Implicit Assessment of Self-Injury Related Outcome Expectancies: A Comparison of Three behavioural Tasks

Abstract

According to Social Cognitive Theory, the anticipated consequences of a behaviour (outcome expectancies), influence the likelihood of engaging in a behaviour. Results from self-report studies suggest that people who have self-injured expect self-injury will regulate emotions while people who have never self-injured expect self-injury to result in pain. In this study we trialled three experimental tasks measuring implicit self-injury related outcome expectancies. 150 Australian university students aged 18–45 (M = 21.45, SD = 3.84) completed the experimental tasks (Sentence Completion Task, Implicit Association Tests, Covariation Bias Task) within a laboratory setting. Results revealed that implicit associations with affect regulation, pain, and communication differentiated people according to self-injury history in the sentence completion task. The strength of implicit associations with affect regulation also predicted the recency of self-injury. People who had self-injured, but not in the past 12 months appeared to have a bias towards associating images of self-injury and neutral words when compared to people who had recently self-injured. Implicit associations, as measured by the Implicit Association Tests did not significantly differentiate participants by self-injury history. Results suggest that the sentence completion task could further research and theoretical understanding of the role of implicit outcome expectancies in facilitating self-injury.

Keywords

self-injury outcome expectancy implicit NSSI

Nonsuicidal self-injury (NSSI), the deliberate damage to one’s own body without suicidal intent, includes behaviours such as cutting, burning, and biting the skin, self-battery, and wound interference (International Society for the Study of Self-Injury, 2018). Culturally sanctioned behaviours such as tattoos and piercings are excluded from the definition, as are behaviours which are a symptom of another diagnosis (e.g. trichotillomania, excoriation). NSSI is prevalent in community samples of adolescents (approximately 17%) and young adults (approximately 13%; Swannell et al., 2014). Approximately 20% of university students report having engaged in self-injury at least once in their life (Swannell et al., 2014). Among university students, NSSI is associated with poor academic outcomes (Kiekens et al., 2016), diagnosis of mental illness (Kiekens, Hasking, Claes, et al., 2018), and suicidal thoughts and behaviours (Kiekens, Hasking, Boyes, et al., 2018). A recent study also suggests that the first two years at university are high risk for the onset of NSSI (Kiekens et al., 2019). Although NSSI is a distinct behaviour from suicide and suicide attempt, people who self-injure are up to five times more likely to engage in suicidal thoughts and behaviours than people with no history of NSSI (Kiekens, Hasking, Boyes, et al., 2018; Whitlock et al, 2013).

People most often report that they self-injure as a way to regulate intense and unwanted emotions (Taylor et al., 2018). Supporting this, empirical studies have found that people who self-injure tend to have more difficulties regulating their emotions (Wolff et al., 2019) and experience a reduction in negative affect and increase in positive affect after engaging in NSSI (Rodriguez-Blanco et al., 2018). As such, models used to explain the onset and maintenance of NSSI have focussed on the roles of the experience and regulation of emotion (Bentley et al., 2014; Chapman et al., 2006; Nock, 2009; Selby & Joiner, 2009). However, less attention has been paid to how thoughts and beliefs about self-injury may play a role in facilitating NSSI.

Within Social Cognitive Theory, Bandura (1986) proposed that the expected consequences of a behaviour (outcome expectancies) influence the likelihood of engaging in the behaviour. Expecting a positive outcome increases the likelihood of engagement, while expecting a negative outcome decreases this likelihood. Outcome expectancies can be influenced by personal experience, observing the consequences of others’ actions, and imagining possible outcomes (Bandura, 1986). As such, outcome expectancies can be held about behaviours that have never previously been engaged in. The recently proposed Cognitive-Emotional Model of NSSI considers the role of specific cognitions about self-injury, including outcome expectancies, alongside the experience and regulation of emotion in facilitating NSSI (Hasking et al., 2017). Only recently have outcome expectancies related to NSSI been explored. Hasking and Boyes (2017) developed the NSSI Expectancies Scale to measure NSSI outcome expectancies, identifying five anticipated consequences of NSSI (affect regulation, negative social outcomes, communication or care from other people, physical pain, and negative self-beliefs). Subsequent studies have found that people with a history of self-injury have stronger expectations that NSSI will result in affect regulation while people who have never self-injured believe self-injury will results in physical pain (Dawkins et al., 2018; Hasking, 2017; Hasking & Boyes, 2017).

Self-report measures require a certain degree of insight and as such can be biased by participant’s lack of insight or by social desirability (Marissen et al., 2005). In contrast, implicit associations are fast, do not require insight, and can be measured indirectly using experimental measures (Wiers & Stacy, 2006). As such, implicit measures avoid the influence of social desirability and can tap into underlying beliefs or associations that people are unaware they hold. Implicit associations often contribute to the prediction of behaviours over and above self-reported associations and can pick up changes in mental associations often before conscious awareness of a change (Hahn & Gawronski, 2015).

Several tasks have been developed to assess implicit associations with outcome expectancies related to other health risk behaviours (e.g. drinking, smoking). These include the Expectancy Task (ETASK; Palfai et al., 1997; Read & Curtin, 2007) and Implicit Association Tests (IAT; Jajodia & Earleywine, 2003). The ETASK is a sentence completion task, where participants indicate whether sentences related to outcome expectancies (e.g. Alcohol helps me… RELAX) are true or false for them. Faster reaction times (RT) are thought to indicate stronger implicit associations with the related outcome expectancy. Implicit association tests are used to measure the strength of associations between different stimuli and associated outcomes. It is expected that participants will have faster reaction times in conditions where associated stimuli and outcomes have paired response keys, when compared to conditions where paired stimuli and outcomes are not associated. Previous studies using Implicit Association Tests (IAT) to explore associations between NSSI and the experience of emotional relief have found that people with a history of self-injury implicitly associate NSSI with relief (over disgust; Gratz et al., 2016, 2018). Within the learning literature, implicit associations between stimuli and anticipated outcomes are often measured by assessing whether a covariation bias is present (Hermann et al., 2004). A covariation bias is the tendency to overestimate the association between a stimulus and positive/negative outcomes (Amrhein et al., 2005). These tasks could easily be adapted to measure implicit associations with NSSI-related outcome expectancies and to examine whether implicit expectancies differ between people with and without a history of NSSI.

Before evaluating whether implicit associations with NSSI related outcome expectancies have incremental predictive value, above that provided by explicit measures, we first need to determine which measures reliably differentiate participants based on their history of NSSI. The aim of this study was to measure implicit NSSI outcome expectancies using a variety of experimental methods (sentence completion task; implicit association tests; covariation bias task) and determine which best differentiates people who have never self-injured from people who have a past history of self-injury and people who have recently engaged in NSSI. Self-report studies have found that expectations regarding the ability of NSSI to either regulate affect and/or induce physical pain are most salient in differentiating people with a history of NSSI from those without (Dawkins et al., 2018; Hasking, 2017; Hasking & Boyes, 2017). We expect a similar pattern of findings to emerge when assessing outcome expectancies with implicit tasks.

Method

Participants

Participants were 150 Australian university students aged 18-45 years (M = 21.45, SD = 3.84). The majority of participants were female (n = 109, 72.7%), 40 were male (26.7%), and 1 identified as agender (.7%). One (.7%) participant identified as Aboriginal or Torres Strait Islander. The majority of participants (138, 92%) were studying full-time and, as is customary in Australia, living at home with their parents or family members (n = 109, 73%).

Experimental tasks

Each experimental task was programmed in DMDX Version 5.1.3.4, a millisecond accurate display system used to measure reaction times to visual and auditory stimuli (Forster, 2002). A Dell 2.1.5 inch LCD screen (resolution: 1920 × 1080) was used to display images at 220 pixels per inch and words were displayed at the default font size for DMDX.

Sentence completion task

The ETASK developed by Palfai et al. (1997) to measure implicit alcohol related outcome expectancies, was modified to measure participant’s implicit associations with five NSSI outcome expectancies. In this task the first half of a sentence was displayed on the computer screen for one second before the second half was displayed. Participants indicated, using the left and right shift keys, whether the sentence was true (left shift key) or false (right shift key) for them (e.g. “If I self-injured I would feel … calm”). The sentences were constructed using the items from Hasking and Boyes’ (2017) NSSI Expectancies Questionnaire. The questionnaire comprises 25 items that tap into five anticipated outcomes of engaging in NSSI: affect regulation (e.g. If I self-injured I would feel calm); negative social expectancies (e.g. If I self-injured my friends would be disgusted); communication (e.g. If I self-injured I would get care from others); pain (e.g. If I self-injured it would cause pain); and negative self-beliefs (e.g. If I self-injured I would feel ashamed). Each item was presented once in the original format and once in reverse format (50 trials; e.g. “If I self-injured I would feel… calm” and “If I self-injured I would NOT feel … calm”). Whether participants indicated that the sentence is true or false for them, and the time taken to respond, was recorded. The time taken to respond indicated the strength of the association between the participant’s response (true or false) and the sentence presented. A faster response indicated a stronger association. Endorsement of an expectancy was indicated by a “True” response to a positively worded item (e.g. “If I self-injured my family would be disgusted”) or a “False” response to the negatively worded items (e.g. “If I self-injured my family would NOT be disgusted”). Mean reaction time for the items of each subscale (positively and negatively worded) were calculated, as was the number of endorsements of each NSSI outcome expectancy. Twenty-five items taken from the Goldberg’s Adjective Scale were randomly interspersed among the NSSI items as control variables in the original and reverse format (50 trials; see Goldberg, 1992). Individual differences in reaction time were assessed using the mean response latency to these items.

Implicit association test – Relief

The Implicit Association Test (IAT; Greenwald et al., 1998) is used to measure differential associations between concepts (e.g. flowers, insects) and attributes (e.g. good, bad). Participants are asked to categorise stimuli (words or pictures) into the associated concepts and attributes. The IAT is a reaction time task based on the premise that responses will be quicker when paired stimuli are more closely associated (Greenwald et al., 1998).

The IAT requires that the attribute of interest is bipolar (e.g. pleasant, unpleasant; good, bad; Greenwald et al., 1998). As this is not always possible, several versions of the IAT have been developed. One version is the unipolar design where the attribute (e.g. good) is compared to neutral words, the other is a single attribute IAT where only one attribute is paired with the concepts with no opposing attribute (Houben et al., 2008). The bipolar and unipolar IATs have greater validity and reliability than the single attribute IAT (Houben et al., 2008).

In developing IATs to assess associations between self-injury and NSSI outcome expectancies, attribute format needed careful consideration. As self-report studies have found affect regulation expectancies and pain expectancies to be the most reliable predictors of NSSI (Dawkins et al., 2018; Hasking, 2017; Hasking & Boyes, 2017), these were chosen as the attributes of interest. Two bipolar IATs were developed: one assessed the association between NSSI and affect regulation (relief) and the other assessed the association between NSSI and pain.

As there is no natural opposing attribute to relief (e.g. black v white; good v bad) IAT format needed to be considered. A unipolar IAT (i.e. attribute vs. neutral) could not be used as neutral words can be similar to relief related words. This would make it difficult for participants to differentiate between attributes. As such, joy-related words were chosen as the opposing attribute to relief in a bipolar IAT due to its opposing nature and intensity to relief. Participants were asked to categorise ‘relief’ (e.g. relax, calm) and ‘joy’ (e.g. cheer, happy) related words as well as images of NSSI (cutting) or neutral images (furniture) by pressing keys on a desktop computer keyboard. The NSSI images were taken from a previous study exploring implicit aversion to NSSI (Franklin et al., 2014) while neutral images were taken from the International Affective Picture System (IAPS; Bradley & Lang, 2007). The IAPS is a database of normative emotional colour photographs which have been standardised according to valance and arousal ratings (Bradley & Lang, 2007). The neutral (furniture) images chosen were of neutral valence (M = 4.93) and low arousal (M = 2.45; IAPS image codes: 7235, 7705, 7020, 7175, 7026, 7025). Six words reflecting relief (Soothe, Calm, Ease, Relax, Peace, Solace) and six words reflecting joy (Bliss, Cheer, Glee, Pleasure, Enjoy, Happy) were used as the attribute stimuli.

The task comprised 7 blocks of trials. For each block, the left and right shift keys were assigned attributes for pictures and words to be categorised into. Participants were directed to respond to the images or words presented by pressing the correct key response (i.e. the assigned shift key) as quickly as possible. In all blocks if an incorrect response was given, an X was displayed on the screen and the participant was required to provide a correct response before continuing with the task. In the first block participants were asked to sort pictures of the concepts (i.e. cutting, furniture) into categories. If the picture fit the cutting category participants pressed the left shift key. If the picture was of furniture they pressed the right shift key. In the second block participants were required to sort attribute words into categories. If the word was associated with relief participants pressed the left shift key and if it was associated with joy they responded by pressing the right shift key. In the third and fourth block, both pictures and words were presented. If stimuli were cutting images or relief words participants pressed the left shift key, while, furniture images and joy words were categorised by pressing the right shift key. For the fifth block, the concepts were switched so that furniture was now categorised by pressing the left shift key and self-injury images by pressing the right shift key. The sixth and seventh blocks were combined, with furniture images and relief words categorised using the left shift key, and cutting images and joy words categorised using the right shift key. The order in which participant categorised the combination of categories (i.e. blocks 3 & 4 and blocks 6 & 7) was counterbalanced between participants.

In line with Greenwald et al.’s (2003) IAT scoring algorithm, data from blocks 3, 4, 6, and 7, were used in the analysis. In line with the scoring algorithm, trials with latencies more than 10,000 ms (Relief n = 7; Pain n = 6) were deleted; no participants had more than 10% of trials with latencies less than 300ms. The IAT effect (D score) was calculated by creating average reaction time scores for each block. Two difference scores (i.e. Block 6 – Block 3; Block 7 – Block 4) of the means were calculated, then each was divided by its associated pooled standard deviation and averaged across the two quotients. Positive scores (i.e. > 0) indicated an association between NSSI and relief, with higher scores indicating a stronger association.

Implicit association test – Pain

Participants also completed an implicit association test measuring the strength of their association between NSSI and expectations of pain. The procedure was identical to the relief task, however the 6 relief words were replaced with 6 pain related words (Hurt, Ouch, Ache, Sore, Agony, Painful). The same furniture images and joy related words were also used in this task.

Covariation bias

This task was used to evaluate whether people overestimate the association between NSSI and NSSI-related outcome expectancies (relief, pain). Participants were presented with a series of paired pictures and words and then asked to estimate the percentage of time each picture type was paired with each word type. Each picture/word pairing were presented an equal number of times (i.e. 33% of the presentation). The NSSI and neutral images, and pain and relief words from the IATs were also used in this task, with negatively valanced images also taken from the IAPS (Bradley & Lang, 2007). Six images for each stimuli picture type were used (NSSI, neutral, negative; negative image IAPS codes: 9903, 9340, 9220, 2703, 9560, 2800) and 6 words for each outcome (pain, relief, neutral; neutral words: Chair, Table, Couch, Lamp, Bench, Desk). The negative images used were of low valence (M = 2.10) and medium arousal (M = 5.27). Negative images were included to ensure that any bias towards NSSI was not the product of a general negativity bias.

Pictures and words were paired and presented randomly in a series: stimulus image (1 sec); outcome word (1 sec); blank screen (1 sec), with each picture paired with each word three times. Participants were told at the beginning of the task that they would be watching a series of pictures and words and instructed on screen to “Just relax and watch the screen, but also pay attention to the pictures and words. You will be asked questions about them afterwards”. Participants were then presented with paired images and words. The entire presentation took approximately 17 minutes.

When the presentation was completed, participants were asked to estimate the percentage of time each picture category was paired with each outcome (e.g. “Please estimate out of 100% how often images of SELF-INJURY were paired with PAIN related words”). As each image/word pair was presented an equal number of times, a bias towards associating concepts was demonstrated by an overestimation of the percentage of times a pairing type was presented (i.e. >33%). It was expected that people who engage in NSSI would overestimate the percentage of time that relief words were presented after an NSSI image. Conversely people with no history of NSSI were expected to overestimate the rate at which pain words were associated with NSSI images.

Questionnaires

All questionnaires were administered online through Qualtrics.

Nonsuicidal self-injury

NSSI engagement was assessed using two questions taken from Part One of the Inventory of Statements About Self-Injury (ISAS; Klonsky & Glenn, 2008). After being provided a definition of NSSI, participants were asked whether they had ever engaged in NSSI, and if they had, at what age they had first engaged and how many times they had self-injured in the past 12 months. Participants were also asked to estimate their lifetime frequency of 12 methods of NSSI and indicate their main form of self-injury. This information was used to determine history of NSSI: people with no history of NSSI; people who have self-injured but not in the past 12 months (past NSSI), and people who have self-injured in the past 12 months (recent NSSI).

NSSI outcome expectancies

The Nonsuicidal Self-Injury Expectancies Questionnaire (NEQ) was used to measure participants’ self-reported NSSI outcome expectancies (Hasking & Boyes, 2017). The measure comprises 25 items assessing five outcome expectancies related to NSSI: affect regulation (e.g., I would feel calm); negative social outcomes (e.g., My friends would be disgusted); communication (e.g., Other people would notice and offer sympathy); physical pain (e.g., It would hurt); and negative self-beliefs (e.g., I would feel like a failure). Participants respond on a four-point Likert scale from 1 (not at all likely) to 4 (extremely likely), how likely they believe the possible outcomes to be if they were to self-injure in the future. The scale has demonstrated strong criterion validity, discriminant validity, and internal consistency (affect regulation $α$ = 0.86, negative social outcomes $α$ = 0.78, communication $α$ = 0.71, pain $α$ = 0.80, negative self-beliefs $α$ = 0.78). In the current sample Cronbach’s alphas for each subscale were: affect regulation $α$ = 0.84, negative social outcomes $α$ = 0.85, communication 0.66, pain $α$ = 0.89, negative self-beliefs $α$ = 0.70).

Procedure

Participants attended a 90 minute laboratory session to complete the four computer-based tasks and the questionnaires. Upon arrival, participants were seated in individual cubicles and asked to read the information sheet, which was presented online in the Qualtrics survey system. If they agreed to participate they were asked to click the “I agree” button to proceed with the study. Participants completed the computer tasks, which were counterbalanced across participants. Participants were then asked to complete the questionnaires. At the end of the study, participants were provided with information about self-injury and where to find support if they were feeling distressed or wished to discuss any concerns with a mental health professional.

Analyses

Sentence completion

To explore the relationship between the endorsement of an outcome expectancy and reaction time to respond, and whether this was related to history of NSSI, Hayes’ (2013) Process model 1 was used to conduct moderated regressions. Moderation models were conducted for each NSSI outcome expectancy, with mean reaction time to that anticipated outcome as the outcome variable. Predictor variables were the number of times the expectancy was endorsed; history of NSSI (i.e. no history of NSSI, past NSSI, recent NSSI) was entered as a moderator. Three moderation models were conducted for each outcome expectancy each using different NSSI groups as the moderating variable. One compared people with no history of NSSI to people with a past history of NSSI, one compared people with no history of NSSI to people with a recent history of NSSI, and the third compared people with a past history of NSSI to people with a recent history of NSSI. Including one categorical variables (3 levels) as the moderator variable in a single anlaysis would not have allowed for these interactions to be explored due to the need to create dummy variables.

Implicit association tests

Two one-way ANOVAs (one for each IAT) were used to compare D scores of people with no history of NSSI, people with a past history of NSSI, and people with a recent history of NSSI, as an indication of the strength of the associations between NSSI and expectancies of both pain and relief.

Covariation bias

The estimated percentages of word/picture pairings reported in the covariation bias task were analysed using a 3(Group: no history of NSSI, past history of NSSI, recent history of NSSI) × 3(Picture type: cutting, negative, neutral) × 3(Word: pain, relief, neutral) ANOVA.

Results

Preliminary analysis

Of the full sample 58 (38.7%) indicated that they had a history of NSSI. Of these, 33 (56.9%) indicated that they had self-injured at least once in the past 12 months. The most commonly reported forms of self-injury were cutting (47.4%), self-battery (17.5%), and severe scratching (15.8%). The mean age of onset was 14.20 (SD = 2.23) years old. Age (r = −0.02, p = 0.78) and gender (χ² = 5.10, p = .28) were not related to history of NSSI and as such were not statistically controlled in the analyses.

Missing Values Analysis (MVA) of the questionnaire responses revealed <5% missing values, which were missing completely at random χ²(9541)=8341.56, p = 1.00. As such, expectation maximisation was used to impute missing data.

Inspection of responses on the Covariation Bias Task revealed that 11 participants’ responses differed from the other participants in format (estimated on a scale out of 0-10 rather than 0-100%). These responses were adjusted to percentages to be proportionate with the rest of the sample (e.g. responding 6 rather than 60, or 2 rather than 20). Five participants did not respond to any of the items. MVA was conducted on the remaining participant data. All items had less than 5% missing values, which were missing completely at random χ²(103)=122.96, p = 0.88; expectation maximisation was used to impute missing values.

Explicit NSSI outcome expectancies

Group differences in self-reported NSSI outcome expectancies revealed that participants with a history of self-injury (recent and past) were more likely to expect NSSI to result in affect regulation than people with no history of self-injury (Table 1). People with no history of self-injury held stronger explicit expectations that self-injury would result in physical pain and communication and care from others than people with a history of self-injury (recent and past). There were no group differences in the self-reported strength of negative social or negative self-belief expectancies.

Table 1.

Group comparisons of self-reported NSSI related outcome expectancies.

	No NSSI(a)M(SD)	Past NSSI(b) M(SD)	12 month NSSI(c) M(SD)	F	Partial $η ²$	Group comparisons^a
Affect regulation expectancies	12.58(2.97)	15.77(2.76)	16.93(3.45)	26.75***	.29	a<b*, a<c*
Negative social expectancies	17.67(3.99)	17.55(3.25)	18.97(3.88)	1.38	.02	–
Communication expectancies	16.26(2.47)	13.91(2.22)	14.63(3.44)	8.92***	.12	a>b*, a>c
Pain expectancies	19.66(2.03)	16.44(3.17)	16.43(3.28)	25.79***	.28	a>b*, a>c*
Negative self-belief expectancies	19.91(2.57)	21.27(2.25)	19.49(3.96)	2.62	.04	–

Note:. *p < 0.05; **p < 0.01; ***p < 0.001 ^aonly significant contrasts reported.

Sentence completion

There was no group difference in mean overall reaction time in response to the personality related items, F (2, 145) = 0.157, p = .855; as such individual differences in reaction time were not statistically controlled in the analyses. Correlations between mean reaction times and endorsements of NSSI expectancy outcomes from the sentence completion task, self-reported NSSI outcome expectancies, and history of NSSI are included in the supplementary material. These revealed a strong correlation between participants self-report NSSI-related outcome expectancies and their implicit expectancies. Group differences in the number of times participants endorsed each outcome expectancy can be found in the supplementary material. People with no history of NSSI endorsed affect regulation expectancies significantly less and pain expectancies significantly more than people with a history of NSSI.

Affect regulation expectancies

History of NSSI and the number of times participants endorsed affect regulation expectancies were not significantly related to reaction time on affect regulation trials (see Table 2). However, there were significant interactions between expectancy endorsement and NSSI in predicting the strength of association with affect regulation expectancies. When comparing people who have never self-injured and people who have self-injured in the past 12 months, there was a negative relationship for people who had recently self-injured that approached significance, b -125.83, t = −1.78, p = .077, but no significant relationship for people who have never self-injured, b = 75.80, t = 1.12, p =.264 (see Figure 1). There was also a significant interaction when comparing people with a past history of NSSI to people who had self-injured in the past 12 months, with a significant negative relationship for people with a recent history of NSSI, b = −161.43, t = −2.06, p < .05, and no significant relationship for people who had self-injured but not in the past 12 months, b = 152.75, t = 1.69, p = .096 (see Figure 1).

Table 2.

Results of moderation models conducted for each NSSI outcome expectancy measured in the sentence completion task with overall reaction time as the outcome variable.

	Never and past NSSI						Never and recent NSSI						Past and recent NSSI
	b	SEB	t	p	95% CI		b	SEB	T	p	95% CI		b	SEB	t	P	95% CI
					Lower	Upper					Lower	Upper					Lower	Upper
Affect regulation expectancies
Constant	1644.45	158.71	10.36	<.001*	1329.93	1958.97	1739.83	143.08	12.16	<.001*	1456.49	2023.17	1419.11	237.48	5.98	<.001*	942.79	1895.43
True responses	91.82	64.20	1.43	.156	−35.41	219.05	21.71	53.50	.46	.686	−84.25	127.66	−26.01	59.96	−.43	.67	−146.28	94.26
NSSI	−192.62	130.84	−1.47	.144	−451.92	66.67	−8.38	58.20	−.14	.886	−123.64	106.87	−115.66	127.98	−.90	.370	−372.36	141.03
True response× NSSI	83.37	123.00	.68	.499	−160.40	327.13	−100.82	48.12	−2.10	.038*	−196.11	−5.52	−314.18	117.83	−2.67	.010*	−550.53	−77.83
Negative social expectancies
Constant	1763.88	39.54	44.61	<.001*	1685.53	1842.23	1742.49	33.24	52.43	<.001*	1676.67	1808.31	1619.33	65.60	24.69	<.001*	1487.75	1750.90
True responses	−29.46	37.56	−.78	.435	−103.90	44.98	−21.24	33.66	−.63	.529	−87.89	45.41	−45.03	55.15	−.82	.418	−191.18	245.25
NSSI	−59.81	95.41	−.63	.532	−248.89	129.27	−70.05	47.82	−1.46	.146	−164.75	24.64	−107.50	117.39	−.92	.364	−342.96	127.97
True response× NSSI	−34.84	87.42	−.40	.691	−208.08	138.39	.44	39.18	.01	.99	−77.14	78.03	27.03	108.79	.25	.805	−191.18	245.25
Communication expectancies
Constant	2006.32	40.01	50.14	<.001*	1927.03	2085.62	2022.08	38.03	53.17	<.001*	1946.78	2097.39	1854.09	59.69	31.06	<.001*	1734.37	1973.81
True responses	−73.10	37.87	−1.93	.056	−148.15	1.95	−57.35	37.73	−1.52	.131	−132.07	17.37	97.23	50.90	1.91	.061	−4.86	199.32
NSSI	−246.85	98.11	−2.52	.013*	−441.29	−52.42	−55.92	55.22	−1.01	.313	−165.27	53.42	94.96	107.01	.89	.379	−119.68	309.60
True response× NSSI	204.11	97.89	2.09	.039*	10.12	398.10	112.08	42.73	2.62	.01*	27.47	196.69	17.61	104.37	.17	.867	−191.72	226.95
Pain expectancies
Constant	1639.52	35.09	46.73	<.001*	1569.99	1709.06	1640.44	37.08	44.24	<.001*	1567.01	1713.86	1576.11	73.29	21.50	<.001*	1429.10	1723.11
True responses	−213.47	60.09	−3.55	<.001*	−332.56	−94.39	−223.73	65.00	−3.44	<.001*	−352.44	−95.02	−135.60	44.72	-3.03	.004*	−225.31	−45.90
NSSI	−151.46	85.91	−1.76	.081	−321.72	18.80	−78.48	52.78	−1.49	.140	−183.00	26.04	30.18	130.64	.23	.82	−231.85	292.21
True response× NSSI	158.07	87.81	1.80	.074	−15.96	332.09	38.50	51.16	.75	.453	−62.81	139.81	−78.71	87.70	−.90	.374	−254.61	97.18
Negative self-belief expectancies
Constant	2036.61	48.44	42.04	<.001*	1940.60	2132.61	2036.68	43.15	47.20	<.001*	1951.24	2122.12	1888.27	64.65	29.21	<.001*	1758.59	2017.94
True responses	−13.97	45.08	−.31	.758	−103.29	75.40	1.84	44.03	.04	.967	−85.35	89.03	−16.12	53.19	−.30	.763	−122.81	90.58
NSSI	−165.11	117.01	−1.41	.161	−397.00	66.79	−52.73	62.59	−.84	.401	−176.67	71.22	−4.60	115.30	−.04	.968	−235.86	226.67
True response× NSSI	−49.06	104.61	−.47	.640	−256.38	158.25	8.73	50.72	.17	.864	−91.70	109.17	64.47	105.60	.61	.544	−147.34	276.27

Note. ***p < .001, **p < .01, *p < .05.

Figure 1.

Reaction time to affect regulation expectancy items dependent on the level of endorsement and history of NSSI.

Negative social expectancies

History of NSSI and the number of times participants endorsed negative social outcome expectancies were not related to reaction time on these trials (see Table 4). There were also no significant interactions when predicting associations with negative social outcome expectancies.

Communication expectancies

The number of times participants endorsed communication expectancies was not related to reaction times (Table 4). However, people with a past history of NSSI responded faster to communication items than people with no history of NSSI. There was a significant interaction when comparing people who have never self-injured to people who had a past history of NSSI, which revealed a significant negative relationship between expectancy endorsement and reaction time for people with no history of NSSI b = −117.47, t = −2.82, p = .006, but no significant relationship for people with a history of NSSI b = 86.64, t = .98, p = .330 (Figure 2). There was also a significant interaction when comparing people with no history of NSSI to people with a recent history of NSSI, which indicated a significant negative relationship between expectancy endorsement and reaction time for people with no history of NSSI b = −117.49, t = −2.67, p = .009, but no significant relationship for people with a history of NSSI b = 106.67, t = 1.46, p = .15 (Figure 2).

Figure 2.

Reaction time to communication expectancy items dependent on the level of endorsement and history of NSSI.

Pain expectancies

People who endorsed pain expectancies had faster reaction times than people who did not endorse pain expectancies. There was no relationship between history of NSSI and reaction time. The interaction between expectancy endorsement and history of NSSI was approaching significance when comparing people with no history of NSSI and people with a past history of NSSI. Exploration of this revealed a significant negative relationship for people with a history of NSSI b = −89.77, t = −2.03, p = .045, but this was stronger for people with no history of NSSI b = −247.84, t = −3.27, p <.001 (Figure 3).

Figure 3.

Reaction time to pain expectancy items dependent on the level of endorsement and history of NSSI.

Negative self-belief expectancies

History of NSSI engagement and number of times participants endorsed negative self-belief expectancies were not significantly related to reaction time on these trials (see Table 4).

Implicit association tests

Correlations between IAT D Scores, self-reported NSSI outcome expectancies, and NSSI can be found in the supplementary material. D scores from the relief IAT were significantly positively correlated with self-reported negative social expectancies. There were no other significant correlations between D scores on either IAT with self-reported NSSI outcome expectancies or NSSI for either IAT.

Relief IAT

Overall mean D scores were greater than 0 indicating an association between NSSI and relief for all groups (see Table 3). There were no significant group differences in the Mean D scores when comparing people who had never self-injured, people who had self-injured in the past, and people who had recently self-injured, F(2, 147) = 1.31, p = .273, partial $η 2$ = .018.

Table 3.

Mean D score for the relief and pain IATs for each NSSI group.

	No history of NSSI	Past history of NSSI	Recent History of NSSI
Relief IAT D M(SD)	.002 (.253)	.025 (.258)	.084 (.244)
Pain IAT DM(SD)	.186 (.285)	.176 (.236)	.127 (.296)

Pain IAT

Overall mean D scores indicated that participants associated NSSI with pain (i.e. = > 0; Table 3). There were no significant group differences in the Mean D scores when comparing people who had never self-injured, people who had self-injured in the past, and people who had recently self-injured, F(2, 147) = .54, p = .586, partial $η 2$ = .007.

Covariation bias

Correlations between Covariation Bias estimates, self-reported NSSI outcome expectancies, and NSSI can be found in the supplementary material. Generally, Covariation Bias estimates did not correlate with self-reported NSSI outcome expectancies or history of NSSI.

Overall, participants more accurately estimated the presentation of neutral images (M = 38.34), than NSSI (M = 43.60), and negative (M = 44.47) images F(2, 568) = 17.181, p < .001 partial $η^{2}$ = .108. There was also a main effect of word type F(2, 568) = 12.55, p < .001, partial $η^{2}$ = .081, with participants more accurately estimating the presentation of neutral words (M = 38.52) than pain (M = 45.16) and relief (M = 42.79) words. There was no main effect of NSSI history F(2, 568) = 1.269, p = .284, partial $η^{2}$ = .018. There were no significant 2-way interactions between image type and NSSI history F(4, 568) = 0.89, p = .468, partial $η^{2}$ = .012; word type and NSSI history F(4, 568) = 0.123, p = .974, partial $η^{2}$ = .002; or image type and word type F(4, 568) = 1.245, p = .29, partial $η^{2}$ = .009. There was a significant 3-way interaction F(8, 564) = 2.361, p = .017, partial $η$ ² = .032. People with a recent history of NSSI more accurately estimated the presentation of NSSI images when paired with neutral words (M = 32.28, 95%CI = 25.59, 38.97) than people with a past history of NSSI (M = 45.69, 95%CI = 37.79, 53.58; Figure 4). There were no other significant differences in estimations (Table 4).

Figure 4.

Differences in estimates of the presentation NSSI images paired with neutral images dependent on NSSI history.

Table 4.

Comparisons of estimates of image-outcome pairings from the covariation bias task.

Pairing Type		No History of NSSI^a	Past History of NSSI^b	Recent History of NSSI^c	Group comparisons p=
Image	Outcome	No History of NSSI^a	Past History of NSSI^b	Recent History of NSSI^c	Ab	ac	bc
NSSI	Pain	46.74	46.87	42.19	1.00	.789	1.00
NSSI	Relief	46.38	47.25	45.42	1.00	1.00	1.00
NSSI	Neutral	41.55	45.69	32.28	1.00	.060	.034*
Negative	Pain	49.04	48.43	43.93	1.000	.527	1.00
Negative	Relief	44.26	50.00	37.34	.634	.263	.058
Negative	Neutral	39.41	39.22	46.25	1.000	.244	.529
Neutral	Pain	41.63	47.01	40.59	.589	1.000	.563
Neutral	Relief	37.27	39.70	37.45	1.00	1.00	1.00
Neutral	Neutral	33.86	37.74	30.67	1.00	1.00	.502

Discussion

The aim of this study was to measure implicit NSSI-related outcome expectancies and to establish whether they differ between people with and without a history of NSSI. Unlike self-report measures, implicit measures can tap into unconscious associations and avoid social desirability bias. Additionally, implicit measures often have predictive power over and above that of explicit measures (Hahn & Gawronski, 2015). We found that while two of the tasks we used did not differentiate NSSI history, the sentence completion task shows promise in measuring implicit NSSI outcome expectancies and capturing different beliefs held by people, based on history of NSSI.

Consistent with self-report studies, participants with a history of self-injury had stronger associations between NSSI and affect regulation, and weaker associations between NSSI and pain, than people with no history of NSSI (Dawkins et al., 2018; Hasking, 2017; Hasking & Boyes, 2017). When endorsing items in the sentence completion task, participants with a history of self-injury demonstrated a stronger implicit association with affect regulation expectancies. While people with a past history of NSSI associated NSSI with affect regulation more than people with no history of NSSI, the implicit association was even stronger for people who had recently self-injured. This may indicate a change in the strength of associations with affect regulation as engagement in the self-injury becomes more distant. It is possible that with less recent experience, the association becomes weaker, an effect that has not previously been observed when using self-report measures. Alternatively, people who have not self-injured within the past 12 months may have held weaker associations with affect regulation to begin with, minimising the chance they would continue to self-injure.

Interestingly, while people with no history of self-injury self-reported stronger communication expectancies than people with a history of NSSI, there was no difference in the endorsement of communication items in the sentence completion task. However, people with no history of self-injury who had explicitly endorsed communication expectancies, had a stronger implicit association than people with a history of engaging in self-injury. This may reflect the common misconception in the community that people who self-injure do so to receive attention from other people (Klonsky et al., 2014). Expecting NSSI to result in friends and family becoming upset with you or experiencing diminished self-worth were associated with NSSI regardless of NSSI history. This may reflect past negative experiences of people who have self-injured, which shape their expectations of anticipated future outcomes.

Although non-significant, the pattern of results in the IATs was in the expected direction. The lack of significant differences may have been influenced by the choice of “Joy” as the opposing outcome to relief and pain. Joy is not likely associated with NSSI and the IAT may not have been powerful enough to measure differences in associations without the pull of an opposing outcome that is associated with NSSI. With previous IAT tasks conducted by Gratz et al. (2016), disgust-related words were used as an opposing attribute; however we felt disgust may be implicitly associated with NSSI for people who have never engaged in the behaviour (Zila & Kiselica, 2001). This may have created larger group differences by weakening the association between NSSI and relief for people who have not self-injured. This could be examined further in future studies comparing different IAT designs within the context of associations with NSSI outcome expectancies.

People who had recently self-injured were more accurate in their estimation of the presentation of NSSI images when related to neutral words, compared to people who had self-injured in the past. This likely suggests that NSSI stimuli were more salient to people who have not recently self-injured. If it had been a while since their last engagement in NSSI, participants may have noticed the images more both because of their personal relevance, as well as their relative novelty. Triggering thoughts about their past self-injury may have influenced the association. However, there was no difference with people who had never self-injured. This may have been because, although images are possibly salient, they may not have had the same level of personal reference that they would have to people who have self-injured in the past.

Implications

Theoretically, the results strengthen the assertation that NSSI outcome expectancies differ between people with and without a history of NSSI. Specifically, it supports the inclusion of social cognitive constructs within the proposed Cognitive-Emotional Model of NSSI (Hasking et al., 2017). We found that implicit associations between NSSI and affect regulation expectancies may change or differ with recency of NSSI. It is possible that the sentence completion task could be used in future research to determine whether explicit and implicit expectancies change over time. This would improve the understanding of how outcome expectancies may play a role in the onset, maintenance, and cessation of NSSI. It is possible that implicit expectancies may be found to be an indicator of change before there is an observable reduction in frequency of NSSI. This could be useful within clinical settings to assess the efficacy of interventions aimed at treating NSSI. If future studies find that implicit NSSI outcome expectancies have predictive utility over and above self-report measures, they could be used in predictive models, or clinical settings, to identify people at risk of initiating or continuing engagement in self-injurious behaviours. This could be used to inform the implementation of prevention and interventions.

Limitations and future research

Our results need to be considered with regards to several limitations. The small number of participants with a history of NSSI needs to be considered. If possible future research may want to consider targeted recruitment of participants with lived experience of NSSI to increase the likelihood of equal group sizes. As this study was cross-sectional we cannot speak to the predictive utility of implicit NSSI-related outcome expectancies. Measuring implicit outcome expectancies across different time points, and differing contexts, may shed light on how these associations change over time. Future research could use these tasks in longitudinal or Ecological Momentary Assessment (EMA) studies to develop a better understanding of the predictive value of implicit NSSI outcome expectancies. Implicit associations may be helpful in predicting when individuals are likely to engage in self-injurious behaviours. EMA studies using a within-person design could identify cognitive changes leading up to engagement in self-injury which could be used to predict future NSSI. Future research could also explore whether there are changes in expectancy strength which coincide with affective experience. For example, it is possible that associations between NSSI and affect regulation expectancies may strengthen when individuals are experiencing acute stress. Within this study images of self-injury exclusively depicted cutting. The majority of participants with lived experience of NSSI (82%) reported using cutting as a method of self-injury at least once. Therefore, these images may not be relevant for participants who use other methods of NSSI. Future research may consider using personalised images which are relevant to the participant.

Conclusion

Our results suggest that the sentence completion task is most likely to be useful in future studies assessing implicit NSSI expectancies. Implicit measures are important with regards to behaviour prediction, but also can pick up on small changes in associations over time, which may give light to the underlying mechanisms involved in NSSI. Additionally, the Sentence Completion Task was able to differentiate between people with a recent and past history of self-injury. Future research into the underlying mechanisms of NSSI and the role of outcome expectancies could inform theory, and future prevention and intervention efforts.

Supplemental Material

sj-pdf-1-prx-10.1177_0033294120961512 - Supplemental material for Implicit Assessment of Self-Injury Related Outcome Expectancies: A Comparison of Three behavioural Tasks

Supplemental material, sj-pdf-1-prx-10.1177_0033294120961512 for Implicit Assessment of Self-Injury Related Outcome Expectancies: A Comparison of Three behavioural Tasks by Jessica Dawkins, Penelope Hasking, Camilla Luck and Mark Boyes in Psychological Reports

Footnotes

Notes

ORCID iDs

Jessica Dawkins

Mark Boyes

References

Amrhein

Pauli

Dengler

Wiedermann

(2005). Covariation bias and its physiological correlates in panic disorder. Journal of Anxiety Disorders, 19, 177–191. https://doi.org/10.1016/j.janxdis.2004.01.004

Bandura

(1986). Social foundations of thought and action: A social cognitive theory. Prentice Hall.

Bentley

Nock

Barlow

(2014). The four-function model of nonsuicidal self-injury: Key directions for future research. Clinical Psychological Science, 2, 638–656. https://doi.org/10.1177/2167702613514563

Bradley

M. M.

Lang

P. J.

(2007). The international affective picture system (IAPS) in the study of emotion and attention. In Coan

Allen

(Eds.), Handbook of emotion elicitation and assessment (pp. 29–46). Oxford University Press.

Chapman

Gratz

Brown

(2006). Solving the puzzle of deliberate self-harm: The experiential avoidance model. Behaviour Research and Therapy, 44, 371–394. https://doi.org/10.1016/j.brat.2005.03.005

Dawkins

Hasking

Boyes

Greene

Passchier

(2018). Applying a cognitive-emotional model to nonsuicidal self-injury. Stress & Health, 35, 39–48. https://doi.org/10.1002/smi.2837

Forster

(2002). DMDX display software. http://www.u.arizona.edu/∼kforster/dmdx/dmdx.htm

Franklin

J. C.

Lee

K. M.

Puzia

M. E.

Prinstein

M. J.

(2014). Recent and frequent nonsuicidal self-injury is associated with diminished implicit and explicit aversion toward self-cutting stimuli. Clinical Psychological Science, 2, 306–318. https://doi.org/10.1177/2167702613503140

Goldberg

L. R.

(1992). The development of markers for the big-five factor structure. Psychological Assessment, 4, 26–42. https://doi.org/10.1037/1040-3590.4.1.26

10.

Gratz

K. L.

Chapman

A. L.

Dixon-Gordon

K. L.

Tull

M. T.

(2016). Exploring the association of deliberate self-harm with emotional relief using a novel implicit association test. Personality Disorders: Theory, Research, and Treatment, 7, 91–102. https://doi.org/10.1037/per0000138

11.

Gratz

K. L.

Tull

Dixon-Gordon

Turner

Chapman

(2018). Is the association of deliberate self-harm with emotional relief stable or dependent on emotional context? Journal of Behaviour Therapy and Experimental Psychiatry, 60, 61–68. https://doi.org/10.1016/j.jbtep.2018.03.003

12.

Greenwald

A. G.

McGhee

D. E.

Schwartz

J. L.

(1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464–1480. https://doi.org/10.1037/0022-3514.74.6.1464

13.

Greenwald

A. G.

Nosek

B. A.

Banaji

M. R.

(2003). Understanding and using the implicit association test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197–216. https://doi.org/10.1037/0022-3514.85.2.197

14.

Hahn

Gawronski

(2015). Implicit social cognition. In Wright

(Ed.), International encyclopaedia of the social & behavioural sciences (pp. 714–720). https://doi.org/10.1016/B978-0-08-097086-8.24066-X

15.

Hasking

(2017). Differentiating non-suicidal self-injury and risky drinking: A role for outcome expectancies and self-efficacy beliefs. Prevention Science, 18, 694–703. https://doi.org/10.1007/s11121-017-0755-7

16.

Hasking

Boyes

(2017). The nonsuicidal self-injury expectancy questionnaire (NEQ): Factor structure and initial validation. Clinical Psychologist, 22, 251–261. https://doi.org/10.1111/cp.12127

17.

Hasking

Whitlock

Voon

Rose

(2017). A cognitive-emotional model of NSSI: Using emotion regulation and cognitive processes to explain why people self-injure. Cognition and Emotion, 31, 1543–1556. https://doi.org/10.1080/02699931.2016.1241219

18.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis. Guilford Press.

19.

Hermann

Ofer

Flor

(2004). Covariation bias for ambiguous social stimuli in generalized social phobia. Journal of Abnormal Psychology, 113, 646–653. https://doi.org/10.1037/0021-843X.113.4.646

20.

Houben

Nosek

Wiers

(2008). Seeing the forest for the trees: A comparison of different IAT variants measuring implicit alcohol associations. Drug and Alcohol Dependence, 106, 204–211. https://doi.org/10.1016/j.drugalcdep.2009.08.016

21.

International Society for the Study of Self-Injury. (2018, May). What is self-injury? http://itriples.org/about-self-injury/what-is-self-injury

22.

Jajodia

Earleywine

(2003). Measuring alcohol expectancies with the Implicit Association Test. Psychology of Addictive Behaviours, 17, 126–133. https://doi.org/10.1037/0893-164X.17.2.126

23.

Kiekens

Claes

Demyttenaere

Auerbach

Green

Kessler

, Mortier, P., Nock, M., & Bruffaert

(2016). Lifetime and 12-month nonsuicidal self-injury and academic performance in college freshmen. Suicide and Life Threatening Behaviour, 46, 563–576. https://doi.org/10.1111/sltb.12237

24.

Kiekens

Hasking

Boyes

Claes

Mortier

Auerbach

, Cuijpers, P., Demyttenaere, K., Green, J., Kessler, R., Myin-Germeys, I., Nock, M., & Bruffaerts

(2018). The associations between non-suicidal self-injury and first onset suicidal thoughts and behaviours. Journal of Affective Disorders, 239, 171–179. https://doi.org/10.1016/j.jad.2018.06.033

25.

Kiekens

Hasking

Claes

Boyes

Mortier

Auerbach

, Cuijpers, P., Demyttenaere, K., Green, J., Kessler, R., Myin-Germeys, I., Nock, M., & Bruffaert

(2019). Predicting the incidence of non-suicidal self-injury in college students. European Psychiatry, 59, 44–51. https://doi.org/10.1016/j.eurpsy.2019.04.002

26.

Kiekens

Hasking

Claes

Mortier

Auerbach

R. P.

Boyes

, Cuijpers, P., Demyttenaere, K., Green, J., Kessler, R., Nock, M., & Bruffaert

(2018). The DSM-5 nonsuicidal self-injury disorder among incoming college students: Prevalence and association with 12-month mental disorders and suicidal thoughts and behaviors. Depression and Anxiety, 35, 629–637. https://doi.org/10.1002/da.22754

27.

Klonsky

E. D.

Glenn

C. R.

(2008). Assessing the functions of non-suicidal self-injury: Psychometric properties of the inventory of statements about self-injury (ISAS). Journal of Psychopathological Behaviour Assessment, 31, 215–214. https://doi.org/10.1007/s10862-008-9107-z

28.

Klonsky

E. D.

Victor

S. E.

Saffer

B. Y.

(2014). Nonsuicidal self-injury: What we know, and what we need to know. The Canadian Journal of Psychiatry, 59, 565–568. https://doi.org/10.1177/070674371405901101

29.

Marissen

M. A.

Frenken

I. H.

Blenken

Hendik

V. M.

, & Van Der Brink. (2005). The relation between social desirability and different measures of heroin craving. Journal of Addictive Diseases, 24, 91–103. https://doi.org/10.1300/J069v24n04_07

30.

Nock

(2009). Why do people hurt themselves? New insights into the nature and functions -of self-injury. Current Directions in Psychological Science, 18, 78–83. https://doi.org/10.1111/j.1467-8721.2009.01613.x

31.

Palfai

Monti

Colby

Rohsenow

(1997). Effects of suppressing the urge to drink on accessibility of alcohol outcome expectancies. Behaviour Research and Therapy, 35, 59–65. https://doi.org/10.1016/S0005-7967(96)00079-4

32.

Read

Curtin

(2007). Contextual influences on alcohol expectancy processes. Journal of Studies on Alcohol and Drugs, 68, 759–770. https://doi.org/10.15288/jsad.2007.68.759

33.

Rodriguez-Blanco

Carbello

Baca-Garcia

(2018). Use of ecological momentary assessment (EMA) in nonsuicidal self-injury (NSSI): A systematic review. Psychiatry Research, 263, 212–219. https://doi.org/10.1016/j.psychres.2018.02.051

34.

Selby

Joiner

(2009). Cascades of emotion: The emergence of borderline personality disorder from emotional and behavioural dysregulation. Review of General Psychology, 13(3), 1–19. https://doi.org/10.1037/a0015687

35.

Swannell

S. V.

Martin

G. E.

Page

Hasking

St John

N. J.

(2014). Prevalence of nonsuicidal self-injury in nonclinical samples: Systematic review, meta-analysis and meta-regression. Suicide and Life-Threatening Behaviour, 44, 273–303. https://doi.org/10.1111/sltb.12070

36.

Taylor

Jomar

Dhingra

Forrester

Shahmalak

Dickson

(2018). A meta-analysis of the prevalence of different functions of non-suicidal self-injury. Journal of Affective Disorders, 227, 759–769. https://doi.org/10.1016/j.jad.2017.11.073

37.

Whitlock

Muehlenkamp

Eckenrode

Purington

Abrams

G. B.

Barreira

Kress

(2013). Nonsuicidal self-injury as a gateway to suicide in young adults. Journal of Adolescent Health, 52, 486–492. https://doi.org/10.1016/j.jadohealth.2012.09.010

38.

Wiers

R. W.

Stacy

A. W.

(2006). Implicit cognition and addiction. Current Directions in Psychological Science, 15, 292–296. https://doi.org/10.1111/j.1467-8721.2006.001455

39.

Wolff

Thompson

Thomas

Nesi

Bettis

Ransford

, Scopelliti, K., Frazier, E., & Liu

(2019). Emotion dysregulation and non-suicidal self-injury: A systematic review and meta-analysis. European Psychiatry, 59, 25–36. https://doi.org/10.1016/j.eurpsy.2019.03.004

40.

Zilla

L. M.

Kiselica

M. S.

(2001). Understanding and counselling self-mutilation in female adolescents and young adults. Journal of Counseling & Development, 79, 46–52. https://doi.org/10.1002/j.1556-6676.2001.tb01942.x

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.23 MB