Psychiatric patients who engage in self-destructive behavior by cutting, burning, or abrading their skin are currently one of the most difficult-to-treat groups in both inpatient and outpatient settings. The complexities of treating these patients, the risk factors associated with this symptom, and the rise in the prevalence of self-mutilation in America's adolescents and young adults provided the impetus for the current study. This article explores aspects of aggression, dependency, object relations, defensive structure, and psychic boundary integrity that may contribute to the genesis and maintenance of self-mutilation. Rorschach protocols from 90 borderline personality-disordered inpatients (48 self-mutilators and 42 non-self-mutilators) were scored using five psychoanalytic content scales. Results indicate that self-mutilating patients exhibit greater incidence of primary process aggression, severe boundary disturbance, pathological object representations, defensive idealization, devaluation, and splitting than did a matched group of non-self-mutilating borderline patients. Clinical theory and technical recommendations are considered in light of the current empirical findings. (Bulletin of the Menninger Clinic, 64[3], 365-385)
For the past 70 years, analysts have worked to comprehend the unconscious structures that are thought to be responsible for masochistic fantasies and behaviors of humiliation and physical pain (Bollas, 1992; Bonaparte, 1952; Deutsch, 1930; Doctors, 1981; Freud, 1923/1961b,1924/1961a; Kernberg, 1984; Menninger, 1938; Reich, 1940; Smirnoff, 1969; Stolorow & Lachmann, 1980). Masochism, in the form of bodily self-destruction, negativism, and negative therapeutic reactions, has consistently presented a perplexing theoretical problem and a serious challenge to effective treatment of neurotic, borderline, and psychotic disorders (Freud, 1924/1961a; Joseph, 1982; Olinick, 1964; Riviere, 1936). Modern clinicians are faced with similar theoretical complications in comprehending the underlying psychic structures of self-mutilating individuals. These patients frequently become embroiled in negative therapeutic reactions and masochistic transference impasses that destroy the frame of the treatment and can, in the most extreme cases, lead to the suicide of patients.
Self-mutilating patients are predominantly adolescent and young adult females who engage in self-destructive behavior by cutting, burning, or abrading their skin. Like masochists of Freud's era, these patients are currently one of the most challenging and treatment-refractory groups of seriously disturbed patients (Favazza & Conterio, 1988; Favazza & Favazza, 1987; Pao, 1969; Simpson, 1980). The impact of their sadomasochistic attacks on their body and on the frame of the treatment can be deeply disturbing to those who treat them. In short, countertransference fear, repulsion, and rage can interfere with the therapist's ability to consider the intrapsychic meaning of cutting as a defensive reaction, as a transference communication, or as a suicide attempt.
When therapists attempt to manage self-destructive actions without a clear interpretive frame for understanding and communicating their interventions, therapist and patient can fall into a reflexive mode of chronic crisis management (Sacksteder, 1997). The potential for transference enactments to go unanalyzed by the therapist is greatest at such moments because the primary task of treatment can be lost in the midst of crises. This type of disengagement from the interpretive task is a serious impediment to the treatment of a condition that is often erroneously considered simply "manipulative," but that has serious consequences for the patient.
Past research has shed light on the dangers and complexities of this phenomenon by exploring demographic and personal characteristics (Favazza, 1989; Feldman, 1988), the phenomenology of patients (Himber, 1994; Leibenluft, Gardner, & Cowdry, 1987), and clinical observations and treatment approaches (Kernberg, 1987; Stone, 1987; Tantam & Whittaker, 1992). The data on suicide rates for self-mutilating borderline patients are twice that for other borderline patients (Stone, Hurt, & Stone, 1987). Other risk factors include the potential for developing stereotyped and chronic self-mutilation due to increased release of endorphin metenkephelin (Cold, Allolio, & Rees 1983), contagion through modeling (Rosen & Walsh 1989; Walsh & Rosen, 1985), and the development of an identity as a "cutter" (Podvolt, 1969).
While the etiology is complex, self-mutilators frequently report childhood experiences of emotional and physical trauma, including sexual abuse, early traumatic invasive surgery, and neglect (Cross, 1993; Favazza & Conterio, 1988; Feldman 1988). These traumatic events are thought to create severe disturbances in object relations (Berliner, 1958; Chasseguet-Smirgel, 1995; Kernberg, 1984), in the capacity to symbolize experience (Muller, 1996), and in maintaining an ongoing experience of self that is free from derealization and dissociation (Favazza, 1992; Feldman, 1988; Rosenthal, Rinzler, Walsh, & Klausner, 1972; Soloff, Lis, Kelly, Cornelius, & Ulrich, 1994).
The purpose of this study is to investigate psychoanalytic concepts of self-mutilation empirically by exploring object representations, affect, and thought disorder manifest in two similar groups of inpatient Rorschach protocols. Assessing unconscious processes manifest in projective test data allows for an analysis of greater depth and breadth than do self-report data, while maintaining the advantages inherent in large-sample investigations. In order to narrow the range of unconscious processes to be investigated, we reviewed the vast array of psychoanalytic explanations of masochism and self-mutilation.
Modern ego psychology, object relations theory, and self psychology provide insights into the structural deficits and the underlying motivations that maintain self-mutilation. We have therefore drawn from several theorists' work (Doctors, 1981; Kernberg, 1984; Stolorow & Lachmann, 1980) as a springboard for constructing a study of the unconscious correlates of self-mutilation using Rorschach responses from borderline inpatients.
Kernberg (1984) proposed that patients who self-mutilate tend to be infantile, narcissistic, and "as if" personalities, functioning at an overt borderline level. In histrionic and infantile personality disorders, self-mutilation occurs in the context of some real or imagined failure or thwarted wish. The act is considered an unconscious coercive effort to gain control over the environment by inducing guilt in the frustrating other. Kernberg observed a more severe form of self-mutilation in "malignant narcissistic" patients. These patients tend to present as aloof and uninterested in others, rather than clinging. When their grandiose self-image is challenged, they experience intense humiliation, and in an effort to rid the self of unbearable humiliation, these patients physically enact, through self-mutilation, an unconscious triumph over the internal world of objects as they cheat death and pain. These patients flaunt their sense of superiority over their therapists when they discover that the therapist is chagrined or frightened by their self-mutilation. A third category of patient suffering from atypical psychoses is far more rare, and can be identified by the bizarre nature of the self-mutilation and the unusual degree of cruelty and idiosyncratic features. Each distinct subgroup suffers from primitive pathological object relations, primitive defenses, rage, and narcissistic vulnerabilities.
Stolorow and Lachmann (1980) have suggested that masochistic activity often occurs in response to a fear of fragmentation and that it represents an abortive (and sometimes sexualized) attempt to restore the cohesion and stability to a fragmenting self-representation. Building on decades of psychoanalytic exploration of masochism, the authors assert that the most violent and primitive expressions of self-directed aggression are found in individuals who are most vulnerable to narcissistic injury. Relatively minor frustrations such as the loss of control over a relationship and failed mirroring may threaten the structurally deficient narcissistic individual. In such cases, the need for experiences of acute pain, localized on the skin surface, may help buttress or protect against the derealization, sensations of falling apart, and dissociation that is commonly associated with a dissolving self-other boundary.
In a similar analysis, Doctors (1981) concluded that self-mutilating adolescents have not obtained emotional object constancy, exposing them to defensive splitting of good and bad part-objects. Furthermore, the lack of sufficient differentiation between self and objects results in an unstable array of internal representations in which even minor frustration can rapidly develop into intense rage that inundates the object world, and thereby spoils precariously maintained self-representations and object representations. This spoiling of the inner world, combined with a fragmenting ego boundary, further blurs the patient's tenuously held distinction between self and other. The breakdown in the capacity to maintain self-other differentiation has been linked to self-reports of falling to pieces, or feeling empty, unreal, and dead inside. According to this theory, a main unconscious motivation of self-mutilation is the effort to repair and restore a sense of self-cohesion and to expunge bad part-objects in order to restore good object representations.
Significant theoretical and clinical differences exist between ego psychology and self psychology, especially in the understanding of narcissistic and borderline disturbances. Despite these differences, there are fundamental similarities across theoretical orientations that are particularly important to this investigation. All three theories assume that the act of cutting is an unconscious effort to reestablish some degree of narcissistic equilibrium and self-cohesion. For Kernberg (1984), pathological affect and object representations set the stage for self-mutilation, especially when patients feel deprived, abandoned, or insulted by the therapist. The motivation to master and control another object is an effort to restore narcissistic balance, either through a dependent yet coercive object relation or through a coercive insistence on mirroring of the patient's grandiose self-organization. For Stolorow and Lachmann (1980), narcissistic injury produces intense rage in the structurally deficient ego, which can lead to sensations of fragmentation and boundary dissolution. For Doctors (1981), a narcissistic injury may precipitate an affective storm that inundates the patient's object world, transforming precariously maintained positive imagoes into malevolent objects.
These clinical theories find some support in descriptive studies and to a lesser extent in large-scale empirical research. For example, Stolorow and Lachmann's assertion that these patients use acute experiences of pain to bolster crumbling ego boundaries finds support in case descriptions (Cross, 1993; Doctors, 1981; Favazza, 1989; Gardner & Gardner, 1975; Miller & Bashkin, 1974; Pao, 1969; Simpson & Porter 1981). Researchers have found that self-mutilators are significantly more prone to derealization and drug-free hallucinations/delusions than are non-self-mutilating borderline patients (Soloff et al., 1994), and self-mutilators have higher rates of serious identity disturbance than the non-self-mutilators (Simeon et al., 1992). The clinical inference that aggression is a key factor in the act of self-mutilation has equivocal research support. Soloff and colleagues (1994) found that self-mutilating patients scored lower on the Buss-Durkee Hostility Inventory than did clinical controls, and that physical assaults on others were negatively correlated with self-mutilation. By contrast, Simeon and colleagues (1992) reported that self-mutilators produced significantly higher scores on the Minnesota Multiphasic Personality Inventory (MMPI) Psychopathic Deviance Scale (PD), yet did not differ significantly from controls on the Buss-Durkee Hostility Inventory. Finally, Rosenthal and colleagues (1972) noted that most of the 24 female patients reviewed in their study feared their own anger, described an inability to cope with these feelings, experienced derealization, then cut themselves in an effort to reintegrate. These equivocal findings may have more bearing on the patients' willingness or ability to report their subjective affect states than it does on the nature of their affective life. Thus aggression may be related to self-mutilation, but it has been unclear to what extent hostility and aggression are consciously registered, contained, and modulated by the ego.

From the preceding clinical and empirical data, we have formulated the following hypotheses:

The initial sample consisted of 224 patients admitted to The Austen Riggs Center. Patient records from January 1993 to June 1997 (including identification numbers, diagnostic codes, detailed descriptions of specific behavioral manifestations of self-mutilation, and medical procedures performed in response to the self-destructive activity) were first masked to disguise patient identity, then downloaded from the Center's database. Behavioral records were then classified by Dr. Fowler into self-mutilating and non-self-mutilating groups prior to collecting archival Rorschach records. The data extracted from the medical records can be considered a reliable and relatively accurate representation of the patients' self-destructive activities during hospitalization because the nursing staff was required to record all incidents of self-inflicted lacerations and burns.
The primary diagnosis of borderline personality disorder (BPD) was established in a consensus case conference at the culmination of the initial evaluation and treatment phase 4-6 weeks following admission. Diagnoses were made using available sources of information, including an integration of interview data from the admission consultation, initial contact with the therapist and psychopharmacologist, consultations with outpatient therapists, prior hospital records, and interviews with relatives to clarify family history of psychiatric disorders, life history, and premorbid level of functioning. All patients were assessed by a board-certified and licensed psychologist and psychiatrist. Diagnoses were assigned according to the diagnostic criteria of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV; American Psychiatric Association, 1994). This method of diagnostic practice approximates the LEAD (longitudinal expert evaluation using all data) standard of diagnosis (Pilkonis, Heape, Ruddy, & Serrao, 1991; Skodol, Rosnik, Kellman, Oldham, & Hyler, 1988; Spitzer, 1983). The diagnosis of BPD was confirmed in 100% of the cases by an independent rating conducted by either a psychiatrist or psychologist as an ongoing aspect of the hospital's performance improvement policy. It is important to note that in all cases the diagnosis of BPD was determined independently of Rorschach data. Although all patients in this study had multiple Axis I and Axis II disorders, those patients with a comorbid Axis I psychotic disorder were excluded from the study.
Criteria for inclusion into this study required a hospital stay of 6 months or more in order to obtain a representative sample of patient behavior. All patients had completed a battery of projective tests administered during the first 30 days of the index hospitalization. In addition, an act of self-mutilation had to occur within 60 days following administration of the Rorschach to ensure the temporal relevance of the Rorschach data. The final sample of 90 adult inpatients consisted of 2 males and 88 females with a mean age of 30.9 years (SD = 9.1) at admission. The average number of years of education completed by the patients was 14.96 years (SD = 1.9). A total of 54 patients were single, 24 were married, and 9 were divorced or widowed.

Administration and original scoring of the Rorschach followed the procedures articulated by Exner (1993). These protocols were later rescored on the Rorschach content scales by Dr. Fowler, who was blind to patient identity, group inclusion, and diagnosis. For the purpose of interrater reliability (Weiner, 1991), 20 Rorschach protocols were chosen at random and scored independently by Dr. Hilsenroth, who was also blind to all patient data. The two sets of scored protocols were then compared, and percentages of correct agreement and intraclass correlations were calculated. The resulting interrater agreement and intraclass coefficients are presented in Table 1.
Aggressive ideation was assessed using Holt's (1977) method for scoring primary and secondary process manifestations on the Rorschach. Holt's system differentiates two levels of aggressive content. Level I scores are related to primary process forms of primitive aggressive themes, specifically measuring murderous or palpably sadomasochistic aggression. Level 2 aggression scores are related to secondary process ideation and specifically measure indirect forms of hostility or aggression expressed in more socially acceptable ways. The Holt system has demonstrated high levels of test-retest (Gray, 1969) and interrater reliability (Fowler, Hilsenroth, & Handler, 1995), as well as construct validity in a number of studies (Blatt & Berman, 1984; Hilsenroth, Hibbard, Nash, & Handler, 1993).
The Rorschach Oral Dependency scale (ROD; Masling, Rabie, & Blondheim, 1967) was developed as a psychoanalytic content scale to assess oral/dependent imagery. A response is defined as oral dependent if it falls into any of the following categories: food and drinks, food sources, food objects, food providers, passive food receivers, food organs, supplicants, nurturers, gifts and gift givers, good luck symbols, oral activity, passivity and helplessness, pregnancy and reproductive anatomy, and negations of oral percepts (e.g., not pregnant; man with no mouth). The construct validity and interrater reliability regarding this measure have proven to be excellent in more than 90 experimental studies utilizing various populations (Bornstein, 1996).
The Mutuality of Autonomy Scale (MOA; Urist, 1977) assesses the thematic content of relationships (stated or implied) between animal, inanimate, and human representations in Rorschach percepts. The scale was developed to assess the degree of differentiation of object representations, focusing primarily on the developmental progression of separation-individuation from engulfing, fused relations to highly differentiated self-other representations. Scale points 1 and 2 reflect the capacity to construe self- and other representations as structurally differentiated and engaged in mutually interactive or parallel activity (e.g., "two people talking about grocery prices, pushing shopping carts"). Scale points 3 and 4 capture dependent and mirroring object relationships and often reveal an emerging loss of autonomy between figures (e.g., "Siamese twins connected at the waist"). Scale points 5, 6, and 7 reflect not only the loss of the capacity for separateness but also increasing malevolence (e.g., "an evil fog engulfing this frog ... smothering it"). Reliability data are excellent (Tuber, 1989), and the scale has demonstrated a high degree of construct validity with behavioral ratings (Ryan, Avery, & Grolnick, 1985; Urist, 1977; Urist & Schill, 1982), assessment of therapeutic change (Blatt & Ford, 1994), and multimethod assessment of the construct (Fowler, Hilsenroth & Handler, 1995; Urist, 1977). For this study, we chose a composite score of all level 5, 6, and 7 pathological scores (PATH; Berg, Packer, & Nunno, 1993) because it has been found to be a robust and stable measure of pathological object relations.
Defensive structures were assessed using the Lerner Defense Scale (LDS; Lerner & Lerner, 1980). This scale is based on Kernberg's (1975) theoretical conceptualizations and other commentators' clinical observations (Holt, 1977; Mayman, 1967; Peebles, 1975). Primitive defenses of splitting, idealization, devaluation, and denial represented in percepts of human, quasi-human, and human detail (Hd) responses were assessed for this study. The LDS has shown good construct validity and high interrater reliability (Lerner, 1991). To use more stringent parametric statistics in the analysis of those defenses that are ranked on a continuum from high to low order (devaluation, 1-5; idealization, 1-5; and denial, 1-3), defenses were weighted according to rank and then were collapsed into an overall score for that category. For example, if there are three instances of idealization on a subject's protocol, one Level 1 and the other two instances at Level 3, the subject would receive a total idealization score of 7 (1 + 3 + 3 = 7).
The Boundary Disturbance and Thought Disorder Scale (BDS; Blatt & Ritzler, 1974) assesses an individual's capacity to maintain distinctions between objects along cognitive/perceptual and affective dimensions. Blatt and Ritzler drew connections between the degree of thought disorder present on the Rorschach and the concomitant degree of ego boundary dysfunction. Drawing on Rapaport's indices of thought disorder, they proposed the following hypotheses: (1) Mild forms of ego boundary fragmentation or looseness of boundary (boundary laxness) could be measured by fabulized combination. (2) More severe problems of differentiating fantasy from reality (inner/outer boundary disturbance) would be represented in responses containing confabulations. (3) The most severe form of boundary fragmentation and disintegration (self/other boundary disturbance) would be captured in the severely thought-disordered responses known as contaminations. Several studies (Blatt & Ritzler, 1974; Lerner, Sugarman, & Barbour, 1985; Wilson, 1985) have found that borderline patients typically have greater difficulty with boundary laxness and inner/outer boundaries, whereas schizophrenic patients typically have greater difficulty distinguishing between self/other boundaries. The more severe self/other boundary disturbance may correspond to what many have described as the crumbling ego boundaries, dissociation, and drug-flee hallucinations observed in many patients who self-mutilate.

Prior to assessing the main hypotheses, analyses of variance (ANOVA) were conducted in order to identify potential confounding variables. Contrasts of the two groups (see Table 2) revealed no significant differences in age (F = .86, p = .36), level of education (F = .26, p = .61), full-scale IQ (F = 1.6, p = .20), or Rorschach productivity (F = .12, p = .72). The groups were also well matched on gender (one male in each group) and marital status (predominantly single, never married). These well-matched groups ensure that results obtained in further analyses are most likely based on actual psychological differences between the groups.
Results of a multivariate analysis of variance (MANOVA) demonstrated significant differences across the two borderline groups using all content scales (F = 2.90, p = .004). Therefore the results of the univariate analyses (controlling for chance significance with Bonferroni adjustment) most likely reflect actual differences between groups. Univariate analyses (see Table 3) demonstrated that self-mutilating patients produced a greater number of primary process aggression responses than did non-self-mutilating borderline patients (F = 15.03, p < .0001), whereas no differences were found in terms of secondary process aggression (F = 1.9, p = .18). The estimated effect size for primary process aggression yields a medium to large effect size (Eta = .38) and accounts for approximately 14% of the variance between the groups.
The degree of oral/dependent imagery was not significantly greater for the self-mutilating borderline patients than for the other borderline patients (F = 2.97, p = .09). In the realm of object relations, self-mutilating borderline patients manifested more instances of malevolently controlling object representations as assessed by the PATH score of the Mutuality of Autonomy Scale (F = 10.63, p = .002). Estimated effect size for the PATH score reflects a medium to large effect size (Eta = .33) and accounts for approximately 10% of the variance between groups.
In assessing how self-mutilating patients manage powerful affects and conflicts, we contrasted the groups' use of primitive defenses measured by the Lerner Defense Scale. Of the four defenses as entered into the analyses, primitive splitting (F = 20.07, p < .0001), idealization (F = 14.0, p < .0001), and devaluation (F = 5.4, p = .02) were manifested in the Rorschach records of self-mutilating borderline patients with greater frequency and intensity than in the records of the non-self-mutilating patients. Primitive denial was not significantly different between groups (F = .35, p = .86). Estimated effect size for splitting reflects a large effect (Eta = .43) and accounts for approximately 19% of the variance between the groups, whereas idealization produced a medium effect size (Eta = .36), accounting for 13% of the variance. Devaluation yielded a small effect size (Eta = .22) and accounted for only 5% of the variance.
The final contrasts involved the boundary disturbance and thought disorder scale. Contrasts highlighted expectable similarities in that the groups were not significantly different in the degree of boundary laxness (F = .01, p = .91) or inner/outer boundary disturbance (F = 3.1, p = .08). A striking difference in the self-mutilating borderline patients' self/other boundary disturbance (F = 12.6, p < .0001) highlighted a higher incidence of self-other boundary disintegration than in other borderline patients. A medium to large effect size (Eta = .36) for self/other boundary disturbance demonstrated and accounted for approximately 13% of the variance between the groups.
A post-hoc stepwise logistic regression analysis was performed in order to identify the set of Rorschach variables that, combined, account for the greatest variance between self-mutilating and non-self-mutilating borderline patients. Results revealed that defensive splitting (R = .44; R2 = .19), primary process aggression (R = .47; R2 = .23), self-other boundary disturbance (R = .49; R2 = .25), and idealization (R = .51; R2 = .26) accounted for the greatest variance in discriminating between these two similar diagnostic groups.

The Rorschach results provide evidence for the wide-ranging psychological differences between self-mutilating borderlines and other inpatient borderlines. The tendency for self-mutilating patients to invoke primitive defenses of splitting, idealization, and devaluation has been observed in various analytic encounters (Doctors, 1981; Kernberg, 1984). Kernberg, for example, observes a preponderance of part-object relations and primitive splitting, idealization, devaluation, and denial in these patients. In our investigation, massive idealization and defensive splitting were the primary modes of defense that differentiated self-mutilators from other borderline patients. Kernberg formulates the defensive functions of idealization and splitting as the primary mechanisms to ward off hate and envy. In the extreme case, these defenses are coupled with serious ego deficits and malevolent object/affect representations (including primitive superego elements); the result often reflects a grandiose self-structure in what Kerberg (1984) refers to as a "malignant narcissistic disorder" (p. 293). Novick and Novick (1991) come to a somewhat similar conclusion when they observe that grandiosity and delusions of omnipotence frequently underlie the masochistic activity of patients. These patients exhibit particular ego deficits in affect regulation that expose them to primary process hate and overstimulated libidinal impulses. These volatile affects then interact with "a fragile defense system and a deficient superego to produce the delusion that only they themselves were powerful enough to inhibit their omnipotent impulses, and then only by resorting to severe masochistic measures" (p. 311).
A second possible interpretation of the breakdown in idealization represented by self-mutilation relates to certain self psychology formulations (Stolorow & Lachmann, 1980). First, idealization of selfobjects is suggestive of a mirroring transference in which idealization is used to ward off potential narcissistic injuries that are inevitable in an exploratory treatment. In the case of an idealizing mirroring transference, a therapist's failure to live up to the idealization may result in massive frustration and rage that can further destabilize precarious ego boundaries. These failures can be as subtle as a poorly timed interpretation or the therapist's refusal to extend the session by a minute or two. At such moments of disappointment, it is common for therapists to feel the effects of splitting as they become the object of fierce rage and devaluation when they fail to live up to the patient's wished-for perfection.
The most consistently held hypothesis for why psychiatric patients self-mutilate is the belief that these patients experience intense episodic rage that is expressed through their masochistic attacks on their bodies. Past research linking aggression and self-mutilation has been inconclusive, most likely because these studies have relied on self-report measures that are transparent, and for which the illusion of affect control can be perpetuated (Shedler, Mayman, & Manis, 1993). In addition, self-report measures of hostility do not assess primary process aggression--a form of aggressive experience that is not readily available for conscious articulation.
When patients' conscious defenses are circumvented with a projective technique such as the Rorschach, it is apparent that self-mutilating borderline patients exhibit significantly greater problems with the control of primary process aggression than a matched group of borderline inpatients. In short, self-mutilators are prone to primary process rage, and whether patients experience such rage through ego-syntonic and consciously registered channels, or experience themselves as the victim of others' rageful attacks (through a process of disavowal and projection), it is clear that they struggle with rage. The repeated breakthrough of unmodulated sadomasochistic aggression can be interpreted as an indication of self-mutilating patients' failing capacity to bind aggressive affects through secondary process modes of expression. As is the case with most people, such intense affective states can tax the ego's capacity to process information through symbolic channels. In the case of the self-mutilating inpatients in our sample, it is conceivable that a predisposition for intense rageful affects can be triggered by frustrations that can lead to a spoiling of their inner world, eliciting primitive defenses and compromising boundary integrity.
The problem of thought disorder and boundary disturbance is evident in the frequency with which clinical reports detail a sequence of intense emotional turmoil, followed by sensations of emptiness, fears of falling to pieces, and experiences of derealization and dissociation (Rosenthal et al., 1972). These reports correspond to the manifestations of severe boundary disturbance demonstrated in the self-mutilator's Rorschach protocols. Whether these patients sustain narcissistic injuries or become overwhelmed by an upsurge in aggressive drive derivatives, it seems plausible to assume that such powerful affects tax their ego capacities for maintaining a clear understanding of internal and external affects, controls, and prohibitions. For the more disturbed borderline patient, this confusion of self and other leads the patient to engage in power struggles with others in an attempt to create some clearer boundary between self and other. Patients struggling to maintain self-cohesion may subject themselves to humiliation, pain, and even the risk death in order to re-create a stable sense of self.
Recent contributions by Muller (1996) and Gedo (1996) support the hypothesis that self-mutilation is enacted in order to ward of fragmentation of the self-organization. Muller (1996) has proposed that patients who compulsively self-mutilate do so in an effort to fend off merger into a dedifferentiated existence by creating a concrete marker of their outer boundary: "signifying that there is a boundary between the subject and the Real, it prevents a dedifferentiated existence" (p. 86). According to Muller, the collapse into dedifferentiation is due to a breakdown in semiotic structures, wherein signs and symbols are often corrupted by incest, violence, and a chaotic existence that disrupts the formation of a stable semiotic system. Gedo's (1996) comments on this phenomenon are also relevant to this discussion: "Clinical experience has taught us that embarrassment--even humiliation--is sometimes insufficient to override motivations stemming from the need to maintain the stability of self-organization.... One common example is that of compulsive self-mutilation" (p. 175). Again, the manifestation of masochistic activity can be linked to underlying structural deficiencies in ego boundary integrity.
Most modern theories (Cross, 1993; Doctors, 1981; Kernberg, 1984) of self-mutilation posit a fundamental disturbance in object relations, with particular emphasis on the self-mutilator's unconscious fear of being controlled by powerful malevolent forces. Often this fear takes the form of merging with, or being overtaken by, dangerous objects. The clinical corollary to this internal imbalance is the development of a hypervigilant and rigidly held fear of being influenced by passive experiences of desire and dependency, and fear of being influenced by others (Piers, 1999; Shapiro, 1981). The fact that self-mutilating borderline patients manifest more poorly differentiated and malevolently controlling objects than do other borderline inpatients provides empirical support to clinical experience of transference enactments in which these patients violently ward off any experiences of being influenced by therapists (Doctors, 1981). This finding also lends credence to Kernberg's (1987) observation that patients self-mutilate in response to "intense feelings of resentment, rage, and impotence in the effort to control an important person (including the therapist), and that the experience of self-mutilation is the relieving enactment of revenge" (p. 344).
Although self-mutilation is an unwelcome but expectable symptom in the treatment of some borderline and narcissistic disorders, the greatest danger in treating these patients occurs when transference enactments degenerate into ever-escalating self-destruction toward suicide. Clinical accounts abound in which patient and therapist become embroiled in struggles to contain and control the patient's self-destructive acts. One consistent observation in treatments plagued with recurrent countertransference impasses is the therapist's intolerance for hate in the transference. Those for whom hatred is an unacceptable response from patients will often resort to sentimentality as a means of dealing with an enraged and enraging patient (Fromm, 1995; Winnicott, 1975). Through their efforts to contain self-mutilation (e.g., through sympathetic offerings of transitional objects), these therapists may inadvertently communicate to the patients that they want them to be different than they are without a clear understanding of the meanings of self-directed aggression.
These results provide interesting evidence for the differences between two groups of inpatient borderlines; nevertheless, too great a reliance on statistical significance can obscure the complexity and the variations found in the self-mutilating group. For example, the great dispersion of scores around the group means is suggestive of great variation within the self-mutilating sample. This finding is not terribly surprising, given the fact that clinical experience has taught us that borderline patients rarely fit a prototypical pattern, and are perhaps best understood as a collective of patients at our boundary of understanding (Fromm, 1995).
In light of these limitations, we propose a tentative working model for understanding compulsive self-mutilation based on the "best fit" among clinical observations, psychoanalytic theory, and the results of the present study. Several preexisting conditions seem to place patients at risk for self-mutilation. A number of chaotic life events, such as physical traumas, and radically inconsistent environments may ultimately lead to overwhelming experiences of a malevolent and chaotic world. Boundary transgressions, lack of stable semiotic structures, and exposure to unbearable emotional experiences overwhelm the individual's capacity to integrate emotional experience and interfere with the establishment of a stable self-organization. Complications arising during puberty may have an adverse impact on the predisposed adolescent's capacity to differentiate from the mother (Chasseguet-Smirgel, 1995; Cross, 1993). Struggles over the psychic ownership of body and self may help to explain the sudden emergence of self-mutilation and eating disorders in adolescence--during puberty--when the young girl's body develops a likeness to her mother's. The formation and subsequent arrest in the development of defenses, and the internalization of precariously differentiated self- and other representations, further complicate the capacity to assume ownership of the body.
Noxious life events and their subsequent pathological adaptations make these patients vulnerable to ego fragmentation and psychotic regression. These vulnerabilities may lead patients to be extremely sensitive to any challenge to their self-integrity, and may lead to a defensive insistence on controlling their environment. Thus any challenge to their sense of control over themselves and the environment may upset their fragile narcissistic equilibrium, leading to rage and unmodulated aggression. Once in the midst of such rageful spoiling, patients may engage a series of primitive defenses such as idealization, splitting, and devaluation in order to ward off fragmentation. When these defenses fail, these patients may self-mutilate in a desperate effort to rid themselves of noxious affects, and unconsciously to assert control over their fused sense of body, the other, and the outer world. By cutting the outer boundary of the body, they may in fact create a concrete marker of their differentiation from the environment.
Such a tentative formulation precludes any definitive statement on treatment; nevertheless some technical recommendations may provoke further thought and inquiry into this treatment-resistant population. Integrating research findings from this study and Plakun's (1994) principles for the psychodynamic treatment of self-destructive borderline patients at serious risk for suicide has proven quite useful. While Plakun focuses primarily on the problem of suicide attempts that seriously threaten the continuity of the treatment alliance, we have successfully applied several principles to the treatment of self-mutilating patients.
At the outset of treatment, we recommend that the therapist set a frame about how crises will be managed, including the exploration of self-destructive acts as part of the interpretive process. These efforts to come to a mutually agreed-upon protocol for handling crises and an agreed-upon interpretive task will help to highlight the differentiated role and task for patient and therapist. We have found that this recommendation helps patients use the clearly articulated frame to temporarily reinforce their limited capacities for self-other differentiation and often eliminates the power struggle over interpreting the self-destructive acts.
A second recommendation, to metabolize countertransference reactions prior to responding to the suicidal or self-destructive behavior, is particularly appropriate when one considers the self-mutilating patient's disturbed object relations, vulnerability to narcissistic injury, and preponderance of primary process aggression. The capacity to acknowledge and tolerate countertransference hate is requisite for treating psychotic and borderline conditions, and has been particularly fruitful in our experience. This is not to suggest that therapists silence their reactions, but rather to modulate their responsiveness carefully to match the task of the treatment. The risks inherent in therapists' unmetabolized self-disclosures of guilt or rage are greatest because they may lead the patient to conclude that the therapist is out of control and dangerous. Equally disastrous is the patient's interpretation of the therapist's unbridled reaction as a challenge to the patient's autonomy.
Defense analysis and an exploration of the patient's experience of self-mutilation before interpreting the latent aggression may be particularly important because the patient's capacity for recognizing somatic sensations as emanating from within his or her own body as a signal of rage is often lacking. The timing and empathic delivery of interpretations of self-destructive acts seem particularly relevant, given our findings of self-mutilating patients' unmodulated primary process aggression on the Rorschach. Such unmodulated internal states need to be interpreted and opened for the fullest range of experience so that patients integrate these experiences of anger through secondary process channels of fantasy, reconstruction, and protest.
Once therapist and patient have explored the ramifications of the recent suicidal/self-destructive act as aggressive, Plakun (1994) recommends that both parties search for the perceived narcissistic injury that precipitated the self-destructive behavior. There are several interpenetrating reasons for this intervention, including the opportunity for the therapist to acknowledge accurately perceived failures (Cooperman, 1989). The crucial element, from our vantage point, is the exploration of the transference and the likely disillusionment and injury that occurred in the preceding sessions. By placing a premium on exploring the unconscious and conscious meanings of the self-destructive act, the therapist is holding to the therapeutic task and, in so doing, is reaffirming the requirements of the therapeutic frame to seek meaning. Following these principles will in no way eliminate the risk of provocative acting out and self-destruction. Rather, we hope these recommendations and the measured use of consultation and hospitalization will aid in the effective treatment of these patients.