A Network Approach to Affect Regulation Dynamics and Personality Trait–Induced Variations: Extraversion and Neuroticism Moderate Reciprocal Influences between Affect and Affect Regulation Strategies

Affective components: Positive and negative affect

Few studies have focused on the possible influence of affective experiences on the subsequent implementation of affect regulation strategies (Brans, Koval, Verduyn, Lim, & Kuppens, 2013; Burns et al., 2008; Fredrickson & Joiner, 2002; Moberly & Watkins, 2008; Pavani et al., 2015). Nevertheless, their findings all stress the importance of considering affective valence (i.e. positive or negative affect). The broaden–and–build theory (Fredrickson, 1998, 2001) and the affect priming theory (Bowers, 1991) outline different mechanisms explaining how affective valence can influence the use of affect regulation strategies.

The broaden–and–build theory (Fredrickson, 1998, 2001) states that the experience of a positive affect momentarily triggers a cognitive and behavioural broadening mechanism. Such a broadening process is manifested by a temporary widening of the attentional scope and thought–action repertoire of the person who has felt this affect. This mechanism is supposed to encourage individuals to explore their environment when they encounter a healthy situation. Several studies confirm this hypothesis, notably revealing that a positive affect experience is followed by an increase in creativity and cognitive flexibility (Davis, 2009; Tsai, Lin, Chen, & Lin, 2014). Interestingly, creativity and flexibility are particularly important in the implementation of so–called broad–minded affect regulation strategies (Fredrickson & Joiner, 2002). These strategies are characterized by a broad perspective on the causes of one's affective feelings (e.g. discovering new ways of interpreting an unpleasant situation, identifying ways of altering it through concrete actions, directing attention away from it). If the experience of positive affect enhances creativity and cognitive flexibility, and if enhanced creativity and cognitive flexibility facilitate the implementation of broad–minded affect regulation strategies, then it is not surprising to observe that positive affect feelings primarily influence the use of this type of strategy (Burns et al., 2008; Fredrickson & Joiner, 2002; Pavani et al., 2015; but for an exception, see Brans et al., 2013).

The broaden–and–build theory (Fredrickson, 1998, 2001) also acknowledges that the experience of a negative affect momentarily engenders a cognitive and behavioural narrowing mechanism. Such a process is associated with an attentional fixation on the threats present in the environment, and a reduction in the repertoire of available thoughts and actions, in favour of ones that promote fight, flight, or freeze. This cognitive and behavioural narrowing following a momentary experience of negative affect has been demonstrated in numerous studies (e.g. Gasper & Clore, 2002; Hamilton, 1989; Matthews & Wells, 2000). Once again, the implementation of certain affect regulation strategies may be particularly sensitive to this process. These strategies can be labelled ‘narrow–minded’ strategies, because they are characterized by the fixation of thoughts and attention on certain stimuli, or by the inhibition of some behaviours. For instance, within a few hours of experiencing more negative affect, individuals are more inclined to ruminate (Brans et al., 2013; Moberly & Watkins, 2008; Pavani et al., 2015).

The mechanisms outlined in the affect priming theory (Bowers, 1991) are based on the specific organization of information in memory. According to this framework, each affect is represented in memory by a concept, and each affective concept in memory is related to the words, propositions and images that accompanied the previous feelings of this affect. Thus, the experience of an affect activates the corresponding affective concept in memory, as well as all the information to which it is related. In other words, feeling a positive affect activates positive information, whereas feeling a negative affect activates negative information. In turn, this information is thought to momentarily shape information processing. For example, several studies have shown that eliciting a particular affect results in the retrieval of memories, biases in attention deployment, and the interpretation of ambiguous situations that are all congruent with the valence of the initial induced affect (Bowers, 1991; Forgas, 1995). These affective priming effects may explain the influence exerted by affective experiences on subsequent engagement in particular affect regulation strategies, most of which involve the display of positive or negative thoughts and interpretations, or the direction of attention toward pleasant or unpleasant stimuli (Gross, 1998).

Affect regulation strategy components: Broad–minded and narrow–minded strategies

Not only is engagement in broad– or narrow–minded affect regulation strategies likely to depend on prior experiences of positive or negative affect, but it is also thought to influence subsequent feelings of positive or negative affect. For instance, the ability to flexibly change one's point of view, behaviour or focus of attention forms the basis of the main psychological interventions designed to promote affect regulation (Beck, 2011; Seligman, Rashid, & Parks, 2006). We chose to focus on three broad–minded strategies (i.e. appreciation, positive reappraisal, and distraction) and two narrow–minded strategies (i.e. expressive suppression and rumination). There were three reasons for this choice. First, each of these five strategies can be described as either broad– or narrow–minded. Second, their implementation, be it habitual or momentary, has been shown to exert a significant influence on affective valence. Third, these five strategies have been widely studied and are thus familiar to people interested in affect regulation research.

Appreciation refers to the initiative to notice and enjoy those aspects of life that individuals usually rush through (Seligman et al., 2006). This strategy can be seen as a broad–minded strategy, as its implementation rests on a change in one's cognitive habits. First, implementing appreciation involves directing attention to the present moment. Second, using this strategy entails a modification in the initial appraisal of a seemingly neutral situation that becomes pleasant once all its small positive aspects have been recognized. Appreciation has been shown to promote mainly positive affect experiences (Erisman & Roemer, 2010; Fagley, 2012).

Positive reappraisal consists in perceiving negative events as beneficial (Folkman, 1997). This strategy involves changing one's initial interpretation of a negative event by considering the positive features of this event. The contributions of creativity and cognitive flexibility to this process have already been acknowledged (e.g. Weber, de Assunção, Martin, Westmeyer, & Geisler, 2014). Positive reappraisal has been shown to influence affective experiences mostly by increasing positive affect feelings (Rood, Roelofs, Bögels, & Arntz, 2012; Shiota & Levenson, 2012). By contrast, the effect of this strategy on affective inertia (i.e. resistance of an affective experience to change from one moment to another; Kuppens, Allen, & Sheeber, 2010a), an important aspect of affective dynamics possibly captured by network approaches, was found to be negligible in one recent study (Koval, Butler, Hollenstein, Lanteigne, & Kuppens, 2015).

Distraction is a strategy whereby an individual's attention is diverted from a negative situation or the unpleasant feelings it elicits (Van Dillen & Koole, 2007). This strategy can be used by directing attention to the most neutral or positive aspects of a critical situation, or even by changing the situation (e.g. engagement in an alternative activity). In the short term, distraction can enhance feelings of positive affect (Brans et al., 2013) and reduce feelings of negative affect (Nolen–Hoeksema & Morrow, 1993), but its chronic use can have the opposite effect.

Expressive suppression consists in inhibiting the behavioural tendencies associated with the elicitation of an affect (John & Gross, 2004). This strategy entails suppressing one's facial expressions or gestures, to hide one's feelings from others. Expressive suppression can be regarded as a narrow–minded strategy, because it constitutes a form of behavioural inhibition, which is supposed to be encouraged by a narrowing mechanism triggered by experiences of negative affect. This strategy influences affective valence mainly by increasing experiences of negative affect (Brans et al., 2013; John & Gross, 2004). Furthermore, like positive reappraisal, it appears to have little effect on the inertia of affective feelings. Nevertheless, expressive suppression seems to be associated with higher inertia of affectively expressive behaviour (Koval et al., 2015).

Rumination refers to the passive and repetitive generation of thoughts about negative events or affects (Nolen–Hoeksema, 2000). This strategy can be viewed as a prototypical narrow–minded strategy, as it is defined by the fixation of thoughts and attention on a single stimulus. Rumination mainly influences affective experience through its impact on feelings of negative affect, either in the short term or the longer term (Nolen–Hoeksema, 2000). Furthermore, its habitual use has been shown to increase the inertia of negative affect feelings (Koval, Kuppens, Allen, & Sheeber, 2012).

Reciprocal influences between affect and affect regulation strategies

To summarize the numerous findings described previously, positive feedback loops could be observed between positive affect and broad–minded affect regulation strategies such as appreciation, positive reappraisal and distraction. Individuals who feel positive affect at a particular time appear to temporarily display increased creativity, cognitive flexibility, and easier access to positive information. This state may facilitate their inclination to see negative or neutral situations in a favourable light and direct their attention to more positive stimuli, thus encouraging fresh feelings of positive affect. The experience of negative affect and engagement in narrow–minded affect regulation strategies, including rumination and expressive suppression, could also reinforce each other. Individuals who feel negative affect at a particular time appear to momentarily tend to focus their attention on negative stimuli, and restrict their behaviours to those required to cope with them. In such a state, there would be a notable increase in the inclination to brood on previous or future problems, as well as to overestimate, then avoid, the risk associated with expressing affect to others. These mechanisms may foster fresh experiences of negative affect.

Importantly, these hypotheses on the interplay between affect and affect regulation strategies, which are based on theories pertaining to the broad impact of affective experiences on cognition and behaviour (Bowers, 1991; Fredrickson, 1998, 2001), differ from hypotheses based entirely on recent models of affect regulation (Gross, 2015; Kuppens, Oravecz, et al., 2010b). For instance, as set out above, the extended process model of emotion regulation (Gross, 2015) claims that affect only influences affect regulation strategies when affective experiences are negatively evaluated, triggering the urge to use strategies to produce experiential changes. As the most negatively evaluated affective experiences appear to be high in negative affect and/or low in positive affect (Riediger, Schmiedek, Wagner, & Lindenberger, 2009), Gross (2015)'s model mainly predicts that a momentary increase in a negative affect experience promotes the use of various affect regulation strategies which, in turn, decreases the initial feeling of negative affect. Conversely, an increase in a positive affect experience reduces the use of affect regulation strategies, resulting in a subsequent decrease in the initial feeling of positive affect. Thus, the extended process model of emotion regulation (Gross, 2015) mainly predicts negative rather than positive feedback loops between affect and affect regulation strategies.

We acknowledge that these negative feedback loops form part of affect regulation dynamics. For instance, although their implementation is likely to depend on prior positive affect experiences as mentioned previously, the use of distraction and positive reappraisal is primarily thought to arise in a context where a negative affect is felt (Folkman, 1997; Van Dillen & Koole, 2007). Moreover, implementing these strategies can sometimes reduce this initial negative feeling (e.g. Nolen–Hoeksema & Morrow, 1993). Nevertheless, our main hypotheses suggest that greater attention should be paid to possible positive feedback loops between affect and affect regulation strategies.

Note that the positive feedback loops between positive affect and broad–minded strategies are sometimes referred to hereafter as virtuous cycles, and those between negative affect and narrow–minded strategies as vicious cycles. Despite their tempting simplicity, these terms must be used with caution, because individuals are not always motivated by feeling more positive affect and less negative affect (Riediger et al., 2009). Moreover, under some circumstances (e.g. the occurrence of frequent unpleasant events), an increased vigilance and negative affectivity can be adaptive.

Extraversion and affect regulation dynamics

To the best of our knowledge, there are no available findings on the impact of extraversion on reciprocal relationships between affect and affect regulation strategies. Nevertheless, some information leads us to hypothesize that a higher level of extraversion is associated with more intensely positive feedback loops between positive affect and broad–minded strategies. First, a pleasant stimulation of the same magnitude triggers a more intense experience of positive affect among extraverted individuals than among introverted ones (Larsen & Ketelaar, 1991). As the use of a broad–minded affect regulation strategy can be perceived of as a pleasant internal stimulation, the enhanced positive affect reactivity displayed by extraverted individuals may lead them to feel more intense positive affect than their introverted counterparts, for the same intensity of strategy use. Second, a higher level of extraversion is associated with greater motivation to seek pleasure and rewards (Smits & Boeck, 2006). Thus, the cognitive and behavioural broadening mechanisms that follow the experience of an initial positive affect may particularly increase the implementation of affect regulation strategies that promote feelings of positive affect among the most extraverted individuals. Initial findings support this hypothesis, suggesting that a higher level of extraversion is accompanied by a greater inclination to continue to seek pleasure after positive affect induction (Hirsh, Guindon, Morisano, & Peterson, 2010).

Neuroticism and affect regulation dynamics

Although the relationship between neuroticism and the dynamics of affective feelings starts to be understood (Bringmann et al., 2016; Suls, Green, & Hillis, 1998), the impact of this trait on the interplay between affect and affect regulation strategies has rarely been analysed. We can nevertheless hypothesize that the positive feedback loops between the experience of negative affect and the use of narrow–minded affect regulation strategies are most intensely displayed by the most neurotic individuals, compared to the less neurotic ones. As the use of a narrow–minded strategy can be perceived of as an unpleasant internal stimulation, and as neurotic individuals react more intensely to negative stimuli than their less neurotic counterparts (Larsen & Ketelaar, 1991), the former may feel more intense negative affect than the latter for the same intensity of strategy use. Furthermore, the stronger inclination to focus on negative stimuli exhibited by neurotic individuals compared with less neurotic ones (Smits & Boeck, 2006) may mean that a higher level of neuroticism is accompanied by a more intense use of narrow–minded strategies in reaction to an initial experience of negative affect. Finally, given the deterioration in affect regulation skills associated with increased neuroticism (Yoon, Maltby, & Joormann, 2013), we can also predict that components of the virtuous cycles studied here have less of an inhibitory effect on vicious cycle components among neurotic individuals than among less neurotic ones.

Affect

We assessed participants’ affective experience with the 12 kinds of affect described by the 12–point affect circumplex model (Yik, Russell, & Steiger, 2011). This model identifies five kinds of positive affect and five kinds of negative affect according to their arousal level (highly activated, activated, neither activated nor deactivated, deactivated and highly deactivated positive and negative affect). This model also contains two neutrally valenced affects: one that is activated, and one that is deactivated. Examples of the statements that participants had to summarize for each type of affect are set out in an openly accessible file available from https://osf.io/project/s3chz/files/osfstorage/58920befb83f6901f40f7ec4/. At each assessment occasion, participants were asked to indicate the intensity with which they currently experienced these 12 kinds of affect, on a Likert scale ranging from 1 (not at all) to 5 (a lot). As explained earlier, several findings had suggested that affective valence is important to consider in affect regulation dynamics. We therefore ignored the two neutrally valenced affects. We then conducted a multilevel confirmatory factor analysis on the 10 remaining items (i.e. the five positively valenced and five negatively valenced affects) to test the extent to which a two within–participants and two between–participants factor structure explained the covariations among these items. Results indicated that the fit of this model was satisfactory (CFI = .97, TLI = .95, RMSEA = .04). Consequently, we calculated an indicator of positive affect and an indicator of negative affect by averaging the scores for the five degrees of positive affect and the scores for the five kinds of negative affect. To estimate the reliability of these indicators, we computed the two coefficients devised for multilevel data by Bolger and Laurenceau (2013): omega (ω) and reliability of change (RC). Results were ω = .732 and RC = .730 for our indicator of positive affect, and ω = .760 and RC = .756 for our indicator of negative affect. The reliability of both indicators was thus satisfactory.

Affect regulation strategies

We assessed participants’ implementation of affect regulation strategies with five strategies: appreciation, positive reappraisal, distraction, expressive suppression, and rumination. Our protocol also contained the assessment of two other strategies (i.e. social sharing of affect and problem–focused coping). However, these two strategies were not examined in the present study, as previous research (e.g. Zech & Rimé, 2005), as well as initial analyses on comparable data, had led us to hypothesize that their short–term influence on affective valence, over and above the effect of the five other strategies, would be negligible. Examples of the statements that participants had to summarize for each affect regulation strategy are set out in an openly accessible file available from https://osf.io/project/s3chz/files/osfstorage/58920befb83f6901f40f7ec4/. At each assessment, participants were asked to report the intensity with which they had implemented each strategy since the last assessment, on a Likert scale ranging from 1 (not at all) to 5 (a lot). As with previous experience sampling studies attempting to examine a variety of strategies (e.g. Brans et al., 2013), the only drawback of measuring strategies with solely one item is that it makes it impossible to assess the factorial validity of such measures. Nevertheless, initial findings based on these items suggest that they have good construct validity (Pavani et al., 2015).

Extraversion and neuroticism

Extraversion and neuroticism were assessed with two 60–item subscales of a French translation of the International Personality Item Pool (IPIP: Goldberg, 1999). Participants were asked to rate the degree to which they agreed with each of the statements in these subscales, on a Likert scale ranging from 1 (not at all) to 5 (a lot). In the IPIP, extraversion is evaluated through items organized around six facets (i.e. activity, assertiveness, warmth, gregariousness, positive emotions, and excitement seeking), while neuroticism is measured through items organized around six other facets (i.e. anxiety, depression, anger, vulnerability to stress, social shyness, and impulsivity). Although this questionnaire has not yet been validated, the reliability of the indicators of both extraversion (α = .90) and neuroticism (α = .92) was satisfactory.

The network of affect regulation dynamics at the sample level

Our first hypothesis was that the network of affect regulation dynamics would be characterized at the sample level by positive feedback loops between positive affect and appreciation, positive affect and positive reappraisal, positive affect and distraction, negative affect and expressive suppression, and negative affect and rumination.

We used multilevel vector autoregressive models to test this hypothesis. More precisely, we computed seven models. Models 1 and 2 contained experience of positive affect and experience of negative affect at t − 1 as predictor variables, as well as implementation of the five strategies of interest between t − 1 and t. The outcome variable was experience of positive affect at t for Model 1 and experience of negative affect at t for Model 2. Models 3, 4, 5, 6 and 7 all contained the same predictor variables (experience of positive and negative affect at t − 1, and implementation of the five strategies between t − 2 and t − 1), but the outcome variable was the implementation, between t − 1 and t, of appreciation, positive reappraisal, distraction, expressive suppression, or rumination.

Results are set out in Table 2 and Figure 1. To make this figure easier to understand and interpret, we concealed effects that were not significant, or whose size did not exceed the absolute value of β = .10. ⁶ These effects were concealed but not removed, even though their removal is recommended in certain network approaches (e.g. Costantini et al., 2015). The main point of removing the spurious edges of a network is to refine the estimation of certain characteristics of that network (e.g. betweenness centrality, clustering coefficients). However, we did not need to estimate these characteristics to test our specific hypotheses.

OPEN IN VIEWER

Table 2

Regression coefficients and statistical significance of the reciprocal influences between affective experience and affect regulation strategy implementation at the sample level

	PAt	NAt	Appt	PRt	Dist	Supt	Rumt
PA t–1	.30 ** [.26, .33]	.06 ** [.03, .09]	.18 ** [.14, .23]	.02 [−.03, .06]	.12 ** [.07, .16]	.05 * [.01, .09]	−.02 [−.06, .02]
NA t–1	.01 [−.02, .04]	.38 ** [.35, .41]	.01 [−.03, .05]	.02 [−.02, .06]	.10 ** [.06, .14]	.06 ** [.02, .10]	.10 ** [.06, .14]
App t–1	.28 ** [.25, .30]	−.19 ** [−.21, −.16]	.20 ** [.17, .24]	−.01 [−.04, .02]	−.03 [−.05, 0]	−.04 * [−.07, −.01]	−.01 [−.04, .02]
PR t–1	.12 ** [.11, .15]	−.10 ** [−.12, −.07]	.01 [−.02, .04]	.23 ** [.20, .26]	.02 [−.01, .05]	.01 [−.02, .04]	−.01 [−.04, .02]
Dis t–1	.05 ** [.02, .07]	−.01 [−.04, .01]	0 [−.03, .04]	.05 ** [.02, .08]	.20 ** [.17, .23]	.02 [−.01, .05]	−.02 [−.06, .01]
Sup t–1	−.02 [−.05, 0]	.07 ** [.05, .10]	−.03 [−.07, 0]	.06 ** [.03, .09]	.06 ** [.03, .09]	.25 ** [.22, .29]	.04 ** [.01, .08]
Rum t–1	−.11 ** [−.13, −.08]	.21 ** [.18, .24]	−.03 [−.06, .01]	.03 [0, .06]	.04 ** [.01, .07]	.04 * [.01, .07]	.26 ** [.23, .29]

**

p < .01,

*

p < .05.

Note: PA, positive affect; NA, negative affect; App, appreciation; PR, positive reappraisal; Dis, distraction; Sup, expressive suppression; Rum, rumination. Predictor variables are displayed in rows and outcome variables in columns. The numbers in square brackets correspond to the lower and upper limits of the 95% confidence interval for each effect. Regression coefficients are standardized.

OPEN IN VIEWER

Figure 1

The network made up of the reciprocal influences between affect and affect regulation strategies at the sample level. Note. PA = positive affect; NA = negative affect; App = appreciation; PR = positive reappraisal; Dis = distraction; Sup = expressive suppression; Rum = rumination. Arrows correspond to the effects that exceeded the significance threshold (p < .05) and whose size exceeded the absolute value of β = .10. Solid lines correspond to positive effects, dashed lines to negative effects.

Results confirmed most of our predictions. First, we clearly observed a virtuous cycle between positive affect and appreciation, and a vicious cycle between negative affect and rumination. More specifically, positive affect felt at a particular time significantly promoted engagement in appreciation (β = .18, p < .001) within the space of a few hours. By the same token, appreciating the present more within this time interval significantly increased the positive affect experienced afterwards (β = .28, p < .001). Likewise, within a few hours of experiencing more negative affect, individuals ruminated more (β = .10, p < .001), and engaging more in rumination within this time interval increased their subsequent experience of negative affect (β = .21, p < .001).

Second, when we focused on the significance of the effects we had examined, we observed another virtuous cycle, between positive affect and distraction, as well as another vicious cycle between negative affect and expressive suppression. Although these effects were significant, the size of some of them was too small for us to interpret them as contributing to the occurrence of such positive feedback loops. More specifically, small effect sizes were found for the effect of implementing distraction on the experience of positive affect (β = .05, p < .001), the influence of engaging in expressive suppression on the experience of negative affect (β = .07, p < .001), and the influence of negative affect feelings on the subsequent use of this strategy (β = .06, p < .01).

Finally, contrary to what we predicted, whereas implementing more positive reappraisal increased feelings of positive affect within a few hours as expected (β = .12, p < .001), feelings of positive affect did not influence subsequent implementation of this strategy (β = .02, p > .05).

A more precise examination of Table 2 and Figure 1 reveals three more interesting results. First, each component of the network displayed self–loops. These autoregressive effects systematically had the largest sizes, reflecting so–called affective inertia (Kuppens, Allen, et al., 2010a) as regards our indicators of positive and negative affect experience. Affective inertia provides information about the rapidity of change in the relevant affect during a given interval. High inertia (i.e. regression coefficient very close to 1) reflects slow change (a form of rigidity) and resistance to change, whereas low inertia (i.e. regression coefficient very close to 0) reflects rapid change (a form of flexibility, susceptibility to change). In the present research, modest inertia was observed not only in affective experience, but also in affect regulation strategy implementation, which has received less attention from researchers.

Second, within a few hours of experiencing more negative affect, individuals tended to implement coping strategies (i.e. strategies designed to regulate negative affect experiences) more intensively. Feeling this type of affect not only enhanced subsequent use of rumination and expressive suppression, but also promoted the implementation of distraction (β = .10, p < .001). This suggests that affect regulation dynamics are partly driven by the motivation to change unwanted feelings after their elicitation, as posited by the most prominent models of affect regulation (e.g. Gross, 2015).

Third, strategies whose implementation is known to exert the strongest influence on affective life impacted feelings of both positive and negative affect. More intensive engagement in appreciation and positive reappraisal was associated not only with enhanced subsequent experience of positive affect, but also with reduced experiences of negative affect (βs = −.19 and −.10, p < .001). Likewise, ruminating more was followed not only by a heightened experience of negative affect, but also by a decreased feeling of positive affect (β = −.11, p < .001).

In summary, regarding our hypothesis, the dynamics of affect regulation at the sample level appeared to be characterized by two systems. Negative affect and narrow–minded strategies reinforced each other, even though only the vicious cycle involving rumination had a large effect size. Likewise, virtuous cycles emerged between positive affect and some broad–minded strategies, but only the cycle featuring appreciation had a large size, the one involving distraction having a smaller one. As regards the positive feedback loop between positive affect and positive reappraisal that we expected to observe at the sample level, the following analyses enabled us to determine whether it could be identified among some individuals, or did not appear at all.

Variations in the network of affect regulation dynamics related to extraversion

Our second hypothesis was that the network of affect regulation dynamics varies according to extraversion. More specifically, we predicted that these variations would mainly be manifested in the positive feedback loops between experience of positive affect and use of broad–minded strategies. These loops would become more intense as the level of extraversion increased.

To test this hypothesis, we calculated the same seven models as for the first hypothesis, adding the multiplicative effect of extraversion to their predictor variables. As neuroticism and extraversion were correlated in our study, we also included the additive effect of neuroticism, to neutralize its potential confounding effect. First, these models enabled us to determine which connection in the network was significantly moderated by extraversion. Second, the coefficients estimated by these models were used to predict the distinctive networks that fictitious individuals with either low extraversion (1.5 standard deviations below the mean) or high extraversion (1.5 standard deviations above the mean) would display. Results for these two networks are set out in Table 3 and on the top of Figure 2. The effects that were significantly moderated by extraversion are shown in bold in Table 3, and in black in the two networks on the top of Figure 2. In these networks, we chose to represent some effects whose size was below the absolute value of β = .10 but above the absolute value of β = .05, as they were significantly moderated by extraversion, and thus relevant to our analyses. Table 3 also provides the AICs of the seven models we calculated, with and without the inclusion of the multiplicative effect of extraversion.

OPEN IN VIEWER

Table 3

Regression coefficients of the reciprocal influences between affective experience and affect regulation strategy implementation among individuals with low versus high extraversion scores

	Low extraversion
	PAt	NAt	Appt	PRt	Dist	Supt	Rumt
PA t–1	.38	.04	.18	−.06	.06	−.01	.02
NA t–1	.05	.36	.05	.02	.09	.05	.10
App t–1	.21	−.17	.24	.05	.02	.03	−.02
PR t–1	.10	−.10	−.01	.20	.02	−.04	.01
Dis t–1	.04	−.05	.04	.01	.22	.01	.01
Sup t–1	.05	.04	−.05	.04	.11	.35	.11
Rum t–1	−.09	.21	−.05	.11	.10	.13	.26
	High extraversion
	PAt	NAt	Appt	PRt	Dist	Supt	Rumt
PA t–1	.21	.08	.19	.08	.16	.09	−.06
NA t–1	−.05	.40	−.03	.01	.11	.07	.11
App t–1	.34	−.20	.17	−.07	−.07	−.09	.01
PR t–1	.14	−.09	.02	.26	.03	.05	−.03
Dis t–1	.06	.02	−.03	.08	.17	.02	−.05
Sup t–1	−.10	.10	0	.07	.02	.16	−.01
Rum t–1	−.13	.22	−.01	−.02	−.02	−.05	.25
AIC Pop	7042	7504	9721	9241	8784	8813	8940
AIC E	6999	7500	9730	9234	8781	8782	8931

Note: PA, positive affect; NA, negative affect; App, appreciation; PR, positive reappraisal; Dis, distraction; Sup, expressive suppression; Rum, rumination. Low extraversion corresponds to a fictitious individual scoring 1.5 standard deviations below the mean on this trait, and high extraversion to a fictitious individual scoring 1.5 standard deviations above the mean on this trait. Predictor variables are displayed in rows and outcome variables in columns. Regression coefficients are standardized. The coefficients in bold represent the ones that were significantly moderated by extraversion. AIC Pop represents the AICs of the seven models that did not include extraversion in their predictors and AIC E the AICs of the seven models that did include it.

OPEN IN VIEWER

Figure 2

The network made up of the reciprocal influences between affect and affect regulation strategies among individuals scoring low or high on extraversion (on the top) and neuroticism (at the bottom). Note. PA = positive affect; NA = negative affect; App = appreciation; PR = positive reappraisal; Dis = distraction; Sup = expressive suppression; Rum = rumination. Arrows are effect sizes above the absolute value of β = .10, or between the absolute values of β = .06 and β = .10, if they were significantly moderated by extraversion for the two networks on the top of this figure, and by neuroticism for the two networks at the bottom of it. Solid lines correspond to positive effects, dashed lines to negative effects. Black arrows refer to the effects that were significantly moderated by extraversion for the two networks on the top of this figure or by neuroticism for the two networks at the bottom of it, grey arrows to the effects that were not.

Results were generally consistent with our hypothesis. Extraverted and introverted individuals differed in the form of their affect regulation dynamics. The AICs in Table 3 notably reveal that six of the seven models we calculated had a better fit when they contained the multiplicative effect of extraversion. Extraverted individuals were characterized by low positive affective inertia (β = .21), indicating rapid change in this affect. This type of low inertia reflects the tendency of the most extraverted individuals to display greater flexibility in their positive affective lives, possibly manifested in the rapid emergence of the virtuous cycles we explored in this study.

Two observations are consistent with this idea. First, extraverted individuals were the only ones to exhibit a positive feedback loop between positive affect and positive reappraisal, in addition to the virtuous cycle between positive affect and appreciation we had already identified at the sample level. Specifically, extraverted individuals were the only participants whose use of positive reappraisal was prompted by a more intense previous experience of positive affect (β = .08), even though the size of this effect was small. Second, the virtuous cycle we had already identified at the sample level tended to be more intense among these participants. In particular, appreciating the present had a greater influence on the feelings of positive affect of the most extraverted individuals (β = .34) than on the feelings of positive affect of their most introverted counterparts (β = .21).

In contrast to their extraverted peers, the most introverted individuals displayed high positive affective inertia within the time interval we chose (β = .38). Their positive affective lives therefore appeared to be more rigid—a rigidity that may have been expressed in their tendency to go through the virtuous cycles of interest more slowly. In particular, introverted individuals failed to display a positive feedback loop between positive affect and positive reappraisal, as they tended to implement less positive reappraisal within a few hours of experiencing more positive affect (β = −.06). Furthermore, as set out earlier, the use of appreciation had a smaller impact on their feelings of positive affect than on the feelings of positive affect of their extraverted counterparts.

Albeit less closely related to our hypotheses, a final interesting observation regarding the positive affect feelings of introverted individuals concerns the particular relationship between this type of feeling and their engagement in expressive suppression. More specifically, suppressing affective expression for a few hours did not reduce their subsequent feelings of positive affect (β = .05), contrary to what we observed among their more extraverted peers (β = −.10). If using expressive suppression does not noticeably impact affective experiences in the short term among introverted individuals, and if the main function of affect regulation strategies is to modify affective experiences, then introverted individuals may be less motivated to modulate expressive suppression (i.e. change the intensity with which this strategy is used) than extraverted individuals. This reduced inclination to control this strategy may explain the high expressive suppression inertia observed among introverted individuals (β = .35), compared with their extraverted peers (β = .16).

Taken together, our results suggest that considering extraversion is important for understanding affect regulation dynamics more fully. As expected, the variations in the network of affect regulation dynamics brought about by this trait mainly concerned the positive feedback loops between the experience of positive affect and the use of broad–minded strategies. Extraverted individuals displayed more–and more intense–virtuous cycles than their more introverted peers.

Variations in the dynamic network of affect regulation related to neuroticism

Our third hypothesis was that the network of affect regulation dynamics varies according to neuroticism. These variations would mostly be identified in the positive feedback loops between experience of negative affect and implementation of narrow–minded strategies. These loops would become stronger as neuroticism increased, and their components (i.e. negative affect, expressive suppression, rumination) would become less inhibited by the components of the virtuous cycles.

We tested this hypothesis by performing the same seven models as before, adding the multiplicative effect of neuroticism and the additive effect of extraversion to their predictor variables, and using the coefficients estimated by these models to predict the distinctive networks that fictitious individuals with low (1.5 standard deviations below the mean) and high (1.5 standard deviations above the mean) neuroticism would exhibit. The results are set out in Table 4 and in the two networks at the bottom of Figure 2. The effects that were significantly moderated by neuroticism are shown in bold in Table 4, and in black in the two networks at the bottom of Figure 2. Table 4 also contained the AICs of the seven models constituting the network with and without the inclusion of the multiplicative effect of neuroticism.

OPEN IN VIEWER

Table 4

Regression coefficients of the reciprocal influences between affective experience and affect regulation strategy implementation among individuals with low versus high neuroticism scores

	Low neuroticism
	PAt	NAt	Appt	PRt	Dist	Supt	Rumt
PA t–1	.22	.03	.17	0	.10	.02	−.10
NA t–1	−.07	.33	−.01	0	.10	.04	.04
App t–1	.34	−.21	.20	−.07	−.02	−.06	0
PR t–1	.15	−.15	0	.26	.05	.05	0
Dis t–1	.07	.01	−.02	.13	.17	.01	−.07
Sup t–1	−.08	.13	−.05	.01	.01	.22	0
Rum t–1	−.20	.25	−.01	.01	.02	−.01	.28
	High neuroticism
	PAt	NAt	Appt	PRt	Dist	Supt	Rumt
PA t–1	.34	.09	.20	.03	.13	.07	.05
NA t–1	.05	.42	.04	.03	.11	.08	.15
App t–1	.22	−.17	.21	.05	−.02	−.01	−.02
PR t–1	.10	−.04	.01	.21	0	−.02	−.03
Dis t–1	.03	−.03	.03	−.02	.22	.02	.02
Sup t–1	.03	.01	−.01	.11	.12	.28	.10
Rum t–1	−.04	.18	−.05	.06	.05	.07	.23
AIC Pop	7042	7504	9721	9241	8784	8813	8940
AIC N	6978	7480	9734	9238	8791	8819	8927

Note: PA, positive affect; NA, negative affect; App, appreciation; PR, positive reappraisal; Dis, distraction; Sup, expressive suppression; Rum, rumination. Low neuroticism corresponds to a fictitious individual scoring 1.5 standard deviations below the mean on this trait and high neuroticism to a fictitious individual scoring 1.5 standard deviations above the mean on this trait. Predictor variables are displayed in rows and outcome variables in columns. Regression coefficients are standardized. Coefficients that were significantly moderated by neuroticism are shown in bold. AIC Pop represents the AICs of the seven models that did not include neuroticism in their predictors and AIC N the AICs of the seven models that did include it.

Results did not entirely support our hypothesis. Differences emerged in the affect regulation dynamics of the most and the least neurotic individuals. In particular, the AICs in Table 4 reveal that four of the seven models constituting the network had a better fit when they contained the multiplicative effect of neuroticism. However, the neuroticism–related differences we observed did not exactly follow the predicted pattern.

One of the main characteristics of the affect regulation dynamics of the most neurotic individuals was the rigidity of their affective feelings, that is, the resistance of these feelings to change during the time interval we examined. First, this rigidity manifested itself in a high level of negative affective inertia (β = .42). This resistance to change may explain why, as compared with what occurs among their less neurotic peers, the engagement in rumination by the most neurotic individuals had a reduced influence on their negative affect experiences (β = .18), contrary to what we had predicted. Therefore, even though the most neurotic individuals were particularly inclined to ruminate after feeling negative affect (β = .15), the reduced effect of rumination on this negative affect meant that there was no intensification in the vicious cycle between negative affect and rumination among the most neurotic individuals. Similarly, the most neurotic individuals showed no intensification of the vicious cycle between negative affect and expressive suppression, owing to the resistance of their negative affect feelings to change brought about by previous use of this strategy (β = .01). Interestingly, their negative affect experiences were also resistant to the intensity with which they used positive reappraisal (β = −.04), confirming our hypothesis that, as neuroticism increases, some vicious cycle components are less inhibited by virtuous cycle components.

Rigidity was also observed in the positive affect feelings of the most neurotic individuals, as suggested by their high positive affective inertia (β = .34). Once again, their positive affect feelings were characterized by low dependence on the strategies they had previously used. More specifically, compared with what we observed among their less neurotic peers, these feelings were quite resistant to the intensity with which they engaged in appreciation (β = .22), expressive suppression (β = .03), and rumination (β = −.04).

Contrary to their more neurotic counterparts, the least neurotic participants displayed low positive (βs = .22) and low negative (βs = .33) affective inertia. This may have reflected their tendency to experience intense virtuous and vicious cycles within the short time interval we chose, mainly owing to the dependence of their affective experiences on the strategies they implemented.

Finally, the specific tendency of positive affect experiences to inhibit subsequent engagement in rumination among the least neurotic individuals (β = −.10) was an interesting finding in relation to our hypothesis. In addition to the enhanced influence of positive reappraisal on feelings of negative affect (β = −.15), it suggests that some of the vicious cycles were particularly inhibited by components of the virtuous cycles among these individuals.

In summary, interindividual differences in neuroticism were associated with several variations in the network of affect regulation dynamics. Contrary to what we had predicted, the positive feedback loops between experience of negative affect and implementation of expressive suppression and rumination did not intensify as the level of neuroticism increased. This may have been due to the increased rigidity (i.e. resistance to change) that the most neurotic individuals displayed in their feelings of negative affect, compared with their less neurotic peers. Nevertheless, the components of these vicious cycles tended to be less inhibited by virtuous cycle components as the level of neuroticism increased, in line with our prediction, even though neuroticism had few significant moderating effects on these inhibitory influences.