Anticipated emotions and personal experience for predicting behavioral intentions and behavioral expectations.
When people imagine experiencing emotions in the future once certain future events have occurred, these "anticipated emotions" (Baumgartner, Pieters, & Bagozzi, 2008) may energize the volitional processes (Bagozzi, et al., 1998) that involve proximal antecedents of behaviors (i.e., behavioral intentions and behavioral expectations). We set out to explore the role of anticipated emotional experience in predictions of the proximal antecedents of two risk behaviors: sex without condom and excessive drinking of alcohol.
The affective component of attitude includes, but is not confined to, emotions (Haddock & zanna, 1999), and people may respond to attitudinal scales considering only semantic relationships, beliefs about emotions which do not always coincide with emotions as current experiences (see Robinson & Clore, 2002). Most events are sufficiently complex to have many implications for the individual; hence, multiple affective responses are common (Caballero et al., 2007; Carrera, Munoz, & Caballero, 2010; Dillard, Plotnick, Godbold, Freimuth, & Edgar, 1996). In the following studies, anticipated emotions will be measured as specific emotional categories (such as joy, sadness or fear) in order to facilitate the anticipation task. In order to avoid individual differences obscuring the role of emotions, we follow two strategies: a) focusing on emotional valence: participants first report their specific anticipated emotions freely within a wide range, and then we categorize their self-reports into three anticipated emotional profiles (AEPs): positive only, negative only and mixed (positive and negative emotions); and b) focusing on emotional intensity: we will calculate a general negative anticipated emotion index averaging all the emotions (including zero reports and reverse-coded joy). The advantage of these two procedures is that first of all people describe their emotional experience in a very detailed way using a wide range of emotional categories, and later (as in an open-ended question) researchers code their answers in several variables, taking into account different affect dimensions (i.e., valence and intensity).
Particular patterns of situational appraisals induce discrete emotional experiences, and these involve specific action tendencies (Frijda, 1986). For the anticipation task we chose four emotions (joy, sadness, fear and anger) proposed by Ekman (1992) as basic emotions or emotion families, and two moral emotions used in previous studies: regret (see Richard, van der Pligt, & de Vries, 1996) and shame, since it is an emotional category close to regret (see Hurtado, Fernandez-Dols, Parrot, & Carrera, 2010). Specific emotions are associated with modes of action readiness and with specific events (such as a specific behavior), so that anticipating them will be easier than anticipating general mood (positive and/or negative), in which actions and antecedents are diffuse (see differences between emotion and mood in Fox, 2008).
Seeking new factors to improve the predictions from TPB (theory-broadening approach) has been a very popular strategy; another widely-used approach has involved the use of different proximal antecedents to behavior (e.g., Rhodes & Matheson, 2005). Behavioral intention is the main proximal antecedent to behavior in TRA and TPB (Ajzen & Fishbein, 2005), and explains 30% of variance in future behavior (see Armitage & Conner, 2001). Gibbons (2006), reviewing recent research, showed that other predictors could be better in some circumstances, such as behavioral willingness (Gibbons et al., 2003) or behavioral expectation (Sheppard, Hartwick, & Warshaw, 1988; Warshaw & Davis, 1984). Behavioral intentions (BI) are plans to engage in or not engage in a behavior, while behavioral expectations (BE) are estimations of one's own likelihood of performing some specific future behavior (see Sheppard et al., 1988). These two measures are not equivalent: "behavioral expectation may act as a perceived proxy measure for future behavior by taking temporal fluctuations in motivation and interactions with volitional control factors into consideration" (Rhodes & Matheson, 2005, p. 64).
Previous studies about differences between proximal antecedents of behavior have suggested that the decision process is different for people with high experience of risk behavior than for people with low experience. In the first group, participants have more knowledge of their motivations and capacities and of situations associated with risk behavior; this increases their awareness of what is likely to happen in the future, and they can anticipate advantages and disadvantages (Beck & Ajzen, 1991; Pomery, Gibbons, Reis-Bergan, & Gerrard, 2009). In this vein, Pomery et al., (2009) found that BI is a better predictor for behavior (skipping lectures) in high-experience students than in low-experience students. The influence of past experience could also be relevant in the role of anticipated emotions in relation to the proximal antecedents of risk behaviors. Davis and Warshaw (1992) defined BE as an estimate or subjective probability that a behavior will actually be performed, considering situational factors and past experience and anticipating changes in intentions or control limitations. Gordon (1990) has pointed out that intention is based on behavioral beliefs but expectations are focused on situational cues and past experiences with behavior. In moderate-low experience samples, situational cues and personal past experiences such as emotions are scarce, so that participants' predictions about BE will not include them. High-experience people can use past experience to fit their anticipated emotions more closely to their predictions about BI and BE.
We test the role of anticipated emotions in the prediction of two proximal antecedents of behavior, behavioral intention and behavioral expectations, and also explore the role of personal experience in those relationships. To achieve these objectives we shall study two behaviors with different incidence in our samples: sexual risk behavior (moderate-high experience) and binge drinking (dividing the sample in two groups: low-experience versus moderate-high experience). We expect anticipated emotions to improve predictions from TPB about BI and BE in moderate-high experience samples (Study 1 and the moderate-high experience sub-sample in Study 2), and that this improvement will be greater for BI than for BE in the low-experience sub-sample (Study 2).
In this first study we tested whether, in a moderately high-experience sample, anticipated emotions improve the predictions from TPB in relation to behavioral expectation and behavioral intention. We chose sex without condom as the risk behavior. In a Spanish survey, it was found that university students did not consider unprotected sex as a very dangerous behavior, and hence practiced it frequently. Although they are sexually active, they use other methods to control pregnancy, and are not worried about sexually-transmitted diseases (STDs). Furthermore, although the epidemiological data from the report for the first semester of 2009 by the Instituto de Salud Carlos III (ISCIII) on AIDS in Spain suggest a stabilization or even a slight fall in the number of cases, Spain continues to be one of the countries most affected by this syndrome in Europe (see, e.g., Bermudez & Teva, 2004; Teva, Bermudez, & Buela-Casal, 2009). As regards other STDs, we can observe an increasing trend in their incidence among the Spanish population in recent years (Centro Nacional de Epidemiologia, ISCIII, 2011). In this vein, some studies indicate that up to 45% of new HIV infections occur in those aged 15 to 24 (ONUSIDA report 2008, in Bermudez, Castro, Madrid, & Buela-Casal, 2010). These data suggest that unprotected sex is a risk behavior whose incidence needs to be reduced.
71 students from the Psychology Department at the Autonoma University in Madrid (Spain) participated in the study, with an average age of 22.35 years (SD = 1.85); 16 were men and 55 were women. Participation was voluntary and anonymous. After signing the consent form, participants completed the questionnaire individually. From this sample we selected a sub-sample with moderate-high experience in the risk behavior, so that the final analysis was carried out on 57 participants (15 men and 42 women).
Material and Procedure
In the questionnaire we asked respondents about their personal past experience in sex without condom, requiring them to answer on a 7-point scale (ranging from never to very frequently). We included two open questions about whether they had a partner at that time, and how long they had been in that relationship. The next questions were based on Theory of Planned Behavior variables, and were measured using unipolar 7-point scales: Attitude was measured via the item "To what extent do you think that sex without condom is ...?", with three positive (good, positive, pleasant) and three negative (bad, negative, unpleasant) items (negative items, and positive items recoded, Cronbach's alpha = .81). To assess the social norm, we asked "To what extent do you think your friends consider that sex without condom is ...?", with two items (positive, negative). When we recoded the positive item, the two showed high correlation (Cronbach's alpha = .76). To assess perceived control, we asked "When you are going to have sex, can you control/decide whether to use a condom?", with a range of responses from not at all to very much so. We included six dichotomous-then-unipolar scales to explore anticipated emotional experience about sex without condom: "If in the next few weeks you were to have sex without condom, how do you think you would feel... angry / sad / ashamed / joyful / regretful / afraid". Respondents were required to first answer yes or no, and in the case of an affirmative answer, to indicate the intensity with which they thought they would feel that emotion, on a 7-point scale (from very low to very intense). We chose four basic emotions (joy, sadness, anger, fear) and two moral emotions (shame and regret). Finally, we asked participants about their "intention" (behavioral intention, BI) and "likelihood" (behavioral expectation, BE) of using a condom if they were to have sex in the next few weeks. We used 7-point scales (ranging from none at all to very strong).
In general, social norm and attitude toward having sexual relations without condom were moderately negative (social norm: M = 4.45, SD = 1.38; attitude: M = 3.67, SD = 1.30), and participants reported high levels of perceived control (M = 6.25, SD = 1.10).
The total sample of participants reported moderate-to-low previous experience in the risk sexual behavior (M = 3.57, SD = 2.69). In order to obtain a sample with higher levels of experience, we removed participants with no experience of sex without condom. This new sample had a size of 57, with a level of experience in the risk behavior above the mid-point of the scale (M = 4.5, SD = 1.9), so that it could be considered moderate-to-high; social norm and attitude were moderately negative (M = 4.28, SD = 1.35 and M = 3.45, SD = 1.26, respectively) and perception of control was high (M = 6.17, SD = 1.08). The relative frequency with which this risk behavior is practiced is explained by the fact that the majority of the sample with experience had a partner at the time of the survey (70.2%). However, the relationships concerned are not long-term ones: 48% of the sample had been in their relationship for less than 1 year, and only 21.5% for more than 3 years; just two participants reported having been in their relationship for 6 years. Although the use of contraceptive methods other than condoms in these sexually active individuals does not provide protection against STDs, they do not perceive how risky their behavior is.
For this sample with personal experience in the risk behavior, the behavioral intention of having sex without condom in the next few weeks was moderate (M = 3.29, SD = 2.25); the reported behavioral expectation was also moderate, though significantly higher (M = 3.94, SD = 2.32; F(1,56) = 9.40, p < .003, [[eta].sup.2] = .14).
As regards anticipated emotional experience, Table 1 shows the number of participants, the mean and the standard deviation for the anticipated emotions reported.
Fear was the anticipated emotion with the highest frequency, and sadness the least often mentioned. We also calculated a general negative anticipated emotion index by averaging intensities (zeros included and reverse coded joy). General intensity was low because many emotions were not anticipated (M = 1.57, SD = 1.44).
We decided to organize this emotional information in more homogeneous patterns, taking into account the emotional valence (positive and/or negative). We recoded anticipated emotional experience in three patterns: anticipated experience of negative emotions only (N = 20) (anger, sadness, shame, regret or/and fear), mixed anticipated emotional experience (N = 19) (positive and negative emotions) and positive only (N = 12) (joy). Six participants who did not anticipate any emotion were omitted from the analysis.
Our next objective was to check whether, in this sample with moderate--high experience in the risk behavior, the anticipated emotional profiles improved the level of explanation provided by the TPB. For this purpose we carried out several hierarchical regression analyses, entering in the first step the attitude, the social norm and the perceived control, and in the second step the three anticipated emotional profiles (as a new categorical variable 1-only negative emotions, 2-mixed emotions and 3-only positive emotion). Likewise, we were interested to see whether such an increase occurred in the two behavioral antecedents assessed: BI and BE.
TPB, anticipated emotions and BI.
We carried out a set of hierarchical regression analyses with standardized scores for controlling factor variability level. In the first step BI was regressed onto personal negative attitude, social norm and perceived control. According to this model, F(3,47) = 7.60, p <.001, the TPB variables explained 28% of intention to perform the risk behavior ([[R.sup.2].sub.ch] = .28), and negative attitude was the most relevant information in that prediction ([beta] = -.54, p < .01). In the second step we added anticipated emotional profiles (six participants did not report emotions so they were removed from this analysis); this second model showed a significant improvement to 45%, [[R.sup.2].sub.ch] = .45, F(1,46) = 15.6, p < .001, and anticipated emotional profile and attitude were significant predictors ([beta] = .44 p < .001 and [beta] = -.40, p < .001, respectively). Anticipated emotional profiles (valence focus, only negative as 1, mixed as 2 and only positive as 3) improved the prediction from TPB variables in relation to BI. Adding new independent variables as parallel predictors of the dependent variables, however, presents some problems (see Perugini & Bagozzi, 2001, Trafimow, 2004). Trafimow (2004) explains that because many predictors are correlated with one another, increasing the variance in the intention or behavior that can be accounted for by the list of independent variables is not enough. To solve this problem, Trafimow (2004) has proposed some statistical and methodological strategies to convert [DELTA][R.sup.2] into probability of successful prediction. Following this proposal, we used binomial effect size display (see Rosenthal & Rosnow, 1991) in order to assess the actual meaning of the change in [DELTA][R.sup.2]. A change from 28% to 45% implies an increase in probability of successful prediction about BI of 6%, which is a moderate value.
Focusing on emotional intensity we also repeated regression analysis with negative anticipated emotion index standardized (now 57 participants were included in the analysis because non-reported emotions were included as zero value). Results were similar: adding negative emotional experience showed a significant improvement from 24% to 32%, [[R.sup.2].sub.ch] = .32, F(1,52) = 7.61, p < .01, and negative anticipated emotion index and attitude were significant predictors ([beta] = -.33 p < .01 and [beta] = -.37, p < .01, respectively). A change from 24% to 32% implies an increase in probability of successful prediction about BI of 4%, which is a moderate-low value.
TPB, anticipated emotions and BE.
We repeated this analysis for BE. In the first step, BE was regressed onto personal negative attitude, social norm and perceived control, and this model was significant, F(3,47) = 5.05, p < .01, the TPB variables explaining 19% of behavioral expectation to perform the risk behavior ([[R.sup.2].sub.ch] = .19), and again negative attitude was the most relevant information in that prediction ([beta] = -.44, p < .01), followed by perceived control ([beta] = -.26, p < .05). In the second step, adding anticipated emotional profiles, we found a significant improvement to 39%, [[R.sup.2].sub.ch] = .40, F(1,46) = 16.9, p < .001, in which anticipated emotional profile, attitude and perceived control were significant ([beta] = .48 p < .001, [beta] = -.29 p < .01 and [beta] = -.22, p < .05, respectively). Once again, anticipated emotional profiles improved the prediction from TPB variables in relation to BE. A change from 19% to 40% implies an increase in probability of successful prediction about BE of 8%, a moderate increase. These results show that anticipated emotional profiles constitute important information for improving predictions from the TPB model about BI and BE: the more positive emotions are anticipated, the higher person's BI and BE.
Regression analysis with negative anticipated emotion index standardized showed a significant improvement to 40%, [[R.sup.2].sub.ch] = .40, F(1,52) = 22.7, p < .001, and this negative anticipated emotion index was significant ([beta] = -.54 p < .001) and attitude and perceived control were significant as a trend ([beta] = -.19, p < .09 and [beta] = -.18, p < .08, respectively). A change from 16% to 40% implies an increase in probability of successful prediction about BE of 11%, which is a moderate value.
These results have shown us that in samples with moderate-high levels of experience in the risk behavior, the predictions made from the TPB about the proximal antecedents of risk behavior (BI and BE) can be improved by adding anticipated emotional experience. Anticipated emotions provide us with richer and more detailed information than that which is provided by the TPB variables. Table 2 shows a summary of the regression analysis.
In the first study we saw how, using a sample with moderate-high experience in a risk behavior, anticipated emotional profiles and negative anticipated emotion index improved the predictions from TPB in relation to behavioral antecedents. In this second study we were interested to know whether anticipated emotions would improve the prediction from TPB in a sample in which we could distinguish different levels of experience (high and low) in the risk behavior (excessive drinking) and to check whether the results found in Study 1 with a sample of moderate-high experience in the risk behavior would be replicated in the case of a different behavior. With less experience in the risk behavior we expected anticipated emotions to correlate significantly with behavioral intention (BI), but not (or to a lesser extent) with behavioral expectations (BE). This hypothesis is based on the fact that behavioral expectations depend to a larger extent than intentions on situational factors and past behavior, information which the sample with low personal experience have less knowledge of, so that they cannot use it to fit expectations and anticipated emotions. We expected in the low-experience condition that anticipated emotions would have higher predictive power in relation to behavioral intentions than in relation to behavioral expectations. In the sample with moderate-high personal experience in the behavior we expected, as in Study 1, a good fit between anticipated emotions and the two types of behavioral antecedents (i.e., BI and BE).
In this study we choose the behavior of occasional excessive drinking as risk behavior in a sample of university students. In Spain, this type of behavior is frequent (Becona, 2000; Caballero et al., 2004), and gives cause for concern, especially because it is combined with the driving of automobiles and the practice of unprotected sex. Alcohol is the psychoactive substance most widely consumed by young Spaniards, this consumption being concentrated basically on weekends and at parties or gatherings of friends. The most recent research indicates that this abusive pattern of alcohol use has been on the increase in recent years (Ingles et al., 2007).
One hundred and forty-five students at the Psychology Department of the Autonoma University of Madrid participated in this second study, with an average age of 21.20 years (SD = 3.21). Participation was voluntary and anonymous. After signing the consent form, participants completed the questionnaire individually. We divided this sample in low experience (N = 61, 52 women and 9 men) versus moderate-high experience (N = 84, 67 women and 17 men).
Material and Procedure
In the questionnaire we asked them about their attitude towards drinking excessive alcohol, without specifying a quantity or quality, so that each individual could apply their own criteria as to what they considered excessive. Previous research (Caballero et al., 2007) had shown that a free personal categorization of excess was useful to increase the sample, because in this way we can include those who drink moderately as compared with general population, but excessively according to their personal standard (i.e., taking into account weight, gender, habit, subjective norm, and so on). The first question was about their personal frequency of drinking excessively, with a 7-point scale (ranging from never to very frequently). We used this item to split the sample into low versus moderate-high experience. The next items were about Theory of Planned Behavior variables: Attitude was measured with six unipolar 7-point scales, "To what extent do you think that drinking excessive alcohol is.?", with three positive (good, positive, pleasant) and three negative (bad, negative, unpleasant) items (negative items, and positive items recoded, Cronbach's alpha = .86). To evaluate subjective norm we asked "To what extent do you think that your friends consider that drinking excessive alcohol is... ?", with two items (positive, negative); when we recoded the positive item the two showed high correlation (Cronbach's alpha = .84). To assess perceived control we averaged two items "Can you control your excessive drinking?" and "Can you stop drinking alcohol whenever you want?" (Cronbach's alpha = .80).
We included six dichotomous-then-unipolar scales to explore anticipated emotional experience about excessive drinking: "If in the next few weeks you were to drink alcohol excessively, how do you think you would feel. angry / sad / ashamed / joyful / regretful / afraid". Respondents were required to first answer yes or no, and in the case of an affirmative answer, to indicate the intensity with which they thought they would feel that emotion, on a 7-point scale (from very low to very intense). Finally, we asked about their "intention" (behavioral intention, BI) and "likelihood" (behavioral expectation, BE) of drinking excessive alcohol in the next few weeks. We used 7-point scales (ranging from none at all to very strong).
Results and Discussion
As in Study 1, no emotional category was reported by the whole sample. Focusing on emotional valence, we recoded anticipated emotions in three profiles in either experience sub-sample (a new categorical variable): negative emotions only (1), mixed emotions (2) and positive emotions only (3). We also calculated negative anticipated emotion index, averaging negative emotions and the reverse-coded positive one (M = 1.54, SD = 1.84).
First of all we carried out a hierarchical multiple regression analysis to examine the hypothesis that personal experience in risk behavior moderates the relations between anticipated emotions and BI and BE (all measures were standardized). Personal experience in drinking excessive alcohol and AEPs was entered in Step 1, followed by their interaction in Step 2 (Experience x AEPs). As might be expected, interaction in relation to BI was not significant ([beta] = .04 p = .55) but interaction in relation to BE was significant ([beta] = .15 p < .05). This result means that anticipated emotional profiles are related to level of personal experience in the prediction of behavioral expectation, but this is not the case for behavioral intention, where anticipated emotional profiles predict at any level of experience. As Figure 1 shows, the greater the personal experience and the more positive the anticipated emotional profiles, the higher the person's BE to engage in binge drinking.
When we used the general negative emotion index we did not find this difference: the interaction was significant for both BI and BE. In the following analyses we attempt to clarify the complex relationship between emotional forecast and personal experience in relation to the prediction of BI and BE.
[FIGURE 1 OMITTED]
In order to replicate the results obtained in Study 1 and evaluate the role played by experience and anticipated emotion, we decided to compare the moderate-high experience group (as Study1) with a low-experience sample in risk behavior. Therefore, we selected a sub-sample who had never drunk alcohol to excess or who had done so only occasionally (lower than 3 on a 7-point scale ranging from never to very frequently). The result of this selection process was a sample of 61 participants (52 women and 9 men) with low experience in this risk behavior (M = 1.73, SD = .44). The moderate-high experience sample included 84 participants (67 women and 17 men) with higher experience than the first group, (M = 4.07, SD = 1.08), F(1,143) = 251.6, p < .001, [[eta].sup.2] = .63.
As Table 3 shows, attitude, perceived control, BI and BE are different for the two sub-samples. In the low-experience group, participants have a more negative attitude and higher perceived control, and their level of intention and expectations are lower than in the moderate-high experience group. Both samples think that their friends have a more positive attitude toward the risk behavior than them, F(1,60) = 78.1, p < .001, [[eta].sup.2] = .56; F(1,82) = 70.6, p < .001, [[eta].sup.2] = .46, low and moderate-high experience respectively. We also carried out a within-test on BI and BE in either group, finding BE to be higher than BI in both groups, F(1,60) = 16.25, p < .001, [[eta].sup.2] = .21; F(1,83) = 11.39, p < .001, [[eta].sup.2] = .12, low and moderate-high experience, respectively.
Table 4 shows emotional intensities and frequencies in specific emotional categories in both experience groups.
We found 8 participants in the low-experience subsample and 14 participants in the moderate-high experience who did not report any anticipated emotion. General negative anticipated emotion index for each sub-sample shows that intensity was higher in low-experience (M = 2.07, SD = 2.12) than in moderate-high experience (M = 1.16, SD = 1.04), F(1,143) = 8.9, p < .01, [[eta].sup.2] = .06.
As in Study 1, our main objective was to check the extent to which the inclusion of the anticipated emotional profiles and general negative anticipated emotion index improved the predictions made from TPB about the two types of behavioral antecedents (BI and BE). To this end we carried out several hierarchical regression analyses, distinguishing this time between the samples with low and moderate-high experience in the risk behavior. In the sample with the higher level of experience we expected to replicate the results of Study 1, given that the sample in that study had a level of experience similar to that of the moderate-high sub-sample in Study 2, F(1,139) = 2.77, p > .05. However, as the moderation analysis had shown, we expected that the improvement through the addition of anticipated emotions to TPB would be lower in the case of BE than in the case of BI.
TPB, anticipated emotions and BI, BE, in low-experience sub-sample
We carried out a set of hierarchical regression analyses with standardized scores. In the low-experience sub-sample, BI was regressed onto personal negative attitude, social norm and perceived control. In this model, F(3,49) = 3.37, p <.05, the TPB variables explained 12% of intention to perform the risk behavior ([[R.sup.2].sub.ch] = .12), and negative attitude was the most relevant information in that prediction ([beta] = -.31, p < .05). In the second step we added anticipated emotional profiles (as a new categorical variable 1-only negative emotions, 2-mixed emotions and 3-only positive emotion); this second model showed a significant improvement to 34%, [[R.sup.2].sub.ch] = .34, F(1,48) = 18.23, p < .001, where anticipated emotional profile was a significant predictor ([beta] = .56 p < .001). Thus, anticipated emotional profiles improved the prediction from TPB variables in relation to BI in the low-experience sub-sample. Following Trafimow (2004), a change from 12% to 34% implies a rise in probability of successful prediction about BI of 11%, which is a moderate increase. These results were replicated (as a trend) using the negative anticipated emotion index, which improved predictions on BI from 9% to 13%, F(1,56) = 3.68, p = .06, a low increase in probability of success of 3.5%, where the negative anticipated emotion index was the main predictor ([beta] = -.25 p = .06).
On repeating this analysis in relation to BE, we did not find that anticipated emotional profiles improved prediction from TPB variables. The first model showed a low but significant explanation of BE, [[R.sup.2].sub.ch] = .11, F(3,49) = 3.34, p < .05, where attitude was the most relevant predictor ([beta] = -.27, p < .05). But in the second model the explanation of BE was not significant, [[R.sup.2].sub.ch] = .14, F(1,48) = 2.56, p = .11. Using general negative emotion index we obtained the same non-significant improvement, F(1,56) = .51, p = .47).
When participants do not have any experience, or only low-level experience, of risky behavior, anticipated emotions do not show a good fit to BE, but their fit to BI is indeed good. Anticipated emotions and BI share assumptions or imagined scenarios, so that they show the same tendency. BE includes contextual cues, which are not considered by low-experience people when they anticipate emotions.
TPB, anticipated emotions and BI, BE, in moderate-high experience sub-sample
We carried out a new set of hierarchical regression analyses in the moderate-high experience group in relation to BI and BE. The first analysis showed that in relation to BI, attitude and perceived control were significant in the first model ([beta] = -.32, and [beta] = -.24, p < .05, respectively), the TPB variables explaining 23% of BI, [[R.sup.2].sub.ch] = .23, F(3,66) = 8.05, p < .001. Anticipated emotional profiles improved this model, [[R.sup.2].sub.ch] = .36, F(1,65) = 13.81, p < .001, in which perceived control ([beta] = -.22, p < .05) and emotional profiles ([beta] = .40, p < .001) were significant. A change from 23% to 36% implies an increase in probability of successful prediction about BI of 5.5%, which is a moderate-low but significant increase. Focusing on level of emotional intensity, we obtained an increase from 25% to 31%, F(1,79) = 7.27, p < .01, where attitude ([beta] = -.36, p < .01) and negative emotion index ([beta] = -.25 p < .01) were significant; this represents 3.5% of improvement in probability of successful prediction about BI.
We found a similar improvement in relation to BE. The TPB model predicted 17% of BE, [[R.sup.2].sub.ch] = .17, F(3,66) = 5.86, p < .01, and attitude was relevant ([beta] = -.35, p <.01), as well as perceived control at the level of a trend ([beta] = -.22, p = .06). The second model, in which emotional profiles were added, improved the prediction to 31%, [[R.sup.2].sub.ch] = .31, F(1,65) = 14.1, p < .001, where perceived control at the level of a trend ([beta] = -.20, p = .06) and emotional profiles ([beta] = .42, p < .001) were significant. A change from 17% to 31% implies an increase in probability of successful prediction about BE of 6.5%, which is a moderate but significant increase. When negative emotion index was added to TPB we again found a significant improvement, from 18% to 24%, F(1,79) = 7.36, p < .01, an increase in probability of successful prediction about BE of 3%, where attitude and negative emotional experience were the main predictors ([beta] = -.29, p < .001 and [beta] = -.27, p < .001, respectively). When participants have moderate-high personal experience in the risk behavior, anticipated emotions, as emotional profiles or as negative emotion index, improve predictions from the TPB model, in relation to BI and to BE. A summary of all these regression analyses can be found in Table 5.
In the two studies presented here our principal aim was to assess whether anticipated emotions, as emotional profiles or as general negative emotion index, could improve the prediction made from TPB in relation to proximal antecedents of behavior such as behavioral intentions and behavioral expectations. Our area of research was focused on risk behaviors, which are difficult to predict given that they are usually associated with both positive and negative consequences. The high attitudinal ambivalence in this type of behavior explains the low percentages of prediction from deliberative models such as those of TPB (Conner & Sparks, 2005; Conner, Sparks, Povey, James, & Shepherd, 1998; Cook & Sheeran, 2004). The choice of anticipated emotions for improving predictions from the TPB is not a novel one, previous research having shown the efficiency of this approach with simple emotions such as regret (Abraham & Sheeran, 2004; Zeelenberg, 1999). However, what we set out to do was enlarge this proposal, using an extended list of anticipated emotions, with four basic emotions (joy, sadness, anger, fear) and two moral emotions (shame and regret). We know from previous studies that risk behaviors, being complex and with multiple implications, are associated with a large set of emotional experiences (see, Caballero et al., 2007; Carrera, et al., 2010). The alternative of measuring specific emotions involved a high percentage of missing values (see Tables 1 and 4). To solve this problem we did two things: first we re-categorized specific emotions into three patterns (AEPs): positive only, mixed, and negative only; and second, focusing on emotional intensity, we calculated a general negative emotion index. These new measures were made post-hoc, after the participants had freely specified their anticipated emotions. We tested whether both recoded variables (i.e., AEPs and negative emotion index) were useful for solving problems related to high percentages of non-reported emotions.
The results of the two studies show that anticipated emotions significantly improve prediction from the TPB in relation to one of the most widely accepted predictors of behavior: behavioral intention. In two risk behaviors relevant for young people -sex without condom and excessive drinking -, the anticipation of emotions showed a good fit with the prediction of behavioral intention. The plans participants made about these behaviors correlated significantly with the patterns of anticipated emotions, the intention being greater when the anticipated emotions were positive only, and smaller when the emotions were negative only; it remained at an intermediate point in cases in which both types of emotion were anticipated. In other words, the higher the level of negative anticipated emotional experience indicated by the emotion index, the lower the intention to perform the risk behavior. Therefore, it is the anticipated emotional information that is the key to the increased level of explanation of behavioral intention. Anticipated emotional experience is coherent with intention and knowing the valence and intensity of future- oriented emotions helps to improve predictions. Emotions once again improve predictions made from attitudes, showing the need to improve deliberative models with emotional information (see Abraham & Sheeran, 2004; Caballero et al., 2007; Parker, Manstead, & Stradling, 1995; Richard, van der Pligt, & de Vries, 1995). This result also highlights the difference between the affective component of attitudes and specific emotions, even when these are affective forecasts and not current emotions (on the differences between affect and cognition, see Forgas, 2001; Giner-Sorolla, 1999; Haddock & Zanna, 1999; Robinson & Clore, 2002).
Prior to our study, research in the field of the prediction of behaviors had shown the need to distinguish behavioral intentions from behavioral expectations (Warshaw & Davis, 1984). Both are proximal antecedents of behavior, but they involve different aspects with important consequences for prediction. In the introduction we pointed out that the difference between these two behavioral antecedents becomes especially relevant for studying people who have different levels of experience. Past experience is crucial in the prediction of behavioral expectations, given that BEs involve personal experience both situational and psychological, whilst intention refers to beliefs and plans that are not necessarily based on personal experience (Gordon, 1990). This difference was replicated in our results as the interaction between anticipated emotions and personal experience for predicting behavioral expectations but not behavioral intentions. In the sub-sample with low personal experience in the risk behavior, anticipated emotions improved the TPB prediction for BI but not for BE. Low-experience people are not familiar with the circumstances and emotions associated with risk behavior, so that they cannot include them in their expectations.
These results lead us to highlight the relevance of including anticipated emotional experience in the prediction of the two behavioral antecedents proposed by Warshaw and Davis (1984). The greater the experience, and hence the more the situational and psychological information (which include the emotions felt in the past on performing the risk behavior), the better the fit of BI and BE with the anticipated emotions. But in low-experience groups, predictions about behavioral expectations are more difficult and poorly defined, so that their fit with anticipated emotions is lower.
These two studies represent a good first approach to the role that anticipated emotions can play in the prediction of proximal behavioral antecedents such as BI and BE. Even so, they clearly have some limitations, the prediction of the behavior itself probably being one of the most relevant. Likewise, it would be necessary to explore a wider range of emotional categories and of risk behaviors, and to include better measures of past experience as objective measures of risk behavior (e.g., actual alcohol consumption) and more items for measuring each predictor. Moreover, further research is on the question of how other influential factors, such as peer pressure or behavioral difficulty, interact with anticipated emotions in the prediction of BI and BE. These studies are a first step to show how anticipated emotions can help us to predict BI and BE. Future experimental research must test how anticipated emotions change BI, BE and actual behavior.
In spite of these limitations, we believe that these two studies reveal the need to take into account anticipated emotions, AEPs and general anticipated emotion index for improving predictions in relation to BI and BE. Likewise, they confirm the findings of previous research (Pomery, et al., 2009; Rhodes & Matheson, 2005; Sheppard, et al., 1988; Warshaw & Davis, 1984), which stresses the need to use different behavioral antecedents, such as behavioral intentions and expectations, taking into account the different levels of experience reported by participants.
We thank David Weston and William Clifton for their help in the preparation of the English version. This research was supported by grant PSI2008-04849
Abraham, Ch., & Sheeran, P. (2004). Deciding to exercise: The role of anticipated regret. British journal of Health Psychology, 9, 269-278. doi:10.1348/135910704773891096
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211. doi:10.1016/0749-5978(91 )90020-T
Ajzen, I., & Fishbein, M. (2005). The influence of attitudes on behavior. In D. Albarracin, B.T. Johnson, & M. P. Zanna (Eds.), The Handbook of Attitudes (pp. 173-221). New Jersey, NJ: Lawrence Erlbaum Associates.
Armitage, C. J., & Conner, M. (2001). Efficacy of the theory of planned behavior: A meta-analytic review. British Journal of Social Psychology, 40, 471-499. doi:10.1348/014466601164939
Bagozzi, R. P., Baumgartner, H., & Pieters, R. (1998). Goal-directed emotions. Cognition and Emotion, 12, 1-26. doi:10.1080/ 026999398379754
Bargh, J. A., & Ferguson, M. J. (2000). Beyond behaviorism: on the automaticity of higher mental processes. Psychology Bulletin, 126, 925-945. doi:10.1037/0033-2909.126.6.925
Baumgartner, H., Pieters, R., & Bagozzi, R. P. (2008). Future-oriented emotions: conceptualization and behavioral effects. European Journal of Social Psychology, 38, 685-696. doi:10.1002/ejsp.467
Beck, L., & Ajzen, I. (1991). Predicting dishonest actions using the theory of planned behavior. Journal of Research in Personality, 25, 285-301. doi:10.1016/0092-6566(91)90021-H
Becona, E. (2000). Los adolescentes y el consumo de drogas [Teens and drugs]. Papeles del Psicologo, 77, 25-32.
Bermudez, M. P., Castro, A., Madrid, J., & Buela Casal, G. (2010). Analisis de la conducta sexual de adolescentes autoctonos e inmigrantes latinoamericanos en Espana [Analysis of sexual behavior of indigenous and Latin American immigrants adolescents in Spain]. International Journal of Clinical and Health Psychology, 10(1), 89-103.
Bermudez, M. P., & Teva, I. (2004). Situacion actual del sida en Espana e inmigracion: analisis epidemiologico [Current situation of AIDS in Spain and immigration: an epidemiological analysis]. Universitas Psychologica, 8, 37-48.
Brehm, J. W. (1999). The intensity of emotion. Personality and Social Psychology Review, 3, 2-22. doi:10.1207/s15327957pspr0301_1
Caballero, A., Carrera, P., Munoz, D., & Sanchez, F. (2007). Emotional ambivalence in risk behaviors: The case of occasional excessive use of alcohol. The Spanish Journal of Psychology, 10, 151-158.
Caballero, A., Carrera, P., Sanchez, F., Shih, P. C., Munoz, D., Toro, I.,... Munoz, A. (2004). Conductas de riesgo en adolescentes [Risk behaviors in adolescents.] Madrid, Spain: Concejalia de Juventud Ayuntamiento de Tres Cantos.
Carrera, P., Munoz, D., & Caballero, A. (2010). Mixed emotional appeals in emotional and danger control processes. Health Communication, 25, 726-736. doi:10.1080/10410236.2010.521914
Centro Nacional de Epidemiologia, ISCIII (2011). Vigilancia epidemiologica de las infecciones de transmision sexual 19952009 [Epidemiological surveillance of sexually transmitted infections 1995-2009]. Madrid, Spain: Centro Nacional de Epidemiologia.
Conner, M. T., & Sparks, P. (2005). Ambivalence and attitudes. In W. Stroebe & M. Hewstone (Eds). European Review of Social Psychology, (Vol 12, pp. 37-70). Chichester, UK: John Wiley & Dons, Ltd. doi: 10.1002/0470013478.ch2
Conner, M. T., Sparks, P., Povey, R., James, R., & Shepherd, R. (1998). Interpretations of healthy and unhealthy eating, and implications for dietary change, Health Education Research, 13(2), 171-183. doi:10.1093/her/13.2.171
Cook, R., & Sheeran, P. (2004). Moderation of cognition-intention and cognition-behavior relations: A meta-analysis of properties of variables from the theory of planned behavior. British Journal of Social Psychology. 43(2), 159-186. doi:10.1348/ 0144666041501688
Davis, F. D., & Warshaw, P. R. (1992). What do intention scales measure? The Journal of General Psychology, 119(4), 391-407.
Dillard, J. P., Plotnick, C. A., Godbold, L. C., Freimuth, V. S., & Edgar, T. (1996). The multiple affective outcomes of AIDS PSAs: Fear appeals do more than scare people. Communication Research, 23, 44-72. doi:10.1177/009365096023001002
Ekman, P. (1992). Are there basic emotions? Psychological Review, 99, 550-553. doi:10.1037/0033-295X.99.3.550
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley.
Forgas, J. P. (2001). Feeling and thinking. The role of affect in social cognition. New York, NY: Cambridge University Press.
Fox, E. (2008). Emotion Science. Basingstoke, NY: Palgrave MacMillan.
Frijda, N. (1986). The emotions. Cambridge, UK: Cambridge University Press.
Gibbons, F. X., Gerrard, M., & Lane, D. J. (2003). A social-reaction model of adolescent health risk. In J. J. Suls & K. A. Wallston (Eds.), Social Psychological Foundations of Health and Illness. Oxford, UK: Blackwell. doi:10.1002/9780470753552.ch5
Gibbons, F. X. (2006). Behavioral intentions, expectations, and willingness. In M. Gerrard & K. D. McCaul (Eds.), Constructs and measures web resource. National Cancer Institute Internet Web Site. Retrieved from: http//cancercontrol.cancer.gov/brp/ constructs/intent-expect-willingness/index.html.
Giner-Sorolla, R. (1999). Affect in attitude. Immediate and deliberative perspectives. In S. Chaiken & Y. Trope (Eds.), Dual process theories in social psychology (pp. 441-461). New York, NY: Guilford.
Gordon, R. A. (1990). Informational bases of behavioral intentions and behavioral expectations or self-predictions. Basic and Applied Social Psychology, 11(4), 433-442. doi:10.1207/ s15324834basp1104_6
Haddock, G., & Zanna, M. (1999). Cognition, affect, and the prediction of social attitudes. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology, (Vol 10, pp. 75-99). Chichester, UK: John Wiley & Dons, Ltd. doi:10.1080/ 14792779943000026
Hurtado, A., Fernandez-Dols, J. M., Parrot, G., & Carrera, P. (2010). Emotion terms, category structure, and the problem of translation: the case of shame and verguenza. Cognition & Emotion, 24, 661-680. doi:10.1080/02699930902958255
Ingles, C. J., Delgado, B., Bautista, R., Torregrosa, M. S., Espada, J. P., Garcia-Fernandez, J. M.,. Garcia-Lopez, L. J. (2007). Factores psicosociales relacionados con el consumo de alcohol y tabaco en adolescentes espanoles. [Psychosocial factors related with alcohol and tobacco use in Spanish adolescents.] International Journal of Clinical and Health Psychology, 7(2), 403-420.
Parker, D., Manstead, A. S. R., & Stradling, S. G. (1995). Extending the theory of planned behavior: The role of personal norm. British Journal of Social Psychology, 34, 127-137.
Perugini, M., & Bagozzi, R. P. (2001). The role of desires and anticipated emotions in goal-directed behaviors: Broadening and deepening the theory of planned behavior. British Journal of Social Psychology, 40, 79-98. doi:10.1348/014466601164704
Pomery, E. A., Gibbons, F. X., Reis-Bergan, M., & Gerrard, M. (2009). From willingness to intention: experience moderates the shift from reactive to reasoned behavior. Personality and Social Psychology Bulletin, 35(7), 894-908. doi:10.1177/ 0146167209335166
Rhodes, R. E., & Matheson, H. (2005). Discrepancies in exercise intention and expectation: theoretical and applied issues. Psychology and Health, 20, 63-78. doi:10.1080/088704404 12331296071
Richard, R., van der Pligt, J., & de Vries, N. (1995). Anticipated affective reactions and prevention of AIDS. British Journal of Social Psychology, 34, 1411-1428.
Richard, R., van der Pligt, J., & de Vries, N. (1996). Anticipated affect and behavioral choice. Basic and Applied Social Psychology, 18, 111-129. doi:10.1207/s15324834basp1802_1
Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for an accessibility model of emotional self-report. Psychological Bulletin, 128, 934-960. doi:10.1037/0033 2909.128.6.934
Rosenthal, R., & Rosnow, R. (1991). Essentials of behavioral research: Methods and data analysis (2nd ed.). New York, NY: McGraw Hill.
Sheeran, P. (2002). Intention-behavior relations: A conceptual and empirical review. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 12, pp. 1-36). Chichester, UK: Wiley.
Sheppard, B. H., Hartwick, J., & Warshaw, P. R. (1988). The theory of reasoned action: A meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research, 15, 325-343. doi:10.1086/209170
Teva, M. P., Bermudez, M. P., & Buela-Casal, G. (2009). Characteristics of sexual behavior in Spanish adolescents. The Spanish Journal of Psychology, 12(2), 471-484.
Trafimow, D. (2004). Problems with change in [R.sup.2] as applied to theory of reasoned action research. British Journal of Social Psychology, 43, 515-530. doi:10.1348/0144666042565344
Warshaw, P. R., & Davis, F. D. (1984). Self-Understanding and the accuracy of behavioral expectations. Personality and Social Psychology Bulletin, 10(1), 111-118. doi:10.1177/0146167284101013
Zeelenberg, M. (1999). Anticipated regret, expected feedback and behavioral decision-making. Journal of Behavioral Decision Making, 12, 93-106. doi:10.1002/(SICI)1099-0771(199906)12:2 <93::AID-BDM311>3 .0.CO;2-S
Pilar Carrera, Amparo Caballero, Dolores Munoz, and Luis Oceja
Universidad Autonoma de Madrid (Spain)
Correspondence should be addressed to: Pilar Carrera. Universidad Autonoma de Madrid. Departamento de Psicologia Social, Facultad de Psicologia, C/ Ivan Pavlov, 6, Cantoblanco, 28049 Madrid. (Spain). E-mail: email@example.com
Received March 18, 2010
Revision received December 10, 2010
Accepted December 16, 2010
Table 1 Mean, Standard Deviation and N of anticipated emotions Mean (SD) N Anger 4.10 (1.99) 28 Sadness 3.75 (1.86) 12 Shame 3.64 (1.54) 14 Regret 4.51(1.88) 29 Fear 4.30 (2.06) 36 Joy 4.67 (1.63) 28 Table 2 Hierarchical regressions of BI and BE on TPB variables and anticipated emotional profiles and index (Study 1) Dependent Model Variable(s) [beta] F [DELTA] Variable entered [R.sup.2] BI a. attitude -.40 *** 15.6 *** .17 perceived control -- social norm -- AEPs .44 *** b. attitude -.33 ** 7.61 ** .08 perceived control -- social norm -- index BE a. attitude -.29 ** 16.9 *** .21 perceived control -.22 * social norm .48 *** AEPs b. attitude -.19 + 22.7 *** .24 perceived control -.18 ++ social norm -.54 *** index *** p < .001; ** p < .01; * p < .05, ++ p < .08, + p < .09 Table 3 Means, Standard Deviations and F tests for TPB variables Low-experience Moderate-high F Mean (SD) experience Mean (SD) Attitude 5.91 (.98) 5.28(1.05) 13.08 *** Social Norm 4.31 (1.29) 3.98 (1.34) 2.33 Perceived Control 6.41 (.77) 5.60 (1.17) 22.42 *** Behavioral Intention 1.52 (1.19) 2.72(1.81) 22.01 *** Behavioral Expectation 1.95 (1.25) 3.13 (1.81) 19.95 *** * p < .05; ** p < .01; *** p < .001 Table 4 Mean, Standard Deviation and N of anticipated emotional categories Low-Experience Moderate-High Experience Mean (SD) N Mean (SD) N Anger 5.04 (1.53) 21 4.09 (1.37) 21 Sadness 4.91 (1.44) 23 3.94 (1.34) 18 Shame 4.41 (1.63) 34 4.58 (1.54) 29 Regret 5.00 (1.72) 36 3.90 (1.91) 42 Fear 4.52 (1.77) 17 3.40 (1.42) 10 Joy 3.86 (1.28) 23 4.20 (1.24) 40 Table 5 Hierarchical regressions of BI and BE on TPB variables and anticipated emotional profiles and index (Study 2) Dependent Model Variable(s) [beta] Variable entered Low experience BI a. attitude -- perceived control -- social norm .56 *** AEPs b. attitude -- perceived control -- social norm .25 + index BE a. attitude -- perceived control -- social norm -- AEPs b. attitude -- perceived control -- social norm -- index Moderate-High BI a. attitude experience perceived control -.22 * social norm -.40 *** AEPs b. attitude -.36 ** perceived control -- social norm -.25 ** index BE a. attitude -- perceived control -.20 + social norm .42 *** AEPs b. attitude -.29 *** perceived control -- social norm -.27 *** index Dependent Model F [DELTA] Variable [R.sup.2] Low experience BI a. 18.23 *** .22 b. 3.68 + .04 BE a. 2.56 ns b. .51 ns Moderate-High BI a. 13.81 *** .13 experience b. 7.27 ** .06 BE a. 14.1 *** .14 b. 7.36 ** .06 *** p < .001; ** p < .01; * p < .05, + p < .06
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||texto en ingles|
|Author:||Carrera, Pilar; Caballero, Amparo; Munoz, Dolores; Oceja, Luis|
|Publication:||Spanish Journal of Psychology|
|Date:||Nov 1, 2011|
|Previous Article:||Learning about faces: effects of trustworthiness on affective evaluation.|
|Next Article:||Cost of mental set reconfiguration between digits and their photisms in synaesthesia.|