La acosadora

The female stalker

J. Reid Meloy Ph.D., Kris Mohandie Ph.D. and Mila Green Ph.D.

Article first published online: 23 FEB 2011

DOI: 10.1002/bsl.976


Se llevó a cabo un estudio de 143 acosadoras, la participación de una amplia muestra de acosadores de América del Norte (N = 1005) reunidos por la policía, la fiscalía, y archivos de seguridad de empresas del entretenimiento (Mohandie, Meloy, Verde McGowan, y Williams, 2006). La típica acosadora era una mujer soltera, separada o divorciada en su mediados de sus 30 años con un diagnóstico psiquiátrico, la mayoría con un trastorno del estado de ánimo. Ella más probablemente perseguirá a un hombre conocido, extraño, o una celebridad, en lugar de una previa pareja íntima sexual. Cuando se compara con los acosadores masculinos, las acosadoras tenían significativamente menos frecuentemente antecedentes penales, y eran significativamente menos amenazantes y violentas. Su comportamiento de seguimiento era de una base de menor cercanía, y sus comunicaciones eran más benignas que los de los varones. La duración media de acecho fue de 17 meses, la duración modal era de dos meses. La reincidencia del acecho fue del 50%, con tiempo modal entre la acción y volver a ponerse en contacto con la víctima de un día. Cualquier relación real anterior (conocido sexual o íntimo) aumentaron significativamente la frecuencia de las amenazas y la violencia, aumentando de gran forma el efecto en la muestra femenina completa. El subgrupo más peligroso fue el de las acosadoras de anteriores parejas íntimas sexualmente, de los cuales la mayoría amenazaban y eran físicamente violentas. Las menos peligrosas eran las acosadoras femeninas de las celebridades de Hollywood. Dos de las variables de predicción de violencia en el acoso de hombres de McEwan, Mullen, MacKenzie y Ogloff (2009b)  fueron certificados con efectos moderados en el efecto para las mujeres externamente validadas: amenazas fueron asociadas con un riesgo mayor de violencia, y la escritura de cartas estaba asociado con una reducción del riesgo de violencia.


A study of 143 female stalkers was conducted, part of a large North American sample of stalkers (N = 1005) gathered from law enforcement, prosecutorial, and entertainment corporate security files (Mohandie, Meloy, Green McGowan, & Williams, 2006). The typical female stalker was a single, separated, or divorced woman in her mid-30s with a psychiatric diagnosis, most often a mood disorder. She was more likely to pursue a male acquaintance, stranger, or celebrity, rather than a prior sexual intimate. When compared with male stalkers, the female stalkers had significantly less frequent criminal histories, and were significantly less threatening and violent. Their pursuit behavior was less proximity based, and their communications were more benign than those of the males. The average duration of stalking was 17 months, but the modal duration was two months. Stalking recidivism was 50%, with modal time between intervention and re-contacting the victim of one day. Any prior actual relationship (sexual intimate or acquaintance) significantly increased the frequency of threats and violence with large effect sizes for the entire female sample. The most dangerous subgroup was the prior sexually intimate stalkers, of whom the majority both threatened and were physically violent. The least dangerous were the female stalkers of Hollywood celebrities. Two of the McEwan, Mullen, MacKenzie, and Ogloff (2009b) predictor variables for stalking violence among men were externally validated with moderate effect sizes for the women: threats were associated with increased risk of violence, and letter writing was associated with decreased risk of violence.


Although most stalkers are men and their victims are women, females who stalk others remain a largely opaque subject among criminologists, sociologists, psychologists, and psychiatrists. The National Crime Victimization Survey (Baum, Catalano, Rand, & Rose, 2009) reported that 24% of female stalking victims were stalked by another female, and 43% of male stalking victims stated that the offender was female. Around 65,270 persons participated in this survey, and these numbers represent the proportion of U.S. female stalkers from annual prevalence estimates for stalking during the year 2005. Most notably, these female gender rates are higher than in previous community-based studies of stalking victimization conducted during the previous decade, which found that 12–13% of stalkers were female (Purcell, Pathé, & Mullen, 2002; Tjaden & Thoennes, 1998), and are also higher than in studies conducted in forensic mental health settings (Meloy, 1998; Mullen, Pathé, & Purcell, 2000) a decade ago, wherein male stalkers outnumbered females at a ratio of four to one. In a large meta-analysis of stalking research (Spitzberg & Cupach, 2006), gender estimates of perpetrators varied, but the patterns remain similar: the majority of stalkers are men, but women comprise a large minority of those who engage in unwanted pursuit which frightens another. Such data pose the provocative question of a possible increase in the proportion of female stalkers over the last decade.

Why study female stalkers? Purcell, Pathé, and Mullen (2001) noted that male victims of female stalkers are often met with indifference, skepticism, or even derision by law enforcement. On the other hand, female victims of female stalkers often have their sexual orientation questioned, and often a homosexual relationship is assumed, contrary to research findings. There is also a striking disparity between the disposition of male and female stalking cases, with the former more likely to be prosecuted than the latter (Hall, 1998). Recent stalking research has also suggested that there are comparable rates of violence between male and female stalkers (Meloy & Boyd, 2003; Purcell et al., 2001); female stalkers are more persistent than male stalkers given certain motivations (McEwan, Mullen, & MacKenzie, 2009a); and in one study, which measured psychopathy among stalkers, males and females had comparable degrees of this character pathology (Reavis, Allen, & Meloy, 2008). In the larger context of domestic violence there is a growing body of research that males and females are equally assaultive, although males cause more physical injury to their partners (Dutton, 2006). Most obviously, gender differences among stalkers have been woefully neglected in the research. To our knowledge there are only two empirical studies that have focused upon female stalkers in the scientific literature (Meloy & Boyd, 2003; Purcell et al., 2001). This is the third such study. It is conducted to advance our knowledge in a specialized area of forensic psychological research, to test the validity of the previous two female stalking studies’ findings, to compare female and male stalkers from the same large archival sample (Mohandie et al., 2006), and to attempt to discriminate between violent and nonviolent female stalkers utilizing some of the stalking violence predictors for men identified by McEwan et al. (2009b). Findings about female stalker pursuit patterns, threats, violence risk, and recidivism could be very helpful to threat management practice and other forensic psychological consultations. Female stalkers, while not as common as male stalkers in forensic practice, are regularly encountered in threat assessment and other criminal contexts. Scientific data specific to this group are essential to inform competent practice in these arenas.


Over a 17 month period between March 2003 and June 2004, two trained researchers (one with a Ph.D. in Forensic Psychology, and the other with a Canadian diploma in Research Analysis) reviewed over 2,300 files dealing with stalking, criminal harassment, menacing, terrorist threats, or domestic violence behaviors. The cases were amassed from six invited nonrandom samples of convenience across North America, including three prosecutorial agencies in California, one police agency in Canada, a corporate security department for a large entertainment firm in Los Angeles, and the first two authors’ files. Potential sample sources were identified by the second author and asked if they would be willing to participate in the study. Approval for access to these files was provided by the appropriate organizational entity within each of the agencies. To protect confidentiality, researchers signed confidentiality agreements when requested, files were reviewed at the sample locations, no identifying information regarding the case was included on the coding sheets, and coding sheets were labeled with a case number. All data were archival, and therefore subject permission for inclusion in the study was not required. From this pool, 1,005 cases were placed in the study sample. In order to be included, each file had to represent a stalking case that had been, or currently was being managed for threat assessment purposes, investigated by law enforcement personnel, and/or prosecuted. Some cases had been reopened due to additional harassment behavior while others were closed due to behavior cessation (Mohandie et al., 2006).

Stalking behavior was defined as two or more unwanted contacts by a subject toward a target that created a reasonable fear in that target. The rejected cases did not meet this definition. One-third of the 1,005 cases were from prosecutorial agencies, another one-third were from law enforcement, and the final one-third were from the entertainment corporation. Stalking inclusion criteria were met by 93% of the prosecutorial agencies’ cases, 56% of the law enforcement cases, and 23% of the corporate cases. Rejected cases typically involved only one contact, acts of domestic violence without stalking, or had no identifiable target or subject.

A six page, 50-variable coding sheet was utilized to structure data gathering, and covered the following general areas: subject and target demographics; general and specific descriptors of stalking behaviors; target, security, law enforcement, and/or judicial responses to the behaviors; and recidivism. A variable was coded as unknown if the data were unavailable. Each case was categorized according to one of four RECON types: Intimate, Acquaintance, Public Figure, and Private Stranger. The development and rationale for the RECON typology are available in the earlier study of this entire sample (Mohandie et al., 2006). This study demonstrated that the RECON typology has a high degree of assignment reliability (ICC = 0.95), and is a very useful operational system that distinguishes threat and violence risk, pursuit patterns, and likely mental health issues within a generalized and non-mental-health-referred large sample of male and female stalkers; many of its findings validated previous stalking studies done by different research groups in different locations at different times (Mullen et al., 2000, 2009; Meloy, 1998).

Specific data were gathered on subjects’ criminal history, past arrests for other obsessional harassment behavior, additional targets of harassment, and the subjects’ legal status when the stalking began. In 76% of the total cases, information was also available on mental health status, including presence of key DSM-IV-TR Axis I or Axis II diagnoses. These were generally categorized by the researchers as thought, mood, substance abuse, personality, or other disorders, or there were suspected diagnoses as evidenced by case documentation of psychosis and/or substance use during the stalking, a history of psychotropic medication use, and/or suicidality (any mention of suicide ideation or attempts).

Specific data collection also included recording the exact relationship between subject and target (12 possible categories), as well as the context in which the behavior took place (public versus private), domestic violence history between the parties, and the target’s prior stalking victimization, if any. We noted the duration of pursuit, forms of pursuit (telephone calls, letters, surveillance, etc.), most frequent pattern of contact, and type of contact including indirect (writing), direct (telephone calling), or proximity based (physically approaching). We coded for escalation in frequency and/or type of contact, as well as presence of any known precipitating event, typically a major loss in work or love. The quality of the communications (love, help seeking, demeaning, etc.) and any interference with the target’s life were noted. Data were collected on threats and violence, how often, how severe, in what form, whether weapons were used, and who was targeted (victim or third party). Threats were defined as “a written or oral communication that implicitly or explicitly states a wish or intent to damage, injure, or kill the target” (Meloy, 1999, p. 90). Violence, defined as acts of intentional physical aggression toward a person or an object, was coded as either affective or predatory (Meloy, 2006) and included incidents of stealing, property damage, assault, abduction, sexual assault, and murder. Case outcome information was recorded, including whether the subject was charged, offense type, and resultant criminal sanctions. Recidivism was only coded when a subject re-contacted the target subsequent to legal or criminal justice intervention, and would include persistence as defined by McEwan and colleagues (2009a). Recidivism data were acquired from the same files as the other data, and involved varying follow-up periods ranging from immediately up to ten years—given that the sample involved both open and closed cases and the nature of stalking is often episodic.

Because two researchers collected data, alpha was utilized to determine interrater reliability for all variables combined on 10% (n = 101) of the sample. An intraclass correlation coefficient (ICC) was used to determine the reliability of assignment to one of the four RECON types, and whether or not the subject was mentally ill. Other inferential statistical comparisons were made utilizing chi-square, independent t-test, or one-way ANOVA analyses as indicated. For each group comparison found to be significant, effect sizes were also computed (i.e., Phi, Cohen’s d, or partial η2, respectively) (Trusty, Thompson, & Petrocelli, 2004; see Table 1). In each significant ANOVA, the appropriate post hoc analysis (depending on the equality of group variance) was undertaken to isolate which group difference(s) contributed to a significant finding. Overall significance was set at p < 0.01.

Table 1. Calculating and interpreting effect size
Analysis Effect size statistic Small Moderate Large
t-test Cohen’s d 0.2–0.3 0.5 0.8
ANOVA Partial η2 0.01 0.06 0.14
Chi square Phi 0.2–0.3 0.4–0.69 0.7–0.89

Comparisons were made between the female and male stalkers from the overall sample (N = 1005), and also between and among the RECON stalking types for both males and females on threats and violence. We also attempted to validate the predictive factors for violence among prior intimate stalkers identified by McEwan and colleagues (2009) in their logistic regression analysis, although our sample size prevented us from running a comparative regression analysis. We instead compared our sample of violent and nonviolent female stalkers on seven variables that McEwan and colleagues (2009b) identified as predictive of violence risk among mostly male stalking groups utilizing the Mullen et al. typology (2000, 2009).



Coefficient alpha for interrater reliability on all collected variables combined was 0.92. The ICC for assignment to one of four RECON types was 0.95, and the ICC for determining presence of mental illness (assignment to one of five general diagnostic categories, a “suspected diagnosis” category, or a “no mental health issues apparent” category) was 0.85.


There were 143 females in the total study sample of 1005 subjects (14.2%). Average age was 35 years, with a range of 12–76 years (SD = 12.01). According to the RECON typology, 6% of the prior intimate stalkers were women (n = 31); 20% of the prior acquaintance stalkers were women (n = 26); 26% of the public figure stalkers were women (n = 71); and 15% of the private stranger stalkers were women (n = 15).

Sixty percent of the women were Caucasian, 11% were African-American, and 6% were Hispanic (23% unknown). Eight percent did not complete high school, 15% graduated from high school, and 15% had earned associate or bachelor degrees (62% unknown). Forty-nine percent of the women were single, 16% separated or divorced, and 11% were married (24% unknown). Sixty-four percent were heterosexual, 9% homosexual, and 3% bisexual (24% unknown sexual orientation). Female stalkers were significantly less likely to be heterosexual than the male stalkers (64% versus 86%; F = 25.542, p < 0.001, partial η2 = 0.056). Comparisons between homosexual and bisexual male and female stalkers did not yield any significant differences.

The female stalkers’ criminal histories were serious, but significantly less so than the male stalkers (F = 15.573, p < 0.001, partial η2 =0.055). Among the females, 11% had violent, 8% had nonviolent, and 39% had no recorded criminal histories. Fifty-eight percent of the male stalkers had a known criminal history. Four percent of the females (n = 6) had a prior arrest for harassing behavior, whereas 20% of the males did so (x2 = 9.447, p < 0.01, Φ = 0.108). The men were also more likely to be in some form of custody (prison, jail, probation, parole) at the time of the stalking when compared with the women (74% versus 44%; F = 5.620, p < 0.001, partial η2 = 0.034).

The female stalkers were also noticeably psychiatrically impaired. Presence or absence of psychiatric history was known in 83% of the cases. There was a suspected diagnosis in 33%, a confirmed diagnosis in 18%, and no diagnosis in 36% of the women when data were available. The most likely general diagnosis for the women was a mood disorder (18%) followed by a thought disorder (15%). Suicidality—any evidence of a desire or intent to end one’s life—was present in 61% of the female cases where data were known, but it was unknown in 78% of cases. During the stalking there was evidence of psychosis in 16% of the female cases, and 13% were on psychotropic medications. One in three women was also abusing drugs at the time of the stalking, but this was the only significant psychiatric difference when compared with the male stalkers, who were abusing drugs in 67% of the cases (x2 = 20.810, p < 0.001, Φ = 0.202). A diagnostic evaluation was completed for 17% of the women (n = 24), but the specifics of these evaluations were unavailable to us.

When frequency of suicide issues was examined among those female cases where there was sufficient information to determine the presence or absence of suicidality, 88% of the Intimate stalkers, 80% of the Acquaintance stalkers, 75% of the Public Figure stalkers, and 20% of the Private Stranger stalkers evidenced suicidal thinking. By comparison, among the male stalkers, 81% of the Intimate stalkers, 64% of the Acquaintance stalkers, none of the Public Figure stalkers (100% missing data), and 53% of the Private Stranger stalkers evidenced suicidality.


Among the female stalkers, 38% of the targets were female, 60% were male (2% unknown). Male stalkers significantly more often targeted the opposite sex when compared with female stalkers (91% versus 60%; x2 = 233.817, p < 0.001, Φ = 0.489). Also, male stalkers were more likely to target present and former romantic/sexual partners (43% versus 19%), while female stalkers targeted celebrities more often than men (49% versus 23%; F = 6.996, p < 0.001, partial η2 = 0.072). There was a domestic violence history between the stalker and the target in 10% of the female stalker cases and 37% of the male stalker cases (x2 = 39.988, p < 0.001, Φ = 0.203). Information regarding the reciprocity of violence between the stalker and the victim was unavailable. The context of the stalking was more public for the women than the men (60% versus 33%; x2 = 37.504, p < 0.001, Φ = 0.193); that is, female stalkers more often than male stalkers pursued public figures or persons they developed an interest in due to their public visibility. Six percent of female stalkers and their victims had a prior or ongoing professional relationship (six had a work or business relationship and in two additional cases the target was a health care/mental health provider who had seen the subject as a patient).

There was a precipitating event in 41% of the female stalking cases, such as a major loss or receiving a signed fan photo, but significantly less so than in male stalking cases (61%; x2 = 7.286, p < 0.01, Φ = 0.086).

In 20% of the cases, the female stalker pursued secondary targets (“target dispersion,” Scalora, Baumgartner, Zimmerman, Callaway, Maillette, Covell et al.,2002) in addition to the primary target. These 29 secondary targets included celebrities or strangers (n = 13), friends, co-workers, or acquaintances of the target or subject (n = 7), family members or former intimates of the target or subject (n = 6), several unknown individuals (n = 2), and one professional assigned to the case.

Pursuit Characteristics

The stalking duration in female cases lasted from one day to ten years, with a mean of 17 months and a mode of two months. Contact frequency, however, was significantly different for the women than the men. There was no gender difference for those stalkers who contacted daily or every two to three days (29% versus 20%), or weekly (36% versus 32%), but women were more likely to contact sporadically every month (27% versus 19%) or every 2–3 months (17% versus 7%; F = 4.487, p < 0.001, partial η2 = 0.027).

The primary pattern of pursuit also varied significantly according to gender. The women did not use third parties at all. They were more likely to send letters or faxes (31% versus 15%) and gifts or packages (8% versus 3%), but less likely to make personal contact with the target (45% versus 52%) or commit burglaries (8% versus 16%; F = 5.124, p < 0.001, partial η2 = 0.035). Overall contact was more proximity based for the men than the women (73% versus 56%; F = 10.444, p < 0.001, partialη2 = 0.020).

The women were also significantly more benign in the content of their communications to the target. They were more likely “just communicating” (14% versus 7%) or help seeking (18% versus 9%) than the men, and also less insulting (15% versus 39%; F = 6.190, p < 0.001, partial η2 = 0.030). There was escalation in 34% of the female cases, meaning an increase in frequency of contact and/or method of contact.

Threats and Violence

In the female stalking cases where threats were made, the number of threats ranged from 1 to 95, with a mode of 1 and a mean of 8. Thirty-seven percent of the women threatened, but significantly less so than men (63%, F = 10.755, p < 0.001, partial η2 = 0.041). Women made fewer implied threats (5% versus 12%) or direct threats (27% versus 46%). Men were more likely to threaten the target than women, who more often threatened the family or friends of the target or themselves (55% versus 32%; F = 6.394, p < 0.001, partial η2 = 0.037).

Sixty percent of the female stalkers did not threaten and were not violent (true negative rate). Twelve percent threatened but were not violent (false positive rate). Three percent did not threaten but were violent (false negative rate). Twenty-five percent both threatened and were violent toward person and/or property (true positive rate).

Violence was coded as either affective (reactive, emotional, impulsive) or predatory (instrumental, premeditated, planned, purposeful) (Meloy, 1988, 2000, 2006). Twenty-eight percent of the female stalkers were violent toward person, property, or both, significantly less so than the male stalkers (50%; F = 7.274, p < 0.001, partialη2 = 0.028). Violent male stalkers were most violent toward the target of their pursuit (33%), whereas the violent female stalkers equally aggressed toward their target (12%) or property of the target (13%). When violence occurred in the female stalking cases, it was more likely to be predatory (18%) than affective (10%). There were not enough data in the female case files to accurately classify the rest of the cases (72%) as either affective or predatory.

Thirteen percent of the violent female stalkers used a weapon, most commonly objects for which there was insufficient detail to describe. The female stalkers did not harm any pets, commit sexual assault, or murder anyone. Male stalkers were more likely to assault when compared with the female stalkers (26% versus 15%; F = 4.634, p < 0.001, partial η2 = 0.027).

Table 2 is a comparison of threats and violence when the female stalkers are divided into those who had a previous relationship with the victim (Type I), either intimate or acquaintance; and those who did not have a previous relationship with the victim (Type II), either a public figure (Hollywood celebrity) or a private stranger. When the women are divided in this manner, a prior relationship is associated with a significantly greater frequency of threats and violence. The effect sizes are large (partial η2 = 0.23–0.33).

Table 2. Comparison of threats and violence among female stalkers who did or did not have a prior relationship with the victim
Relationship (n = 57) No relationship (n = 86) Effect size
  1. All findings significant at p < 0.001.
 None 33% 83%
 Direct 46% 14%
 Indirect 11% 0
 Implied 11% 0 partial η2 = 0.254
Threats to whom
 Target 56% 16%
 Family or friends 7% 1%
 Self 4% 0 partial η2 = 0.255
 None 74% 89%
 Target 21% 6%
 Others 5% 0
 Property 0% 4% partial η2 = 0.235
Threats and violence
 None 26% 83%
 Threats only 19% 7%
 Violence only 7% 0
 Both 47% 11% partial η2 = 0.329
Acts of violence
 Assault 29% 6%
 Property damage 26% 4% partial η2 = 0.235

Table 3 is a comparison of threats and violence by gender and RECON type. There was no significance difference between men and women in threats and violence when compared within each type. However, when female stalkers are compared with each other across RECON types, there is a significant difference for both threats (F = 22.896, p < 0.001, partial η2 =0.331) and violence (F = 28.044, p < 0.001, partial η2 = 0.377). When male stalkers are compared with each other across RECON types, there is also a significant difference for both threats (F = 117.200, p < 0.001, partial η2 = 0.291) and violence (F = 173.789, p < 0.001, partial η2 = 0.378). These are large effect sizes (see Table 1). Table 3 allows for a more precise look at threats and violence among the four RECON types for the women rather than grouping subjects by relationship/no relationship as in Table 2.

Table 3. Violence and threats among males and females according to RECON types
Female (n = 143) Male (n = 862)
  1. Chi-square analyses across genders were not significant for either threats or violence in any RECON group: IA = Prior Sexual Intimate, IB = Acquaintance, IIA = Public Figure, IIB = Private Stranger.
Type IA N 31 471
Threats 23 (74%) 392 (83%)
Violence 21 (68%) 350 (74%)
Type IB N 26 103
Threats 15 (58%) 70 (68%)
Violence 10 (39%) 53 (51%)
Type IIA N 71 200
Threats 7 (10%) 40 (20%)
Violence 2 (3%) 3 (1%)
Type IIB N 15 88
Threats 8 (53%) 43 (49%)
Violence 7 (47%) 25 (28%)

Criminal Justice Involvement and Recidivism

When all the female stalkers are viewed as a group, any kind of criminal justice involvement was significantly less likely than for the male stalkers (47% versus 77%;x2 = 56.888, p < 0.001, Φ = 0.238). For the women, charges were dismissed (3%), a restraining order was issued (3%), probation was instituted (13%), mental health treatment was ordered (1%), hospitalization occurred (1%), or detention resulted (6%). While more male cases were dismissed (7% versus 3%), more men were also incarcerated (21% versus 6%; F = 9.668, p < 0.001, partial η2 = 0.064).

There were 133 cases in which recidivism was possible across the entire female sample. In our inquiry, for recidivism to have occurred, the subject must have been told in some way to stop her harassing behavior (i.e., a cease and desist letter, a restraining order, police notification, or formal criminal sanctions). Then, despite the warning, the stalker re-contacted the same victim again. In 57% of these applicable cases (n = 76), recidivism occurred. The time range between notification and actual re-contact was anywhere from two days to three years, but the modal time was one day. The mean time was 2.3 months. This pattern of reoffense was not significantly different than that demonstrated by our male stalkers. Our definition of recidivism for these statistics is the same as the definition of persistence by McEwan and colleagues (2009a).

Group Comparisons

Prior Sexually Intimate Female Stalkers

Due to the widely researched and acknowledged danger of prior sexually intimate male stalkers when compared with other groups (Mohandie et al., 2006; Mullen et al., 2009), we compared this subgroup (n = 31) of female stalkers with the other three groups (n = 112). These women had a significantly greater likelihood of a violent criminal history (29% versus 6%; F = 6.242, p < 0.005, partial η2 = 0.135). They were more likely to physically pursue the victim and burglarize his or her home (65% versus 39% for pursuit, 32% versus 2% for burglary; F = 11.375, p < 0.001, partial η2 = 0.334). They were more likely to make direct threats (55% versus 18%; F = 9.601, p < 0.001, partial η2 = 0.172). They were also more likely to both threaten and be violent (58% versus 16%; F = 16.204, p < 0.001, partial η2 = 0.259), and they were significantly more likely to assault (39% versus 9%) and damage property (32% versus 7%; F = 10.879, p < 0.001, partial η2 = 0.240). All of these stalking behaviors were significantly more often preceded by a precipitating event (84%) when compared with the other three groups (39%; x2 = 19.862, p < 0.001, Φ = 0.377).

Public Figure (Hollywood Celebrity) Female Stalkers

On the other hand, research indicates that some of the most benign stalkers are those who pursue public figures who are celebrities, often seeking a relationship and evidencing mental illness (Meloy, Mohandie, & Green, 2008). The sample of public figure female stalkers (n = 71) was compared with the other three groups (n = 72). These women were significantly less likely to have either a violent or a nonviolent criminal history (5% versus 34%; F = 8.439, p < 0.001, partial η2 = 0.174). They were less likely to pursue in person (25% versus 64%), relying significantly more often solely on letters, cards, and faxes (54% versus 8%; F = 13.265, p < 0.001, partial η2 = 0.369). They sought love (42%) or help (23%), or were “just communicating” (24%) significantly more frequently than the other groups (89% versus 50% total for three categories; F = 10.417, p < 0.001, partial η2 = 0.275). Any type of threat was unusual (10% versus 64%; F = 21.742, p < 0.001, partial η2 = 0.319). If they did threaten, 20% of the time it was directed at the family or friends of the celebrity figure, not the identified target. Two percent of these women assaulted compared with 28% of the other three groups combined (F = 15.905, p < 0.001, partial η2 = 0.316). None damaged property.

These women also tended to escalate significantly less often (23% versus 44%; x2 = 7.392, p < 0.01, Φ = 0.228), and only 20% had a precipitating event that could be identified before they began stalking compared with 77% in the other three groups (x2 = 45.752, p < 0.001, Φ = 0.572). The criminal justice system was involved in only 17% of these cases, compared with 76% of the other groups (x2 = 50.803, p < 0.001,Φ = 0.596).

Violence Risk Prediction

Table 4 is an investigation of the usefulness of the violence risk predictors found to be significant by McEwan and colleagues (2009b) in our sample of female stalkers. Using only the female stalkers in the Prior Intimate, Acquaintance, and Private Stranger groups (n = 72)—since their study included virtually no public figure stalkers—chi-square analyses were run on each predictor to assess for a significant relationship to violence during the stalking behaviors. Data were available on seven variables that McEwan and colleagues (2009b) found to predict violence in various regression analyses that they conducted; however, our sample sizes were too small to perform our own regression analysis. Significance between the violent and nonviolent females was found for two variables at p < 0.01: those who threatened the victim were more likely to be violent, and those who wrote to the victim were less likely to be violent. Both findings had moderate effect sizes (Φ = 0.437–0.535).

Table 4. Significance testing of violence risk predictors per McEwan et al. (2009b) in a female stalking sample
Violence No violence Effect size
  • +p < 0.001.
Rejected motive (n = 72)
 Prior sexual intimates 71% 29%
 No prior sexual intimates 41% 59%
Age (n = 63)
 Under age 30 60% 40%
 30 yrs and older 56% 44%
Psychosis (n = 59)
 Yes 33% 67%
 No 52% 48%
Writing to the victim+ (n = 72)
 Yes 0% 100%
 No 63% 37% Φ = 0.437
Prior violence (n = 72)
 Yes 60% 40%
 No 53% 37%
Substance abuse (n = 34)
 Yes 55% 45%
 No 74% 26%
Threats+ (n = 72)
 Yes 76% 24%
 No 15% 85% Φ = 0.585


This study validates the findings of the two earlier studies (Meloy & Boyd, 2003; Purcell et al., 2001) that female stalkers in general are heterosexual women with an average age in their mid-30s who are single, separated, or divorced. Psychiatric impairment is present in a majority of known cases, with the most likely suggested diagnosis being a mood disorder. No reliable data on personality disorder were available for this sample, but a previous study strongly suggested that borderline personality disorder was prevalent among female stalkers (Meloy & Boyd, 2003). One in six women were psychotic at the time of the stalking and on psychotropic medications. Both nonviolent and violent criminal histories were present, but significantly less so than male stalkers. Drug abuse is also present in one out of three women, significantly less than men, but with a small effect size. As in virtually all stalking studies that have gathered clinical and forensic data, female stalkers are multi-problem individuals who display criminal, psychiatric, and drug abuse difficulties and also engage in the crime of stalking. Recent research has not supported the theory that stalkers are more intelligent than other criminals (MacKenzie, James, McEwan, Mullen, & Ogloff, 2010), and they typically have a standardized IQ in the low average range with a significant verbal compared with performance deficit.

The female gender prevalence of 6–26%, depending on the RECON type, is similar to that in previous studies (Spitzberg & Cupach, 2006), with the largest proportion of women among the celebrity public figure stalkers. Although there is no other data set with which to compare this finding, target selection of a celebrity figure may be a product of the female stalker’s general desire to establish a relationship rather than repair one, and engage in more distal and indirect stalking behaviors than men (Meloy & Boyd, 2003; Purcell et al., 2001). This study validates the finding in the previous two studies that female stalkers are less likely to pursue a prior sexual intimate than male stalkers, but evidence for the notion that the nature of the acquaintance relationship prior to the stalking would be a professional one was not supported (Purcell et al., 2001). They are also more likely than men to select targets of the same gender, but the majority do not. Their pursuit characteristics are more likely to be benign communications that are sent through multiple means that do not involve seeking proximity to the object of attention when compared with men. Stalking recidivism frequencies—what McEwan and colleagues (2009a) refer to as persistence—were about 50%, an apparently robust finding when compared with the few studies that have been conducted in this area (Mohandie et al., 2006; Rosenfeld, 2003). The most frequent elapsed time for recidivism was one daybetween intervention and re-contact of the victim.

Once again, there is a striking difference between average length of stalking (17 months) and the most frequent length of stalking (two months). Previous research has rarely made this distinction (Mohandie et al., 2006), but it underscores the effect of outliers on some statistical computations, and confirms the fact that most stalking is intense and short lived, with the early pursuit behavior of the stalker being the best predictor of duration (Purcell, Pathé, & Mullen, 2004).

Although other problems caused by stalkers, such as persistence, psychological trauma, and embarrassment, are concerning (Mullen et al., 2009), the Holy Grail for both mental health and criminal justice professionals remains a complete understanding of violence in this population.

One out of three female stalkers will threaten, and one out of four will be violent toward the target or his property in equal measure. Female stalkers had significantly lower rates of threats and violence than male stalkers when the overall sample was examined. However, female stalkers did threaten and engage in both personal and property violence at frequencies that justify moderate violence risk concern in female stalking cases. Female patterns of violence and threats paralleled male patterns across the RECON typology, with the highest frequency of violence and threats inflicted by female prior intimate stalkers (IA), and the lowest frequency found among the female stalkers of public figures (IIA), perhaps an artifact of their personal security and protective intelligence operations (Table 3).

The root of these differences is likely found in the presence or absence of an actual prior attachment or bond to the object of pursuit (see Table 2). There is now a substantial empirical literature on the pathology of attachment among stalkers (Dutton, Winstead, & Mongeau, 2006; MacKenzie, Mullen, Ogloff, McEwan, & James,2008), supporting the original hypothesis of Meloy (1989, 1992) that stalking was fundamentally driven by a pathology of attachment. Adult attachments vary with different types of stalker. Most notably when comparing relational and nonrelational stalking and violence risk among women, the effect sizes for all the dependent variables were significant and large.

Moreover, when female stalkers were compared with male stalkers within each category of the RECON typology—Intimate, Acquaintance, Public Figure, and Private Stranger—no statistically significant differences were noted for threat or violence frequency (Table 3). This finding suggests that female stalkers, while somewhat less likely overall to be threatening or violent, are not significantly different from males in frequency of threats and actual violence when compared within subtypes. It may partially reflect the lower numbers of female offenders in the sample, as well as potential sampling bias. Nonetheless, it underscores that there is a very real violence and threat danger posed by some female stalkers that must be considered by threat assessment professionals, principally among the female prior sexually intimate stalkers who usually threatened and were violent toward person or property 68% of the time. Palarea, Zona, Lane & Langhenrichsen-Rohling (1999) underscored the danger of prior intimate stalkers, a finding that has been subsequently validated in a number of studies (Mullen et al., 2009).

Sixty-one per cent of the overall female cases evidenced suicidality, but with much missing data. The highest suicidality frequency appeared among the Intimate stalkers of both genders, suggesting that there is a desperation aspect to these cases, which occurs hand in hand with the noted violence risk. Prominent suicidality is also consistent with the common diagnosis of a mood disorder. Future research should address the issue of suicidal thinking and behavior in this population, as the noted trend in our cases where there were sufficient data suggests that it is a very real aspect of these cases, and may well be a harbinger of risk to the stalker and to others.

We also attempted to validate the predictors identified by McEwan and colleagues (2009b) in their sample of 211 stalkers in forensic mental health treatment in the community. Although we did not have enough subjects to conduct logistic regressions, two of their seven predictor variables for which data existed did distinguish between the violent and nonviolent women: writing to the victim was associated with no violence, and threats were associated with violence. The effect sizes were both moderate. Rejected motive was only significant at the 0.05 level. This external validation is important since these were all female stalkers and were not sampled from a mental health treatment setting.

The female public figure stalkers, in this case targeting Hollywood celebrities, are of a different ilk. Relatively benign from a violence risk standpoint, they stalk indirectly, appear to prefer no proximal contact, and are not motivated by a significant loss in work or love. Most were seeking love or help. Entitled reciprocity—the belief that one’s devotion in time and effort toward the public figure creates an obligation in the public figure to do likewise—and grandiosity—the disparity between the facts of one’s life and the fantasy of one’s importance—have recently emerged as important psychodynamics in public figure stalking (Meloy, Sheridan, & Hoffmann, 2008). Although these variables were not measured among the female stalkers, much of their behavior appears to parallel the behavior of the male stalkers of celebrity public figures.

Gender plays a role in shaping the behavior of the female stalker, although the relationship between the stalker and the victim appears paramount. Female gender mitigates aggression, but attachment aggravates it. This study underscores the importance of determining the type of stalker when conducting research or planning interventions, drawing from the two most useful stalking typologies for operational (Mohandie et al., 2006) and clinical purposes (Mullen et al., 2009). As McEwan and colleagues (2009b) noted, “There were only marginal differences in accuracy [between the two typologies] when predicting violence. Given the apparent empirical equivalence of the two typologies, the choice of classification scheme should vary according to the needs of the user” (p. 8).

Clearly the most dangerous subgroup among both male and female stalkers is those who have had a prior sexually intimate relationship with the victim.


There are several sources of methodological weakness in this study. Selection bias is evident in the nonrandom samples of convenience utilized, and limits the generalizability of the findings to those female stalking subjects who have had contact with law enforcement or private security firms. Observation bias may be present in the differential recall and data gathering strategies for each of the six samples when the cases were first identified. For example, the authors’ files likely contained much more psychiatric and psychological information than the law enforcement, prosecutorial, and security files. Nonresponse bias is evident in the relatively large proportion of unknown data concerning mental health status, suicidality, and drug abuse in this sample. Methodological weakness concerning chance—the likelihood that significant differences among groups is the result of chance—has been minimized, however, due to the sample size, chosen significance value, and use of effect sizes. Confounding variables may exist in this study that influence the significant findings, but are unknown to the researchers since they were not measured.


This study lends further support to the notion that both male and female stalkers stand equidistant between psychiatric patients and criminal offenders. In both groups psychological and psychiatric assessment and treatment should be a highly desired, if not a mandatory, component of any risk management strategy. Stalking is a complex behavior that unfolds over time, and the risk of physical violence toward the target is considerable, especially if the stalker is a prior sexual intimate. Threat assessment for both stalking and violence in any given case needs to appreciate the dynamic and continuous behavior of the stalker, while mental health treatment considerations need to be as nuanced and comprehensive as the particular diagnosis of the stalker dictates (MacKenzie et al., 2009).

Although the stalking research is considerable, and now spans a period of more than 20 years, the very few studies of female stalking parallel the research concerning female criminality in general—that is, efforts are earnestly made but rarely undertaken by anyone. Given many of the gender differences among male and female stalkers, future researchers in any aspect of stalking should attempt to compare their samples by gender on both lateral and longitudinal variables, with an eye toward improving both risk management and treatment.