Deception and culture

Cultural differences can be defined as ‘the collective programming’ that distinguish one group from another [34]. Hence, culture typically refers to customary beliefs, values and characteristic societal markers that determine individual attitudes, and shape individual and social behaviours [35–37]. Individuals from different cultures behave and communicate variously in forensic investigative contexts. For example, event details reported can differ across cultures irrespective of veracity, likewise interruptions, assertions of normality and interrupting behaviours which can all be more or less common in both liars and truthtellers as a function of culture [32,33,38,39]. However, all deceivers face a similar challenge, that is how to appear credible. In forensic interview contexts, deceivers may embed lies in the truth and so offer a partially truthful account that includes some deceptive elements, and/or they withhold chunks of information completely, often suggesting a lack of recall or denying having seen or heard elements of the events in question [8,40,41]. Cross cultural understanding of similarities and differences in truth and lie behaviours in interview contexts continues to emerge. However, irrespective of culture, lying is often more cognitively demanding than telling the truth [8,42–46] and so the cognitive challenges of formulating and maintaining a deceptive account in response to investigative questioning appear robust and pancultural.

To appear credible, all deceivers must first formulate a sensible lie and maintain their lie script in response to questions, while holding back the truth to avoid slips of the tongue. Liars must also respond appropriately to follow-up clarification or challenge questions as an interview progresses. Clarification can be in the form of asking for further event details or asking interviewees to repeat answers to monitor consistency. Challenges in an interview typically concern directly questioning whether a response is veridical, or the interviewer might introduce information that has not previously been revealed which casts doubt on the veracity of previous responses. Either way, deceptive interviewees must manage several enduring concurrent cognitive processes [46–49] throughout an interview. In contrast, telling the truth requires the interviewee to recall and verbalise an experienced event, which invokes fewer cognitive resources.

Despite widespread perceptions to the contrary, physical behaviours such as gaze avoidance, self-adaptors, and nervousness have not been found to be reliable cues to veracity because of individual, contextual and cultural variances [50–52]. However, several observable verbal behaviours predicated by cognitive load theories of deception and the impact of attempting to manage several concurrent cognitive processes to appear credible have begun to emerge as potentially stable cues, irrespective of culture and/or language. Deceiver verbal accounts are often impoverished and simple, whereby they provide fewer event specific details than truth-tellers [31,42,50–52]. Although appearing to provide informative accounts, in fact deceivers often offer general knowledge information rather than detailed event details that could be verified, for example. Guiding observers to rate detailedness and the amount of verifiable event information has been found to support improved detection of deception performance beyond chance for observers in some circumstances [38].

Truthful accounts are also often more plausible, that is they sound more sensible, likely, or believable. Judgements of plausibility, in terms of how ‘believable’ an account is, have been found to distinguish truth tellers from liars in some contexts whereby plausibility ratings for deceptive accounts are typically lower than for truthful accounts [30,31,50,51]. Furthermore, observer plausibility ratings have been reported to positively predict the detailedness of an account [40,41]. Lay-observers may, therefore, be able to recognise variable verbal behaviours such as detailedness and plausibility for guiding them in making real-time truth and lie judgments. However, questions arise centred on the real-world challenges for lay-observers.

Cut-off indicators do not exist for the amount of detail an interviewee may or may not provide in response to individual questions or across the interview as a whole, and so subjective assessments are necessary [40,41]. Likewise, plausibility and believability also require subjective assessments concerning how individuals understand how sensible or likely an account is [41]. Subjective assessments are guided by cultural norms, expectations and understandings and so it is important to investigate whether these emerging cues to truth and lies are intersubjective [42]. That is, whether context germane to culture leverages space that might make detailedness and plausibility less important or less valid subjective cues for some cultures.

Second, applied researchers use various paradigms whereby lay-observer participants often view multiple interviews or read multiple interview transcripts where some participants have been incentivised to be deceptive while others are instructed to be truthful [31,40,43,47,53,54]. This approach has been vital for improved understanding of the efficacy of various types of interview approaches for amplifying cues to truth and lies, and how veracity judgments are made. However, in reality lay observers may only have access to a single interview (audio, transcript or video), or one or two relevant portions of an interview which does not allow group performance comparisons. Access to just one interview does not support relative nor comparative verbal behaviour judgements as often occurs in laboratory research paradigms where anomalous verbal behaviour can sometimes be easier to spot across a group, or cohort [55,56]. In such instances, observers can be alerted to truth and lies in a manner that does not map onto the real world where they may be asked to make an absolute judgment based on the behaviours apparent in one interview or transcript.

The current research

Using first language as a proxy for cultural origins, the research reported here advances understanding of the utility of detailedness, believability and plausibility as veracity cues in forensic interview contexts for lay-observers from a non-western culture, and investigates truth and lie cues attended too. To reflect real-world conditions, observers have access to just one video interview conducted in Arabic, following which they are immediately asked to make a real-time truth or lie judgment. Despite Arabic being one of the most spoken languages worldwide, as far as we can ascertain, lay observer truth and lie judgments of this significant cultural group have received little attention using single culture paradigms. Here, interviewees are mock persons of interest, known to have IRI. Interviews are conducted in Arabic using an information gathering interview forensic protocol. Lay observers all have Arabic as their first language and are asked to self-identify their ethnic/cultural background.

The psychological literature of direct relevance to the to the focus of this research is extremely sparce, and research findings are mixed. However, cultural differences in perceptions of truth and lies behaviours are often reported which can undermine credibility in cross-cultural interview contexts, whereby misunderstanding and misinterpretation typically triggers a lie bias [28,32,33,57–59], although not always [60,61]. To date, it appears researchers have yet to investigate the detection of truth and lies by lay-observers in non-western culture and language matched forensic contexts where observers are unable to make comparative judgments, as is typically the case in the real-world.

The cognitive challenges associated with verbal deception are believed to be pancultural, and so verbal behaviours triggered by cognitive demand, such as levels of event detailedness, believability and plausibility, for example, may transcend culture and language, or they may not. While individuals from Western individualist cultures tend to provide more information than those from non-Western collectivist cultures [62–65], some research has indicated that all truth tellers generally report more details than deceivers, including in Arab and south Asian cultures [30,31,56,65]. Given the novelty of this research, rather than hypothesising, we formulated a series of research questions towards advancing pancultural understanding with reference to the real-world challenges, current empirical understanding and associated empirical questions raised by those concerned in developing guidance for practice. It is these questions that guided our paradigm and analysis approach, as follows.

First, we examined lay-persons absolute judgments of truth and lies having seen a single within culture interview conducted in matched first language. Consistent with findings from research with other within cultural groups, we expected a truth bias to emerge, and that confidence would differ across veracity conditions.

Second, we examined subjective assessments of plausibility, believability and detailedness as veracity cues, asking whether variations across deceiver and truthteller accounts are discernible and impactful for guiding veracity judgements having been exposed to just one interview. Consistent with a very limited amount of previous research, we expected lay-persons to rate liars as less plausible, less believable and less detailed than truthtellers. We use the terms plausible and believable since although virtually synonymous, statements that are believable are likely to be correct or true whereas plausible refers to accounts being reasonable in terms of being correct or true. It is unclear whether plausible would be robustly and similarly interpreted across cultural groups, since the term has emerged from Western research paradigms. Research from other domains suggests this may not necessarily be the case [66,67] and so we include believable to cross validate and as a potential additional cue.

Third, we examined the types of cues attended to. We recognise the limited amount of culture specific prior research in this area, nonetheless we expected lay-observers attention would be drawn to behaviours that have previously been found to be pseudo diagnostic.

Ethics

This research was ethically approved by the University of Westminster Research Ethics Committee: ETH1617-0528 and the US Dept of Justice FBI institutional Review Board Docket No. 353–16 and was conducted in accordance with the British Psychological Society and Health and Care Practitioner Council codes of ethical conduct. All participants were adults (18 years and older). In line with internationally recognised ethical best practice, opt-in consent was gained from each lay observer participant via the Qualtrics research platform. Participants first read an information screen explaining information about anonymity, what they would be asked to do and how they were able to withdraw only prior to submission of their responses since the study was anonymous, withdrawal post submission would be impossible. Participants were offered the chance to contact the researchers with any additional questions. Participants then moved forward to a consent screen. Participants who did not consent (by clicking the ‘I do not consent’ button), were not able to take part in the research and so were directed to a generic ‘thank you and debrief screen’. Those who consented moved through the research process in Qualtrics as described. Digital records of consent are stored on the Qualtrics platform. The lay-observer data reported in this research was collected between 1^st June 2023 and 28^th August 2023.

Forensic interview videos

The two videos used for this research are of interviewees who had been involved in a live event as described below, which was designed to mimic the experiences of some individuals who might later be interviewed as a person of interest. Interviewee performance data and performance has been previously fully reported and the paradigm has also been reported in full and [8], but for replication purposes and manuscript clarity a brief description of the paradigm is provided.

Culturally matched participants experienced the target event in groups of six. Embedded in each participant group was a culturally matched confederate (C1) playing the role of a participant. Confederate 2 (C2) played the role of researcher (C2). C2 greeted participants and ran the session during which the target event occurred. Participants were primed to be sympathetic to C1 as follows: i) placing C1 within the participant group (C1 arrived and interacted with participants, completed the same tasks, etc.) created conditions for in-group favouritism; ii) the scenario (the manner in which the event unfolded, and verbal exchanges that took place between C1 and C2 during the event); and iii) C2 acting as researcher in charge of the session, creating a perceived imbalance of social influence, in which C2 was portrayed as more powerful than C1.

Both interviewees were male aged 37 (interviewee A) and 38 (interviewee B) years old. Both self-identified as Arabic and Middle Eastern, were bi-lingual and born outside of the UK, with Arabic as their first/native language they had used at home with their parents and family from birth and used at school. English was a second language that interviewees learned from the age of 12 years onwards at school. Both interviewees were paid $30 each to participate and were further incentivised to withhold information (be deceptive) with the offer of an additional payment of $60 dependent upon their interview performance. Of the two interviewee participants taking part in the research reported here, one complied with this instruction (interviewee B), the other did not (interviewee A), thus the interview data comprised one interview with a truthteller one with a deceiver. The interviewee research procedure involved four phases:

1. Interviewees were provided with an information and consent sheet following. Signed consent and audio recorded consent was collected. Following consent, interviewees moved as a group to a seminar room to complete two research questionnaires.
2. While completing the first questionnaire, an unexpected event took place, involving C1 and C2 comprising a verbally aggressive altercation during which a laptop computer was seriously damaged–the entire session, from entering to leaving the event room lasted approximately 25 minutes.
3. Following the event, interviewees were then individually interviewed, and as occurs during real criminal investigations in many jurisdictions (e.g., UK, across the EU and some US States, Australia, and Ireland) and for applied research of this nature, interviews were digitally audio and video recorded. Interviews followed an interview protocol (see Appendix A in S1 Appendix).

Interviewer

The interviewer was born in the UAE and spoke Arabic as a first language but was fluent in English as a second language. The interviewer underwent a half day of classroom training (given by the first author, an experienced interviewer), which included a detailed explanation of the relevant interview protocol and role-play practice. The interviewer also took part in an additional half-day practice session prior to conducting interviews, which was audio recorded. Detailed verbal feedback was provided on adherence to the protocol. The interviewer was naïve to the design and experimental hypotheses but was provided with the following instructions ‘The researcher’s computer was seriously damaged during the data collection session. Your job is to interview the people in the room and find out exactly how the damage happened, using the interview protocol’.

Interview protocol

Both interviewees were interviewed using the Framed Controlled Cognitive Engagement (Framed—CCE) protocol [3,37]. Interviews comprised five discrete phases as follows: 1) explain and build rapport, 2) free account, 3) probed questioning, 4) challenge, and 5) closure (for protocol see Appendix A in S1 Appendix).

Lay observer participants

The data collection procedure for lay observers was identical across conditions other than in the cue direction condition where participants were provided with one additional clear instruction prior to viewing the interview, which was to take notice of the amount of detail provided by the interviewee about the event, and to rate the amount of detail on a scale (see materials and Appendix B in S1 Appendix). Further, that the detail rating should then be used to guide their veracity decision. The research was hosted on Qualtrics. Participants were recruited via Prolific platform where participants were screened for age (+ 18 years) and language (reading and speaking Modern Standard Arabic as first language and English as second). Qualifying participants were then directed to Qualtrics via a one-time link where further demographic questions were included (see Appendix B in S1 Appendix).

In both conditions (cue direction and no cue direction) all participants were provided with full information about the study, followed by consent information (in Modern Standard Arabic). All participants then completed a series of demographic questions (age, country of birth, parents first language and schooling). Although we used first language (here Arabic) as a proxy for culture, we also asked participants to indicate their cultural identity in terms of their self-identity and self-perception (see Appendix C in S1 Appendix for self-identity data). Participants were provided with the following written information (in Arabic): ‘that the person being interviewed was known to be present during a serious incident where a university laptop was seriously damaged’. Participants in the cue direction condition were further instructed to ‘take note of the degree to which the interviewee includes details such as descriptions of people, places, actions, objects, events, and the timing of events. Think about the degree to which the message of the interviewee seemed complete, concrete, striking, or rich in details’ [35] It was further explained that they would be asked to rate the interviewee’s verbal responses to questions for detailedness after they had seen the video, but before making a veracity decision.

All participants then watched the video (either A or B according to veracity condition randomly assigned, conducted in Arabic). Immediately after having viewed the video interview, participants in the cue direction condition were asked to i) make a detailedness judgement ranging from 1 (no detailed) to 10 (very detailed) and ii) were instructed to use the detailedness judgement score to guide their truth or lie dichotomous decision (where 5 and below should trigger a deception decision). Participants in the no cue direction condition did not receive these instructions and so did not make a detailedness judgment. Rather, they simply made a dichotomous truth or lie decision (see Appendix B in S1 Appendix for materials).

All participants completed plausibility and believability scales (all ranging from 1 to 10) and were asked to indicate the types of behaviours (verbal only, behavioural only or a mix of both) that had been important in helping them to make a veracity decision. Participants who had indicted verbal behaviours had been important/partially important were asked to think about the way in which the interviewee had verbally responded to questions and to indicate all that applied from a range of known present verbal behaviours. (see Appendix C in S1 Appendix for materials).

Stimulus interviews

Each participant listened and watched only one of two information-gathering style interviews. Each interview was conducted in Arabic by the same bilingual interviewer whose first language was Arabic. Both interviews followed the same verbal protocol (see Appendix A in S1 Appendix). In contravention of the confederate (C3) researcher instructions (see procedure), the interviewee in interview A answered all questions truthfully and so did not withhold information nor formulate a deceptive account of the target event. In interview B (deceiver), as instructed by C3 the interviewee was deceptive by withholding elements of the target event, and so provided an account that was deceptive in part (see procedure). Interview A lasted 11.38 minutes. Interview B lasted 11.35 minutes.

The interviews selected for this research are part of a larger, part-published dataset and were selected to control for potentially confounding variances. First all interviews were translated, transcribed and then coded for target event details. We broadly defined details as verbalised event information provided by the interviewee in response to questions, such as (but not limited to) descriptions of the people involved in the incident in question, clear descriptions of the place and time that the incident took place, information regarding the two confederates actions (including speech), the actions of others present (including speech), the objects involved, and the timing of events. Interview A and B were selected because they differed for event detail verbalised by the interviewee whereby interviewee A provided 41 event details, whereas interviewee B provided 17. Hence, there was an 141% difference in the number of event specific details provided across the two interviews.

Second, both interviewees were well matched. Both were Arabic men, aged 33 and 35 respectively and of a similar west Asian appearance with short dark/black hair. Both were clean shaven but with some dark beard regrowth/shadow, and both were wearing casual clothes (t-shirt & shirt) with no visible tattoos, scars nor other distinguishing markings. Both spoke in Arabic from the start to the end of the interview, were interviewed in the same room and at the same position/angle to the interviewer. Both had been born and raised in Libya and had completed their formal schooling in Libya with English as a second language learned after the age of 16 years. Both interviews were of a similar duration (approx. 11 mins 30 seconds each).

Design and analysis approach

Between subjects 2 Veracity (truth; lie) X 2 Direction (No cue direction; Cue direction) design was employed, collecting dichotomous veracity decision data and scale data for plausibility, confidence and believability from each participant (see procedure and materials). To examine lay-persons absolute judgments of truth and lies we conducted a three-way loglinear analysis of veracity decisions to provide an odds ratio measure of association between an exposure (veracity condition) and outcome (veracity decision). We conducted a series of 2 X 2 ANOVAs to investigate subjective assessments of plausibility, believability and detailedness as veracity cues as a function of veracity condition and cue condition, reporting main effects and interactions. To investigate the relative importance of veracity cues, we conducted Freidman test repeated measures analyses of variance by ranks, followed by Mann-Whitney U across cue conditions (Cue Vs. No Cue).

Lay observer participants

A total of 217 bilingual participants, fluent in Modern Standard Arabic as a native/first language, took part as lay observers. One hundred and one females (46.5%), 112 (51.6%) males, and one participant who preferred not to say. Mean age was 27.14 years (SD = 7.12), ranging from 18 to 63 years. While sex and gender are reported, neither are theorized to impact our findings and so sex nor gender-based analyses were not carried out. A-priori power analysis conducted using G*Power 3.1 to determine minimum sample size estimation to detect a practically meaningful medium effect size for applied research of this nature was N  =  179 (assuming power  =  0.80, effect size F 0.25, a  =  0.05). Thus, the obtained sample size of N  =  217 was adequate given resource constraints and access to Arabic speaking populations and is in line with sample size norms described in many empirical cross-cultural studies similar to the one reported here [28,31,60,61,68,69].

Observer participants were only able to take part in one condition and were blind to the existence of the additional studies and conditions and were excluded/screened from other conditions via the Prolific targeted screening and exclusion facility. Two hundred participants were recruited from the general population via Prolific and participated during 1^st July and 11^th August 2023. The remaining 17 were recruited via social media, word of mouth and snowballing, and participated in the first week in August 2023. Mean age as a function of condition veracity X direction condition did not differ significantly, F(3, 216) = .149, p = .930, likewise gender distribution, X² (3, 216)  =  5.049, p  =  .536.

Veracity decisions

Overall, the dichotomous truth/lie decision data revealed a truth bias. Irrespective of condition (truthteller, deceiver, cue direction, no cue direction), 143 (66%) of observers made a truth decision whereas 74 (34%) made a lie decision. Across conditions, counts of correct and incorrect deception and truth decisions are shown in Fig 1 (below).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Veracity decision count as a function of condition (N = 217).

https://doi.org/10.1371/journal.pone.0310384.g001

A three-way loglinear analysis of veracity decisions produced a final model that retained all effects. The likelihood ratio of this model was χ² (0) = 0, p = 1 indicating the higher order model (veracity condition X cue direction condition X veracity decision) was significant, χ² (1) = 14.105, p < .001. There was a significant association between veracity condition (truth, deception) and correct veracity decisions, χ² (1, 217) = 16.471, p < .001. Eighty-six lay-observers (79%) in the truth condition correctly judged the interviewee was truthful, whereas in the deception condition 51(47%) correctly judged the interviewee as deceptive. The odds ratio of making a correct veracity decision were 19.32 times higher in the truth condition than the deception condition.

Chi-square (χ²) probability distribution revealed a significant association between cue direction (no cue direction, cue direction) and correct veracity decision, χ² (1, 217) = 11.471, p < .001. Twenty-six lay-observers (23%) in the cue direction condition correctly judged the interviewee as deceptive whereas in the no cue direction condition 48 (43%) correctly judged the interviewee as deceptive. The odds ratio of correctly deciding the interviewee had been deceptive was 2.56 times higher in the no cue direction condition than the cue direction condition. Loglinear analysis indicated lay observers were more likely to make a correct veracity decision when observing a truthteller than a deceiver and were less likely to correctly judge the interviewee as deceptive when provided with cue directions to make a heuristic detailedness judgement to guide their veracity decisions.

Confidence scale

There was a significant main effect of veracity condition for confidence ratings (on a scale from 1 to 10, where 1 = at all confident and 10 = completely confident), F(1, 213) = 7.938, p = .005, η_p² = .36. Participants in the lie condition were less confident (M = 6.60, SD = 1.93, 95% CI, 6.27, 6.94) than those in the truth condition (M = 7.28, SD = 1.60, 95% CI, 6.95, 7.62). The main effect of cue direction (M _{no direction} = 6.76, SD = 1.89, 95% CI, 6.42, 7.10; M _{cue direction} = 7.12, SD = 1.70, 95% CI, 6.79, 7.45) and the veracity X direction interaction were non-significant, all Fs < 2.243, all ps > .136 (see Table 1 below for interaction means, SDs and 95% CIs).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mean interaction confidence scale ratings (N = 217).

https://doi.org/10.1371/journal.pone.0310384.t001

Believability scale

The main effects of veracity and cue direction for believability scale ratings (on a scale from 1 to 10, where 1 = not at all believable and 10 = completely believable) were significant, F(1, 212) = 18.701, p < .001, η_p² = .78, and, F(1, 212) = 6.263, p = .013, η_p² = .29, respectively. Participants in the truth condition rated the interviewee as more believable (M = 6.83, SD = 2.31, 95% CI, 6.40, 7.26) than those in the deception condition (M = 5.50, SD = 2.26, 95% CI, 5.07, 5.93) and participants in the cue direction condition rated the interviewee as more believable (M = 6.55, SD = 2.48, 95% CI, 6.13, 6.97) than those in the no direction condition (M = 5.78, SD = 2.21, 95% CI, 5.35, 6.22). The veracity X cue direction interaction was non-significant, F = .062 p = .804 (see Fig 2 below for interaction means).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Believability and plausibility scale interaction means (N = 217).

https://doi.org/10.1371/journal.pone.0310384.g002

Plausibility scale

The main effect of veracity for plausibility scale ratings (on a scale from 1 to 10, where 1 = not at all plausible and 10 = completely plausible) was significant, F(1, 212) = 15.257, p < .001, η_p² = .67. Participants in the truth condition rated the interviewee as more plausible (M = 6.75, SD = 2.10, 95% CI, 6.34, 7.16) than those in the deception condition (M = 5.59, SD = 2.27, 95% CI, 5.18, 6.01). The main effect of cue direction (M _{no direction} = 5.93, SD = 2.15, 95% CI, 5.51, 6.35; M _direction = 6.41, SD = 2.35, 95% CI, 6.00, 6.82) and the veracity X cue direction interaction were both non-significant, all Fs < 2.575, all ps > .110 (see Fig 2 below for interaction means).

A repeated measures ANOVA comparing plausibility and believability ratings across conditions revealed non-significant main effects and a non-significant interaction, all Fs < 5.790, all ps, > .017 across conditions.

Detailedness

Only participants in the cue direction condition completed a detailedness scale. Analysis as a function of veracity decision revealed participants in the truth condition rated the interviewee’s account as significantly more detailed (M = 7.07, SD = 1.46, 95% CI 6.63, 7.46) than those in the deception condition (M = 5.96, SD = 1.78, 95% CI 5.43, 6.39), F(1, 109) = 14.078, p < .001, η_p² = .11.

Veracity cue scale

Analysis of the cues lay observers indicated were important when making a veracity decision revealed significant main effects of veracity and cue direction for cue scale ratings (1 = verbal only, 2 = mainly verbal but some behaviour, 3 = verbal and behaviour equally, 4 = mainly behaviour but some verbal, 5 = only behaviour), F(1, 213) = 18.775, p < .001, η_p² = .81, and, F(1, 213) = 17.705, p < .001, η_p² = .92, respectively. Irrespective of cue direction condition, participants in the truth condition (M = 2.79, SD = .97, 95% CI, 2.62, 2.96) indicated that they had mainly listened to what the interviewee was saying but took some notice of how the interviewee behaved. Conversely, participants in the lie condition had mainly taken notice of how the interviewee behaved but took some notice of what the interviewee said (M = 3.32, SD = .91, 95% CI, 3.15, 3.49).

In a similar vein, irrespective of veracity condition, participants in the cue direction condition (who had been cued to note how much detail the interviewee provided and to rate detailedness) indicated they had mainly listened to what the interviewee was saying but also took some notice of how the interviewee behaved when making a veracity decision (M _{cue direction} = 2.77, SD = .82, 95% CI, 2.60, 2.94 –see Table 2 below for interaction means).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Cue scale interaction means, SDs and 95% CIs (N = 217).

https://doi.org/10.1371/journal.pone.0310384.t002

Participants in the no cue direction condition mainly took notice of how the interviewee behaved but took some notice of what the interviewee said (M _{no direction} = 3.34, SD = 1.03, 95% CI, 3.16, 3.51). The cue condition X veracity interaction was non-significant F = .001, p = .974.

Verbal behaviours

Overall, of the 217 lay-observers, 156 (72%) responded that they had primarily or equally attended to verbal behaviours (by responding 1, 2 or 3 on the veracity cue) of which just 10 (4%) responded they had only attended to verbal behaviour, 50 (23%) that they had mainly listened to what the interviewee said but had taken some notice of how the interviewee behaved, while 96 (44%) stated they had attended equally to what the interviewee said and how they behaved.

Participants indicated which verbal behaviours had been influential when making veracity decision (they were able to indicate more than one behaviour). A Freidman test to rank the verbal behaviours attended to was significant, χ² (156) = 21.232, p < .001 (see Table 3 below). Overall, self-corrections were ranked as most influential and providing a complicated account was ranked as least influential.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Mean ranking, numbers of responses and percentages for verbal behaviours (n = 156).

https://doi.org/10.1371/journal.pone.0310384.t003

Further Freidman tests to rank the verbal behaviours attended as a function of cue direction condition was significant for the no cue direction condition, χ² (46) = 20.181, p < .001, (see Table 4) but non-significant for the cue direction group, χ² (110) = 9.032, p = .108 (see Table 4). Again, self-corrections were ranked as most influential for the non-cue direction group and providing a complicated account was ranked least influential.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Mean ranking for verbal behaviours as a function of no cue (n = 46) and cue direction condition (n = 110).

https://doi.org/10.1371/journal.pone.0310384.t004

Mann-Whitney tests revealed significant differences in verbal behaviours attended too as a function of cue direction condition (direction; no direction) for consistency, U = 1920.00 p = .006 (Mean Rank _{No direction} = 65.24; Mean Rank _{Cue direction} = 84.05), complexity, U = 1967, p = .008 (Mean Rank _{No direction} = 66.26; Mean Rank _{Cue direction} = 83.62), and detailedness, U = 2044.00, p = .029 (Mean Rank _{No direction} = 67.93; Mean Rank _direction = 82.94). Participants in the cue direction group attended to all three verbal behaviours more than those in the no cue direction. Mean rankings for all other verbal behaviours were non-significant across the two groups, all Us > 2136.00 ps, > .075, indicating participants attended equally across conditions.

Results summary

The truthful interviewee was accurately detected by 79% of observers whereas accuracy dropped to just below chance levels (47%) in the deceptive interviewee condition. Overall, the truthful interviewee was rated as providing a more detailed account than the deceptive interview. However, giving specific instructions to look for the detailed responses for cueing truth and lies significantly undermined observer ability to accurately detect the deceptive interviewee, falling to 23% accuracy. Comparatively, detection of deception performance improved in the absence of cue instructions, but again performance was still below chance at 43% accuracy.

Confidence ratings were higher in the truthful interviewee condition and confidence remained stable irrespective of detailedness cue instructions. Overall, the truthful interviewee was rated as more believable and more plausible than the deceptive interviewee. Detailedness cueing did impact believability ratings whereby lay observers in the cue condition rated the interviewee as more believable.

Participants viewing the truthful interviewee stated that they mainly took account of verbal behaviour when making their veracity decision but took some notice of nonverbal behaviours. Conversely, those viewing the deceptive interviewee mainly took note of nonverbal behaviour but took some account of what the interviewee said. This pattern of results was mirrored across the cue conditions whereby participants in the cued condition mainly took account of verbal behaviour but took some notice of nonverbal behaviour while those in the no cue condition mainly noted nonverbal behaviour but took some note of verbal behaviours.

Overall, the verbal behaviour used by all lay observers to cue their veracity decision was ‘correcting answers’ whereas the least important was ‘giving a complicated account’. The pattern of cues attended too across all conditions was similar.