Non-suicidal self-injury (NSI) refers to the intentional destruction of one’s own body tissue without suicidal intent and for purposes not socially sanctioned (ISSS ). Approximately 4–6 % of adults in the general population report having engaged in NSI at least once [16, 20], and this figure increases to approximately 14–18 % in community samples of adolescents and young adults [24, 25, 29, 32]. NSI is of concern due to its association with a variety of psychological disorders, as well as both its concurrent and prospective relationship to suicidal behavior [1, 2, 18, 20, 33].
Whereas early research tended to focus on psychosocial and diagnostic correlates of NSI, many studies from the last 10 years have addressed the functions of NSI [5, 14, 22, 27]. A functional perspective emphasizes variables that may be conceptualized as motivating or reinforcing the behavior . Research on NSI functions has greatly advanced understanding of NSI. For example, it is now well established that affect regulation—using NSI to alleviate intense negative emotions—is the most common function of NSI, endorsed by more than 90 % of those who engage in the behavior [4, 15, 14]. It is also well documented that 50 % or more of those who self-injure endorse self-punishment, or self-directed anger, as a motivation for NSI , a pattern that has led subsequent studies to elucidate the role of self-criticism in NSI . Many other NSI functions have also been identified including anti-dissociation (e.g., causing pain to stop feeling numb), anti-suicide (e.g., stopping suicidal thoughts), peer bonding (e.g., fitting in with others), interpersonal influence (e.g., letting others know the extent of emotional pain), and sensation seeking (e.g., doing something to generate excitement) [14, 17].
Despite the high endorsement of affective regulation functions of NSI, most individuals who self-injure endorse multiple functions [14, 17, 26]. Therefore, it is important to understand the extent to which different functions overlap or co-occur. For example, reducing negative feelings (affect regulation) may help reduce suicidal thoughts (anti-suicide), as well as reduce dissociation (anti-dissociation) for those who feel numb or unreal when overwhelmed by intense negative emotions. Similarly, using NSI to influence others (interpersonal influence) may include using the behavior to improve relationships with others who self-injure (peer bonding), as well as using NSI in social circles as an ‘extreme’ or exciting activity (sensation seeking). In addition, there is accumulating evidence that different NSI functions have different implications for treatment, prognosis, and suicide risk [17, 19, 27]. Thus, understanding the conceptual and empirical overlap among functions is critical both for theory development in research contexts and for case conceptualization and treatment planning in clinical contexts.
One study in particular has been influential in addressing covariation among NSI functions. Nock and Prinstein  administered the Functional Assessment of Self-Mutilation (FASM; ) to a sample of 89 adolescent patients with histories of NSI. The FASM is a self-report questionnaire that includes 22 reasons for engaging in NSI. Nock and Prinstein  utilized confirmatory factor analyses (CFA) to examine the structure of the 22 reasons and concluded that the motivations were best conceptualized as falling into one of four different categories: Automatic-Negative (use of NSI to reduce unpleasant internal states), Automatic-Positive (use of NSI to produce desirable internal states), Social-Negative (use of NSI to escape from interpersonal demands), and Social-Positive (use of NSI to gain attention or desirable responses from others). Importantly, Nock and Prinstein  also found a good fit for a two-factor model of NSI functions: Automatic and Social. This two-factor model fit the data as well as the less parsimonious four-factor model; however, the authors retained the latter on theoretical grounds.
The four-factor model advocated by Nock and Prinstein  has been extremely influential, as evidenced in part by a Google Scholar citation count exceeding 600. It is thus important to consider limitations of the evidence supporting the four-factor structure. First, the sample size was relatively small, reducing power to detect differences in fit between competing models (e.g., two-factor vs. four-factor). Second, some correlations between factors were high. For example, the Social-Negative and Social-Positive factors correlated .78, a magnitude high enough to suggest they represent the same latent factor . Similarly, the Automatic-Negative and Automatic-Positive factors correlated .52, which is high considering that the low coefficient alphas for these two factors (.62 and .69, respectively) limit the extent to which these variables can correlate. Third, the Automatic-Negative factor consisted of just two items, which presents a challenge to its reliability and replicability. Perhaps as a consequence, in a subsequent study, one of the two Automatic-Negative items was switched to the Automatic-Positive factor for both empirical and conceptual reasons , leaving just a single item on the Automatic-Negative scale. Finally, Nock and Prinstein  utilized a CFA rather than an exploratory factor analysis (EFA). CFA is indeed useful for evaluating a theoretically derived structure. At the same time, because CFA requires identifying item-factor loadings a priori, the use of CFA places limits on the number and nature of factors that may emerge. Therefore, EFA, which places no such factor restrictions, may be especially appropriate for early stages of structural research (for elaboration see ).
Indeed, a recent spate of studies has examined the factor structure of the FASM and found solutions that diverge from that reported in Nock and Prinstein . A study of a Chinese version of the FASM found that the four-factor structure reported by Nock and Prinstein  provided inadequate fit . Two other studies of the FASM have found empirical support for a three-factor solution: (1) automatic, (2) social influence/communication, (3) peer identification/conformity. Specifically, Young et al.  found this structure utilizing principal components analysis of 170 15-year old students, and Dahlström et al.  found this structure using both EFA and CFA in 836 adolescents. Dahlstrom et al. also found excellent fit for a theoretically driven four factor solution consisting of one automatic factor and three social factors (social influence, peer identification, and avoiding demands).
The research described so far has focused on the structure of NSI functions as assessed by a particular measure, the FASM. Of course, any structure that emerges from research on this measure may reflect particular properties of the FASM rather than of NSI functions more generally. It is therefore important to note a separate line of research on NSI functions that has focused on another measure: the Inventory of Statements About Self-injury (ISAS; ). The ISAS is a self-report questionnaire consisting of 39 reasons for engaging in NSI, which are organized into 13 rationally derived functional scales. Klonsky and Glenn  utilized EFA to examine the structure of the 13 scales in a sample of 235 university students with histories of NSI and found that they were best conceptualized as representing two superordinate factors: Intrapersonal and Interpersonal functions. The Intrapersonal factor included self-focused functions, such as affect regulation and self-punishment, whereas the Interpersonal factor included other-focused functions, such as interpersonal influence and peer bonding. Klonsky and Glenn  concluded that these Intrapersonal and Interpersonal factors were conceptually equivalent to Nock and Prinstein’s  Automatic and Social factors, respectively. This two-factor structure was later further supported by a confirmatory factor analysis in a large (n = 529) Turkish sample of high school students with NSI histories .
However, two important limitations of both Klonsky and Glenn  and Bildik et al.  deserve note. First, both studies factor-analyzed the 13 ISAS scales rather than the 39 ISAS items. Thus, research has yet to empirically examine the structure of the ISAS at the item-level. Second, both studies utilized non-clinical samples; many participants may have engaged in infrequent or sub-clinical NSI, which may limit generalizability to treatment-seeking populations.
The present study was conceived to address ambiguity regarding the structure of NSI functions. Specifically, in two large samples of patients receiving acute-care treatment for NSI, we utilized EFA to investigate the structure of NSI functions as assessed by both the ISAS and the FASM. Use of two different measures helps ensure that findings will be generalizable, rather than artifacts of a particular questionnaire, and the large sample sizes provide sufficient power for item-level EFAs. In addition, this will be the first investigation of the structure of NSI functions to use large samples of patients. Based on findings from both Nock and Prinstein  and Klonsky and Glenn , we suspect a two-factor structure will best characterize NSI functions: Intrapersonal (Automatic) and Social (Interpersonal).Footnote 1 However, because neither the FASM nor ISAS items have been examined using an exploratory approach in patient populations, and because recent studies on the FASM have produced both three and four-factor structures, we utilized EFA so as not to constrain the number and nature of functional factors that could emerge.