Behavioral Science and the Secret Service: Toward the Prevention of Assassination (1981)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

PROBLEMS IN ASSESSING AND MANAGING DANGEROUS BEHAVIOR Shervert H. Frazier, M.D.* Psychiatrist-in-Chief McLean Hospital Belmont, Massachusetts McLean Hospital is a private non-profit psychiatric hospital affiliated with the Massachusetts General Hospital and is a teaching hospital of the Harvard Medical School. Founded in the 19th Century, the hospital is noted for providing evaluation and treatment for severely disturbed men and women, especially those who need secure and comprehensive inpatient care. In 1975, the Massachusetts Department of Correction initiated an innovative contract with McLean Hospital under which McLean would be responsible for providing psychiatric and psychological expertise in the evaluation and treatment of male patients at the Massachusetts maximum security psychiatric facility, the Bridgewater State Hospital for the Criminally Insane. Since 1975, McLean clinicians have been evaluating and treating the most psychotic and violent men in Massachusetts. Currently, a staff of 16 McLean psychiatrists, psychologists, and social workers provide mental health services at Bridgewater. Recently the Department of Correction expanded the contract so that McLean clinicians now provide psychiatric services throughout the state correctional system. The United States Secret Service (USSS) requested that the National Academy of Sciences convene a conference on the question of how behavioral research can aid the Secret Service to protect governmental leaders. As a result, a team of McLean/Bridgewater psychologists and psychiatrists considered these questions. The Secret Service seeks further knowledge to aid in its identification, assessment, and management of persons who might be dangerous to those it protects. Specifically, the USSS asks about objective indicators of "dangerousness," decision-making priorities for assessing an *In collaboration with Park E. Dietz, M.D., M.P.H.; Ronald S. Ebert, Ph.D.; Sara Eddy, Ed.D.; Robert A. Fein, Ph.D.; William S. James, M.D.; Wesley E. Profit, Ph.D.; Renatt C. Wack, Dipl. Psych., M.P.H., all of the McLean/Bridgewater Program, McLean Hospital, Belmont, Massachusetts. 93

individual's dangerousness, interviewing styles and strategies that are effective with individuals who threaten the protected, the relationship of "dangerousness" to social factors such as group affiliations, family, and local or national events, the management of individuals who are seen as dangerous, and improvement of relationships between the Secret Service and mental health and behavioral science professionals. To consider these questions, we reviewed thoroughly a substantial literature on "dangerousness" and the few rare articles on assassins. We met with Secret Service agents from Washington and Boston. We evaluated our experience in assessing, managing, and treating men and women who assault and murder. Although the literature on "dangerousness" is substantial, the concept of dangerousness is still not adequately defined. Because anyone can be dangerous, the Secret Service is concerned only with persons whose specific aim is to harm those it protects. It works to identify, assess, and manage the small particular subset of people who might act on thoughts or impulses they have about injuring or killing government leaders and/or their families. Secret Service agents in Boston report that increasing numbers of mentally disordered individuals threaten the president primarily to get attention and care. These people might or might not be dangerous to a protected person and often need sophisticated assessment by mental health professionals. The demands on the Secret Service are growing. "Deinstitutionalization" of large, antiquated state mental hospitals has sent thousands of mentally disordered men and women into the community. However laudable the goals of deinstitutionalization, it is clear that mental health services (and controls) for the mentally ill have changed dramatically. In Massachusetts, police, judges, and criminal justice agencies report that increasing numbers of mentally ill persons are arrested for crimes of all kinds, minor and major. The identification, assessment, and management of individuals who threaten government leaders is a difficult and complex task. In this report, we present factors we consider in determining whether a person warrants detailed clinical evaluation for dangerousness. We describe the outline of a detailed assessment. We discuss the selection of Secret Service agents, the composition of agent corps and squads, the need for male-female interviewing teams, the support of agents, liaison with mental health and behavioral science professionals, and the collection, storage, and analysis of data relevant to the Secret Service's mission. We include a memorandum on potential female assassins and a set of recommendations. 94

I. Identification of Potential Assassins In our work with psychotic assailants and murderers of both the famous and the not-so-famous, we have identified some criteria we use when considering an individual's likelihood of future violent behavior. We are aware that no valid, reliable, objective criteria exist which will definitively predict an individual's "dangerousness." We have, however, identified several factors that could lead us to do a preliminary assessment of such an individual. We consider the following factors: A. History of mental illness 1. Has he/she a history of acute psychosis, especially with paranoid symptoms? 2. Has he/she a history of suicide attempts? 3. Has he/she a history of delusional ideas in which there is has a sense of mission or of being extraordinarily special, such as receiving messages from God? 4. Has he/she heard voices that command to behave in a directive fashion? B. Present mental illness (see above symptoms) C. Relationships with others 1. Does he/she have a history of being a loner? 2. Does he/she have any sustaining ties with other persons? 3. Has he/she experienced the recent loss of an important relationship? 4. Is he/she affiliated with militant political groups or "fringe" religious sects? D. Perceived grievance that has recently increased in significance. E. History of attempts to contact a protected person F. Current mental and physical capacity to develop a plan and act in a sustained, organized fashion to carry out that plan. II. Assessment of a Potential Assassin If the above preliminary assessment suggests that a more formal assessment is in order, a variety of other factors should be considered. Long-range prediction of what a particular individual will do may not be possible; it is possible, however, to anticipate short-range behavior in specific cases. Rather than think about 95

dangerous people, we prefer to think about dangerous situations—situations involving a specific subject, a victim, and an act under specific circumstances. Therefore, our comprehensive evaluation attempts to ask questions about these factors; moreover, we have added items particular to identifying persons who might be dangerous to a president. A. History and background of subject 1. Family history a. Relationships with parents and siblings b. Relationship of parents to one another and to the family c. Family history of mental illness or violence d. History of subject's having been beaten or abused as a child e. History of subject's having had physical fights with parents f . Cultural background of subject and significance and importance of violence and/or assassination in that culture. 2. Psychiatric history a. Attitude toward illness and helping agencies b. Wish to return to the hospital 3. Present symptoms of mental illness 4. History of suicide attempts 5. Sexual history 6. Criminal history B. Action-producing conditions 1. Episodes of excitement, accompanied by uncontrolled motor activity 2. Delusions of having special powers or missions 3. Voices commanding subject to behave in a particular fashion 4. Delusions of transmitting and/or receiving devices having been implanted in one's brain 5. Impulsivity a. History of accidents, automobile violations, bedwetting, fire setting b. History of antisocial behavior 6. Substance abuse (alcohol, amphetamines, PCP [angel dust], barbiturates or "downers") 7. Organic problems a. Temporal lobe epilepsy 8. Episodes in which person feels separated from his/her actions ("It was like watching myself in a movie.") 9. Existence of a delusional world inhabited by threatening persons 96

C. Interpersonal relationships 1. History of only very few close friends, if any 2. Inability to care for others 3. Recent rejection by or loss of a loved one 4. Perceived rejections and betrayals by persons in governmental organizations 5. Difficulty at work with co-workers and/or superiors D. Means 1. Ability to travel to the protected person 2. Ability to form and carry out a plan 3. Easy access to a weapon E. Setting 1. Recent, well-publicized violent acts in the same geographic area as the protected person F. Current life events 1. Present stresses (social, economic, interpersonal, job-related) G. Action 1. Threatening letter(s) from subject to president or to Secret Service 2. Visit from subject to White House or approach to president An interview is the means by which clinicians and Secret Service agents obtain data about subjects; we have some specific recommendations about the setting, the interview itself, and the interview team. Setting Agents who visit a subject at his/her place of residence will have an opportunity to see the living situation. Agents can sometimes make inferences about a subject's psychological state by seeing his/her personal environment. Such information can be quite valuable. However, we suggest that all interviews after the initial interview take place in the agents' offices. A setting more familiar to the agents, such as their offices, will allow them to be more comfortable and better able to elicit information from the subject. Whether the interview takes place at the subject's home or the agent's office, both agents and subjects should sit with equal access to doorways and exits. Agents should not sit between the subject and the exit, and should be careful to stay within the subject's view. 97

Conducting the interview The agents first should address the subject's interests and complaints. They will obtain more information if they agree with the subject's complaints and are appropriately sympathetic; however, they should avoid entering into the subject's psychosis or delusional system. They should follow the subject's verbal leads, listening but asking few questions. At the conclusion of the interview, direct questions about the subject's threats, intentions, plans, and means for carrying out the threats are important. Confronting and "stressing" the subject are generally unproductive; such techniques should be used only late in the interview and with the purpose of eliciting more information about the subject's thoughts and feelings about the protected person. The agents should be available to help the subject obtain assistance for his/her complaints, and may need to arrange for suitable psychiatric/psychological treatment. Following the interview, agents should record their impressions of the subject, paying particular attention to their own thoughts and feelings about the individual. These impressions are extremely important in drawing conclusions about that individual's potential dangerousness. Personnel Male-female interview teams will provide substantially increased cooperation from the subject and will lead to richer data. We recommend that each subject, male or female, be interviewed by such a team. Consultation with mental health professionals can help these agents to develop the techniques of a team approach. III. Selection and Training of Agents If we generalize from the agents we have met, we would conclude that the Secret Service chooses individuals who are likely to be effective in assessing subjects. In our view, an agent should be someone who is stable, relates well to others, has friends, possesses some self-knowledge, and is capable of tolerating and dealing with anxiety. He or she should be flexible in solving problems, curious about what makes people tick, and basically empathic. It is important that agents be people who can be appropriately suspicious and vigilant. The agent must be able to see his/her work as professional, having clear ideas about the job and keeping a sense of perspective. Moreover, agents should be able to think and respond quickly. In our view, an agent should be somewhat compulsive, worrying about doing well enough, never too satisfied with yesterday's accomplishments. 98

Training for agents should include didactic education concerning mental illness. The content of training should include information about psychosis, affective disorders, character pathology, organic pathology, and drug and alcohol abuse. But no amount of didactic training can substitute for supervised, direct contact with mentally ill people. Ways of training agents to assess mentally ill subjects include the following. Hospital-based training Agents can participate in courses, interviews, and other clinical interventions to acquire basic diagnostic and interviewing skills. Training Academy-based instruction Agents participate in time-limited workshops offered by mental health professionals. Workshops should be designed both for new and for more experienced agents. On-the-job training Clinicians can join agents on occasional assignments and provide assistance, supervision, and consultation. The educational objectives of hospital-based training are to provide the agent with: A. basic knowledge of different forms of psychopathology associated with unlawful behavior B. basic knowledge and skills in clinical interviewing and intervention C. understanding of the effects of mental illness on social and intellectual functioning D. basic knowledge of factors causing impulse dyscontrol. Agents should attend a seminar on mental health, mental illness, and dangerousness, observe clinical interviews, perform supervised clinical interviews, observe court-ordered evaluations of patients who have acted violently, and work on treatments units. We suggest that hospital-based training occur in maximum security psychiatric facilities, which exist in most states, such as Bridgewater State Hospital in Massachusetts. 99

IV. Composition of Agents Corps, Squads, and Interviewing Teams The ethnic, racial, and gender composition of the agent corps should be a key issue for the Secret Service. Same-sex and same-race interviewers are likely to obtain different information (often more comprehensive and accurate) from subjects than are-opposite sex and other-race interviewers. Cross-racial and cross-sex squads and interviewing teams provide maximum flexibility with subjects, particularly with paranoid persons. Some women will talk more freely to a female agent; some women will talk more freely with a male agent. The same is true for male subjects. The Secret Service should consider how to recruit, select, train, and keep sufficient numbers of female, black, and Hispanic agents to meet the criteria a bove. V. Support of Agents There is no question but that agents have exceedingly difficult and stressful jobs. As one agent explained to us, "If somebody calls the White House and threatens the president, we can get a call within minutes. We may have to go out immediately, interview the subject, talk with whatever family members we can find, and, within a matter of hours, report to Washington about whether the subject is dangerous to the president." Agents provide peer group consultation and support for each other. Currently, there are too few opportunities for agents to consult with mental health professionals who are experienced at assessing and managing dangerous patients. If available, such consultation could: A. assist in quality control by adding a clinical voice to the decision-making process B. reduce some of the stress of the agent's job by decreasing the feeling (not necessarily valid) that one is making decisions outside one's training and expertise C. provide up-to-date knowledge for agents about local mental health resources D. build a network of mental health professionals who might hear of individuals who should come to the attention of the Secret Service We propose that the Secret Service contract with outside agencies for mental health resource groups. Such groups could provide important consultation to the Secret Service in interviewing, diagnosing, assessing dangerousness, developing dispositional alternatives, training, and supervision. Members of a mental health 100

resource group would be available to agents on an as-needed basis. Need would be determined by the Secret Service, whose agents would initiate contact with the resource group. Ten to twelve mental health resource groups would function as a national system. Each would be based in a major population center and would work with the local Secret Service office. Each group should be based in a mental health teaching institution. Mental health and behavioral science professionals with experience with violent individuals would be available for around-the-clock consultation. Resource groups should have adequate female and minority representation. Groups should consist of sufficient numbers of persons to allow for reasonable sharing of responsibility, representation of race and sex, and speed of movement. The broader the geographical area to be covered, the larger the group needs to be. Resource group members should receive an initial orientation through the Secret Service and should meet together on an infrequent basis (perhaps no more than once a year) to discuss techniques, strategies, problems, new knowledge, and experiences. VI. A National Mental Health and Behavioral Science Advisory Group In recent years the Secret Service has, by necessity, become a front-line mental health screening group. Much of the work of agents involves interviewing, assessing, and attempting to find treatment for individuals who are mentally ill and who lack adequate treatment. The current national policy of deinstitutionalization has placed many mentally ill individuals with histories of dangerous behavior into the community. Many of these persons, in the past, would have been effectively contained in state-supported mental hospitals. These patients often are difficult to treat. They may be reluctant to accept needed medication. Frequently they have neither treatment nor social supports. Our experience suggests that this population will more and more frequently come to the attention of the Secret Service. We think Massachusetts' experience can be generalized to the nation. We propose, therefore, that the Secret Service appoint a national mental health and behavioral science advisory group. This group should meet regularly with senior Secret Service staff to advise the Service on mental health issues and policies that affect its work, particularly policies concerning staff training and education. It should keep the Service abreast of relevant recent research and should be a sounding board for the agency's internal research and assessment efforts. It would stand as a symbol of the Service's willingness to maintain contact with mental health and behavioral science professionals. A national advisory group should consider recommendations for changes in national and state policies 101

concerning individuals who are convicted of threatening Secret Service protectees. It should advise the Service on how best to share information with state mental health officials about individuals who come to the Service's attention and should aid the Service in formulating policies on information-sharing. VII. Storage and Analysis of Data Relevant to the Service's Function Although we are not intimately familiar with the Service's current method of organizing the information its agents collect, we have been told that there is need for change. Agents report that several years ago, when the agent pool was much smaller, one knew most other agents and could gather and interpret information on a personal knowledge basis. With the Service's dramatic growth in the past ten to fifteen years, such personalized data assessment is not possible. Further, while the Service has thousands of files with information about subjects who have come to its attention, there has been no systematic effort to analyze these data in ways which might improve the Service's ability to find, assess, and manage such subjects. We recommend that the Secret Service develop a standard investigation record and computer capability for making records accessible to agents. We suggest that the Service ask its consultants to develop a research plan which would utilize both its internal data and data available from other sources. The Service should have both internal research'capability and the ability to contract for specific research. We have outlined one model for a standardized investigation record and have included it in the Appendix. This record would organize information directly related to the Service's goals of assessment and management. The standardized investigation record should have the following elements: A. pre-coded format to allow efficient computer entry and analysis B. narrative entry spaces to permit concise verbal accounts necessary for reporting details, exceptions, qualifiers, and other important information not conducive to pre-coding C. ease of operation D. potential for periodic updating E. data source coding F. data verification coding. Properly pre-coded, standard investigation record data could be entered on a computer within hours to days of collection, thereby quickly becoming available to Secret Service agents worldwide. Such 102

data can be gathered for each investigator to identify blind spots (such as missing data in the same categories too often) or bias (such as dispositions disproportionately correlated with one variable). These data, computer stored, could be a permanent and cumulative record of Secret Service investigations. VIII. Memorandum on Potential Female Assassins An interdisciplinary group of professionals met at McLean Hospital to discuss women as potential presidential assassins. Past efforts have focused on developing "psychological profiles" only of male assassins and threateners; our preliminary search of the literature has neither yielded such information about women nor references to the three women who have recently made attempts to kill a president. We know that the Secret Service must, by federal statute, investigate each threat upon the president and be concerned with identifying those persons and situations which are potentially dangerous. We believe, however, that a "psychological profile" of the potential female presidential assassin would not be useful; it would lead to overprediction, which is fruitless. Also in our consideration is the fact that when women murder, they are more likely than men to kill their spouses. Almost half of their victims are spouses, and nearly all are family, friends, and acquaintances. Women kill people with whom they have a relationship. Although few women actually have a relationship with a president, he is not necessarily safe from attempts on his life by women. A woman may act in a dangerous fashion toward a president on behalf of a man with whom she has a relationship and/or on the basis of her fantasied or delusional relationship with the president. When interviewing women suspects, Secret Service agents need to question these women about their relationships. They should focus on the woman's past and present family and love relationships and on her current living situation. Has she or a loved one recently experienced losses, rejections, financial or health problems for which she might hold a president responsible? The interviewers will also need to examine her real and/or unreal notions about her relationship with a president, and learn whether she links it to other important relationships from her past. These interviews should be conducted in settings that are relatively free from tension or fear. A woman subject will generally be more comfortable, less guarded and suspicious, and more likely to disclose private thoughts and fantasies about men and about a president if she is interviewed by a woman agent; some women subjects, however, will respond more fully and frankly to a male interviewer. Because of this, each female subject should be 103

interviewed by a man-woman team. These interview teams will provide substantially greater opportunities for an agent to use his/her particular skill, will elicit more productive data, and will greatly enhance the process of gathering clinical information for subsequent decision-making. To implement these recommendations, the Secret Service will need to recruit sufficient numbers of female agents so that the teams can be developed. Frequent consultation and case review with a male-female mental health team will be helpful to the agents in maximizing the unique contribution of each agent while enabling them to work smoothly together. Because women murder within the context of a situation and a relationship, the Secret Service should learn more about the target a woman selects. It is likely that some women may be dangerous not to presidents in general but to a president in particular. Since the actual numbers of women who have attempted to kill a president are so few, we suggest that the Secret Service examine attempts upon and threats toward other prominent public figures by women, followed by an investigation of threats toward and/or attempts upon the lives of congressmen, senators, and governors across the country. The investigation would involve interviewing women threateners and learning about their real and/or fantasied relationships with their targets, researching the behavior of the targets, and reviewing the characteristics of those prominent public figures who have never been threatened. A comprehensive investigation of women and their targets will yield important information for use by the Secret Service. IX. Recommendations It is recommended that the Secret Service 1. expand its training of agents with curricula to include hospital-based training in understanding mental illness and more knowledge of clinical interviewing 2. develop teams of male and female interviewers for better understanding of women who threaten a protected person and for better understanding of male paranoid subjects 3. increase the number of minority and female agents for purposes of expanding interview techniques and to communicate better with minority subjects who threaten a protected person 4. develop mental health resource groups for consultation, disposition of threateners, training and supervision in the educational process of agents 104

5. develop a national mental health and behavioral science advisory group for study of mental health issues and policies to keep the Secret Service abreast of recent research data, and to recommend pertinent new research 6. develop a standard investigational record computer capability 7. improve research capability. APPENDIX Standardized Investigation Record A. Administrative (process) data, such as 1. identifiers (name, social security number, identifying marks, aliases, date of birth) 2. demographic data (age, race, sex, social class) 3. address, phone numbers 4. schools attended 5. jobs held 6. relatives' names and locations 7. contacts' names and locations 8. stage of investigation 9. investigating agents 10. dispositions 11. initiation of report 12. type of target of threats B. Personal and family history data, such as 1. exposure to violence in the home 2. childhood behavioral patterns 3. adolescent behavioral patterns 4. personal psychiatric history 5. personal medical history 6. family stability data 7. family medical history 8. family psychiatric history 9. family criminal history C. Diagnostic category 1. antisocial personality disorder 2. schizoid personality disorder 3. paranoid personality disorder 4. narcissistic personality disorder 105

5. borderline personality disorder 6. sadism 7. pyromania 8. explosive disorder 9. obsessive-compulsive disorder 10. post-traumatic stress disorder 11. manic'episode 12. major depressive episode 13. paranoid disorder 14. schizophrenia D. Criminal history data including 1. juvenile record 2. school delinquency history 3. military disciplinary history 4. arrest histories (state, federal, Interpol) 5. conviction history 6. correctional history E. Substance use and abuse history, including 1. alcohol 2. opiates 3. stimulants 4. sedative-hypnotics 5. hallucinogens F. Injustice history, including 1. civil suit history 2. complaining letter history 3. complaining phone call history 4. perceived injustice G. Armament inventory, including 1. knowledge of firearms and explosives 2. books and manuals obtained from libraries 3. magazines, books and manuals purchased 4. subscriptions 5. imagery of pictures owned and displayed 6. weapons owned 7. firearm registration and licensure 8. access to explosives and munitions H. Situational variables, including 1. recent life stress 2. recent substance use 3. recent mood cycles 4. geographic drifting or movement 5. access to information on government leaders 6. recent absenteeism from work 7. current social network 8. recent help-seeking behavior 106

NOTES ON THE SECRET SERVICE AND THE PROTECTION OF THE PRESIDENT Robert Mlchels, M.D.* Barklie McKee Henry Professor of Psychiatry Cornell University Medical College Psychiatrist-in-Chief The New York Hospital New York, New York I. The protection of the president, his family, and a number of other designated persons is a vital function of the Secret Service. The conceptualization of this task appears to have evolved since its initial formulation, starting with the immediate personal protection, or "bodyguard" function, and then expanding to include the response to threats and the screening and management of threatening persons. These latter activities now involve a considerable amount of Secret Service time and effort. However, before discussing this task and some possible approaches to improving its efficacy and efficiency, it might be interesting to speculate about the future possibilities in the evolution of the Service's mission. If the first phase was immediate protection and the second the screening and management of threats, a plausible third phase would be the identification of possible sources of danger and their evaluation. This would represent a continuation of the shift from a passive to a reactive and then to a proactive stance, with the task defined as maximizing the security of the person being protected rather than evaluating dangerous persons that come to the attention of the Service or thwarting actual attacks. Metaphorically, this would be analogous to a shift from first aid to the treatment of disease, and then to the public health concern with the cause and prevention of illness. It is possible that, as in the medical metaphor, the major threats to the president's security are not associated with those persons who come to the Service's attention, but rather with those who are not "referred" by the current network, and that an active program of identifying potential threats would be more productive than improved screening and evaluation of those threats that are now detected. We will return to this issue later. II. Focusing for the moment on the current concept of the Service's mission, there are three functions that will be discussed: (1) the Commissioned paper 107

delineation of specific tasks and the formulation of job descriptions; (2) selection, retention, and promotion procedures for agents; and (3) training methods and curricula. The question of tasks and job descriptions arises because of the extremely varied and difficult tasks assigned to the Service. At various times agents must physically protect the president and others; interview and evaluate mentally ill individuals; maintain liaison with police, mental health institutions, the penal system, and others; and so forth. These tasks are sufficiently heterogeneous that the Service must decide whether its work will best be accomplished by "generalists" or "specialists," or, more likely, what should be the optimal ratio between the two. The answer to this question will, of course, change as the activities of the Service evolve, and care must be taken that the answer be responsive to the current definition of the Service's tasks, rather than reflecting an earlier, narrower concept or, conversely, anticipating future developments that have not yet occurred. The question can be illustrated by considering the advantages for the screening function of having agents with a wide variety of personal backgrounds and interests, and with extensive training in psychosocial and interpersonal skills, while the direct protective function would clearly lead to quite different priorities in selection. It is undoubtedly desirable to have a significant number of agents who combine the capacities necessary for both functions, but it might limit the excellence of the program if the roles were formulated in a totally "homogenized" way. Shifting from job description to recruitment and selection, most of the experience of others in the selection and training of persons who are to interview, evaluate, or manage the mentally ill would suggest that recruitment and selection are at least as important as training. We are much more successful in identifying persons who have interpersonal sensitivity, empathy, or psychological mindedness than in training those who do not possess them. Put in other terms, training enhances these traits in those who already have them, but it has little impact on those who are initially low in them. Most persons who are experienced in teaching psychiatric interviewing place more faith in selecting the trainees than in constructing their curriculum. Psychologic and psychiatric interviews, group training exercises, assessment of psychosocial skills, and psychodiagnostic procedures should all be considered as selection tools. They would also provide the basis for an important future research project—the development of predictive criteria for identifying potential agents who will perform well or poorly at various tasks—and, therefore, for selecting future agents and assigning them to specific tasks. Indeed, if at present one agent were discovered to be particularly good or bad at screening potentially dangerous persons, it would be difficult to generalize this information usefully unless there were descriptive measures that would relate him to other agents in some consistent way. One might 108

try to study what he did, rather than who he was, but the nature of the task is such that the study of "what" without "who" would be of limited value. Training procedures will be discussed in three categories: education in psychologic and psychiatric knowledge; the development of skills in interviewing, assessment, and evaluation; and programs of continuing education. Although there is little certainty concerning what kind of knowledge would be most relevant, there would probably be reasonable consensus about where to start. Persons who will be evaluating the mentally disturbed should have some basic exposure to psychopathology, the major psychiatric syndromes, and simple principles of psychodynamics, with particular emphasis on factors related to impulsive or violent behavior—organic disturbances, sociopathy, manic, paranoid, and acute psychotic states, drug-related states, and so forth. They would also require knowledge concerning those conditions in which questions of impulsive behavior or violence might enter the differential diagnosis— obsessional or panic episodes, fugue states, and so forth. They would learn about suicide, homicide, violence, criminal behavior, and the group, social, and cultural aspects of such behavior. They would learn about the social networks in which persons with these problems exist—family and community, treatment institutions, penal institutions, and also about the various caretakers and professionals with whom they have contact. They would visit prisons and mental hospitals, and learn what they are like. They would learn about treatment programs, and the natural history of the major disorders, with particular emphasis on the persistence of paranoid fantasies in covert form. Finally, they would ledrn about the theoretical basis of interviewing—transference, defenses, projection, and so on. Simultaneous with these educational experiences would be training exercises involving exposure to real subjects, similar to those who will later be evaluated, practice interviews, and so forth. It is essential that some of these exercises be conducted with real people, not actors simulating subjects or videotapes, although both of these may be useful adjuncts. The emotional stress of an interview on the interviewer cannot be replicated without involvement with real subjects. These interviewing experiences coulc take place in a variety of settings. One possibility would be a practicum, perhaps similar to the exposure of medical students on a psychiatry clerkship; another would be contact with selected patients, perhaps institutionalized, who are drawn from the population that will later be the focus of interest; a third would b, accompanying experienced agents and participating in tandem with thei as trainees. There would probably be an advantage to combining these, allowing the trainee a number of different views of the population under consideration. 109

An incidental problem is that of access to the subject populations necessary for training. I would see no ethical problem in providing this access. The difficulties that have been encountered might represent negative attitudes toward the Service among the various professions and disciplines who care for the subjects in institutional and other settings. If this is true, it probably has ramifications far beyond the question of training, which will be explored in more detail below. Some of the obstacles might be overcome by compensating volunteer subjects drawn from the population in question, a procedure that is consistent with ethical practice in the health professions. Any comprehensive approach to training must include some type of continuing training, involving agents after they have completed their initial educational experiences. This could be in the form of periodic "brush-up" courses, ongoing clinical experiences while working in the field, visiting consultants, and so forth. There have been a sufficient number of new developments in our knowledge of psychopathology in the past decade to suggest that without such continuing input the effectiveness of agents in the field is likely to decay significantly over time. III. A program of this scope and importance obviously requires ongoing development and evaluation. However, there are special difficulties here. One is attempting to prevent what is already a rare event, therefore, it is difficult to evaluate the effect of the prevention program. However, without some version of such an evaluation, there is no way to improve the program or to determine how many resources should be devoted to it. Therefore, an immediate problem concerns what types of partial evaluations or analog systems for evaluation can be developed. The performance of individual agents can be evaluated by the use of standardized video interviews, direct observation in the field, or the use of actors unknown to the agent who simulate subjects in random field evaluations. These methods would determine whether agents are doing what they "should" be doing, probably to be defined as what experts and experienced agents think that they ought to be doing. It would not determine whether that consensus is valid, whether what they "should" be doing is in fact effective. A second component that could be evaluated involves the selection and training procedures. The agents who emerge from given selection and training processes could be tested and compared with a control group selected by different criteria or trained by different methods. Systematic differences could then help in planning future selection criteria or curricula. 110

A third type of evaluation would involve the methods and criteria used in assessing subjects in the field. It should be relatively easy to establish reliability of various procedures, measure their redundancy, and so forth. Here again, the most difficult problem would be validity, and it is difficult to improve a process if one cannot determine how well it is working. However, it may still be possible to increase its efficiency, and probably to detect certain procedures that miss information that seems to have face validity. For example, if it were determined that challenging or confronting interviews with paranoid persons lead to an under-reporting of previous acts of violence, this would be of interest even if it were not possible to demonstrate that these persons are subsequently involved in additional violent acts. Analog systems for testing procedures might be developed from the study of homicide, suicide, violent offenders, or terrorism. Each of these has problems, but each might provide an opportunity for studying various aspects of the prediction involved—impulse control, the specificity of targets, the time course of risk, psychosocial and stress "triggers," and so forth. For example, knowledge concerning what type of material in the mass media precipitated flurries of social violence might alert one to a subject who reported recent exposure to such material, while measures of impulse control that failed to predict when chronically depressed individuals made suicide attempts might be considered suspect in their ability to predict other types of violence. IV. Finally, I will return to my initial comments, the potential shift from responding to threats to identifying potential sources of danger. In doing so, I will present a relatively radical proposal, one that I myself respond to rather negatively, but one that seems sufficiently interesting to warrant discussion. At present, although most of the "referrals" to the Service who are eventually considered to be significant risks are persons who have or have had contact with the mental health system, there are few referrals from the mental health system itself. If this system knows, cares for, and follows the same persons who are seen and evaluated by the Service, it is interesting to consider why there is so little interaction between the two. It is, of course, possible that "referrals" from the mental health system would be of little value, but even this could not be established without study. It is also possible that such referrals would overrun the capacity of the Service. If such referrals were of little relevance, the easy solution would be to ignore them. However, if they were comparable to those now seen by the Service, but represented a significantly larger number of subjects, it would raise questions about the value of current procedures. It would be possible to establish answers to such questions by studying 111

relatively small communities and extrapolating to the rest of the country. What factors deter the mental health system from making such "referrals"? One involves the ethical problems relating to confidentiality, though most mental health professionals believe that there is no major ethical dilemma if it is in the patient's interest to violate his confidentiality, and that it is generally in the patient's (as well as society's) interest to prevent a major crime of violence. While there may be no basic ethical issue here, there may be a problem of consciousness-raising, one that might be solved by statements from leading figures in the profession known to be interested in such issues. Secret Service policies that protect the confidentiality of information from informants, especially if it could be detrimental to the subject, might be helpful. Related factors are the actual or perceived legal barriers to such "referrals." Professionals might fear that their patients will have legal claims against them if the patients' identities are disclosed to the Secret Service. The confidentiality and security of the Secret Service record system would be relevant here. In addition, legislation limiting such claims might be effective. Probably the most important factor relates to prejudices and attitudes concerning the Secret Service. If the Service is seen as a police agency, it will be regarded with skepticism by mental health professionals; if it is viewed as concerned and competent in regard to the needs of those it evaluates, as well as those it protects—in effect as a kind of social service agency—the attitudes of mental health professionals might change. It is conceivable that the most critical information regarding potential danger to the president and related individuals is located in the mental health system, and that the most modifiable step in the protective process does not involve improving the selection and training of agents or the efficacy of screening procedures, but rather improving the referral system and the flow of information into the agency itself. A central issue might be the image of the agency, particularly with mental health professionals, and their willingness or even eagerness to cooperate with it. A modification at this step would have the effect of greatly increasing the scope of the agency's knowledge and extending its network of information, with the agency itself having a central monitoring function. There are clearly potential dangers to civil liberties in such a system, and it is probably the fear of such dangers that has prevented it from evolving. However, if these dangers are only potential and can be controlled, there are possibilities to be explored. Perhaps the most important thing that mental health consultants can do for the Secret Service is not to teach it how to train agents or evaluate and manage subjects, but rather to facilitate its becoming a specialized part of the mental health system, linked (with appropriate controls) to the rest of the system. 112

LEARNING FROM EXPERIENCE, THE SECRET SERVICE, AND CONTROLLED EXPERIMENTATION Lincoln E. Moses, Ph.D. Professor, Department of Statistics Stanford University Stanford, California 1. A True Fable In the second half of 1980 a manager in a Washington statistical agency noted that there was a small room, always filled with four clerks using telephones. Inquiry revealed that they were telephoning recipients of questionnaires who were late in responding, to remind them of their legal obligation to reply. The clerks were engaged in "follow-up of non-response." The manager inquired, "Why is this done?" She was told, "To improve the response rate"—but no facts existed about how well it worked. So she (the manager) required that half the names on the non-respondent list—chosen with a table of random numbers—not be phoned at all that month. Then the timeliness of response in the two groups, called and uncalled, could be compared and the benefits of phone follow-up weighed against the costs. The result was surprising. Response was no better at all in the group receiving telephone follow-upJ The four clerks were assigned to more useful work. This true story briefly illustrates the investigative and convincing power of the experimental approach to resolving practical questions. The alternative, theoretical disputation over the merits and demerits, can be endless; and when it does come to a conclusion (right or wrong), it is likely to leave many unresolved doubts. So there is wisdom in asking about practical questions of importance, "Is there some way to investigate this problem by a controlled experiment?" The answer necessarily is often, "No." (The lung cancer and smoking controversy would have been resolved much sooner if the experimental approach could have been applied, but it could not.) But the answer is sometimes, "Yes," and each such instance is an opportunity that should be carefully explored. Perhaps a direct, convincing answer can be obtained to a nagging, unresolved doubt. 113

2. A Second Example and Some Discussion The following idea was put to test in a celebrated experiment done in New York City in the 1960s. Idea: Arrested persons with community ties, such as jobs or homes in the locality, may safely be released before trial without posting bail. All new arrest cases (excluding arrests for very severe crimes), over a period of many months, were interviewed before arraignment about their community ties (using a scoring system). Thousands of defendants were selected as "eligible," having high enough scores. The experimental staff then recommended, for pre-trial release without bail, a random half of those eligibles. The court accepted about 50 percent of these recommendations, and, in addition, released without bail about 16 percent of the eligibles in the random control half who had not been recommended. The results were that about seven-tenths of one percent of all the eligibles released without bail failed to show up in court. (Not everyone who does post bail appears in court.) This experiment showed so plainly the workability of the idea under test that it was directly applied routinely in New York City, and rapidly taken up in cities all over the country. There was startling additional information turned up by the experiment. Among the eligibles, those chosen by lot to be recommended for pre-trial release without bail had a conviction rate of 40 percent, while those eligibles who were by lot not recommended had a 77 percent conviction rate'. Further, of those who were convicted, 16 percent were sentenced to prison in the first group and 96 percent in the second I These unexpected results show a powerful handicap to arrestees who are not at liberty before the trial to arrange for their legal defense. In both these experiments, the answers that emerged were convincing and had practical consequences. At base, both had the power to answer the question and to convince the skeptical, because their method was simultaneous comparison of two well-defined different procedures on cases of unquestioned comparability. When a difference occurs under these conditions, it is not easily explained away; it is likely to eventuate in action. 3. A More Detailed Look at the Method A fuller analysis of the widely-used experimental approach reveals several essentials which deserve examining here. They can be listed and taken in turn. A. There must be two (or sometimes more) alternate ways of proceeding.In the first example,these were "phone" and "wait." In the second, they were "release without bond" and "detain if no bond is posted." 114

B. The two procedures must be definitely described. Otherwise the results are hard to apply. To define the procedures in terms of "the way Jones does it and the way Smith does it" won't do. Even if Smith's results are clearly much better than Jones', we are left wondering how Smith does it, and how we shall adopt his method for wider use. C. The class of subjects to whom the treatments are applicable must be clearly defined. This was particularly important in the bail bond experiment; persons arrested for very severe crimes were not eligible for possible release without bond. D. The relevant outcomes must be agreed upon in advance. These outcomes were the date of receipt of the questionnaire (in the first example) and appearing in court on the date set (in the second example). The choice of a relevant outcome can sometimes be much more difficult. In comparing training programs, shall we count test performance, job performance, job satisfaction, all three, or something else? E. Fair ways to measure outcomes must be specified in advance. If the measurement involves judgment or any subjective assessment, it can be very important that the judge, or assessor, not know to which group a subject belongs. (This is called "blind" assessment; it is not necessary with very objective measurements—such as scores in target practice, or results of laboratory tests.) F. The alternate procedures should be applied simultaneously. This simple precaution saves the investigation from being weakened by some unforeseeable change in a crucial condition part way through. It assures that all kinds of things, associated with weather, work load, epidemics, hiring freezes, computer breakdowns, are automatically similar under the two conditions. It may prevent endless dispute later because a host of possible explanatory differences are simply barred from entering the picture. Both of the examples used simultaneous comparison. If either of them had, instead, compared one period on one treatment to a later period on the other, there would be doubts. In the bail bond experiment, many things could have changed over time—the kinds of cases, the personnel making the assessment of community ties, some of the judges. Every such doubt would make the outcome less trustworthy, and less useful. G. A mechanically random choice should determine (separately for each subject) the choice of procedure applied. This 115

is a key requirement. It best assures that the subjects will be comparable in the two treatment groups; it simply prevents conscious or unconscious bias from favoring one of the treatments. In the bail bond experiment, it was known that the arrestees recommended for release were no better (or worse) than those not recommended, because it was literally the luck of the draw that decided which particular eligible subjects received the recommendation. Randomization was also crucial to the telephoning example. Without that precaution, the supervisor of the clerks might have chosen for no-phone-call those respondents who usually had prompt responses, or-those involving the largest telephone tolls, or the respondents which were smallest; in any of these cases, the outcome of the experiment would have been debated and the whole undertaking likely undermined by concerns about group non-comparability. These concerns were actually prevented from arising by the use of physical randomization. 4. Possible Applications in the Secret Service It is best to begin by recognizing that some kinds of questions, vexing to the Service, are not likely to be susceptible to resolution by experiment. "Who is dangerous?" "How does interaction with a fanatical group affect the probability of dangerous behavior?" Tractable questions, instead, may relate to those things the Service does and how it does them. The essence of the method is to try out two or more methods in a manner that allows firm conclusions to emerge. It is a way—a very good way—of learning from experience. Some early steps (not necessarily in order): A. Choose a problem. Identify cases of existing intra-service diversity of practice on the same matter, and ask, "Why does this diversity occur?" Is it possible that some of these practices are definitely better than others? Since any well-run federal agency is likely to strive for uniformity of practice, ingenuity may be needed to find diversity. But there are ways to try: (1) When a subject is transferred from one office to another, there is likely to be a change in the way that subject is handled. Are there patterns in such changes? Do they point to hypotheses worth examining? (2) There may be certain decisions that explicitly call for the judgment of two or more agents. What are the operational controversies that are sometimes settled in this process? Do they raise any questions deserving systematic study for trustworthy answers? (3) There may be very similar (but not identical) classes of situations in which different procedures are used. Two 116

or more such classes might be the subject of an experimental check-out of all the methods, used in combination with each of the closely-similar situations. Is there widespread internal dissatisfaction with any of the Services' present routines? Widely shared misgivings of a group of professionals, if objectively considered, may produce questions well worth experimental study. If work of some kinds is widely felt to be needlessly extensive, then it is likely to be widely ill-done. An experiment might settle the question of necessity and whether the result showed that an unpopular practice was, indeed, necessary or not. The answer, being definite, would be useful. B . Identify measurable tokens of success or failure for outcomes. This step is best pursued in connection with the effort to explore a specific question. Still, some general observations may be possible. First, in attacking a problem experimentally, ask, "What outcomes do we hope will occur? What do we hope will not?" And "How can we track these outcomes?" Second, important outcomes may be of varied types, possibly ranging over such diverse matters as agent learning, quality of service from other law enforcement agencies, quality of contact with subjects, change in life style of subject, average time to complete an investigation, agent job satisfaction. Third, at points where reviews now occur (reviews of cases, of agents, of offices, of files) there may be opportunity to measure outcomes, and in terms already used by the Service. C. Decide on fair outcome measures. Consider how to measure the outcomes in ways that cannot be distorted by opinions, beliefs, or wishes of participants. A frequently useful device is to have outcomes judged by qualified persons who do not know which experimental group any case is in (or, even better, do not know that there is an experiment going on) . D. Identify eligible subjects, cases, or opportunities. Specify the class of instances where either of the treatments under comparison may be legitimately applied in compliance with a random choice. Thoughts about this may lead to excluding in advance certain kinds of subjects (or cases, or agents) from the experiment. E. Define the treatment precisely. Describe carefully the treatments to be compared. There can be allowance for flexibility, provided it is spelled out. For example, in medical experiments the dosage of a medication may be 117

defined not as so many milligrams, but rather as the largest dose the patient can accept without nausea. Clear definitions of treatments can be challenging! A study about "the stress interview" isn't doable until that concept has been made more precise and also given a useful operational definition. Is the key feature the use of threat? Or refusing to be conciliatory when the subject voices distress? Eventually a characterization of two kinds of interviewing procedures may be achievable, where one clearly captures important components of the notion "stress interview" and the other lacks those, and where both types are definite enough to do or not do, unambiguously. The foregoing list may seem formidable, and indeed the tasks there may demand much ingenuity and work. But the potential rewards are large. In the last 30 years, clinical medicine has made great strides, and it is fair to say that a large share of the success must be attributed to the widespread and persistent use of randomized, controlled clinical trials to securely establish good innovations and to drop poor ones. The complexities in medical applications are probably as great as those in the Secret Service setting. The Secret Service may do well to consider trying out the method. 5. Some Strategic Suggestions Start small and safe. Although an experimental study can be a simple matter (as with the telephone example), that is not likely. Murphy's Law will lurk. Hence the advice, take a "small" first problem, though a worthwhile one. If it has good prospects for success, so much the better, for that will facilitate attempting more controversial, more difficult, more important studies in later stages. Even if a "small" problem is well chosen for its likelihood of success, and its realistic interest, that first attempt will be challenging enough! There are so many ways to go wrong: the treatment definitions can turn out to be fatally flawed, or some important outcome measure may only be obtainable in 70 percent of the cases, or the eligibility criteria for inclusion/exclusion in the experiment may overlook some numerous class of special cases. So method and discipline are essential throughout. But there is yet more challenge. It will be observed that the whole thing is easier to do if the treatments are not compared simultaneously, but in tandem. Yielding to this notion shoots the program in the head. Someone will notice that it is easier to use odd or even file numbers of new cases than to use random assignment. Yielding to this notion will probably allow someone somewhere to put new cases at choice into either treatment group, by jockeying the order of file number issuance. 118

Vigilance will be essential to prevent these and other modifications from ruining the first experiment. The point is that the essentials of the method are essential; there lies the challenge, to maintain those essentials* There is so much potential value in the experimental mode of learning from experience that the Service may, after consideration, wish to ask interested members of staff, and some of its consultants, to propose one or more possible first studies. From among such proposals it would be very likely to see at least one attractive, initial study emerge. 119

PROBLEMS IN ASSESSING AND MANAGING DANGEROUS BEHAVIOR: SOME COMMENTS AND REFLECTIONS Saleem A. Shah, Ph.D.* Chief, Center for Studies of Crime and Delinquency National Institute of Mental Health Rockville, Maryland One purpose of the Institute of Medicine Workshop on Behavioral Research and the Secret Service is to help the Service augment its knowledge pertinent to its specific mission of protecting the safety of designated political figures and their families. To this end, it is necessary that the focus of attention not be limited solely to the prediction of dangerous behavior and the identification, assessment, surveillance, and management of potential assailants. We may have to direct our special attention and emphasis to potential assailants, but we should try initially to view the task in a broader perspective. We are presumably in agreement that the behaviors of concern to this workshop are serious and potentially lethal attacks against the persons protected by the Service. What follows is a rough conceptual scheme that I would like to propose to the workshop participants as a framework for discussion. The crucial behaviors of interest have three major elements, shown below, each of which is associated with key variables. A. TARGETS (protected persons) Characteristics Availability Vulnerability (e.g., risk-taking behaviors) (Others) B. ASSAILANTS (potential assassins and others) Characteristics Credibility and seriousness of threat posed Presence and proximity Lethal capability (Others) *I would like to acknowledge the helpful comments and suggestions of my colleagues Christopher Dunn, Thomas L. Lalley and Ecford S. Voit, Jr. 121

C. SETTINGS, SITUATIONS, AND TEMPORAL FACTORS Characteristics of settings and situations (e.g., the protected's usual place of work and residence, temporary place of work and residence; in transit situations, public and well publicized versus private and unpublicized travel) and the risks associated with each. Temporary situations (e.g., political and other crises during which risk potentials for some governmental leaders may be appreciably increased). Protectors (U.S. Secret Service and other protective agencies, their key personnel, and their capabilities) and the extent to which organizational structure and operations of the agency facilitate effective and efficient use of available resources; availability of ongoing research and evaluation to test the validity and effectiveness of assumptions underlying various agency policies, procedures, decision rules, training programs; factors that seem to distinguish the "better" protective agents with respect to specific roles, skills, and functions. The following are some notions that follow from the above conceptual scheme: • Unless all three elements in the scheme (targets, assailants, and settings/situations/temporal factors) come together in space and time, there cannot be an assassination attempt. Moreover, even when all three elements do come together, it is possible that the presence and efforts of the protectors could help prevent or thwart the attempt. Therefore, appropriate control and manipulation of any of the three elements in the scheme could, in theory, serve to prevent an attack. • Since there is no a priori reason for focusing attention on any single element in the proposed conceptual scheme (except for reasons of feasibility or practicality), all three elements should be considered and kept in view in any efforts to prevent an attack. • If in particular circumstances one or another element and its key variables is not amenable to needed manipulation or control, then for purposes of preventing attack, attention must necessarily focus on one or both of the other elements and its key variables. Targets and assailants move about in space and time and must be 122

monitored and tracked. As for potential assailants, they must be identified beforehand. • There are many technical, legal, constitutional, public policy and related constraints on the accurate identification and control of "suspected" potential assailants. The Service has (or should have) a considerable advantage over potential assailants in having both advance and highly accurate information about the location and movement of protected persons (at least in regard to events that are not publicized or part of public events such as inaugural parades and other public appearances of targets). Moreover, the Service can (or should) influence to some degree the nature and extent of advance publicity distributed about the movements of the protected, especially when unreasonably high risks are involved. What follows are some further comments and elaborations on the elements of the above conceptual framework and a few of the key variables. The workshop agenda indicates that the topics to be discussed focus primarily on factors associated with assailants. I have some serious reservations about the desirability and wisdom of this emphasis if the primary concern is to prevent assault on the protected. As the conceptual scheme makes evident, in order to prevent the criterion behaviors one must also remain very alert to the several other possible ways in which the Service's mission of protecting the safety of government leaders might be accomplished. For example, I would argue that the characteristics (especially with respect to risk-taking behaviors) of the targets must also be considered when trying to assess the risks of assault. A quick review of the few previous assassination attempts against presidents and presidential candidates, such as Robert Kennedy, makes it evident that these targets have almost never been attacked while in their regular or even temporary place of work or residence. Obviously, in such "regular" locations it is much easier to guard against assassination attempts. Similarly, in all the assassination attempts (with the single exception of the attempt on President Truman, when he was residing in Blair House), the targets were in some public setting. It would appear, therefore, that when the protected move around in public places they not only put themselves at considerably greater risk, but they may also be viewed by potential assassins as more vulnerable. In light of the above observations, and considering the assessment of risk probabilities associated with the targets, 123

especially when in public settings, one might wish to know: (a) Do the protected provide adequate information and notice to the Service about their various trips? (b) How much media and other advance publicity do the protected seem to desire in regard to trips that are associated with greater risk probabilities? (c) How willing are the protected to be guided by, or at least to seriously consider, the advice and suggestions of the Service with regard to reducing unreasonable risks associated with particular proposed activities? Because our societal values pose various limitations in regard to controlling access to firearms and the kinds of surveillance measures that can be used, and because of the immense technical difficulties in making reliable and accurate predictions of events that have extremely low base rates, we should give much greater attention to the characteristics of the targets and to those situations in which assassination risks are greatly increased. With regard to the foregoing, we might also distinguish between various types of constraints and limitations in terms of the available options for preventing or reducing risk probabilities of potentially lethal assaults on protected persons. For example, the constraints could be classified as technical (the state of predictive technology in regard to events that occur infrequently), legal and constitutional (such as prohibitions against "stop and frisk" searches, wiretaps, and the use of preventive detention of suspects), and political (the fact that elected government leaders would be reluctant to markedly reduce their exposure to the public). The above classification makes it quite evident that there is not much that could be done about greatly improving predictive technology with respect to the accurate identification of potential assassins. There are also serious though not absolute limitations with respect to the legal and constitutional constraints. In theory, at least, assassination risks could be appreciably reduced by exerting greater control over the key variables under settings/situations/temporal factors, for example, by reducing and/or more closely controlling opportunities for assasination in public settings and situations. Considering the tendency over recent years to increase the number of the protected, some attention should be given to identifying those who are of more critical "protective interest" and who are also at greater risk—at least judging from past data in this regard. For example, presidents appear, by far, to be at much greater risk while in office than after they leave office and as compared to vice presidents—whether during or following their terms in office. Such distinctions (which the Service probably already uses) are very essential in order that limited resources, personnel, and facilities be allocated in some relationship to degree of risk. 124

Along the same lines, it is unclear to me just how "major" presidential and vice presidential candidates are determined. Here, again, there are questions with respect to both national interest, and risk probabilities. The more widely known candidates might be at greater risk of lethal attacks, while some other candidates might be more at risk for verbal abuse and beer cans. There is yet another relevant issue of concern here. It is my impression that Congress has in recent years significantly increased the number of persons the Service has responsibility for protecting. Considerations other than perceived potential risk of dangerous assault seem to have motivated at least some of these enlargements. It would be worthwhile to consider whether the list of protected persons might properly be reduced—at least in reference to the responsibilities of the Service—in order that available resources be more carefully allocated for persons whose protection is much more a matter of national concern and who are also likely to be at greater risk. To the extent that sentiments of sympathy, concern, gratitude, and the like, appear to have motivated at least some of the additions by Congress to the list of protected persons, might not other and more appropriate expressions of such feelings be explored? For example, funds could be provided to designated persons in order that they might obtain desired or indicated private protective services; however, there would be no requirement that the funds so allocated could be used only for protective services. Given the immense technical problems in trying to improve the predictive accuracy for events with extremely low base rates, Frank Zimring* has offered the excellent idea of considering "proxy validation" by the use of "plausible proxies." While this strikes me as a very good notion, the problem I have with the proxies presented is that their "plausibility" seems poor. Acts of interpersonal violence selected as proxies must have some functional equivalence, or at least close similarity, to the criterion behaviors of concern; otherwise, we would simply have exchanged one serious problem (very low base rates, or low frequency events) for another one (inability to extrapolate findings to the criterion behaviors of prime concern). Violent crimes in general, child abuse, spouse abuse, and suicide attempts strike me as being rather different and distinct from violence that is directed at national political leaders. There are several other approaches that might be considered for dealing with the base rate problem and the difficulties in trying to analyze and make more sense of the very few cases available. *Professor of Law, University of Chicago, in a memorandum to conference participants, included in this summary, page 187. A member of the planning committee, Mr. Zimring was unable to attend the conference. 125

First, assuming that violence directed at political leaders is in some ways rather distinct (i.e., assassination attempts), why not increase the rate by including serious assaults directed at state and perhaps even local political leaders? Of course, even though the targets would remain in the same descriptive category (political leaders), it is quite possible that only some of the attacks against state level politicians, and even fewer against local politicians, would be sufficiently similar to those against national political leaders. Second, because protecting national political leaders is a concern that most assuredly is widely shared with most (if not all) other countries, one could undertake some cross-national studies and pool relevant data from several other countries, preferably from the Western democracies with socio-political characteristics similar to those in the United States. Moreover, once assassination attempts are classified into specific categories, such as mentally disordered assassins and politically-motivated assassins, it is possible that an even wider range of functionally equivalent incidents and cases could be pooled for study and analysis. Third, might not the New York City Police Department have some relevant data in regard to their protective responsibilities for members of the various United Nations delegations and also with respect to foreign officials at the various consulates? Fourth, it would appear that various other national leaders (non-politicians) may well have some of the same stimulus qualities as political leaders in attracting certain types of assassins (e.g., Martin Luther King, Vernon Jordan). Similarly, might attacks against some other public figures, such as John Lennon, also have some relevance? There is one other conceptual approach that occurs to me in regard to efforts to deal with the base rate problem. We could try to consider the criterion behavior along a spectrum, where the "core" or prime behavior is first defined and identified, as in a well-planned and potentially lethal attack on a political leader. Then one could move out to less lethal attacks and attempts directed at the same general target group. For example, various types of attacks could be scaled and rated along some dimensions relevant to the critical behaviors. The seriousness or lethality of the attack might be one such dimension; being within good shooting range of the target with a loaded firearm would receive a very high rating, as would being within throwing range with a live grenade or some other explosive. An assault with a rock or other heavy object would have a lower rating, depending upon the pitching arm strength and range of the attacker, and so forth. Attacks with fists, sticks, or empty beer cans would be too far removed from the prime criterion behaviors and might either receive a very low rating or be excluded. 126

Finally, some brief comments about the protective and related capabilities of the U. S. Secret Service. It will have been noted that this appears as one large sub-category of variables under settings/situations/temporal factors in the conceptual scheme. As I had occasion to indicate in my first meeting with Mr. Robert Kyanko of the Secret Service, the agency simply must significantly increase and improve its capability for undertaking a range of in-house research efforts. There must certainly be a wealth of information that has been accumulated over the years with respect to threats, attempts, attacks, and so forth, against various persons protected by the Service. To what extent has such data systematically been analyzed and studied? To what extent have the operating procedures, various decision rules, intuitive judgments, training procedures for agents, and so on, been subjected to careful empirical assessment? For example, a comparison of threats, attempts and completed assault in reference to a host of operational, situational, and other dependent variables might help provide some better and clearer idea of the risk probabilities associated with certain types of serious attempts and potentially lethal attacks. It might also be possible to try to pinpoint certain factors, such as lack of prior information, ineffective response by various protectors, or risk elements that have been overlooked, that seem to contribute to less successful protective efforts. The Service also should have funds to support a variety of extramural research efforts that are directly relevant to its mission. As far as more basic studies are concerned on predictive technology more generally and related topics of broader concern, the Service might well wish to consider cooperative or joint efforts with other federal research support agencies. In order to markedly strengthen its internal research capabilities in the behavioral and social science areas, and also to give this function much greater priority and emphasis than appears to have been the case in the past, there might be established a research advisory committee to assist the agency in developing, further improving, and maintaining at a high level its internal research capabilities. Members of such an advisory committee would be separate from the regular and direct research consultants that have been used from time to time and who should continue to be used as needed by the Service. Finally, I would like to note that, in order that the available resources and capabilities of the protective branches of the U. S. Secret Service be used in the most effective and efficient fashion, it might be necessary to consider whether the organizational structure of the agency actually facilitates efficient functioning in regard to the key mission of concern. The analogy that comes to mind is that the potential horsepower of an automobile engine may not matter very much if, in actuality, the power is not efficiently getting to where—as the saying goes—"the rubber meets the road." 127