Juveniles and Adults Differ in Their Beliefs about Cues to Deception and Strategies during a Hypothetical Police Interview

Police officers are often trained to use the Behavior Analysis Interview (BAI) to detect deceit, but it is based on faulty indicators of lying that may be especially problematic for juveniles due to developmental immaturities. Juveniles, young adults, and adults were assigned to guilt or innocence conditions, read a criminal scenario, and self‐reported their likelihood of providing truthful and deceitful responses during a hypothetical BAI. All participants indicated they would give more truthful than deceptive responses. Guilty participants reported more use of strategies to appear innocent, while innocent participants said they would behave naturally. Juveniles were more likely to choose deceitful responses and say they would use strategies to appear innocent during a police interview but endorsed fewer stereotypical cues of deception compared to adults. Juveniles may not recognize how certain behaviors could be seen as cues to deception, which could put them at risk of being misidentified as guilty.

Keywords: adolescence; behavior analysis interview; deception detection; juveniles; reid technique

The Reid Technique (Inbau et al., [33]) is among the most widely implemented interrogation trainings for law enforcement in the United States (Meyer & Reppucci, [51]) and is frequently used with both juvenile and adult suspects (Cleary & Warner, [7]). The Behavior Analysis Interview (BAI) is the pre‐interrogation interview protocol of the Reid Technique. BAI interviewers rely on methods of detecting truth‐ and lie‐telling behaviors to determine if a formal interrogation of the suspect is necessary; those judged as deceitful are moved into an accusatory interrogation phase where the goal is to obtain a confession (Inbau et al., [33]). The basis of this initial judgment of deceit in the interview is a crucial step in the investigative process, as any error lays the groundwork for an innocent individual to be subjected to further accusatory and coercive interrogation tactics that could lead to a false confession (Gould & Leo, [22]; Meissner & Kassin, [50]; see also Scherr et al., [60]). This is especially true for juveniles, who are uniquely vulnerable to the pressures put on them by police officers, more susceptible to commonly used interrogation techniques, and more likely to falsely confess than adults (e.g., Cleary, [6]; Drizin et al., [13]; Redlich & Goodman, [54]). To promote accurate judgments of deceit and guilt, and therefore protect innocent juveniles from unnecessary interrogations, it is essential to investigate the BAI with juvenile suspects. More specifically, we examined adults', young adults', and juveniles' beliefs about cues to deception, self‐reported strategies, and planned behaviors during a hypothetical BAI.

The BAI consists of 15 questions developed to evoke specific verbal and nonverbal behavioral responses claimed to indicate lie‐ or truth‐telling. Nonverbal behaviors said to indicate lying include shifting in the chair, crossing legs, gaze aversion, or grooming mannerisms; verbal behaviors claimed to indicate lying include evasive, delayed, vague responses, or increased response latency (Inbau et al., [33]; see Table 1). However, the BAI has been criticized due to its over‐reliance on observations of "deceitful" behaviors that do not accurately indicate deceit (Vrij, Mann, & Fisher, [76]). In a meta‐analysis of behavioral cues in adults, DePaulo et al. ([12]) demonstrated that the verbal and nonverbal cues of deception used in the BAI are only weakly correlated with actual deception, at best.

1 TABLE Behavior analysis interview questions and claimed deceitful and truthful responses

Vrij, Mann, and Fisher ([76]) interviewed adults who were either guilty or innocent of a mock crime, finding that many of the BAI's claimed cues to deceit were more often observed in the truth‐tellers. Specifically, truth‐tellers exhibited more naïve and evasive responses, were more likely to cross their legs and shift their posture, and exhibit more nervous behaviors. There were no differences between liars and truth‐tellers for the remaining observed behaviors, demonstrating that the BAI's cues did not accurately distinguish liars from truth‐tellers (see also Vrij et al., [77]). Masip and Herrero ([45]) adapted this methodology to a hypothetical police interview by randomly assigning young adult participants to 'imagined guilt' or 'imagined innocence' conditions, presenting a criminal scenario, and asking them to imagine they would be interviewed by a police officer whom they needed to convince of their innocence. When provided a full list of responses considered by the BAI manual to show honesty and deceit, both "innocent" and "guilty" participants more often chose responses the BAI would consider truthful, suggesting an ability to manipulate these indicators to appear innocent.

Proponents of the Reid Technique claim that interviewers trained in the BAI can distinguish between truthfulness and deceitfulness 85% of the time (Inbau et al., [33]), yet extensive research shows that individuals typically cannot detect deception at better than chance rates (Bond & DePaulo, [3]). Moreover, BAI training does not typically improve lie detection skills; officers trained in the Reid Technique show increased confidence in their judgments of deception, but no increase in the accuracy of their judgments (Meissner & Kassin, [50]). While limited research has investigated police officers' ability to detect deceit in juveniles, Vrij, Akehurst, and Knight ([74]) found police officers, social workers, teachers, and laypersons believed children, adolescents, and adults typically exhibit the same cues to deception (e.g., nervousness, gaze aversion, and evasive responses). Additionally, adults were believed to be more likely to manipulate their verbal and nonverbal behaviors to appear truthful, compared to children and teens.

Using the BAI method of deception detection may be especially problematic for juveniles, who exhibit many of the claimed indicators of deceit regardless of their truthfulness or the context. Compared to adults, juveniles slouch more, make less eye contact, take longer to respond (Meyer & Reppucci, [51]), may pause or hesitate more during interviews due to limited understanding of legal processes (Birckhead, [1]), and may have a harder time managing their stress due to poorer self‐regulation abilities (Cleary, [6]). Thus, previous research suggests that juveniles are more likely to exhibit behaviors that may be misguidedly judged as indications of deceit during a police interview.

An additional criticism of the BAI is that the behaviors claimed to indicate deceit are congruent with common‐sense stereotypes of how liars behave, suggesting that guilty suspects can easily and consciously manipulate these behaviors to appear truthful to investigators (Masip et al., [47], [44]). Research from the Global Deception Research Team (2006) found survey respondents believed that liars avoid eye contact, are nervous, incoherent, and that lying can be detected through facial expressions, speech fillers, and pauses. Police officers have been shown to hold similar stereotypes when reporting the behaviors they believe to signal lying (Bogaard & Meijer, [2]; Masip & Herrero, [46]; Strömwall et al., [66]) and are taught to observe for these and related behaviors in a BAI. However, the majority of these commonly held deception cue stereotypes are inaccurate (in that they do not distinguish liars from truth‐tellers), and therefore judgments based on them are flawed (Hartwig & Bond, [26]).

In a series of studies, Masip et al. ([47], [44]) demonstrated that the BAI indicators of deceit are congruent with common stereotypes of how people behave when lying. When presented with an example BAI, lay people with no training classified the suspect who provided the BAI's claimed deceitful responses as guilty, and the suspect who provided the BAI's claimed truthful responses as innocent. Lay participants also rated the BAI's deceitful responses as significantly more indicative of deceit than the "truthful" responses. Furthermore, both novice and experienced police officers identified the BAI's claimed deceitful responses as deceitful (Masip et al., [44]). Taken together, their results demonstrate that the BAI's recommendations are simply common‐sense, stereotypical notions of what liars look like. Thus, using these stereotyped notions, guilty suspects may attempt to actively manipulate their behaviors during an interview to appear more truthful. In fact, research has shown that guilty suspects apply more strategies to appear truthful than innocent suspects do in interrogations, including avoiding signs of nervousness, planning the content of their responses, and maintaining eye contact (Hartwig et al., [28]; Hartwig & Bond, [26]; Hines et al., [30]; Masip & Herrero, [45]). On the other hand, innocent suspects may not actively manipulate their behaviors because they have an overwhelming belief that their innocence will easily be seen by interrogators (Kassin, [36]). Consequently, more innocent suspects may be mistakenly judged as deceitful by police, as officers have been shown to have a bias towards judging suspects as guilty (e.g., Kassin, [36]; Meissner & Kassin, [50]; Narchet et al., [52]).

While research has documented flaws in the use of the BAI with adults, it remains unknown how this method applies to juveniles. However, substantial research has investigated how adolescents' psychosocial and neurobiological development uniquely impacts their interactions with other facets of the legal system (Cleary, [6]; Steinberg, [63]). Adolescents' under‐developed brains impact their ability to appropriately plan, control impulses, weigh immediate rewards with future consequences, and resist peer influence (Bonnie & Scott, [4]; Cleary, [6]; Steinberg et al., [65]). Adolescents' decision‐making, self‐regulation skills, response inhibition, and cognitive flexibility are particularly impaired during emotionally charged situations (i.e., "hot cognition" situations), such as an interview or interrogation (Bonnie & Scott, [4]; Icenogle et al., [32]). Furthermore, adolescents are more susceptible to external pressures, including peer influence, compliance with authority figures, coercive tactics, and the stress associated with an interrogation (Bonnie & Scott, [4]; Gould & Leo, [22]; Scott‐Hayward, [59]). In contrast, in non‐emotional situations (i.e., "cold cognition"), the development of cognitive skills such as response inhibition, cognitive flexibility, and working memory, often plateaus around age 15 (see Icenogle et al., [32]).

As juveniles have been shown to be distinctly different from adults in their cognitive and psychosocial development, differences in lying behaviors and strategies emerge as well. Although children as young as four‐years‐old demonstrate the ability to intentionally lie, the ability to tell and maintain sophisticated lies increases with age, even into adulthood, as executive functioning matures (Evans et al., [15]; Icenogle et al., [32]; Luna et al., [42]; Steinberg et al., [65]). As children's perspective taking, emotional self‐regulation, and self‐presentation skills improve, so does their ability to manipulate expressive behaviors to disguise internal states (DePaulo, [11]). For example, at age 8, children have been shown to control their nonverbal behaviors while lying by using strategies like making eye contact, smiling, having relaxed and confident expressions, and a positive tone of voice in efforts to appear truthful (Talwar & Lee, [68]). Moreover, 11‐to‐15‐year‐olds can use strategies to appear truthful (Strömwall et al., [67]), including maintaining eye contact when lying (McCarthy & Lee, [48]), and manipulating their verbal statements to improve their credibility during an interview (Vrij et al., [75]). Further, adults are more accurate at detecting children's true statements than their false statements, but adults' veracity judgments get less accurate as the child gets older (Gongola et al., [21]), as older adolescents are likely better at concealing lies than younger adolescents (Feldman et al., [18]). Taken together, studies show that by early adolescence, juveniles are beginning to be aware of the way their behaviors can impact others and manipulate their reactions to adjust those judgments. However, it remains unclear how this awareness translates to juveniles' intentions to manipulate certain behaviors in a criminal scenario to avoid detection when lying.

Although police officers acknowledge some developmental differences, they often believe juveniles can be treated similarly to adults (Reppucci et al., [57]) and report using similar interrogation techniques with youth and adult suspects (Cleary & Warner, [7]; Meyer & Reppucci, [51]). Interrogation manuals advise similar techniques with juveniles as with adults (Feld, [17]), and the Reid Technique manual, specifically, offers little insight on how to deal with juvenile suspects, stating that the same techniques can be applied to both juveniles and adults. Additionally, Reid‐trained officers have been shown to perceive no difference between adolescents' and adults' developmental maturity and sensitivity to coercive techniques (Kostelnik & Reppucci, [38]). John E. Reid and Associates ([34]) even offer Reid training for school administrators in interviewing and interrogating students who engage in misconduct—including use of verbal and nonverbal behaviors as indicators of truth/deceit, deviating little from their techniques used with adult criminal suspects. As substantial research has shown the use of BAI techniques with adults to be flawed (Bond & DePaulo, [3]; Kassin & Fong, [37]; Masip & Herrero, [45]), the use of these techniques may be especially inappropriate for juveniles given their known immaturities. Common behaviors exhibited by juveniles are congruent with stereotypical behavioral cues of lying (Birckhead, [1]; Jupe et al., [35]; Meyer & Reppucci, [51]; The Global Deception Research, [69]), therefore, juveniles may be at an increased risk of mistaken judgments of deceit if they are unaware of and do not actively control for the presentation of those behavioral cues during a police interview.

The present study was designed to determine whether juveniles, young adults, and adults share similar stereotypes of cues to lying and would choose similar strategies and responses during a BAI‐styled hypothetical police interview. Prior studies on the BAI have used only young adult or college student samples (Masip & Herrero, [45]; Vrij, Mann, & Fisher, [76]). We sampled three age groups: juveniles (12–16 years), young adults (18–25 years), and adults (34–50 years). We included a young adult sample to allow for an appropriate comparison of results to Masip and Herrero's ([45]) college‐aged sample (from which this study's methods were adapted). Moreover, even though the legal system defines 18‐year‐olds as adults, developmental science recognizes that the brains of 18‐ to 25‐year‐olds are still developing, and considerable research indicates that many factors relevant to impaired interview performance, including self‐regulation in emotionally negative situations, continue to develop into the early‐to‐mid‐twenties (e.g., Cohen et al., [8]; Steinberg, [63]). We additionally included a middle‐aged adult sample to determine whether there are further changes in strategies, responses, and stereotypes with development.

Using a 3 (age: juveniles v. young adults v. adults) × 2 (guilt condition: imagined innocence v. imagined guilt) by 2 (response type: truthful responses v. deceptive responses) mixed factorial design, we compared juveniles', young adults', and adults' choice of responses to a hypothetical BAI about an imagined crime, their reported strategies to appear innocent, and views on stereotypical lying behavior.

As research has shown that the BAI's claimed deceitful behaviors are congruent with common stereotypes of how liars behave (Masip et al., [47], [44]), we predicted that regardless of age or guilt condition, participants would select the BAI's claimed truthful responses to a greater extent than the BAI's claimed deceitful responses during a police interview (H1). We additionally predicted that juveniles would select deceitful responses to a greater extent than young adults and adults, while young adults would select more deceitful responses compared to adults (H2). Moreover, in line with previous research (e.g., Hartwig et al., [28]; Hartwig et al., [29]; Hines et al., [30]), we predicted that guilty suspects would choose more strategies to appear innocent than innocent suspects, regardless of age (H3), and finally, that adults would select more strategies to appear innocent than juveniles and young adults, regardless of guilt status (H4).

We further explored variability in responses to the hypothetical BAI according to demographic factors (gender and race) and prior experience being questioned by police. Previous research has shown that cues to deception and the ability to conceal deception can differ between cultures (see Castillo, [5]) and genders (e.g., Feldman et al., [18]; McQuaid et al., [49]). Participants who have experience being questioned by police may differ in their ability to immerse themselves in the hypothetical interview scenario presented in this study and in their understanding of the goals of a police interview (see Granhag et al., [23]; Vidal et al., [70]). Thus, we examined if some individuals are more adept at distinguishing between claimed truthful and deceitful responses, though we report no a priori predictions regarding these variables. Further, we included age as a continuous measure in these exploratory analyses to assess linear changes in responses from adolescence through middle adulthood. Lastly, we explored associations between our primary outcomes of likelihood to use the BAI truthful or deceitful responses, strategies to appear innocent, and stereotypical cues to deceit. Specifically, several unique behaviors appeared on all three measures (eye contact, detailed story, and pausing), so we examined how the likelihood of using those responses in the BAI related to using that behavior as a strategy to appear innocent, and belief that the behavior signaled lying.

Full study materials, additional demographic information, and supplemental results are available on the Open Science Framework at http://doi.org/10.17605/OSF.IO/B6FZ4.

Participants were recruited from three age groups: juveniles (aged 12–16), young adults (aged 18–25), and adults (aged 34–49). Participation was limited to English‐speaking residents of the U.S. who could read at a 7th‐grade level.

An a priori power analysis using G*Power (Faul et al., [16]) anticipating medium effect sizes of f = 0.25 (based on findings from Masip & Herrero, [45]) with a repeated measures MANOVA to detect between‐within interactions with six groups indicated a sample size of 270 (45 per group) would be sufficient for a power of 0.90 at alpha of 0.05 (Cohen, [9]). Thus, we collected data until 270 valid participants (90 per age group) were obtained. Juveniles were recruited via social media, word‐of‐mouth, and ChildrenHelpingScience.com, an online collaboration where researchers can post studies aimed at recruiting children. Juvenile participants and their parent or legal guardian first met with a member of the research team via Zoom to ensure voluntary participation and gain informed consent/assent. They received a $10 Amazon gift card upon completion. Additional young adult and adult samples were recruited from Amazon's Mechanical Turk using Cloud Research (Litman et al., [40]) and were compensated $3.00 for completing the study appropriately. Only participants who had at least a 99% task approval rating and 500+ HITs approved were recruited, and data from those who accurately completed all attention checks were retained to ensure high quality data.

For the juvenile sample, 94 12–16‐year‐olds participated but four were excluded due to completing less than 75% of the study. Thus, the final juvenile sample consisted of 90 participants aged 12–16‐years‐old (M = 13.59, SD = 1.31); 50% identified as female and 83.3% as White. For the young adult sample, 91 18–25‐year‐olds participated but one was excluded due to failing attention checks. Thus, the final young adult sample comprised 90 participants with ages ranging from 18 to 25 (M = 22.41, SD = 1.84): 45.6% identified as female and 66.7% as White. For the adult sample, 91 34‐49‐year‐olds participated, and one was excluded for being beyond the maximum age cutoff of 50. The final adult sample consisted of 90 participants with ages ranging from 34 to 49 (M = 40.87, SD = 4.32): 51.1% identified as male and 67.8% as White. See the OSF for full demographic data.

Materials used for this study included a hypothetical theft scenario, imagined BAI, and a post‐interview questionnaire. This questionnaire contained measures on strategies to appear innocent and stereotypes of behaviors that indicate lying. Four open‐ended, free‐recall attention check questions were also asked throughout the study. All materials presented to participants were written at a 7th grade reading‐level (Flesch–Kincaid grade‐level) and available on the OSF.

After obtaining consent (parental consent for juveniles) and assent, all participants read a high‐stakes criminal scenario vignette depicting a theft of an expensive computer from school (work for adult participants). The scenario was modeled after the Reid Technique manual's example BAI (Inbau et al., [33]) and adapted from Masip and Herrero's ([45]) study. Participants were randomly assigned to the innocent or guilty condition and instructed to imagine they would be questioned by a police officer regarding their involvement in the case and that they need to convince the officer that they are innocent.

Participants were then presented the 15 BAI questions, in a fixed order and one at a time, with all potential truthful and all potential deceptive responses outlined by Inbau et al. ([33]) presented under each question. Each potential response to the officer's question started with the prompt "If you were innocent [guilty, dependent on randomly assigned guilt condition], would you give a ..." and continued into the specific verbal or nonverbal behavioral response. BAI questions vary between 2 and 11 possible responses (M = 4.26, SD = 2.46). All BAI questions and the corresponding responses are shown in Table 1. Participants were instructed to rate on a 1 ("I definitely would not give this answer") to 6 ("I definitely would give this answer") scale the extent which they believed they would give each potential response based on their involvement in the case (see Table 2 for example question).

2 TABLE Example BAI question and responses

Participants answered questions regarding the strategies they would use during an interview to convince an officer that they are innocent of the crime in question. All questions and response scales were adapted from Masip and Herrero ([45]) to allow direct comparisons to their results with young adults. They first read a list of five general strategies and indicated on a Likert scale of 1–6 the extent to which they would actively engage in each of the strategies, with lower scores indicating more use of that strategy and higher scores indicating acting naturally. These strategies included making a plan for how they would behave during the interview, trying to appear innocent, trying to change physical/nonverbal behaviors to look innocent, trying to control their words to looks innocent, and giving an alibi. Participants then were presented with seven specific strategies rated on a 1–3 Likert scale, with the midpoint indicating acting natural, and the two end points indicating opposite behaviors (e.g., 1 = "Make eye contact," 2 = "Act natural," 3 = "Not make eye contact"). These included making body movements, looking nervous or relaxed, telling a detailed or simple story, denying or admitting guilt, including truthful or deceptive information, answering all questions, and making eye contact.

Ten behaviors identified by The Global Deception Research Team ([69]) as being most stereotypically revealing of lying were presented to participants. Participants read a prompt (e.g., "When people are lying, they act...") and then selected the behavior they associated with lying (e.g., calm, nervous, or neither calm nor nervous). Each of the 10 behaviors was then recoded into new variables so that a 1 indicated the behavior was stereotypically thought to be associated with liars, or a 0 indicated the behaviors not stereotypically thought to be associated with liars. Summed scores were then created across the 10 recoded variables such that higher scores indicated perceptions of behaviors that cue deceit being more congruent with global stereotypes.

All participants were prompted to indicate their age, gender, and race/ethnicity. The three age groups were not significantly different from each other in terms of their gender or race, χ2(4) = 4.06, p = .40, Cramer's V = .09; χ2(6) = 12.10, p = .06, Cramer's V = .15, respectively. Additionally, participants were asked whether they have had prior interactions with police, specifically if they had been formally interrogated about their involvement in a crime or informally interviewed by an officer. The three age groups were not significantly different from each other in terms of their prior experience being questioned by police either formally or informally (18.9% of juveniles, 27.8% of young adults, 30% of adults reported having been questioned by police; χ2[6] = 6.05, p = .42, Cramer's V = .11).

To test the effect of the two independent variables (age and guilt/innocence condition) on the dependent variables (rated likelihood of use of deceitful and truthful BAI responses), two response type means were first computed for each of the 15 BAI questions. One mean was computed for the average on all truthful responses per question (resulting in 15 truthful response type means) and one for the average on all deceitful responses per question (resulting in 15 deceitful response type means). For example, if a question had five total responses, and three of those are claimed to indicate deceit, ratings on those three items were averaged together for a mean of deceitful responses for that question and the remaining two items were averaged for mean of truthful responses. A mixed MANOVA with response type (i.e., truthful or deceptive) as the within‐subjects variable and age and guilt condition as the between‐subjects variables was conducted to investigate differences in the mean truthful and mean deceptive scores on the BAI. Any significant multivariate effects were followed‐up with univariate analyses for each BAI item with Bonferroni corrections to control for the overall Type I error rate (critical alpha levels are reported for respective sets of univariate tests).

Results showed a significant multivariate main effect of response type, F(15, 250) = 60.05, p < .001, partial η2 = .78. When averaged across the 15 BAI questions, regardless of age or guilt condition, participants gave significantly higher likelihood ratings to the BAI's claimed truthful responses (M = 3.51, SE = 0.04, 95% CI [3.42, 3.59]) than deceptive responses (M = 2.96, SE = 0.04, 95% CI [2.89, 3.04]), supporting Hypothesis 1. Univariate analyses for the 15 BAI items (critical α = .003) revealed participants rated their likelihood of giving truthful responses significantly higher than deceptive responses for questions 2, 6, 8, 9, 10, 13, and 14. For questions 11 and 12, participants rated their likelihood of giving deceptive responses as significantly greater than truthful responses.

The multivariate main effect of guilt condition was significant, F(15, 246) = 5.27, p < .001, partial η2 = .24. This effect was qualified by a significant multivariate response type × guilt condition interaction, F(15, 250) = 4.83, p < .001, partial η2 = .23. When averaged across the 15 BAI questions, the size of the difference between rated likelihood of use of truthful and deceptive responses was larger for innocent participants (M = 3.59, SE = 0.06, 95% CI [3.47, 3.71]; M = 2.91, SE = 0.05, 95% CI [2.81, 3.01], respectively) compared to guilty participants (M = 3.42, SE = 0.06, 95% CI [3.30, 3.55]; M = 3.02, SE = 0.05, 95% CI [2.92, 3.12], respectively). We used independent samples t‐tests with Bonferroni adjustments to further assess this interaction. Countering predictions from Inbau et al. ([33]), guilty participants were not significantly more likely to select deceptive responses than innocent participants, t(268) = 1.49, p = .14, d = 0.18, 95% CI [−0.06, 0.42], nor were innocent participants significantly more likely to select truthful responses than guilty participants, t(268) = −1.97, p = .05, d = 0.24, 95% CI [−0.47, 0.00]. Significant univariate analyses for response type x guilt condition interactions on the 15 BAI items (critical α = .002) showed guilty participants reported more use of deceitful responses for questions 1 and 5, while innocent participants reported more use of truthful questions for questions 1, 4, and 15. See the OSF for means by BAI question, response type, and guilt condition.[1]

The multivariate main effect of age was also significant, F(30, 502) = 2.20, p < .001, partial η2 = .17. This effect was qualified by a significant multivariate response type x age interaction, F(30, 502) = 3.11, p < .001, partial η2 = .16. When averaged across the 15 BAI questions, post‐hoc analyses with Bonferroni adjustments showed that juveniles reported a significantly greater likelihood of giving deceptive responses (M = 3.21, SE = .06, 95% CI [3.09, 3.34]) compared to young adults (M = 2.92, SE = .06, 95% CI [2.80, 3.04]; p = .003; d = 0.52, 95% CI d [0.21, 0.82]) and adults (M = 2.76, SE = .06, 95% CI [2.64, 2.89]; p < .001; d = 0.78, 95% CI d [0.48, 1.08]), thus supporting Hypothesis 2. When averaged across the 15 BAI questions, juveniles did not significantly differ in their rated likelihood of using truthful responses (M = 3.65, SE = .08, 95% CI [3.50, 3.80]) from young adults (M = 3.49, SE = .08, 95% CI [3.33, 3.63]; p = .39; d = 0.26, 95% CI d [0.04, 0.55]) nor adults (M = 3.39, SE = .08, 95% CI [3.25, 3.54]; p = .06; d = 0.36, 95% CI d [0.06, 0.65]; see Figure 1). Young adults and adults were not significantly different from each other in their likelihood of use of deceptive (p = .24; d = 0.25, 95% CI d [0.05, 0.54]) nor truthful responses (p = .99; d = 0.11, 95% CI d [0.18, 0.41]). At the BAI item level, juveniles reported significantly (critical α = .001) greater likelihood of using deceitful responses than young adults for questions 2 and 10 and adults for questions 2, 7, 8, 9, 10, and 12. Juveniles also rated their likelihood of using truthful responses significantly higher than young adults and adults for questions 4, 5, and 10 (see Table 3).

3 TABLE Cell means for BAI response type by age interactions

1 Note: Means with different superscripts differ at the p = .001 level (critical α after Bonferroni corrections).

Note that BAI items 2 and 10 were the only two BAI questions with nonverbal behaviors included in the potential responses, and that juveniles consistently differed from both adult groups on these items. Therefore, we explored differences between the three age groups in their likelihood of using the BAI claimed deceptive nonverbal responses to these questions (i.e., we computed two averages: one for likelihood of using the claimed deceptive nonverbal responses for question 2 and one for question 10; see Table 1 for list of responses) using two 3 (age) × 2 (guilt) ANOVAs (critical α = .008). Main effects of age for question 2 (F[2, 264] = 21.29, p < .001, partial η2 = .14) and question 10 (F[2, 264] = 12.64, p < .001, partial η2 = .09) showed that juveniles reported a significantly higher likelihood of using the claimed deceptive nonverbal responses for both questions (M = 2.82, SE = .11, 95% CI [2.60, 3.04]; M = 2.74, SE = .14, 95% CI [2.47, 3.02], respectively) compared to adults (M = 1.85, SE = .11, 95% CI [1.63, 2.07], p < .001; M = 1.78, SE = .14, 95% CI [1.50, 2.06], p < .001, respectively) and young adults (M = 2.04, SE = .11, 95% CI [1.82, 2.26], p < .001; M = 2.02, SE = .14, 95% CI [1.75, 2.30], p < .001, respectively). There were non‐significant main effects of guilt for question 2 (F[1, 264] = 1.06, p = .30, partial η2 = .004) and 10 (F[1, 264] = 5.91, p = .02, partial η2 = .002), and non‐significant guilt by age interactions (F[2, 264] = 1.53, p = .22, partial η2 = .01; F[1, 264] = 0.33 p = .72, partial η2 = .002, respectively).

A 3 (age) × 2 (guilt condition) MANOVA was used to examine the effects of age and guilt condition on participants' general strategies to appear innocent. Scores on the first five general strategies (of the 12 total) were first analyzed together because they were measured on the same scale of 1–6. Lower ratings indicated that participants were more likely to use that strategy to appear innocent and higher ratings indicated they were more likely to act natural (i.e., less likely to use that strategy).

The multivariate effect of guilt condition was significant, F(5, 260) = 13.00, p < .001, partial η2 = .20. Univariate analyses for each strategy (critical α = .01) showed that guilty participants indicated they would be more likely to make a plan for how to act during the interview (p < .001; d = 0.63, 95% CI d [0.39, 0.88]), try to change their physical/nonverbal behaviors to look innocent (p < .001; d = 0.55, 95% CI d [0.31, 0.79]), and try to control their words to appear innocent (p < .001; d = 0.66, 95% CI d [0.41, 0.90]). Guilty and innocent participants did not significantly differ (after Bonferroni adjustments) in their reports of trying to appear innocent (p = .02; d = 0.29, 95% CI d [0.05, 0.52]) nor giving an alibi (p = .05; d = 0.25, 95% CI d [0.01, 0.48]). Providing partial support for Hypothesis 3, guilty participants were more likely to report using strategies to appear innocent compared to innocent participants (see Table 4).

4 TABLE General strategies to appear innocent by guilt condition and age

• 2 Note: Higher means indicate less likely to use each of the strategies (i.e., more likely to act naturally). Means with different subscripts differ at the p < .001 level (critical α after Bonferroni corrections).
• 3 *** p < .001.

There was also a significant multivariate main effect of age group, F(10, 522) = 4.08, p < .001, partial η2 = .07. In contrast to our expectations in Hypothesis 4, post‐hoc analyses (critical α = .003) revealed that compared to adults, juveniles indicated they would be significantly more likely to make a plan for how to act during the interview (p < .001; d = 0.51, 95% CI d [0.22, 0.81]) and try to change their physical/nonverbal behaviors to look innocent (p = .001; d = 0.52, 95% CI d [0.22, 0.81]). Additionally, juveniles were significantly more likely to try to appear innocent than young adults (p < .001, d = 0.62, 95% CI d [0.32, 0.92]) and adults (p < .001; d = 0.70, 95% CI d [0.40, 1.00]). Juveniles were also significantly more likely to try to control their words to appear innocent than young adults (p < .001, d = 0.61, 95% CI d [0.31, 0.91]) and adults (p < .001; d = 0.64, 95% CI d [0.34, 0.94]; see Table 4).

The second set of seven specific strategies to appear innocent were analyzed for differences by guilt condition and age groups using a series of chi‐squares. Providing additional partial support for Hypothesis 3, guilty participants were more likely to say they would try to tell a story with few details (χ2[2, _I_N_i_ = 270] = 21.17, p < .001, Cramer's V = .28) and try to include as much false information as possible (χ2[2, _I_N_i_ = 270] = 8.62, p = .01, Cramer's V = .18), while innocent participants said they would be more likely to try to tell their story naturally.

Again, refuting hypothesis 4, juveniles said they would use several specific strategies to appear innocent more often than young adults and adults. More juveniles said they would try to look calm and relaxed (χ2[4, _I_N_i_ = 270] = 19.71, p = .001, Cramer's V = .27), they would tell a very detailed story (χ2[4, _I_N_i_ = 270] = 24.40, p < .001, Cramer's V = .30), and they would try to include as much truthful information as possible (χ2[4, _I_N_i_ = 270] = 14.34, p = .006, Cramer's V = .23), whereas more young adults and adults said they would tell their story/act naturally.

A univariate ANOVA was used to test for age differences in behaviors stereotypically associated with lying. There was a significant main effect of age on the sum variable of behaviors stereotypically associated with lying, F(2, 267) = 5.04, p = .007, partial η2 = .04. Bonferroni post‐hoc analyses show that adults' perceptions of cues to deception (M = 7.21, SE = .22, 95% CI [6.78, 7.64]) are significantly more congruent with global stereotypes compared to juveniles (M = 6.23, SE = .22, 95% CI [5.80, 6.66]; p = .005; d = 0.48, 95% CI d [0.18, 0.77]). Young adults (M = 6.76, SE = .22, 95% CI [6.33, 7.19]) did not significantly differ from adults (p = .42, d = 0.22, 95% CI d [0.07, 0.51]) nor juveniles (p = .27, d = 0.25, 95% CI d [0.04, 0.55]). This indicates that juveniles were less likely to agree with the stereotypes of what a liar looks like than adult respondents.

Chi‐square analyses were conducted to further explore this age difference in endorsement of the 10 stereotypes of liars' behaviors. Of the three age groups, juveniles were least likely to endorse the stereotype that liars are more nervous than usual (χ2[4, _I_N_i_ = 270] = 20.34, p < .001, Cramer's V = .27), that liars have less consistent stories than usual (χ2[4, _I_N_i_ = 270] = 10.15, p = .04, Cramer's V = .19), that liars have longer stories than usual (χ2[4, _I_N_i_ = 270] = 23.38, p < .001, Cramer's V = .29), and that liars use more hand gestures than usual (χ2[4, _I_N_i_ = 270] = 12.61, p = .01, Cramer's V = .22).

We explored whether individual demographic factors explained additional variance in responses to our dependent measures beyond that attributable to guilt condition and age. Hierarchical regressions were conducted to predict overall use of the BAI's deceitful responses (averaged across the 15 items), overall use of the BAI's truthful responses (averaged across the 15 items), averaged likelihood of using the general strategies to appear innocent (i.e., the first five strategies measured on 1–6 scales), and the summed variable of stereotypical lying behaviors. Guilt condition was entered in step 1, age (as a continuous measure) was entered in step 2, and gender, race, and prior police questioning experience were entered in a stepwise manner in step 3 (note that for the lying stereotypes analysis, guilt condition was irrelevant so was not included as a predictor).

As shown in Table 5, age significantly accounted for additional unique variance beyond that of guilt status for both deceitful and truthful responses (R2 = .075; R2 = .03, respectively). Gender was the only additional significant demographic predictor of BAI responses and accounted for additional unique variance in both deceitful and truthful responses (ΔR2 = .02; ΔR2 = .04, respectively); males were significantly less likely than females to use deceptive and truthful responses, when holding age and guilt status constant.[2] None of the demographic factors were significant predictors (ps > .05) of reported use of general strategies to appear innocent. Rather, age added significant variance explained beyond guilt status (F[2, 256] = 24.63, p < .001, ΔR2 = .081), and guilt condition and age significantly predicted use of strategies (β = −0.29, t[254] = −4.98, p < .001; β = 0.29, t[254] = 4.96, p < .001, respectively), consistent with prior analyses. Lastly, age alone significantly predicted endorsement of lying behavior stereotypes (F[1, 255] = 9.86, p = .001, R2 = .037; β = 0.19, t[255] = 3.14, p = .002), but no other demographic factors were significant (ps > .05).

5 TABLE Exploratory stepwise regression analyses with predictors of BAI deceitful responses and BAI truthful responses

• 4 Note: Higher scores indicate greater likelihood of using the presented responses (1–6 scales). Condition was coded such that 0 = innocent condition, 1 = guilty condition. Gender was coded such that 0 = male, 1 = female.
• 5 *** p < .001.
• 6 ** p < .01.
• 7 * p < .05.

Finally, we explored whether participants' responses were related across similar behaviors assessed on our dependent measures. First, partial correlations were calculated to determine the extent to which overall use of deceitful responses (averaged across the 15 BAI questions) corresponded to stereotypes about lying behavior, while controlling for age. Overall use of deceitful responses and stereotypes about lying behavior were not significantly correlated (r = −.04, p = .47), suggesting an increase in use of deceptive responses is unrelated to common stereotypes about how a liar behaves. However, the BAI questions present more, and more specific, verbal and nonverbal behaviors beyond that of the 10 stereotypical behaviors, which may obscure the relationship between the two constructs. Thus, we examined partial correlations for specific cues (eye contact, response length, and anxiety) that were presented in all three measures, while controlling for age group and guilt condition. An increased likelihood in participants reporting they would make eye contact when responding to the officer's question (claimed truthful response to BAI 2) was significantly correlated with eye contact as a strategy to appear innocent (r = .45, p < .001) and believing that lack of eye contact signals lying (r = .15, p = .02). Next, an increased likelihood of participants reporting they would give a shorter response when answering the officer's question (claimed deceitful response to BAI 14) was significantly correlated with telling a simple story as a strategy to appear innocent (r = .15, p = .02) and believing that long stories signal lying (r = .15, p = .01). Finally, a decrease in participants reporting they would engage in anxiety‐reducing behaviors when answering the officer's question (claimed deceitful response to BAI 10) was correlated with believing that liars act nervous (r = −.16, p = .01) but was not correlated with the strategy of looking calm to appear innocent (r = .02, p = .79).

Research has shown that use of the Behavior Analysis Interview (BAI) is problematic for adults in that (1) it does not reliably differentiate truthful from deceitful suspects (DePaulo et al., [12]; Masip & Herrero, [45]; Vrij, Mann, & Fisher, [76]), (2) BAI training does not improve lie‐detection (Vrij, [72]), and (3) guilty suspects can manipulate the behaviors said to indicate deceit because they are congruent with global stereotypes on how a liar behaves (Masip et al., [47], [44]; The Global Deception Research, [69]). However, the use of this method in application to juvenile suspects and juveniles' strategies during a police interview had not been empirically addressed previously, which was the primary goal of this study. This is an important undertaking as nearly 700,000 juveniles in the United States were arrested and potentially subjected to interrogation in 2019 alone (Office of Juvenile Justice and Delinquency Prevention, [53]).

Participants self‐reported greater likelihood of using truthful than deceptive responses, regardless of age or guilt condition. For just one of the 15 BAI questions, innocent participants indicated they would give more truthful responses and guilty participants more deceitful responses. Although these trends support the BAI's predicted pattern of responses for guilty and innocent suspects for this question, there were no consistent differences in response type according to guilt‐status for the remaining 14 questions. Thus, we did not find support for the BAI's predictions of the truthful and deceptive responses that innocent and guilty suspects would give. These results were expected, as prior research has consistently shown that the BAI does not accurately distinguish between liars and truth‐tellers (e.g., Vrij, Mann, & Fisher, [76]) because the behaviors thought to indicate lying are typically unrelated to actual deception (e.g., DePaulo et al., [12]; Hartwig & Bond, [26]; Vrij, [72]) and are congruent with global stereotypes of lying (Masip et al., [47], [44]; The Global Deception Research Team, [69]). Thus, guilty suspects can attempt to control these behaviors to avoid detection (Hartwig et al., [28]; Hartwig & Bond, [26]; Hines et al., [30]). Our exploratory analyses showcased this effect with the behaviors of making eye contact and response length. Participants' increased self‐report of making eye contact with the officer and giving shorter responses when answering questions in the BAI were significantly related to using eye contact and telling a simple story as strategies to appear innocent, and significantly related to the beliefs that lack of eye contact and long stories signal lying (when controlling for age and guilt). These examples highlight that participants who recognized that these behaviors are believed to indicate lying were more likely to adjust their hypothetical responses to avoid these stereotypical cues to deceit.

Guilty suspects self‐reported that they would be more likely to use strategies to appear innocent in comparison to innocent suspects. Specifically, guilty suspects reported that they would make a plan for how to act in the interview, change their physical and nonverbal behaviors to look innocent, control their words to appear innocent, tell a story with few details, and include false information in their responses. In contrast, innocent condition suspects indicated they would not use the strategies to appear credible, but rather would just act naturally in an interview. It appears that innocent suspects believe that their innocence will be obvious to an officer, and as a result, they do not feel the need to change their verbal and nonverbal behaviors to be congruent with how truth‐tellers are thought to act. The phenomenology of innocence posits that 'innocence places innocents at risk' (Kassin, [36]) because innocent suspects believe that their internal states are highly visible to others (illusion of transparency) and that they will not be punished because the truth will emerge (belief in a just world; Hartwig et al., [28]; Kassin, [36]).

Two theoretical approaches can help to explain these observed differences in guilty and innocent particpants. First is attempted control, which suggests that liars put more effort into supressing indicators of nervousness to avoid being caught and appear credible, and thus actively engage in impression management (Hocking & Leathers, [31]). However, this is not an easy task and depends on the liar's successful ability to suppress behaviors and act in a credible way (see Vrij, [73]). Second, the self‐presentation perspective proposes that people will change their self‐presentation in efforts to be perceived by others in a more desirable way; thus both liars and truth tellers will adjust their behaviors to appear honest and credible (DePaulo, [11]; DePaulo et al., [12]). However, truthtellers do not have to engage in information management in the same manner that liars do, thus explaining why liars attempt to manipulate their behavior to a greater extent (Hartwig et al., [29]).

Unique trends according to participant age emerged: juveniles self‐reported that they would be significantly more likely to give the BAI's claimed deceitful responses than both groups of adults, including nonverbal behaviors. These results suggest that juveniles may be less aware of how behaviors during an interview are perceived, whether they plan to lie or tell the truth. Past research shows adolescents often perform comparably to adults on "cold" cognition tasks, but their limitations are revealed in "hot" cognition situations (Icenogle et al., [32]). Our results provide evidence for age differences in a cold cognition task; even in the absence of pressure from an authority figure or the possibility of negative consequences, and with ample time to strategize on how to best respond, juveniles still did not perform comparably to young adults nor adults.

Juveniles indicated they would use strategies to appear innocent to a greater extent than adults and young adults, but their intentions did not translate into their chosen responses to the BAI, as they reported higher likelihood to use deceptive behaviors (overall and for six of the 15 individual questions) compared to the older age groups. These results show a disconnect in how juveniles intend to use strategies to manipulate their behaviors to appear innocent, and the actual responses they think they would give that indicate innocence and guilt. One potential explanation for this disconnect could lie in their beliefs of how liars stereotypically behave; juveniles did not hold the same stereotypes of what a liar looks like in comparison to older adults, and perhaps as a result, they did not differentiate between claimed truthful and deceitful responses as well as adults did. For example, juveniles were not as likely to believe the stereotypes that liars are nervous, tell inconsistent and long stories, and use hand gestures—and were more likely to select these and similar claimed deceitful behaviors in the BAI compared to young adults and adults. Likewise, juveniles seemed to over‐estimate their abilitiy to appear innocent; they indicated use of deliberate self‐presentation strategies in efforts to appear credibile, but failed to accurately recognize how the behaviors presented in the hypothetical police interview would be interpreted as lying by others. These results provide insight into how juveniles may approach a police interview or interrogation situations, specifically in how their knowledge and perceptions of lying behaviors may impact how they plan to respond. Future research should further investigate how juveniles' stereotypes of how liars act translates into their actual behaviors exhibited when lying and how their self‐presentation skills are impacted by the expectations of the receiver.

No significant differences between young adults and older adults were found in our primary analyses, suggesting these two age groups had similar engagement with the BAI, strategies to appear innocent, and stereotypes about lying behavior. This pattern is consistent with research on the development of "cold" cognition, showing that mental processes during emotionally neutral situations tend to improve up until mid‐adolescence, and then remain stagnant throughout adulthood (Icenogle et al., [32]). Specifically, young adults (aged 18–21) have been shown to perform comparably to older adults on cognitive control tasks in nonemotional states (Cohen et al., [8]), and in legal settings, older adolescents are shown to be comparable to adults in terms of their understanding of Miranda rights (Viljoen et al., [71]) and competence to stand trial (Grisso et al., [24]).

When age was examined as a continuous measure in exploratory analyses, age accounted for additional variance in overall likelihood of using the BAI's truthful responses, even when controlling for guilt condition. This further implies that younger participants were less selective in their reported use any of the BAI's responses. Although this finding should be viewed with caution, it suggests that using age as a continuous variable allows for greater sensitivity in detecting differences across the lifespan. Future studies can use age as a continuous measure, without gaps between discrete age groups, to more precisely examine patterns of development. Given that 18‐ to 25‐year‐olds (defined here as young adults) are still maturing in their cognitive control and self‐regulation (see Cohen et al., [8]; Steinberg, [63]), and that younger juveniles (under 15) differ from older juveniles in important ways (e.g., Goldstein et al., [19]; Grisso et al., [24]; Viljoen et al., [71]), researchers could explore the full range of adolescence and young adulthood to determine whether changes are gradual and incremental or more abrupt and discontinuous.

In addition to interviewing and interrogating criminal suspects, the John E. Reid and Associates group ([34]) also offers interviewing and interrogation trainings for school administrators, who frequently question students about criminal involvement with and without School Resource Officers present (see Crane, [10]). This training includes use of the BAI, with a particular focus on how students' nonverbal and verbal behaviors can indicate truth or deception. In a survey of School Resource Officers (SROs), 41.5% reported attending a Reid training and 84% reported using behavioral cues to judge the legitimacy of juvenile students' statements (Snow et al., [62]). Moreover, SROs estimated they could reliability detect deception 66% of the time by "analyzing children's and youth's eye movements, speech patterns, and body language" (pg. 366). Yet professionals (including officers and teachers) are, on average, accurate in detecting deceit in children only 56% of the time (Gongola et al., [21]). Results from the current study provide preliminary evidence for urging caution in using the BAI with juveniles in any setting.

The largest limitation of the current study is that participants were responding to a low‐stakes, hypothetical police interview rather than a real‐life, high‐stakes interview. Although previous deception detection research has shown low‐stakes or "cold" cognition laboratory paradigms to be representative of high‐stakes or "hot" cognition scenarios (Hartwig & Bond, [27]; Mann et al., [43]), any generalizations of these results to high‐stakes scenarios should be made with caution. However, we based our materials and procedures on those developed by Masip and Herrero ([45]) who found results consistent with studies using in‐person mock‐interview methods (e.g., Vrij, Mann, & Fisher, [76]). Further, although juveniles might have greater difficulty than adults imagining themselves or predicting their responses in a hypothetical situation, we adapted the crime and questions to be more appropriate for a situation in which the juveniles might be involved. Moreover, the ability to think abstractly and consider hypotheticals tends to mature around middle adolescence (Steinberg, [64]), and other studies investigating juveniles in various legal scenarios, such as Miranda decision‐making (Sharf et al., [61]), plea decision‐making (Redlich & Shteynberg, [55]), and falsely confessing (Haney‐Caron et al., [25]), have successfully utilized hypothetical mock‐crime scenarios to approximate juveniles' reactions to real‐world scenarios. Thus, while this study does not directly assess behaviors in a stressful, actual police interview, it does provide initial evidence that juveniles poorly strategize and misperceive behaviors exhibited in a police interview. It would be interesting to see if juveniles could carry out their planned interview responses in the face of actual questioning. Future studies should therefore seek to investigate how these results replicate in a mock‐interview paradigm such as that used by Vrij, Mann, and Fisher ([76]) to determine whether and how developmental differences shape strategies and behaviors during police interviews.

A further limitation of this study is the use of restricted scales. Specifically, it is unclear how results would change if participants were presented open‐ended response formats on the BAI rather than pre‐determined scales. An open‐ended response format may shed additional light on how juveniles' beliefs about lying behaviors influence their own strategies and behaviors to avoid detection in a police interview. Further, the 3‐point scale used to measure some of the strategies to appear innocent and stereotype of lying behaviors may have artificially simplified complex behaviors that occur throughout a police interview. Although this allowed us to appropriately compare our results to those of Masip and Herrero ([45]) and research by The Global Deception Research Team ([69]), future studies may seek to use scales with a wider range to increase sensitivity.

Our exploratory analyses also showed that males reported a lower likelihood of using both truthful and deceptive responses, when controlling for age and guilt status. Males and females may differ in their beliefs of nonverbal cues that indicate deceit (e.g., Rosip & Hall, [58]), but research is generally mixed on whether men or women are better at concealing and detecting lies (see Lloyd et al., [41]). Research should further examine individual differences that may impact beliefs of cues to deception and approaches to interviews. For example, cues to deceit typically differ between cultures and can quickly lead to errors in deception detection during cross‐cultural communication (e.g., Leal et al., [39]). Further, as the ability to tell and maintain lies requires complex mechanisms in executive functioning (see Gombos, [20]), future studies can investigate how differences in abilities such as response inhibition and perspective taking impact cues to deception during a police interview.

Despite these limitations, the present study offers an important first look at how age can uniquely impact strategies to appear innocent and perceptions of lying behavior during a pre‐interrogation criminal interview. This study provides insight into how juveniles plan to behave during a hypothetical, "cold" cognition situation. If played out in an actual interview, findings from the current study have several implications. Guilty suspects who plan to purposefully adjust their verbal and nonverbal behaviors to appear truthful may be able to successfully go undetected by an officer, while innocent suspects may unknowingly engage in deceit‐assumptive behaviors, and thus be inappropriately interrogated, increasing the risk of false confession. This pattern could be especially true for juveniles, who may intend to manipulate their responses and behaviors to appear innocent, but may not realize how those behaviors are believed to cue lying. Further, juveniles may lack the self‐regulation and impulse control skills needed to successfully adjust their responses to appear innocent during a "hot" cognition situation (Cleary, [6]; Steinberg et al., [65]). If inaccurately judged as deceptive, innocent juveniles could be subjected to an unnecessary, accusatory interrogation. Research has shown that individuals under the age of 25 possess developmental immaturities in an interview or interrogation that place them at an increased risk of being coerced into a false confession (e.g., Drizin & Leo, [14]) and/or waiving their Miranda rights (e.g., Redlich et al., [56]). The present findings contribute to that body of work by uniquely addressing how juveniles' beliefs about cues to deception and reported strategies and behaviors may make them vulnerable to an incorrect judgment of deceit during a BAI.

Support was provided by Akera Williams, Julianna Schau, and Stephanie George at the University of Tennessee at Chattanooga, and ChildrenHelpingScience.com.

Financial support was provided by the Scholarship, Engagement, the Arts, Research, Creativity, and Humanities Grant and Faculty Research and Creative Activity Grant from the University of Tennessee at Chattanooga.

The authors declare no known conflicts of interest.

Study materials and supplemental results are available at the Open Science Foundation repository at http://doi.org/10.17605/OSF.IO/B6FZ4