The Future of Emotion Regulation Research

How are demographic characteristics, personality facets, and psychological processes examined?

Several investigations over the past decade have examined the moderating effects of various demographic characteristics (e.g., age, gender, ethnicity, life stressors), personality facets (e.g., neuroticism), and dispositional and state-level psychological processes (e.g., trait affect, habitual use of strategies, mood) on the regulatory process. These include correlational (e.g., John & Gross, 2004; Kokkonen & Pulkkinen, 2001; Nolen- Hoeksema & Aldao, 2011; Soto, Perez, Kim, Lee, & Minnick, 2011; Zlomke & Hahn, 2010) as well as experimental studies (e.g., age: Larcom & Isaacowitz, 2009; McRae, Ciesielski, & Gross, 2011; Winecoff, LaBar, Madden, Cabeza, & Huettel, 2010; gender: McRae, Ochsner, Mauss, Gabrieli, & Gross, 2008; ethnicity: Gross & John, 2003; Roberts, Levenson, & Gross, 2008; trait affect: Dalgleish, Yiend, Schweizer, & Dunn, 2009; life stressors: O’Mara, McNulty, & Karney, 2011; Shallcross, Troy, Boland, & Mauss, 2010; and habitual use of emotion regulation strategies: Burns, Quartana, & Bruehl, 2007; Geraerts, Merckelbach, Jelicic, & Smeets, 2006; Gilliam et al., 2010; Ray et al., 2005; Wolgast, Lundh, & Viborg, 2011). However, most of these investigations have examined only one or two person-level characteristics at a time, therefore falling short of modeling the complex interactions among the various aspects of demographic characteristics, personality facets, and psychological processes. Moreover, the number of studies examining the same person-level characteristics is quite small, which has limited the scope of between-studies comparisons.

To address this limitation, it will first be important to recruit samples that are large enough to have sufficient power for the simultaneous examination of multiple person-level variables. Because there might be a practical limitation to how many individuals can participate in a laboratory-based assessment (relative to survey studies), it will be essential to ensure that there is a sufficient number of participants at each level of the moderator variable. This can be achieved by oversampling for those characteristics that have a low base rate in the population. Second, it will be critical to test the interactions among person-level characteristics. In this respect, a recent study showed that the habitual implementation of certain regulation strategies was a function of the interaction between age and gender (Nolen-Hoeksema & Aldao, 2011). For example, older women reported using suppression to a greater extent than did young and middle-age women, and there were no age differences within men. These findings underscore the necessity to simultaneously use multiple dimensions to characterize individuals and their regulation repertoire. Third, it will be particularly important to adopt a developmental approach to comprehensively model changes over time in children (e.g., Cicchetti, Ackerman, & Izard, 1995; Cole, Martin, & Dennis, 2004) and older adults (e.g., Isaacowitz & Blanchard-Fields, 2012).

How is psychopathology assessed?

Many of the early studies on emotion regulation in psychopathology included one clinical group and omitted a comparison group of healthy controls (e.g., Borkovec & Hu, 1990; Borkovec, Lyonfields, Wiser, & Deihl, 1993; Campbell-Sills, Barlow, Brown, & Hofmann, 2006; Levitt et al., 2004; Liverant, Brown, Barlow, & Roemer, 2008). Needless to say, this constituted a major limitation that prevented the identification of the pathological processes in emotion regulation in psychopathology. Fortunately, more recent investigations have included a control group. However, they have mostly examined only one psychopathology group, which has precluded the differentiation of disorder-specific versus transdiagnostic deficits (Harvey, Watkins, Mansell, & Shafran, 2004). Examples include reappraisal in depression (e.g., Johnstone, van Reekum, Urry, Kalin, & Davidson, 2007), anxiety (e.g., Campbell-Sills et al., 2011), and borderline personality disorder (e.g., Schulze et al., 2011); suppression in depression (e.g., Dalgleish & Yiend, 2006) and anxiety (e.g., Kingsep & Page, 2010); suppression and reappraisal in binge eating disorder (e.g., Svaldi, Caffier, & Tuschen-Caffier, 2010); and acceptance and reappraisal in anxiety (e.g., Aldao & Mennin, 2012). The lack of multiple pathological groups is an issue that transcends the study of emotion regulation, as it applies to most studies of psychological processes that include only one psychopathology group. However, given arguments that emotion regulation is a transdiagnostic factor (e.g., Aldao & Nolen-Hoeksema, 2010; Kring & Sloan, 2010), it is particularly important that investigators conduct the type of studies that will allow for the identification of which emotion regulation deficits are broad markers of psychopathology and which are specific to the particular phenomenology of individual disorders. As a related point, a vast majority of the laboratory-based studies have been primarily conducted on anxiety and mood disorders, further limiting our understanding of emotion regulation across disorders (for a review of a similar asymmetry in correlational studies, see Aldao et al., 2010).

The most evident solution to the issues delineated above is the inclusion of more than one psychopathology group. However, this poses a series of additional complications: (a) the need for larger samples that meet specific inclusion and exclusion criteria and (b) the difficulties modeling comorbidity between discrete clinical groups. For example, in a study that includes two psychopathology groups (diagnoses of social anxiety disorder and major depressive disorder, respectively), what is the best way to categorize those participants who have both diagnoses? Excluding such participants altogether would limit the external validity of the investigation; conversely, creating a separate “comorbid” group might obscure between-group differences. This point resonates with the recent interest in dimensional models of psychopathology (e.g., Clark, 2005; Cuthbert, 2005; Krueger, Markon, Patrick, & Iacono, 2005; Watson, 2005; Widiger & Samuel, 2005; see Research Domain Criteria: Insel et al., 2010). Therefore, a necessary next step in the study of emotion regulation in psychopathology will involve the dimensional measurement of symptoms (e.g., self-report questionnaires; diagnostic interviews such as the Anxiety Disorders Interview Schedule, which includes a dimensional scoring system; DiNardo, Brown, & Barlow, 1994). From a statistical standpoint, the use of structural equation modeling (Arbuckle, 2007) could provide an advantage over the more commonly used linear regression, as it would offer different options to model covariance structures among symptoms of disorders that have high diagnostic overlap (e.g., covariance among error terms, latent factors).

What is the ecological and external validity of the emotion-eliciting stimuli and the paradigms?

Although a good number of the emotion-eliciting stimuli have been shown to reliably elicit emotion and, in many cases, have even been standardized (e.g., pictures: Lang et al., 2008; film clips: Gross & Levenson, 1995; Rottenberg et al., 2007), their use has led to limitations in the external and ecological validity of the experimental manipulations. This is particularly the case for those stimuli that result in a relatively passive viewing by participants, as they induce a state that could be conceptualized as the opposite of the active engagement with the environment that emotions are believed to facilitate (e.g., Frijda, 1986). Indeed, watching television could be described as a behavior serving an avoidant function in certain contexts (Endler & Parker, 1990), and it is common to hear from patients that they sit in front of the television for hours in order to escape the experience of difficult emotions.

Relatedly, only a handful of investigations have measured the implementation of strategies in response to naturally occurring stimuli via the use of ecological momentary assessments (EMA; e.g., Moberly & Watkins, 2008; Shahar & Herr, 2011; Silk, Steinberg, & Morris, 2003). Such naturalistic interactions are particularly informative in the context of psychopathology because they can allow researchers to identify the extent to which regulation follows the same patterns of rigidity as the emotional experiences (e.g., Rottenberg et al., 2005). For example, an individual who suffers from social anxiety disorder in the context of dating situations might be able to perform relatively well in a laboratory paradigm consisting of an impromptu speech, a frequent manipulation used to elicit fears of negative evaluation (e.g., Beidel, Turner, Jacob, & Cooley, 1989). Conversely, examining the day-to-day interactions of this individual with peers and potential romantic partners might provide investigators with a much richer picture of his or her pattern of emotion dysregulation.

It will be important that the field reconsiders the selection of emotion-eliciting stimuli to favor those that produce more complex emotional and motivational states that are idiographically relevant and can therefore enhance the external and ecological validity of the experimental approach. One such example is the presentation of film clips in which participants themselves are discussing emotional topics (e.g., Rottenberg et al., 2005). However, as discussed previously, there is an ingrained passivity in the watching of stimuli that likely dampens the extent to which they elicit motivational engagement. Therefore, it will be important for idiographic stimuli to stimulate motivational states by requiring active engagement from participants. One example of such stimuli consists of worry and rumination inductions in which participants are guided to engage in worry and ruminative processes (e.g., Nolen- Hoeksema, Wisco, & Lyubomirsky, 2008). In addition, a relatively small number of studies have examined emotion regulation strategies in relation to goal-oriented tasks, such as playing video games (e.g., Tamir et al., 2008), writing emotionally evocative personal stories (e.g., Dalgleish et al., 2009), and tolerating distress (e.g., Feldner, Zvolensky, Eifert, & Spira, 2003; Szasz, Szentagotai, & Hofmann, 2011). Such paradigms have been used largely with normative samples; therefore, an important next step will be the use of these paradigms in studies with participants who experience clinical levels of psychopathology. A related point that might allow researchers to better capture contextual variability is the modeling of stimuli as random rather than fixed factors (for a comprehensive review of this issue, see Judd, Westfall, & Kenny, 2012).

In addition, it will be important to increase the variability of the settings in which the assessments take place by using EMA designs. In addition, it will be informative to experimentally induce the implementation of regulation strategies via EMA. Such an approach could be extremely useful in helping researchers to teach participants how to implement strategies online and might eventually result in the development of trainings that could be used as stand-alone interventions or as augmentations for traditional psychotherapeutic approaches.

What about the interpersonal context?

One of the most important contextual factors that influence human behavior is the presence of other individuals. The process of emotion regulation is no exception to this rule. From a crying toddler seeking a parent’s hug to an adult requesting advice from a friend, people constantly rely on others for the regulation of their affective states (e.g., Coan, Schaefer, & Davidson, 2006; Lakey & Orehek, 2011; Marroquín, 2011; Rime, 2009). Similarly, the interpersonal context constitutes one of the main spheres of dysfunction across psychological disorders (e.g., Cain et al., 2011; Erickson & Newman, 2007; Joiner, Van Orden, Witte, Selby, & Rudd, 2009; Linehan, 1993; Lynch, Trost, Salsman, & Linehan, 2007; Marroquín, 2011). Therefore, such interpersonal processes are vital to the understanding of the deficits in emotion regulation that characterize psychopathology. However, the empirical work is still in its infancy, and much remains to be understood about the basic and pathological mechanisms by which interpersonal regulation takes place. One exception is the process of co-rumination, conceptualized as the tendency to excessively discuss problems within a relationship (e.g., Rose, Carlson, & Waller, 2007). Yet, this process has been examined in only a handful of studies, and much remains to be understood regarding its relationship to various forms of psychopathology and to the implementation of other emotion regulation strategies.

Given the scarcity of experimental paradigms that systematically model interpersonal processes, a first step toward a solution will consist of functional analyses of interactions between the participant and other participants and/or confederates. For example, when participants are having difficulties implementing a strategy to manage their anxiety, do they reach out to the other participant? If so, do they seek distraction or validation? In this respect, the literature on emotion regulation in psychopathology has much to learn from the dyadic studies that are so central to the developmental literature (see Cole et al., 2004; Eisenberg, Spinrad, & Eggum, 2010). Second, it will be important to expand on the operationalization of the strategies to incorporate joint processes, similarly to the case of co-rumination. For example, it would be useful to operationalize co-avoidance or co-reappraisal. Third, the use of EMA methodologies would allow researchers to assess naturalistic interpersonal processes that are both elicitors and regulators of emotion. A fourth suggestion involves matching the nature of the interpersonal scenario to those that are particularly problematic within the context of specific disorders, such as an interpersonal conflict script in borderline personality disorder (Suvak et al., 2012).

How are participants instructed to regulate?

There is substantial variability in the operationalization of emotion regulation strategies. This issue is perhaps best illustrated in the case of suppression, which encompasses activity in various domains: facial expressions (e.g., Gross, 1998a), internal experiences (e.g., Jackson, Malmstadt, Larson, & Davidson, 2000), or thoughts (e.g., Wenzlaff & Wegner, 2000). For example, some investigators have defined experiential suppression broadly (e.g., “please try your best to push these emotions away so that you do not feel them. Do your best to remain neutral during the film”; Tull, Jakupcak, & Roemer, 2010, p. 119). Others have equated it with thought suppression (e.g., “do not think about any of your feelings at all; push them out of your mind”; Quartana & Burns, 2007, p. 404). Still others have combined it with expressive suppression (e.g., “[do] not let your feelings take over or get out of control. In fact, you should try to behave in a way in which others, if they were in the room, would have no idea of how and what you are feeling”; Leen-Feldner, Zvolensky, & Feldner, 2004, p. 787). Recent work has also pointed out the heterogeneity of ways in which people engage in cognitive reappraisal (McRae et al., 2011). Needless to say, such variability interferes with the ability to make cross-study comparisons. This is particularly problematic in the context of psychopathology for two reasons. First, the emotional disturbances seen in psychiatric disorders are highly variable from one disorder to the next (e.g., positive affect is blunted in depression and social anxiety, elevated in mania, and within normal ranges in most anxiety disorders; Brown, 2007; Gruber, Harvey, & Gross, 2012). Second, people suffering from psychological disorders have been shown to have difficulties identifying their emotional states and their attempts at regulation of such states (e.g., Declercq, Vanheule, & Deheegher, 2010; Flynn & Rudolph, 2010; Gratz & Roemer, 2004; Marchesi, Brusamonti, & Maggini, 2000; Mennin, Holaway, Fresco, Moore, & Heimberg, 2007). Therefore, it is possible that individuals with psychopathology might show higher variability in their methods of implementing a given strategy (e.g., worry; Roemer, Molina, & Borkovec, 1997) or might confuse the strategies (e.g., by thinking they are problem solving when they are actually worrying and ruminating; Borkovec & Roemer, 1995; Nolen-Hoeksema et al., 2008; Wells, 2005).

The first step in addressing the heterogeneity in the operationalization of strategies consists of the development of a standardized coding scheme to functionally classify regulation instructions. Some of the proposed dimensions to code in such a system include whether instructions (a) pertain to activity in subjective, behavioral (i.e., expressive), and/or physiological domains; (b) provide examples on how to use the target strategy; (c) specifically instruct against the use of other strategies; and (d) include a practice period (and, if so, how long). Using these dimensions would facilitate cross-study comparisons and would allow for a thorough evaluation of the relationship between the type of strategy the investigators intended to induce and how participants actually implemented it. This could be examined via a meta-analytic review of the published data (although such a review would require a substantial degree of cooperation from the authors, given that a good part of this information is not available in published reports). It could also be approached by developing a series of guidelines for future investigations. A common language in the study of such complex phenomena as emotions is certainly an important step in the growth of the field over the next decade.

What do we know about spontaneous regulation?

The majority of experimental studies on emotion regulation have instructed participants to implement strategies, instead of assessing which strategies individuals implement spontaneously (cf. Egloff, Schmukle, Burns, & Schwerdtfeger, 2006; Ehring, Tuschen-Caffier, Schnulle, Fischer, & Gross, 2010; Gruber et al., 2012; Sheppes, Scheibe, Suri, & Gross, 2011). This omission has been problematic because it has interfered with the development of an understanding of the mechanisms by which individuals naturally select strategies. Indeed, the handful of studies assessing spontaneous selection (Egloff et al., 2006; Sheppes, Scheibe, Suri, & Gross, 2011; in psychopathology, Ehring et al., 2010; Gruber et al., 2012) has produced interesting findings that underscore the importance of such an approach. For example, Sheppes and colleagues have found that the spontaneous implementation of certain strategies can be a function of varying contextual demands, such that healthy participants use reappraisal in situations that elicit low emotional intensity and distraction when situations are high in affective intensity.

The assessment of spontaneously selected strategies need not be at odds with the use of paradigms in which participants are instructed to implement certain strategies. Indeed, valuable information could be obtained regarding the process by which participants deviate from their instructions, as this would mimic situations frequently encountered in psychotherapeutic settings when patients have difficulties carrying out homework assignments that consist of practicing regulation strategies at home (e.g., Mennin & Fresco, 2010; Roemer, Orsillo, & Salters-Pedneault, 2008). Measuring the spontaneous implementation of strategies can be easily achieved by incorporating questions that ask about the implementation of additional, noninstructed, strategies to the manipulation checks. Moreover, such manipulation checks could also incorporate a temporal dimension, by asking about (a) the order in which different strategies were implemented and (b) the amount of time an individual spent on a strategy before moving onto the next one. In order to reduce the demand characteristics of such manipulation checks, participants could be asked open-ended questions in which they are prompted to describe the ways in which they sought to regulate their emotions. Although these reports might still be influenced by reporting biases, it has been shown that when individuals report on their affect within 2 hr, such reports tend to be based on episodic or experiential knowledge, which is less susceptible to biases than semantic memory (Robinson & Clore, 2002). Another option that might allow investigators to minimize reporting biases would consist of asking participants to provide online ratings of the extent to which they are using each regulation strategy via the use of affect rating dials (see Ruef & Levenson, 2007). However, by doing so, investigators might increase participants’ cognitive load, which, in turn, might interfere with the ability to carry out the regulation effort (e.g., Wegner, Erber, & Zanakos, 1993)—and/or to report on it. ⁴ In addition, researchers risk priming participants to use certain strategies. Given that each of these approaches poses a potential limitation, the choice of which one to use should follow the theoretical considerations of a given investigation.

In terms of quantifying responses, I propose the calculation of several scores that could help capture the complex relationship among strategies in the emotion regulation repertoire: (a) specificity of target strategy, calculated as either the number of additional strategies used (by dichotomizing use of each strategy) or as the difference score between the use of the target strategy and the average of the other strategies; (b) regulatory effort, computed as the average use of all strategies; (c) flexibility, calculated as the number of times participants switch strategies (for additional scores, see Bonanno, Pat-Horenczyk, & Noll, 2011); (d) perseveration, computed as the continued use of a strategy when it is not producing desired effects; and (e) repetition, calculated as the extent to which participants go back to previously abandoned strategies. These proposed scores constitute a starting point that I hope will influence the development of a more complex approach to the assessment of spontaneous emotion regulation.

Can regulation be automatic?

A related issue pertains to the overreliance on study designs that manipulate and assess the conscious implementation of emotion regulation strategies and therefore ignore automatic or implicit processes. This is problematic because, although emotion regulation has been largely conceptualized as a conscious process (Gross, 1998b), there is a growing theoretical literature (Bargh & Williams, 2007; Gyurak, Gross, & Etkin, 2011; Koole & Rothermund, 2011) and empirical evidence from normative samples (e.g., Jostmann, Karremans, & Finkenauer, 2011; Mauss, Cook, & Gross, 2007; Mikulincer, Shaver, & Rom, 2011; Quirin, Bode, & Kuhl, 2011; Williams, Bargh, Nocera, & Gray, 2009) demonstrating the importance of studying emotion regulation at the automatic or implicit level. Indeed, this issue is particularly relevant to clinical disorders, as pathological states are characterized by disconnects between an individual’s experience and his or her conscious interpretation it (e.g., Barlow, 2002; Beck, 1976; Freud, 1920/1966). For example, difficulties identifying and labeling emotional states (e.g., Bagby, Parker, & Taylor, 1994; Mayer, Salovey, & Caruso, 2004) have been associated with a wide range of disorders (e.g., Declercq et al., 2010; Flynn & Rudolph, 2010; Gratz & Roemer, 2004; Marchesi et al., 2000; Mennin et al., 2007). Therefore, the systematic evaluation of automatic or implicit emotion regulation has the potential to shed light onto core processes of psychopathology.

The majority of the empirical studies on automatic emotion regulation have been conducted in normative populations, so the first step toward a solution will consist of evaluating such manipulations in clinical samples. A second step will consist of the examination of interactions between conscious and unconscious regulation. For example, in a normative sample, Williams and colleagues (2009) showed that the effects of nonconscious reappraisal were more pronounced in those individuals who did not habitually use reappraisal. Given that individuals with psychopathology tend to have difficulties using this strategy (see Aldao et al., 2010), evaluating the extent to which they could implement it in an automatic or implicit way could have important clinical implications. Third, it will be useful to develop experimental designs to examine how the conscious implementation of strategies becomes automatic and ingrained in the repertoire of strategies over time. For example, researchers could design paradigms including multiple follow-up periods and laboratory visits as well as take advantage of EMA methodology. Such an approach would allow the field to answer important theoretical questions and inform clinical practice. For example, what is the average number of times an individual needs to consciously practice implementing acceptance until it starts becoming “second nature”? Relatedly, how does the automatization of a certain strategy affect the conscious and unconscious implementation of other strategies?

How are participants actually regulating?

A related issue pertains to how participants actually implement the instructed (or spontaneous, for that matter) strategies. As mentioned earlier, investigators typically administer a series of questions at the end of the experiment asking participants to rate the extent to which they used certain strategies to regulate their affect. Not surprisingly, participants tend to rate themselves as implementing the strategy they were asked to use to a greater extent than any other strategies (e.g., Campbell-Sills et al., 2006; Wolgast et al., 2011), which is problematic because it is likely that demand characteristics play a role in how these ratings are provided. This issue is, of course, not exclusive to empirical research on emotion regulation; rather, it reflects the complex methodological problems surrounding manipulation checks in experimental studies, which is beyond the scope of this review. Nonetheless, an overly simplistic approach to manipulation checks poses a substantial methodological shortcoming given that individuals likely implement the same strategy via the use of a wide range of tactics (e.g., McRae et al., 2011). This problem is particularly accentuated in the case of clinical populations, because, as described earlier, individuals with psychological disorders might show higher variability in their methods of implementing a given strategy or might confuse strategies.

One way of reducing the demand characteristics of the manipulation checks is to ask participants open-ended questions about the ways in which they sought to regulate their emotions. Such an approach would force participants to generate responses and therefore reduce the likelihood that they might just endorse a high number on the strategy that they were required to implement. In addition, this approach would have the added benefit of allowing researchers to quantify the extent to which participants diverged from the intended regulation strategy and thus allow for more fine-grained analyses of their data (see McRae et al., 2011). Relatedly, it would provide investigators with the opportunity of capturing regulatory flexibility at two levels in tandem: strategies and their tactics. For example, investigators could assess whether individuals tend to implement a given strategy with the same set of tactics across situations or whether these individuals can modulate the selection of tactics on the basis of contextual demands. In addition, given the work by Sheppes and colleagues showing that when reappraisal is implemented in a response-focused fashion it leads to deleterious outcomes, such as elevated sympathetic activation (Sheppes, Catran, & Meiran, 2009) and increased expenditure of self-control resources (Sheppes & Meiran, 2008), it will be important to assess the point in time, relative to the elicitation of the affective state, in which the strategies are implemented. Lastly, given that individuals, particularly those with psychopathology, might have difficulties providing reliable reports of their regulation process, it might be useful to administer manipulation checks that assess implicit or automatic emotion regulation (see Koole & Rothermond, 2011).

How many strategies are being compared?

Empirical examinations have largely consisted of comparisons between only one or two strategies (cf. Hofmann et al., 2009, and Szasz et al., 2011, who examined three strategies). Therefore, much of our knowledge of the effects of strategies is relative to the other strategy with which they are compared and does not reflect the complexities of the emotion regulation repertoire. This problem is particularly exacerbated by the fact that there is an asymmetry in the number of studies conducting certain comparisons that further limits the scope of the approach. For example, a strategy such as reappraisal is much more frequently compared with expressive suppression (e.g., Butler et al., 2003; Denson, Grisham, & Moulds, 2011; Dillon, Ritchey, Johnson, & LaBar, 2007; Goldin et al., 2007; Gross, 1998a; Ochsner, Bunge, Gross, & Gabrieli, 2002; Richards & Gross, 2000) than with acceptance (e.g., Aldao & Mennin, 2012; Hofmann et al., 2009; Szasz et al., 2011; Wolgast et al., 2011).

To address this limitation, it will be important to increase the number of strategies compared with one another within each experimental session. This is important because it will allow researchers to delineate the effects of Strategy X not only with respect to Strategy Y but also in relation to a wider set of strategies, which would result in a richer delineation of the functional role of each strategy. In addition, in the experimental psychopathology literature, the inclusion of only one strategy does not allow investigators to draw conclusions regarding whether the deficits observed pertain to the implementation of a particular strategy or to broader deficits in the implementation of any strategy. Relatedly, a more frequent utilization of within-subjects designs would yield important additional information, as this would allow researchers to measure participants’ ability to flexibly implement various strategies. This is particularly important, because the ability to flexibly switch from one strategy to another (e.g., suppressing and enhancing facial expressions) has been associated with low levels of psychopathology (e.g., Bonanno, Papa, O’Neill, Westphal, & Coifman, 2004; Gupta & Bonanno, 2011; Westphal, Seivert, & Bonanno, 2011), and psychological flexibility is a key component of current therapeutic approaches, particularly acceptance and commitment therapy (e.g., Hayes et al., 1999), mindfulness-based cognitive therapy (Segal, Williams, & Teasdale, 2002), dialectical behavior therapy (Linehan, 1993), emotion regulation therapy (Mennin & Fresco, 2010), and acceptance-based behavioral therapy (Roemer et al., 2008). However, it is also the case that using within-subjects designs in the elicitation of affect might lead to spillover effects that contaminate the subsequent manipulation. Investigators are then encouraged to counterbalance the order and use precise measurements to assess the presence of such effects. For instance, it would be interesting to assess subsequent implicit mood and regulation tendencies (see Koole & Rothermund, 2011). Moreover, it would be useful to seek to model such effects as delayed regulation effects (see What is the time frame of effects? below).

What is the relationship among strategies in the repertoire?

Another example of omission of contextual factors pertains to the relatively nonexistent examination of the interaction among strategies. This omission is problematic because individuals possess an emotion regulation repertoire in which the selection of strategies likely influences the selection and effects (or lack thereof) of others. For example, in one study, participants instructed to suppress the facial expression of emotions elicited by a film clip were more likely to also use a cognitive strategy than those instructed to exaggerate their responses; moreover, the implementation of cognitive strategies in the context of suppression led to reductions in negative affect (Demaree, Robinson, Pu, & Allen, 2006). Similarly, in another study, participants failing to reduce aversive emotion by using reappraisal were more likely to engage in avoidance (Wolgast et al., 2011). However, both studies examined healthy participants, therefore limiting the clinical implications of their findings. From an experimental psychopathology perspective, Bonanno and colleagues have produced an extensive line of work examining flexibility in alternating between facial expressions and suppression and have identified such flexibility as a marker of psychological health (i.e., negative associations with psychopathology; Bonanno et al., 2004; Coifman & Bonanno, 2010; Gupta & Bonanno, 2011; Westphal et al., 2010). However, this work has focused on the regulation of facial expressions rather than on that of internal experiences. A more recent survey study examining a wider range of regulation strategies showed that the habitual use of adaptive strategies interacts with that of maladaptive strategies in the prediction of psychopathology, such that only those individuals who endorse using high (but not low) levels of maladaptive strategies display the expected negative association between adaptive strategies and symptoms (Aldao & Nolen-Hoeksema, 2012b). However, this study is correlational, which prevents us from drawing causal inferences. Beyond these investigations, the empirical examination of the complex interactions among strategies has been scarce, and therefore little is known about the processes by which individuals with psychopathology develop and maintain rigid patterns of dysregulation (e.g., Aldao et al., 2010; Kring & Sloan, 2010). This issue is particularly relevant for those treatment approaches that focus on the teaching of flexibility in the implementation of regulatory strategies (e.g., Hayes et al., 1999; Linehan, 1993; Mennin & Fresco, 2010; Roemer et al., 2008; Segal et al., 2002).

It will be important to build on the interaction findings between trait-level strategies (Aldao & Nolen-Hoeksema, 2012b) by examining interactions at the state level and in relation to a wider range of outcomes. It will also be useful to evaluate whether the implementation of Strategy X at Time 1 is associated with an increase or a decrease in the probability of using Strategy X at Time 2 (or Strategy Y at Time 2). Researchers could examine the spillover effects of regulation on subsequent tasks in the laboratory as well as on naturalistic interactions with the environment (via delayed surveys or EMA designs). Relatedly, it will be useful to use longitudinal designs to measure the process by which strategies implemented sporadically become ingrained in the emotion regulation repertoire. Of particular interest will be to conduct such work within a developmental framework to delineate the process by which children and adolescents develop their repertoire as they become adults. A third question pertains to whether individuals can implement multiple strategies simultaneously or whether regulatory attempts can take place only sequentially. Two studies suggest that people can implement multiple strategies at once (Demaree et al., 2006; Wolgast et al., 2011). It will be important to examine whether this simultaneous implementation reflects regulatory flexibility or haphazard attempts at regulation. Examining this process in normative and healthy populations will constitute a first step in the evaluation of its functional adaptiveness. These and related questions have the potential to spur the development of experimental paradigms that will allow us to achieve a more thorough understanding of the patterns in the implementation of emotion regulation strategies and help identify processes that might break the vicious cycles that underlie the rigidity seen across clinical disorders.

Are all emotional domains equally affected?

Emotions are the result of activation in multiple domains, namely, the subjective, behavioral, and physiological (e.g., Bradley & Lang, 2000; Ekman, 1992). Of note, the physiological domain encompasses activity in both the peripheral and central nervous systems (e.g., affective neuroscience; Davidson et al., 2000; Ochsner & Gross, 2008). Despite this recognition of the multifaceted nature of emotion, many questions remain unanswered. For example, we do not know how much activity in each domain is required for an emotion to take place. For example, if an individual is having an anxious thought, how much physiological reactivity is necessary (if any at all) for that thought to be considered part of an emotion? The answer to this question belongs to the field of philosophy and is, therefore, well beyond the scope of this review. However, I mention it here because it influences how we think about the effects of emotion regulation strategies. When people seek to regulate an emotion, they tend to use one domain to influence other domains. For example, when individuals implement reappraisal, they use the subjective domain (i.e., modify their thoughts) to reduce arousal in the physiological domain and produce a given behavioral response. Therefore, to fully examine the influence of regulation strategies on each domain, it is important that we develop a more complex approach to modeling the relationships among domains. To this date, a handful of investigations have sought such an endeavor by calculating the coherence (i.e., correlation) among domains during emotional reactivity (e.g., Mauss et al., 2007; Sze, Gyurak, Yuan, & Levenson, 2010). However, such an approach has not yet been implemented in relation to the regulation of such reactivity. This scarcity of studies is likely the result of the methodological (e.g., specialized equipment) and statistical (e.g., lagged correlations) challenges involving the continuous measurement of affect. Despite these complexities, identifying the processes that constitute a normative orchestrated response for a given strategy in a particular context might be extremely beneficial in delineating disruptions in regulation within the context of psychopathology.

Experimental investigations of emotion regulation usually include a baseline and, less frequently, a recovery period. Therefore, they consist of at least three discrete time points: baseline, stimulus presentation, and recovery. It is common practice to conduct repeated measures multivariate analyses of variance to predict activity at various time points. However, very little attention has been paid to an equally simple yet conceptually important type of analysis: whether activity in Domain X during the stimulus presentation predicts changes in Domain Y from stimulus presentation to recovery. Or, even more simply, it would be useful to calculate correlations among domains at each point in time and then statistically compare whether they are different from each other via the calculation of Fisher’s Z coefficients. Conducting this type of analysis would constitute a critical first step in delineating the complex relationship between emotional domains during the regulatory process.

What constitutes a beneficial outcome?

The emotion regulation literature has evaluated the adaptiveness of outcomes by focusing on hedonic goals, that is, the ability of strategies to (a) downregulate negative affect and/or (b) maintain or increase positive affect. It is not surprising that such a narrow conceptualization of the multifaceted transactions between the individual and his or her environment that ignores additional, nonhedonic, goals has resulted in mixed and puzzling findings. From the basic literature, one clear example is the comparison between the effects of acceptance and reappraisal, both putatively adaptive strategies. In one study examining regulation in response to emotions elicited by film clips, acceptance and reappraisal led to lower subjective distress, physiological activation, and behavioral avoidance than a no-instructions condition; however, for some of the film clips, reappraisal produced larger subjective and physiological effects than acceptance (Wolgast et al., 2011). Similarly, in another study, during an impromptu speech task, acceptance and reappraisal produced smaller increases in heart rate from baseline to speech task than the suppression condition (and did not differ from each other), yet reappraisal resulted in smaller increases in subjective anxiety than suppression, with acceptance falling nonsignificantly in between (Hofmann et al., 2009). Lastly, in an experiment consisting of a frustration task, reappraisal resulted in larger decreases in anger and greater behavioral persistence than did acceptance and experiential suppression (Szasz et al., 2011). Overall, these findings would suggest that reappraisal might be more beneficial for the regulation of subjective states than acceptance; however, such findings do not provide a clear picture regarding physiological states or behavior. More important, before making conclusive claims pertaining to the greater effectiveness of one strategy versus the other, it will be critical to integrate these findings within the purported function of each strategy. In other words, insofar as acceptance entails the allowance of affective states (e.g., Hayes et al., 1999), it is to be expected that it would lead to lower decreases in emotional experience than reappraisal (e.g., Aldao & Mennin, 2012). Indeed, in a study comparing acceptance with experiential suppression and a no-instructions condition, acceptance was found to result in ongoing emotional activation, whereas suppression led to downregulation of emotion (Dunn, Billotti, Murphy, & Dalgleish, 2009).

In the clinical literature, this issue is clearly exemplified in the demarcation of how much downregulation of affect is “sufficient.” Two recent studies have shown that anxious children and adults seeking to downregulate negative affect achieved similar reductions as nonanxious controls (Campbell-Sills et al., 2011; Carthy, Horesh, Apter, Edge, & Gross, 2010). However, it is also the case that highly anxious individuals begin the regulation process at higher absolute levels of negative affect. In this respect, a comparable reduction to that achieved by controls might still leave the anxious individuals experiencing a substantial amount of negative affect. In other words, focusing on the slope of change alone, rather than in tandem with the intercepts, provides a limited grasp on the complex dynamics of regulation.

Recent theoretical work (Koole, 2009; for a philosophical discussion, see Charland, 2011) and a handful of empirical studies by Tamir and colleagues (e.g., Tamir & Ford, 2009; Tamir et al., 2008) have proposed to broaden the conceptualization of emotion regulation beyond hedonic goals to incorporate idiographically relevant instrumental goals. For example, Tamir and colleagues (2008) have shown that when preparing to have a confrontation, participants preferred activities that increased anger. In other words, individuals were choosing to endure a difficult and unpleasant state in the present for the sake of a valued benefit in the future. This is reminiscent of the delayed gratification literature (Mischel, Shoda, & Rodriguez, 1989) and, to a lesser extent, of the work on distress tolerance (see Zvolensky, Bernstein, & Vujanovic, 2010). Relatedly, there has been a recent interest in identifying the paradoxical negative effects of positive affective states (e.g., Gruber, Mauss, & Tamir, 2011; Mauss, Tamir, Anderson, & Savino, 2011).

I propose a further delineation of regulation goals by incorporating an idiographic approach in which participants provide information regarding their motivation at baseline. For example, in an experimental manipulation consisting of a cold pressor in exchange for a certain amount of money, it will be useful to obtain information regarding participants’ goals (e.g., how important it is to earn the money, how concerned they are about the water being almost freezing, how much they fear negative evaluation by the experimenters). Then, their performance on the task (e.g., latency to quit, reports of pain) could be evaluated within a richer idiographic and motivational context. However, the notion of relying on participants’ idiographic description of what constitutes adaptive regulation should be evaluated with caution, particularly in the case of individuals experiencing elevated symptoms of psychopathology. As described above, most clinical disorders are characterized by patterns of rigidity and avoidance (e.g., Cheng, 2001; Davidson et al., 2000; Grillon, 2002; Hayes et al., 1999; Kashdan & Rottenberg, 2010; Rottenberg et al., 2005); therefore, individuals with psychopathology are likely to identify goals that serve an avoidant function. To address this potential catch-22, it will be important to use a multimethod approach to evaluate regulation outcomes across multiple domains and, ideally, over repeated instances. In this respect, investigators will be able to develop a more nuanced understanding of such patterns of rigidity and avoidance.

What is the time frame for effects?

In addition to focusing on hedonic goals, most of the empirical investigations on emotion regulation have examined the immediate effects of regulation on affect and behavior, and relatively little attention has been devoted to long-term or spillover effects of a regulatory effort. This is particularly problematic because the outcome of a given regulation event likely influences the next event in which affect is generated and regulated. For example, if an individual is anxious about confronting his or her employees about something they did wrong and avoids them altogether, the person might experience an immediate reduction in anxiety. We would be tempted to call this regulation effort “successful.” However, if the person then gets home and continues to worry about the situation and experiences higher anxiety about this situation as well as the other things going on at home that evening, we might be tempted to question the success of this regulatory effort. Therefore, when examining the influence of emotion regulation on affect and behavior, it becomes extremely important to juxtapose the immediate and delayed effects. This point becomes more central for the study of emotion regulation in psychopathology, as psychological disorders are characterized by vicious cycles in which individuals rigidly favor the short- over the long-term consequences of their behavior.

The primary way of addressing this limitation in the literature would consist of the development of paradigms that would allow for the joint examination of immediate and delayed effects of regulation. The simplest approach would involve the systematic incorporation of affective recovery periods, that is, a period following the regulation task in which participants sit quietly (e.g., Campbell-Sills et al., 2006). A more complex approach would consist of the incorporation of an additional task, the performance of which would be analyzed in relation to the type of regulation the participant conducted in the initial task. Of note, such spillover task need not be an emotion regulation task. For example, one could measure the influence of emotion regulation on attentional bias (e.g., Jamieson, Nock, & Mendes, 2011). To increase the ecological validity, researchers could collect data from the participant after the laboratory assessment has taken place. This could be done via a follow-up survey, a second visit to the laboratory, a diary study, or an EMA design (e.g., Ebner-Priemer & Trull, 2009). In this respect, Kuppens and colleagues have shown that emotional inertia, that is, the degree to which a person’s current emotional state is predicted by the prior one, is positively associated with psychopathology (e.g., Kuppens et al., 2012).