“Most folks are about as happy as they make up their minds to be” (quote attributed to Abraham Lincoln by Dr. Frank Crane in 1914).

The assumption in New Zealand (the site of the present study) and other Western countries is that most people strive to experience happiness most of the time (Oishi et al., 2004; Howell et al., 2016). However, recent research on what is called “happiness aversion” (Agbo & Ngwu, 2017; Joshanloo & Weijers, 2014) suggests that a segment of the population actually seeks to avoid situations where they might experience happiness. In an attempt to understand this counterintuitive finding, theory and research has sought to identify different dimensions of motivational systems as well as emotion regulation (ER) strategies to explain why someone might want, or not want, to feel the positive emotion of happiness (Krasko et al., 2020). ER describes the actions we take to try to influence which emotions we have, when we have them, and how we experience and express them (Gross, 1998). In essence, ER is the process by which people try to change an existing emotion into a more desired emotion (Thompson, 1994). People often regulate emotions to change what they feel or might feel into what they want to feel (Gross, 2015; Tamir, 2016), and, importantly, several lines of research indicate that people vary considerably in what they want to feel. Research on happiness aversion, for instance, focuses on motives to experience negative emotions as well as motives to avoid experiencing positive emotions (Gilbert et al., 2012), and some research suggests that accepting negative emotions may be beneficial (Ford et al., 2017). Although the happiness aversion research has highlighted the importance of motives to experience or to avoid experiencing certain emotions, the current literature remains thin on understanding the general role of such emotion motives within the study of emotion regulation.

Research supports the view that some ER strategies tend to be associated with negative mood outcomes, while other ER strategies tend to be associated with positive mood outcomes. For example, the ER strategy of “expressive suppression” (the intentional reduction of facial expressions that would convey emotions) is considered by Gross and Thompson (2007) to be maladaptive, i.e., positively correlated with negative outcomes; whereas, they see cognitive reappraisal, e.g., seeing the silver lining in a bad situation, as effective in promoting positive outcomes. Critically, research is lacking on what motives people have that lead them to use either expressive suppression or cognitive reappraisal (or any of the many other ER strategies). On this point, Gross (2015) noted that “Emotion regulation goals may include decreasing or increasing either negative emotion or positive emotion” (p. 5). He pointed out that although most people, most of the time, want to increase positive emotions and decrease negative emotions, the opposite goals are sometimes pursued, particularly in situations requiring instrumental action, e.g., expressing anger at someone thwarting one’s goal. To account for this occasional inconsistency, the field needs to have a clear distinction between emotion motives and goals and associated empirical research to detail how they constrain and predispose someone to feel certain emotions and not others.

Emotion motives and goals

As Tamir et al. (2020) have succinctly stated, “emotion regulation is motivated” (p. 115). They provide several key terms that we agree should come into common use in the literature. Emotion motives refer to higher-order goals, such as “I want to be a generally happy person.” Emotion goals are lower-order goals that serve to support higher-order goals, e.g., “I want to feel pride in my academic work so that I will be a happy person.” Both motives and goals stand as intentions to experience or to not experience certain emotions (or classes of emotions), but the key distinction between the two seems to be that motives are conceived of as more stable and trait-like compared to goals, which are thought to be more context-sensitive and state-like. With this distinction in mind, the present study was designed to explore whether emotion motives significantly predict their corresponding emotion goals in a daily diary assessment.

Following Gross’s (2015) and Tamir’s (2020) observations about motives and goals that underlie emotion regulation efforts, we seek to empirically test what we call the two level motivational theory of emotion regulation motives. Consistent with Gross and Tamir, we theorise that people hold at an abstract level general trait-level motives concerning approaching and avoiding different valenced emotional states. Individuals will consistently strive according to these motives to obtain their desired emotions, and we aver that these motives are largely based on reasonably stable personality characteristics and past experiences. However, these general trait-level guidelines may sometimes be altered within particular contexts in order to obtain different desired outcomes in the moment. For example, we predict that someone who seeks to avoid negative emotions in general is more likely to avoid specific situations in their life in which they might be provoked to feel negative emotions. Thus, our theoretical stance is that people will generally extend their trait-level emotion motives into particular contexts in the form of consistent state-level emotion goals. However, we qualify this initial prediction with the axiom that trait-level motives will not predict state-level motives in all cases all of the time because people understand that they may need to adjust their customary stance in order to address pressing situation-specific needs. For example, a person who generally avoids negative emotions may find themself becoming incensed by a particularly egregious violation of their autonomy, and they may embrace a sense of righteous anger because it serves an instrumental purpose to rectify the situation. Despite general intentions to feel emotions of a particular valence and intensity/duration, individuals are likely to make exceptions in the moment to experience typically undesired emotions in order to optimally address the pressing needs of the situation. Thus, emotion motives will usually guide and constrain emotion goals for everyday living, but exceptions will occur due to the usefulness for a person to experience and act upon the full range of emotional responses that all humans possess.

Although Gross (2015) theorized that people may be motivated to increase or decrease either negative or positive emotions, no self-report measure designed to capture this range of ER motives existed until a new measure of emotion regulation was developed by Bloore et al. (2020). Their General Emotion Regulation Measure (GERM) assesses four distinct (but related) emotion motives, two of which are hedonic and two of which are contra-hedonic. The two hedonic motives captured by the GERM are “trying to experience positive emotions” and “trying to avoid experiencing negative emotions,” whereas the contra-hedonic motives are “trying to experience negative emotions” and “trying to avoid experiencing positive emotions.” Although Bloore et al. (2020) have shown that GERM motives are strongly associated with concurrent mood outcomes (e.g., subjective happiness, depressive symptoms) in a single-occasion survey, no studies have examined whether GERM motives (assessed in a trait-based form or with momentary assessments within a daily diary study) predict momentary positive and negative affect. To this end, we obtained data from individuals in a single-occasion survey in conjunction with a diary study which obtained reports of state-based emotion goals and positive and negative affect once daily over 14 days. Based on the findings of Bloore et al. (2020), we expected to find that the two hedonic motives would predict greater positive affect, whereas the two contra-hedonic motives, although less commonly endorsed, would predict greater negative affect. In other words, we predicted that participants would generally achieve their emotion motives (trait-based) and emotion goals (state-based) in their everyday lives.

A related question that we sought to explore was the degree of overlap among hedonic and contra-hedonic emotion motives, which has been termed the ‘bivariate vs. bipolar’ issue with regard to positive and negative emotions (Huppert & Whittington, 2003; Zhao & Tay, 2022). That is, should we understand the motives to regulate positive emotions as operating in a distinct and separable way from the motives to regulate negative emotion, or should we understand the desire to increase one as necessarily involving the desire to decrease the other? Although a correlation of approximately − 0.50 between positive and negative emotion-based constructs has been reported in previous research (e.g., Keyes, 2005), much of the available data in the literature are concurrent (i.e., collected at a single point in time) and little work has examined the separability or overlap of valenced emotion motives. Bloore et al.’s (2020) concurrent results on emotion motives revealed a weak negative correlation between ‘trying to experience positive emotions’ and ‘trying to experience negative emotions (r = − 0.21, signifying shared variance of about 4%). The relatively modest correlation suggests that valenced emotion motives may support the bivariate view rather than the bipolar view, but longitudinal data are needed to better capture the amount of overlap manifested over time. We sought in the present study to test whether more support would be found for the bivariate view (i.e., little overlap) between hedonic and contra-hedonic motives over time.

Aims and hypotheses of the current study

Our chief aims in the present research were: (1) to determine whether trait-based emotion motives would reliably predict state-based (momentary) emotion goals in a daily diary study; (2) to ascertain whether these state-based emotion goals, in turn, would predict momentary positive and negative mood states in a multilevel analysis; and (3) to test whether the bivariate or bipolar view of valenced emotion motives would be supported.

Based on theory and research (Bloore et al., 2020) cited above, we proposed five hypotheses:

H1) Trait emotion motives would predict daily affect in the diary study in valence-consistent ways, i.e., hedonic trait motives would positively predict positive daily mood, and contra-hedonic trait motives would positively predict negative daily mood.

H2) The four trait emotion motives would predict corresponding state emotion goals assessed with momentary assessments, i.e., hedonic trait motives would positively predict hedonic state goals and similarly for contra-hedonic motives and goals.

H3) The hedonic state emotion goals would positively predict positive momentary mood, and the two contra-hedonic state emotion goals would positively predict negative momentary mood.

H4) Valence-consistent state emotion goals would mediate between trait-based emotion motives and the relevant mood outcome. For example, we expected that hedonic state goals would mediate between hedonic trait motives and positive emotional mood.

H5) We would find weak to null associations between hedonic and contra-hedonic motives over time, suggesting relative independence between these two emotion motive systems, thereby supporting the bivariate theory perspective over the bipolar perspective.

Method

Participants

Two hundred and sixty-eight participants were recruited from a mid-sized university in New Zealand as part of a data collection exercise within an undergraduate Introductory Psychology course. The only inclusion criterion was being a student in the course, and no one who volunteered to participate was excluded from participating. Participants were recruited for the study in three separate academic terms. The sample was composed of 88 students in Term 1 of 2020, 137 students in Term 2 of 2020, and 43 in Term 1 of 2021. The self-reported gender breakdown was 72 males, 190 females, and 6 ‘other gender’, which, although asymmetrical, is typical of undergraduate psychology courses. Participants ranged from 18 to 42 years old, with 73% reporting being between 18 and 20 years.

Analyses were performed to determine if the three samples differed in terms of mood outcomes, as data were collected over 18 months when the coronavirus pandemic waxed and then waned in its effects upon New Zealand residents. No differences were found for momentary negative affect between the three groups, and only a small effect size was found for momentary positive affect (partial η2 = 0.01). The latter indicated a small increase in positive affect over time (coincident with the easing of pandemic restrictions). Since these three datasets were very similar, we combined them in order to make analyses simpler and to obtain greater statistical power. The size of the dataset allowed for the identification of small effect sizes (Cohen, 1992). Participation in the study was voluntary, informed consent was collected from all participants, and ethics approval was obtained from the university.

Procedure

The participants provided responses in two separate but related data collections: a single-occasion survey followed by a 14-day daily diary study. All participants completed the survey first, in which they responded to the trait version of the General Emotion Regulation Measure (Bloore et al., 2020). The GERM questions were phrased to assess what individuals strive for in general (see wordings of question stems below), and we therefore considered these assessments to capture more trait-like aspects of emotion regulation.

The following day, participants began the 14-day daily diary study. Once each day in the evening, participants answered questions about their goals to try to experience and to try to avoid experiencing both positive and negative emotions in the preceding 24 h, as well as the momentary positive and negative affect that they were experiencing “now”. Since these momentary reports occurred within the context of daily living and could vary from day to day, they were considered to assess more state-like emotion states.

Measures

Trait assessment of GERM motives

The General Emotion Regulation Measure is an assessment tool of trait-like general emotion regulation motives developed by Bloore et al. (2020). The GERM was created in order to measure the intensity of motives concerning how much individuals generally try to experience and try to avoid experiencing the two groups of valenced emotions: positive and negative emotions. The GERM in its full form uses 24 distinct emotion terms as emotion prompts: the twelve positive emotions are happiness, gratitude, hope, love, liking a person, relief, pride, peacefulness, determination, joy, enthusiasm, and compassion; and the twelve negative emotions are fear, anxiety, guilt, contempt, frustration, disgust, anger, regret, sadness, disliking a person, shame, and distress. The large number of emotion terms encompasses a wide range of affect within the two broad groups of positive and negative emotions. Participants are asked: (1) “How often do you try to experience”, and (2) “How often do you try to AVOID experiencing” the specifically named emotions on a 5-point Likert scale (1 = “never”, 2 = “occasionally”, 3 = “about half the time”, 4 = “most of the time” and 5 = “all of the time”). Selection of emotions used in the scale were based on key emotion terms identified in Roseman’s appraisal theory of emotions (Roseman et al., 1990), and a comprehensive review of the emotion literature.

Average scores for trying to experience positive emotions was termed ExpPosT (T = trait); trying to experience negative emotions was termed ExpNegT; trying to avoid experiencing positive emotions was termed AvdPosT; and trying to avoid experiencing negative emotions was termed AvdNegT. Cronbach’s alphas for the four trait GERM motives were found, consistent with alphas reported by Bloore et al. (2020), to be excellent: trying to experience positive emotions (ExpPosT) = 0.90; trying to experience negative emotions (ExpNegT) = 0.92; trying to avoid positive emotions (AvdPosT) = 0.89; and trying to avoid negative emotions (AvdNegT) = 0.95.

Momentary reports of GERM goals

The full 24-item measure was too long to use on a daily basis, so in order to shorten it for momentary assessments, we chose five positive emotions (i.e., happiness, hope, peacefulness, gratitude, and love) and five negative emotions (i.e., shame, frustration, sadness, anxiety, and anger). These terms captured a range of frequently endorsed and easily recognised positive and negative emotions (Bloore et al., 2020). To contextualise the question for the daily diary study, we slightly modified the trait-based emotion prompts to “Over the last 24 hours, to what extent did you try to experience [emotion]?” and “Over the last 24 hours, to what extent did you try to AVOID experiencing [emotion]?” The 10 emotion terms were presented randomly after these questions, yielding four groupings of state-assessed GERM goals analogous to the trait-assessed motives. We sought to constrain participants’ memories to the last 24 h because we wanted to make it easier for the participants to accurately remember events and goals that had occurred since the last assessment. Labels for these four emotion goals were similar to the trait motives except we replaced the suffix of T with S for ‘state’, e.g., ExpPosS.

To obtain estimates of internal consistency reliability, we computed occasion-specific Cronbach’s alphas for the 14 separate days for each of the four emotion goals. The Cronbach’s alphas for trying to experience positive (ExpPosS) varied from 0.86 to 0.94; trying to experience negative (ExpNegS) varied from 0.82 to 0.90; trying to avoid experiencing positive (AvdPosS) varied from 0.70 to 0.91; and trying to avoid experiencing negative (AvdNegS) varied from 0.92 to 0.96. Thus, all four emotion goals yielded good to excellent internal reliability across the 14 days. Further, estimates of ICC for these variables were found to be excellent as well, ranging from 0.21 to 0.37, which showed that considerable within-subject variance existed in the predictor variables.

Momentary reports of positive and negative affect

Positive and negative affect were assessed by fourteen daily mood self-report diary reports. Daily mood was assessed with two valenced subcategories: positive and negative mood. For momentary positive mood, participants responded to this question: “How do you feel NOW in reaction to the events over the last 24 hours?” Participants used a 3-point Likert scale (1 = “not at all”; 2 = “somewhat”; and 3 = “a lot”) to rate their momentary positive affect on four positive emotions: contented, relieved, happy, and grateful. For momentary negative mood participants responded to the same question for four negative emotions: disappointed, upset, angry, and frustrated.

Subsequently, we recoded the data to create averages of positive and negative momentary emotions called NowPos and NowNeg. The occasion-specific Cronbach’s alphas for NowPos varied from 0.71 to 0.84 and NowNeg varied from 0.75 to 0.83. Thus estimates of internal consistency reliability within days for these measures ranged from adequate to good. Further, estimates of ICC were 0.50 and 0.55 for NowPos and NowNeg respectively, which supported a key assumption of multilevel modeling, namely that significant within-subject variance existed in the outcome variables.

Planned analyses

The first three steps in the linear mixed model analyses were to: (1) test the predicted relationships of trait motives at Level 2 with valenced mood states at Level 1 (the c path in a mediation); (2) test the predicted relationships between trait motives at level 2 with state goals at Level 1 (the a path in a mediation), and (3) test the predicted relationships between state goals at Level 1 with valenced mood states at Level 1 (the b path in a mediation). In the fourth step, we performed mediation analyses within this multilevel modeling framework to determine whether the ability of Level 2 scores to predict valenced mood states would be mediated by the state goals at Level 1 in predicted ways (products of a paths by b paths). And finally, we conducted generalized linear mixed model (GLMM) analyses on each of the two outcome variables: by including all trait-level motives and all state-level goals simultaneously, we sought to ascertain the strongest relationships across these levels in predicting daily mood.

Data availability

Readers may access the stored data on the OSF website for the project named “Multilevel analysis of how trait-assessed emotion motives predict valenced affective states in daily life” under the name of the first author.

Results

Missing data and analytic procedures

Missing values constituted on average 17.6% of the daily diary dataset and 26.9% of the survey dataset. When using linear mixed model estimation, we employed the Full Information Maximum Likelihood (FIML) estimation algorithm to compensate for the missing data. FIML is an effective way to maximize statistical power (Enders & Bandalos, 2001) as it uses the dataset’s covariance matrix to perform all analyses rather than imputing individual missing values.

Descriptive statistics

Table 1 presents the means, standard deviations, indices of normality, and missing data for the dataset. Consistent with descriptive statistics reported in Bloore et al. (2020), means for ExpPos and AvdNeg were noticeably higher than for ExpNeg and AvdPos, and this apparent difference was mirrored between trait and state variables. Distributional characteristics of variables fell within acceptable ranges except for AvdPosS, which was excessively kurtotic. The mean for this variable was low, so the elevated skewness and kurtosis was probably due to a floor effect. Rather than transforming this variable, we left it unchanged to retain a comparable metric with all other variables. Results based on AvdPosS should be viewed with caution, however. The mean for daily positive mood fell about at the mid-point of the three-point scale, and the mean for daily negative mood was about one standard deviation lower, mirroring somewhat higher motives and goals for positive emotions compared to negative emotions.

Table 1 Means and standard deviations of all variables and frequencies of missing data
Full size table

In addition, Table 2 below reports the means, within- and between-person variances, and the ICCs, after FIML imputation, for the six Level 1 variables. Notable is that the ICCs for the two mood outcomes were large, indicating considerable within-person variability relative to the emotion goals, which yielded lower within-person variability. Also of interest is that the two hedonic emotion goals yielded the lowest ICCs, suggesting that these motives were more strongly determined by between-person characteristics than situational aspects compared to the contra-hedonic motives. Correlations among trait variables and covariances among state variables (see Table 3) largely evidenced positive associations within the two groups of hedonic and contra-hedonic motives and largely null or weak associations across the two groups of variables.

Table 2 Means, within- and between-person variances, and ICCs for all Level 1 variables
Full size table
Table 3 Correlations among State Variables and Covariances among Trait Variables
Full size table

Relationships between Trait GERM motives, state GERM goals, and daily levels of positive and negative affect

Since the data took the form of a nested dataset (daily reports nested within individual participants), linear mixed model estimation was utilized to compute the following regressions. We specified random intercepts for the predictor variables in the regressions in order to obtain the most generalizable results. Further, we covaried out gender and age in all analyses in order to generalise to the population of university students.

Prediction of daily mood by trait GERM motives

We first sought to test Hypothesis 1, that the two hedonic trait GERM motives of ExpPos and AvdNeg would positively predict residualised positive daily mood, and that the two contra-hedonic trait GERM motives of ExpNeg and AvdPos would positively predict residualised negative daily mood.

Table 4 Linear mixed model analyses for the relationships between trait emotion motives (Level 2) and daily emotion states (Level 1). Note. Gray shading indicates a result that was consistent with predictions, and lack of shading indicates an analysis for which no prediction had been made. B = unstandardised regression coefficient; SE = standard error; T = trait motive measure
Full size table

All four predicted relationships yielded statistically significant (at p < 0.05) positive relationships between trait motives and residualised valenced daily affect (see Table 4). We next considered a Bonferroni correction to adjust for repeated hypothesis testing on portions of the same dataset. We adopted the Holm-Bonferroni method, however, as it is less conservative than the typically used Bonferroni correction because it controls for family-wise error rate (Wikipedia, 2023). Sequentially applying adjusted p-levels in a step-down fashion, we determined that we could reject the null hypothesis for all four relationships. Pseudo-R2 values of the four significant effects, in order from top, suggested excellent model fit in all cases: 0.58, 0.54, 0.53, and 0.55. Gender and age were not significant predictors of either outcome. Thus, Hypothesis 1 was supported, the results suggest that trait-level emotion motives robustly predicted levels of the expected state-level valenced daily emotional state, namely hedonic motives predicted higher levels of daily positive affect and contra-hedonic motives predicted higher levels of daily negative affect.

Prediction of state emotion goals by trait emotion motives

We next tested Hypothesis 2, namely that each of the four trait emotion motives would positively predict its state emotion goal counterpart (e.g., ExpPosT = > ExpPosS).

Table 5 Linear mixed model analyses for the relationships between trait emotion motives (Level 2) and state emotion goals (Level 1). Note. Gray shading indicates a predicted association, and lack of shading indicates an analysis for which no prediction had been made. B = unstandardised regression coefficient; SE = standard error; T = trait motive; S = state goal
Full size table

Analyses for all four predicted relationships (gray shading) yielded statistically significant positive relationships between each trait emotion motive and its state-based goal counterpart (see Table 5). In other words, the assessed emotion motives demonstrated significant consistency from trait to state versions. After applying the Holm-Bonferroni adjustment for repeated analyses, four other unpredicted relationships were deemed to be statistically significant as well. Importantly, these four additional relationships verified that hedonic motives predicted hedonic goals, while contra-hedonic motives predicted contra-hedonic goals. The last finding (AvdNegT = > ExpNegS) was rejected because it failed the Holm-Bonferroni method. No significant predictions of outcomes by gender or age were noted. And pseudo-R squared values ranged between 0.60 and 0.79, indicating superior model fit in all cases. Thus, Hypothesis 2 was supported in full, albeit with additional theoretically sensible findings. The findings support the bivariate view in that no significant associations were noted between hedonic and contra-hedonic constructs.

Prediction of daily mood by state emotion goals

We next examined Hypothesis 3, namely that the two state-based hedonic emotion goals of ExpPos and AvdNeg would positively predict positive daily mood, and that the two state-based contra-hedonic emotion goals of ExpNeg and AvdPos would positively predict negative daily mood. As Level 1 predictors can be partitioned into within-person vs. between-person variance components (Chang & Kwok, 2022), we did so for the Level 1 emotion goals, and used both as predictors of Level 1 valenced daily mood.

Table 6 Linear mixed model analyses for the relationships between emotion goals (partitioned into between- and within-variance) and daily emotion states (all at Level 1). Note. Gray shading indicates a significant result consistent with predictions. No predictions were made for unshaded cells. B = unstandardised regression coefficient; SE = standard error; S = state goal; ns = nonsignificant
Full size table

All four predicted relationships evidenced statistical significance in a positive direction between particular between- and within-person emotion goal motives and valenced mood (see Table 6). The fact that both variance components predicted the outcomes speaks to the robustness of these relationships. As predicted, the two hedonic state goals (ExpPosS and AvdNegS) both significantly positively predicted positive affect. In addition, both contra-hedonic goals (ExpNegS and AvdPosS) significantly predicted negative affect. Again, no significant predictions of outcomes by gender or age were noted. And pseudo-R squared values ranged between 0.57 and 0.70, indicating superior model fit. Thus, Hypothesis 3 was uniformly supported, i.e., a person’s valenced emotion goal during a previous 24-hour period predicted corresponding positive or negative emotion in the present for all four emotion goals.

Notably, several unpredicted significant findings were obtained as well. In three cases, the within-person variance component of the emotion goals negatively predicted the opposite-valenced outcome (ExpNegS and AvdPosS predicted lower positive affect and ExpPosS predicted lower negative affect). These three findings at Level 1 support the bipolar view of valenced emotion. The fourth unanticipated result showed that AvdNegS predicted negative affect, which may indicate that some state-based efforts to avoid negative emotions may not work, i.e., these efforts may backfire and unintentionally sustain negative affect.

Mediation analyses

The next set of analyses tested Hypothesis 4, namely that trait emotion motives would predict their relevant state emotion goal counterparts, which, in turn, would predict corresponding daily diary moods (see Table 7). In particular, the two hedonic trait motives were expected to positively predict the two hedonic state goals, which, in turn, were expected to positively predict an increase in positive affect from one day to the next. Four corresponding mediations were also proposed for the two contra-hedonic motives and goals predicting an increase in negative affect. The predictions were evaluated by computing multilevel model mediations using a 2-1-1 framework in the R package lavaan (Crowson, 2020; Rosseel, 2012).

Table 7 State emotion goals (within-subject variance) mediating between trait emotion motives and daily diary mood. Note. IV = independent variable; DV = dependent variable; T = trait motive; S = state goal; B = unstandardised regression coefficient; SE = standard error
Full size table

After applying a Holm-Bonferroni procedure, all eight predicted mediations were statistically supported. In addition, based on significant a and b paths reported in Tables 5 and 6, we explored in a post hoc fashion eight unpredicted but theoretically plausible mediations that would illuminate the bipolar vs. bivariate question, and we found that six additional mediations yielded statistical significance (see Table 8) after applying the Holm-Bonferroni adjustment.

Table 8 Exploratory mediations: state emotion goals (within-subject variance) mediating between trait emotion motives and daily diary mood
Full size table

Taken together, the predicted mediation findings suggest that hedonic trait-based emotion motives significantly predicted hedonic state-based emotion goals, which, in turn, predicted greater positive daily affect, and all four combinations among contra-hedonic motives, goals, and negative affect were supported as well. In addition, the exploratory analyses suggested that consistent contra-hedonic motives and goals significantly diminished daily positive affect and consistent hedonic motives and goals somewhat diminished daily negative affect in agreement with the bipolar perspective.

Full models of trait and state motivations predicting the two mood outcomes

As informative as the specific planned analyses we presented above are, we also thought that insight could be obtained by pitting the four trait emotion motives against the four state emotion goals in predicting the two valenced mood outcomes. We utilized the GLMM (generalized linear mixed modeling) analytic technique because it models random effects longitudinally more robustly than linear mixed modeling (de Melo et al., 2022). Further, we performed the analyses in a hierarchical fashion, beginning with a block of the trait emotion motives, followed by a block of the state emotion goals for both outcomes separately.

Outcome of daily positive mood

Table 9 shows that the initial block of emotion motives identified ExpPosT as the only significant predictor, B = 0.182, SE = 0.03, 95% CI = [0.12, 0.24], p < 0.001. With the addition of the second block, it is notable that ExpPosT lost its ability to predict the outcome (p = 0.20), while two emotion goals demonstrated the ability to significantly predict the outcome: ExpPosS, B = 0.173, SE = 0.01, 95% CI = [0.15, 0.20], p < 0.001, and AvdPosS, B = − 0.048, SE = 0.02, 95% CI = [-0.08, − 0.02], p = 0.004. The pseudo-R square value with the second block of predictors included indicated good model fit. This pattern suggests that the trait emotion motive of trying to experience positive emotions initially explained significant variance in daily positive moods, but that when the state emotion goals were introduced, they did a better job of predicting the outcome. Daily goals of trying to experience positive emotions positively predicted, and the absence of trying to avoid experiencing positive emotions positively predicted, variance in daily positive moods.

Outcome of daily negative mood

Table 9 also shows that the initial block of emotion motives revealed, as expected, ExpNegT as a significant predictor, B = 0.153, SE = 0.03, 95% CI = [0.09, 0.22], p < 0.001, as well as AvdPosT, B = 0.078, SE = 0.04, 95% CI = [0.01, 0.15], p = 0.03. Unexpectedly, ExpPosT predicted daily negative affect in a positive direction, B = 0.099, SE = 0.03, 95% CI = [0.05, 0.15], p < 0.001. This positive relationship may indicate that striving for positive emotions may sometimes fail and lead to greater negative affect.

As above, with the addition of the second block of predictors, it is notable that the two contra-hedonic trait motives lost their ability to predict the outcome (ps = 0.19 and 0.90), while all four of the emotion goals demonstrated the ability to significantly predict the negative mood outcome: ExpPosS, B = − 0.038, SE = 0.01, 95% CI = [-0.06, − 0.02], p < 0.001, ExpNegS, B = 0.230, SE = 0.01, 95% CI = [0.20, 0.26], p < 0.001, AvdPosS, B = 0.056, SE = 0.01, 95% CI = [0.03, 0.09], p < 0.001, AvdNegS, B = 0.035, SE = 0.01, 95% CI = [0.02, 0.05], p < 0.001. As with the regression with positive affect, the pseudo-R square value with the second block of predictors included yielded good model fit. This pattern suggests that the state motives were more successful predictors of negative mood than trait motives because the daily goals are temporally closer and likely more contextually relevant than abstract motives. The directions of association were sensible in all cases but two. It makes sense that both trait-level contra-hedonic motives positively predicted daily negative mood, and further at the daily level, it makes sense that ExpPosS would negatively predict and ExpNegS and AvdPosS would positively predict negative mood. The fact that both ExpPosT and AvdNegS were positively predictive of negative mood is counter-intuitive. It may have been that individuals were striving to avoid experiencing negative emotions that persisted over time, or it may have been that trying to avoid negative affect boomeranged in an undesired direction, leading to the identification of a positive association.

In terms of identifying the most robust relationships among all possible associations, it seems that trying to experience daily positive emotions and not trying to avoid experiencing daily positive emotions were the strongest predictors of daily positive mood. In contrast, the strongest predictors of daily negative emotions were trying to experience daily negative emotions and trying to avoid experiencing daily positive emotions, although the absence of striving for positive daily emotions contributed as well.

Table 9 GLMM analysis of trait emotion motives and state emotion goals predicting daily diary mood
Full size table

Did we find bivariate or bipolar associations between valenced emotion motives?

Throughout the analyses performed to test the first four hypotheses, a clear pattern of bivariate association between hedonic and contra-hedonic motives emerged; findings indicated a marked separability between these two modes of seeking valenced emotions. Consistent with Hypothesis 5, no significant cross-motive association (e.g., from hedonic motive to contra-hedonic goal) was identified, and only three significant associations were noted (in Tables 3 and 5) between valenced motives and goals and the oppositely-valenced mood. We did find some evidence in favour of the bipolar view among the mediation results, but, on balance, most of the findings of the study seem to favour the bivariate view. These findings are notable because the bipolar view (Zhao & Tay, 2022) would postulate that pursuing hedonic emotion goals is likely to diminish daily negative mood and striving for contra-hedonic emotion goals is likely to diminish positive mood. Scant evidence of these processes were found. The bivariate approach, in contrast, suggests that hedonic and contra-hedonic goals and associated valenced emotions can co-exist reasonably well in a person’s emotional experience. The present evidence suggests that people’s pursuit of hedonic and contra-hedonic emotion motives and goals largely resulted in the desired emotion outcomes without much affecting the opposite-valenced mood states.

Discussion

This research tested whether emotion motives assessed in a general (trait-based) sense would evidence significant ability to predict daily valenced mood states. Table 4 showed that our data supported the hypotheses that trait-based emotion motives significantly and substantially predicted desired daily emotion states. The motives of ‘trying to experience positive emotions’ and ‘trying to avoid experiencing negative emotions’ significantly predicted higher levels of daily reports of positive affect, and the motives of ‘trying to experience negative emotions’ and ‘trying to avoid experiencing positive emotions’ both significantly predicted higher levels of daily reports of negative affect. In sum, participants’ trait-based emotion motives were generally consistent with and predictive of their subsequent experienced daily mood states, i.e., hedonic striving predicted higher positive affect and contra-hedonic striving predicted higher negative affect. Importantly, the concurrent relationships identified by the GERM self-report measure in Bloore et al. (2020) have been replicated and extended to the context of a longitudinal/daily diary study.

The second hypothesis proposed that the four trait-based emotion motives would consistently predict their state-based emotion goal counterpart in the daily diary study. Table 5 presented support for this prediction: each of the four emotion motives strongly and positively predicted the corresponding goal in the daily context. The strongest relationship was found for ExpPos (trying to experience positive emotions), but the other trait-to-state relationships were quite robust as well. In addition, four other theoretically consistent associations were noted, i.e., hedonic traits predicted hedonic goals and contra-hedonic traits predicted contra-hedonic goals. These findings, taken together, strongly support the inference that trait-based emotion motives substantially inform and predict state-based emotion goals in daily reports.

The third hypothesis concerned whether state-based emotion goals would predict experienced daily positive and negative affect (see Table 6). We found strong support for this prediction in that all four predicted relationships were statistically supported (for both between- and within-variance predictors). Thus, we obtained support for the view that participants who reported hedonic or contra-hedonic momentary intentions to pursue particular valenced emotion goals duly reported valenced affect consistent with those goals. These findings suggest that daily emotion goals functioned to predict subsequent daily emotion outcomes.

Fourth, we tested whether empirical support for eight proposed mediations from motives to goals to mood outcomes could be obtained (see Table 7). All of the eight mediations achieved statistical significance, and the results suggest that trait-based emotion motives significantly constrained state-based emotion goals, which, in turn, significantly predicted consistent mood outcomes.

And fifth, we largely found support for the bivariate perspective as opposed to the bipolar view in terms of associations between hedonic and contra-hedonic motivations. In only a few cases in these analyses did hedonic motives significantly predict negative affect, and in only a few cases did contra-hedonic motives significantly predict positive affect. Based on Bloore et al.’s (2020) findings showing weak associations of hedonic motives with contra-hedonic motives, we anticipated weak to null associations in the present data, and our present findings support this general prediction. The question of whether hedonic and contra-hedonic motives are bipolar or bivariate (see Zhao & Tay, 2022, for a discussion of this issue with concurrent mood outcome data), to our knowledge, has not been broached before, because a valid tool capable of comprehensively assessing these two types of motives has not previously existed. We think that individuals likely pursue multiple emotional outcomes simultaneously that are heterogeneous for hedonic tone. For example, one might be planning a surprise birthday party for a friend, which brings happiness, while at the same time being fearful that only a few people will agree to attend.

The GLMM analysis revealed the strongest associations among the many associations examined in this study. Two conclusions emerged from this examination. First, daily mood was better predicted by state-level emotion goals than by trait-level emotion motives, which makes sense given the temporal proximity and contextual relevance of state goals. And second, both positive and negative daily mood variables were predicted by a mixture of state-level hedonic and contra-hedonic goals. Valenced daily mood seems to not be entirely determined by striving for the corresponding emotions, but the absence of countervailing goals is also important. Thus, positive daily mood was strongly predicted by striving for the experience of positive emotions and, at the same time, not striving to avoid the experience of positive emotions.

Emotion motives, goals, and striving

Emotion regulation has been described as the set of efforts people make to move from a less desired emotional state to a more desired one (Thompson, 1994). A more inclusive conceptualization is that ER “requires the activation of a goal to up- or down-regulate either the magnitude or duration of the emotional response” (Gross, 2013, p. 359). Tamir (2009) would call this focus on an emotion goal the why of emotion regulation. Most of the research in the ER literature, in contrast, has focused on Gross and Thompson’s (2007) influential process model of emotion regulation as it describes in detail how a person seeks to fulfill their emotion goal. The current study elucidates aspects of the why question.

Not much work has been devoted to explicating emotion motives and goals, and most of the extant work is theoretical. Tamir and colleagues (2008, 2009) have made a distinction between instrumental and hedonic motivations for emotion regulation. For example, people generally want to experience positive emotions because those emotions feel pleasant, but sometimes want to experience negative emotions “to promote the attainment of long-term goals” (Tamir, 2009, p. 101). In this theoretical account, pursuing negative emotions may be instrumental: people “may prefer to feel useful emotions, even if they are unpleasant” (Tamir, 2009, p. 101). Indeed, Gross (2015), suggested a 2 × 2 matrix of emotion regulation goals: (1) the hedonic goals of increasing positive emotions and of decreasing negative emotions; and (2) the ‘counterhedonic’ goals of decreasing positive emotions and of increasing negative emotions. Notably, Gross’s (2015, p. 5) examples of counterhedonic ER efforts, e.g., “firing oneself up before a big game”, were instrumental in nature, in agreement with Tamir’s characterization of negative emotions.

Above we cited Tamir et al.’s (2020) pronouncement that “emotion regulation is motivated” (p. 115), and described their distinction between emotion motives and emotion goals. They also made a useful contrast between goal-setting and goal-striving. Goal-setting refers to specifying which particular emotions, such as pride, are targeted as desirable to feel, whereas goal-striving refers to focusing one’s energies and behaviours toward actually trying to achieve the emotion goal. Goal-striving has been much studied within the context of Gross’s process model of ER, but relatively little work has been devoted to understanding how goal-setting occurs. We would argue that findings of the present study begin to shed light on how general emotion motives inform and motivate the formation of specific emotion goals within one’s lived experiences.

We think it is germane to compare the present study with Tamir et al.’s (2019) study which explored the linkages between emotion goals and emotion regulation strategies. In their study, some participants were instructed to hold an hedonic emotion goal (i.e., “feel less negative emotion”), while others were urged to use the ER strategy of cognitive reappraisal (CR). Participants in the first condition succeeded in diminishing negative affect–regardless of whether they spontaneously used CR or some other strategy–to a similar level as the participants who were instructed to use the specific strategy of CR. Thus, it seems that the instantiation of emotion goals was sufficient to create changes in emotions in the desired direction.

The present study has both similarities to and differences from the Tamir et al. (2019) study. We similarly investigated linkages between emotion motives and emotion goals. One key difference, however, is that we sought to determine whether naturally occurring emotion motives, as opposed to experimentally manipulated goals or strategies, would significantly predict the desired mood states. A second difference is that, unlike Tamir et al. (2019), we did not examine the use of ER strategies to operationalise and instrumentally enact the motives. We would argue that the trait-based GERM scores captured emotion motives, in the sense that respondents told us how much they wanted to experience (or avoid experiencing) the two broad classes of positive and negative emotions. We would argue that these are motives because they are higher-order objectives on par with “I want to be a person who experiences a range of [positive/negative] emotions.” In contrast, we would also argue that the state-based GERM goals assessed in the daily diary study should be considered to be more specific than general emotion motives, because participants reported how much they wanted to experience particular emotions in a 24-hour window. In other words, the emotion goals measured by the GERM changed to some degree over time, probably due to changing circumstances in the person’s everyday life.

In sum we think that the results from the present study supported our expectations that we could identify predicted associations among emotion motives, emotion goals, and daily lived mood outcomes. This verification of expected linkages suggests that people, in general, are largely efficacious in feeling what they want to feel (in agreement with the adage attributed to Abraham Lincoln). In a theoretical vein, we argue that the present findings add more detail and nuance to the models posed by Gross (2015) and Tamir and colleagues (2008, 2009) as to why and how people regulate their emotions to achieve a ‘desired’ emotion.

Limitations and suggestions for future research

We did not measure emotion regulation efforts, e.g., cognitive reappraisal, at the daily level, so we were unable to provide comparable evidence to what was described in Tamir et al.’s (2019) experimental study. We did, however, assess trait-based ER strategies in the present research project (results were not included here due to space limitations), and preliminary findings suggest that particular ER strategies are multiply and complexly predicted by mixtures of emotion motives. As Tamir suggests (2009), the paths and linkages between emotion motives and goals to the choice of ER strategies are richly deserving of further research attention because this work will clarify how human motivations inform and drive strivings for emotion states.

Second, although there is a literature on ambivalence and conflict within personal strivings (Emmons & King, 1988), we did not pursue analyses that would elucidate the extent to which hedonic and contra-hedonic goal motives conflict or complement each other. Our findings suggest that our participants largely pursued hedonic or contra-hedonic motives in separate channels with little to no conflict with each other (based on the lack of negative associations across these two domains). A useful future direction would be to discern individual differences in this presumed conflict using the GERM measure.

Third, our findings are based on a single study, with most participants falling in a narrow age range (young adulthood), having a similar educational level, and being located within a specific culture. Thus, more research is needed to replicate these results in other cultural contexts, and with participants of different ages and of different economic backgrounds. The lack of cultural diversity means that our sample is unlikely to adequately reflect the range of cultural differences in various aspects of ER (e.g., Joshanloo & Weijers, 2014; Eid & Diener, 2001; Oishi et al., 2004; Tamir et al., 2016). Therefore, an important next step would be to develop and validate translated versions of the GERM scale and to investigate whether the present findings differ or are replicated among disparate cultures.

Fourth, our data collection was conducted during the emergence of the COVID-19 pandemic, which was a significant stressor for many people. The lockdown occurring in New Zealand during Term 1 of 2020 (the time of our first data collection) may have affected recruitment into the study and may have biased reports of emotion motives and mood states. For these reasons, it would be important to test whether or not results would be similar under more normal circumstances with a similar sample.

And finally, the sole use of self-report measures for valenced affect could represent another limitation, given that people may unconsciously bias or misreport their motives and the valence of their emotions. Further, we acknowledge that the particular emotion terms used in our momentary assessments might not be representative of the population of all emotion terms. Therefore, future studies replicating and extending these findings might include other methods of assessment, such as physiological measures of reported daily affect, including daily cortisol, blood pressure, or stress levels, and future EMA studies might utilise a wider or different range of emotion terms.

Conclusions

Our study found, first, that trait-based emotion motives and state-based emotion goals significantly predicted subsequent desired momentary affective states. On the basis of these findings we can draw the important conclusion that, on average, people were reasonably successful in achieving their emotion motives and goals. Consistent with other theory and research (Tamir, 2009, 2014; Tamir & Bigman, 2014; Tamir et al., 2020) concerning emotion goals, it seems that our participants much of the time were successful in regulating their experience of emotions in the desired direction. Second, our findings suggest that the two channels of hedonic and contra-hedonic motives and goals did not affect each other over time in any significant fashion. In particular, striving to achieve a hedonic emotion goal did not necessarily diminish striving to achieve a contra-hedonic goal or vice versa. Overall, people seem to take a proactive stance toward trying to have desired rather than undesired emotional states, and it seems, based on these data, that many people, much of the time, are successful in achieving these goals.