Psychological Perspectives on Interrogation

Torture

Several researchers have evaluated claims of “effectiveness” with respect to the use of torture or coercion to overcome resistance and yield compliance with an interrogator’s requests (Arrigo, 2004; Bell, 2008; Costanzo & Gerrity, 2009; Hartwig, Meissner, & Semel, 2014; Pfiffner, 2014; Rejali, 2009). The consensus view is that the use of torture often fails as an effective means of successfully moving a resistant subject into a state of cooperation that may yield information of criminal or intelligence value. In 2006, the U.S. Intelligence Science Board conducted a systematic review of research underlying U.S. military and intelligence interrogation practices, including the use of torture (Fein, Lehner, & Vossekuil, 2006). It concluded that the preponderance of studies and reports regarding the use torture tactics “weigh against their effectiveness” and that both theory and related research “suggest that coercion or pressure can actually increase a source’s resistance and determination not to comply” (p. 35).

Cases in which torture does appear to “work” typically reflect a form of compliance in which the subject simply confirms an interrogator’s belief and therein yields a confession statement of dubious accuracy. For example, the SSCI’s (2014) Study on the Central Intelligence Agency’s Detention and Interrogation Program examined 20 of the most notable “successes” offered by the CIA’s program, concluding that there was little or no relationship between the use of “enhanced interrogation techniques” and the elicitation of intelligence. Further, detainees were shown to have provided false information or offered speculations in response to the application of torture, creating challenges to the use of any “intelligence” gathered in such contexts. We note that several former U.S. government officials and contractors who have acknowledged their direct or indirect involvement in the CIA’s interrogation program dispute the SSCI findings with respect to the effectiveness of coercive practices (Mitchell & Harlow, 2016; Morell & Harlow, 2015; Rodriguez, 2012). Their arguments, however, lack any measure of scientific rigor.

Although it is impossible to ethically evaluate the effectiveness of torture as an interrogation tactic, anecdotal reports and case studies of the operational use of such tactics offer only limited support for its purported efficacy in decreasing resistance and increasing compliance with an interrogator’s requests (cf. Arrigo, 2004; Bell, 2008). For example, Stockdale (2001) estimated that more than 95% of U.S. personnel successfully resisted torture by the North Vietnamese, whereas Hoffman (1977) documented that torture by Nazi interrogators failed to gain compliance from high-level officials allegedly involved in plots to assassinate Adolf Hitler. Silverman (2001) examined more than 600 cases of judicial torture in France from the 1500s to the 1700s, finding that torture produced confessions only 5% to 33% of the time. British interrogators similarly achieved success in only about 30% of interrogations at the London Cage using a variety of tactics that included torture (Cobain, 2005). Andrews (2001) describes experiments conducted by Nazi scientists in which they subjected concentration camp inmates to severe pain, extreme temperatures, and various drugs. Despite the variety of methods examined, they found no reliable means of gaining compliance. Finally, in a study supported by the U.S. Air Force, Biderman (1960) assessed the influence of interrogation tactics used by Communist Korea and China against more than 200 U.S. military prisoners of war. Biderman concluded that cooperation was more likely “in situations in which the ex-prisoner reported he was not subjected to overt threats or violence than in situations in which such coercion was reported” (pp. 143–144).

Recent systematic interviews with military and intelligence interrogators, including those who interrogated high-value targets, confirm these findings—professionals frequently reference the use of torture as the least effective technique for gaining cooperation, with such tactics seen as more often producing resistance (Narchet, Russano, Kleinman, & Meissner, 2016; Russano, Narchet, Kleinman, & Meissner, 2014). It is notable that these findings are consistent with the views of Markus Wolf, chief of the East German foreign intelligence service during much of the Cold War, who asserted that “interrogation . . . should serve to extract useful information from the prisoner . . . not to exact revenge by means of intimidation and torture” (Wolf & McElvoy, 1997, pp. 261–262).

Psychological theory offers a perspective on why the use of torture or physical coercion so often fails to engender cooperation. For example, terror management theory demonstrates that people become more extreme in their belief systems (and therein resistant to change) when they are reminded of their mortality (cf. Burke, Martens, & Faucher, 2010; Hirschberger & Ein-Dor, 2006). Reactance theory also suggests that resistance is greater when one’s freedom of action is constrained (Brehm, 1966), and recent modifications of this theory offer a nuanced perspective on the conditions under which the use of coercion or torture might yield compliance versus increased resistance (Miron & Brehm, 2006). First, motivation to resist appears to be key: When an expectation of success in resistance is maintained, a subject’s resistance will remain strong (Seligman, 1975; Wortman & Brehm, 1975). Second, if the subject lacks knowledge regarding the difficulty of maintaining resistance, he or she will continue to “mobilize as much energy as the goal of restoring freedom is seen to be worth” (Miron & Brehm, 2006, p. 6). Finally, resistance can be maintained even when the difficulty of restoring the freedom is perceived to increase, though this motivation to maintain resistance will decline as the subject’s perceived ability to reinstate freedom is lost.

Together, theory and data suggest that a subject’s resistance to torture and physical coercion is likely determined by the perception of his or her ability to restore the freedom and to therein control his or her situation, and this perception will likely vary across individuals and situations. Khalid Sheikh Mohammed (KSM), the alleged mastermind of the 9/11 attacks who was waterboarded by the CIA at least 183 times, was said to have known the limits of the procedure—counting the seconds with raised fingers until it was completed. KSM maintained a high degree of resistance throughout his interrogations despite the repeated use of tactics regarded as torture. Even when he ultimately provided information to his interrogators, KSM remained uncooperative—as the SSCI (2014) determined that the information he provided was false.

As noted above, a small percentage of individuals subjected to torture will comply with an interrogator’s requests. Although torturers often equate the pain or intensity of a technique with its “efficiency” (see Rejali, 2009), research suggests that the perception of pain is subjective and can vary as a function of gender, culture, and life experience (cf. Hadjistavropoulos & Craig, 2004). Further, pain intensity has been shown to be influenced by a variety of cognitive, emotional, environmental, and behavioral factors (see Gatchel, Peng, Peters, Fuchs, & Turk, 2007). Of relevance to the influence of intense pain in torture, prior research has shown (a) that one’s beliefs about the extent of pain can influence perceived intensity (Arntz & Claassens, 2004), (b) that catastrophizing (“an exaggerated negative ‘mental set’ brought to bear during painful experiences”; Sullivan et al., 2001, p. 52) is associated with an increased perception of pain intensity, and (iii) that perceived control over pain can influence both perceived intensity and tolerance (Samwel, Evers, Crul, & Kraaimaat, 2006). As suggested by Arrigo (2004), an interviewee’s motivational interpretation of his or her physical sensations (and the potential reactance one draws from it) could also likely determine the influence of torture in producing cooperation or increasing resistance.

Accusatorial approaches

As described previously, harsh interrogation tactics also have a history within the U.S. criminal justice system, where “third degree” approaches involving physical abuse; incommunicado detention; deprivation of food, sleep, and medical attention; and explicit threats of harm were regularly used against subjects. Reforms in the United States during the 1930s and 1940s, including a Presidential Commission’s “Report on Lawlessness in Law Enforcement,” eventually diminished their use and led to the development of accusatorial interrogation approaches that emphasized psychological manipulation (Leo, 2008; Meissner & Albrechtsen, 2007). This psychological approach is most popularly embodied in the Reid Technique of interrogation, first formalized by Inbau and Reid (1962) and highly influential within U.S. federal, state, and local law enforcement to this day (Inbau, Reid, Buckley, & Jayne, 2013; see Kelly & Meissner, 2015). Over the past decade, however, psychological and socio-legal research has demonstrated that the types of trickery and deception regularly used by law enforcement in some countries can produce a significant cost to social justice—they increase the likelihood of eliciting false confessions by the innocent (Kassin, 1997; Kassin et al., 2010; Lassiter & Meissner, 2010; Meissner et al., 2014).

Accusatorial tactics operate by manipulating a subject’s beliefs about the relative consequences of confession and denial and often involve the use of emotional “themes” that diminish feelings of guilt and lessen perceived culpability. Trickery and deception are the foundation of these approaches: In some countries, police can lie to suspects, present false evidence of their guilt, and manipulate their expectations regarding the potential consequences associated with confession (Kelly & Meissner, 2015; for critiques of this approach, see Kassin, 1997, 2006; Kassin & Gudjonsson, 2004; for social-psychological perspectives on the ensuing process of influence, see Davis & Leo, 2012; Kassin, 2015; Madon, Guyll, Scherr, Greathouse, & Wells, 2012).

Though lacking many of the physical elements of torture, the powerful effects of accusatorial approaches have been shown to produce both truthful confessions from the guilty and false confessions from the innocent (Kassin et al., 2010; Lassiter & Meissner, 2010; Meissner et al., 2015; Meissner et al., 2014). Such tactics as presenting subjects with false incriminating evidence (Kassin & Kiechel, 1996; Nash & Wade, 2009; Perillo & Kassin, 2011) and minimizing the potential consequences associated with confession have been shown to increase the likelihood that both guilty and innocent individuals will confess (Horgan, Russano, Meissner, & Evans, 2012; Russano, Meissner, Narchet, & Kassin, 2005). Further, investigators who believe in a subject’s “guilt” have been shown to conduct longer, more psychologically coercive interrogations, leading to a cycle of behavioral confirmation that encourages confession, particularly by the innocent (Kassin, Goldstein, & Savitsky, 2003; Kassin, Meissner, & Norwick, 2005; Meissner & Kassin, 2002, 2004; Narchet, Meissner, & Russano, 2011).

Although accusatorial tactics are largely viewed as psychologically manipulative and therein successful at achieving compliance, a recent field study by Kelly, Redlich, and Miller (2015) has suggested that the use of certain accusatorial tactics, such as the presentation of evidence, emotion provocation, and confrontation, can actually enhance resistance, leading to a subject’s denials and refusal to cooperate. Goodman-Delahunty, Martschuk, and Dhami (2014) have similarly found that the use of accusatorial tactics failed to enhance either cooperation or disclosure in a sample of high-value interrogations and that an accusatorial framing of evidence reduced cooperation (see also Walsh & Bull, 2012, 2015).

In short, accusatorial tactics can, at times, induce compliance with an interrogator’s request for an admission or confession—however, the diagnostic value of this information is often diminished (Kassin et al., 2010; Meissner et al., 2014). Minimization tactics that seek to lessen a subject’s perceived culpability (and therein the consequences associated with confession) are particularly problematic for inducing false confessions by the innocent (Horgan et al., 2012). Recent research also suggests that certain maximization strategies (involving evidence presentation, emotion provocation, and confrontation) can produce resis-tance and therein diminish cooperation (Kelly et al., 2015). Although U.S. law enforcement has embraced the use of these methods since the 1960s (Kelly & Meissner, 2015), the aforementioned research (together with recent studies on alternative, rapport-based approaches described later) is only now beginning to influence the modern practice of interrogation in some countries (see Balko, 2017; Kolker, 2016).

Rapport-based, information-gathering approaches

In Great Britain, public response to a spate of miscarriages of justice involving false confessions became the catalyst for change, leading to the Royal Commission on Criminal Procedure in 1981. The commission determined that the factors producing these miscarriages of justice were occurring in the interrogation room—police frequently relied on techniques that were both physically and psychologically manipulative, and they lacked an existing protocol or training for noncoercive interrogations (Irving, 1980; Irving & Hilgendorf, 1980). To address this situation, the Police and Criminal Evidence Act (PACE) of 1984 was created, which expressly limited the use of psychologically manipulative tactics and required that all interrogations be audio recorded (Bull & Milne, 2004a). In 1993, the Royal Commission on Criminal Justice further reformed British interrogation methods by proposing the PEACE model, developed by a team of experienced detectives, informed by the available psychology (British Psychological Society, 2016). Each letter of the acronym “PEACE” represents a phase of interrogation to which investigators should adhere. In the “preparation and planning” phase, interrogators focus on organizing evidence and constructing a plan for the interview. During the “engage and explain” phase, the goal is to build rapport and to make the interviewee aware of the purpose of the interview. The third phase, “account,” is the core of the interview. For compliant interviewees, investigators are encouraged to use the Cognitive Interview (Fisher & Geiselman, 1992; Memon et al., 2010). For noncompliant interviewees, officers are instructed to use Conversation Management (Bull & Milne, 2004b; Mortimer & Shepherd, 1999) to encourage cooperation and discussion.

Contrary to an accusatorial style of interrogation, this approach has the goal of “fact finding” rather than that of obtaining a confession, and investigators are not permitted to lie to interviewees. After initiating the interview, subjects are encouraged to provide a complete account of their involvement or relation to the crime, and they are encouraged to speak freely, while close-ended questions are kept to a minimum (Bull & Milne, 2004b; Mortimer & Shepherd, 1999). Once the individual has provided his or her narrative and then been questioned about this (and other issues) while being challenged with evidence information known to the interviewer, the investigator offers the opportunity to correct any discrepancies (the “closure” phase). Finally, the investigator again compares the interviewee’s statements to evidence, tries to clear up any remaining inconsistencies, and draws conclusions.

Over the past decade, researchers have begun to systematically evaluate the efficacy of rapport-based, information-gathering approaches such as the PEACE model (Meissner et al., 2015; Meissner, Russano, & Narchet, 2010). Preliminary evidence suggests that utilizing such an approach not only reduces the likelihood of false confessions but also increases the elicitation of accurate information (Meissner et al., 2014). Field studies suggest that when investigators properly demonstrate the elements of the PEACE model, interviewees are more likely to provide complete accounts of their crimes (Walsh & Bull, 2010b) and investigators are able to overcome initial denials (Walsh & Bull, 2012). Laboratory studies also demonstrate that, when compared with accusatorial approaches, information-gathering approaches facilitate more cooperation and yield more accurate information from interviewees who are initially resistant (Evans, Meissner et al., 2013).

Central to the success of an information-gathering approach is the development of rapport—defined as “a positive and productive affect between people that facilitates mutuality of attention and harmony” (Bernieri & Gillis, 2001, p. 69; for a review, see Vallano & Schreiber Compo, 2015). In fact, surveys and interviews of law enforcement demonstrate wide support for the use of rapport and relationship-building approaches (Kassin et al., 2007; Redlich, Kelly, & Miller, 2014; Russano et al., 2014; Vallano, Evans, Schreiber Compo, & Kieckhaefer, 2015). Evaluations of rapport development in the interrogation point to its critical role in both developing cooperation and eliciting accurate information. For example, Holmberg and Christianson (2002) interviewed 83 sexual offenders about their interrogation, finding that interviews that involved an empathic and humanitarian perspective were associated with the elicitation of full confessions, whereas those who were viewed as judgmental and dominant were more likely to elicit denials (see also Kebbell, Alison, Hurren, & Mazerolle, 2010). Kelly et al. (2015) coded a sample of U.S. law enforcement interrogations and found that rapport and relationship tactics were associated with an interviewee’s cooperation and ultimately with confession to the crime (see also Walsh & Bull, 2012). Through systematic interviews of high-value detainees and interrogation professionals, Goodman-Delahunty et al. (2014) demonstrated that rapport strategies were significantly associated with (early) disclosure and the elicitation of accurate information. And finally, in what is likely the most complete evaluation of rapport and relationship building in an interrogation context, Alison, Alison, Noone, Elntib, and Christiansen (2013; Alison et al., 2014b) evaluated five facets of rapport (autonomy, adaptation, evocation, empathy, and autonomy) drawn from the motivational interviewing literature (Miller & Rollnick, 2013) in a sample of 418 separate interviews of 29 terrorism suspects in the United Kingdom. The authors found that an interrogator’s ability to exhibit both conversational rapport and adaptive interpersonal skills (Birtchnell, 2002; Leary & Coffey, 1954) was associated with a significant reduction in the likelihood of counterinterrogation (resistance) strategies by the subjects and an increase in the elicitation of investigative information. These empirical findings are consistent with the reports of several interrogators who have chronicled their experiences in questioning terrorist subjects (see Alexander & Bruning, 2008; Soufan, 2011).

How is rapport developed in an interrogative context? A cogent review of rapport tactics has been offered by Abbe and Brandon (2014). In general, active listening and positive communication skills appear to be critical to facilitate a perception of openness, to express empathy and respect, and to humanize the conversation (Alison et al., 2013; Holmberg & Christianson, 2002; Kebbell et al., 2010; Oxburgh & Ost, 2011; Walsh & Bull, 2012). Evans, Meissner et al. (2013) experimentally demonstrated that positive emotional approaches, such as self-affirmations, expressing interest, and instilling calm, significantly reduced anxiety and increased feelings of rapport in a suspect interrogation (see also Davis, Soref, Villalobos, & Mikulincer, 2016). Abbe and Brandon (2013) also note that rapport may be developed via selective principles of social influence (Cialdini, 2006), particularly by addressing interest, identity, or relational motivations (Kelman, 2006). In this context, Goodman-Delahunty and colleagues have found that liking and reciprocity were most frequently used to develop rapport in a sample of high-value detainees (i.e., subjects detained for terror-related activities) and that such tactics were significantly associated with information disclosure (Goodman-Delahunty & Howes, 2014; Goodman-Delahunty et al., 2014).

Kleinman (2006) summarizes the distinction between methods of torture, coercion, or accusatorial approaches and that of a rapport-based, information-gathering approach by conceptually describing them as “control” and “rapport” models of interrogation: “The control model would obtain information only in direct response to the specific questions posed . . . in contrast, the rapport model is more likely to obtain not only similar kinds of information, but also additional information within the scope of the source’s knowledgeability that was not necessarily addressed by the interrogator” (p. 136). Evidence is accumulating for the effectiveness of a rapport-based model for encouraging a productive exchange and minimizing reactance while offering empathy and autonomy to the interviewee. Rapport-based approaches also appear to facilitate the timely and appropriate (positive) confrontation of a subject to clarify statement-evidence inconsistencies (Alison et al., 2013; Alison et al., 2014b; Evans, Meissner et al., 2013), including facilitating an investigator’s ability to leverage evidence in a strategic manner to promote disclosure (Tekin et al., 2015; Walsh & Bull, 2015).

Conclusion

Taken together, the available research suggests that the use of torture or physical coercion fails to produce timely and accurate information from a subject; instead, such tactics are less productive and less diagnostic than their advocates might lead us to believe and also appear more likely to increase resistance by the subject. Accusatorial tactics can, at times, induce compliance with an interrogator’s request for an admission or confession—however, the diagnostic value of this information is diminished (Meissner et al., 2014). Minimization tactics that seek to lessen a subject’s perceived culpability are particularly problematic for inducing false confessions by the innocent and certain maximization strategies can produce resistance and therein diminish cooperation. In contrast, evidence is accumulating for the effectiveness of a rapport-based model for facilitating cooperation and minimizing resistance.

Torture

Experimental data concerning the effect of torture on the elicitation of information from memory do exist, showing a clear pattern that such conditions have a negative effect on memory and recall. A particularly interesting experiment was carried out with 184 Special Operation warfighters in the U.S. Army enrolled in a Military Survival School (Morgan, Doran, Steffian, Hazlett, & Southwick, 2006). Military Survival School training is one of the most rigorous forms of training experienced by special operations personnel. It is modelled after experiences of American prisoners of war in World War II and the Korean, Vietnam, and Gulf wars. The exercises that the trainees underwent included confinement, food and sleep deprivation, extreme temperature variations, and exposure to stressful interrogations under intensive and unremitting conditions. Unlike real-life interrogations in which subjects are often motivated not to report all the information they know and/or are motivated to lie, in this training the subjects’ task was not to withhold any information but to be as complete and accurate as possible in their recall. The findings revealed that the physical, psychological, and emotional pressures experienced during the exercise resulted in significant memory impairment, including an inability to accurately recall previously learned information.

In another experiment, researchers evaluated the effect of a combination of standard torture techniques such as sleep and food deprivation, as well as temperature manipulation, on individuals’ cognitive functioning (Lieberman et al., 2005). After being exposed to these stressors for a period of 5 days,, the volunteers (with an average of 9 years of active military service) showed severe impairment in their cognitive functioning, including their memory. Similarly, studies measuring the effects of sleep deprivation (e.g., Fenn, Gallo, Margoliash, Roediger, & Nusbaum, 2009; Payne et al., 2012; Ratcliff & Van Dongen, 2009) and extremes of cold and heat (e.g., Pilcher, Nadler, & Busch, 2002) have also shown that such factors have a negative effect on the recall of learned material. In his cogent review of this research, O’Mara (2009) concluded that “prolonged and sustained sleep deprivation, in part because it results in a substantial increase in cortisol levels, has a deleterious effect on memory” (pp. 497–498).

Torture is a threatening experience for subjects, and the general cognitive problems generated by these threatening conditions are problematic when conducting an interview for the purposes of eliciting accurate and comprehensive information. First, because memory retrieval is impaired under stress (see also Stawski, Sliwinski, & Smyth, 2009), subjects under threat will either retrieve fewer experiences when they are asked open-ended questions or they will have more difficulty answering specific, closed questions. Second, when threatened, subjects are likely to pay attention to the interviewer’s actions and demeanor, leading to divided attention or multitasking that is known to disrupt cognition, in general, and memory retrieval, specifically (Johnston, Greenberg, Fisher, & Martin, 1970; Rohrer & Pashler, 2003; Vredeveldt, Hitch, & Baddeley, 2011).

Another potential limitation of interviews conducted under high-stress circumstances is that a subject’s ability to monitor the quality of his or her own recollections may be impaired (e.g., Nelson et al., 1990), which can increase the likelihood of reporting false events. Such false recollections may come about because subjects base their recollections more on constructing from a schema (what might normally take place) than from the actual event or information to be recalled. Alternatively, subjects may commit source-monitoring errors and confuse information associated with one task or event with a different task or event (Johnson, Hashtroudi, & Lindsay, 1993). In sum, our cognitive machinery simply does not function well under conditions of threat or high stress.

The physiological processes associated with torture can also undermine reliable recall. O’Mara (2015), for example, applied a neuroscience perspective to the effect of torture and described how the brain reacts to fear, extreme temperatures, starvation, thirst, and sleep deprivation. All these factors severely impair the brain systems responsible for memory, mood, and cognition. In particular, “chronic, prolonged and extreme stress . . . inhibits long-term potentiation (LTP; the biological process believed to underlie memory formation in the brain) and facilitates long-term depression (the inverse of LTP) . . . [and] causes hippocampal atrophy and, hence, impairs learning in humans and animals” (O’Mara, 2009, p. 498). Further, severe stressors at the time of retrieval can lead to increased cortisol levels that impair hippocampal function, producing impaired memory recall for both semantic and episodic information (for a review, see Hoscheidt, Dongaonkar, Payne, & Nadel, 2013). Against this background, O’Mara (2015) pondered why anyone would imagine that the significant degradation in cognitive performance and mood imposed by such stressors would in some way facilitate recall, enhance memory, and improve motivation.

Accusatorial approaches

In accusatorial interviews (and perhaps also in interrogations involving the use of torture), investigators often seek confirmation of facts they believe to be true and frequently suggest themes or narratives that the subject is simply asked to verify (Meissner & Kassin, 2004; Meissner et al., 2014; Narchet et al., 2011). Unfortunately, the use of such confirmatory and leading questioning tactics (particularly when an investigator’s assumptions are wrong) can have significant negative effects both for the memory of the subject and the ultimate conclusions of an investigation.

Once exposed to misleading information after an event, subjects can begin to make systematic errors when reporting what they have experienced (Frenda, Nichols, & Loftus, 2011; Loftus, 2011), leading to the corruption or alteration of memory. The types of errors participants can make in misinformation studies include recalling the perception of nonexistent items or offering incorrect descriptions of items they previously observed (clean-shaven man was remembered as having a moustache and an individual with straight hair was remembered as having curly hair). Leading and confirmatory questioning can also facilitate the construction of false memories for events or experiences that never occurred (Loftus, 2011; Newman & Garry, 2013).

In a misinformation study utilizing the Special Operations Training School paradigm introduced in the previous section (Morgan, Southwick, Steffian, Hazlett, & Loftus, 2013), some military personnel were given inaccurate information about their interrogator following a 30-min interrogation. Although they could clearly see the interrogator during the interrogation and were interviewed only 1 h after this interrogation, a large percentage of the interviewees exposed to the postevent misinformation (and many in the control condition) falsely identified a different individual as their interrogator.

Studies have also shown that false memories can be created in a manner that leads subjects to offer false criminal accusations against another individual (Loney & Cutler, 2016) and that suggestive questioning approaches can induce subjects to generate false memories of a crime they had never committed (Shaw & Porter, 2015; though see Wade, Garry, & Pezdek, in press). The presentation of false or misleading evidence (see Nash & Wade, 2009; Wade, Green, & Nash, 2010) and suggesting that a subject has “memory problems” when he or she fails to recollect an event (Van Bergen, Jelicic, & Merckelbach, 2008) are common accusatorial interrogation tactics that also play an important role in producing false memories. Such accusatorial approaches are believed to lead a subject to “distrust” his or her own memory and therein facilitate the production of a (false) confession (see Gudjonsson & MacKeith, 1982). When used in combination with approaches that involve “shutting down denials” and preventing the subject from providing an account of the event that distances or exonerates him or her (Inbau et al., 2013), it becomes clear that accusatorial approaches are counterproductive from a memory elicitation standpoint. As discussed below, these approaches also hinder an investigator’s ability to assess credibility and instead facilitate a biased perception of deception or guilt that is independent of veracity.

Information-gathering approaches

The Cognitive Interview (Fisher & Geiselman, 1992) is a particularly effective information-gathering technique to obtain complete and accurate accounts from interviewees. In brief, the Cognitive Interview incorporates research-based principles to enhance three underlying psychological processes within an information-gathering interview: (a) the social dynamics between the interviewer and the respondent, (b) the interviewer’s and the interviewee’s cognitive processes, and (c) communication between the interviewer and the interviewee. Extensive laboratory and field testing conducted by several different laboratories around the world has shown the Cognitive Interview to be highly effective, eliciting approximately 35% to 50% more information than either typical police interviews or Structured Interviews (see Memon et al., 2010, for a meta-analysis, and Fisher, Schreiber Compo, Rivard, & Hirn, 2014, for a recent review.) A recent experimental study has also examined the effectiveness of an information-gathering approach with Cognitive Interview elements in eliciting guilty knowledge from resistant interviewees (Evans, Meissner et al., 2013). The authors demonstrated that such tactics led to significantly greater cooperation and information disclosure when compared with accusatorial tactics. Finally, research focusing on the recall by interviewers found that the interviewers’ recall of the information provided by the interviewee was enhanced when a Cognitive Interview was used (Köhnken, Thurer, & Zorberbier, 1994).

Given the robust effectiveness of the Cognitive Interview, some may wonder whether such interviews can be incorporated into harsh interrogations (e.g., torture or accusatory interviews). We believe this is not possible. The context of a harsh interrogation (the use of physical or psychologically coercive techniques aimed at gaining compliance with respect to directed responses regarding information objectives) differs markedly from that of a Cognitive Interview (a cooperative interview context in which the interviewee is recognized as holding the critical information in memory and is offered autonomy in providing it). Crucially, several elements that contribute to the Cognitive Interview’s success may be difficult or impossible to implement in—or even adapt to—a threatening context. We first address the problem from the subject’s perspective and then from the investigator’s perspective.

Conclusion

Overall, the threatening and adversarial nature of harsh interrogation is often inimical to the goals of fostering efficient cognition. Harsh interrogation will therefore reduce the likelihood that respondents will provide reports that are extensive, detailed, and accurate—even in instances where that is the respondent’s self-determined objective. If the interrogator’s goal is to seek confirmation of facts, enhanced interrogation will only serve to validate the narrow set of assumptions that an interrogator holds. Confirmatory, leading, and suggestive questioning tactics associated with accusatorial approaches are also likely to corrupt a subject’s memory and lead to false confessions (particularly if the interrogator’s assumptions about the facts are inaccurate). In this way, both harsh interrogation tactics and accusatorial approaches are counterproductive to the elicitation of extensive, detailed, highly accurate information. A research-based alternative involving rapport-based, information-gathering approaches that incorporate elements of the Cognitive Interview is more likely to yield robust and reliable information from a resistant interviewee. Further, as discussed in the next section, the use of information-gathering approaches offers a corollary benefit by enhancing the assessment of credibility in suspect interviews (Geiselman, 2012; Meissner et al., 2015).

Torture

With the exception of one study (Houck & Conway, 2015), we are not aware of any systematic studies that assess the effect of torture on a subject’s decision to lie. However, numerous real-life cases are available in which subjects decided to lie rather than tell the truth after being exposed to torture. As described above, KSM, the alleged mastermind of the 9/11 attacks, allegedly provided false information when he eventually began to talk during his waterboarding sessions (SSCI, 2014). Mark Fallon, the chief investigator of a Department of Defense task force with forward deployed elements in Afghanistan, Iraq, and Guantanamo Bay, Cuba, oversaw thousands of interrogations of terrorist suspects, including an unknown number through direct observations. He concluded that resorting to abusive techniques was likely to greatly increase the chance that subjects would lie to appease their interrogators, leading to the collection of inaccurate and unreliable information (Fallon, 2015). Fallon notes that these false leads, coerced from subjects, were often believed and that the comprehensive efforts to follow up on them resulted in a significant waste of resources.

For interrogators to distinguish between truthful and deceitful responses, they must elicit diagnostic nonverbal or speech-related cues to deceit. The problem, of course, is that physical, emotional, or psychological abuse almost certainly will have a pronounced effect on a subject’s nonverbal behavior, and this effect is likely to overshadow any small effect that the subject’s veracity status may have on his or her nonverbal behavior. The ability to detect lies by assessing speech content depends upon whether subjects exposed to harsh interrogation methods are likely to give short or long answers. Longer narratives are more likely to reveal verbal cues to deceit, as we will discuss later.

Accusatorial tactics—Anxiety and nonverbal behavior

A central tenant of accusatorial approaches to interrogation is that, given a context in which the subject is confronted by a confident investigator with significant evidence of culpability, the individual will experience anxiety, and any subsequent attempts to lie or conceal information will result in nonverbal cues of deceit. The few studies examining behavioral responses in accusatory interviews present a bleak picture regarding the veracity of this proposition, particularly regarding the “nervous responses” (e.g., gaze aversion and behavioral fidgeting) to which advocates of accusatorial approaches generally attend (Strömwall, Granhag, & Hartwig, 2004). In several experimental laboratory studies using an accusatory setting, indicators of nervous behaviors did not differentiate between liars and truth tellers (Vrij, 1995, 2006; Vrij, Mann, & Fisher, 2006). More generally, a robust literature on cues to deception similarly offers a pessimistic view on the relationship between nonverbal behavior and deception. Meta-analyses summarizing the findings of over more than 100 separate research studies conclude that nonverbal cues to deceit, particularly those promoted in interrogation training manuals (e.g., gaze aversion, shifting position, and fidgeting) are faint and unreliable (DePaulo et al., 2003; Sporer & Schwandt, 2006, 2007).

Of course, the stakes for liars (negative consequences of being disbelieved and positive consequences of being believed) are rather low in laboratory studies, and perhaps more pronounced differences in nonverbal behaviors between truth tellers and liars will emerge in high-stakes situations. Although a reasonable proposition, a recent meta-analysis showed that such differences in nonverbal behaviors are equally small in both low- and high-stakes situations (Hartwig & Bond, 2014). Liars may well display nonverbal cues indicative of anxiety during high-stakes interviews, but truth tellers are also anxious in such interviews and, consequently, display the very same cues as liars (Bond & Fahey, 1987; Evans, Meissner et al., 2013). As a strategic framework for assessing credibility, anxiety-based lie detection techniques are problematic—they lack a sound theoretical underpinning as to why truth tellers and liars would differ from each other in their anxiety-related responses (National Research Council, 2003).

Research examining people’s ability to detect deceit by observing other people’s behavior has reached a similar dead end. The fundamental nature of accusatory interviews, according to the few studies in this area, makes it difficult to accurately distinguish truth tellers from liars. In none of the lie detection studies in which an accusatory interviewing setting was employed were observers able to distinguish between truth tellers and liars (Vrij, 1994; Vrij, Mann, Kristen, & Fisher, 2007; Zimmerman, Veinott, Meissner, Fallon, & Mueller, 2010). Moreover, Vrij et al. (2007) found that observers made more false positives (false accusation of truth tellers) when observing accusatory interviews compared to information-gathering interviews. One reason for this finding is that truth tellers display nervous behaviors when falsely accused of wrongdoing (Bond & Fahey, 1987), which may make them look suspicious. Further, research has found that investigators who are trained in accusatorial interrogation methods are more likely to demonstrate a bias toward perceiving deception and guilt in forensic interviews (Kassin et al., 2005; Meissner & Kassin, 2002, 2004). This investigative bias is an important finding because once innocent interviewees are mistakenly identified as guilty, they are more likely to endure longer and more pressure-filled interrogations (Kassin et al., 2003; for a review, see Kassin, 2005), leading to an increased risk of false confessions (Narchet et al., 2011).

Bond and DePaulo (2006) used a meta-analytic approach to examine people’s ability to detect truth and lies. Across nearly 25,000 observers, studies demonstrated an average accuracy rate of 52% in correctly classifying truth tellers and liars when someone could only see (but not hear) the target person, whereby a 50% accuracy rate would be obtained by chance alone. These accuracy rates are significantly less than the 63% accuracy rate obtained when participants could only listen to (but not see) the target person (Bond & DePaulo, 2006)—suggesting, once again, that verbal content cues are more diagnostic in discriminating liars and truth tellers when compared with nonverbal behavior. Whether someone is a professional lie catcher or a layperson has no effect on accuracy, though professionals (e.g., police investigators) are more confident in their judgments than laypersons (Aamodt & Custer, 2006; Vrij, 2008).

In theory, there are two possible explanations for the low accuracy rates when assessing nonverbal behavior. First, there is little difference in the nonverbal behaviors displayed by truth tellers and liars. Second, people look for the wrong cues and fail to spot the differences that exist. In their meta-analytic lens analysis, Hartwig and Bond (2011) examined these two possibilities and concluded that people perform poorly because the behavioral differences between truth tellers and liars are too small to make the task achievable. This finding also explains why training people to detect lies by informing them about “diagnostic nonverbal cues to deceit” has hardly any positive effect (Hauch, Sporer, Michael, & Meissner, 2014).

It is a lamentable state of affairs—and an alarming indicator of how little science has been permitted to inform conventional interviewing practices—that claims regarding the efficacy of nonverbal behavior in revealing deception are so widespread. A simple internet search will produce an expansive number of popular articles expressing this idea, whereas many books also seek to convey this idea, including Lie Spotting (Meyer, 2010) and Spy the Lie (Houston, Floyd, & Carnicero, 2012). Nonverbal lie detection tools such as the Behavior Analysis Interview (BAI) (Horvath, Blair, & Buckley, 2008; Horvath, Jayne, & Buckley, 1994) and Ekman’s (1985) approach of observing facial expressions and involuntary body language, the approach utilized by the fictitious character Cal Lightman in the Fox network TV series Lie to Me, are frequently taught to practitioners in many (but not all) countries, including law enforcement, military, and intelligence personnel. There is no evidence that BAI or lie detection through observation of microexpressions work. Inbau et al. (2013) cited the Horvath et al. (1994) study to support the efficacy of the BAI; however, that field study is, according to Horvath and colleagues themselves, problematic due to the lack of ground truth (i.e., uncertainty about which of the 60 interviewees examined in the study were actually lying). A laboratory experiment testing the BAI procedure found no support for its efficacy (Vrij et al., 2006). Further undermining support for credibility assessment based on nonverbal cues, Paul Ekman (who introduced the microexpression lie detection approach) has never published data showing that it works. Porter and ten Brinke (2008) examined the relationship between microexpressions and deception in a laboratory experiment and found that microexpressions very rarely occurred (only in 14 out of 697 video fragments, or 2% of the segments examined) and that both liars and truth tellers displayed them.

Accusatory interviews also offer little opportunity for verbal cues to deception to emerge primarily because interviewees say relatively little and/or are given very few chances to speak in such interviews (Meissner et al., 2014). Of key importance here is that verbal cues are more likely to occur when interviewees are encouraged to provide larger volumes of information, as their words are the carriers of verbal cues to deceit (Vrij et al., 2007). The finding that subjects say relatively little in accusatory interviews is likely the result of U.S. interrogation training that promotes confirmatory questioning strategies and the “shutting down” of denials, while also encouraging investigators to induce anxiety and identify nonverbal signs of deception (often related to anxiety) ((Vrij & Granhag, 2007; Vrij, Granhag, & Porter, 2010).

Training manuals frequently encourage this behavioral assessment by making the claim that more than 70% of a message communicated between persons occurs at the nonverbal level (Vrij, 2014). This claim is primarily based on Mehrabian’s (1971) work, which studied the communication of single spoken words. If people say little, speech content cannot play a primary role in the exchange of information. It creates what can only be described as a vicious circle: Because the subject is given little opportunity to speak, the interviewer has few options to assess credibility by means of an evidence-based strategy (verbal cues) and, instead, embraces a method (nonverbal cues) that empirical research has disputed as reliable.

Verbal cues to deceit are typically more diagnostic than nonverbal cues to deceit (DePaulo et al., 2003; Vrij, 2008). This explains the higher accuracy rates when listening to speech compared to observing behavior reported above. It also explains why in one of the rare studies of ecologically valid, high-stakes lying/truth-telling, it was found that police officers assessing video-recorded interviews with suspects in real-life interviews achieved detection accuracy rates significantly above chance level, especially those who indicated that they relied on speech cues (Mann, Vrij, & Bull, 2004). And in a recent study, it was found that when Japanese police officers accurately judged the participants’ veracity, they were more likely to rely upon verbal than nonverbal cues (Wachi et al., 2017). It also explains why training people to detect lies by informing them about “diagnostic verbal cues to deceit” has a larger effect than informing them about nonverbal cues (Hauch et al., 2014).

Taking into account the different strategies employed by truth tellers and liars explains why nonverbal cues are less diagnostic than verbal cues to deception. Nonverbally, truth tellers and liars will try to suppress behaviors they believe appear suspicious (mostly signs of nervousness). This means that truth tellers and liars will both use the same strategy as far as their behavior is concerned (Hartwig, Granhag, Strömwall, & Doering, 2010; Vrij, Mann, Leal, & Granhag, 2010).

It is a different story for speech content (Hartwig et al., 2010; Vrij, Mann et al., 2010). A truth teller’s strategy is to “tell it all” and to give as much detail as he or she can remember, whereas a liar’s strategy is to avoid incriminating himself or herself by, for example, being vague, providing evasive information, or offering little detail (Hartwig, Granhag, & Strömwall, 2007). The different strategies used by truth tellers and liars are reflected in the content of their speech, leading to differences between truth tellers and liars, that strategic interviewing tactics, inherent to an information-gathering approach, can elicit.

Information-gathering approaches—Analyzing speech

Two studies have directly compared the influence of information-gathering and accusatorial approaches in their ability to detect truths and lies. One study demonstrated a higher accuracy rate in information-gathering interviews (Zimmerman et al., 2010), whereas the other study found no differences in accuracy rates but a higher percentage of false positives in the accusatory interviews ((Vrij et al., 2007). We addressed the negative consequences of false positives for innocent interviewees earlier (see Meissner & Kassin, 2004). In terms of nonverbal cues, no clear picture emerged between information-gathering and accusatory interviews, but this is perhaps not surprising given the generally weak relationship between nonverbal behaviors and deception more generally. What is clear is that the information provided, and therein the speech content of the interviewee, significantly increases when an information-gathering approach is employed (Evans, Meissner et al., 2013).

Recent research has demonstrated that information-gathering interviews can actually enhance the elicitation of diagnostic cues to deceit, particularly verbal cues, when specific questioning protocols are employed (Vrij & Granhag, 2012). The protocols that are particularly promising include the Strategic Use of Evidence (SUE) (Granhag & Hartwig, 2008, 2015; Hartwig, Granhag, & Luke, 2014), cognitive credibility assessment (Vrij, Fisher, & Blank, 2017; Vrij, Fisher, Blank, Leal, & Mann, 2016; Vrij, Granhag, Mann, & Leal, 2011), and the Verifiability Approach (Nahari, Vrij, & Fisher, 2014a, 2014b). A common theme across these questioning techniques is that they aim to make lying a more difficult task.

SUE is premised on the operational presumption (supported by experimental research) that truth tellers are forthcoming in interviews, whereas liars do not wish to be linked to incriminating evidence and thereby use an “avoid and escape” strategy. The core of the SUE technique is to ask questions related to the evidence without specifically mentioning the evidence—for example, “When you were in the shopping mall, did you visit the book store?” (not referring to the closed-circuit television [CCTV] evidence that the person visited the book store). According to a meta-analysis of the SUE technique (Hartwig, Granhag et al., 2014), liars are more likely than truth tellers to provide a statement that contradicts the evidence (statement-evidence inconsistencies; e.g., denying having been at a certain place at a certain time). Further, when liars begin to realize during the interview that investigators possess evidence (such as the CCTV evidence), they will attempt to change their statement so that it will accommodate this evidence (e.g., from a denial to having visited the book store to an admission to having done so), resulting in more within-statement inconsistencies (Hartwig, Granhag et al., 2014).

The cognitive credibility assessment approach contends that certain instructions can be more difficult to follow for liars than truth tellers. This technique comprises three key elements. First, lying is often more difficult than truth telling in interview settings (Christ, Van Essen, Watson, Brubaker, & McDermott, 2009), and investigators can exploit this by making the interview setting more difficult by making additional requests of the interviewee designed to increase “cognitive load.” If lying already requires more cognitive resources than truth telling, liars will have fewer cognitive resources left over to deal with such additional requests. For example, when interviewees were asked to recall their story in reverse order—a difficult task—lie detection was better than when they recalled their stories in chronological order (Evans, Michael, Meissner, & Brandon, 2013). In another study, some truth tellers and liars were asked to squeeze a spring-loaded handgrip as long as possible—an exercise which makes people fatigued over time—whereas other truth tellers and liars did not have to do this. Under these circumstances, reaction times were slower for liars than for truth tellers (Debey, Verschuere, & Crombez, 2012).

The second element of the cognitive credibility assessment approach encourages interviewees to say more by, for example, using a “model statement” of a detailed response. According to social comparison theory (Festinger, 1954; see also Cialdini, 2006), in the absence of objective information, individuals will compare themselves to others. In interview settings, interviewees are often uncertain about what is required of them—for example, with respect to the amount of information they need to provide (Vrij et al., 2014). In such contexts, individuals are likely to use other sources as a point of reference (Lawrence, 2017). In this case, providing interviewees with a detailed model statement leads them to recognize that what they planned to provide is less detailed than what the interviewer is expecting from them (Ewens, Vrij, Leal, Mann, Jo, Shaboltas et al., 2016). This results in truth tellers adding more plausible detail to their narratives (Leal, Vrij, Warmelink, Vernham, & Fisher, 2015). By comparison, liars lack the necessary imagination or creativity to add the same amount of plausible detail as truth tellers, or they may be reluctant to provide the additional information given that it may expose possible deception to investigators.

The third element of the cognitive credibility assessment approach is to ask unexpected questions. Liars prepare themselves for interviews by thinking of plausible answers to possible questions (Hartwig et al., 2007). The difficulty liars face is that they do not know what questions will be asked. Investigators can exploit this by asking a mixture of questions that liars have likely expected along with those that are likely unexpected yet relevant to the given context (e.g., questions about the planning of activities). Typically, truth tellers and liars provide the same amount of detail when answering expected questions, whereas liars are less detailed than truth tellers when answering unexpected questions (Knieps, Granhag, & Vrij, 2013).

A meta-analysis of 38 studies examining “cues to deceit” revealed that the cognitive credibility assessment approach was more effective in eliciting diagnostic cues to deceit (e.g., lack of detail, plausibility, and consistency) than a standard approach (Vrij et al., 2016). Similarly, a meta-analysis of 14 studies examining “accuracy rate” in distinguishing liars and truth tellers (accuracy rates based on human observers and computer software combined) revealed a superior accuracy rate in the cognitive credibility assessment approach (71%) compared to the standard approach (56%) (Vrij et al., 2017). Finally, in a training study, experienced law enforcement personnel in England and Wales successfully increased their ability to distinguish between truth tellers and liars after using the cognitive credibility assessment approach ((Vrij, Leal, Mann, Vernham, & Brankaert, 2015).

The Verifiability Approach is based upon different strategies truth tellers and liars employ in investigative interviews (Granhag & Hartwig, 2008). Truth tellers are inclined to be open and to tell all they remember about their activities. In contrast, liars are motivated to withhold key information from the investigator, particularly information they believe incriminates them or information that could reveal that what they have provided is false. If liars are motivated to omit information that could reveal their deception, they will be especially likely to avoid details that an investigator could check—so-called verifiable details (e.g., “I phoned my friend John at 10:30 this morning”). The Verifiability Approach encourages investigators to look for details that can be checked: (a) activities carried out with or witnessed by identifiable or named persons who the investigator can consult; (b) activities that, according to the interviewee, can be shown on CCTV cameras; or (c) activities involving technology that can be traced (e.g., the use of debit or credit cards, mobile phone, tablets, or computers). Research confirms that truth tellers include more verifiable details into their accounts than liars do (Nahari et al., 2014a), and this effect can be strengthened if interviewees are asked prior to the interview to include details that an investigator can check (Harvey, Vrij, Nahari, & Ludwig, in press; Nahari et al., 2014b).

All of the techniques described in this section are situated within an information-gathering context. Whether they could also be implemented in interrogative contexts that induce torture or apply accusatorial approaches will depend on how truth tellers respond in such contexts. The techniques should work best when truth tellers have a vivid and detailed memory of the experiences about which they are interviewed. As we saw above, truth tellers’ memory could be negatively impaired when they are stressed, which is more likely to happen in torture and accusatorial interviews than in information-gathering interviews. The techniques are also more efficient when truth tellers receive the opportunity and encouragement to provide a complete account of their experiences. The techniques are thus likely to be less effective in interview settings in which investigators offer themes and narratives that interviewees simply confirm or deny and in which interviewees are given few chances to speak (e.g., accusatorial interviews).

Conclusion

Deception studies examining torture or accusatorial tactics are rare; however, the available evidence shows that they are not beneficial in terms of cues to deceit and lie detection. They focus on analyzing nonverbal cues to deceit, which research has shown to be faint and unreliable. Analyzing speech content is the most reliable method of lie detection; however, interviewees must give extensive narratives for verbal cues to deceit to occur, something that is unlikely to occur in harsh interrogations. Information-gathering interviews are the preferred method for verbal lie detection and the ability to detect deceit. They can be further enhanced by introducing cognitive-based interview protocols.