Recidivism in Stalking and Obsessional Harassment

Transcription

1 Law and Human Behavior, Vol. 27, No. 3, June 2003 ( C 2003) Recidivism in Stalking and Obsessional Harassment Barry Rosenfeld 1 Despite the rapidly growth of mental health attention focused on the phenomenon of stalking, no empirical research to date has attempted to assess the frequency of repeat offending or attempted to identify predictors of recidivism. A total of 148 stalking and harassment offenders who were court-ordered to undergo a mental health evaluation were followed for a period of years in order to assess the frequency of repeat offenses and the variables that differentiated high versus low risk offenders. Recidivism data were obtained from a variety of sources, including criminal justice records, mental health records, and reports from probation officers and victims. A number of potential predictor variables were selected on the basis of the existing recidivism literature in other criminal justice populations. Frequency analysis were used to identify variables that significantly differentiated offenders who did and did not reoffend while survival analysis was used to analyze the impact of these covariates on time to reoffense. A total of 49% of the offenders reoffended during the follow-up period, 80% of whom reoffended during the first year. The strongest predictors of recidivism included the presence of a personality disorder, and in particular, a Cluster B personality disorder (i.e., antisocial, borderline, and/or narcissistic). In addition, those offenders with both a personality disorder and a history of substance abuse were significantly more likely to reoffend compared to either of these risk factors alone. Surprisingly, the presence of a delusional disorder (e.g., erotomania) was associated with a lower risk of reoffender. The findings are discussed in terms of the legal system and treatment implications. KEY WORDS: stalking; recidivism; crime; mental disorder. INTRODUCTION Few types of criminal behavior evoke as much pessimism on the part of victims, mental health providers, and the criminal justice system as stalking and obsessional harassment cases. Estimates of the frequency of stalking cases exceed 1,000,000 per year (Tjaden & Theones, 1998) and many of these victims assume that nothing short of death (their own or the stalker s) will curb the stalker s pursuit of them. These concerns prompt many victims to take extreme measures, such as moving to a new 1 Department of Psychology, Dealy Hall, Fordham University, Bronx, New York 10458; rosenfeld@fordham.edu /03/ /1 C 2003 American Psychology-Law Society/Division 41 of the American Psychology Association

2 252 Rosenfeld state or community or even changing their identity. This fatalistic perspective is also reflected in the criminal justice system s treatment of stalking and harassment crimes, as many of the stalking laws passed over the past decade have included increasingly lengthy sentences for continued harassment after an initial conviction. Yet despite the frequent attention paid to the topic of recidivism in stalking, not a single study has focused on the frequency of repeat offending or attempted to identify predictors of recidivism. The rapidly growing literature devoted to stalking and harassment has primarily focused on either descriptive studies, detailing the demographic and clinical characteristics of this population (e.g., Harmon, Rosner, & Owens, 1995; Meloy & Gothard, 1995; Schwartz-Watts, Morgan, & Barnes, 1997) or comparisons of particular subgroups of stalkers (e.g., Harmon, Rosner, & Owens, 1998; Kienlen, Birmingham, Solberg, O Regan, & Meloy, 1997; Mullen, Pathé, Purcell, & Stuart, 1999). More recently, an increasing number of studies have focused on identifying predictors of violence (e.g., Brewster, 2000; Meloy, Davis, & Lovette, 2001; Palerea, Zona, Lane, & Langhinrichsen-Rohling, 1999; Rosenfeld & Harmon, 2002) and documenting victim reactions to stalking (Kamphius & Emmelkamp, 2001; Pathé & Mullen, 1997, Westrup, Fremouw, Thompson, & Lewis, 1999). These two literatures have also generated a number of conclusions, albeit tentative ones (Meloy, 1998; Rosenfeld, in press). Several variables have emerged as seemingly important, either because they denote an increased risk of violence or distinguish stalkers from the typical offender population. Among the variables that have interested stalking researchers are a history of intimate relationship between victim and offender ( prior intimates ), the presence of a delusional disorder (or, in the case of violence risk, a psychotic disorder diagnosis more generally), and the diagnosis of a personality disorder. Although never focused on stalking cases, the problem of criminal recidivism has been extensively researched in a number of other criminal justice populations including sex offenders, spouse/child abusers, juveniles, mentally ill NGRI acquittees, and paroled felons in general (e.g., Bonta, Law, & Hanson, 1998; Cottle, Lee, & Heilbrun, 2001; Furby, Weinrott, & Blackshaw, 1989; Hall, 1995; Rosenfeld, 1992). These literatures, while often confounded by addressing multiple types of recidivism (e.g., general criminal recidivism, repeat offending of a specific crime such as a sex offense, violence recidivism, etc.), have generated many divergent findings but a number of consistent trends. The most common correlates of recidivism across many of these literatures have included age (with younger offenders more likely to reoffend), having a prior criminal record, a violent index offense, diagnosis of antisocial personality disorder or psychosis (the absence of), a history of substance abuse, and previous psychiatric hospitalizations (e.g., Bonta et al., 1998; Cottle et al., 2001; Furby et al., 1989). Unfortunately, much of the literature focused on recidivism has utilized only simplistic research methodologies, comparing those offenders who do or do not reoffend in simple univariate analyses. More sophisticated statistical methods exist that can incorporate the length of time an offender is at risk for reoffending (e.g., survival analysis), and have been increasingly recognized as a more appropriate method for studying criminal recidivism, but are still the exception rather than the norm (e.g., Albonette & Hephurn, 1997; Quinsey, Rice, & Harris, 1995; Winner, Lanza-Kaduce,

3 Recidivism in Stalking 253 Bishop, & Frazier, 1997). Survival analysis offers an important advantage over the more frequent comparisons of offenders who do and do not reoffend by this inclusion of the timing of the reoffense. By examining the rate and timing of reoffending, important differences between groups might emerge that are obscured by analyses that only focus on whether a reoffense has occurred. Another crucial question in studies of recidivism concerns the choice of dependent variable. Many studies have relied solely on rearrest or reconviction in determining whether a repeat offense has occurred, although official records necessarily underestimate the true frequency of reoffending. In populations where rearrest is a relatively low likelihood event (e.g., sex offenders) the degree of underestimation can be dramatic (Prentky, Lee, Knight, & Cerce, 1997). However in the case of stalking and harassment, this source of bias may be far more modest given that virtually all stalking victims know the identity of the offender, particularly at the point of possible recidivism, since charges have already been filed in an initial offense. Although offenders who choose a new victim or who have successfully kept their identity secret may evade detection as a recidivist, the likelihood of a second prosecution following a first stalking offense is likely quite high. This study represents the first attempt to explore the frequency of recidivism among a large sample of stalking and harassment offenders 2 as well as to identify predictors of differential recidivism. In order to account for the impact of opportunity on recidivism, survival analysis (i.e., Cox Proportional Hazards models) were used to assess whether a set of preselected variables predicted stalking recidivism. These results were compared with traditional nonparametric models (e.g., frequency analysis based on whether or not an individual reoffended). Several of the traditional correlates of recidivism were expected to significantly predict reoffending in this population as well as some of the variables that are unique to stalking cases (e.g., victim offender relationship, presence of a delusional disorder). METHOD Procedure Demographic and clinical data were derived from several sources: case files of stalking offenders referred to the New York City Forensic Psychiatry Clinic for mental health evaluations between January 1, 1991, and December 31, 1997 (described in more detail later), court documents from these (and often, earlier) cases (e.g., criminal complaint forms detailing the circumstances of the instant offense), records from the Department of Probation pertaining to the defendants being evaluated (available in many, but not all cases and often including statements from the stalking victim). The primary source of recidivism data was an electronic database compiled by the New York State Division of Criminal Justice Statistics database (DCJS). This database contains information of all criminal defendants and offenses that occur in New York State as of June 8, 2000, corresponding to a follow-up period of between 2 Although not all subjects had been convicted, the term offender is used throughout this paper for the sake of consistency.

4 254 Rosenfeld 2.5 and 8.5 years. However, because some offenders were evaluated as much as 3 4 years after their first stalking-related arrest, follow-up periods for some offenders exceeded 12 years (with a maximum of 153 months). The Forensic Psychiatry clinic, a division of Bellevue Hospital, evaluates virtually all criminal defendants referred for pretrial mental health evaluations (e.g., competence to stand trial evaluations) in Manhattan as well as most defendants referred by the Department of Probation from all five boroughs of New York City. Evaluations in this clinic are conducted by either psychiatrists or clinical psychologists, most of whom have extensive experience evaluating criminal defendants (see Rosenfeld & Ritchie, 1998). Records from all defendants referred to this clinic between January 1, 1991, and December 31, 1997 (approximately 1400 per year), were reviewed and those cases reflecting stalking or obsessional harassment were selected for further examination and matched with data from the other sources. DCJS records were matched using the defendants New York State Identification number (NYSID), a number assigned to all individuals arrested or charged with a crime in New York state. In order to be classified as a stalking case, the available data had to indicate a persistent and repetitive pattern (with a minimum of three incidents reported) of unwanted contact or harassment encompassing multiple attempts to contact the victim. This definition was chosen in order to maximize the comparability of this sample with previous studies of stalking offenders (e.g., Mullen et al., 2000), because New York State did not have a stalking law at the time this study was conducted. Instead, most offenders were charged with Aggravated Harassment or Contempt of Court (for violating an order of protection), although all cases would have met the definition of stalking in the New York statute enacted in Demographic correlates of many of these cases (approximately 75%; all cases between 1991 through 1996) were described in a report by Harmon et al. (1998), although all data were recoded after additional information was obtained (e.g., Clinic records from prior or subsequent evaluations, Probation Department records, DCJS files). Records were reviewed by a research assistant and coded for a number of demographic, clinical, and offense-related variables. Demographic variables collected including age, gender, race, place of birth, years of education, marital status, reported history of physical/sexual abuse, and mental health history (i.e., number of psychiatric hospitalizations, whether past outpatient treatment and treatment psychotropic medications were reported). Clinical data was elicited from the psychiatric reports, clinician s handwritten notes, and other records available at the time of the evaluation. These data included DSM diagnosis recorded by the evaluating clinician (see below for a description of how conflicting or ambiguous diagnoses were resolved; these diagnoses were ultimately collapsed into psychotic/nonpsychotic based on the presence of a diagnosis of schizophrenia, delusional disorder, or mania), estimated intelligence (based on clinician assessments; psychological testing was infrequently available), history of substance abuse (based on self-report and Probation Department records), and the presence or absence of several psychiatric symptoms (anxiety, depression, suicidal ideation, hallucinations, delusional beliefs, paranoid ideation). Offense-related data, derived from defendant s self-report during the clinical interview and official records from the Department of Probation and the DCJS records,

5 Recidivism in Stalking 255 included the types of harassment reported (presence/absence of threats, attempted face-to-face contact, following, phone calls, mail, unwanted gifts), number and gender of victims, motivation of the offender (offenders were classified as revenge-motivated if they expressed anger or retaliation as their primary motive and romantically motivated if they expressed a desire to reunite with the victim or believed they were involved in a relationship; some offenders did not fit these broad categories and were classified as other/mixed), victim/offender relationship (friend/acquaintance, former intimate, business associate/employer, relative, stranger), history of prior criminal behavior and/or previous violence unrelated to the current stalking case (referred to as the instant offense even though many cases were ongoing for several months or years). Offenders were classified as having reoffended if there was any indication of a second arrest or renewed harassment (e.g., based on victim or Probation reports) following the first recorded stalking-related arrest. 3 Although New York state did not have a charge labeled Stalking at the time this study was conducted, most stalking cases were indicated by charges of harassment or criminal contempt (for violating an order of protection). Because of the timing of the clinical evaluations, many offenders had reoffended before they were either sentenced or referred for psychiatric evaluation. The determination of whether a reoffense had occurred were based on data from all of the sources noted earlier, including (and primarily) DCJS records, 4 data from Department of Probation reports and victim statements, and clinic records (which occasionally, although rarely, indicated a new harassment incident even though no formal complaint had been filed). Although there is little doubt that some reoffenses might not have been detected through these multiple sources of information, for the reasons described earlier (and later, in the Discussion section) this source of bias is probably neither substantial nor systematic. In addition, cases in which the offender was found incompetent to stand trial for the instant offense were not included in these analyses because data were not available regarding the length of time at risk (i.e., because these offenders were hospitalized for an indefinite period following the finding of incompetence, it was not possible to determine when, or even if, they had an opportunity to reoffend). Approximately 15% of the files were coded by two raters (the principal investigator and a research assistant), blind to the ratings of one another, in order to assess the reliability of variable coding. Kappa coefficients for ratings were high (>.90) for all variables coded except clinical diagnosis, which could not be calculated (because uncertain diagnoses were resolved through a review of records and consultation with the first author 5 ) indicating that the coding system was reliably applied. 3 Many individuals were referred for more than one evaluation. In these cases, the index evaluation was considered to be the first contact with this clinic during the study period. 4 This database only includes cases in which prosecution was sought and excludes sealed cases (e.g., juvenile offenses). Given the nature of stalking offenses (i.e., in which the victim typically knows the identity of the stalker and has already established a relationship with the District Attorney s Office through the instant offense), and the probationer status of most defendants, it appears unlikely that a substantial number of cases were lost due to reliance on this data source. 5 In cases where diagnostic conclusions were ambiguous or conflicting, such as those when different evaluators recorded different diagnoses at different times, available data were reviewed by the author to determine the most appropriate clinical diagnoses.

6 256 Rosenfeld Statistical Analysis Univariate analyses (Chi-square, t test) were initially used to assess whether any relationship existed between recidivism (a new charge related to stalking/harassment after the first reported offense) and the various demographic, historical, and clinical variables measured. For these analyses, several continuous variables (e.g., age, years of education) were recoded into categorical variables for both interpretive considerations as well as to facilitate use in subsequent survival analysis models. Time to reoffend was analyzed using Cox Proportional Hazards models with recidivism and time at risk considered along with selected covariates discussed later (Allison, 1995). Model significance was tested using the log-rank test (a nonparametric test statistic) and the Log-likelihood test (a less conservative test of association). On the basis of a review of the recidivism literature in general, as well as the stalking literature, the following variables were selected for analysis as possible predictors: gender, offender age, nature of the victim offender relationship, length of harassment prior to the first known contact with the criminal justice system (typically based on victim reports); existence of a previous criminal record before the first stalking arrest, history of violence unrelated to the current stalking case, previous psychiatric hospitalizations, presence of delusional beliefs, and diagnosis. Specific comparisons were made between those offenders with and without the following disorders: personality disorder and cluster B personality disorder/traits in particular, psychotic disorder and delusional disorder in particular, and the interactions between personality disorder, psychosis and substance abuse. RESULTS Sample Characteristics During the 7-year period in which cases were reviewed, 202 individuals were evaluated for crimes reflecting stalking or harassment and complete clinical and demographic data were available for 189 of these cases. 6 The sample included 147 men (77.8%) and 42 women (22.2%), with an average age of 38.7 (SD = 10.4) at the time of the instant offense and an average of 13.2 years of education (SD = 3.0). Racial background of the sample was 49.7% Caucasian (n = 92), 27.6% black (n = 51), and 17.3% Hispanic (n = 32); 5.4% (n = 10) were of other (primarily Asian) descent (data were missing for four cases). The majority of stalking defendants (54.0%, n = 102) had a prior criminal history unrelated to stalking or harassment and 32.3% (n = 61) acknowledged or were reported to have a history of violent behavior. The most common diagnostic categories ascribed to these offenders were personality disorder, diagnosed in 69 cases (36.5%; 11 of these individuals were also diagnosed with an Axis I disorder), schizophrenia, diagnosed in 32 cases (17.2%), delusional disorders (typically erotomanic or persecutory type), diagnosed in 29 cases (15.6%), and mood disorders (including bipolar disorders, major depressive episode, dysthymia, and depressive disorder not otherwise specified), diagnosed in 22 cases 6 Ten cases could not be identified because they did not have NYSID numbers, three offenders were incarcerated throughout the follow-up period.

7 Recidivism in Stalking 257 (11.8%). Further analysis of the subgroup diagnosed with a personality disorder (roughly half of whom were diagnosed with personality disorder not otherwise specified) revealed that Borderline, Antisocial, and Narcissistic personality disorders/traits were the most frequently ascribed, present in 36 of these 69 cases (52.1%). Characteristics of Harassment The victims of harassment were largely, although not exclusively, female, with 112 offenders (59.6%) only known to have harassed one woman and an additional 15 (8.0%) having harassed more than one different woman; 107 of these 127 offenders were male. Men were the victim of harassment in 18.8% of cases (n = 35) and in all but two of these cases there was only one victim; 16 of the offenders who harassed men were female. In 26 cases (14.1 %) both men and women were harassed by the same offender and these offenders were largely male (n = 20, 76.9%). The average length of time that these victims had been harassed prior to arrest was 16.5 months (SD = 28.6). Former intimate partners were the victims of harassment in 59 cases (33.3%; these data were missing in 12 cases). Acquaintances, friends (nonromantic relationships), and former business associates (supervisors, coworkers, attorneys, etc.) were victimized in 48 cases (27.1%). Family members (primarily parents, siblings, and children) were the victims of harassment in 25 cases (14.1%) and strangers were targeted in 45 cases (25.4%). The motive behind the harassment was romantic in 60 cases (39.2%), anger or revenge in 69 cases (45.1%), and other motives (including a mixture of anger and romantic aspirations) were present in 24 cases (15.7%); motive could not be determined in 36 cases. In 114 cases (60.3%), direct threats were made toward the victim or a third party in the course of the stalking case and violence occurred in 32.8% of cases (n = 62). Recidivism Data Of the 189 offenders originally studied, 93 were known to have reoffended (49%) and the average number of stalking-related offenses committed (separate charges) was 2.3 (SD = 1.9). Of the 93 offenders who reoffended during the follow-up period, fully 80% reoffended within the year following their first arrest. The average length of time to reoffense, for those offenders who reoffended, was 11.9 months whereas offenders who did not reoffend were at risk for an average of 71.6 months. Because 41 of these offenders were found incompetent to stand trial and hospitalized, these cases were excluded from the analysis of recidivism because it was not possible to accurately determine whether or for how long they were at risk of reoffending (i.e., when they were released from the hospital). Of the 148 cases used in the following analyses, 72 (49%) were known to have reoffended during the follow-up period (see Fig. 1). There were no differences, on any of the variables studied, between the 148 offenders included in these analyses and the 41 who were found incompetent to stand trial, suggesting that the exclusion of these cases was not likely to have led to any systematic bias.

8 258 Rosenfeld Fig. 1. Survival curve for total sample. Univariate analyses were first to ascertain which variables were significantly associated with recidivism (see Table 1). These analyses revealed that several of the expected correlates of recidivism were significantly associated including age, r(n = 141) =.19, p =.02, prior intimate relationship, χ 2 (df = 1) = 6.35, p =.01, and personality disorder diagnosis, χ 2 (df = 1) = 7.81, p =.005. A psychotic disorder diagnosis was negatively associated with recidivism, χ 2 (df = 1) = 3.57, p =.06, although this analysis did not reach significance. Subsequent analyses examined whether offenders with a delusional disorder diagnosis were at significantly higher Table 1. Predictors of Recidivism Variable Total sample Reoffense No reoffense φ p Gender (% Male) 76.9% 78.6% 75.3% Age (mean in years) a.02 Length of harassment (months) a.65 Prior Intimate relationship 31.8% 41.7% 22.4% Prior criminal history 56.1% 61.1% 51.3% Prior unrelated violence 34.5% 41.7% 27.6% Psychiatric hospitalizations 48.7% 41.7% 55.3% Delusional beliefs 40.7% 33.8% 47.8% Personality disorder 38.1% 50.0% 27.6% Cluster B disorder/traits 17.6% 29.2% 6.6% Psychotic disorder 41.0% 32.8% 48.6% Delusional disorder 13.5% 6.9% 19.7% Substance abuse 50.4% 56.6% 44.6% a point-biserial r.

9 Recidivism in Stalking 259 risk of reoffending yet the reverse was found, as these offenders were significantly less likely to reoffend, χ 2 (df = 1) = 5.18, p =.02. On the other hand, offenders with a Cluster B personality disorder (antisocial, borderline, narcissistic) were significantly more likely to reoffend than were other offenders (including those with other personality disorder diagnoses), χ 2 (df = 1) = 13.03, p = Several variables that were anticipated to differentiate offenders who did and did not reoffend were not significantly different among these two groups. For example, a prior criminal record unrelated to stalking was unrelated to recidivism in this sample, despite being among the more frequently observed correlates of recidivism in other populations, χ 2 (df = 1) = 1.44, p =.23. History of previous violent behavior was also not significantly more common among reoffenders, although this difference approached significance, χ 2 (df = 1) = 3.22, p =.07. Similarly, history of previous psychiatric hospitalizations, which has often differentiated mentally ill repeat offenders from their one-time offender counterparts, approached significance, χ 2 (df = 1) = 2.74, p =.10. Finally, neither gender, χ 2 (df = 1) = 0.22, p =.64, nor history of substance abuse, χ 2 (df = 1) = 2.03, p =.15, were significantly associated with recidivism in our dataset. Survival Analysis Cumulative survival estimates were plotted for time at risk by reoffense status incorporating the variables previously described as covariates using a Cox Proportional Hazards model. The results of these analyses are presented in Table 2. Virtually all of the variables that were significantly associated with reoffending in univariate analyses significantly influenced survival curves in these analyses as well, including age, victim offender relationship in general and prior intimate relationship in particular (see Fig. 2) and personality disorder diagnosis. In addition, history of previous violence, which approached significance in the univariate analysis, also appeared to Table 2. Factors Influencing Recidivism Survival Curve Variable df Log Rank χ 2 p 2 LL p Gender Age a Victim-offender relationship Prior intimate relationship Prior criminal history Prior unrelated violence Psychiatric hospitalizations Personality disorder Cluster B disorder/traits b Psychotic disorder Delusional disorder c Substance abuse Personality disorder Substance abuse Psychotic disorder Substance abuse a Under 30 vs vs. over 40 years old. b Cluster B personality disorder vs. other personality disorders vs. no personality disorder. c Delusional disorder vs. other psychotic disorder vs. nonpsychotic.

10 260 Rosenfeld Fig. 2. Personality disorder type and recidivism. significantly influence survival curve shape ( p =.06 based on one indicator [Log Rank χ 2 ] and p =.02 based on the second [ 2 Log Likelihood]). We also conducted several analyses intended to break down the personality disorder effect seen in both survival analysis and frequency analysis, assessing the relative risk of a Cluster B personality disorder (including those diagnosed with personality disorder not otherwise specified with borderline, antisocial, and/or narcissistic traits) compared to other types of personality disorder and offenders without a personality disorder diagnosis. This analysis, displayed in Fig. 3, showed that Cluster B diagnosis posed a substantially greater risk of reoffense relative to other personality disorder diagnosis while the latter were not substantially more likely to reoffend than were offenders without a personality disorder. In addition, while history of substance abuse was only marginally associated with survival (p =.13 based on Log Rank χ 2 and p =.05 based on 2 Log Likelihood), the interaction between substance abuse and personality disorder was significant. Inspection of the survival curve for this analysis demonstrates that reoffense was most rapid and most likely for those offenders with a history of substance abuse and a personality disorder diagnosis (see Fig. 4). Either one of these two risk factors alone resulted in a somewhat lower rate of recidivism while offenders with neither risk factor were at the lowest risk for reoffending. A similar analysis modeling the interaction between psychotic disorder diagnosis and substance abuse did not reach significance and the analysis comparing offenders with a delusional disorder compared to those with other psychotic disorders and nonpsychotic offenders was only marginally significant ( p =.09 based on Log Rank

11 Recidivism in Stalking 261 Fig. 3. Victim offender relationship and recidivism. Fig. 4. Personality disorder, substance abuse, and recidivism.

12 262 Rosenfeld Fig. 5. Psychotic disorder type and recidivism. χ 2 and p =.04 based on -2 Log Likelihood). Nevertheless, an inspection of the survival curves for this analysis (Fig. 5) reveals an interesting pattern, with those offenders diagnosed with a delusional disorder demonstrating an apparent plateau in reoffending (i.e., all reoffending within this group occurred within the first few months; those offenders who did not reoffend quickly did not reoffend at all). There was no apparent interaction between psychosis and substance abuse with regard to reoffending. DISCUSSION This study, which represents the first empirical study of recidivism among stalking offenders, revealed a number of unexpected findings. First, despite common folklore that all stalkers continue pursuing their victim indefinitely, we found a recidivism rate of roughly 50%. Although this rate may initially appear high, it is certainly substantially lower than many stalking victims, mental health professionals, and legal scholars might have predicted. The emphasis (although not exclusive reliance) on official records in this study may have underestimated the true recidivism rate to some extent, but the unique nature of stalking offenses likely bolsters the validity of official records as an indicator of recidivism because few repeat stalkers go undetected if they resume stalking an individual who has previously filed criminal charges. Thus, although some potential for underestimation of true recidivism rates still exists, this potential source of bias is probably quite modest. Nevertheless, because only

13 Recidivism in Stalking 263 individuals who were arrested were included in this study, the possibility exists that different correlates of recidivism might exist in that subgroup of stalkers who are not arrested. A number of surprising findings also emerged in the analysis of predictors of recidivism. For example, despite an expectation that offenders with a delusional disorder would be the most persistent subgroup of offenders (i.e., most likely to reoffend), these data demonstrated the opposite; delusional offenders were less likely to reoffend in general (based on frequency analysis) although their overall survival curves were only marginally different than those for offenders with other psychotic disorders and nonpsychotic offenders. Likewise, although the interaction between substance abuse and psychosis has often been described as highly problematic in studies of violent and criminal behavior, this combination of risk factors did not significantly influence recidivism in our sample of stalking offenders (although it approached significance in the less conservative analysis of the Cox proportional hazards model). It should be noted, though, that by excluding a portion of the most severely ill offenders (i.e., those found incompetent to stand trial), these important variables (delusional disorder and substance abuse) may have been masked to some extent. Although a reanalysis of the data using all 189 cases generated comparable results, some caution is nevertheless warranted in generalizing from these findings to the larger population of stalking offenders because the most severely ill offenders may not have been detected (e.g., if an offender was hospitalized throughout the follow-up period because of psychotic symptoms). Clinical experience, however, suggests that the likelihood of this is relatively low, as few delusional stalking offenders are hospitalized absent a finding of incompetence and hospitalizations after a finding of incompetence are typically brief (an oft-heard complaint from stalking victims). Whereas some expected correlates of recidivism were not supported by these analyses, the presence of a personality disorder diagnosis and, more specifically, a Cluster B personality disorder (e.g., antisocial, borderline, narcissistic), was strongly associated with an increased risk of reoffending. In addition, the combination of a personality disorder and a history of substance abuse raised the risk of reoffending even further, far surpassing the risk of either of these two factors alone; virtually all of the offenders with both a personality disorder and a history of substance abuse reoffended during the follow-up period. Also, offenders with a prior intimate relationship were also at significantly greater risk of reoffending, followed closely by those who victimized strangers while other offenders were somewhat less likely to reoffend. Although these findings may not seem obviously counterintuitive, conventional wisdom would likely suggest that psychotic offenders, and those with delusional disorders in particular, are most prone to reoffend. Thus, the finding that romantically estranged, personality disordered individuals represent a higher risk for reoffending than those with a delusional attachment to a stranger or acquaintance is somewhat surprising. One possible (albeit hypothetical) explanation for these findings is that psychotic offenders may have been more likely to receive aggressive psychiatric treatment (e.g., hospitalization, treatment with antipsychotic medications) and that this treatment was the reason why reoffending was less likely among delusional offenders. Unfortunately, without additional information regarding treatment compliance this hypothesis remains largely speculative.

14 264 Rosenfeld On the other hand, several variables that have consistently emerged as predictors of recidivism in other offender populations, such as prior criminal history and prior psychiatric hospitalizations, were unrelated to recidivism in our sample. Further, the length of time harassment had gone on for prior to the first (known) involvement of the legal system was unrelated to whether the offender continued or ceased their harassment. These null findings, which are all the more striking given the numerous significant findings that were observed, highlight the need to assess risk factors for recidivism in the specific population of interest rather than generalizing from other populations and research literatures. These results are also noteworthy in that they highlight the overlap between stalkers who reoffend and the emerging literature on stalking-related violence. Recent studies have consistently identified stalkers who target former intimates and those with a history of substance abuse as representing an increased risk of violence (Rosenfeld, in press). On the other hand, delusional stalkers were somewhat less likely to reoffend in the present study and have been found to be less likely to be violent in a number of prior studies (Rosenfeld, in press). These converging literatures may ultimately help identify a subgroup of stalking offenders that represent the most problematic stalkers in terms of both potential violence as well as reoffending. Successful intervention strategies focused on such a subgroup of offenders might therefore have a substantial impact on reducing the most problematic aspects of stalking offenses. Although this study is admittedly preliminary, and the modest sample size (N = 148 in whom reliable follow-up data were available) limits the use of more sophisticated multivariate modeling (e.g., multivariable proportional hazard regression models), our findings are likely encouraging for the victims of stalking and harassment crimes. Despite the substantial recidivism rate, common sense might suggest that delusional offenders who have engaged in an extended pattern of harassment will simply never stop. Our data suggest that this pessimism may be somewhat exaggerated. Particularly given the possibility that interventions may have significantly influenced the course of stalking behaviors, the potential for relief clearly exists. Further research must strive to clarify which, if any, interventions are most effective and which whom are particular types of interventions (e.g., mental health treatment, criminal justice sanctions, pharmacotherapy) best suited. Only with continued efforts will we be able to offer a modicum of relief to the victims of stalking and harassment. ACKNOWLEDGMENTS Several individuals have been instrumental in developing this database including Ronnie Harmon of the New York City Forensic Psychiatry Clinic, without whose assistance this dataset might never have been compiled. In addition, research assistants Justine Schmullinger and Dana Levin, as well as Dr Bruce Frederick and Susan Jacobson from the New York State Division of Criminal Justice Services were crucial in compiling and coding the relevant data.

15 Recidivism in Stalking 265 REFERENCES Albonetti, C. A., & Hepburn, J. R. (1997). Probation revocation: A proportional hazards model of the conditioning effects of social disadvantage. Social Problems, 44, Allison, P. D. (1995). Survival analysis using the SAS system. Cary, NC: SAS Institute Inc. Bonta, J., Law, M., & Hanson, K. (1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 123, Brewster, M. P. (2000). Stalking by former intimates: Verbal threats and other predictors of physical violence. Violence and Victims, 15, Cottle, C. C., Lee, R. J., & Heilbrun, K. (2001). The prediction of criminal recidivism in juveniles: A meta-analysis. Criminal Justice and Behavior, 28, Furby, L., Weinrott, M. R., & Blackshaw, L. (1989). Sex offender recidivism: A review. Psychological Bulletin, 105, Hall, G. C. N. (1995). Sex offender recidivision revisited: A meta-analysis of recent treatment studies. Journal of Consulting and Clinical Psychology, 63, Harmon, R. B., Rosner, R., & Owens, H. (1995). Obsessional harassment and erotomania in a criminal court population. Journal of Forensic Sciences, 40, Harmon, R. B., Rosner, R., & Owens, H. (1998). Sex and violence in a forensic population of obsessional harassers. Psychology, Public Policy, and Law, 4, Kamphius, J. H., & Emmelkamp, P. M. G. (2001). Traumatic distress among support-seeking female victims of stalking. American Journal of Psychiatry, 158, Kienlen, K. K., Birmingham, D. L., Solberg, K. B., O Regan, J. T., & Meloy, J. R. (1997). A comparative study of psychotic and nonpsychotic stalking. Journal of the American Academy of Psychiatry and Law, 25, Meloy, J. R. (1998). The psychology of stalking. San Diego: Academic Press. Meloy, J. R., Davis, B., & Lovette, J. (2001). Risk factors for violence among stalkers. Journal of Threat Assessment, 1, Meloy, J. R., & Gothard, S. (1995). Demographic and clinical comparison of obsessional followers and offenders with mental disorders. American Journal of Psychiatry, 152, Mullen, P. E., Pathé, M., & Purcell, R. (2000). Stalkers and their victims. Cambridge, UK: Cambridge University Press. Mullen, P. E., Pathe, M., Purcell, R., & Stuart, G. W. (1999). Study of stalkers. American Journal of Psychiatry, 156, Palerea, R., Zona, M. A., Lane, J. C., & Langhinrichsen-Rohling, J. (1999). The dangerous nature of intimate relationship stalking: Threats, violence and associated risk factors. Behavioral Sciences and the Law, 17, Pathé, M., & Mullen, P. (1997). The impact of stalkers on their victims. British Journal of Psychiatry, 170, Prentky, R. A., Lee, A. F. S., Knight, R. A., & Cerce, D. (1997). Recidivism rates among child molesters and rapists: A methodological analysis. Law and Human Behavior, 21, Quinsey, V. L., Rice, M. E., & Harris, G. T. (1995). The actuarial prediction of sexual recidivism. Journal of Interpersonal Violence, 10, Rosenfeld, B. (1992). Court-ordered treatment of spouse abuse. Clinical Psychology Review, 12, Rosenfeld, B. (in press). When stalking turns violent: Developments in the assessment of stalking risks. In M. Brewster (Ed.), Perspectives in stalking. Kingston, NJ: Civic Research Institute. Rosenfeld, B., & Harmon, R. (2002). Factors associated with violence in stalking and obsessional harassment cases. Criminal Justice and Behavior, 29(6), Rosenfeld, B., & Ritchie, K. (1998). Competence to stand trial: Clinician reliability and the role of offense severity. Journal of Forensic Sciences, 43, Schwartz-Watts, D., Morgan, D. W., & Barnes, C. J. (1997). Stalkers: The South Carolina experience. Journal of the American Academy of Psychiatry and Law, 25, Tjaden, P., & Thoennes, N. (1998). Stalking in America: Findings from the National Violence Against Women Survey. Washington, DC: U.S. Department of Justice Westrup, D., Fremouw, W. J., Thompson, R. N., & Lewis, S. F. (1999). The psychological impact of stalking on female undergraduates. Journal of Forensic Sciences, 44, Winner, L., Lanza-Kaduce, L., Bishop, D. M., & Frazier, C. E. (1997). The transfer of juveniles to criminal court: Re-examining recidivism over the long term. Crime and Delinquency, 43,