Self-Injurious Behavior

Abstract

Self-injurious behavior is a type of behavioral symptom: the deliberate, direct destruction of one's own body tissue without suicidal intent, spanning the stereotypic self-injury of intellectual disability and the non-suicidal self-injury (NSSI) of adolescents and adults. This article unifies both literatures around one idea, that self-injury is not emitted at random but is an operant maintained by its consequences, which the same four-function scheme (positive or negative reinforcement, delivered socially or automatically) organizes. It covers the clinical taxonomy, the experimental functional analysis that established the reinforcement account, the affect-regulation model of NSSI, and the boundary with suicide. Three demonstrations make the mechanisms manipulable: a functional-analysis classifier, an affect-regulation negative-reinforcement loop, and a function-profile distribution. What remains contested is whether NSSI warrants its own diagnosis and how repeated self-injury raises the capability for later suicide.

Keywords: behavioral symptoms, non-suicidal self-injury, functional analysis, reinforcement

What Self-Injurious Behavior Is

In MeSH, Self-Injurious Behavior is defined as behavior in which a person deliberately and directly injures their own body tissue without the intent to die, and the descriptor is filed under behavioral symptoms — the parent category that frames it not as a diagnosis but as an observable, troubling display. The definition does two kinds of work at once. By requiring that the injury be deliberate and direct, it excludes accidental harm and the indirect, slow self-destruction of substance misuse or disordered eating. By requiring the absence of suicidal intent, it separates self-injury from a suicide attempt, even though the two can involve the same act.

That single descriptor spans two research traditions that grew up apart. The older tradition studied the stereotypic, repetitive self-injury — head-banging, self-biting, skin-picking, eye-poking — seen in a substantial minority of people with intellectual disability and autism, where it can be frequent, severe, and medically dangerous. A population study of adults with intellectual disabilities found self-injurious behaviour in roughly 4–5% at any one time, rising with the severity of the disability (Cooper et al., 2009). The younger tradition studied non-suicidal self-injury (NSSI) — most often self-cutting — in adolescents and young adults without developmental disability, where lifetime prevalence estimates cluster around 17% in adolescents and fall across adulthood (Swannell et al., 2014). Favazza's clinical synthesis (1998) gave the field its enduring taxonomy, sorting self-injury into major (rare, severe, often psychotic acts such as self-enucleation), stereotypic (the repetitive developmental form), and superficial-to-moderate (the common cutting and burning), the last of which is what most clinical and research attention now addresses.

The unifying claim of the cognitive-behavioral account is that, across both traditions and all three types, self-injury is not a symptom emitted at random but a behavior that persists because it works — because it reliably produces a consequence the person finds reinforcing. A modern systematic review of NSSI consolidates the definition, prevalence, functions, and correlates on which this account rests (Cipriano et al., 2017). The rest of this article develops that claim: first the functional scheme that classifies the consequences, then the experimental method that established it, then the affect-regulation mechanism that dominates in NSSI, and finally the contested boundary with suicide.

Types of Self-Injurious Behavior

MeSH places Self-Injurious Behavior (D016728) directly beneath Behavioral Symptoms and gives it three narrower descriptors. These children are classificatory conveniences for indexing the literature, not a mechanistic partition: the categories overlap (self-mutilation and the proposed excoriation disorder both fall under the superficial-to-moderate band of Favazza's taxonomy), and a MeSH child marks where papers are filed rather than a distinct causal kind. Only self-mutilation currently has its own article here; the other two are glossed from their scope in the tree. Table 1 lists them.

The placement of Suicide as a child of Self-Injurious Behavior is the clearest sign that the tree is an indexing classification rather than a theory: by the intent-based definition used throughout this article, a suicide attempt is precisely not non-suicidal self-injury. The tree groups them because both involve self-directed physical harm, which is a reasonable basis for retrieving papers and a poor basis for explaining behavior.

The Four Functions of Self-Injury

The organizing framework of the whole field is a two-by-two classification of why self-injury is maintained. Nock and Prinstein (2004) crossed two dimensions of operant reinforcement. The first dimension is the contingency: whether the behavior is followed by the addition of a stimulus (positive reinforcement) or the removal of one (negative reinforcement). The second is the source: whether the reinforcing consequence is delivered by other people (social, or interpersonal) or by the person's own physiology (automatic, or intrapersonal). Crossing them yields four functions, each a distinct answer to why an individual continues to self-injure.

- Automatic negative reinforcement — the self-injury reduces an aversive internal state, most often unbearable negative emotion. This is the dominant function in NSSI. - Automatic positive reinforcement — the self-injury generates a desired internal state, such as feeling something when numb or dissociated. - Social negative reinforcement — the self-injury removes an unwanted social demand, such as ending an argument or escaping a task. - Social positive reinforcement — the self-injury produces a desired social response, such as care, attention, or communication of distress.

The scheme matters because it is prescriptive, not merely descriptive: function-based assessment identifies which contingency maintains a given person's behavior and points treatment at that contingency, so that a person who self-injures to escape overwhelming affect needs a different intervention from one whose behavior is socially reinforced (Klonsky, 2007). The weight of evidence across both the developmental and the NSSI literatures is that the two automatic (intrapersonal) functions, and the affect-regulation function in particular, are endorsed far more often than the social ones — a conclusion the demonstrations below make concrete. Figure 1 lays out the scheme as the two-by-two it is.

Figure 1

The Four-Function Scheme of Self-Injury

Figure 1. The two-by-two of contingency by source after Nock and Prinstein (2004). The automatic-negative cell, affect regulation, is the function endorsed most often in non-suicidal self-injury.

The first demonstration classifies a vignette by its contingency and source into one of the four cells.

Four-function classifier

Automatic negative reinforcement — self-injury that removes an aversive internal state. Cutting that reduces unbearable negative emotion — the dominant NSSI function.

The two-by-two of contingency (negative/positive) by source (automatic/social) after Nock & Prinstein (2004); cell labels abbreviate the source and contingency initials.

The Experimental Functional Analysis

The four-function scheme did not begin as a questionnaire; it began as an experiment. The behavior-analytic study of developmental self-injury established that the maintaining function of a person's self-injury can be identified not by asking but by systematically varying the environment and watching the rate of behavior respond. Iwata and colleagues (Iwata et al., 1994) conducted the field-defining study: 152 individuals with self-injurious behavior, each exposed to a series of standardized analogue conditions — a demand condition (to test escape/social-negative reinforcement), an attention condition (social-positive), an alone condition (automatic, since no social consequence is possible), and a play control. The function is read off the condition in which self-injury is highest: elevated rates in the alone condition implicate automatic reinforcement; elevated rates under demand implicate escape.

The result was decisive for the whole field. Across the 152 cases the behavior proved orderly — in the large majority a clear, differentiated pattern emerged identifying a specific maintaining variable, and the single most common pattern was sensitivity to social-negative reinforcement (escape from demands), with substantial subsets maintained by social-positive reinforcement and by automatic reinforcement. The demonstration that a behavior as apparently senseless as repetitive head-banging is in fact lawfully controlled by its consequences — and that the controlling consequence can be isolated experimentally — is what made function-based treatment possible: if escape maintains the behavior, one can teach a functional communication response that earns the same break without the injury. The experimental functional analysis is the methodological bedrock on which the later questionnaire measures of function were built and validated.

Affect Regulation: The Dominant Function in NSSI

When the functional lens is turned on non-suicidal self-injury in adolescents and adults, one cell of the two-by-two dominates: automatic negative reinforcement, the reduction of aversive emotional arousal. Chapman and colleagues (2006) formalized this as the experiential-avoidance model: self-injury is negatively reinforced because it terminates or dampens unwanted internal experiences — intense anger, anxiety, shame, or the painful blankness of dissociation — and the relief that follows makes the behavior more probable the next time such arousal builds. The model treats NSSI as an emotion-regulation strategy that is effective in the short term and corrosive in the long term, the same logic that explains other escape-maintained behaviors.

The strongest test of the affect-regulation account is not retrospective self-report but real-time measurement. Houben and colleagues (2017) used ecological momentary assessment — repeated momentary ratings in daily life — in inpatients with borderline personality features and found that negative affect rose before episodes of self-injurious urges or acts and fell afterward, exactly the before-and-after signature the negative-reinforcement model predicts. Because the ratings are collected in the moment rather than recalled, they rule out the memory distortions that weaken retrospective designs, and they show the reinforcement loop operating on the timescale of minutes to hours. This affect-regulation mechanism is why NSSI is understood as a disorder of emotional regulation: the person has learned a reliable, fast-acting, bodily way to down-regulate affect that other strategies fail to touch.

The second demonstration simulates this loop. Negative affect accumulates under stress; a self-injury act sharply reduces it, delivering the relief that negatively reinforces the behavior and raises its probability under the next build-up of arousal.

Affect-regulation reinforcement loop

Affect climbs from a baseline of 40 to 76 over 3 stressed intervals, then an act drops it to 21 — a 55-point relief that negatively reinforces the behavior. Each reinforced act raises the probability of self-injuring at the next build-up by 0.10 from a baseline of 0.30, reaching 0.60 after 3 episodes.

A deterministic illustration of the experiential-avoidance / automatic- negative-reinforcement loop (Chapman et al., 2006); the linear increments are a didactic simplification.

The Self-Injury–Suicide Boundary

Self-injury and suicide are distinguished by intent: non-suicidal self-injury is, by definition, performed without the intent to die, whereas a suicide attempt is performed with at least some intent to end life. The distinction is clinically essential — the two have different functions, different methods, and different management — and it is why MeSH's filing of Suicide as a child of Self-Injurious Behavior is an indexing artifact rather than a claim about mechanism. A recent global meta-analysis of children and adolescents kept the three categories separate precisely because they behave differently in the data, reporting distinct prevalence estimates for non-suicidal self-injury, deliberate self-harm, and suicidal behavior (Lim et al., 2019).

Yet the boundary is permeable in one direction that matters enormously. The interpersonal theory of suicide (Van Orden et al., 2010) argues that a suicide attempt requires not only the desire to die but the acquired capability to carry it out — the capacity to override the powerful self-preservation instinct and tolerate the pain and fear of a lethal act. Capability is not innate; it is acquired through repeated exposure to painful and provocative experiences, and repeated non-suicidal self-injury is one of the most direct such exposures. On this account NSSI is a robust predictor of later suicide attempts not because it is suicidal but because it habituates the person to self-inflicted pain, lowering the barrier to a later act performed with lethal intent. The relationship is developmental and mechanistic, not categorical: self-injury and suicide sit at different points on a pathway linked by acquired capability, which is why a history of NSSI is treated as a serious risk marker even when each individual act is non-suicidal.

The third demonstration makes the function-distribution evidence concrete, splitting a cohort's endorsed functions into the automatic (intrapersonal) and social (interpersonal) categories to show how strongly affect regulation dominates.

Function-endorsement profile

Automatic (intrapersonal) functions account for 73.0% of the endorsed distribution and social (interpersonal) functions for 27.0%. Across the literature the automatic-negative cell — affect regulation — is endorsed far more often than the social functions, which is the empirical basis for treating NSSI as a disorder of emotion regulation (Klonsky, 2007).

Default shares are illustrative of the typical function-endorsement pattern, not values from a specific sample; percentages are normalized to sum to 100.

Worked Example

Consider two cases — Dana, a 16-year-old who cuts, and Reza, an adult with severe intellectual disability who head-bangs — and trace the common functional logic quantitatively.

Classifying the function (Dana). Dana reports that she cuts when overwhelmed by anger or shame and that the distress drops sharply afterward, with no social audience and no change in anyone's demands. The behavior removes an aversive internal state (negative contingency) generated by her own physiology (automatic source): it is automatic negative reinforcement, the affect-regulation function Chapman et al. (2006) describe.

The reinforcement loop (Dana). Model her momentary negative affect on a 0–100 scale. Stress drives it from a baseline of 40 up by 12 points per interval until an act occurs; a self-injury act drops it by 55 points. Starting at 40, after three stressed intervals affect is 40 + 3 × 12 = 76; the act drops it to 76 − 55 = 21. That 55-point relief is the reinforcer. If each reinforced act raises the probability of self-injuring at the next build-up by 0.10 from a baseline of 0.30, then after three reinforced episodes the probability is 0.30 + 3 × 0.10 = 0.60 — the escalation the experiential-avoidance model predicts.

Classifying the function (Reza). An experimental functional analysis exposes Reza to analogue conditions. His head-banging per minute is 0.4 in play (control), 0.6 under attention, 1.1 under demand, and 2.8 when alone. The highest rate, in the alone condition where no social consequence is possible, identifies automatic reinforcement — the pattern Iwata et al. (1994) found in a substantial subset of cases. The rate under demand (1.1) is nearly triple the control rate, so escape (social-negative) is a secondary contributor; attention (social-positive) is near control and is ruled out.

The shared conclusion. Both cases resolve to automatic reinforcement identified by a difference in rate across conditions — Dana's by the before-and-after drop in affect, Reza's by the alone-condition elevation. The two literatures, developmental and NSSI, converge on the same four-function scheme and the same dominance of automatic (intrapersonal) maintenance. The three numbers — a 55-point affect drop, a 0.60 escalated probability, and a 2.8-versus-0.4 alone-to-control rate ratio — are the maintenance signatures the demonstrations above make manipulable.

Discussion

The cognitive-behavioral account of self-injurious behavior is unified by a single commitment: treat the behavior as an operant and ask what maintains it. That commitment dissolves the apparent gulf between the repetitive head-banging of a nonverbal person with intellectual disability and the private cutting of a distressed adolescent. Both are lawful, both are maintained by their consequences, and both are classified by the same two-by-two of contingency and source. The experimental functional analysis supplies the method, the four-function scheme supplies the taxonomy, and the affect-regulation model supplies the content of the most common function. Function-based assessment is not an academic exercise: it is what lets a clinician replace a self-injurious act with a functionally equivalent alternative rather than merely suppressing the behavior.

Two caveats temper the account. First, function is not fixed: a given person's self-injury can serve more than one function, and the function can change over time, so an assessment is a snapshot of a dynamic system rather than a permanent label. Second, identifying a maintaining contingency does not identify a cause — it explains why the behavior persists, not why it began, which may involve temperamental vulnerability, trauma, developmental disability, or modeling. The functional account is a theory of maintenance, and its clinical power comes precisely from the fact that maintenance, unlike origin, is manipulable in the present.

Current Directions

Two fronts are active. The first is the real-time study of the affect-regulation function using ecological momentary assessment: Houben and colleagues' (2017) demonstration that negative affect rises before and falls after self-injury in daily life is being extended with wearable physiological sensors and just-in-time mobile interventions that aim to interrupt the reinforcement loop at the moment of peak arousal. The second is the contested question of whether NSSI should be its own diagnosis. The DSM-5 placed non-suicidal self-injury disorder among the conditions for further study rather than as an established diagnosis, and the review of the empirical case (Zetterqvist, 2015) found reasonable support for the proposed criteria but unresolved questions about the frequency threshold, the boundary with borderline personality disorder, and whether a distinct disorder adds predictive value over the behavior itself. Across both fronts the trend is away from self-injury as a unitary sign and toward a measurable, function-specific, real-time target.

Key Researchers

Armando R. Favazza

(b. 1941, University of Missouri). Authored Bodies Under Siege and built the clinical classification of self-injury into major, stereotypic, and superficial/moderate types — the taxonomy that organized the field and separated self-mutilation from suicide. Wikipedia

Jill M. Hooley

(Harvard University). Co-developed the benefits-and-barriers model of non-suicidal self-injury, asking not only why self-injury regulates affect but why it is chosen over other strategies and why most people never do it. ORCID · Wikipedia · Faculty page

Brian A. Iwata

(1948–2023, University of Florida). Pioneered the experimental functional analysis of self-injurious behavior, establishing through a 152-case study that developmental self-injury is maintained by identifiable reinforcement contingencies and laying the foundation for function-based assessment and treatment. Wikipedia · Faculty page · Google Scholar

E. David Klonsky

(University of British Columbia). Reviewed and weighed the competing functions of self-injury to establish affect regulation as the best-supported, and developed the Three-Step Theory bridging self-injury and suicide risk. ORCID · Faculty page · Google Scholar

Matthew K. Nock

(b. 1973, Harvard University). Crossed contingency and source into the four-function model of self-injury and built the integrated theoretical account of why people hurt themselves, anchoring the modern self-injury and suicide research program. ORCID · Wikipedia · Faculty page

Glossary

Acquired capability.

In the interpersonal theory of suicide, the learned capacity to override self-preservation and tolerate the pain and fear of a lethal act, raised by repeated exposure to self-injury.

Affect regulation.

The down-regulation of aversive emotional arousal; the automatic-negative-reinforcement function that dominates non-suicidal self-injury.

Automatic reinforcement.

Reinforcement delivered by the person's own physiology rather than by other people; the intrapersonal source in the four-function scheme.

Ecological momentary assessment (EMA).

Repeated real-time sampling of experience in daily life, used to test the affect-regulation account by measuring negative affect before and after self-injury.

Experiential avoidance.

The attempt to escape or suppress unwanted internal experiences; Chapman's model casts self-injury as negatively reinforced experiential avoidance.

Four-function scheme.

The two-by-two classification of self-injury's maintaining function, crossing contingency (positive vs negative reinforcement) with source (automatic vs social).

Functional analysis.

The experimental identification of a behavior's maintaining function by systematically varying environmental conditions and observing the rate of behavior.

Negative reinforcement.

The strengthening of a behavior by the removal of an aversive stimulus; the contingency behind the affect-regulation function of self-injury.

Non-suicidal self-injury (NSSI).

Deliberate, direct destruction of one's own body tissue without suicidal intent, most often self-cutting, in people without developmental disability.

Operant.

A behavior controlled by its consequences rather than by an eliciting stimulus; the central claim of the cognitive-behavioral account is that self-injury is an operant.

Positive reinforcement.

The strengthening of a behavior by the addition of a stimulus; in self-injury, generating a desired internal state or social response.

Social reinforcement.

Reinforcement delivered by other people — attention, care, or escape from demands; the interpersonal source in the four-function scheme.

Stereotypic self-injury.

The repetitive, rhythmic self-injury (head-banging, self-biting) seen in intellectual disability and autism, often maintained by automatic reinforcement.

Superficial/moderate self-injury.

Favazza's band of common, low-lethality self-injury (cutting, burning, scratching) that encompasses most NSSI.

Frequently Asked Questions

Is self-injurious behavior the same as a suicide attempt?

No. Self-injurious behavior, as the term is used here, is performed without the intent to die, whereas a suicide attempt is performed with at least some intent to end life (Lim et al., 2019). The two differ in intent, function, and method, which is why they are managed differently, even though repeated self-injury raises the risk of a later attempt.

Why do people injure themselves if it hurts?

Because it works: the pain is outweighed by a reinforcing consequence, most often the reduction of overwhelming negative emotion. Self-injury is maintained by automatic negative reinforcement, in that it rapidly lowers aversive arousal, which is why it is understood as a maladaptive emotion-regulation strategy (Chapman et al., 2006).

What are the four functions of self-injury?

Self-injury is classified by crossing two dimensions: whether it adds or removes a stimulus (positive vs negative reinforcement) and whether the consequence comes from others or from the body (social vs automatic). The four cells are automatic-negative, automatic-positive, social-negative, and social-positive reinforcement (Nock & Prinstein, 2004).

How is the function of someone's self-injury identified?

Experimentally, by functional analysis: the person is exposed to standardized conditions (alone, attention, demand, and play) and the function is read off the condition in which self-injury is highest. Iwata and colleagues established this method in a 152-case study (Iwata et al., 1994).

How common is self-injury?

It depends on the population. In adults with intellectual disabilities, point prevalence of self-injurious behaviour is around 4 to 5%, rising with severity of disability (Cooper et al., 2009). In the general adolescent population, lifetime non-suicidal self-injury is estimated at roughly 17% (Swannell et al., 2014).

Is self-injury in intellectual disability the same as cutting in teenagers?

They are different in form and context but share a common functional logic. The repetitive, stereotypic self-injury of intellectual disability and the self-cutting of adolescents are both operants maintained by their consequences, classified by the same four-function scheme, and both most often maintained by automatic reinforcement (Nock, 2010).

Is non-suicidal self-injury a mental disorder?

Not an established one. The DSM-5 lists non-suicidal self-injury disorder among the conditions for further study; the empirical review finds reasonable but incomplete support, with open questions about thresholds and the boundary with borderline personality disorder (Zetterqvist, 2015).

Does self-injury predict suicide?

Yes, as a risk marker. Repeated self-injury is one of the strongest predictors of later suicide attempts, not because it is suicidal but because it builds the acquired capability (habituation to self-inflicted pain and fear) that the interpersonal theory of suicide identifies as necessary for a lethal act (Van Orden et al., 2010).

References