivation and cognition: Foundations of social behaviour, Vol. 2 (pp. 527–561). New York: Guilford Press.
Schwarz, N., & Clore, G.L. (1988) How do I feel about it? The informative function of affective states. In K. Fiedler & J.P. Forgas (Eds.), Affect, cognition, and social behavior (pp. 44–62). Toronto: Hogrefe.
Sedikides, C. (1992) Changes in the valence of self as a function of mood. Review of Personality and Social Psychology, 14, 271–311.
Sedikides, C. (1994) Incongruent effects of sad mood on self-conception valence: It's a matter of time. European Journal of Social Psychology, 24, 161–172.
Sedikides, C. (1995) Central and peripheral self-conceptions are differentially influenced by mood: Tests of the differential sensitivity hypothesis. Journal of Personality and Social Psychology, 69, 759–777.
Trope, Y. Ferguson, M., & Raghunanthan, R. (2001) Mood as a resource in processing self-revalant information. In J.P. Forgas (Ed.). The handbook of affect and social cognition. Mahwah, NJ: Lawrence Erlbaum Associates Inc.
Watson, J.B., & Rayner, R. (1920) Conditioned emotional reactions. Journal of Experimental Psychology, 3, 1–14.
Zajonc, R.B. (1980) Feeling and thinking: Preferences need no inferences. American Psychologist, 35, 151–175.
Zajonc, R.B. (2000) Feeling and thinking: Closing the debate over the independence of affect. In J.P. Forgas (Ed.), Feeling and thinking: The role of affect in social cognition. New York: Cambridge University Press.