Available online at www.sciencedirect.com

ScienceDirect Behavior Therapy 45 (2014) 56 – 66 www.elsevier.com/locate/bt

Five Indices of Emotion Regulation in Participants With a History of Nonsuicidal Self-Injury: A Daily Diary Study Konrad Bresin University of Illinois at Urbana–Champaign

Theory has proposed that nonsuicidal self-injury (NSSI) is a response to intense frequent negative affect (NA) that is difficult to control. Therefore, individual differences that are related to emotion dysregulation should be higher in individuals who engage in NSSI compared to healthy controls. Though current research supports this prediction, this research could be strengthened by corroborating evidence from daily diary studies. The current study used a daily diary protocol to thoroughly examine the emotional correlates of NSSI. Individuals with and without a history of NSSI rated their affect daily for 14 days. This information was used to score multiple indices of emotionality (e.g., mean level, within-person variation, reactivity). The results showed that compared to controls, individuals who engaged in NSSI had higher mean levels, within-person variation, and lower emotional differentiation of NA, but groups did not differ on inertia of NA or reactivity of NA. Moreover, individuals with a history of NSSI reported lower levels of positive affect and lower inertia of positive affect. These results are discussed in terms of affect regulation models of NSSI and treatment implications.

Keywords: self-injury; self-harm; negative affect; diary study

NONSUICIDAL SELF-INJURY (NSSI) IS DEFINED as the intentional destruction of one’s own body tissue

The author would like to thank Yara Mekawi and M. Sima Finy for comments on earlier drafts of this manuscript. The author also would like to thank Kathryn H. Gordon for her supervision during data collection. Address correspondence to Konrad Bresin, Department of Psychology, University of Illinois at Urbana–Champaign, 603 E. Daniel Street, Champaign, IL 61820; e-mail: [email protected] 0005-7894/45/65-75/$1.00/0 © 2013 Association for Behavioral and Cognitive Therapies. Published by Elsevier Ltd. All rights reserved.

without suicidal intent (Nock, 2009). Prominent theories of NSSI focus on the role of emotions, particularly emotional dysregulation, as a vulnerability factor for engagement in NSSI (Chapman, Gratz, & Brown, 2006; Linehan, 1993; Nock). Therefore, understanding facets of emotion regulation that differ between individuals with and without a history of NSSI may be a way to identify candidate vulnerability factors that are involved in the onset and maintenance of NSSI. Though some research has found that individuals who engage in NSSI report higher levels of individual differences related to emotion dysregulation (e.g., neuroticism; Brown, 2008) than controls, these studies have relied on retrospective self-report measures of long time frames (e.g., “How do you feel in general?”), which are known to be influenced by multiple biases, particularly in regard to individual differences in emotion and emotion regulation (Robinson & Clore, 2002). Currently, there is an absence of studies examining the difference between individuals with and without a history of NSSI on indices of emotion dysregulation based on daily ratings of affect. In this vein, the goal of the current study was to examine the emotional correlates of NSSI in an ecologically valid manner using a daily diary study.

affect and emotion dysregulation Affect has generally been conceptualized to consist of two components: valence (unpleasant to pleasant) and arousal (low to high; Larsen & Diener, 1992; Russell, 1980; Watson & Tellegen, 1985). Within this two-dimensional affective space, negative affect (NA) is defined as unpleasant valence that is high in arousal (with low arousal, pleasant affect [e.g., relaxed] as the opposite pole), whereas positive affect (PA) is defined as pleasant valence that is high in arousal (with low arousal, unpleasant affect [e.g., sluggish] as the opposite pole; Watson,

emotion-regulation in participants with a history of nssi Clark, & Tellegen, 1988). Though in a true circumplex structure the correlation between NA and PA would be zero, the relatio\nship between them is negative and moderate to small (Watson, 2000). However, the strength of the relationship depends on the time frame assessed, with shorter time frames (e.g., “right now”) leading to larger correlations than longer time frames (e.g., “in general”; Watson). Still, NA and PA are generally considered to be unique constructs (Watson; though see Russell & Carroll, 1999). This suggests that studying NA and PA separately is important because dysregulation in one (e.g., NA) does not necessarily imply dysregulation in the other (e.g., PA). Berenbaum, Raghavan, Le, Vernon, and Gomez (2003) provide a useful taxonomy for the organization of emotional dysregulation, which includes three broad categories of disturbances. First, emotional valence disturbances occur when an individual experiences extremely high or extremely low levels of pleasant or unpleasant affect. For example, internalizing psychopathology may be characterized by high levels of NA (Watson, O'Hara, & Stuart, 2008). Second, emotional intensity disturbances occur when an individual is exceptionally over- or underreactive to emotional situations. For instance, it has been proposed that individuals with borderline personality disorder (BPD) are overly sensitive to emotional cues (Linehan, 1993). Finally, emotional disconnections occur when the individual has a particularly low awareness of his or her own emotional experience. Awareness of emotion can be parsed into two components: attention to emotion (i.e., valuing and attending to emotions) and clarity of emotion (i.e., the ability to distinguish and identify emotions; Gohm & Clore, 2000). It is likely that being on either extreme of these emotional disturbances (e.g., overreactive, underreactive) can be maladaptive, but varying levels in between these extremes are likely adaptive. Each of these disturbances could play a role in NSSI. For example, theory suggests that individuals who engage in NSSI frequently experience high levels of NA (Chapman et al., 2006; Nock, 2009), consistent with a valence disturbance in unpleasant emotions (i.e., too much high arousal, unpleasant emotion). This prediction has been supported by research showing that individuals with a history of NSSI report higher trait NA than individuals without a history of NSSI (Baetens, Claes, Willem, Muehlenkamp, & Bijttebier, 2011; Brown, 2008; Kamphuis, Ruyling, & Reijntjes, 2007; Maclaren & Best, 2010). Though other research also suggests that low-arousal unpleasant emotions are important in NSSI (Kamphuis et al., 2007; Nock, Prinstein, & Sterba, 2009), most research suggests that high-arousal unpleasant emo-

57

tions are more important in NSSI, given that they predict future NSSI (Nock et al., 2009) and are reduced more than low-arousal unpleasant emotions following engagement in NSSI (Klonsky, 2009). In terms of positive valence, it is unclear if individuals who engage in NSSI also report fewer experiences of PA compared to controls, as previous studies have not consistently found group differences in trait PA. In addition to valence disturbances, it has been proposed that individuals who engage in NSSI are overreactive to emotional stimuli (Chapman et al., 2006; Linehan, 1993). Consistent with this, individuals who engage in NSSI have higher scores than controls on trait self-report measures of emotional reactivity (e.g., Glenn, Blumenthal, Klonsky, & Hajcak, 2011). However, as detailed below, experimental studies of emotional reactions in the laboratory have found mixed results. To date, studies have mostly focused on reactivity to negative stimuli; thus, it is unclear if individuals who engage in NSSI are over- or underreactive to positive stimuli. Finally, previous cross-sectional research has found that individuals with a history of NSSI report lower levels of awareness and clarity of their emotions than controls (e.g., Gratz & Roemer, 2008; Polk & Liss, 2007). These results make sense in light of other research, which suggests that a lack of emotional awareness is related to poor mental health. For instance, lower levels of emotional differentiation, similar to emotional clarity, are related to the use of poor emotion-regulation strategies (Feldman Barrett, Gross, Christensen, & Benvenuto, 2001). Conversely, high emotional differentiation has been related to a decreased likelihood of engaging in impulsive behaviors (e.g., aggression) when experiencing intense emotions (e.g., anger; Pond et al., 2012). Therefore, it is reasonable to assume that individuals who engage in NSSI would have poor emotional awareness. Because previous studies have only looked at the awareness of emotions in general, not awareness of NA or PA specifically, it is unclear whether the emotional awareness abilities of individuals who engage in NSSI would be general (i.e., affect overall) or specific (e.g., only for NA). One aspect of emotion dysregulation that is not covered in the Berenbaum et al. (2003) model is variability. For example, two individuals could have the same valence disturbance (e.g., extremely high, unpleasant affect), but one may always have a high level, whereas the other may fluctuate greatly. This type of intra-individual variability could be important to understanding NSSI in at least two ways. On one hand, it is possible that individuals who engage in NSSI have high mean levels of NA and low variability. Thus, it might be hypothesized that they engage in NSSI as an attempt to regulate persistent

58

bresin

negative mood. On the other hand, it is possible that individuals who engage in NSSI have highly variable moods, and NSSI is a response to large shifts in NA (or PA). One previous cross-sectional study found partial support for the latter proposal, such that self-reported affective lability, or the tendency to switch between emotional states, was correlated with NSSI frequency (Kamen, Pryor, Gaughan, & Miller, 2010). Though the current study did not examine the onset of NSSI, understanding how intra-individual variation in affect differs between individuals with and without a history of NSSI can begin to narrow down the plausible roles of intra-individual variability in the onset of NSSI. Despite the evidence supporting the roles of trait NA, emotional reactivity, emotional awareness, and affective lability in NSSI, there are still gaps in the literature. Given that individual differences are considered to be general tendencies (Pytlik Zillig, Hemenover, & Dienstbier, 2002), there are multiple ways to assess individual differences in emotion and emotion regulation. The most common approach is to have participants provide judgments about their average affective experience. This requires participants to retrospectively aggregate across multiple events (e.g., his/her entire life), which can be cognitively demanding (Piasecki, Hufford, Solhan, & Trull, 2007). Therefore, participants rely on heuristics to make judgments, which introduces a variety of biases into their reports, including placing a greater weight on some experiences over others (e.g., mood-congruent memories; Mayer, McCormick, & Strong, 1995) and being overly influenced by recent events (Gorin & Stone, 2001; Kahneman, Fredrickson, Schreiber, & Redelmeier, 1993). Hence, given the current literature, it is unclear whether individuals who engage in NSSI experience emotion dysregulation or whether they are more biased in their retrospective reports than individuals who do not engage in NSSI. An alternative measurement strategy is to have participants provide multiple assessments of affect across a variety of situations (e.g., ecological momentary assessment; EMA; Shiffman, Stone, & Hufford, 2008) and aggregate the events statistically. This strategy is desirable because the effects of biases are decreased as the time frame assessed decreases (Robinson & Clore, 2002), and thus shorter time frames (e.g., a day) will be a better representation of the individual’s affective experiences than longer time frames (e.g., in general). Consistent with this idea, previous research has found only moderate correlations between trait self-report measures and aggregated momentary measures (e.g., Anestis et al., 2012; Solhan, Trull, Jahng, & Wood, 2009). This is not to say that one method is more important than the other,

as they both provide unique information. In fact, more scientific credence should be put into results that replicate across measurement types. Though previous research on NSSI has used EMA (Armey, Crowther, & Miller, 2011; Muehlenkamp et al., 2009; Nock et al., 2009), these studies have not focused on individual differences in affective experiences. Instead, they have been focused on affective states preceding and following NSSI engagement. Hence, they have not taken advantage of the multiple indices that reflect different processes, which can be calculated from repeated assessments.

momentary indices of individual differences in affect The most straightforward index of individual differences in emotion would be an average of NA or PA over multiple assessments in multiple environments, with higher levels indicating a propensity to experience a given type of affect, corresponding to valence disturbances, at least at extreme levels. Therefore, it is likely that individuals with a history of NSSI have higher mean levels of NA, but based on the current literature, there is no clear prediction to be made about mean level of PA. In addition to calculating mean levels within an individual, there is also the ability to calculate variation within a person. Intra-individual variability can be calculated in multiple ways. Theorists have proposed that these measures can be divided into two broad categories: one that reflects spontaneous change irrelevant to time or environment, and one that reflects systematic change over time (Fiske & Rice, 1955; Ram & Gestorf, 2009). These distinct categories of intra-individual variability are thought to reflect unique processes. Spontaneous changes, which can be indexed by within-person variance, are thought to reflect the capacity for change (e.g., rigidity versus lability). Given that one previous study found a positive cross-sectional correlation between selfreported affective lability and NSSI frequency (Kamen et al., 2010), individuals who engage in NSSI may have more within-person variance than individuals who do not. In contrast, time-dependent measures of variability are thought to reflect adaptive processes in response to an environment (e.g., a return to homeostasis; Ram & Gestorf, 2009). For example, emotional inertia, or the autocorrelation between affect at one time point and the next, can be thought of as a response to emotional situations (Suls, Green, & Hillis, 1998). The smaller this relationship is (i.e., as it approaches zero), the quicker the individual returns to homeostasis following emotionally eliciting incidents. Conversely, the larger this relationship is (i.e., as it approaches one), the more likely the

emotion-regulation in participants with a history of nssi individual is to have situation-insensitive affect. That is, affect from one assessment period will carry over into another. Because theories of NSSI posit that individuals who engage in NSSI are highly sensitive to emotional situations (Chapman et al., 2006; Linehan, 1993), it might indicate that individuals with a history of NSSI may have low levels of inertia (i.e., affect at conjoining time points are unrelated). Compared to other individual differences in emotion regulation, emotional reactivity is amendable to laboratory experimental paradigms. Despite the increased experimental control gained by using these methods, there is the possibility of losing ecological validity. That is, laboratory settings may be too artificial to create emotions of the same intensity as those experienced in real life. As mentioned above, the results from experimental studies on NSSI have provided limited support for the idea that individuals who engage in NSSI are more emotionally reactive than controls. Though Nock and Mendes (2008) found that compared to controls, adolescents with a history of NSSI displayed greater skin conductance during a frustrating card-sorting task, other studies have found null results with both self-report and psychophysiological measures of affect (Glenn et al., 2011; Gratz et al., 2011; Kaess et al., 2012; Weinberg & Klonsky, 2011). Moreover, studies finding null results have used both standardized emotional stimuli (e.g., images; Glenn et al.) and personalized stimuli (e.g., recalling a distressing situation; Gratz et al.). These results are difficult to interpret, given that they may be due to a methodological limitation. However, it is equally likely that individuals who engage in NSSI are not more reactive to emotional stimuli compared to individuals who do not engage in NSSI. One way to possibly provide clarity to the field would be to examine affective response to everyday events in close proximity to when they happen. If individuals who engage in NSSI are overreactive, they should have a larger increase in NA on days that negative events happen in relation to individuals who do not have a history of NSSI. Measures of clarity of emotion ask participants to respond to items such as, “I am usually clear about my feelings” (Salovey, Mayer, Goldman, Turvey, & Palfai, 1995). This requires a high level of introspection, which may not be present in individuals who engage in NSSI. An alternative to this strategy is to examine emotional differentiation, or the extent to which different emotional states are correlated within an individual. Individuals with low differentiation will have a high correlation between unique emotional states (e.g., anger, fear, sadness), whereas individuals with high levels of differentiation tend to experience emotional states as more discrete. Because previous research has shown that NSSI is

59

related to low awareness of emotion cross-sectionally (e.g., Gratz & Roemer, 2008), it is possible that individuals who engage in NSSI will show low emotional differentiation.

current study To summarize, theory has proposed that NSSI is a response to intense, frequent NA that is difficult to control (Chapman et al., 2006; Linehan, 1993). Therefore, the general prediction would be that individual differences in emotion dysregulation should be higher in individuals who engage in NSSI compared to healthy controls. Though current research supports this prediction (Brown, 2008; Kamen et al., 2010), this research could be strengthened by corroborating evidence from daily diary designs. However, the few EMA studies on NSSI have not examined individual differences in emotional responding (e.g., Armey et al., 2011). To address this gap in the literature, a daily diary study was conducted. Participants with and without a history of NSSI made daily ratings of NA and PA, along with emotionally eliciting events, once a day for 14 days. These ratings were used to calculate five indices of emotional experience: mean level, withinperson variance, inertia, reactivity, and differentiation. 1 Based on previous research (Brown, 2008; Glenn et al., 2011; Gratz & Roemer, 2008), it was predicted that compared to individuals with no history of NSSI, individuals with a history of NSSI would report higher mean level and reactivity for NA and lower emotional differentiation for NA. Given the lack of theory and research in the area, the predictions for within-person variability of NA and NA inertia were somewhat exploratory. Still, it was assumed that individuals with a history of NSSI would have greater levels of within-person variation 1 Some researchers also suggest using measures that combine the temporal aspects and magnitude of change (Ebner-Priemer, Eid, Kleindienst, Stabenow, & Trull, 2008; Jahng, Wood, & Trull, 2008). One such measure is the mean squared successive difference (MSSD; von Neumann, Kent, Bellinson, & Hart, 1941), which is calculated as the average of the squared differences between time points. This measure has been criticized for confounding temporal and magnitude effects (Wang, Hamaker, & Bergeman, 2012). For completeness, analyses for MSSD were also run. The NSSI group had significantly greater MSSD in NA (M = .27) than the non-NSSI group (M = .13), t(116) = 3.29, p = .001, pseudo-r2 = .10. However, the two groups did not significantly differ on instability in PA, t(116) = 1.56, p = .121, pseudo-r2 = .03. These results were unchanged when adjusting for mean level of affect (cf. Ebner-Priemer et al., 2008). However, when controlling for within-person variation, NSSI groups did not significantly differ on MSSD of NA, t(117) = .94, p = .351, pseudo-r2 = 01. Hence, the MSSD results appear to be driven by within-person variability, not inertia. Given that these results were largely redundant, they were not reported in the main text.

60

bresin

but lower levels of inertia in comparison to controls. No a priori predictions concerning PA were made due to the limited research in this area.

Method participants Across two semesters, college undergraduates (N = 1,612) were screened for participation in the study by completing the Deliberate Self-Harm Inventory (DSHI; Gratz, 2001). The DSHI is a 17-item questionnaire that is used to assess whether participants had ever engaged in a variety of methods of NSSI (e.g., cutting, biting) as well as the last time this had happened (in years). Similar to the prevalence rates in other college samples (Whitlock, Eckenrode, & Silverman, 2006), 23% of participants endorsed lifetime presence of at least one NSSI incident. Of those with a lifetime history of NSSI, 8% reported that their most recent incident occurred in the last year. Participants with at least one NSSI incident that occurred in the last year and participants who reported no history of NSSI were invited to participate in the study. A total of 118 participants (94% Caucasian) were recruited for the study for course credit. The non-NSSI group consisted of 57 participants (33 women) with a mean age of 19.73 (SD = 2.94) years. The NSSI group consisted of 61 participants (34 women) with a mean age of 19.67 (SD = 3.03) years. Among those recruited, the most common method of NSSI was cutting (49%) and the least common was biting (7%). The median number of NSSI episodes was 15 (range = 1–1,000), and participants reported using an average of 3.00 (SD = 2.00) methods. The average time since the most recent NSSI incident was .49 (SD = .40) years. Taken together, these results indicate that this sample is comparable to samples used in previous research in terms of methods and frequency of NSSI (Gratz et al., 2011; Nock & Mendes, 2008). daily assessments Affect was measured using the Positive and Negative Affective Schedule (PANAS; Watson et al., 1988). Participants rated a total of 20 items, 10 related to NA (e.g., “distressed”) and 10 related to PA (e.g., “proud”), with the instructions to rate how they generally felt “today” on a 5-point Likert scale (1 = very slightly/not at all, 5 = extremely). The PANAS has been used in multiple EMA studies (e.g., Armey et al., 2011; Berg et al., 2013). The psychometric properties (e.g., factor structure correlation between the scales) of the “today” instructions are similar to that of other instructions (e.g., “in general”; Berg et al.; Watson & Clark, 1994). Alphas in this sample (NA = .89; PA = .80) were calculated

using generalizability theory (Cranford et al., 2006), which decomposes the variance into between-subject, between-day, and between-item components and two-way interactions. Therefore, the alphas appropriately account for the nested nature of the data. Similar to previous studies, there was a small to moderately negative correlation between the scales at both the between-subject, r(118) = -.15 p = .086, and within-subject levels, r(1122) = -.20, p b .001. Each day, participants reported whether nine unpleasant events and/or four pleasant events had happened to them that day. These events were taken from the Daily Events Survey originally created by Butler, Hokanson, and Flynn (1994) and later adapted by Nezlek and Plesko (2003). A smaller subset of events from the original scale was used to reduce participant burden with the diary protocol. Events were both interpersonally (e.g., “was excluded or left out by my group of friends”; “went out socializing with friends/date”) and achievement (e.g., “fell behind in course work or duties”; “did well on a school or work task”) oriented. Because the results were the same for both event types, scores were collapsed across categories. The events were summed within days to create unpleasant and pleasant event scores. The alphas (also calculated using generalizability theory) in the current sample were .68 for unpleasant events and .75 for pleasant events, which are similar to those reported in previous research using diary designs (Nezlek & Plesko). Previous research supports the construct validity of these events in that NA is higher on days with more unpleasant events, and PA is higher on days with more pleasant events (Nezlek & Plesko).

procedure Participants were recruited via email to complete a laboratory study unrelated to the diary study (see Bresin & Gordon, 2013). At the end of the laboratory session, participants were given instructions for the daily diary protocol. Starting the Monday after the laboratory portion, participants completed 14 days of diary ratings. Each evening, participants were sent an email at approximately 7 P.M. with a link and password for that day’s diary entry using a secure website. Participants were instructed to fill out the diary as close to the time that they went to bed as possible but were given until 9 A.M. the next morning to complete the questions. This sampling frequency was chosen based on previous research in this area (Feldman Barrett et al., 2001; Nezlek & Plesko, 2003; Suls, Green, & Hillis, 1998) and available technological resources. To encourage compliance, participants were given additional course credit for completing at least 11 days. Across groups, participants completed

61

emotion-regulation in participants with a history of nssi an average of 10.51 (SD = 2.97) days. Though the two groups did not significantly differ on the number of days completed, t(116) = -1.79, p = .076, d = -.32, the non-NSSI group completed on average one more day (M = 11.01, SD = 10.32) than the NSSI group (10.04, SD = 9.22). Controlling for number of days completed, however, did not affect any of the results.

Results Due to the nested nature of the data (days within-subjects), multilevel modeling (MLM) was used for all analyses (aside from emotional differentiation, see below). MLM is able to accurately parse variance apart at the between- and withinsubject levels to allow for accurate parameter estimates of nested data (Singer & Willett, 2003). Moreover, given that MLM is robust to missing data (Singer, 1998), this allowed for the use of data from all participants who completed more than one day of diary entries. All models were run in SAS 9.3 (SAS Institute Inc., 2011) PROC MIXED (or PROC NLMIXED where appropriate), with the settings recommended by Singer (1998). To accommodate the nested nature of the data, an unstructured variance-covariance matrix was used. This type of variance-covariance matrix applies no restrictions to the variance components, which can provide a better fit to the data than methods with restrictions (e.g., autocorrelation, compound symmetry; Singer & Willett, 2003). Alternative variance-covariance structures were explored; however, none provided a significantly better fit than the unstructured. More importantly, the results for the fixed effects remained unchanged. The parameter estimates for all models are displayed in Table 1. Effect sizes are reported as pseudo-r 2 values, which indicate the percentage of the explainable variance explained by the addition of a given parameter over a model without that parameter (Bryk & Raudenbush, 1992). This effect size was chosen for multiple reasons. First, it allowed for consistency across analyses, some of which involved group differences of a mean, and some of which involved group differences in a slope. Second, it allows for the calculation of effect sizes for both between- and within-person variability separately, which was desirable as some analyses involved between-subject predictors (e.g., NSSI group) and others involved cross-level interactions (e.g., NSSI group by unpleasant events). To save space, pseudo-r 2 values are only reported for the most relevant level (e.g., within-subjects for a cross-level interaction).

mean level and variability Group differences in mean level and variability of affect were examined using the mixed-effects location scale model proposed by Hedeker, Mermelstein,

Table 1

Unstandardized Regression Coefficients and Standard Errors for Multi-Level Modeling Results Negative Affect Positive Affect

Mean Level Fixed Effects 1.60 (.06)⁎⁎ Intercept (γ00) .29 (.09)⁎⁎ NSSI Group (γ01) Variance Components (in natural log units) -1.57 (.18)⁎⁎ Between Intercept (α00) NSSI group (α01) .10 (.20) Within Intercept (τ00) -2.57 (.16)⁎⁎ .80 (.22) NSSI group (τ01) Inertia Fixed Effects 1.58 (.06)⁎⁎ Intercept (γ00) .30 (.09)⁎⁎ NSSI Group (γ01) Lag NA/PA (γ10) NSSI Group by Lag NA/PA (γ11) Variance Components 2 ) Between (σBw 2 Within (σWn) Emotional Reactivity Fixed Effects Intercept (γ00) NSSI Group (γ01) Events (γ10) NSSI Group by Events (γ11) Variance Components 2 ) Between (σBw 2 Within (σWn)

.06 (.05) -.04 (.06) .23 (.03)⁎⁎ .19 (.008)⁎⁎

2.87 (.08)⁎⁎ -.25 (.11)⁎ -1.08 (.19)⁎⁎ .09 (.28) -1.37 (.11)⁎⁎ .15 (.15) 2.85 (.08)⁎⁎ -.25 (.11)⁎ .16 (.04)⁎⁎ -.15 (.06)⁎ .37 (.01)⁎⁎ .32 (.05)⁎⁎

1.60 (.06)⁎⁎ .30 (.09)⁎⁎ .10 (.01)⁎⁎

2.86 (.08)⁎⁎ -.24 (.11)⁎ .22 (.02)⁎⁎

.02 (.02)

.01 (.03)

.23 (.03)⁎⁎ .16 (.007)⁎⁎

.35 (.05)⁎⁎ .28 (.01)⁎⁎

Note. ⁎⁎ p b .01, ⁎ p b .05. NSSI = nonsuicidal self-injury; NSSI Group: 0 = Non-NSSI, 1 = NSSI; NA = Negative Affect; PA = Positive Affect; Lag = previous day; Events = Unpleasant events for negative affect and pleasant events for positive affect.

and Demirats (2008). This model allows for the simultaneous modeling of location (mean) and scale (variance) parameters. Mean level, between-person variation, and within-person variations were modeled as a function of NSSI group. Specifically, the Level 1 (within-person) model was yij ¼ b0 j þ rij

ð1Þ

and the Level 2 (between-person) model was b0 j ¼ γ00 þ γ01 ðNSSI Group j Þ þ u0 j

ð2Þ

Hence, the combined MLM model was yij ¼ γ00 þ γ01 ðNSSI Group j Þ þ u0 j þ rij

ð3Þ

where yij is the NA/PA of the ith day for participant j, γ00 is the grand mean for the non-NSSI group, γ10 is the difference between the NSSI group and non-NSSI group (i.e., NSSI group was coded as 0 = non-NSSI,

62

bresin

1 = NSSI), u0j is the between-person residual, and rij is the within-person residual. Further, because we 2 2 assume that u0j ~ N(0, σBw ) and rij ~ N(0, σWn ), both variance components can be modeled as a function of NSSI group using the log link function.
2 ηBw ¼ log σBw ð4Þ ηBw ¼ α00 þ α01 ðNSSI Group j Þ 2 ηW n ¼ log σW n




ηW n ¼ τ00 þ τ10 ðNSSI Group j Þ þ wij

ð5Þ ð6Þ ð7Þ

where α00 is the between-person variance for the non-NSSI group in natural log units, α01 is the difference in the between-person variation between the NSSI group and non-NSSI group in natural log units, τ00 is the within-person variance for the non-NSSI group in natural log units, τ10 is the difference in within-person variation between the NSSI group and non-NSSI group in natural log units, and wij is the within-person residual. Therefore, the significance of α01 and τ10 indicate the presence of group differences in between-person and withinperson variability, respectively. As predicted, across the 14-day period, the NSSI group reported significantly more NA (M = 1.89) than the non-NSSI group (M = 1.60), t(116) = 3.27, p = .001, pseudo-r 2 = .09. Conversely, the NSSI group reported significantly less PA (M = 2.61) compared to the non-NSSI group (M = 2.88), t(116) = -2.34, p = .021, pseudo-r 2 = .05. In terms of variability of NA, the NSSI group did not have significantly more between-person variability (M = .22) than the non-NSSI group (M = .20), t(116) = .5, p = .619, pseudo-r 2 = .05. However, the NSSI group did have significantly more within-person variation (M = .17) than the non-NSSI group (M = .07), t(116) = .5, p = .619, pseudo-r 2 = .33. For PA, there were no significant differences between the two groups for between-person variability (NSSI M = .37, non-NSSI M = .33), t(116) = .33, p = .741, pseudo-r 2 = .04, or within-person variability (NSSI: M = .29, non-NSSI: M = .25), t(116) = .96, p = .339, pseudo-r 2 = .07.

inertia To examine group differences in emotional inertia, the combined model was yij ¼ γ00 þ γ01 ðNSSI Group j Þ
þ γ10 Lag Affect ij
þ γ11 NSSI Group  Lag Affect ij þ u0 j þ rij

ð10Þ

where γ00 and γ01 can be interpreted as before, γ10 is the unstandardized slope relating affect on the previous day to affect on the current day for the non-NSSI group, and γ11 is the difference in this slope between the two groups. In this model, lag affect was centered within-person (Enders & Tofighi, 2008), which removes between-person variation. Hence, the slope can be interpreted as changes from an individual’s mean. Surprisingly, there was not a significant relationship between NA on the previous day and NA on the current day for the non-NSSI group, γ = .06, t(1003) = 1.25, p = .211, pseudo-r 2 = .00. Moreover, this relationship was not significantly different for the NSSI group, γ = -.04, t(1003) = 0.70, p = .482, pseudo-r 2 = .00. There was a significant relation between PA on the previous day and the current day for the non-NSSI group, γ = .29, t(1003) = 6.86, p b .001, pseudo-r 2 = .01. Interestingly, the relation was significantly smaller for the NSSI group, γ = -.15, t(1003) = -2.64, p = .008, pseudo-r 2 = .01, indicating reduced inertia of PA.

reactivity The combined model for reactivity was
yij ¼ γ00 þ γ01 ðNSSI Group j Þ þ γ10 Eventsij þ γ11 ðNSSI Group  Eventsij Þ þ u0 j þ rij

ð11Þ

where the parameters can be interpreted similarly to those from the inertia model. Events were centered within-subject. There was a significant positive relation between unpleasant events and NA, γ = .10, t(1121) = 6.46, p b .001, pseudo-r 2 = .03 for the non-NSSI group. However, individuals with a history of NSSI were not significantly more reactive to unpleasant events, γ = .02, t(1121) = 1.60, p = .290, pseudo-r 2 = .00. The results were similar for pleasant events: there was a significant positive relation between pleasant events and PA, γ = .22, t(1121) = 10.05, p b .001, pseudo-r 2 = .08, but this relation did not differ between groups, γ = .01, t(1121) = .40, p = .690, pseudo-r 2 = .03.

differentiation The analyses for emotional differentiation took a slightly different approach from those above. First, for each participant, the correlation between each PANAS item with all other PANAS items for a subscale were calculated (e.g., distress and afraid, distressed and upset, etc.; Feldman Barrett et al., 2001). These correlations were then transformed using Fisher’s r to Z and then summed within the appropriate affect scale (e.g., NA). The scores were then transformed back to correlations and subtracted

emotion-regulation in participants with a history of nssi from 1 (to make them differentiation) before being compared with an independent samples t-test. For NA, the NSSI group had significantly less differentiation (M = .75) than the non-NSSI group (M = .82), t(112) = -2.08, p = .039, r 2 = .03. The groups did not, however, differ on differentiation of PA, t(112) = -.22, p = .826, r 2 = .00. For completeness, correlations among the measures (averaged across days) are reported in Table 2. Given that there were moderate to large correlations among some of the measures, several follow-up tests were conducted. Not all combinations were examined because in some cases, both affect measures were in the same model (e.g., mean level and variance, mean level and inertia) and in other cases the correlation was not significant (e.g., inertia). Therefore, the focus of the follow-up tests was on the variables that had the largest correlations that were not modeled simultaneously. In separate analyses, the results for emotional differentiation of NA were conducted controlling for mean level, within-person variability, and reactivity. Though the effect of group was only statistically significant when mean level was also in the model, the effect size (partial η 2) was similar across models (range = .02–.05) and similar to the size of the effect without covariates. Hence, in spite of the large correlations among the measures, the results appear to be at least partially unique, though the correlations among the measures reduce the size of the effects.

Discussion The goal of this study was to examine the emotional correlates of NSSI based on multiple assessments of affect in participants’ natural environments. For NA, three of the five indices measured showed significant group differences: mean level, withinperson variance, and emotional differentiation, with individuals with a history of NSSI having greater levels compared to healthy controls. Moreover, even though there were no a priori predictions for PA, the NSSI group evidenced lower mean levels

Table 2

Correlations Between Emotionality Indices Averaged Across Days

1. 2. 3. 4. 5.

Mean Level Variance Inertia Reactivity Differentiation

1

2

3

4

5

__ .65⁎⁎

.73⁎⁎ __ .03 .38⁎⁎ .72⁎⁎

.27⁎⁎ .18⁎

.41⁎⁎ .45⁎⁎

__ .01 .01

.08 __ -.32⁎⁎

.96⁎⁎ .76⁎⁎ -.19⁎ -.44⁎⁎ __

.02 .26⁎⁎ -.83⁎⁎

-

Note. ⁎ p b .05, ⁎⁎ p b .001. Values below the diagonal reflect negative affect and values above the diagonal reflect positive affect. Variance is in natural log units to reduce skew.

63

and lower inertia than the non-NSSI group. These results expand on the previous literature, which has relied on retrospective reports over a long time frame (e.g., in general) aggregated by the participant, by using statistical aggregation of daily affect that is less subject to biases (Robinson & Clore, 2002). The ability to replicate findings from previous studies with different methods should allow researchers to have more confidence in these findings, which are likely to add to the understanding and treatment of NSSI. The results for mean level of NA are largely a replication of previous research, which has found that individuals who engage in NSSI report higher levels of trait NA than individuals who do not engage in NSSI (e.g., Brown, 2008). These results also extend previous EMA studies on NSSI (e.g., Muehlenkamp et al., 2009), which have not compared controls and individuals with a history of NSSI on mean level of daily NA. Counter to previous studies (Brown, 2008), individuals in the NSSI group reported lower levels of PA than individuals in the non-NSSI group. Given the discordant findings in relation to the current literature, it is possible the results for PA are spurious. Taken together, these results suggest that NSSI may be characterized by valence disturbances in NA and PA. However, it should be noted that just because individuals with a history of NSSI had higher levels of NA and lower levels of PA compared to healthy controls, it does not necessarily mean that the difference is clinically meaningful. Though it is also unclear from these data what role NA and PA play in the onset and maintenance of NSSI, it may be useful for clinical interventions to aim to decrease levels of NA (e.g., via emotionregulation skills) and increase PA (e.g., via behavioral activation). There were divergent results across the two indices of intra-individual variability. The NSSI group displayed greater within-person variability of NA compared to the non-NSSI group. However, the groups did not differ on PA variability. Conversely, the two groups did not differ on NA inertia, but the NSSI group had less PA inertia in comparison to the non-NSSI group. Given the conceptualization of the different indices of intra-individual variability, these results may suggest two very different processes for the experience of NA and PA in individuals who engage in NSSI. Because measures of within-person variability are thought to indicate the potential for change (e.g., lability), it could be assumed the individuals who engage in NSSI do not experience persistent NA. Rather, they experience many fluctuations around a high mean level. The lack of significant NA inertia results suggests that these deviations are short-lived

64

bresin

(i.e., affect from one day does not spill over into another). For PA, however, individuals who engage in NSSI do seem to have relatively persistent low PA, in that they have low mean levels with little variability. Moreover, the lower inertia of PA for the NSSI group compared to the non-NSSI group may suggest that even when individuals who engage in NSSI experience PA for one day, it is short-lived, as they are quicker to return to their low PA homeostatic state the next day. Though more work is needed to understand what role intra-individual variability plays in future NSSI engagement, the tentative clinical implications of these results may be a focus on increasing stability of NA, and a focus on increasing the inertia of PA. Counter to theories of NSSI (Linehan, 1993; Nock, 2009) and cross-sectional research (Glenn et al., 2011), individuals with a history of NSSI were not more emotionally reactive to unpleasant or pleasant events compared to controls. These results are generally in line with previous experimental work, which has largely failed to find group differences in emotional reaction to affective stimuli (Glenn et al., 2011; Gratz et al., 2011). The current results may add to this literature in that ecologically valid events still failed to elicit greater emotional reactivity. However, there are two methodological limitations. First, given the small set of events sampled, it is possible that the events individuals who engage in NSSI are sensitive to were not assessed (e.g., intimate partner violence). Second, measuring both affect and events at the day level makes it difficult to know how participants reacted in the moments immediately after the event. It is possible that more assessments within a day are necessary to detect reactivity in individuals who engage in NSSI. Still, given that multiple studies have failed to find that individuals who engage in NSSI are more emotionally reactive than controls, it is possible that these individuals do not have an emotional intensity disturbance. The results for emotional differentiation are concordant with other studies based on crosssectional reports of attention to and clarity of emotion (e.g., Gratz & Roemer, 2008), suggesting that NSSI may be characterized by an emotional disconnection disturbance. These results add to the literature by showing that individuals who engage in NSSI not only perceive themselves to have a lack of clarity of NA, but they also report experiencing many unpleasant emotions at the same time, at least within a day. These findings support the focus of empirically supported treatments for NSSI on teaching clients to identify emotions and use the information that they provide (Gratz & Gunderson, 2006; Linehan, 1993). Further research is needed to understand what role emotional differentiation plays in the manifestation of

NSSI, particularly given the strong correlation between emotional differentiation and other measures.

limitations Though these results make a novel contribution to the literature on NSSI, they should be considered in the context of their limitations. First, as mentioned above, assessments were only taken once a day. Changes in affect occur multiple times a day, and only taking one assessment per day may not have been enough to detect some emotional processes (e.g., reactivity to negative events). Second, though participants reported on daily affect, there was still a retrospective component in that participants had to aggregate experiences throughout the day. Thus, both issues in remembering emotions throughout the day and properly weighting them in aggregation likely affected the results. Moreover, the fact that all the results are self-report leaves room for biases. Future studies assessing other modalities of emotion (e.g., psychophysiology) are needed. Third, both groups in this study are likely heterogeneous; therefore, it is difficult to infer that NSSI status is the cause for group differences. Disorders related to both emotion dysregulation and NSSI (e.g., depression, BPD) may account for the current findings. Replication studies should include more thorough assessments in order to better characterize the groups and rule out such alternative explanations. Fourth, this study used one model of affect that comes with a set of assumptions. It is unclear if these results can be generalized to other models of affect. This is particularly important because low arousal emotions (e.g., sadness) have been implicated in NSSI (Nock et al., 2009). Future studies should assess affect using the full affective circumplex. Finally, though EMA adds a small longitudinal component, this study is largely crosssectional. Because it is unknown if changes in emotional processes (e.g., low mean level NA) come before NSSI or whether they are a product of NSSI, future studies with longitudinal designs are necessary to determine causation. Still, this study provides an incremental step toward identifying facets of emotion regulation that may be important in the onset and maintenance of NSSI. Conflict of Interest Statement The author declares that there are no conflicts of interest.

References Anestis, M. D., Silva, C., Lavender, J. M., Crosby, R. D., Wonderlich, S. A., Engel, S. G., & Joiner, T. E. (2012). Predicting nonsuicidal self‐injury episodes over a discrete period of time in a sample of women diagnosed with bulimia nervosa: An analysis of self‐reported trait and ecological momentary assessment based affective lability and previous suicide attempts. International Journal of Eating Disorders, 45, 808–811.

emotion-regulation in participants with a history of nssi Armey, M. F., Crowther, J. H., & Miller, I. W. (2011). Changes in ecological momentary assessment reported affect associated with episodes of nonsuicidal self-injury. Behavior Therapy, 42, 579–588. Baetens, I., Claes, L., Willem, L., Muehlenkamp, J. J., & Bijttebier, P. (2011). The relationship between non-suicidal self-injury and temperament in male and female adolescents based on child- and parent-report. Personality and Individual Differences, 50, 527–530. Berenbaum, H., Raghavan, C., Le, H. N., Vernon, L. L., & Gomez, J. J. (2003). A taxonomy of emotional disturbances. Clinical Psychology: Science and Practice, 10, 206–226. Berg, K. C., Crosby, R. D., Peterson, C. B., Engel, S. G., Mitchell, J. E., & Wonderlich, S. A. (2013). Facets of negative affect prior to and following binge-only, purge only, and binge/purge events in women with bulimia nervosa. Journal of Abnormal Psychology, 122, 111–118. Bresin, K., & Gordon, K. H. (2013). Changes in negative affect following pain (vs. nonpainful) stimulation in individuals with and without a history of nonsuicidal self-injury. Personality Disorders: Theory, Research, and Treatment, 4, 62–66. Brown, S. A. (2008). Personality and non-suicidal deliberate self-harm: Trait differences among a non-clinical population. Psychiatry Research, 169, 28–32. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage. Butler, A. C., Hokanson, J. E., & Flynn, H. A. (1994). A comparison of self-esteem lability and low trait self-esteem as vulnerability factors for depression. Journal of Personality and Social Psychology, 66, 166. Chapman, A. L., Gratz, K. L., & Brown, M. Z. (2006). Solving the puzzle of deliberate self-harm: The experiential avoidance model. Behavior Research and Therapy, 44, 371–394. Cranford, J. A., Shrout, P. E., Lida, M., Rafaeli, E., Yip, T., & Bolger, N. (2006). A procedure for evaluating sensitivity to within-person change: Can mood measures in diary studies detect change reliably? Personality and Social Psychology Bulletin, 32, 917–929. Ebner-Priemer, U. W., Eid, M., Kleindienst, N., Stabenow, S., & Trull, T. J. (2008). Analytic strategies for understanding affective (in)stability and other dynamic processes in psychopathology. Journal of Abnormal Psychology, 118, 195–202. Enders, C. K., & Tofighi, D. (2008). Centering predictor variables in cross-sectional multilevel models: A new look at an old issue. Psychological Methods, 12, 121–138. Feldman Barrett, L., Gross, J. J., Christensen, T. C., & Benvenuto, M. (2001). Knowing what you’re feeling and knowing what to do about it: Mapping the relation between emotion differentiation and emotion regulation. Cognition and Emotion, 15, 713–724. Fiske, D. W., & Rice, L. (1955). Intra-individual response variability. Psychological Bulletin, 52, 217–250. Glenn, C. R., Blumenthal, T. D., Klonsky, E. D., & Hajcak (2011). Emotional reactivity in nonsuicidal self-injury: Divergence between self-report and startle measures. International Journal of Psychophysiology, 80, 166–170. Gohm, C. L., & Clore, G. L. (2000). Individual differences in emotional experience: Mapping available scales to processes. Personality and Social Psychology Bulletin, 26, 679–697. Gorin, A. A., & Stone, A. A. (2001). Recall biases and cognitive errors in retrospective self-reports: A call for momentary assessments. In A. Baum, T. A. Revenson, & J. E. Singer (Eds.), Handbook of health psychology (pp. 405–413). Mahwah, NJ: Erlbaum.

65

Gratz, K. L. (2001). Measurement of deliberate self-harm: Preliminary data on the Deliberate Self-Harm Inventory. Journal of Psychopathology and Behavioral Assessment, 23, 253–263. Gratz, K. L., & Gunderson, J. G. (2006). Preliminary data on an acceptance-based emotion regulation group intervention for deliberate self-harm among women with borderline personality disorder. Behavior Therapy, 37, 25–35. Gratz, K. L., Hepworth, C., Tull, M. T., Paulson, A., Clarke, S., Remington, B., & Lejuez, C. W. (2011). An experimental investigation of emotional willingness and physical pain tolerance in deliberate self-harm: The moderating role of interpersonal distress. Comprehensive Psychiatry, 52, 63–74. Gratz, K. L., & Roemer, L. (2008). The relationship between emotion dysregulation and deliberate self-harm among female undergraduate students at an urban commuter university. Cognitive Behavior Therapy, 37, 14–25. Hedeker, D., Mermelstein, R. J., & Demirats, H. (2008). An application of a mixed-effects location scale model for analysis of ecological momentary assessment (EMA) data. Biometrics, 64, 627–634. Jahng, S., Wood, P. K., & Trull, T. J. (2008). Analysis of affective instability in ecological momentary assessment: Indices using successive difference and group comparison via multilevel modeling. Psychological Methods, 13, 354–375. Kaess, M., Hille, M., Parzer, P., Maser-Gluth, C., Resch, F., & Brunner, R. (2012). Alterations in the neuroendocrinological stress response to acute psychological stress in adolescents engaging in nonsuicidal self-injury. Psychoneuroendocrinology, 37, 157–161. Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). When more pain is preferred to less: Adding a better end. Psychological Science, 4, 401–405. Kamphuis, J. H., Ruyling, S. B., & Reijntjes, A. H. (2007). Testing the emotion regulation hypothesis among self-injuring females: Evidence for differences across mood states. Journal of Nervous and Mental Disease, 195, 912–918. Kamen, C., Pryor, L. R., Gaughan, E. T., & Miller, J. D. (2010). Affective lability: Separable from neuroticism and the other big four. Psychiatry Research, 176, 202–207. Klonsky, E. D. (2009). The functions of self-injury in young adults who cut themselves: Clarifying the evidence for affect-regulation. Psychiatry Research, 166, 260–268. Larsen, R. J., & Diener, E. (1992). Promises and problems with the circumplex model of emotion. In M. S. Clark (Ed.), Review of personality and social psychology: Emotion (Vol. 13, pp. 25–59). Newbury Park, CA: Sage. Linehan, M. M. (1993). Cognitive-behavioral therapy of borderline personality disorders. New York, NY: Guilford Press. Maclaren, V. V., & Best, L. A. (2010). Nonsuicidal self-injury, potentially addictive behaviors, and the Five Factor Model in undergraduates. Personality and Individual Differences, 49, 521–525. Mayer, J. D., McCormick, L. J., & Strong, S. E. (1995). Mood-congruent memory and natural mood: New evidence. Personality and Social Psychology Bulletin, 21, 736–746. Muehlenkamp, J. J., Engel, S. G., Wadeson, A., Crosby, R. D., Wonderlich, S. A., Simonich, H., & Mitchell, J. E. (2009). Emotional states preceding and following acts of nonsuicidal self-injury in bulimia nervosa patients. Behavior Research and Therapy, 47, 83–87. Nezlek, J. B., & Plesko, R. M. (2003). Affect- and self-based models of relationships between daily events and daily well-being. Personality and Social Psychology Bulletin, 29, 584–596. Nock, M. K. (2009). Why do people hurt themselves? New insights into the nature and functions of self-injury. Current Directions in Psychological Science, 18, 78–83.

66

bresin

Nock, M. K., & Mendes, W. B. (2008). Physiological arousal, distress tolerance, and social problem-solving deficits among adolescent self-injurers. Journal of Consulting and Clinical Psychology, 76, 28–38. Nock, M. K., Prinstein, M. J., & Sterba, S. K. (2009). Revealing the form and functions of self-injurious thoughts and behaviors: A real-time ecological assessment study among adolescents and young adults. Journal of Abnormal Psychology, 118, 816–827. Piasecki, T. M., Hufford, M. R., Solhan, M., & Trull, T. J. (2007). Assessing clients in their natural environment with electronic diaries: Rationale, benefits, limitations, and barriers. Psychological Assessment, 19, 25–43. Polk, E., & Liss, M. (2007). Psychological characteristics of self-injurious behavior. Personality and Individual Differences, 43, 567–577. Pond, R. S., Kashdan, T. B., Dewall, C. N., Savostyanova, A., Lambert, N. M., & Fincham, F. D. (2012). Emotional differentiation moderates aggressive tendencies in angry people: A daily diary analysis. Emotion, 12, 336–337. Pytlik Zillig, L. M., Hemenover, S. H., & Dienstbier, R. A. (2002). What do we assess when we assess a Big 5 trait? A content analysis of the affective, behavioral, and cognitive processes represented in Big 5 personality inventories. Personality and Social Psychology Bulletin, 28, 847–858. Ram, N., & Gestorf, D. (2009). Time-structured and net intraindividual variability: Tools for examining the development of dynamic characteristics and processes. Psychology and Aging, 24, 778–791. Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for an accessibility model of emotional self-report. Psychological Bulletin, 128, 934–960. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161–1178. Russell, J. A., & Carroll, J. M. (1999). On the bipolarity of positive and negative affect. Psychological Bulletin, 125, 1–30. Salovey, P., Mayer, J. D., Goldman, S., Turvey, C., & Palfai, T. P. (1995). Emotional attention, clarity, and repair: Exploring emotional intelligence using the Trait Meta-Mood Scale. In J. W. Pennebaker (Ed.), Emotion, disclosure, and health (pp. 125–154). Washington, DC: American Psychological Association. SAS Institute Inc. (2011). SAS/STAT user’s guide, Version 9.3. Cary, NC: Author. Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological momentary assessment. Annual Review of Clinical Psychology, 4, 1–32.

Singer, J. D. (1998). Using SAS Proc Mixed to fit multilevel models, hierarchical models, and individual growth models. Journal of Education and Behavioral Statistics, 23, 323–355. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occurrence. New York, NY: Oxford Press. Solhan, M. B., Trull, T. J., Jahng, S., & Wood, P. K. (2009). Clinical assessment of affective instability: Comparing EMA indices, questionnaire reports, and retrospective recall. Psychological Assessment, 23, 425–436. Suls, J., Green, P., & Hillis, S. (1998). Emotional reactivity to everyday problems, affective inertia, and neuroticism. Personality and Social Psychology Bulletin, 24, 127–136. von Neumann, J., Kent, R. H., Bellinson, H. R., & Hart, B. I. (1941). The mean square successive difference. The Annals of Mathematical Statistics, 12, 153–162. Wang, L. P., Hamaker, E., & Bergeman, C. S. (2012). Investigating inter-individual differences in short-term intra-individual variability. Psychological Methods, 17, 567–579. Watson, D. (2000). Mood and temperament. New York, NY: Guilford Press. Watson, D., & Clark, L. A. (1994). The PANAS-X: Manual for the Positive and Negative Affect Schedule—Expanded form. Unpublished manuscript, The University of Iowa. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Watson, D., O'Hara, M. W., & Stuart, S. (2008). Hierarchical structures of affect and psychopathology and their implications for the classification of emotional disorders. Depression and Anxiety, 25, 282–288. Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219–235. Weinberg, A., & Klonsky, E. D. (2011). The effects of self-injury on acute negative arousal: A laboratory simulation. Motivation and Emotion, 36, 242–254. Whitlock, J., Eckenrode, J., & Silverman, D. (2006). Selfinjurious behaviors in a college population. Pediatrics, 117, 1939–1948.

R E C E I V E D : January 19, 2013 A C C E P T E D : September 19, 2013 Available online 27 September 2013