Attentional Capture by Angry Faces in Girls Who Self-Injure: Evidence from Steady State Visual Evoked Potentials

Abstract

Background:

Non-suicidal self-injury (NSSI) is a significant public health concern, not only because of the personal and social cost of the behavior itself, but also because it increases risk for future self-injurious behaviors, including suicide attempts. NSSI is increasingly prevalent during adolescence, which highlights the need for research aimed at identifying modifiable risk factors that can be targeted to reduce future risk. Building from theoretical models that highlight interpersonal processes, this study examined whether adolescents with an NSSI history exhibit greater difficulty inhibiting attention to emotionally-salient interpersonal stimuli (face), indexed via steady state visual evoked potentials (SSVEPs), which provide a direct neural index of the ability to inhibit attention to task-irrelevant stimuli.

Methods:

Adolescent girls aged 13–17 with (n = 26) and without (n = 28) an NSSI history completed a change-detection computer task during which frequency-tagged SSVEPs were used to assess adolescents’ ability to inhibit attention to affectively-salient stimuli from spatially superimposed targets.

Results:

Compared to adolescents with no NSSI history, adolescents with NSSI demonstrated difficulty inhibiting attention to angry adult faces.

Conclusions:

These findings underscore specific deficits in attentional filtering among girls with an NSSI history, which, if replicated and extended, could be a promising intervention target for reducing risk for future NSSI.

Non-suicidal self-injury (NSSI), defined as intentional self-injury without the intent to die, is alarmingly common in adolescents, with a lifetime prevalence rate of approximately 18% (Swannell et al., 2014). NSSI is a transdiagnostic phenomenon, occurring across various psychiatric disorders as well as in the absence of diagnosis (Nock et al., 2006). In addition to the distress and physical damage associated with NSSI itself, NSSI increases risk for future self-injury and suicide attempts (SA; Ribeiro et al., 2016). Moreover, rates of NSSI increase dramatically during adolescence, highlighting this period as a key window of risk (Barrocas et al., 2012).

Models of NSSI underscore the interpersonal and emotion regulatory components involved in self-injury. According to the Benefits and Barriers Model (Hooley & Franklin, 2018), engagement in NSSI is determined by both benefits of the behaviors (e.g., improved affect, self-punishment, peer group affiliation, communication) and factors that dissuade the behavior (e.g., desire to avoid pain/blood, positive self-view, social norms). This model builds upon earlier models, including the Four Functions Model and emotion regulation models, that emphasize similar interpersonal (e.g., increasing attention, reducing conflict) and emotion regulation (Hasking et al., 2016; Nock & Prinstein, 2004; Selby & Joiner, 2009) benefits. The Four Functions Model underscores the potential role of social environment in NSSI, presuming individuals with undesirable or negative social environments are more motivated to seek perceived interpersonal and regulatory benefits of NSSI than individuals without such social stress (Nock & Prinstein, 2004). Together, these theories suggest those at risk for NSSI exhibit heightened reactivity to stress, perhaps particularly interpersonal stress, and use NSSI as a maladaptive method of regulating their social environment or emotions.

Importantly, compared to individuals at other developmental stages, adolescents report greater emotional reactivity to social stressors such as interpersonal conflict (Larson & Ham, 1993). Moreover, parent-related stressors (e.g., parental criticism, perceived lack of support, parent-child relationship conflict) are associated with youth NSSI (see James & Gibb, in press, for a review). Consistent with conceptual models, this research suggests parental criticism or anger may be particularly salient for adolescents who self-injure. Therefore, examining adolescents’ processing of, and responses to, threat-relevant interpersonal cues may aid identification of modifiable risk factors that can be targeted to reduce future risk.

The ability to filter out task-irrelevant emotional stimuli (e.g., distracting threat-relevant emotional faces) and attend, instead, to goal- or task-relevant stimuli may be one specific component of attention critically involved in NSSI risk. Whereas stimulus-driven attention to emotional distractors is associated with heightened activation in limbic regions, attention to goal- or task-relevant stimuli is driven by regions within the prefrontal cortex (PFC; e.g., anterior cingulate cortex; (Dominquez Borras & Vuilleumier, 2013; Pessoa et al., 2013), which play an important role in downregulating limbic reactivity and resolving emotional interference in attentional control (Etkin et al., 2011; Kanske & Kotz, 2011; Stevens et al., 2011). In addition, interference models of affect-biased attention underscore the potential role of connectivity between fronto-limbic areas and lower-order sensory areas (e.g., visual cortex) in resolving conflict in attention to multiple competing stimuli (i.e., by allocating attention selectively to stimuli; Carretié, 2014; Deweese et al., 2016). To this end, attentional biases can be assessed using steady state visual evoked potentials (SSVEPs), derived from EEG, which provide a direct neural index of attentional capture for salient emotional distractors relative to a neutral visual target (Wieser et al., 2016). Neurophysiologically, SSVEPs are large-scale oscillatory field responses, with frequency signatures identical to a rhythmically modulated visual stimulus. The high signal-to-noise ratio of SSVEPs maximizes utility in clinical populations, particularly when it may not be feasible to collect large samples. This driven oscillation can be further exploited by using a frequency-tagging approach, which presents two or more competing stimuli at distinct frequencies as a way of segregating neural responses. In attention studies, this driven oscillation can be used to discriminate attention to competing stimuli that are spatially and temporally overlapping. Since the target and distractor stimuli are modulated at different frequencies (e.g., 12Hz vs. 15Hz), SSVEPs provide a direct measure of competition effects between attention to goal-directed neutral targets and affective distractors by quantifying neuronal population engagement in the visual cortex specific to each stimulus (Wieser et al., 2016). Moreover, SSVEPs are thought to reflect re-entrant modulation from neural networks involving fronto-limbic regions (Petro et al., 2017; Pourtois et al., 2013).

SSVEPs have been used to document attentional biases associated with psychopathology in youth (Pei et al., 2014; Silberstein et al., 2016) and adults (McTeague et al., 2018; Woody et al., 2017); however, no studies have used SSVEPs to examine attentional filtering associated with NSSI. Nonetheless, there is preliminary evidence from neuroimaging studies for differences in neural activation in fronto-limbic regions in response to social exclusion in adolescents with and without NSSI. Specifically, during a social exclusion paradigm, adolescents with an NSSI history demonstrated increased activation in the PFC (Groschwitz, Plener, Groen, Bonenberger, & Abler, 2016a) and the ventral anterior cingulate cortex (vACC; Brown et al., 2017) relative to adolescents with no such history. The PFC is implicated in the processing, interpretation, and regulation of emotions (Dewall et al., 2012; Eisenberger et al., 2003; Goldin et al., 2008; Gunther Moor et al., 2012; Ochsner et al., 2004), suggesting adolescents who self-injure may require greater recruitment of PFC resources when attempting to regulate affect associated with interpersonal stress. Notably, adolescents who self-injure also reported feeling more helpless in response to a social exclusion task than adolescents who do not, despite exhibiting increased PFC activation (Groschwitz et al., 2016b). The ACC is one of the key brain regions involved in responsiveness to social feedback (Somerville et al., 2006), and the vACC is specifically implicated in social and emotional processes, including the evaluation of social feedback such as rejection. Accordingly, enhanced activation in the vACC among adolescents who engage in NSSI relative to those who do not suggests heightened sensitivity to socially threatening situations such as social exclusion as well as overvaluation of social feedback. These studies, coupled with the well-established link between interpersonal stress and NSSI, indicate adolescents who self-injure may experience difficulty inhibiting attention to salient interpersonal stimuli, particularly stimuli indicative of social threat (e.g., angry faces).

The goal of this project was to examine attentional biases for facial displays of emotion in adolescent girls with and without an NSSI history. Specifically, this study examined whether girls with recurrent NSSI exhibit greater difficulty inhibiting their attention to salient emotional distractors than girls who have never self-injured. To reduce the heterogeneity of our sample, we focused exclusively on girls given that girls exhibit both increased reactivity to interpersonal stress and greater difficulty inhibiting attention to negatively-valenced stimuli (Rose & Rudolph, 2006; Whittle et al., 2011). We hypothesized girls with an NSSI history would exhibit greater difficulty inhibiting their attention to threat-relevant emotional distractors (i.e., angry faces) than girls with no NSSI history.

Materials and Methods

Participants

Participants were 58 adolescent girls recruited from the community based on girls’ NSSI history. Twenty-six girls had a history of recurrent NSSI and 32 had no NSSI history. Four girls from the No NSSI group were excluded based on performance on the SSVEP task (three performed worse than chance on the task’s behavioral component; one was an outlier on the task’s neural component). The final sample comprised 26 girls with recurrent NSSI and 28 with no NSSI history. To qualify for the NSSI group, adolescents had to report recurrent NSSI (specifically, recurrent episodes of cutting). Adolescents who engaged in other forms of NSSI (e.g., burning) in addition to cutting were also included in this group. Adolescents in the No NSSI group could not report any history of NSSI. The average age of adolescents in the sample was 15.25 years (SD = 1.32, Range = 13–17). In terms of race, 81.5% identified as Caucasian/White, 5.6% identified as African American/Black, 3.7% identified as Asian, and 9.3% identified as biracial or from another racial group. Additionally, 9.3% identified as Hispanic. The median annual family income was $70,001-$75,000. Demographic and clinical characteristics of the two groups are presented in Table 1.

Table 1.

Descriptive statistics

Measure	NSSI (n = 26)	No NSSI (n = 28)	r effect size
Adolescent Age	15.23 (1.39)	15.26 (1.28)	−.01
Adolescent Race (% Caucasian)	76.9%	88.5%	−.29*
Household Income	$60,001–$65,000	$80,001–$85,000	−.14
Adolescent Current MDD Dx	38.5%	7.1%	.38**
Adolescent Lifetime MDD Dx	88.5%	46.4%	.45**
CDI	19.58 (8.63)	10.86 (7.05)	.49**
MASC	62.63 (16.14)	47.98 (14.60)	.44**
BPFSC	37.00 (7.13)	26.29 (6.19)	.63**
Adolescent Current SI	38.5%	10.7%	.32*
Adolescent Lifetime SI	84.6%	39.3%	.47**
Adolescent Lifetime SA	15.6%	3.6%	.20
BDI-II	12.69 (7.80)	10.18 (10.98)	.23
BAI	10.40 (8.96)	8.86 (9.53)	.10
Mother Lifetime SA	23.1%	19.2%	.07

Note. MDD = Major Depressive Disorder. Dx = diagnosis. CDI = Children’s Depression Inventory. MASC = Multidimensional Anxiety Scale for Children. BPFSC = Borderline Personality Features Scale for Children. SI = suicidal ideation. BDI = Beck Depression Inventory. BAI = Beck Anxiety Inventory. SA = suicide attempt.

^*p < .05.

^**p <.01.

Measures

NSSI

Adolescents’ NSSI was first identified during a phone screen with the adolescents’ mothers, in which mothers were asked, “Has your daughter ever hurt herself without wanting to die, such as cutting or burning herself?” This screening question was followed by questions about method, recency, and frequency of NSSI. During the laboratory visit, adolescents’ lifetime NSSI was further assessed using the Self-Injurious Thoughts and Behaviors Interview (SITBI; Nock, Holmberg, Photos, & Michel, 2007a). Adolescents were asked “Have you ever actually engaged in NSSI?” where NSSI is defined as “purposely hurting yourself without wanting to die.” Twenty-six adolescents endorsed recurrent NSSI. The median reported number of NSSI episodes was twelve (Range: 2–400) with 15 adolescents reporting past year NSSI (Range of time since last NSSI episode: 5 days-4 years). All adolescents endorsed cutting themselves on at least two occasions; eight (31%) also endorsed additional NSSI methods (e.g., burning, hitting).

Change-Detection Task

Adolescents’ ability to inhibit attention toward emotional distractors relative to a task-relevant neutral stimulus was assessed using a change-detection paradigm (Figure 1; cf. Woody et al., 2017). During the task, adolescents sat 65cm away from a 27-inch computer monitor with a vertical refresh rate of 60Hz. In each experimental trial, a picture of an emotional facial expression was presented at the center of the screen for 5000ms. A semi-transparent Gabor patch was superimposed over the face. Stimuli were flickered at distinct frequencies (i.e., 12 and 15Hz) to evoke SSVEPs frequency-tagged to a given stimuli. Faces were flickered at a frequency of 12Hz to evoke SSVEPs frequency-tagged to the face, and Gabor patches were flickered at 15Hz to evoke SSVEPs frequency-tagged to the Gabor. The flickering frequencies of the face and Gabor were counterbalanced across participants within each group. Trials ended with a variable (2000–4000ms) intertrial interval. On a random 50% of the trials, the Gabor rotated 10° during the trial before rotating back 600ms later. On rotation trials, the Gabor rotated clockwise half of the time and counterclockwise half of the time. At the end of each trial, adolescents were instructed to indicate if the Gabor shifted orientation by clicking the mouse. Following exclusion of three participants in the No NSSI group with less than chance performance, there were no significant group differences in accuracy in detecting the Gabor shift (lowest p = .30).

Figure 1.

Schematic Layout of the Change Detection Task.

Stimuli included 90 pictures of 30 actors (15 males and 15 females) from the Karolinska Directed Emotional Faces stimulus set (KDEF;Flykt, Lundqvist, Flykt, & Öhman, 1998). Pictures portrayed actors gazing directly at the camera and each of the actors was shown in an angry, sad, or happy expression. All pictures were converted to grayscale and equated for luminance, trimmed of all non-facial features (i.e., hair, neck), cropped into an oval, and displayed against a gray background. Faces were randomly selected without replacement during each trial. Consistent with previous research (Woody et al., 2017), we focused upon proportion scores reflecting attentional capture toward the emotional face relative to the Gabor in our analyses (rather than a neutral face contrast) to reduce the possible influence of interpretation biases on our effects. The Gabor provides an unambiguously neutral contrast whereas a neutral face may be perceived negatively by participants, particularly given preliminary evidence adolescents who engage in NSSI interpret neutral faces more negatively than adolescents who do not (Lutz et al., 2020).

EEG Data Recording and Analysis

Continuous electroencephalogram (EEG) was recorded during the change-detection task using a custom cap and the BioSemi ActiveTwo system. The signal was pre-amplified at the electrode with a gain of 16x; the EEG was digitized at 24-bit resolution with a sampling rate of 512Hz using a low-pass fifth-order sinc filter with a half-power cutoff of 104Hz. Recordings were taken from 34 scalp electrodes based on the 10/20 system. Two additional electrodes, an active Common Mode Sense (CMS) and a passive Driven Right Leg (DRL) electrode were used in the study. Raw EEG was recorded relative to CMS. The CMS/DRL electrodes replaced the ground for recordings through a feedback loop which drove the average potential of the subject (i.e. the Common Mode voltage) as close as possible to the “zero” ADC reference voltage in the AD-box (please see http://www.biosemi.com/faq/cms&drl.htm for further details). Electrooculogram was recorded from four facial electrodes.

Off-line EEG analysis was performed using EEGLAB software in MATLAB. All data were band-pass filtered with cutoffs of 0.01Hz and 40Hz and re-referenced to the average of the left and right mastoid electrodes. EEG data was processed using both artifact rejection and correction. Large and stereotypical ocular components were identified and removed using independent component analysis (ICA) scalp maps (Jung et al., 2001). Epochs were extracted from raw EEG and included 1000ms pre-stimulus onset and 5000ms post-stimulus onset, with the interval from −1000ms to 0ms serving as the baseline for each trial. Epochs with large artifacts (greater than 200μV) were excluded from analysis. The number of rejected trials for each participant in each emotion ranged from 0 to 15 (0%−50% of total trials per emotion) with an average of 4.28 happy trials, 4.19 sad trials, and 4.46 angry trials rejected. Rejection rates did not significantly differ across emotions or based on NSSI history for any emotion. Artifact-free EEG trials were averaged in the time domain prior to extraction of frequency spectra. A SSVEP time domain segment and average frequency power spectrum are depicted in Figure 2.

Figure 2.

A steady-state visual evoked potential (SSVEP) time domain segment (800 ms to 5000 ms post-stimulus onset) is depicted in (A) for both single subjects (thin grey lines) as well as the grand mean (thick red line). The data is from electrode Oz. These time domain segments were submitted to a Fourier transform. The average power spectrum is shown in inset (B), depicting a number of clear spectral peaks. Note that in addition to the discrete peaks matching the input frequency of the face (f1: 12 Hz) and the Gabor patch (f2: 15 Hz) there are also various intermodulation frequencies (f2-f1: 3 Hz) and higher harmonics of the fundamental frequencies (e.g., 24 Hz, 30 Hz) and intermodulation ones (e.g., 6 Hz, 9 Hz, 18 Hz, 21 Hz).

Mean SSVEP amplitudes for each stimulus were calculated separately for each facial expression (angry, sad, happy) during the change-detection task by conducting a fast Fourier transform at the occipital electrode sites (O1, Oz, O2) for segments extracted from 800ms to 5000ms post-stimulus onset. This epoch was chosen to exclude the initial nonstationary components of steady-state entrainment from the power spectrum. The split-half reliabilities of SSVEPs to the emotional faces and Gabor were good to excellent with Guttman split-half coefficients ranging from 0.84 to 0.99.

To directly examine neural competition between the overlapping stimuli, a proportion score was created to reflect the relative visual cortical activity evoked by each facial expression (angry, sad, happy) relative to the concurrent Gabor patch. Consistent with previous research (Woody et al., 2017), SSVEP amplitudes to faces and the Gabor were first transformed to T scores and then proportion scores were calculated for each face type: T-SSVEP_face/(T-SSVEP_face + T-SSVEP_Gabor). Proportion scores above .50 reflect greater attentional capture toward the emotional face relative to the Gabor, whereas scores below .50 indicate greater attentional capture toward the Gabor relative to the emotional face. Scalp topographies illustrating prototypical SSVEP responses centralized around the occipital electrodes (O1, Oz, O2) were obtained (Figure 3).

Figure 3.

Scalp topographies (spline interpolated) for SSVEPs to Gabor patches and faces, as a function of NSSI history collapsed across all emotions.

Symptoms and Diagnoses

To more fully characterize the sample, adolescents’ current and lifetime episodes of major depressive disorder were assessed using the Schedule for Affective Disorders and Schizophrenia for School-Age Children – Present and Lifetime Version (K-SADS-PL; Kaufman et al., 1997). Adolescents’ depression and anxiety symptoms were assessed using the Children’s Depression Inventory (CDI; Kovacs, 1981) and the Multidimensional Anxiety Scale for Children (MASC; March, Parker, Sullivan, Stallings, & Conners, 1997), which demonstrated excellent internal consistency (αs = .91 and .91, respectively). Adolescents’ borderline personality disorder traits were assessed using the Borderline Personality Features Scale for Children – Short Version (BPFSC-11; Sharp, Steinberg, Temple, & Newlin, 2014), which demonstrated good internal consistency (α = .85). Mothers’ depression and anxiety symptoms were assessed using the Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996) and the Beck Anxiety Inventory (BAI; Steer & Beck, 1993), which both exhibited excellent internal consistency (αs = .92 and .92, respectively). Adolescents’ and mothers’ current and lifetime suicidal ideation (SI) and behaviors were assessed using the Self-Injurious Thoughts and Behaviors Interview (SITBI; Nock, Holmberg, Photos, & Michel, 2007a).

Procedure and Ethical Considerations

Participants were fully informed about the process and purpose of the study. After providing written consent/assent, mothers completed a series of interviews and questionnaires while adolescents completed the change detection task (Woody et al., 2017) during which SSVEPs were collected. Next, adolescents completed a series of interviews and questionnaires. Adolescents were compensated $30, mothers were compensated $25, and each dyad was provided a $10 gas card. This project was approved by the University’s Institutional Review Board.

Results

A description of how we handled missing questionnaire data is provided in the Supplement. To examine attentional filtering of emotional faces in girls with and without an NSSI history, we conducted a 2 (NSSI: yes, no) × 3 (Emotion: angry, sad, happy) repeated measures ANOVA with SSVEP competition scores serving as the dependent variable. SSVEP competition scores are presented in Figure 4. Although the main effects of NSSI group, F(1, 52) = 3.27, p = .08, η_p² =.06, and emotion, F(2, 104) = .10, p = .91, η_p² < .01, were both nonsignificant, there was a significant NSSI × Emotion interaction, F(2, 104) = 3.56, p = .03, η_p² =.06.

Figure 4.

Steady-state visual evoked potential (SSVEP) competition scores as a function of emotion and NSSI. Competition scores greater than 0.50 reflect heightened attentional capture for the emotional face relative to the Gabor. Competition scores less than 0.50 indicate heightened attentional capture for the Gabor relative to the emotional face. Error bars represent standard error of the mean.

To examine the form of this interaction, we assessed NSSI group differences in adolescents’ SSVEP competition scores for each emotion separately. Although there were no significant group differences for sad, F(1, 52) = 2.83, p = .10, η_p² =.05, or happy, F(1, 52) = 2.10, p = .15, η_p² =.04, faces, there was a significant NSSI group difference in SSVEP competition scores for angry faces, F(1, 52) = 4.69, p = .04, η_p² =.08. As hypothesized, adolescents in the NSSI group exhibited more difficulty inhibiting attention to angry faces than adolescents with no NSSI history. Scalp topographies for SSVEPs to Gabor patches and emotional faces are presented in the Supplement (Supplemental Figure 1). The group difference was maintained after statistically controlling for the influence of adolescents’ current symptoms of depression, F(1, 51) = 7.34, p = .01, η_p² =.13, anxiety, F(1, 51) = 3.94, p = .05, η_p² =.07, borderline personality traits, F(1, 51) = 8.32, p = .01, η_p² =.14, current SI, F(1, 51) = 5.85, p = .02, η_p² =.10, and lifetime SA, F(1, 51) = 4.15, p = .05, η_p² =.08. It was also maintained after statistically controlling mothers’ current symptoms of depression, F(1, 51) = 4.36, p = .04, η_p² =.08, anxiety, F(1, 51) = 4.70, p = .04, η_p² =.08, and history of SA, F(1, 51) = 5.15, p = .03, η_p² =.09. These results indicate that our result was not due simply to current psychopathology in the girls or their mothers.

Discussion

The goal of this study was to examine attentional biases for emotional faces in adolescent girls with and without an NSSI history. We examined adolescents’ ability to inhibit attention to task-irrelevant angry, sad, and happy faces. Supporting our hypothesis, girls with a recurrent NSSI history, compared to those without an NSSI history, exhibited greater difficulty inhibiting attention to angry faces. The group difference was maintained after statistically controlling for the influence of adolescents’ current symptoms of depression, anxiety, traits of borderline personality disorder, SI, and history of SA, as well as their mothers’ current symptoms of depression, anxiety, and history of SA, suggesting this finding was at least partially independent of the influence of these variables.

Despite substantial evidence that different forms of psychopathology are characterized by disorder specific-information processing biases (e.g., attentional biases for sad faces in depression or threat-relevant stimuli such as angry or fearful faces in anxiety disorders; for reviews, see Bar-Haim et al., 2007; Gotlib & Joormann, 2010), the current study is the first to examine attentional biases for emotional faces in individuals who self-injure, a behavior present across myriad disorders. That girls who self-injure demonstrate enhanced attentional capture for angry faces supports theories that highlight negative interpersonal influences, such as punishment or conflict, in NSSI risk (Nock & Prinstein, 2004). This result suggests threat-relevant stimuli (e.g., angry faces) may be especially salient for these adolescents.

Not only do these results contribute foundational information about the general presence of attentional biases for angry faces in adolescents who self-injure, they also help to clarify how a precise attentional component and marker of cognitive control, namely attentional filtering, may be associated with risk for NSSI in adolescent girls. This study focused on competition effects between “bottom-up” attention toward task-irrelevant emotional distractors (i.e., angry, sad, happy faces) and “top-down” attention toward neutral stimuli involved in a cognitive task (i.e., detecting shifts in orientation of a Gabor patch). This specific question may be of particular relevance during adolescence when prevalence rates of NSSI are highest (Barrocas et al., 2012), and prefrontal regions involved in “top-down” attention toward task-relevant stimuli are still developing and may be more susceptible to competition from “bottom-up” attention toward emotional distractors driven by the limbic regions (Casey et al., 2019; Ladouceur, 2012). Moreover, given that attention allocation is one form of emotion regulation (Ma et al., 2017; Todd et al., 2012), these findings may indicate girls who engage in recurrent NSSI exhibit deficits in effective emotion regulation (e.g., impaired ability to disengage attention from threat-relevant facial displays of emotion).

The current study demonstrated several strengths, including the use of SSVEPs to measure a specific component of attentional capture and direct neural index of competition between attention toward task-irrelevant emotional distractors and goal-oriented stimuli involved in a neutral cognitive task. As a direct neural index, SSVEPs are advantageous in this line of inquiry, overcoming the limitations of other behavioral measures of attention (e.g., dot probe; Thigpen, Gruss, Garcia, Herring, & Keil, 2018). Additionally, the study was enhanced by the inclusion of adolescents in both groups with current and past psychopathology. Although the two groups still differed on some of these variables, this decision was made in to increase the specificity of any significant relations to NSSI rather than other sources of distress, and results were maintained when we statistically controlled for these influences in our model.

Nonetheless, several limitations provide important direction for future research. The primary limitation of the current study is its cross-sectional design. Although our hypotheses were motivated by theories proposing mechanisms of risk for NSSI, a prospective design is necessary to examine attentional biases to interpersonal stimuli as potential predictors of first-onset or recurrence of NSSI. Second, our sample was limited to girls. Although the exclusion of boys increased the homogeneity of the sample and specificity of the findings, it will be important for future studies to determine the generalizability of these results to boys. Third, adult faces were used as stimuli for the change detection task. Therefore, to more fully understand the nature of this attention bias, research including adolescent stimuli is necessary to disentangle the roles of parent- and peer-related interpersonal influences in NSSI.

Finally, although the high signal-to-noise ratio is a known strength of SSVEPs, the sample size of the current study was relatively small and may have limited power for some analyses (e.g., the main effect of group for SSVEP competition scores, which was a nonsignificant trend). Therefore, studies with larger sample sizes are needed to more definitively test whether NSSI risk is characterized specifically by deficits in attentional filtering of facial displays of anger rather than negatively-valenced interpersonal stimuli (e.g., angry, fearful, and sad faces) or emotional stimuli more broadly.

Conclusion

The current study helps to clarify how one specific factor involved in processing of, and responding to, interpersonal stimuli may increase risk for NSSI in adolescent girls. Adolescents with recurrent NSSI were characterized by difficulties inhibiting attention to threat-relevant facial displays of emotion as indexed via SSVEPs, a direct neural measure of attentional capture. If replicated and extended, our results could clarify a specific deficit in processing of interpersonal stimuli that may increase future risk. Such findings could also delineate precise targets for prevention and intervention efforts, which could interrupt a broader mechanism of risk (e.g., emotion regulation difficulties), potentially reducing future engagement in NSSI and SA.

Data Availability Statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.