From imposing cognitive load to exploiting different strategies: A reply to Brimbal et al. (2023)

The reasoning behind lie detection based on imposing cognitive load is as follows. In interview settings lying is typically more mentally taxing than truth telling (Gombos, 2006; Walczyk et al., 2013). Theoretically, interviewers could exploit this difference by imposing cognitive load. This should affect lie tellers more than truth tellers because lie tellers will have fewer cognitive resources left over than truth tellers. Asking interviewees to recall their stories in reverse order is one way of imposing cognitive load. Although we (Vrij et al., 2012) and others (Evans et al., 2013) found support for the reverse order lie detection technique, Brimbal et al. (2023) found no support. After failing to replicate our findings, Brimbal et al. (2023) conclude that there may not be sufficient evidence to train the reverse order technique in the field. We came to this conclusion in Vrij and Fisher (2016)—an article cited by Brimbal et al. (2023)—amongst other reasons because the accuracy rates obtained by imposing cognitive load lie detection techniques are quite low. Two meta-analyses supported this conclusion: 58% accuracy rate for truth tellers and 63% for lie tellers in Vrij et al. (2017) and 58.39% (total accuracy rate) in Mac Giolla and Luke (2021). In other words, we are in full agreement with Brimbal et al. (2023) about the usefulness of using reverse order recall as an imposing cognitive load lie detection technique.

We still believe in the principle of imposing cognitive load lie detection but have difficulty translating it into a valuable lie detection technique. Apart from reverse order recall, we examined other ways to impose cognitive load, such as instructing interviewees to maintain eye contact with the interviewer (Vrij et al., 2010) or to remember a car registration number plate during the recall (Vrij, Deeb, et al., 2022). We do not recommend introducing such techniques either. The observable differences were not particularly strong and, although we can give interviewees such imposing cognitive load instructions in the lab, we do not consider it feasible in real life. The closest we came to a successful implementation of imposing cognitive load is in a collective interviewing scenario (interviewing pairs of interviewees together) where we used the forced turn-taking technique (Vernham et al., 2014). In forced turn-taking, the interviewer asks one interviewee to starts answering a question. After a short period of time the interviewer will then interrupt the interviewee and will ask the second interviewee to continue with the story. After again a short period of time that person is interrupted, and the first interviewee is asked to continue. Compared to lie telling pairs, truth telling pairs had more fluent continuations in their accounts, whereas lying pairs were more likely to repeat what their partner last said before continuing. Lie telling pairs also waited longer than truth telling pairs with speaking after a turn-take request. A lie detection test revealed an 76.7% accuracy rate when observers were asked to pay attention to these three cues (continuations, repetitions, waiting).

Brimbal et al. (2023) replicated our finding that truth tellers sounded more plausible than lie tellers. We think this is worth mentioning because we are increasingly interested in plausibility. It typically emerges as a strong (and often the strongest) veracity indicator in our research, see Vrij, Deeb, et al. (2021) for a review, but also Vrij, Leal, et al. (2022) and Chandler et al. (2023). Other researchers also found plausibility to be the strongest veracity indicator (Sporer et al., 2021). Plausibility is often used as a veracity cue by laypersons (Hartwig & Bond, 2011) and by practitioners (Vrij et al., 2023), including in asylum interviews (UNHCR, 2013). Several verbal deception researchers, however, do not examine plausibility, although their deception scenarios are well suited for plausibility measurements. We encourage them to start examining plausibility. We think that the verbal lie detection domain cannot ignore examining a cue that works well as a veracity cue in research and is frequently used in the field.

We found some positive effects of reverse order lie detection recall, but Brimbal et al. (2023) did not. Comparing experiments is problematic and explaining the differences between Vrij et al. (2012) and Brimbal et al. (2023) is speculative. First, the effects Vrij et al. (2012) found were rather weak so there was always a fair chance that it would fail the replication test. In addition, truth tellers and lie tellers in Vrij et al. (2012) were considerably more motivated than those in Brimbal et al. (2023). According to Brimbal et al. (2023, p. ????) motivation is unlikely to have impacted their results, citing Hartwig and Bond (2011). We disagree. Hartwig and Bond (2011) did not measure whether motivation affects interviewees' responses in strategic interview protocols. We can easily reason why it would. Truth tellers employ a ‘tell it all’ strategy in interviews (Granhag & Hartwig, 2008). This requires considerable effort, even more so in reverse order recall because that is mentally taxing. Truth tellers will put less effort in reporting information if they are less motivated. The less they say, the more they will sound like lie tellers.

Brimbal et al.'s (2023) lie detection test took place online via MTurk whereas ours took place in-class. MTurk lie detection studies are easy to run and inexpensive, but we find them remote from a real life lie detection experience. Our experience with running in-class lie detection experiments is that observers quickly lose attention. Did the observers in Brimbal et al. (2023) pay attention to the stimulus materials? No check was introduced but attention checks are nowadays recommended (Hauser & Schwartz, 2016). The MTurk participants may not have paid much attention because they could make veracity judgements regardless. Of course, attention checks should also be introduced during in-class lie detection tests.

Our thinking about lie detection has developed since Vrij et al. (2012). Imposing cognitive load is no longer prominent. The rationale of cognitive credibility assessment (CCA) is that the strategies used by truth tellers (tell it all) and lie tellers (keep it simple) are different (Strömwall & Willén, 2011) and that these differences can be exploited. Truth tellers do not spontaneously report all information (Vrij et al., 2014) because (i) they do not know how much they are required to say, (ii) are not motivated to tell it all or (iii) find it difficult to retrieve information from memory. CCA uses techniques to facilitate these three aspects and that includes reverse order recall, a memory-enhancement tool (Vrij, Granhag, et al., 2022; Vrij, Mann, et al., 2021). These techniques should have a larger effect on truth tellers than on lie tellers because lie tellers prefer to keep their stories simple to (i) avoid reporting incriminating evidence, (ii) remain consistent if interviewed multiple times and (iii) make story telling easier.

Others also include reverse order recall in their protocols (Bogaard et al., 2019; Colwell et al., 2009). The reverse order recall is part of the cognitive interview (Fisher & Geiselman, 2019), an interview tool that facilitates memory recall in cooperative witnesses (truth tellers) (Memon et al., 2010). Similar to the cognitive interview, in CCA, after some recall attempts, interviewees are invited to report the event again but this time in reverse order. The reverse order recall instruction invites truth tellers to think about the event again from a different perspective, which often leads to new information. A reverse order recall is less likely to lead to new information in lie tellers due to their inclination to keep their stories simple. In other words, nowadays we—and others (Bogaard et al., 2019; Colwell et al., 2009)—use reverse order recall as a memory-enhancement tool in a cluster of ‘encouraging interviewees to say more’ techniques (Vrij, Mann, et al., 2021). We predicted and found that truth tellers report more additional information after a reverse order recall instruction than lie tellers (Ewens et al., 2016; Shaw et al., 2014; Vrij, Mann, et al., 2021).

We started our cognitive lie detection research with imposing cognitive load manipulations (e.g. Vrij et al., 2008) but moved on from that. We asked researchers to explore cognitive lie detection (Vrij et al., 2008) and many researchers took this up. Brimbal et al. (2023) propose more replication studies. Others prefer to carry out innovative research by designing new techniques, such as the strategic use of evidence (Granhag & Hartwig, 2015; Oleszkiewicz & Watson, 2021) and the verifiability approach (Nahari, 2019).