The role of desires and anticipated emotions in goal-directed behaviours: Broadening and deepening the theory of planned behaviour
Perugini, Marco; Bagozzi, Richard P. (2001) The British Journal of Social Psychology, 40, 79-98.
Building on the theory of planned behaviour (TPB), we develop a new model of purposive behaviour which suggests that desires are the proximal causes of intentions, and the traditional antecedents in the TPB work through desires. In addition, perceived consequences of goal achievement and goal failure are modelled as anticipated emotions, which also function as determinants of desires. The new model is tested in two studies: an investigation of bodyweight regulation by 108 Italians at the University of Rome and an investigation of effort expended in studying by 122 students at the University of Rome. Frequency and recency of past behaviour are controlled for in tests of hypotheses. The findings show that desires fully mediated the effects of attitudes, subjective norms, perceived behavioural control and anticipated emotions on intentions. Significantly greater amounts of variance are explained in intentions and behaviour by the new model in comparison to the TPB and variants of the TPB that include either anticipated emotions and/or past behaviour.
The origin of action--its efficient, not its final cause--is choice, and that of choice is desire and reasoning with a view to an end.
Aristole (Nicomachean Ethics, 1139a 31-2)
The theory of reasoned action (TRA) has been a mainstay in social psychology for many years (Fishbein & Ajzen, 1975). One advantage of the theory is its parsimony, so it is understandable that changes have occurred infrequently. Perhaps the most important revision of the TRA has been the introduction of perceived behavioural control (PBC) with the theory of planned behaviour (TPB; Ajzen, 1991). The approach taken by Ajzen with this revision was to add an independent variable as a parallel predictor of the dependent variables, along with established predictors. The idea is that more variance can be accounted for by specifying processes formally contained in error terms in tests of the theory. This general approach might be characterized as theory broadening.
A second approach to the revision of any theory is to introduce a variable that explains how existing predictors function to influence intentions. The idea is that certain theoretical mechanisms can be better understood and their effects better qualified by introducing a new construct that mediates or moderates the effects of existing variables. This general approach might be characterized as theory deepening. This study presents a model that expands and deepens the TPB by introducing new constructs which recently have been shown to play important roles in decisionmaking. Figure 1 presents our model, which we term the model of goal-directed behaviour (MGB) for expository purposes.
The MGB posits that desires provide the direct impetus for intentions and transform the motivational content to act embedded in attitudes towards the act (Aact), anticipated emotions (AE), subjective norms (SN) and PBC. Frequency of past behaviour is further assumed to be a predictor of desires, intentions and behaviour, whereas recency of past behaviour predicts behaviour only. As argued below, the introduction of anticipated emotions broadens the TPB by including new decision criteria with respect to a person's goals. The incorporation of desires deepens the TPB by reinterpreting how existing antecedents in the theory function. The inclusion of frequency and recency of past behaviour allows the researcher to incorporate information concerning automatic aspects of goal-directed behaviours not reflected in the variables included in the TPB.
Derivation of the MGB
The TPB seeks to account for actions and maintains that these are a direct function of behavioural intentions (BI) and PBC, and indirect functions, through BI, of Aact, SN and PBC.
A recent meta-analysis examined 142 empirical tests of the TPB and found that the TPB accounted on average for 40 % of the variance in intention and 29 % of the variance in behaviour (Armitage & Conner, in press). Although there is little question that the TPB offers a parsimonious account of purposive behaviour, its sufficiency can be questioned. Researchers recently proposed additions to the TPB that address self-identity processes (Sparks & Shephard, 1992), moral norms (Beck & Ajzen, 1991; Parker, Manstead, & Stradling, 1995), the distinction between perceptions of control and self-efficacy (Armitage &Conner, 1999; Terry & O'Leary, 1995), and anticipated emotions (AEs; e.g. Parker et al., 1995; Richard, van der Pligt, & de Vries, 1995).
We also propose that the TPB can be improved by adding predictors to it. Specifically, we posit that AEs function as important antecedents in decision-making processes. An early consideration of AEs was done by Parker et al. (1995) who found that anticipated regret (measured with `make me feel sorry' and `make me feel good' items) tempered behavioural expectations that one would commit certain automotive driving violations. Likewise, Richard et al. (1995) found that negative AEs (called anticipated post-behavioural affective reactions and measured with `worried-not worried', `regret-no regret' and `tense-relaxed' items) increased expectations that people would refrain from sexual intercourse or would use condoms in casual encounters.
A question that can be raised about attempts to introduce AEs into the TPB is whether the conceptualization of AEs and its measures overlap with Aact and its measures.1 Our theoretical specification and empirical measurement of AEs differ fundamentally from the specification and measurement of both Aact and AEs as proposed heretofore. Consider first Aact. Attitude is conceived as 'a psychological tendency that is expressed by evaluating a particular entity with some degree of favor or disfavor' (Eagly & Chaiken, 1993, p. 1, emphasis in original removed). Under the TRA and TPB, Aact is formulated to refer to a target behaviour and is measured by such bipolar semantic differential items as good-bad, harmful-beneficial, rewardingpunishing, and unpleasant-pleasant (Ajzen & Fishbein, 1980, pp. 261-262). The psychological tendency embodied by an attitude is sometimes termed an `acquired behavioural disposition' (Campbell, 1963). Acquisition of an attitude is thought to be primarily via learning, although biological bases are sometimes acknowledged as well (Eagly & Chaiken, 1993, p. 3).
The nature and role of AEs we introduce differ from that entailed by Aact in attitude theories in three main respects, and concern the referent, underlying process and measurement of attitudes, respectively. First, Aact under the TRA and TPB focuses upon what one does or can do, whereas our specification of AEs focuses not upon action, but rather upon achievement of personal goals. Heretofore, researchers have taken one approach or the other. Under the TRA and TPB, actions are the referents for all antecedents. Indeed, goals are purposefully excluded from the theories (e.g. Ajzen & Fishbein, 1980, pp. 29-30, 111; Fishbein & Stasson, 1990, p. 177). An alternative approach was taken by Bagozzi, Baumgartner, and Pieters (1998) who argued that people, when deliberating to act or not in goal-directed situations, take into account the emotional consequences of both achieving and not achieving a sought after goal (cf. Parker et al., 1995; Richard et al., 1995). In both cases, the processes are believed to be predicated upon a type of thought process analogous to counterfactual thinking, but perhaps better termed `prefactual appraisals' (Gleicher et al. 1995), whereby a decisionmaker imagines the affective consequences of goal attainment and goal failure before deciding to perform instrumental acts. Bagozzi et al. (1998) identified 17 emotions (7 positive, 10 negative: see Method below) that influenced intentions to diet and exercise, in the service of losing or maintaining one's body weight. Unlike Bagozzi et al. (1998), who only considered goal criteria, we also incorporate the TPB and therefore include action referents in our integrative approach.
A second difference between our model and the TPB, beyond divergence in referents, concerns the theoretical processes underlying the effects of Aact vs. AEs. As a disposition to respond in a favourable or unfavourable manner, Aact arises through learning, whereby a person acquires a reaction to an object or action over a period of time or through repeated contact accompanied by reinforcement. An attitude is an evaluative response towards an object or act that, once learned, is triggered automatically when one is exposed to the object or act or thinks about it (Fazio, 1995). By contrast, the processes behind the functioning of AEs are more dynamic and entail self-regulation in response to feedback (Bagozzi, 1992; Carver & Scheier, 1990, 1998). That is, one first has a goal, then appraises the consequences of achieving and not achieving that goal, with corresponding positive and negative emotions arising. An attitude is typically constant over reasonable periods of time and is not formulated as a response contingent on the occurrence of particular happenings to be appraised. The proposed functioning of AEs is, in contrast, specifically contingent on one's appraisal of goal achievement/goal failure, which changes from time to time, depending on the context. An implicit comparison is made between one's goal as a standard or reference value and achieving and failing to achieve that goal, with anticipated emotions as consequences. Attitudes do not function in this manner but are more disposition-like responses to a fixed object or act.
A third distinction we wish to make between Aact and AEs occurs at the level of measurement. When people are asked to respond with their attitudes, they are forced to make a choice of favourability or unfavourability. This is a consequence of the common practice of using bipolar items to indicate respondents' attitudes. By contrast, for the measurement of AEs, we argue that it is important to use unipolar items (e.g. the experience of excitement along a `not at all' to `very much' continuum). In a study of affect, Bagozzi, Wong, and Yi (1999) found that positive and negative affect can be positively or negatively related (or unrelated) to each other, depending on the circumstances. This finding is in line with related evidence for basic differences between positive and negative emotional reactions, in terms of static structural representations of affect (Russell & Barrett, 1999; Watson & Tellegen, 1985), basic functioning of physiological systems (e.g. Davidson, 1992; Gray, 1990), behavioural strategies (Carver & White, 1994; Higgins, 1996), anticipation of future consumption (Lowenstein, 1987), decision-making (van der Pligt, Zeelenberg, van Dijk, de Vries, & Richard, 1998) and effects on behaviour (Cacioppo & Bernston, 1994; Cacioppo, Gardner, & Berntson, 1999). To use a bipolar scale to measure emotions (as is done in attitude theory) would make positive and negative affect mutually exclusive by definition and not permit respondents the opportunity to express their differential relevance. Previous research found that positive and negative anticipated emotions are positively correlated yet clearly differentiated .50; Bagozzi et al., 1998).2
In summary, one type of revision to the TPB proposed herein adds AEs as parallel predictors along with the traditional independent variables found in the TPB. Whereas the TPB specifies that action is the target referent of all independent variables, defines Aact as a disposition-like tendency to respond favourably or unfavourably towards the action, and uses bipolar items to measure Aact, our specification of AEs stipulates that the referents of AEs are personal goals, AEs function as independent variables based upon a decision process that takes into account judged consequences of goal achievement and goal failure, and AEs are measured as unipolar reactions.
A second revision of the TPB is related to the introduction of a variable explaining how existing predictors influence intentions. It has been argued that the TPB fails to consider how intentions become energized (Bagozzi, 1992; Calder & Ross, 1973, p. 7; Fazio, 1995, pp. 271-272). Attitudes, SN and PBC provide reasons for acting, it is claimed, but do not incorporate explicit motivational content needed to induce an intention to act. Similar to Gollwitzer's designation for wants and wishes in his action phases model (e.g. Gollwitzer, Heckhausen, & Steller, 1990, p. 1119), Bagozzi (1992, pp. 184-6) proposed that desires provide the motivational impetus for intentions and suggested that Aact, SN and PBC work through desires enroute to influencing intentions.' This happens in one of two ways. With volitive desires (Davis, 1984a), Aact, SN and PBC provide reasons for acting that a decisionmaker takes into account to form a self-commitment to act. Some philosophers maintain that desires have a particular kind of relationship to intentions in the sense that, once one is aware of and accepts his or her desire to act, this will motivate him or her to form an intention. Davis (1984b, p. 53) calls this the `connection condition' for intentions. With appetitive desires (Davis, 1984a), Aact, SN and PBC serve as catalysts to release or free-up a hidden or latent desire related to such biological needs as food or sex, in contrast to their arousing function for volitive desires (Bagozzi, 1992). Some empirical support for the role of volitive desires in decisionmaking has been found in the contexts of exercising, dieting and studying (Bagozzi & Edwards, 1998; Bagozzi & Kimmel, 1995; Leone, Perugini, & Ercolani, 1999). In the MGB, however, the target behaviour is instrumental to goal achievement. Thus, the specification of desire is relative to performance of a given behaviour (e.g. dieting) because it is conducive to goal attainment (e.g. body-weight regulation). This motivational stance is often referred to in the philosophical literature as the functioning of an extrinsic desire, that is 'a desire for something for its believed conduciveness to something else that one desires' (Mele, 1995, p. 391).
Thus, desires represent the motivational state of mind wherein appraisals and reasons to act are transformed into a motivation to do so. This motivation or desire is hypothesized as the most proximal determinant of intentions in the MGB.
A final shortcoming of the TPB is that past behaviour is not incorporated in the model. Ajzen (1991) criticized the use of past behaviour on grounds that it offers no explanatory content, although he allowed that past behaviour provides a methodological control in tests of any theory: '... past behaviour can be used to test the sufficiency of any model designed to predict future behavior' (p. 202). Furthermore, he argued that the inclusion of PBC in the TPB should preclude the need for past behaviour, in that PBC should mediate any residual effects of past behaviour. However, recent empirical work finds that past behaviour still predicts intentions and/or behaviour in tests of the TPB (e.g. Ajzen & Driver, 1992; Bagozzi & Kimmel, 1995; Beck & Ajzen, 1991; Leone, Perugini, & Ercolani, 1999; Norman & Conner, 1996; Norman & Smith, 1995).
The question remains what the effects of past behaviour might represent. A recent meta-analysis examined 64 studies and found robust evidence for the impact of frequency of past behaviour on both intentions and future behaviours (Ouellette & Wood, 1998). The authors proposed two processes through which frequency of past behaviour guides future behaviour. When a behaviour is well-practised in a constant environment, frequency of past behaviour reflects habit strength and has therefore a direct effect on future behaviour. However, when behaviours are not well-learned or when they are performed in unstable contexts, frequency of past behaviour contributes directly to intentions because `people are likely to form favourable intentions about acts they have frequently performed in the past' (Ouellette & Wood, 1998, p. 56). The latter rationale may also be extended to the direct effects of frequency of past behaviour on desires. In both cases, a reasonable assumption is that not all desires or intentions are explicitly formed consciously or well-formed (e.g. Bagozzi & Yi, 1989; Greenwald & Banaji, 1995). Another proposal has been to partition the effects of past behaviour into frequency and recency effects (e.g. Bagozzi & Warshaw, 1990). Although seemingly related, frequency and recency effects are conceptually distinct and therefore might carry independent information. For instance, one may have a long history of performing a given behaviour without having performed it recently (e.g. a person who long ago gave up purchasing lottery tickets, after experiencing a string of frustrated hopes), or one may have recently taken up an activity with no prior experience with it (e.g. a first-time skier).
Recency of behaviour performance should influence future behaviour to the degree that availability and anchoring/adjustment biases occur in information processing (e.g. Tversky & Kahneman, 1974) and to the degree that an activity, whether established or not, has been recently initiated. The recent initiation of an activity may carry implicit information about intentions over and above the degree to which intentions are accessible to conscious awareness (Bargh & Barndollar, 1996; Greenwald & Banaji, 1995; Verplanken, Aarts, & van Knippenberg, 1997). Recency may serve as an indirect indicator that an intention has been activated and therefore be positively associated with subsequent performance of the behaviour.
To recap, we postulate a new model, the MGB, by proposing revisions to the TPB based upon the addition of new independent variables, the introduction of a mediating variable, and explicit provision for control of past behaviour. We added AEs and past behaviour to the TPB as co-predictors along with Aact, SN and PBC. Desires were hypothesized to provide the motivational impetus channelling the effects of the predictors on intentions. The sufficiency of desires as mediators were tested explicitly by formally comparing the model in Fig. 1 to a model where the above-mentioned antecedents also have direct effects on intentions. The present study tested the MGB in two field settings: losing/maintaining body weight and increasing/maintaining efforts at studying.
Participants and procedure
In Study 1, 108 students (63 women and 45 men, mean age = 22.0, SD = 6.6) at the University of Rome participated in a longitudinal study concerning body weight regulation. In Study 2, 122 students (56 women and 66 men, mean age = 20.5, SD = 3.5) at the University of Rome were asked to participate in an investigation concerning studying. All participants were contacted individually. At Phase 1, participants provided background information, including a statement on their body weight/studying goals. They were asked to write the initials of their names on the questionnaire and were assured of anonymity. Participants in Study 1 were asked: `Which of the following statements best expresses your personal goal over the next four weeks?' The alternatives were: (1) 'I want to decrease my body weight over the next four weeks', (2) 'I want to stay at the same body weight over the next four weeks' (3) 'I want to increase my body weight over the next four weeks'; and (4) 'I do not have a goal with respect to my body weight'. Participants in Study 2 were asked: `Which of the following statements best expresses your personal goal over the next four weeks?' The alternatives were: (1) 'I want to decrease my studying effort over the next four weeks'; (2) 'I want to stay at the same level of effort for studying over the next four weeks; (3) 'I want to increase my studying effort over the next four weeks'; and (4) 'I do not have a goal with respect to my studying effort'. Participants who chose (1) or (2) in Study 1 and (2) or (3) in Study 2 were asked to answer questions worded in accordance with their selected goal for AEs, PBC, Aact, SN, desires, BI and past behaviour. Four weeks later, participants were recontacted and asked to answer a questionnaire containing measures of instrumental behaviours used to achieve their goals.' To minimize the chances for across time self-presentational and consistency biases, participants were not told at Phase 1 that they would be contacted at a later time.
Choice of instrumental behaviour. Whereas in Study 1 two instrumental behaviours (dieting and exercising) were examined for each respondent, participants in Study 2 were asked to choose their own preferred instrumental behaviour. Five behaviours were listed on the questionnaire, and participants were asked to choose one of them or to provide a different one if they so preferred. The five behaviours were selected based on a pilot study (Leone, 1995). The five instrumental behaviours were `stick to a fixed daily schedule', `avoid any type of distractions', `study in a placid place', `refrain from doing anything but studying in the afternoon' and `choose a quiet environment'. Respondents chose the instrumental acts in the following respective proportions: 27.9 %, 19.7 %, 18.0 %, 6.6 % and 11.5 %, respectively. Other behaviours not on the list were chosen by 16.3 % of the participants. With this procedure, we wished to be as certain as possible that the instrumental behaviours were those that participants considered personally relevant to reach their goals. Participants were asked to write their selected instrumental behaviour in a box at the top of each remaining page of the questionnaire. The rest of the questionnaire referred to the chosen instrumental behaviour as `Activity X'.
Anticipated emotions. Anticipated emotions in Study 1 were measured on 11-point items, with response alternatives from `not at all' to `very much'. For the positive emotions, participants were asked to express the felt intensity of each emotion expressed in the subjunctive conditional: `If I succeed to achieve my goal of [decreasing my body weight, staying at the same body weight] over the next four weeks, I will feel [excited, delighted, happy, glad, satisfied, proud, self-assured].' The wording for the negative emotions was: `If I do not succeed to achieve my goal of [decreasing my body weight, staying at the same body weight] over the next four weeks, I will feel [angry, frustrated, guilty, ashamed, sad, disappointed, depressed, worried, uncomfortable, fearful]'. Similar wording was used in Study 2 for `increasing my studying effort [staying at the same level of studying effort]'.
Attitudes. Attitudes were assessed for each behaviour in similar ways in both studies. For example, the dieting item began: 'I think that to keep to a diet in order to decrease my body weight [stay at the same body weight] during the next four weeks is...' Participants responded to 11 semantic differential items on 7-point scales, defined by the pairs useless-useful, ineffective-effective, disadvantageous-advantageous, stupid-intelligent, punishing-rewarding, foolish-wise, unpleasant-pleasant, joyless-joyful, boring-exciting, unattractive-attractive, and unenjoyable-enjoyable.
Subjective norms. Participants were asked to `List the three most important persons for you and indicate how much each of them would approve or disapprove of you doing physical exercise in order to decrease your body weight [stay at the same body weight] during the next four weeks'. Each response was measured on a 7-point scale anchored by 'disapprove' and 'approve'. Similar items were used for dieting and studying. We used these items in order to provide multiple measures but acknowledge that this could introduce error to the extent that they are not highly correlated with the traditional direct way of measuring SN.
Perceived behavioural control. This variable was measured with three items for each instrumental behaviour. For example, the first item for dieting was, 'How much control do you have over keeping to a diet in order to decrease your body weight [stay at the same body weight] during the next four weeks', followed by an 11-point scale going from 'no control' to 'full control'. The second item was, 'For me to keep to a diet during the next four weeks in order to decrease my body weight [stay at the same body weight] is...', followed by an 11-point scale from 'difficult' to 'e2sv'. Finally, the third item asked participants to respond on an 11-point scale from 'very unlikely' to 'very likely' to the following statement: 'If I wanted to, it would be easy for me to keep to a diet during the next four weeks in order to decrease my body weight [stay at the same body weight].' Analogous items were used to measure PBC for exercising. For studying, only the second and third items were used.
Desire. Desires were measured by three items for each instrumental behaviour. For physical exercise, the three items were: (1) 'I desire to do physical exercise in the next four weeks in order to decrease by body weight [stay at the same body weight]', followed by an 11-point scale anchored by 'false' and 'true'; (2) `My desire for doing physical exercise in the next four weeks in order to decrease by body weight [stay at the same body weight] can be described as...', where participants selected one of the following: (a) `no desire', (b) `very weak desire', (c) `weak desire', (d) `moderate desire', (e) `strong desire', and (f) `very strong desire'; and (3) 'I want to do physical exercise in the next four weeks in order to decrease my body weight [stay at the same body weight]', followed by an 11-point scale from 'false' to 'true'. Similar items were employed for dieting. For studying, only the first and third items were used.
Intentions. Intentions were indicated by three items. For example, the item measuring dieting plans was, 'I am planning to keep to a diet in order to decrease my body weight [stay at the same body weight] during the next four weeks', followed by a 7-point scale with 'very unlikely', 'unlikely', 'somewhat unlikely', `neither unlikely nor likely', `somewhat likely', 'likely' and `very likely' as response alternatives. For intentions, the item read, 'I intend to keep to a diet in order to decrease my body weight [stay at the same body weight] during the next four weeks'. Responses were measured on a 7-point scale anchored by `completely disagree' and `completely disagree' and 'completely agree'. Finally, for effort, the item read, 'I will expend effort on dieting to decrease my body weight stay at the same body weight [stay at the same body weight] during the next four weeks', and a 7-point scale going from `completely disagree' to `completely agree' was used to record responses. Parallel items were used to measure intentions to exercise and to study.
Past behaviour frequenct and recency). The measure of frequency was: 'How often did you do physical exercise during the past year to decrease your body weight [stay at the same body weight]?' Participants answered on a 7-point scale with 'never', 'almost never', 'a very few times, 'occasionally', 'often', ,quite often' and 'very many times'. For recency, the item read, 'How often did you do physical exercise during the past four weeks to decrease your body weight [stay at the same body weight] ?', where responses were recorded by a 5-point scale with 'never', 'sometimes', 'regularly', 'often' and 'always' as alternatives. Similar items were used to assess frequency and recency for dieting and for studying.
Behavior. In Phase 2, participants were re-contacted and asked to respond to the following items, depending on the study they were in: 'In order to decrease by body weight [stay at the same body weight], I have done physical exercise during the previous four weeks.; In order to decrease my body weight [stay at the same body weight], I have kept to a diet during the previous four weeks'; and 'In order to increase my studying effort [stay at the same studying effort], I have done [Activity X] during the previous four weeks'. Five-point scales were used in all cases, with 'never', sometimes', 'regularly', 'often' and 'always' as response alternatives.
Specific information concerning the statistical analyses which have been used for the two studies are given in the Results section.
In both studies, Structural Equations Models (SEM) with LISREL 8 were used to test the relevant statistical models (Joreskog & Sorbom, 1993). The goodness-of-fit of the models was evaluated with the chi-square test, the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI) and the Root Mean Square Error of Approximation (RMSEA). Satisfactory fits are obtained when the chi-square test is non-significant, but given the dependence of the chi-square test on sample size, and the need for fit indices normed from 0 to 1, the other indices were examined as well. Good fits are obtained when the CFI and TLI are greater than or equal to .90 and the RMSEA is less than or equal to .08. Further discussion of goodness-of-fit issues can be found in Bentler (1990), Marsh, Balla, and Hau (1996) and Steiger (1990). Note further that the TLI and RMSEA indices reward for model parsimony and penalise for model complexity.
Alternative models were compared with chi-square difference tests for nested models and with the other fit indices for non-nested models. In addition, key parameters were compared and R' values inspected to learn about the predictive power of the alternative models and the sufficiency of desires.
Because Study 1 employed separate measures of two instrumental acts, we examined hypotheses separately for the acts and also for a summary model where exercising and dieting were treated as first-order factors loading on a single second-order factor that can be interpreted as a latent representation of general means to body-weight control. A three-step strategy for analysis was employed beginning with principal components analysis, followed by higher-order confirmatory factor analysis (CFA), and ending with hierarchical regression analysis. The analyses supported our decision to consider the summary model and refer to the specific acts only when statistically significant differences were present. A complete description of the procedures and the findings are available on request from the authors, but is not presented here in the interest of brevity.
Predictions under the proposed MGB were largely supported. Desires, in particular, mediated the effects of the antecedents on intentions in both studies. These findings not only suggest that the TPB is insufficient for explaining intentions, but the results imply that the mechanism behind the effects of the predictors are more complex than hypothesized under the TPB. Desires apparently provide sufficient impetus for intention formation and channel the effects of the antecedents. In this sense, desires constitute important motivators in decision-making. At the same time, desires were found to be functions of AEs and the independent variables specified under the TPB, as qualified below.
By the same token, the MGB accounted for significantly more variance in intentions to diet, exercise and study than the TPB. In this sense, the MGB meets Ajzen's (1991, p. 199) criterion for revising the TPB:
... the theory of planned behavior is, in principle, open to the inclusion of additional predictors if it can be shown that they capture a significant proportion of the variance in intention or behavior after the theory's current variables have been taken into account.
However, in addition to prediction, the MGB refines the explanatory mechanisms underlying intention formation. Desires are hypothesized to provide the direct impetus for intentions and to convey the effects of Aact, SN, AEs and PBC on intentions. It might be argued that the gain in prediction by the MGB comes at the expense of parsimony afforded by the TPB. However, as can be seen in Figs 2 and 3, the explained variance in intentions is owing to desires, and not Aact, SN, AEs and PBC. A model with only desires as antecedents to intentions would be in fact both more parsimonious and predictive than the TPB. Of course, to understand the basis for desires, the MGB provides insight and introduces antecedents based on both action and goal consequences and thus is more explanatory.
Therefore, we believe that the MGB represents a substantial improvement over the TPB, while retaining its key concepts. Should one abandon the TPB? We believe this would be premature for a number of reasons. First, the TPB has been supported by many studies, whereas the MGB is new. More evidence should be produced before one considers abandoning the TPB or accepting the MGB. Second, the MGB was developed to apply to decision-making where behaviours are performed with the manifest purpose to reach a specific goal. Thus, the addition of AEs injects explicit content into the TPB that takes into account the importance or implications of a goal to which an act is contributory. By contrast, the antecedents under the TPB focus on the act as referents and not its specific goal consequences. On the other hand, for target acts or behaviours that are performed as ends in and of themselves (e.g. exercising for pure kinaesthetic pleasure), we would anticipate that the TPB would predict intentions as well as the MGB. One could speculate that the TPB works best when actions are ends, and the MGB provides the biggest improvement when actions are means to end-state goals. More research is needed to verify these speculations and discover other boundary conditions under which the theories apply.
A number of differences between the findings in Studies 1 and 2 should be mentioned. Positive, but not negative, AEs predicted desires to exercise and diet in Study 1 ; but negative, and not positive, AEs predicted desires to study in Study 2. What could account for these differences? Why should desires to exercise and diet be driven by the positive consequences of goal attainment, and studying be driven by the negative consequences of goal failure? One possibility is the different contexts and their meaning for approach and avoidance goals (cf. Carver & White, 1994; Higgins, 1996; Leone, Bagozzi, & Perugini,2000). Consider first, losing and maintaining one's body weight: losing and maintaining weight are concrete goals. Likewise, the connection or linkage between these concrete goals and other concrete or higher level ends (e.g. fit into clothes better, look good, feel good, boost selfesteem, enhance health, promote one's social life) are readily inferred (e.g. Bagozzi & Edwards, 1998). Moreover, losing or maintaining body weight is conceived with relatively high likelihood to lead to other concrete and higher-order ends and to do so in a relatively short period of time, compared to studying (see below). As a consequence, losing and maintaining body weight, at least for people most able to do so (e.g. students), may come to be seen primarily as approachable goals and associated with positive emotions. The negative consequences of not losing or maintaining one's body weight may not be sufficiently strong for students in order to stimulate a desire to engage in goal-directed behaviours. Furthermore, the end states which most likely are associated with the goal concern ideal aspects of the self, especially in an Italian context. According to Higgins (1996; Higgins, Shah, & Friedman, 1997), the presence of an ideal regulatory focus should be associated with an approach strategy and with accentuated importance of positive affect. To increase or maintain one's studying efforts, however, may be less easy to conceive, and in any case the perceived connection between such goals and higher-level ends (e.g. lead to a high paying, challenging career) are weak in Italian society. Furthermore, such consequences are further removed in time than consequences of losing or maintaining one's body weight. Moreover, studying is likely to be viewed, in an Italian context, not as a goal of intrinsic worth, but rather as something which ought to be done. The prevalence of an ought focus should be associated with an avoidance strategy and with an accentuated role played by negative affect such as guilt, shame or anxiety. Further research should focus on deeper investigations of the relation between anticipated emotions, regulatory focuses and behavioural strategies.
An area in need of future research is specification of how self-caused and circumstances-caused factors are taken into account in decisionmaking with respect to goals. We focused herein on the emotional implications of self-caused goal achievement and goal failure as outcomes. However, in addition to these aspects of purposive behaviour, it is important to take into account cases when a plan is progressing well or poorly, when subgoals have been achieved or thwarted, when expectations in the likelihood of goal attainment change (e.g. Oatley, 1992, p. 48), and when personality styles related to confidence, doubt and coping, and happenings beyond one's control, occur that facilitate or inhibit goal attainment.
Finally, the present study leaves open the question of whether other explanatory variables can be introduced between intention and behaviour. To the extent that a particular behaviour constitutes a goal in the mind of a decisionmaker, a number of variables can be suggested as possibilities for future study. For example, implementation intentions, planning, decisions with respect to means and related criteria dealing with the monitoring of progress towards goal attainment and the maintenance of commitment to goal pursuit are processes that might be examined, given that recent research into goal-directed behaviours has found that these variables influence action (e.g. Bagozzi, 1992; Bagozzi & Edwards, 1998, 2000; Gollwitzer & Schaal, 1998; Orbell, Hodgkins, & Sheeran, 1997).
Acknowledgements The authors wish to express their gratitude to the editor and anonymous reviewers for comments made on an earlier draft of this paper.
1. This issue was raised by an anonymous reviewer and the editor, to whom we are grateful. See alsoConner and Armitage (1998).
2 Note that negative antipaticated emotions to failures and postive refer to success in achieving a goal, A positive correlation therefore is what one would normally expect.
For empical and therical evidence about the distinction between desire and intention see, among others
4 In Study 1, 60 % of all respondents (N = 200) had specific body weight goals and were included in the study, for a sample of 120. At Phase 2 we were able to contact 90 % of this sample, thus yielding 108 people for study. Every participant contacted at Phase 2 agreed to participate. In Study 2, 70 % of the sample of 185 initially contacted had a goal for studying. At Phase 2, we were able to contact about 95 % of the sample of 130 people who expressed a goal to increase their effort or keep the same effort for studying, thus yielding 122 people for study. Every participant contacted at Phase 2 agreed to participate.
' Studies 1 and 2 were conducted in Italy and the questionnaire presented in Italian. The Italian versions are available on request.
' A measurement model without this first item of intentions was tested as well. The fit in fact was better (X2(111) = 145.38, p = .02; CFI = .97, TLI = .95, RMSEA = .051), and the loadings of the two remaining items were also higher (.96 and .67) and reliability increased (.81). However, the differences were not dramatic, and the conservative decision was taken to retain the first measure of intentions.
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Process, 50,179-211
Armitage, C.J., & Conner, M. (I9W). Distinguishing perceptions of control from self-efficacy Predicting consumption!of a low fat diet using the theory of planned behaviour. Jofa*ral eJ' Aid Saw Ps, 259, 72-90.
Armitage, C. J., & Conner, M. (in press. Efficacy of the theory of planned behaviour: A mem-analytic review. Bri&b Jolt -of .'j`xial PjyhiiIW.
BagoT RP. (199;',.The sttf rcguhtiaa of attitudes, intentions, and behaviour. Social Pvt y b 55, 178-.
Baga Ft, P., sH., & Piettrs, R. (1998).;Goal-cte emotions. Lo8nitioa mil F.ssgtrwvgmtot - F-mat 12, 1998).
Bagozzi, R. P., & Edwards, E. A. (1998). Goal setting and goal pursuit in the regulation of body weight. Psychology and Health, 13, 593-621.
Bagozzi, R. P., & Edwards, E. A. (2000). Goal-striving and the implementation of goal intentions in the regulation of body weight. Psychology and Health, 15, 255-270.
Bagozzi, R. P., & Heatherton, T. F. (1994). A general approach to representing multifaceted personality constructs: Application to state self-esteem. Structural Equation Modelling, 133, 35-67.
Bagozzi, R. P., & Kimmel, S. K. (1995). A comparison of leading theories for the prediction of goaldirected behaviours. British Journal of Social Psychology, 34, 437-461.
Bagozzi, R. P., & Warshaw, P. R. (1990). Trying to consume. Journal of Consumer Research, 17, 127-140. Bagozzi, R. P., Wong, N., & Yi, Y. (1999). The role of culture and gender in the relationship between positive and negative affect. Cognition and Emotion, 13, 641-672.
Bagozzi, R. P., & Yi, Y. (1989). The degree of intention formation as a moderator of the attitude-behaviour relationship. Social Psychology Quarterly, 52, 266-279.
Bargh, J. A., & Barndollar, K. (1996). Automaticity in action: The unconscious as repository of chronic goals and motives. In P. M. Gollwitzer & J. A. Bargh (Eds.), The psychology of action: Linking cognition and motivation to behaviour (pp. 457-481). New York: Guilford.
Beck, L., & Ajzen, L. (1991). Predicting dishonest actions using the theory of planned behaviour. Journal of Research in Personality, 25, 285-301.
Bentley, P. iMf. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238-246. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Bratman, M. E. (1987). Intentions, plans, and practical reason. Cambridge, MA: Harvard University Press. Cacioppo, J. T., & Berntson, G. G. (1994). Relationship between attitudes and evaluative space: A critical review, with emphasis on the separability of positive and negative substrates. Psychological Bulletin, 115, 401-423.
Cacioppo, J. T., Gardner, Vf. L., & Berntson, G. G. (1999). The affect system has parallel and integrative processing components: Form follows function. Journal of Personality and Social Psychology, 76, 839-855.
Calder, B. J., & Ross, M. (1973). Attitudes and behavior. Morristown, NJ: General Learning Press. Campbell, D. T. (1963). Social attitudes and other acquired behavioral dispositions. In S. Koch (Ed.),
Psychology: A study of a science (Vol. 6, pp. 94-172). New York: McGraw-Hill.
Carver, C. S., & Scheier, M. F. (1990). Origins and functions of positive and negative affect: A controlprocess view. Psychological Review, 97, 19-35.
Carver, C. S., & Scheier, M. F. (1998). On the self-regulation of behavior. Cambridge: Cambridge University Press.
Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS scales. Journal of Personality and Social Psychology, 67, 319-333.
Conner, M., & Armitage, C. J. (1998). Theory of planned behaviour: A review and avenues for future research. Journal of Applied Social Psychology, 28, 1430-1464.
Davidson, R. J. (1992). Emotions and affective style. Hemispheric substrates. Pychological Science, 3, 39-43.
Davis, W. (1984a). The two sense of desire correlates of attitude accessibility. In R. E. Petty & J. A. Krosknik (Eds.), Attitude strength: Antecedents and consequences (pp. 247-282). Mahwah, NJ: Erlbaum.
Davis, W.A. (19984b). A casual theory of intending. American Philosphy Quarterly, 21, 43-54
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behaviour: An introduction to theory and research. Reading, MA: Addison-Wesley.
Fishbein, M., & Stasson, M. (1990). The role of desires, self-predictions, and perceived control in the prediction of training session attendance. Journal of Applied Social Psychology, 20, 173-198.
Gleicher, F., Boninger, D. S., Strathman, A., Armor, D., Hetts, J., & Ahn, M. (1995). With an eye toward the future: The impact of counterfactual thinking on affect, attitudes, and behaviour. In N. J. Roese & M. M. Olson (Eds.), What might have been: The social psychology of counterfactual thinking (pp. 283-304). Mahwah, NJ: Erlbaum.
Gollwitzer, P. M., Heckhausen, H., & Steller, B. (1990). Deliberative and implemental mind-sets: Cognitive tuning toward congruous thoughts and information. Journal of Personality and Social Psychology, 59, 1119-1127.
Gollwitzer, P. M., & Schaal, B. (1998). Metacognition in action: The importance of implementation intentions. Personality and Social Psychology Review, 2, 124-136.
Gray, J. A. (1990). Brain systems that mediate both emotion and cognition. Cognition and Emotion, 4, 269-288.
Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4-27.
Higgins, E. T. (1996). Ideals, oughts, and regulatory focus: Affect and motivation from distinct pains and pleasures. In P. M. Gollwitzer & J. A. Bargh (Eds.), The psychology of action: Linking cognition and motivation to behavior (pp. 91-114). New York: Guilford.
Higgins, E. T., Shah, J., & Friedman, R. (1997). Emotional responses to goal attainment: Strength of regulatory focus as moderator. Journal of Personality and Social Psychology, 72, 515-525.
Joreskog, K. G., & Sorbom, D. (1993). LISREL 8 user's reference guide. Chicago: Scientific Software. Leone, L. (1995). Le odierne teorie sulla relaZione atteggiamento-comportamento: Applica.ione sul comportamento di studio [Current theories on attitude-behaviour relationships: An application on studying . Unpublished manuscript, University of Rome.
Leone, L., Bagozzi, R. P., & Perugini, M. (2000). Anticipated emotions as predictors of intentions and attitudes of engaging in dieting behaviors: The moderator effects of two distinct self-regulatory systems. Unpublished manuscript, Rice University, USA.
Leone, L., Perugini, M., & Ercolani, A. P. (1999). A comparison of three models of attitude-behaviour relationships in studying behaviour domain. European Journal of Social Psychology, 29, 161-189. Loewenstein, G. (1987). Anticipation and the valuation of delayed consumption. The Economic Journal, 97,666-684.
Malle, B. F., & Knobe, J. (1997). The folk concept of intentionality. Journal of Experimental Social Psychology, 33, 101-121.
Marsh, H., Balla, J., & Hau, K. T. (1996). An evaluation of incremental fit indices: A clarification of mathematical and empirical properties. In G. A. Marcoulides & R. E. Schumacker (Eds.), Advanced structural equation modelling: Issues and techniques (pp. 315-353). Mahwah, NJ: Erlbaum.
Mele, A. R. (1992). Springs of action: Understanding intentional behavior. New York: Oxford University Press.
Mele, A. R. (1995). Motivation: Essentially motivation-constituting attitudes. The Philosophical Review, 104, 387-423.
Norman, P., & Conner, M. (1996). Predicting health-check attendance among prior attenders and nonattenders: The role of prior behaviour in the theory of planned behaviour. Journal of Applied Social Psychology, 26, 1010-1026.
Norman, P., & Smith, L. (1995). The theory of planned behaviour and exercise: An investigation into the role of prior behaviour, behavioural intentions and attitude variability. European Journal of Social Psychology, 25, 403-415.
Oatley, K. (1992). Best laid schemes: Psychology of emotions. Cambridge: Cambridge University Press. Orbell, S., Hodgkin, S., & Sheeran, P. (1997). Implementation intentions and the theory of planed behaviour. Personality and Social Psychology Bulletin, 23, 945-954.
Ouellette, J. A., & Wood, W. (1998). Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior. Psychological Bulletin, 124, 54-74.
Parker, D., Manstead, A. S. R., & Stradling, S. G. (1995). Extending the theory of planned behaviour: The role of personal norm. British Journal of Social Psychology, 34, 127-137.
Richard, R., van der Pligt, J., & de Vries, N. (1995). Anticipated affective reactions and prevention of AIDS. British Journal of Social Psychology, 34, 9-21.
Russell, J. A., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes, and other things called emotion: Dissecting the elephant. Journal of Personality and Social Psychology, 76, 805-819. Sparks, P., & Shepherd, R. (1992). Self-identity and the theory of planned behaviour: Assessing the role
of identification with green consumerism. Social Psychology Quarterly, 55, 388-399.
Steiger, J. H. (1990). Structural model evaluation and modification: An interval estimation approach. Multivariate Behavioural Research, 25, 173-180.
Terry, D. J., & O'Leary, J. E. (1995). The theory of planned behaviour: The effects of perceived behavioural control and self-efficacy. British Journal of Social Psychology, 34, 199-220.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185, 1124-1131.
van der Pligt, J., Zeelenberg, M., van Dijk, W. W., de Vries, N. K., & Richard, R. (1998). Affect, attitudes and decisions: Let's be more specific. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 8, pp. 33-66). London: Wiley.
Verplanken, B., Aarts, H., & van Knippenberg, A. (1997). Habit, information acquisition, and the process of making travel mode choices. European Journal of Social Psychology, 27, 539-560.
Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219-235.