The Goldwater Rule: Perspectives From,and Implications for,Psychological Science

The Fact magazine article and the origins of the Goldwater Rule

The 1964 “Barry Lyndon” Presidential Election, which pitted Republican Barry Goldwater against incumbent president Lyndon Baines Johnson in the midst of the impending war in Vietnam, was enormously contentious. Goldwater, a U.S. Senator from Arizona, was a libertarian conservative who favored balancing the budget, massive tax cuts, and an extremely strong national defense; while in the Senate he had voted against the Civil Rights Act of 1964. Johnson was a vice president who had assumed the presidency following the assassination of John F. Kennedy in November of 1963; he was an ardent advocate of social programs and civil rights who had criticized Goldwater for being overly hawkish and insinuated that Goldwater would place America at risk for nuclear war if elected.

In the waning months of the 1964 presidential campaign, Ralph Ginzburg, the openly politically liberal editor of Fact magazine (now defunct), published a 64-page issue entitled “The Unconscious of a Conservative: A Special Issue on the Mind of Barry Goldwater,” an apparent wordplay on the title of Goldwater’s (1960) book, The Conscience of a Conservative. In addition to an article by Ginzburg attacking Goldwater (Ginzburg, 1964), the issue contained the results of a survey regarding Goldwater’s mental health (Boroson, 1964). The survey had been sent to 12,356 psychiatrists, 2,417 of whom responded (Slovenko, 2000). Of these respondents, 1,189 deemed Goldwater to be unfit to be president and 657 deemed him to be fit; 571 said that they did not possess sufficient information to answer. A few psychiatric respondents protested, including Walter Barton, Medical Director of the American Psychiatric Association, who insisted that psychiatric opinions of public figures require a “thorough clinical examination” (Martin-Joy, 2015). Although less than 10% of psychiatrists who had been mailed the survey deemed Goldwater to be unfit for the Oval Office, the cover of the Fact issue proclaimed that “Fact: 1,189 Psychiatrists Say Goldwater Is Psychologically Unfit to Be President.” Although technically true, this headline omitted the rest of the story.

Many of the negative comments about Goldwater were damming. The psychiatrists who responded variously described Goldwater as “emotionally unstable,” “immature,” “cowardly,” “grossly psychotic,” “paranoid,” “delusional,” “amoral and immoral,” and a “dangerous lunatic” (Boroson, 1964). One wrote that “a megalomaniacal, grandiose omnipotence appears to pervade Mr. Goldwater’s personality, giving further evidence of his denial and lack of recognition of his own feelings of insecurity and ineffectiveness.” Others likened him psychologically to brutal dictators, including Hitler and Stalin. Numerous respondents proffered diagnoses of Goldwater, including schizophrenia, paranoia (now termed delusional disorder), obsessive-compulsive disorder, and narcissistic character disorder (Friedman, 2017; Levine, 2017), even though the modern incarnation of the latter (NPD) was not a formal psychiatric diagnosis at the time. One psychiatrist speculated that he suffered from an anal character, another that he harbored a deep-seated emotional identification with Hitler, and still others questioned his masculinity and his relationship with his wife. Even some well-known psychiatrists questioned Goldwater’s psychological adjustment. For example, Jerome Frank, Chairperson of the Department of Psychiatry at Johns Hopkins University, famed for his classic book on the role of nonspecific factors in psychotherapy (Frank, 1961), opined that the “ill-considered, impulsive quality of many of Goldwater’s public utterances is, in my mind, sufficient to disqualify him from the presidency.” Adding insult to injury, the Fact piece featured a phrenological drawing of Goldwater’s skull depicting brain regions labeled with such terms as “destructiveness,” “short-sightedness,” and “deaf ear to the poor” (Boroson, 1964).

Goldwater was not amused (Martin-Joy, 2015). In 1966, he filed a libel suit in U.S. District Court against Fact, maintaining that the magazine had published “false, scandalous and defamatory statements referring to and concerning [the] plaintiff.” Appearing on the witness stand during the trial, he defended himself forcefully against the allegations in Fact, insisting that he was well-adjusted, that he had never been to a psychiatrist, and that his marriage to his wife was healthy. In 1968, the court awarded him $1 in compensatory damages but $75,000 in punitive damages (Black, 1969), an amount that would be equivalent to about a half-million dollars today.

Ginzburg appealed the ruling to the U.S. Court of Appeals and, eventually, the U.S. Supreme Court, but his petitions were rejected. Interestingly, the Supreme Court denial was not unanimous, as Justice Hugo Black was joined in dissent by Justice William O. Douglas. Some of Black’s (1969) comments foreshadowed later criticisms of the Goldwater Rule:

This suit was brought by a man who was then the nominee of his party for the Presidency of the United States. In our times, the person who holds that high office has an almost unbounded power for good or evil. The public has an unqualified right to have the character and fitness of anyone who aspires to the Presidency held up for the closest scrutiny. Extravagant, reckless statements and even claims that may not be true seem to me an inevitable and perhaps essential part of the process by which the voting public informs itself of the qualities of a man who would be President.

In the wake of the Fact controversy, the American Psychiatric Association released several statements warning members against issuing diagnostic pronouncements regarding public figures. It was not until 1973, however, that it formally responded to the concerns raised by the Goldwater affair. It did so by enshrining Rule 7.3 into its first-ever formal code of ethics. The rule, which was to become known colloquially as the Goldwater Rule, read as follows (see American Psychiatric Association, 2017):

On occasion psychiatrists are asked for an opinion about an individual who is in the light of public attention or who has disclosed information about himself/herself through public media. In such circumstances, a psychiatrist may share with the public his or her expertise about psychiatric issues in general. However, it is unethical for a psychiatrist to offer a professional opinion unless he or she has conducted an examination and has been granted proper authorization for such a statement.

Hence, Rule 7.3 mandates that psychiatrists cannot comment on the mental health of public officials unless they have (a) examined these individuals directly and (b) obtained informed consent from them to release this information. Although the Goldwater Rule was initially vague with regard to whether it referred to professional opinions as opposed to formal diagnoses, a statement by the American Psychiatric Association Ethics Committee in March of 2017 made clear that it applies to both and even extends to public statements that a given political figure does not meet criteria for a specific mental disorder (American Psychiatric Association Ethics Committee, 2017). Although the precise boundaries of the Goldwater Rule continue to be debated within psychiatric circles (Martin-Joy, 2017a), most scholars concur that the rule does not prohibit psychiatrists from speaking in generalities about the features of a given mental disorder and their broad relevance to a public figure (Friedman, 2008). For example, if asked by a reporter whether a specific politician fulfills DSM–5 criteria for NPD, the psychiatrist could educate the public regarding the core features of this condition and allow laypersons to draw their own conclusions regarding this politician’s diagnostic status.

A clarification to the Goldwater Rule

A challenge to the Goldwater Rule arose 27 years later when psychiatrist Jerrold Post of George Washington University, acting in his position as founder of the now-defunct Central Intelligence Agency’s Center for the Analysis of Personality and Political Behavior, constructed a detailed psychiatric profile of Iraqi dictator Saddam Hussein despite never having met him. After assembling this profile from a careful examination of multiple biographies of Hussein and interviews of individuals who had known him well, Post concluded that Hussein was a “malignant narcissist.” Questioning the contention that Hussein was a “madman,” Post argued that Hussein was rational but that his extreme narcissism put him at risk for dangerously erratic decisions. Post presented the fruits of his research on Hussein’s psychiatric status at a hearing of a U.S. Senate Committee in 1991 (Goleman, 1991).

In response, the American Psychiatric Association received a slew of complaints from its members, who argued that Post had violated the Goldwater Rule given that he had not met Hussein (Moran, 2008). Echoing these complaints, Post was rebuked by American Psychiatric Association officials for his report (Mayer & Leichtman, 2012). Post retorted that his report constituted a “political personality profile” (Post, 2015) of Hussein rather than a formal psychiatric diagnostic workup. Furthermore, he maintained that he possessed a “duty to warn” the U.S. Congress and the public regarding Hussein’s dangerous personality traits.

In response to the controversies raised by Post’s report on Hussein, the American Psychiatric Association appointed a task force called The Psychiatrist as Psychohistorian, which addressed the ethics of constructing personality profiles on living political figures and on deceased historical figures. The task force recommended a clarification to the Goldwater Rule, which allows for the “psychological profiling of historical figures designed to enhance public and governmental understanding of these individuals” so long as this profiling is not accompanied by a formal psychiatric diagnosis and is based on adequate, peer-reviewed research (Cooke, Goddard, Werner, Cooke, & Griffith, 2014). Nevertheless, this amendment has not been formally adopted by the American Psychiatric Association, so its exact status remains unclear.

This potential amendment is a nod to the substantial body of research on “significant samples”—that is, samples of historically noteworthy figures (Song & Simonton, 2007). Indeed, an entire discipline of study, namely psychohistory, focuses on retrospective psychological analyses of historical figures, typically those who are no longer alive (Runyan, 1982; Schultz, 2005). Freud (1916), for instance, authored psychohistorical analyses of artist Leonardo da Vinci and of U.S. President Woodrow Wilson (Freud & Bullitt, 1966); Erik Erikson (1969) authored a psychohistorical analysis of Mahatma Gandhi; and psychiatrist Nassir Ghaemi (2011) authored psychohistorical analyses of numerous political leaders, including Abraham Lincoln, General William Tecumseh Sherman, and Winston Churchill. Even Paul Meehl, one of the most highly respected clinical psychological scientists of the second half of the 20th century, offered various armchair “diagnoses” of major political figures, including Adolph Hitler, whom he described as a schizotype (Meehl, 1978). Psychohistorical analyses have often acquired a bad name in mainstream psychology, largely owing to their (a) susceptibility to hindsight bias and (b) longstanding connection with Freudian theory, which substantially shaped many early psychohistorical analyses. Nevertheless, the inferential problems with such analyses can be partly overcome by leveraging large quantities of publicly available data (Elms, 1994), a point to which we return.

Number 1: The necessity of a direct personal examination

Number 2: Safeguarding the reputation of mental health professions

Number 3: Abolishing the Goldwater Rule would generate chaos

Number 4: Damaging the reputation of public figures

Number 5: The irrelevance of psychiatric information to political performance

Number 6: Politicization of the profession

Number 7: Lack of incremental validity of psychiatric information beyond other information

The physical examination

In many respects, the direct interview assumption appears to be an heir to the standard physical examination, which is familiar to anyone who has undergone a routine medical check-up. The physical examination has a lengthy history originating with Hippocrates and traces its roots to such developments as the inventions of percussion by Austrian physician Leopold Auenbrugger in 1761 and the stethoscope by French physician Rene Laennec in 1816 (Walker, Hall, & Hurst, 1990). Although the standard physical examination has evolved over time and may consist of varying procedures, it has long rested on five central elements (Campbell & Lynn, 1990): inspection (observation of body parts), clinical inquiry (history-taking), palpation (using the fingers to apply pressure to a body part to assess the condition of underlying body parts), percussion (tapping a body part, most often the thorax or abdomen, to ascertain the condition of the underlying organs), and auscultation (listening to the functioning of the internal organs, typically by means of a stethoscope).

There is little question that many physical disorders, such as aortic valve regurgitation (a leaky aortic valve), are difficult or impossible to diagnose accurately without a direct physical examination (Knox, 2010). For example, in 2005 U.S. Senator and former surgeon Bill Frist of Tennessee was excoriated by many medical colleagues when he concluded, exclusively on the basis of observing videotapes, that Terri Schiavo, a Florida woman who had experienced severe brain damage following a cardiac arrest, was not in a persistent vegetative state, as many physicians had asserted (Annas, 2005). Among those who criticized Frist in print were 31 of his former Harvard Medical School classmates, who contended that videotapes alone did not afford a sufficient basis for a judgment of Schiavo’s neurological condition (Babington, 2005).

Psychiatrists, who are physicians by virtue of their training, may be tempted to assume that a clinical interview possesses the same level of validity for psychiatric diagnostic purposes as does a standard physical examination. Does the research evidence support this presumption?

The unwarranted mystique of the clinical interview

Few psychiatric authors who have commented on the Goldwater Rule (but see Kroll & Pouncey, 2016, for an exception) have acknowledged the limitations of in-person interviews as a means of accessing diagnostically relevant information. This omission is surprising given the substantial corpus of psychological literature bearing on the psychometric shortcomings of interviews. Such evidence is especially problematic for unstructured interviews, which are by far the most commonly used interviewing methods in standard psychiatric clinical practice (Aboraya, 2009). Indeed, Miller, Dasher, Collins, Griffiths, and Brown (2001) noted that the traditional method of diagnostic assessment is the “unstructured interview that is the standard of practice for that task throughout psychiatry” (p. 256); the same state of affairs very likely holds for clinical psychology and allied fields (see Mihura, Roy, & Graceffo, 2017, for survey data demonstrating marked variability in the coverage of structured interviews in clinical psychology doctoral programs).

Nevertheless, the preponderance of the evidence indicates that unstructured interviews, although providing potentially useful information, are characterized by pronounced psychometric limitations. For example, unstructured diagnostic interviews display low to modest levels of interrater reliability (Aboraya, Rankin, France, El-Missiry, & John, 2006; Dreessen, Hildebrand, & Arntz, 1998; Garb, 1998; Spitzer & Fleiss, 1974), most likely because they lack standardized questions, probes, and scoring criteria. Parallel findings have emerged in the industrial-organizational literature, which reveals that the reliability of unstructured interviews tends to be lower than that of structured interviews (Huffcutt, Culbertson, & Weyhrauch, 2013; McDaniel, Whetzel, Schmidt, & Maurer, 1994). Needless to say, when two interviewers disagree on the diagnosis of an individual, it is unclear which diagnostic conclusion, if either, to accept. Further, because classical test theory reminds us that validity is limited by the square root of reliability (Nunnally, 1978), these reliability findings suggest that the validity of unstructured interviews may be considerably lower than desired. Accordingly, data from the psychiatric (Kranzler et al., 1995; Miller et al., 2001) and industrial-organizational psychology (Macan, 2009) literatures reveal that structured interviews usually yield higher levels of validity than do unstructured interviews.

In a helpful review, Samuel, Suzuki, and Griffin (2016; see also Samuel, 2015, and Samuel & Bucher, 2017) observed only modest levels of agreement between clinician-based and research-based (e.g., semistructured interview) measures of personality disorders, regardless of whether categorical or dimensional assessments are used. The same conclusion holds for psychiatric conditions other than personality disorders (Rettew, Lynch, Achenbach, Dumenci, & Ivanova, 2009). Although these findings do not demonstrate that clinician-based data are of lower validity than research-based data, recent data point in this direction. In a study of 320 patients (Samuel et al., 2016), clinician diagnoses of personality disorder displayed relatively low levels of convergence with both semi-structured interview and self-report measures of personality disorders. Furthermore, whereas the semistructured interview and self-reported data each consistently provided incremental validity beyond clinician ratings of personality pathology in statistically predicting psychosocial functioning, clinician ratings provided little incremental validity.

There are at least three reasons to be doubtful on an a priori basis that clinical interviews afford a “gold standard” (e.g., Pierre, 2016) for psychiatric diagnosis. First, the diagnosis of personality disorders in particular requires evidence of pervasiveness—that is, of cross-situational consistency of the behaviors relevant to the condition in question—and of persistence—that is, of temporal stability—of these behaviors (Carlson, Vazire, & Oltmanns, 2013). For example, the diagnosis of NPD requires evidence of grandiose self-importance, entitlement, exploitativeness, lack of empathy, and so on, across multiple situations and extended time periods. Although skilled interviewers can sometimes elicit such information with careful questioning, the accuracy of such information can be compromised by respondents’ poor memories, lack of insight, and defensiveness (Samuel & Bucher, 2017). In such cases, it is likely that observer (informant) reports can be valuable in supplementing information from clinical interviews (Miller & Lynam, 2015; see “The validity and incremental validity of informant reports of personality and personality disorders”).

Second, the limitations of in-person psychiatric examinations in research settings are likely to be greatly exacerbated in real-world settings in which a politician’s future is at stake. In such settings, potent demand characteristics (Orne, 1962) are surely present. Such in-person examinations almost surely constitute “strong situations,” those in which there are clear-cut expectations regarding how and how not to behave (Monson & Snyder, 1977). Imagine a politician being evaluated by a psychologist for suitability for the presidency. He or she is well aware that any indications of a serious mental disorder will raise questions concerning his or her qualifications and will be likely to deny or minimize psychopathology. This problem is likely to apply just as readily to structured as to unstructured psychiatric interviews. Growing data suggest that situational strength moderates the validity of measures of personality traits, with lower validity in “stronger” situations (Meyer, Dalal, & Hermida, 2010). Such findings raise the distinct possibility that the validity of in-person interviews for high-stakes positions, which are mandated by the Goldwater Rule, will be considerably lower than hoped.

In fact, one might question how helpful interview data would be in the face of the overwhelming archival data available for many contemporary politicians. For instance, if confronted with substantial data suggesting that a presidential candidate may meet criteria for antisocial personality disorder (official reports; informant data indicating a longstanding pattern of irresponsible and even illegal behavior beginning in childhood), would a clinician seriously consider information gleaned from an interview that contradicted those data (e.g., the interviewee reporting that these behaviors did not occur)? There is ample reason for doubt.

Third, the large literature on cognitive biases (Garb, 1998, 2005; Samuel & Bucher, 2017) demonstrates that interviewers are not immune from judgmental errors that afflict people in general. For example, clinicians are vulnerable to “diagnostic overshadowing,” whereby the presence of a salient and dramatic diagnosis, such as borderline personality disorder (BPD), often leads them to overlook less salient and dramatic diagnoses, such as major depressive disorder (Garb, 2005; Zimmerman & Mattia, 1999). In addition, when making diagnoses, clinicians often overweight criteria that fit into their schemas of given disorders (Kim & Ahn, 2002). In one study, clinicians tended to diagnose BPD when self-injurious behavior and affective instability were present, even when patients did not satisfy the requisite number of criteria for BPD (Morey & Ochoa, 1989). Some interviewer biases can be attenuated by the use of structured interviews, although as noted earlier these measures are used uncommonly in routine clinical practice.

The interview illusion

Despite these well-documented limitations, many clinicians continue to maintain that unstructured interviews possess high levels of validity and contribute useful information beyond other sources of information (see Kahneman, 2011, and Kahneman & Tversky, 1973, on the “illusion of validity”). In a striking example of the shortcomings of unstructured interviews, in 1979, the Texas State legislature mandated that the University of Texas Medical School at Houston increase its class size from 150 to 200; as a consequence, the program belatedly admitted 50 candidates who had been interviewed but were initially rejected. Follow-up analyses revealed that the candidates rejected largely on the basis of their interviews did not differ significantly from initially accepted applicants in dropout rates or performance during either medical school or 1 year post-medical school (Dawes, 1994; DeVaul et al., 1987; Milstein, Wilkinson, Burrow, & Kessen, 1981).

Some authors have suggested that the addition of unstructured interviews may in some cases actually harm the accuracy of clinical judgments and predictions (Dawes, 1994). This paradoxical effect, whereby more information contributes to lower accuracy, can be explained in part by the dilution effect (Nisbett, Zukier, & Lemley, 1981). In this well-documented effect, the introduction of salient but low-validity information, such as subjective impressions from interviews or job talks, leads raters to rearrange their mental weighting of the predictor variables and place undue emphasis on low-validity data at the expense of high-validity but less memorable data, such as information gleaned from years of past performance.

In a classic review, Sawyer (1966) found that clinicians’ predictions of relevant outcomes, such as length of hospitalization and criminal recidivism, were slightly better when interview data were ignored. More recently, Dana, Dawes, and Peterson (2013) asked individuals to forecast college students’ grade point averages (GPAs). Some participants received biographical information regarding each student along with his or her current GPA; others received this information but were also asked to interview each student. Predictions were substantially worse when interviews were added into the mix. Remarkably, the same pattern held in a separate experimental condition in which participants emitted random responses to yes-no interview questions. Just as strikingly, interviewers found the random interviews to be as informative as the nonrandom interviews.

Similarly, for certain disorders that may be particularly concerning among holders of high public office, such as psychopathic personality (psychopathy), interview-based data may at times be of negative predictive validity. In one study (Porter, Brinke, & Wilson, 2009), psychopathic offenders were 2.5 times more likely than nonpsychopathic offenders to be recommended by parole officers for conditional release, suggesting that the former offenders had left a positive impression during interviews and other interpersonal interactions. Nevertheless, psychopathic offenders’ subsequent rates of recidivism were considerably higher than those of other offenders, suggesting that parole officers had been fooled (see also Ruback & Hopper, 1986, for evidence that the addition of parole interviews to other prisoner biographical information may decrease the accuracy of parole decisions).

Susceptibility of interviews to impression management

A counterargument to concerns regarding the demand characteristics of in-person psychiatric evaluations of public figures is that experienced interviewers can detect interviewees’ propensities to minimize their psychopathology. As noted earlier, Appelbaum (see Rehm, 2016) proposed that direct interviews are needed to determine whether respondents are crafting a misleading impression regarding their adjustment. To the contrary, an insufficiently appreciated limitation of interviews is their vulnerability to response sets, especially impression management. In contrast to self-report inventories, some of which contain embedded validity scales designed to detect response sets (e.g., Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989), the validity of interviews hinges in large measure on the assumption of veridical reporting on the part of respondents (Widiger & Frances, 1987).

Evidence indicates that most interviewers believe that they can detect impression management on the part of interviewees (Robie, Tuzinski, & Bly, 2006). Although no data bear on the veracity of this assumption in the case of psychiatric interviews, two studies have examined this question in the context of pre-employment interviews (see also Barrick, Shaffer, & DeGrassi, 2009). In a study of 164 employment interviews, Roulin, Bangerter, and Levashina (2014) found that interviewers were typically unable to detect interviewees’ efforts at impression management, as measured by applicants’ self-reported attempts at ingratiation, self-promotion, creation of a positive image, and other tactics. Roulin, Bangerter, and Levashina (2015) further examined interviewers’ ability to detect interviewee impression management across five laboratory studies. They again found that interviewers fared poorly, only slightly exceeding chance. Of potential relevance to the Goldwater Rule, detection accuracy was especially poor for such deceptive tactics as image creation and image protection, which constitute efforts to construct a misleading impression of who one is. Furthermore, across studies, interviewer experience was nonsignificantly associated with the accuracy of detection. To the extent to which these findings are generalizable to psychiatric interviews, they raise serious questions concerning interviewers’ ability to detect interviewee efforts at impression management.

Differing sources of data in the assessment of personality and psychopathology

Psychologists have long distinguished among differing sources of data relevant to personality and psychopathology, each with their own strengths and weaknesses. Specifically, psychologists distinguish among S data (self-reported data, such as those from questionnaires), O data (data from observers or raters, sometimes also called R data), T data (data from laboratory tests), and L data (lifetime or archival data, such as participants’ speeches, diaries, school report cards, traffic tickets, and criminal records; Block, 1977; Cattell, 1957). The modal psychiatric interview is an admixture of S and (to a lesser extent) O and L data, as it relies largely on respondents’ self-reports but also in part on interviewers’ (a) ratings of respondents’ personality and pathological features and (b) elicitation of longstanding behaviors.

Although the importance of multimethod approaches has received considerable attention in the childhood and adolescent psychopathology literatures (Achenbach, McConaughy, & Howell, 1987; De Los Reyes et al., 2015), it has received far less attention in the adult psychopathology literature (Achenbach, 2006; Achenbach, Krukowski, Dumenci, & Ivannva, 2005). Nonetheless, in many or most cases, differing sources probably contribute nonoverlapping and potentially useful information for adult disorders. In this regard, Spitzer (1983) introduced the LEAD (longitudinal expert evaluation of all-data) standard as a rubric for establishing psychiatric diagnoses. Using this standard, information from all available sources is collected and integrated into a consensus diagnosis by a panel of experts (Pilkonis, Heape, Ruddy, & Serrao, 1991). Although few studies have examined the validity of the LEAD standard—most investigators have treated it as a criterion rather than as a predictor—data from 100 patients recruited from substance abuse programs demonstrated that the convergent validity of this standard for external criteria (e.g., frequency of substance abuse) generally exceeded that of an expert clinical interview (Kranzler, Tennen, Babor, Kadden, & Rounsaville, 1997). Given that the LEAD-based diagnoses included the results of the expert-based interview, these findings suggest that the addition of other information sources, such as informant reports, self-report data, and chart records, incremented the interview’s validity.

The Goldwater Rule presumes that direct interviews are inherently more valid than are other information sources, including extensive amounts of high-quality O and L data that have been integrated by experts. This assumption is dubious. To see why, consider Donald Trump and Hillary Clinton, the two major candidates for the 2016 U.S. presidential election. As of this writing, the Internet Archive’s Trump Archive contains over 520 hours of videotapes of Trump dating to 2009, including interviews, speeches, and other news footage (https://blog.archive.org/2017/01/05/internet-archives-trump-archive-launches-today/). Trump has also been the subject of several biographies authored by journalists who have conducted interviews with scores of individuals who have known him well for decades (e.g., Kranish & Fisher, 2017). Trump has also tweeted his thoughts about an enormous array of issues over 35,000 times as of this writing (Resnick, 2017). In addition, one videographer alone filmed Clinton for over 6,000 hours during the last presidential campaign (Zuckerbrod, 2016); again, this treasure trove of material does not include the many biographies of Clinton or extended interviews with numerous people who have worked closely with her (Bernstein, 2013; Wakefield, 2016).

We contend that the Goldwater Rule was perhaps defensible in previous decades, when much less was known about political candidates than is the case now, owing to the rise in the internet, social media, the near ubiquity of audio and video-recording devices, as well as a different attitude toward politicians and candidates by reporters today than in the past. For instance, potentially relevant behavior (e.g., sexual infidelities, substance use) and sensitive information (e.g., medical infirmities) were often not revealed about politicians in the past as part of a greater ethos of confidentiality among journalists than is normative today. As president, John F. Kennedy managed to hide his severely impairing Addison’s disease from the press (Dallek, 2017), a feat that would be nearly impossible today. As such, the public in general and mental health clinicians in particular are privy to a great deal of information that would have remained concealed in past decades.

To justify adoption of the Goldwater Rule, its proponents would need to maintain that an assessor could somehow glean more valid information from a relatively brief (typically 1–2 hours and, in rare cases, perhaps 3 or 4 hours in length) in-person interview-based examination of a politician, such as Trump or Clinton, than from assiduous inspection of dozens or hundreds of hours of videotapes of his or her informal interviews with television hosts, speeches, off-the-cut remarks, debates, and so on. Of course, this enormous archival body of L data does not even consider numerous other sources of potentially relevant L and O data about Trump or Clinton, including their autobiographies, biographies, radio interviews, emails, tweets, or in-depth discussions with friends and reporters who have interacted with them frequently over the years.

Data also suggest that high-quality L data can often provide information of approximately equal validity to interview-based data. For example, studies of the Psychopathy Checklist (PCL) and its revised version (PCL-R; Hare, 1982/1991), a well validated interview-based measure of psychopathic personality (psychopathy), demonstrate that this measure can be scored with high validity from detailed file information alone. A meta-analysis of 95 PCL studies (N = 15,826) revealed significantly higher validity for PCL and PCL-R Factor 2 (antisocial lifestyle) dimensions when the measure was scored using file data alone; the findings for PCL and PCL-R Factor 1 (interpersonal and affective features of psychopathy) were comparable and did not differ significantly (Leistico, Salekin, DeCoster, & Rogers, 2008; see Edens, Campbell, & Weir, 2007, for similar results in a meta-analysis of youth PCL measures). Other data suggest that diagnoses of DSM personality disorders, such as BPD, can be made with adequate validity from psychiatric chart data alone (e.g., Blais, Hilsenroth, & Castlebury, 1997; Fyer, Frances, Sullivan, Hurt, & Clarkin, 1988). In aggregate, these findings raise questions regarding the empirical underpinnings of the Goldwater Rule, especially with respect to personality disorders, some of which (e.g., psychopathy) are of obvious relevance to performance in high political office.

The validity and incremental validity of informant reports of personality and personality disorders

Considerable data also attest to the construct and incremental validity (Meehl, 1959; Sechrest, 1963) of data from informants with respect to personality traits and personality disorders. Such findings bear on the Goldwater Rule, as they address the question of whether O data can be equally or more informative than data obtained from respondents’ themselves. To put it in colloquial terms, is it better to get diagnostically relevant information “from the horse’s mouth” or to get it from well-acquainted informants?

There are ample theoretical reasons for expecting O data to contribute substantially to the diagnosis of at least some conditions, such as personality disorders (Widiger & Samuel, 2005). Such data may be helpful in filling in the “blind spots” in self-reporting that characterize individuals with some personality disorders, especially Cluster B conditions (antisocial, histrionic, borderline, narcissistic), which tend to be more ego-syntonic than most mental disorders (Grove & Tellegen, 1991; Lilienfeld & Fowler, 2006). DSM–5 issues similar recommendations with respect to personality disorder diagnosis. Although asserting that “a single interview with the individual is sometimes sufficient for making the diagnosis,” the manual acknowledges that “assessment can also be complicated by the fact that the characteristics that define a personality disorder may not be considered problematic by the individual. . . . To overcome this difficulty, supplemental information from other informants may be helpful” (American Psychiatric Association, 2013, p. 647). Ideally, this information would come from multiple individuals who occupy a broad range of positions in the person’s life (e.g., coworkers, supervisors, friends, romantic partners), akin to the 360-degree evaluations routinely conducted in industrial/organizational psychology (e.g., Beehr, Ivanitskaya, Hansen, Erofeev, & Gudanowski, 2001).

With respect to general personality traits, burgeoning evidence indicates that informant data tend to possess higher validity than do self-report data for a range of external criteria. A meta-analysis across 263 samples (N = 44,178) revealed that O data for personality traits yielded considerably higher validities than S data for these traits with respect to job performance and academic achievement; for predicting strangers’ first impressions, S and O data were of approximately equal validity (Connelly & Ones, 2010; see also Connelly & Hülsheger, 2012). In addition, O data generally displayed substantial incremental validity beyond self-reports for these two criteria but not vice versa. Broadly converging conclusions have emerged in the industrial-organizational (Oh, Wang, & Mount, 2011) and psychopathology literatures (Miller & Lynam, 2015; Oltmanns & Turkheimer, 2009). With respect to the latter, in a 7.5-year longitudinal study of 85 clinically depressed outpatients, O and S data contributed about equally to the prediction of depressive symptoms and overall adjustment, but only O data afforded incremental value in the prediction of psychosocial functioning. In addition, in a study of 1,080 military recruits, O data outperformed S data for 8 of the 10 DSM personality disorders in forecasting premature discharge from the military; for narcissistic and obsessive-compulsive personality disorders, the difference between O and S data did not attain statistical significance (Fiedler, Oltmanns, & Turkheimer, 2004). Although we are unaware of studies examining the incremental validity of O data above and beyond interview data or vice versa, these results demonstrate that O data afford useful information that is routinely overlooked by merely asking individuals about their personality and personality disorder features.

The personality literature further suggests that O data are especially likely to afford incremental validity beyond S data when the traits being assessed are highly evaluative in nature (e.g., dishonesty, manipulativeness), as such traits are likely to be underreported by target individuals themselves (Vazire & Carlson, 2011). It perhaps goes without saying that these are among the very traits that should most concern citizens. In addition, not surprisingly, although this literature suggests that O data may be especially informative beyond S data when the traits in question are readily observable (e.g., extraversion) or highly evaluative in nature (e.g., interpersonal antagonism), they may be less helpful for largely unobservable traits, such as neuroticism (Beer & Vazire, 2017). Extending this line of work to psychopathology, it seems plausible that O data may be especially helpful for more externalizing and less socially desirable forms of psychopathology, such as antisocial personality disorder and NPD, whereas S data may be especially helpful for more internalizing forms, such as mood and anxiety disorders.

Another key advantage of O data that has been largely overlooked by Goldwater Rule proponents is their capacity to capitalize on the power of statistical aggregation by drawing on information from multiple informants (Oh et al., 2011). Such aggregation tends to minimize random error by cancelling out the idiosyncratic perspectives of individual observers and converging on their shared perceptions (Rushton, Brainerd, & Pressley, 1983). This point bears on the Goldwater Rule given that the substantial majority of in-person psychiatric examinations hinge on the evaluation of only a single interviewer.

Family history data

Although not acknowledged by Goldwater Rule advocates, research in clinical psychology and psychiatry has long made profitable use of family history data. As distinguished from family study data, in which family members (probands) are interviewed directly for the presence or absence of psychiatric disorders, family history data derive from second-hand structured interviews of family members, typically first-degree relatives, or probands (Andreasen, Endicott, Spitzer, & Winokur, 1977). Considerable data attest to the validity of family history data for many psychiatric diagnoses, strongly suggesting that direct interviews are not invariably necessary to obtain accurate information concerning these diagnoses in probands. In a meta-analysis of 13 studies comparing diagnoses established using family history methods with those established using family study methods as “quasi-criteria” (Hardt & Franke, 2007) revealed significantly higher levels than chance across all mental disorder categories. The odds ratios (ORs) were much higher for some disorders, such as schizophrenia (OR = 148) and bipolar disorder (OR = 64), than for others, such as depression (ORs ranged from 3 to 37), perhaps reflecting the greater observability of the core features of the two former conditions.

The error rates of family history data tend to be asymmetrical, probably because family members are at times unaware of their relatives’ subtle psychiatric symptoms, past psychiatric history, or both. When family study data are used as quasi-criteria, family history data tend to be characterized by a significant number of false negatives, although their false-positive rate is generally low (Thompson, Orvaschel, Prusoff, & Kidd, 1982). That is, when researchers using family history data conclude that certain diagnoses are present, these diagnoses are typically accurate; in contrast, when they conclude that certain diagnoses are absent, they are frequently mistaken. To the extent that these findings bear on the Goldwater Rule, they indicate that the principal risk of second-hand interviews is underestimating individuals’ genuine psychopathology, not in concluding erroneously that they possess psychopathology. In any case, research demonstrates that diagnoses of substantial validity can often be derived by interviewing relatives of target individuals.

The validity of videotaped and audiotaped interview information

The Goldwater Rule presumes that an in-person interview is needed to yield accurate diagnostic information; as some authors (e.g., Kroll & Pouncey, 2016) have observed, this stricture implies that videotaped interviews cannot be used in lieu of in-person interviews to formulate diagnostic opinions regarding public figures. Nevertheless, there are substantial grounds to question this assumption. For example, in the classic U.S.-U.K. cross-national study, diagnosticians in both countries rated videotaped interviews of the same patients, leading to the well-known conclusion that the same individuals who were diagnosed with schizophrenia in America were commonly diagnosed with manic-depression (now called bipolar disorder) in England (Cooper et al., 1972). The results of this study were considered sufficiently compelling that they contributed to a substantial narrowing of the diagnostic criteria for schizophrenia in DSM–III and its progeny (Andreasen & Flaum, 1991; see also Kroll & Pouncey, 2016). Data also point to high levels of interrater reliability when comparing audiotaped structured interviews with in-person structured interviews. For example, when comparing audiotaped versus in-person Structured Clinical Interview for DSM–IV (SCID) diagnoses in a mixed sample of inpatients and outpatients, Lobbestael, Leurgans, and Arntz (2011) found mean interrater reliability (Kappa) coefficients of .71 for Axis I disorders and .84 for Axis II disorders; they did not, however, compare the construct validity of in-person and audiotaped interviews. Nevertheless, these and numerous other findings (e.g., Kaufman et al., 1997) call into serious question the presumption that in-person interviews are needed to achieve reliable diagnoses.

Summary

Data from the personality, psychopathology, and industrial-organizational literatures, little of which has been cited in psychiatric discussions, converge on a shared conclusion: Unstructured interviews, which are far and away the modal means of acquiring information regarding psychiatric diagnoses in clinical practice (Aboraya, 2009), tend to be of lower reliability and validity than are structured interviews. Furthermore, unstructured interviews can in some cases actually harm the validity of clinical judgments (Dana et al., 2013), although it is unknown whether this conclusion also applies to psychiatric interviews. The notion that diagnoses cannot be made in reference to individuals one has not interviewed, even in the presence of plentiful data from many other information sources, accords false primacy to interview-based approaches while erroneously underestimating the utility of non-interview-based sources of data, such as questionnaires, interviews with knowledgeable informants, extended behavioral observations; and official records. These data were almost surely far less available at the time of the creation of the Goldwater Rule several decades ago but are now often readily obtainable by any psychologist with access to the internet. More broadly, substantial bodies of data in the personality and psychopathology literatures demonstrate that sources of data other than those obtained from direct interviews, including informant reports and high-quality file and videotape data, can provide diagnostically useful information. For personality and personality disorders at least, informant data in particular are of higher validity than are self-reported data (Miller & Lynam, 2015), which constitute the primary source of information from clinical interviews.

The Goldwater Rule in historical context

There are at least two ways in which the Goldwater Rule, although arguably defensible in the 1960s and 1970s, is anachronistic. First, as already observed, times have changed. As merely one example, consider the exponential growth of the internet. When Bill Clinton assumed the presidency in early 1993, there were approximately 50 websites; there are now over 1.2 billion and counting (http://www.internetlivestats.com/total-number-of-websites/). As a consequence of the massive information explosion over the past two and a half decades, psychologists can gather enormous amounts of useful archival data regarding public figures. This quantity and quality of data were unavailable in the days of the Johnson-Goldwater election.

Second, in the mid-1960s, when over 1,000 psychiatrists declared Barry Goldwater unfit for the presidency, psychiatry and even much of clinical psychology was still largely in the thrall of Freudian thinking. Indeed, many of the negative comments in Fact Magazine (1964) reflected the psychoanalytic zeitgeist. For example, one psychiatrist commented on Goldwater’s “projection” of hostility, another on his stronger unconscious identification with his mother than with his father, and still another on “his primitive instinctual feelings which are generally suppressed” (p. 39). Some psychodynamic scholars might maintain that such subjective inferences would indeed necessitate a direct interview. Yet with the advent of DSM–III (American Psychiatric Association, 1980), which introduced a theoretically agnostic model of classification that was essentially free of psychodynamic inference, the necessity of such an interview was largely obviated. When commenting on the possible presence of a psychiatric disorder, there is no longer any requirement that it comes with an accounting of its etiology.

Recommendations

In light of the research we have reviewed, we recommend that the Goldwater Rule be abandoned and that it be reformulated in two major ways (see also Ghaemi, 2016; Pies, 2018). First, we propose that although psychologists should typically refrain from offering diagnoses of public figures, they should be able to do so when these individuals hold positions of substantial power over others, as is the case for most high-profile politicians. That is, we believe that psychologists should avoid proffering diagnoses of celebrities, athletes, and the like for the mere sake of prurient interest. In sharp contrast to celebrities and athletes, candidates for and holders of high public office bear substantial responsibilities to the general public. In these cases, there are often more than adequate grounds for being able to comment on public figures’ psychiatric status given that this information may be relevant to maintaining an informed citizenry. In addition, such information could inform the decisions of politicians entrusted with making high-level appointments. ¹

Indeed, rather than conceptualize the ethical prerogative of psychologists and other mental health professionals as a “duty to warn” (e.g., Post, 2015), which may imply erroneously that the presence of a given psychiatric diagnosis is highly predictive of politicians’ problematic behaviors in office, we instead propose conceptualizing this right as a “duty to inform” (Glancy, Regehr, & Bryant, 1998; Westen, 2017). That is, when psychologists (a) possess considerable expertise concerning a specific disorder and (b) have thoroughly examined a large body of O and L data on a public figure, they can with reasonable justification inform the public about this figure’s likely state of mental health.

Second, although we believe there are cases in which psychologists can ethically inform laypersons regarding a candidate’s or politician’s mental health, we believe that this freedom should not extend to judgments regarding a public figure’s fitness for office. Rarely if ever are mental illnesses absolute disqualifications for serving in office, and as we have already observed, many politicians with clear-cut mental illnesses have surely served capably in positions of high authority. An analogy to the insanity defense may be apt here. Although psychologists and psychiatrists in insanity cases are frequently called on to render judgments regarding a defendant’s likely psychiatric state at the time of the crime, the 1984 Federal Rules of Evidence prohibit them from weighing in on the ultimate issue—that is, the question of whether the defendant is legally sane or insane (Slobogin, 1989). The rationale is that sanity and insanity are ultimately legal, not psychiatric, matters of criminal responsibility that can only be informed, but not fully adjudicated, by psychological or psychiatric expert testimony. Similarly, we believe that psychologists should refrain from offering judgments regarding the ultimate issue of individuals’ fitness for public office. Such fitness is a function of a plethora of factors, both psychological and situational, and cannot be determined by means of knowledge of an individual’s mental health status alone.

Third, when psychologists elect to weigh in on the diagnostic status of public figures, they should do so judiciously and with an explicit acknowledgement of (a) the nature and quality of data sources they have examined (what specific information have they consulted?), (b) the strengths and weaknesses of these sources, and (c) the nature of their background and training in psychiatric diagnosis. When a psychologist has never personally examined a political candidate, the public must be informed that his or her diagnostic formulations were made without the benefit of this information. Such recommendations, we contend, are broadly consistent with the spirit of Standard 9.01 of the APA Ethics Code in requiring psychologists who cannot directly examine a person about whom they are reporting to “clarify the probable impact of their limited information on the reliability and validity of their opinions, and appropriately limit the nature and extent of their conclusions” (http://www.apa.org/ethics/code/). Although an in-person interview of a public figure is generally desirable for establishing the diagnosis of a public figure, it is not invariably necessary, especially when plentiful high-quality data from alternative sources, such as extensive biographical data, large amounts of videotaped data, and data from informants who have interacted closely with this individual for years or decades, are available.

Simply put, there is no gold standard for psychiatric diagnosis, as all diagnoses are fallible and provisional inferences regarding constructs—that is, latent entities that can only be inferred indirectly (Cronbach & Meehl, 1955; Faraone & Tsuang, 1994; Skeem & Cooke, 2010). As such, the notion that any single source of information, be it a direct interview, a questionnaire, or an observer report, is necessary, let alone sufficient, to establish a psychiatric diagnosis is almost certainly unwarranted.

Can versus should

Finally, we should be clear about what we are not saying. Merely because we maintain that the Goldwater Rule is difficult to justify on scientific grounds does not imply that we believe that advancing diagnoses of public figures is generally advisable. To the contrary, we believe that psychological scientists should have an extremely high threshold for proffering such diagnoses. They should do so only when (a) the public figure in question holds, or has the potential to hold, a position of power that bears implications for public welfare and security and (b) they have carefully scrutinized large quantities of high-validity information, ideally from S, O, and L data. When such conditions are not met, which will usually be the case, withholding judgment on public figures is the prudent choice.

Psychologists who elect to comment on the mental health of public figures should do so with a keen awareness of the risks, including the potential harm to the reputation of public figures and to the reputation of the field at large, as well as with a full appreciation of the lengthy history of the abuse of psychiatric diagnoses. Only a matter of a few decades ago, such dubious labels as “sluggish schizophrenia” and “reformist delusions” were used to malign and in some cases imprison dissidents in the Soviet Union (Fulford, Smirnoff, & Snow, 1993), and comparable abuses appear to be occurring in contemporary China (van Voren, 2010). Psychiatric diagnoses are ripe for misuse, and psychologists should take particular precautions against such exploitation. Moreover, large bodies of research remind us that clinical judgment and decision making are hardly infallible and that diagnostic confidence rarely translates into diagnostic accuracy (Garb, 2005). When it comes to rendering opinions on the mental health of public figures, humility should be the watchword. Many if not most of the psychiatrists who denigrated Goldwater’s character on the basis of inadequate information ran afoul of these crucial caveats, and they opened themselves to justified criticism.

Certainly, psychological scientists and other mental professionals should exercise extreme discretion when commenting on the mental health of politicians and others entrusted with important societal responsibilities. Still, they on occasion possess pertinent and important knowledge to impart to laypersons and policy makers in this regard. So long as they draw on an adequate body of well-validated assessment data, openly acknowledge the limitations of their evidence base, and are explicit about their training and credentials, they should not feel hamstrung in doing so.