THE ASSESSMENT, TREATMENT AND COMMUNITY OUTCOME OF INSANITY

ACQUITTEES : FORENSIC HISTORY AND RESPONSE TO TREATMENT

STEPHEN L. GOLDING, Ph. D.

Department of Psychology

University of Utah

DEREK EAVES, F.R.C.P.

Forensic Psychiatric Services Commission

Burnaby, British Columbia

ANDREA KOWAZ Ph. D.

Department of Psychology

Simon Fraser University

Burnaby, British Columbia

International Journal of Law and Psychiatry

1989, 12, 149-179

The rather startling similarities in the controversies surrounding the insanity defense over several centuries (Golding & Roesch, 1987; Hans, 1986; Hermann, 1983; Shah, 1986) underscore the transhistorical and deeply rooted problems involved in the development of a rational and coherent policy for adjudicating criminal non-responsibility on grounds of mental disorder. The importance of developing a clear and coherent set of social, legislative and clinical policies with respect to insanity acquittees, regardless of the statistical infrequency of the insanity verdict (Pasewark & McGinley, 1985), is made patently clear by the continuing legal, public, and scholarly controversy which surrounds the defense. Modern sentiments following John Hinckley's acquittal by reason of insanity differ little from those of Queen Victoria, "We have seen the trials of Oxford and MacNaughten conducted by the ablest lawyers of the day - and they allow and advise the Jury to pronounce the verdict of not guilty on account of insanity, whilst everybody is morally convinced that both malefactors were perfectly conscious and aware of what they did" (cited in Walker, 1968, p. 168). It is quite clear that the "bundle" of problems surrounding the insanity defense and the disposition of insanity acquittees will remain a focal problem, whether or not one abolishes or modifies the plea. Until both social science researchers and political decision-makers, as well as the lay public, have available a longitudinally based patient tracking and evaluation system that allows the systems-wide impact of various changes to be understood, measured, and predicted, the consequences of such changes are likely to remain chaotic and often incongruent with the intentions (whether reformist or abolitionist) of the social policy reformers. The failure to base such policy analyses and reforms on a comprehensive understanding of the impact of such policies forces us to constantly repeat the errors of the past and almost guarantees non-optimal decisions (see Morrissey & Goldman, 1986; Roesch & Golding, 1985; and Stone, 1982 for examples in the area of civil commitment). As even a brief critical review of outcome research on insanity acquittees makes clear, our knowledge of the long range antecedents, consequences and clinical outcome of decision making in the area of insanity acquittees is limited and is primarily sociological in focus (Steadman & Morrissey, 1986). Thus, we have some crude data on the infrequency of the plea and its success (Pasewark & McGinley, 1985); demographic, criminological and diagnostic characteristics of acquittees; and length of hospitalization and subsequent criminal activity (see Pasewark, 1986; Pasewark, 1982; Steadman & Braff, 1983 for reviews). While this limited literature is useful, the most serious problem is that critical policy and clinical issues cannot be addressed without individualized and longitudinally based follow-up data. Thus, the limited data on length of hospitalization or recidivism (Pasewark, 1986; Steadman & Braff, 1983) is ambiguous unless we can know an individual's response to treatment, the grounds for their release, and the nature of the follow up services provided. Measuring the "failure" rate without concern for how the decisions for release were made or followed-up do not answer the most interesting and policy-relevant questions.

A clinical as well as a legal and criminological focus is, in fact, dictated by the Supreme Court's decision in Jones v. United States (1983),

An insanity acquittee is not entitled to his release merely because he has been hospitalized for a period longer than he could have been incarcerated if convicted. The length of a sentence for a particular criminal offense is based on a variety of considerations, including retribution, deterrence, and rehabilitation. However, because an insanity acquittee was not convicted, he may not be punished. The purpose of his commitment is to treat his mental illness and protect him and society from his potential dangerousness. There simply is no necessary correlation between the length of the acquittee's hypothetical criminal sentence and the length of time necessary for his recovery. (p. 355)

and further,

A verdict of not guilty by reason of insanity is sufficiently probative of mental illness and dangerousness to justify commitment of the acquittee for the purposes of treatment and the protection of society. (p. 354)

Thus, in order to fashion a coherent system (where the grounds for commitment and release are strictly rehabilitative and protective of society), we need to know how the prior treatment received by a defendant to restore fitness relates to treatment as an NGRI acquittee, when acquittees have maximally responded to in-hospital treatment, and so forth. At present, little systematic data that address the patterns of recovery and deterioration of NGRI patients as a function of different treatment modalities and decisional policies exist. The need for such a longitudinally based data base is also clear when one considers how the legal and political issues here are firmly tied to empirical and clinical issues. The research reported here is based upon a long-term longitudinal data collection effort that sought to track a cohort of NGRI acquittees both retrospectively, based upon archival data, and prospectively, based upon clinical and agency data. Because of the complexity and the on-going nature of the research, this study will focus on the cohort's a) prior forensic history, and b) treatment while initially hospitalized. A subsequent report will focus on a) long-term follow-up of absolute discharge, b) community experience while on conditional discharge and c) predictive patterns of response to treatment, clinical and social problems during conditional discharge, and subsequent clinical and community adjustment following absolute discharge. We now turn our attention to a brief review of previous research.

Forensic history

A number of studies have reported the prior criminological and hospitalization experiences of NGRI cohorts. In both Canadian and United States samples, considerable variability in simple prior arrest or conviction rates for NGRI acquittees has been reported. Rates generally fall into the 30 - 45% range (Bogenberger, Pasewark, Gudeman & Beiber, 1987; Cooke & Sikorski, 1974; Criss & Racine, 1980; Greenland, 1979; Hawkins & Pasewark, 1983; Hodgins, 1983; Pantle, Pasewark & Steadman, 1980; Pasewark, Pantle & Steadman, 1979a; Pasewark, Pantle & Steadman, 1979b; Pasewark, Bieber & Bosten, 1982). Other studies (summarized in Pasewark, 1986; and Steadman & Braff, 1983) show a similar degree of variability. The highest figure reported for United States jurisdictions is 64% in a small Connecticut sample (Phillips & Pasewark, 1980) and 89% in a Canadian maximum security population (Quinsey, Pruesse & Fernley, 1975). No cogent explanations for the variation among reported studies are available, however it is obvious that jurisdictional differences in the type of individual acquitted are considerable (Keilitz, 1987; Steadman, 1987). It is important to note, however, that when reported, it seems clear that most prior charges in these cohorts are for relatively non-serious crimes. For example, Pasewark (Pasewark et al., 1979b) reports a New York sample in which 2.0% of prior charges are for murder, 14.0% are for assault (mostly common assault) and 3.9% are for sexual crimes.

It should come as no surprise that most studies of NGRI cohorts have reported relatively high rates of prior hospitalizations and mental health contacts. Typically, more than two-thirds of samples have prior hospitalizations, with rates ranging from 20 to 80% (Bogenberger et al., 1987; Cooke & Sikorski, 1974; Criss & Racine, 1980; Hawkins & Pasewark, 1983; Pasewark et al., 1982). Fewer studies report types or outcomes of prior hospitalizations, other mental health contacts, or a breakdown by prior fitness hospitalizations. Steadman, Keitner & Braff (1983) report that prior hospitalizations, both civil and criminal, do not relate to length of NGRI hospitalizations, but do not report the number, length or relationship to current NGRI offense for these prior contacts. Pasewark et al. (1979b), reporting on an extension of the Steadman sample, found that 44% of their male subjects had prior hospitalizations; the average individual had two prior civil and 0.6 prior criminal hospitalizations. Phillips and Pasewark (1980) report that 61% of their sample had prior hospitalizations, and that 60% of these were "criminal" hospitalizations (competency, lateral transfer, etc.). None of these studies attempt to analyze the chronological or treatment-outcome relationship between past hospitalizations and NGRI index offense. A few studies do address the issue of concurrent fitness evaluations or determinations relative to the index NGRI offense, finding that approximately 40% are found unfit prior to their NGRI adjudication (see Golding & Roesch, 1987 for a review).

Characteristics of hospitalization

As noted previously, the Supreme Court of the United States, in Jones v. United States (1983), held that an insanity acquittee is not to be treated as a criminal convictee: the issues pertaining to commitment and release must focus on the individual's mental illness and their dangerousness, not on the "crime" they committed. However, because their dangerous acts have been found to occur "beyond a reasonable doubt," the Court would allow the determination of their continuing mental illness and dangerousness to be decided according to a "preponderance of the evidence" standard rather than the "clear and convincing" evidence standard which holds for ordinary civil commitment (Addington v. Texas, 1979). Nevertheless, the most comprehensive analysis of the comparative length of hospitalizations for NGRI acquittees, conducted by Braff, Arvanites & Steadman, (1983) found clear evidence that nature of offense was the primary determinant of length of incarceration or hospitalization; there were no significant differences in length of hospitalization or incarceration between successful and unsuccessful NGRI pleaders, or in comparison to non-NGRI convictees.

Clearly, NGRI acquittees are held for long periods of time (public perceptions to the contrary), but these data, by themselves, have little clinical or social policy significance without knowing what treatments were provided, and with what outcome. Thus, these data have a very different meaning under interpretation A:

The defendants received maximum clinical benefit from psychotropic medication after an average of 3 months, as evidence by the plateau achieved in their pre-post adjustment and psychopathology ratings, but were held for an additional 120 months for social policy reasons.

and interpretation B:

The defendants psychotic symptomatology did not respond satisfactorily to psychotropic medication and/or there were instances of acute psychotic behavior that led staff to hold them until a period of at least one year of "non-episode" had past. On average, the typical defendant took 120 months to achieve this level of adjustment.

While we do not know how clinical and social policy factors interact in the decision making process, we do know that response to treatment or the absence of active psychotic symptomatology is likely to enter into such decision-making. Thus, these length of hospitalization data are quite ambiguous without knowledge of treatment history and response. Do they speak to insanity acquittees being treated more harshly (being held to equivalent "sentences" even though they have "recovered")? Do they conceal being actually held considerably longer when one takes pre-trial treatment versus detention into account? Do they reflect the effectiveness of psychotropic medication in subduing "behavioral disturbances" that are likely to affect release decisions?

In order to understand the NGRI acquittal and dispositional system, we have to be able to assess the pathways by which these individuals enter the system (including the effectiveness of prior treatment), how they progress through the system and respond to various treatments, and how they function upon release depending upon the treatment and follow-up services and supervision provided.

There is thus an urgent need to examine the criminal justice - mental health interface with respect to insanity acquittees with a longitudinally based research design that incorporates a clinical as well as a sociological or demographic focus. The research reported here, and in a subsequent report, focusses primarily upon the following issues:

1. It is extremely rare for the instant charge to which the defendant has pled "not guilty by reason of insanity" to be the first charge against the defendant, or, more generally, for it to be the first time societal concern has been raised about the individual (as indexed by prior involuntary hospitalizations). Moreover, it appears that approximately 40% of all successful insanity acquittees have been either evaluated or treated (as unfit) for the same mental illness that forms part of their insanity plea. Clinical research and legal policies surrounding the adjudication of fitness, the treatment of unfit defendants, civil commitment, the right to treatment, the prediction of treatment need and responsiveness, and the adjudication and treatment of insanity acquittees stand in obvious relationship each other. Thus, one critical aspect of the current research is to examine, clinically and chronologically, an individual's past contact with the "megasystem," in an attempt to fashion more responsive, rational and clinically sensible policies.

2. As Jones v. United States makes clear, we will need to know how they have been treated and with what result before we can make rational decisions about their adjudication following a successful insanity plea. Once adjudicated, a complex series of political, social and clinical decisional processes begins. We have no knowledge of those processes. The critical question, about which all others revolve, is knowledge of the response of insanity acquittees to various treatment modalities, and the peculiarities of their treatment. We are have little understanding of how such individuals respond to treatments, and how those treatments relate to other critical aspects of their release -- i.e. future potential for violence, future needs as a chronic mental patient, and so forth.

3. As insanity acquittees are released into the community, a process that is not well understood, we need to know how various decisional models function, and why certain individuals are successful, and why others fail. Other than gross recidivism rates, we have very little data regarding how this population really fares, either as modified dischargees or as absolute dischargees. Such data are of obvious importance to reformulation of legislative and social policies.

4. When one contemplates the length of time such defendants spend in hospital prior to adjudication and in prior hospitalizations, the length of time in hospital following adjudication, and the intensity of contact while on modified or unconditional discharge status, the costs are overwhelming. Thus, a final goal of the present research is to provide data that may form the basis of alternative systems that are both more humane, increase the security of other citizens, and are cost effective.

Methodology

Subjects. The subjects in our study consisted of all persons found not guilty by reason of insanity (NGRI) between November, 1975 and January 1, 1984 in British Columbia. In addition, 30 subjects who were already adjudicated as NGRI prior to the formation of the Forensic Psychiatric Services Commission (in 1975) and transferred to the Forensic Psychiatric Institute were included. Our subjects consist of three main cohorts: a) OIC subjects who, by 1/1/84, had not yet been discharged from their initial hospitalizations under an order-in-council (N=50), b) MOIC subjects who, by 1/1/84, had experienced at least one discharge into the community under a modified order-in-council (N=75), and c) ROIC subjects who, by 1/1/84, had their orders-in-council officially rescinded (N=38). In addition, 25 subjects were coded as special because of circumstances that affected their ultimate disposition. Of these special subjects, nine were officially listed as having committed suicide, two died of natural causes, four died of causes officially unclassified that may have been suicides (e.g. over-dosage of lithium and alcohol) or accidental over-dosages, two were transferred to federal prisons for security reasons, and seven had unique outcomes (transfer to another province, NGRI overturned on appeal, etc.). For some analyses, these individuals are included where appropriate. In most analyses, the nature of the logical subject pool is clear from the context. For example, by definition, all 188 subjects are included in the analyses of past hospitalizations and criminal charges, but only MOIC and ROIC and some special subjects are included in the analysis of initial hospitalization period following NGRI since the OIC cohort had not yet had at least one discharge to the community (MOIC) or rescission of their OIC (ROIC). Five individuals found NGRI but deported were not included in any analyses. Of the 188 subjects, 90.4% were male; 9% were under the age of 22 when they offended; 36.1% were 22-30; 32.4% were 30-40, and the rest over 40 when they offended. As is typical in such populations, almost all subjects carried psychotic diagnoses.

Overview of coding procedures. All subjects' prior forensic history, hospitalization and outpatient records were coded as of their status on 1/1/84. Follow-up coding of mental health and arrest records continues, and will be discussed primarily in a subsequent report. After that date, significant changes in status are not reliably coded but were available for some subjects (e.g. discharge, suicide). In such cases, an effort was made to appropriate adjust a patient's status. Coders worked on all available documents (as described below) and used special codes when the archival material would not permit coding or was missing.

Offense coding. The nature of an individual's NGRI offense, and prior and subsequent offenses was coded according to a uniform system based on the Criminal Code of Canada. In general, we attempted to code the nature and date of the offense, weapons used, number of victims, type of victim and disposition where appropriate. All codes were based upon information obtained from the Royal Canadian Mounted Police's computer system using their form C480. In most analyses, we collapse groups of offenses into categories reflecting descending offense seriousness (Murder or attempted murder, Sexual assault, Assault, Robbery, Weapons, Property, Nuisance, Theft, and Drugs). Our coding of subsequent charges and convictions currently extends until March of 1986, the last cycle for which we obtained data.

Psychopathology coding. A profile of psychopathological signs and symptoms was obtained for each subject at each legally significant point in time following arrest for the index offense. Where applicable, the profile was coded at each of the following time points:

-on remand (jail screening as potentially unfit)

-on admission and discharge as unfit to stand trial

-on admission as NGRI

-on first discharge to the community (as a modified order-in-council) or status as of 1/1/84 if not yet discharged

-on subsequent inpatient hospitalization admissions and discharges while under modified order-in-council

-for each community tenure period (moic cycle marked by readmission to hospital) and/or as of 1/1/84 if in the community

-on final recision of the order-in-council.

Each profile of psychopathology was divided into three major sections: a) Psychopathological signs and symptoms, rated from a psychiatrist's mental status examination and organized into 11 areas; b) ward observations derived primarily from nursing notes; and c) available psychological test data. The eleven areas into which psychopathological signs and symptoms are organized were derived from our previous experience in coding such archival data and from our semi-structured interview protocol, the Interdisciplinary Fitness Interview (Roesch & Golding, 1980; Golding, Roesch & Schreiber, 1984). This protocol organizes psychopathological signs and symptoms into ten hierarchical areas: 1) Disturbance of Orientation / Consciousness, 2) Disturbance of Mood / Affect, 3)Ideation, 4) Primary speech disturbance (e.g. incoherence), 5) Secondary speech disturbance (e.g. pressure), 6) Disturbance of memory, 7) Formal thought disorder (e.g. thought broadcast), 8) Present delusions, 9) Present hallucinations, 10) Schizotypal / borderline signs. An eleventh area consists of various behavioral manifestations of pathology such as behaving as if hallucinated, agitation or excitement, bizarre gestures, etc. Each area and sub-categories are coded as present, absent or uncodable.

The ward observations were based upon an adaptation of Lorr and Vestre's (1968) Psychotic Inpatient Profile. Originally this instrument was designed as a behavior check list with items subsumed into 12 areas. We changed the format of the instrument to allow for direct rating of the 12 areas by the coders of the nursing notes. Each area was coded on a 0-3 scale (0 = not at all characteristic; 1 = occasionally true; 2= fairly often true; 3 = nearly always true) that represented the sense of the nursing notes for a two week period. The areas coded were: 1) Excitement, 2) Hostile Belligerence, 3) Paranoid projection, 4) Anxious depression, 5) Retardation, 6) Seclusiveness, 7) Care needed, 8)

Psychotic Disorganization, 9) Grandiosity, 10) Perceptual disorganization, 11) Depressed mood, and 12) Disorientation. Finally, psychological test scores (partial MMPI, WAIS and Rorschach protocols) were coded as available.

Medication history. While other treatments may have been made available to subjects in our sample, only psychotropic medications were recorded with any reliability. Furthermore, psychotropic treatment remains the primary treatment modality in most institutions. The physicians, psychologists, nurses and social workers who have daily contact with the patients undoubtedly provide valuable counseling, support and social skills training, but this is completely undocumented with any regularity or precision. During each patient's remand, unfit hospitalization, initial hospitalization, outpatient treatment or re-hospitalization periods, all psychotropic medications administered were coded on a time sampling basis. In addition to this time-sampling procedure, medications were event sampled on the basis of the week preceding and following a critical incident (see below). For each medication administered, its type and dosage on a weekly basis were recorded. Where possible, medications were converted to approximate chlorpromazine equivalents.

Critical incident coding. Critical incidents were coded for both all inpatient and outpatient periods whenever they were recorded in either nursing notes or in entries in a special official log. A critical incident was defined as either a) an entry into the official critical incident register not pertaining to a routine injury, b) any clinically significant entry in nursing notes involving serious concern for the patient's own safety or the safety of others, or c) any attempted or actual escape. Major categories of critical incidents were a) assaults, b) self-injury, c) property damage, d) medical crises or e) escapes.

Forensic history and psychopathology. As detailed in context, a patient's forensic history and psychopathology were coded according to a complex protocol that included the following major areas: 1) basic demographics, 2) complete date chronology of NGRI charges (offense date, remands, unfit hospitalization, NGRI admission, discharge and re-cycle dates, escapes, final status), 3) all codable previous psychiatric admissions (including dates, diagnoses and treatments received), 4) previous outpatient care, 5) premorbid social adjustment, 6) psychiatric and forensic history, 7) onset of current condition, and 8) drug/alcohol/weapons usage history.

Outpatient history. For each applicable community discharge, all patient's records at the Vancouver Clinic (which supervises outpatients) were coded. A community tenure period was marked by a discharge from outpatient status and subsequent re-admission. If no re-admission had occurred, then there would be one open-ended community tenure period. For each community tenure, the nature of the patient's adjustment, medications, events leading to re-admission, and contacts with the Clinic's staff were recorded. In addition, a sub-sample of patients were interviewed by nursing and social work staff to provide data on the nature of their community adjustment.

Rater/coder training. During the course of the project seven different individuals served as coders, although for most of the project, the same four were involved. All mental status data coders were post-Masters level students from Simon Fraser University's Psychology Department and were trained in the following manner. The senior author conducted a three day training workshop in which the forms and their specific items were discussed in detail. Coders then worked on two trial protocols under the supervision of the senior project co-ordinator and received detailed feedback as to coding problems. Thereafter, while working on data coding, raters worked in pairs and used each other for consultation with problems. Any significant problem was referred to the project co-ordinator and senior author for resolution by conference. If needed, coding manuals were altered for future cases as new rules or modifications became necessary, and prior cases were recoded, if needed to reflect any "mid-stream" changes.

Reliability analyses. Assessing the reliability of our coding processes is a difficult undertaking because of the numerous rating schedules employed and the complexity of each schedule. To estimate a reasonable lower bound of the various reliabilities, we undertook an analysis of the most complex and difficult protocol, involving coding psychiatric and nursing records of psychopathological signs and symptoms. This protocol involves the most rater judgment, and, insofar as it is applied to archival data that was originally recorded with varying degrees of completeness and thoroughness and depended upon the skill of the psychiatric examiner or ward nurses, it provides the most challenging task for a rater. We drew a random sample of four protocols from four raters for both an original NGRI admission and for an unfit admission, for a total of 32 original records. Each rater then rated the NGRI and unfit admission protocols of the other three raters from a complete review of the patient's chart, as would have been done by the original rater. The repeated measures pairs then formed the sample from which we calculated reliability statistics. This resulted in a total "sample size of 6 x 32 = 192 since there are (4x3)/2 unique combinations among the four raters. This method of "stringing out" rater pairs is the only reasonable way to approach a sufficient sample size without asking raters to perform many additional ratings. Its major drawback is that the actual number of patients sampled is small so that variables or codes that occur with very low frequencies are not adequately estimated (see below).

Two forms of reliability are depicted in Table 1 for the sector scores and ward observations (the principle variables used in subsequent statistical analyses of patient psychopathological profiles and change between various time points). The first four columns display the raw percentage of agreement between judgments of presence (+) or absence (-) of the particular symptom-cluster (sector). The remaining two columns depict the Anderberg (1973) statistic for agreement on presence (+) or absence (-) judgments. This statistic reflects the average% of agreement given that one of the raters made the judgment. Thus, for presence judgments it is calculated as

[+/+/(+/+ plus +/-) plus +/+/(+/+ plus -/+)]/2

and represents the mean percent of agreement for raters 1 & 2 given that one or both made a presence judgment. Other statistics such as correlation coefficients or kappa were not calculated because of the extreme skew in the marginals in cases where the symptom is frequently not present in much of the population (Janes, 1979; Maxwell & Pilliner, 1968). The Anderberg statistic adjusts for base rate and random agreement phenomena because it reflects joint agreement given a decision. For the ward observations, the percentage of exact agreements and agreement within one scale point are reported (see tabular note for a clarification).

---------------------------------------------------------- Insert Table 1 about here--------------------

Inspection of the reliability data reveals that the ranges are in accord with what has been previously reported for data of these type in the psychiatric literature (Golding et al., 1984). Major symptoms are coded with reasonable reliability and are often easier to code as to absence than presence. Some symptoms such as disturbance of orientation or consciousness or primary communication disturbance are so infrequent in the sample (as they are in the population) that they cannot be reliably estimated. Within the ward observation data the reliabilities are somewhat uniformly higher as is expected given the more routine noting of these data in daily nursing notes.

Results

CRIMINOLOGICAL ASPECTS OF NGRI OFFENSES

Offenses. Table 2 depicts the basic distribution of the charges against the 188 offenders in our sample. These offenders had a total of 263 charges laid against them. For ease of interpretation, the charges are presented in terms of the most serious charge, followed by the second most serious up to the fourth most serious charge.

---------------------------------------------------------- Insert Table 2 about here--------------------

The distribution of charges against offenders is in general accord with the results obtained by other investigators in both the United States and Canada although there is obvious variability. Slightly more than 40% of our sample had murder or attempted murder as their most serious charge. While in accord with other studies in terms of rank order, this percentage as noted previously, varies considerably.

While the distribution of serious charges indicates that a large portion of the sample was found NGRI for seriously aggressive behavior (murder, attempted murder, sexual assault, or arson [coded as a property offense]), it is important to note the significant numbers of individuals found NGRI for relatively minor charges. Most of the assault charges were for common assaults or other relatively non-serious assaults, the weapons charges are primarily for possession, and the other charges (with the exception of robbery) are by definition minor charges. Thus, approximately 30% of the sample is found NGRI for quite minor offenses. Leaving aside all other complexities of the histories of these individuals, it is clear that this has important policy implications.

Table 3 presents information on the nature of the victims and weapons used by the defendants for their index offense. As would be expected from the nature of index offenses, one-third of the crimes are victimless. The distribution of victims roughly in accord with national statistics in the United States and Canada for serious crimes, i.e. single known acquaintances or relatives are the major class of victim. The "other" category of weapon is quite frequent, as might be expected of the rather bizarre nature of many of the crimes committed by these defendants.

FORENSIC HISTORY

Past charges and convictions. Slightly over one-third of our sample had no recorded past charges. The remaining individuals (61.7% of the sample) had among themselves 832 recorded charges or convictions. The number of charges per individual ranged from 1 to 50. The details of these past charges are presented graphically in Figure 1. Most strikingly, the vast majority of

---------------------------------------------------------- Insert Figure 1 about here-------------------

this group have relatively benign criminal histories. Only sixteen individuals (8.5%) have prior charges involving murder or sexual charges. Most individuals have single charges of either assault, theft or nuisance crimes. For example, 32.4% (N=61) of the total sample have prior assault charges. Of these sixty-one individuals, 33 have single charges, 15 have two charges, and 13 have three or more assault charges. Most of these assault charges are for common assault, resisting arrest, or impaired driving: of 153 total charges in the assault category, 29 are for common assault, 20 are for resisting arrest, and 68 are for impaired driving. Only 36 charges were for serious assaults (assault or common assault causing bodily harm), and most of these are attributable to a few individuals.

Prior criminal activity-Chronology. It is well known that future criminal activity is best predicted by prior criminality (Monahan, 1978; Monahan, 1981; Monahan & Steadman, 1983; Steadman & Cocozza, 1978; Rabkin, 1979; Teplin, 1983). However, as previously discussed it is clear that our sample has little in the way of significant criminal activity, especially of the sort that is reflected in the more serious index offenses. In contrast to the median gap of 9.4 months between prior discharge and index offense, the median time interval between last (and minor) charge and index offense is 21.4 months (for those having a prior; 38.3% do not have a prior). Even when one adjusts for the possible effects of intervening hospitalizations, there is still a substantial difference. Thus, among the sub-sample having no intervening hospitalizations, the median is 11.4 months, and for those having such an intervening hospitalization, the median is 15.3 when that hospital time is removed. These data add to the strength of our observation that the sample is primarily one with repetitive mental disorder problems, largely, we speculate, unaddressed by the treatments and interventions they have previously received.

In sum, almost two-thirds of our sample have prior criminal records, primarily for minor assaults, property and theft charges, and nuisance charges. A series of statistical analyses performed failed to indicate any pattern of relationship between prior type or number of charges and NGRI index offense.

Prior hospitalizations-Frequency. As one might expect in this type of sample, 78.7% had been previously hospitalized and an additional 5.3% had received outpatient treatment only. Thus, only 16% of the sample had no recorded prior mental health contacts of any kind. Of those who had prior admissions, the range was between 1 and 17, with a mean of 4.11 hospitalizations. Of the aggregate of 609 prior admissions, 43.4% (264) were for forensic reasons involving remand, fitness or involuntary hospitalization. In terms of individuals, of the 148 having prior admissions, 108 individuals (73%; 57 percent of the total sample) had at least one of these admissions for forensic reasons. Thus, almost 80% of the sample had significant mental health histories involving frequent hospitalizations, and approximately three-quarters of the sample having prior contacts were hospitalized at least once for reasons that tied their mental disorder to either criminal charges or concerns over their dangerousness to self or others.

Prior hospitalizations-Chronology. Figure 2 depicts the time interval between last known discharge and NGRI index offense. The mean interval is 16.7 months, but the distribution is somewhat skewed and the median (the point at which 50% of the subjects have higher intervals and 50% have lower intervals) is more representative of the series. The median is 9.4 months. Put more simply, over half of the subjects with prior admissions were discharged within one year of their index offense. In fact, of those having a codable prior admission, 44.6 committed their index

------------------- Insert Figure 2 about here-------------------

offense within six months of their last discharge.

A quantitative and qualitative analysis of the nature of these "just prior hospitalizations" is quite revealing. Only 22.5% were voluntary hospitalizations. The majority were involuntary commitments for dangerous acts towards self or others (42.3%). The remaining reasons for admission all include legally significant aspects of the individual's mental disorder (16.2% as mentally disordered offender transferred from jail or prison, 8.5% as unfit to stand trial, and 2.8% for fitness evaluations). In terms of diagnostic categories, fully 63% of those hospitalized were schizophrenic, with the modal individual receiving a paranoid schizophrenic diagnosis. Only 3.1% of the individuals were diagnosed as anti-social personality disorder. Over three-quarters of the subjects had been previously hospitalized one or more times prior to this "just prior" hospitalization. The median admission lasted 40 days, and the primary mode of treatment for almost the entire sample was psychotropic medication. Qualitatively, an analysis of the existing records indicated that the individual's predominantly psychotic symptomatology (76% were psychotic) was substantially related to their legally relevant behavior.

The logical implication of these data is that the vast majority of individuals who were ultimately found NGRI for their index offenses were identified as suffering from legally relevant mental disorders prior to their index offenses. This implication is supported by data on the chronological relationships between prior offenses, hospitalizations and the index offense. It should be clear from the data presented that a large group of the NGRI acquittees were within the forensic system and/or the mental health system prior to their offense, and that a substantial probability exists that certain characteristics may exist that would have identified them as in need of sustained treatment, which treatment may have prevented the index offense. Pursuing this question should be a major research agenda in the future. Can one find ways to predict for an individual in the system that this person is a likely candidate for further re-hospitalizations or legally relevant mental health incidents (civil commitment, remand, unfit, NGRI, dangerousness, etc.). Obviously, we are limited to an analysis of retrospective data on type of admission, discharge and diagnosis; the more likely candidates for prediction are response to treatment, paranoid ideation, delusions of harm, and so forth. The study would have to sample from an unconstrained sample of individuals having hospitalizations of various types and look for predictors of subsequent forensic contacts.

Forensic contacts prior to adjudication

Remands. A total of 85 individuals in our sample (45.2%) were remanded at least once for fitness evaluations. The mean remand period for first remand was almost two months (58.7 days), while the median was one month (29.0). However, the data are highly skewed: the minimum period is 6 days, and the maximum is 890 days.

Unfit hospitalizations. A total of 113 individuals were adjudicated as unfit to stand trial prior to their ultimate adjudication as NGRI. This is 61.1% of the sample, a figure that is considerably higher than those few studies that report such data averaging 40% (see above). There is a considerable discrepancy between the mean length of unfitness hospitalization (428.9 days) and the median (172 days). There are a small number of extremely long unfit admissions of 3087, 4061, 4120 and 4383 days (!) which distort the mean for the entire sample. If one ignores these distorting figures, it appears that the representatively unfit individual is treated for a period of approximately six months prior to be adjudicated as NGRI. As we will see, this brief time period of treatment is critical in that most of the change in measurable psychopathology, occurs during this period and approximately two-thirds of the entire sample have this treatment period prior to their NGRI adjudication.

Characteristics of hospitalizations

Length of initial hospitalization. During the time frame of our study, 124 of the 188 individuals had been discharged at least once to the community (hence constituting the MOIC and ROIC samples). The length of time for this sub-sample is presented graphically in Figure 3 which breaks this down by category of index offense. The mean length of time for the entire sub-sample is 49.9 months,

---------------------------------------------------------- Insert Figure 3 about here -------------------

while the median is 38.2 months. Individuals charged with murder or attempted murder spend an average of 71.9 months in their initial hospitalization. This is significantly different from the average of the entire group (F = 6.28, p < .0001), and is significantly different from the means for the property and nuisance groups by Scheffe's test (p < .05). However, there is little indication in our data, from whatever source, that these individuals were significantly more disturbed than the individuals in these other groups. As of now, we have been unable to discover any differences among the index offense groups in age, sex, marital status, ethnic group, prior charges (type or number), juvenile record, number of prior serious offenses or diagnostic category. Nor do the groups differ significantly with respect to the clinical profiles described below. While the issue merits extended discussion, at present it appears that they are initially held longer simply as a function of the index offense, not as a function of psychopathology, as was predicted.

Clinical profiles significant time points

We now attempt to examine some critical aspects of the question of how and when individuals change during the long process of their adjudication and hospitalization. Tables 4 and 5 present the relevant data for sector scores and nurses' observational ratings respectively. These data were analyzed in the following manner. Since we coded defendant's mental status at the various time points explained in Section II, we are able to determine the% of individuals who have significant pathology in each of the sectors at various time points. Thus, each pair of columns represents a repeated measurement, although different column groups represent different sub-sets of subjects who experienced the various combinations of legally significant events. The notes to Tables 4 and 5 explain which columns pertain to which groups of subjects.

---------------------------------------------------------- Insert Tables 4 and 5 about here----------------

Careful inspection of Tables 4 and 5 reveal converging data from two clinical sources (psychiatric examiners and nurses' observations). (For convenience, Table 4 is presented graphically in Figure 4.) As might be expected given the fact of little or

---------------------------------------------------------- Insert Figure 4 about here-------------------

no treatment offered on remand, there is no substantive change in the psychopathological profile of the cohorts between these time points. The second group of columns in each table, however, reveals that significant and substantial change occurs between fitness admission and fitness discharge in almost all areas of measurement from both sources. In fact, the change is so dramatic that practically no change occurs between any other time points -- to NGRI admission, discharge to the community or between OIC and 1/1/84 for those individuals not yet having their first community discharge. The implications of this finding are dramatic. It appears that almost all individuals whose psychopathology responds to the psychotropic medications offered have made most or all of their significant change before they ever enter the hospital for their first post-adjudication hospitalization which lasts, on average, 49.9 months.

During the remand period (remand admission to fitness admission), there appears to be little change in the level of psychopathology of the sub-sample with the exception of disturbances of consciousness and memory which can be expected to resolve partially with the passage of time. In the mental status data, all variables change significantly from fitness admission to fitness discharge. For nurses' observational data, all but one of the variables changes significantly. Between fitness discharge and OIC admission (the trial phase) almost no changes occur. While the initial hospitalization period, as reported above lasting a mean of 49.9 months, shows some degree of change in psychopathological profile (reducing the levels even further, it is noteworthy that the length of time far exceeds any needed to produce change, if it will occur, with psychotropic medication.

In our subsequent report we will analyze medication patterns in detail and will attempt to use two criteria to identify the point at which the "maximum clinical benefit" from the psychotropic medication is evident from the pattern of medication administration. These criteria are a) no critical incident that reflects a deterioration of mental status or increase in aggressive behavior and b) stability in the individual "time-series" of chlorpromazine equivalents received. We will also have data on the nature of the hospital staff's response to critical incidents in terms of medication and changes in security level.

Community adjustment and tenure

In a subsequent paper, we will present detailed analyses of the community and discharge experience of our sample. These analyses are now being completed. However, some initial data are now available on community tenure periods. Of the 124 individuals having been discharged at least once to the community, 37.1% have not experienced any significant difficulty in community tenure that led to their re-hospitalization. The remaining individuals have been re-hospitalized an average of 2.38 times. Figure 5 presents initial data on the total re-hospitalization time and outpatient time for various cohorts.

---------------------------------------------------------- Insert Figure 5 about here-------------------

The total length of re-hospitalization, averaged across all subjects experiencing re-hospitalization, lasts 518 days or 17.1 months. Subjects in a sub-sample of this cohort spent an average of 47.7 months under outpatient supervision before absolute discharge. Thus, the average individual (regardless of initial charge) is under some form of hospitalization or supervision for 49.9 months (initial hospitalization) + 17.1 months (re-hospitalization) + 47.7 months (outpatient supervision) = 114.7 months, or slightly over nine and one-half years.

Prediction of patterns

The most complex of the analyses which we are attempting is a cluster or pattern search for those individuals whose prior histories, response to treatment, and course of events while under treatment (including critical incidents) is similar. This will enable us to produce indications of what types of individuals may be processed by the system in ways that differ from the average length of hospitalization of nine years, at considerable savings without increase of risk to society. As a preliminary step, we have begun analyses of the pattern of prior psychopathological history of the individuals in our sample. This will, of course, be supplemented by their prior hospitalization and/or outpatient treatment histories, and their complete histories within the period of their adjudication and treatment prior to final recision. Table 6 presents the initial data on major aspects of the sample's

---------------------------------------------------------- Insert Table 6 about here--------------------

past psychopathological history.

Clearly this sample has had a great deal of psychopathological difficulty in the past. Almost 80% have been previously hospitalized and an additional 5.3% received outpatient treatment only. Over half the sample had prior forensic contacts, with most of those occurring within a short period of time prior to the NGRI charge. Seventy% of the sample has had prior delusional difficulties, and roughly half of those with delusions of persecution which are known, clinically, to increase the risk of aggressive behavior. Most of the sample experienced these difficulties for extended time periods prior to their offending, and the precipitating events were not of any major import when viewed from the perspective of an undisturbed individual. Significantly, the modal stressor was a drug reaction. These individuals have a high probability of self-medicating, an indication that they are attempting to solve their problems in ways that prove self-defeating, but also an indication that the treatments available to them at this point in time (prior to the charge) were not effective in controlling their level of distress. In subsequent analyses, we will be able to formulate specific patterns of prior difficulty and response to those difficulties that have policy implications for allocation of funds to the civil and forensic mental health systems.

Discussion

It is clear that the problems surrounding the insanity defense and the disposition of insanity acquittees will remain a national problem, whether or not one abolishes the plea, modifies it, experiments with a "guilty but mentally ill" alternative verdict, or modifies the structure of disposition statutes to place adjudication under court control or to alter the process in other ways. Unlike prior attempts to fashion a morally consistent and pragmatic social-legislative policy, however, we are increasingly in a position to take advantage of empirical data to inform and modify our beliefs about this problem (Keilitz, 1987; Steadman, 1987).

A large number of insanity acquittees are evaluated for trial fitness and are treated, if found unfit, before their successful insanity acquittal and subsequent re-commitment. In previous studies (reviewed by Golding & Roesch, 1987), this group was estimated at 40% of insanity acquittees. In this study we found that 60% of our sample was hospitalized and treated as unfit prior to their adjudication as NGRI. This has obvious social policy implications because of the logic behind the United States Supreme Court's recent decision in Jones v. United States (1983). The constitutionally permitted purpose of post-adjudication commitment is to treat a defendant's mental illness and protect him and society from his potential future dangerousness. Thus, the evolving standard for discharge (Golding, 1989; Hermann, 1983b) is that the individual is no longer suffering from a mental disorder likely to result in a substantial risk to the safety of society. As reported, we have rather strong evidence that this criterion occurs at a point in time quite close to discharge from unfit admission for a substantial portion of the population of defendants. Our data indicate that the major psychopathological change occurs during these fitness treatment periods, which a median of 172 days. Given what we know about probability of response to psychotropic medication, these are reasonable bounds. The largest expenditure of financial and mental health resources that could be targeted for re-allocation has to do with criminologically based periods of initial hospitalization and community supervision. As of now, we have been unable to discover any differences among the index offense groups in response to treatment, diagnostic category, or problems as outpatient. Individuals with more serious NGRI index offenses are held longer simply as a function of the index offense, as was predicted. As detailed above, most individuals have shown their major response to treatment before they ever enter the hospital for their first post-adjudication hospitalization. The average individual in our sample has a 49.9 month initial hospitalization, followed by 47.7 months of intensive outpatient supervision and 17.1 months of re-hospitalization. Thus, the average individual is under some form of secure hospitalization or supervision for 114.7 months, or slightly over nine and one-half years. Additional attention should be paid to the approximately 30% of individuals who found NGRI for quite minor offenses. Here the concern over public safety is a relatively insignificant issue, and the time lengths (even though shorter) are even more out of bounds of pragmatic or humane consideration.

Bloom and his colleagues (Bloom, Rogers & Manson, 1982) report from Oregon that, of 126 individuals released in one form or another to the community, 32% were considered "failures" in that there was some breach of the conditional release plan or a deterioration in mental condition. Bogenbeger et al. (1987), Pasewark (1986), and Steadman and Braff (1983) have all reported similar data. In our study, 62.9% of the individuals experience a "failure" of community tenure during outpatient supervision, marked by 2.4 re-hospitalizations. When the nature and causes of these re-hospitalizations are analyzed fully in our subsequent report, we expect that these data will show that these re-hospitalizations represent the natural re-cycling of the severe mental disorder from which these defendants tend to suffer. In fact, the supervision of outpatients in our sample is likely to be more intense than the Oregon system, and there is reason to believe that the re-hospitalizations are in part a reflection of this. Thus, the social and legal policy question becomes one of the wisdom of viewing these defendants as needing the more expensive intensive criminologically based supervision for what is, in fact, a mental health issue. These patients have been shown to exhibit their major response to treatment even before their initial NGRI hospitalization. The critical question, about which all others revolve, is knowledge of the response of insanity acquittees to various treatment modalities, and the peculiarities of their treatment. Our study as well as the others cited suggest that the criteria for response to treatment may not relate to ultimate "success" in the community defined from a different perspective. As much success as is possible may occur many years prior to release.

Our data on the prior mental health and forensic contacts of NGRI acquittees interacts with the treatment response question in provocative ways. Our modal NGRI acquittee was involuntary committed and discharged six months prior to their index offense. That offense and their acquittal by reason of mental illness finally became the context for a period of prolonged treatment and supervision. If this cohort had received such intensive treatment and supervision within the civil system, would the costly and criminologically based NGRI supervisory system have been necessary? Is it possible to identify such a high-risk group within the repeated contact civil committee population? The data reported here begin to suggest that such possibilities exist. Unfortunately, our study design, and the design of most forensic tracking systems (Steadman, 1987; Steadman & Morrissey, 1986) do not permit a full answer at the current time. Nevertheless, our study suggests that if such a clinically based longitudinal tracking system were routine in both the criminal and civil forensic systems, then we would be able to examine this issue.

Secure hospitalization of "not guilty" and "unfit" defendants is an expensive undertaking. In British Columbia, such hospitalization costs approximately $139.00 per day. Based on our data, with a mean initial hospitalization of 49.9 months, each insanity acquittee is held and treated initially at a (conservatively) estimated of $211,000. Homicide acquittees are held for an average of 71.9 months for an initial cost of $304,000, and they make up a substantial proportion of this group. However, homicide acquittees show no difference, as of the data yet analyzed in pattern of response or risk to the community. These estimates, however, do not consider the cost of maintaining the individual in the community under intensive outpatient supervision, the costs of re-hospitalization during the outpatient period, or the costs of psychiatric and psychological evaluation and treatment prior to the hospitalization (where our data indicate that the major observable change occurs). The average period of modification in the community is 64.8 months, of which 17.1 are spent re-hospitalized. The re-hospitalization cost is thus $72,000. The costs of estimating the outpatient treatment of these individuals is more difficult. In our subsequent report, because of the completeness of the community tenure coding forms, we will have accurate estimates of the nature of the resources (living subsidies, medical treatment, social assistance, outpatient contacts, etc.) provided to these patients. A reasonable estimate, however, is 40% of the cost of inpatient treatment or $56.00 per day. This adds another $81,000. This sums to $364,000 for an average defendant between admission as NGRI and final discharge as rescinded. The figure is $457,000 for a homicide acquittee. Moreover, these estimates disregard the average fitness remand of 59 days and the average unfit admission period of 429 days for a sizeable group of the population. If both of these occur, an additional estimate of $68,000 of expenditures will occur. Thus the typical insanity acquittee (who is charged with murder and is treated as unfit) consumes over one-half of a million dollars in financial resources under the current adjudicational scheme.

These data provide an indication of a cogent financial, as well as legal and social policy basis for reforming the adjudication of these defendants. We do not believe that these data should be used to reduce the amount of monies expended on this population. Quite the contrary, our data seem to clearly imply that this population has strong mental health treatment needs, which should be addressed in the larger mentally ill population prior to a sub-population of this group engaging in high levels of aggressive behavior which are a function of their psychopathology more than their criminality. Thus, we believe that our data argue for a re-allocation of expenditures to achieve a higher level of safety for society as well as more adequate treatment for a sub-population of the severely mentally ill.

In the our report, we expect to have a full set of analyses of causes of "community failure" that will enable us to argue that a long period of close community tenure, and a very short period of inpatient hospitalization will accomplish the same degree of treatment and safety for society. The monies "saved" should be re-allocated to other aspects of the forensic mental health delivery system that targets severely mentally disordered individuals with the kinds of delusions, hallucinations, and other problems that significantly raise the probability of their offending aggressively as a function of their mental disorder. To re-emphasize our point about the need for re-allocation, it is most that the vast majority of this group have relatively benign criminal histories, but extremely significant mental health histories with little indication of significant treatment interventions. The logical implication of these data is that the vast majority of individuals who were ultimately found NGRI for their index offenses were identified as suffering from legally relevant mental disorders prior to their index offenses. This should be the target of the re-allocation.

References

Anderberg, M. R. (1973). Cluster analysis for applications. New York: Academic Press.

Bloom, J. D., Rogers, J. L., & Manson, S. M. (1982). After Oregon's insanity defense: A comparison of conditional release and hospitalization. International Journal of Law and Psychiatry, 5, 391-402.

Bogenberger, R. P., Pasewark, R. A., Gudeman, H., & Beiber, S. L. (1987). Follow-up of insanity acquittees in Hawaii. International Journal of Law and Psychiatry, 10, 283-295.

Braff, J., Arvanites, T., & Steadman H. J. (1983). Detention

patterns of successful and unsuccessful insanity defendants.

Criminology, 21, 439-448.

Cooke, G., & Sikorski, C. R. (1974). Factors affecting length of hospitalization in persons adjudicated not guilty by reason of insanity. Bulletin of the American Academy of Psychiatry and the Law, 2 , 251-261.

Golding, S. L. (in preparation). The adjudication of criminal responsibility: A review of theory and research. In D. K. Kagehiro and W. S. Laufer (Eds.), Handbook of psychology and law. New York : Springer-Verlag.

Golding, S. L., & Roesch, R. (1987). The assessment of criminal responsibility : A historical approach to a current controversy. In I. B. Weiner & A. K. Hess (Eds.), Handbook of forensic psychology (pp. 378-394). New York : Wiley.

Golding, S. L., Roesch, R., & Schreiber, J. (1984). Assessment and conceptualization of competency to stand trial: Preliminary data on the Interdisciplinary Fitness Interview.

Law and Human Behavior, 8, 321 -334.

Greenland, C. (1979). Crime and the insanity defense, An international comparison: Ontario and New York state. Bulletin of the American Academy of Psychiatry and the Law, 7, 125 -138.

Hans, V. P. (1986). An analysis of public attitudes toward the insanity defense. Criminology, 24, 393-414.

Hermann, D. H. J. (1983a). The insanity defense : Philosphical, historical and legal perspectives Springfield, Ill : Charles Thomas.

Hermann, D. H. J. (1983b). Automatic commitment and release of insanity acquittees. Rutgers Law Journal, 14, 667-682.

Hodgins, S. (1983). A follow-up study of persons found incompetent to stand trial and/or not guilty by reason of insanity in Quebec. International Journal of Psychiatry and Law, 6, 399-411.

Janes, C. L. (1979). Agreement measurement and the judgment process. Journal of Nervous and Mental Disease, 167, 343-347.

Jones v. United States. 463 U.S. 354 (1983).

Lorr, M., & Veste, N. D. (1968). Psychotic inpatient profile. Los Angeles, Ca. : Western Psychological Services

Maxwell, A. E, & Pilliner, A. E. G. (l968). Deriving coefficients of reliability and agreement for ratings. British Journal of Mathematical and Statistical Psychology, 21, 105-116.

Menzies, R. J. (1987). Cycles of control: the trancarceral careers of forensic patients. International Journal of Law and Psychiatry, 10, 233-249.

Monahan, J. (1978). The prediction and control of violent behavior. United States House of Representatives Committee on Science and Technology. Subcommittee on Domestic and International Scientific Planning, Analysis and Cooperation, Rep. James Scheuer, Chair January 10, 1978, New York, New York

Monahan, J. (1981). The clinical prediction of violent behavior. Rockville, Md. : National Institutes of Mental Health.

Monahan, J., & Steadman, H. J. (1983). Crime and mental disorder: An epidemiological approach. Crime and Justice: An annual review of research, 4, 145 -189.

Pantle, M. L., Pasewark, R. A., & Steadman H. (1980). Comparing institutionalization periods and subsequent arrests of insanity acquittees and convicted felons. Journal of Psychiatry and Law, 8, 305-316.

Pasewark, R. (1986). A review of research on the insanity defense. In S. A. Shah (Ed.), The law and mental health : Annals of the American Academy of Political and Social Science, 484, 100-114.

Pasewark, R. A. (1982). Insanity plea: A review of the research literature. Journal of Psychiatry and Law, 1982, 357-402.

Pasewark, R. A., Bieber, S., Bosten, K. J., Kiser, M., & Steadman, H. J. (1982). Criminal recidivism among insanity acquittees. International Journal of Law and Psychiatry, 5, 365-375.

Pasewark, R. A., Pantle, M. L., & Steadman, H J. (1979a). The insanity plea in New York State, 1965-1976. New York State Bar Journal, 53, 186 -225.

Pasewark, R. A., Pantle, M. L., & Steadman, H J. (1979b). Characteristics and disposition of persons found Not Guilty by Reason of Insanity in New York State, 1971-1976. American Journal of Psychiatry, 136, 655-660.

Petrila, J., Selle, J., & Rouse, P. C. (1981).

The pre-trial examination process in Missouri:a descriptive study. Bulletin of the American Academy of Psychiatry and the Law, 9(1), 60-85.

Phillips, B. L., & Pasewark, R. A. (1980). Insanity plea in Connecticut. Bulletin of the American Academy of Psychiatry and the Law, 8, 335-344.

Quinsey, V. L., Pruesse, M., & Fernley, R. (1975). Patients found Unfit to Stand Trial or Not Guilty because of insanity. Canadian Psychiatric Association Journal, 20, 461 -467.

Rabkin, J. G. (1979). Criminal behavior of discharged mental patients: A critical appraisal of the research. Psychological Bulletin, 86, 1 -27.

Roesch, R., & Golding, S. L. (1985). The impact of deinstitutionalization. In D. P. Farrington and J. Gunn (Eds.), Current research in forensic psychiatry and psychology: Aggression and dangerousness (pp. 209-239). Chichester, U. K. : Wiley.

Roesch, R., Golding, S. L. (1980). Competency to stand trial. Urbana, Ill: University of Illinois Press

Shah, S. A. (1986). Criminal responsibility.In W. J. Curran, A. L. M. McGarry and S. A. Shah (Eds.), Forensic psychiatry and psychology: Perspectives and standards for interdisciplinary practice. Philadelphia, PA : Davis.

Steadman, H. J. (1987). Mental health law and the criminal offender: research directions for the 1990's. Rutgers Law Review, 39, 323-337.

Steadman, H. J., & Braff, J. (1983). Defendants found not

guilty by reason of insanity. In In J. Monahan & H. Steadman

(Eds.), Mentally disordered offenders: Perspectives from law and

social science New York : Plenum

Steadman, H. J., & Cocozza, J. (1978). Psychiatry, dangerousness and the repetitively violent offender. Journal of Criminal Law and Criminology, 69, 226 -231.

Steadman, H. J., Keitner, L., Braff, J., & Arvanites, T. M. (1983). Factors associated with a successful insanity plea. American Journal of Psychiatry, 140, 401-405.

Teplin, L. A. (1983). The criminalization of the mentally ill: Speculation in search of data. Psychological Bulletin, 94, 54 -67.

Walker, N. (1968). Crime and insanity in England, Vol. 1. Edinburgh, Scot. : Edinburgh University Press.

Winick, B. J. (1987). Incompetency to stand trial: An assessment of costs and benefits, and a proposal for reform. Rutgers Law Review, 39, 243-287.

Footnotes