Evaluation of the NITV CVSA

The purpose of this study was to evaluate a commonly used voice stress analyzer, the National Institute of Truth Verification's (NITV) Computer Voice Stress Analyzer (CVSA), using a speech database containing materials recorded (i) in the laboratory, while highly controlled deceptive and shock-induced stress levels were systematically varied, and (ii) during a field procedure. Subjects were 24 each males/females (age range 18-63 years) drawn from a representative population. All held strong views on an issue and were required to make sharply derogatory statements about it. The CVSA system was then evaluated in a double-blind study using three sets of examiners: (i) two UF scientists trained/certified by NITV in CVSA operation, (ii) three experienced NITV operators provided by the manufacturer and (iii) five experimental phoneticians. The results showed that the "true positive" (or hit) rates for all examiners ranged from chance to somewhat higher levels (c. 50-65%) for all conditions and types of materials (e.g., stress vs. unstressed, truth vs. deception). However, the false-positive rate was just as high - often higher. Sensitivity statistics demonstrated that the CVSA system operated at about chance level.     

The Accuracy of Auditors' and Layered Voice Analysis (LVA) Operators' Judgments of Truth and Deception During Police Questioning*

The purpose of this study was to determine if auditors could identify truthful and deceptive persons in a sample (n = 74) of audio recordings used to assess the effectiveness of layered voice analysis (LVA). The LVA employs an automated algorithm to detect deception, but it was not effective here. There were 31 truthful and 43 deceptive persons in the sample and two LVA operators averaged 48% correct decisions on truth‐tellers and 25% on deceivers. Subsequent to the LVA analysis the recordings were audited by three interviewers, each independently rendering a decision of truthful or deceptive and indicating their confidence. Auditors' judgments averaged 68% correct decisions on truth‐tellers and 71% on deceivers. Auditors' detection rates, generally, exceeded chance and there was significantly (p < 0.05) greater confidence on correct than incorrect judgments of deceivers but not on truth‐tellers. These results suggest that the success reported for LVA analysis may be due to operator's judgment.     

Voice stress evaluators and lie detection

Proponents of vocal stress analysis systems argue that they are able to detect spoken deception by analysis of "stress" in the voice signal. Presumably, they do so by examining traces made by laryngeal microtremors which, they claim exist in the voice, are associated with stress, and ultimately are associated with lying. However, most research that seeks to identify the relationships between microtremors and laryngeal function has produced negative results, and data on the ability of voice analyzers to detect stress from speech--or to identify spoken deception--have been negative or "mixed" in nature. Since perspectives based on available results leave a number of questions unanswered, a series of experiments has been undertaken. The first was focused on the basic acoustic/temporal correlates of stress in voice (the subject of an earlier report), the second on examination of stress by commercial voice analyzers, and the third on the detection of relatively high-risk lies by this same type of voice analysis procedure. It was found that correct stress/nonstress identifications occurred only at chance levels; the lie/nonlie identification scores were quite similar with professional "examiners" performing at about the same level of accuracy as other auditors. The following review is divided into two parts: a history of the controversy and a presentation of the two cited experiments.     

基于语音情感分析系统的语音压力测谎

目的探究语音情感分析系统(Layered Voice Analysis,LVA)在不同情况下的表现,并通过分析数据找出影响系统识别率的因素。方法设置两种不同场景对20名研究对象进行提问,并录音。通过LVA语音系统对语音样本进行分析,使用SPSS统计软件对系统生成的参数进行统计分析。结果在未告知研究对象实验真相的情况下LVA语音系统平均识别率为91.75%,要远高于研究对象在知情情况下的识别率。在询问过程中,当未涉及敏感问题时,随着时间的推移研究对象的心理压力逐渐减小。结论LVA语音系统具有良好的情感分析和识别能力,可为司法工作提供参考。     

Detecting deception in suspects: verbal cues as a function of interview strategy

Research on deception has consistently shown that people are poor at detecting deception, partly due to lack of consistent cues to deception. This research focuses on eliciting verbal cues to deception when questioning suspects who deny crime and how such cues differ due to type of questioning. An experiment examined verbal differences between innocent and guilty mock suspects (N=96) as a function of veracity and interview style (Free recall, Probes, or Free recall plus Probes). Guilty (vs innocent) suspects omitted more crime-relevant information and their statements were more likely to contradict the evidence, showing that statement–evidence inconsistency was a cue to deception. This cue to deception was more pronounced when the interview contained probes. Lie-catchers (N=192) obtained an accuracy rate higher than chance (61.5%) for detecting deceptive denials. Implications for further research on verbal cues to deception are discussed.     

Repeatability and reproducibility of comparison decisions by firearms examiners

In a comprehensive study to assess various aspects of the performance of qualified forensic firearms examiners, volunteer examiners compared both bullets and cartridge cases fired from three different types of firearms. They rendered opinions on each comparison according to the Association of Firearm & Tool Mark Examiners (AFTE) Range of Conclusions, as Identification, Inconclusive (A, B, or C), Elimination, or Unsuitable. In this part of the study, comparison sets used previously to characterize the overall accuracy of examiners were blindly resubmitted to examiners to assess the repeatability (105 examiners; 5700 comparisons of bullets and cartridge cases) and reproducibility (191 examiners of bullets, 193 of cartridge cases; 5790 comparisons) of firearms examinations. Data gathered using the prevailing AFTE Range were also recategorized into two hypothetical scoring systems. Consistently positive differences between observed agreement and expected agreement indicate that the repeatability and reproducibility of examiners exceed chance agreement. When averaged over bullets and cartridge cases, the repeatability of comparison decisions (involving all five levels of the AFTE Range) was 78.3% for known matches and 64.5% for known nonmatches. Similarly averaged reproducibility was 67.3%% for known matches and 36.5% for known nonmatches. For both repeatability and reproducibility, many of the observed disagreements were between a definitive and inconclusive category. Examiner decisions are reliable and trustworthy in the sense that identifications are unlikely when examiners are comparing non-matching items, and eliminations are unlikely when they are comparing matching items.     

Organizational and Human Factors Affecting Forensic Decision-Making: Workplace Stress and Feedback

Although forensic examiners operate in a stressful environment, there is a lack of understanding about workplace stress and feedback. These organizational and human factors can potentially impact forensic science judgments. In this study, 150 practicing forensic examiners from one laboratory were surveyed about their experiences of workplace stress, and the explicit and implicit feedback they receive. Forensic examiners reported that their high stress levels originated more from workplace-related factors (management and/or supervision, backlogs, and the pressure to do many cases) than from personal related factors (family, medical, and/or financial). The findings showed that a few (8%) of the forensic examiners sometimes felt strong implicit feedback about what conclusions were expected from them and that some (14%) also strongly felt that they were more appreciated when they helped to solve a case (e.g., by reaching a “match” as opposed to an “inconclusive” conclusion). Differences were found when comparing workplace stress and feedback levels across three core forensic science fields (forensic biology, chemistry, and latent prints) and across career stages (early, mid, and late). Gaining insights into the stress factors within a workplace and explicit and implicit feedback has implications for developing policies to improve the well-being, motivation, and performance of forensic examiners.     

Cold‐blooded lie catchers? An investigation of psychopathy,emotional processing,and deception detection

Purpose. The process of catching liars is challenging, though evidence suggests that deception detection abilities are influenced by the characteristics of the judge. This study examined individual differences in emotional processing and levels of psychopathic traits on the ability to judge the veracity of written narratives varying in emotional valence. Methods. Undergraduate participants (N= 251) judged the veracity of 12 written narratives (truthful/deceptive) across three emotional categories: positive, negative, and neutral events. Levels of psychopathy were assessed to investigate its relation to accuracy and cue use. Results. Overall accuracy was close to chance, although participants were more accurate in determining the veracity of truthful relative to deceptive narratives. Accuracy was impaired for emotional (positive and negative) relative to neutral narratives. Psychopathy was not associated with levels of overall accuracy, but related to discriminative ability, and differential use of cues in decision making. Reported cue use also differed across emotional narrative conditions. Conclusions. We speculated that an emotive truth bias may have detracted judges from attending to valid cues that are indicative of the deceptive nature of stimuli because they were distracted by the emotional content of the report. Implications for deception detection in forensic settings are discussed.     

Meta-analysis of mock crime studies of the control question polygraph technique

A review of results obtained from standard guilty and innocent treatment conditions in 14 mock crime studies of the control question polygraph technique revealed accuracies ranging from chance to 100% correct. The present study examined several factors that may have contributed to the observed variability in detection rates across studies. Those included sampling error, differences in the populations from which subjects were drawn (Subjects), differences in the nature of incentives provided to subjects for passing the polygraph test (Incentives), and differences in the methods for diagnosing truth or deception (Decision Policy). A meta-analysis revealed that approximately 24% of the variance in detection rates could be attributed to sampling error, and detection rates were correlated with types of Subjects (r=.61). Incentives (r=.73), and Decision Policies (r=.67). The highest diagnostic accuracies were obtained from nonstudent subject samples, when both guilty and innocent subjects were offered monetary incentives to convince the examiner of their innocence, and when conventional field methods were used for interpreting the physiological recordings and diagnosing truth and deception. Together, differences in Subjects, Incentives, and Decision Policies may account for as much as 65% of the observed variance in detection rates. The present findings highlight the importance of conducting mock crime experiments that closely approximate field conditions.     

Children's identification of unfamiliar voices on both target-present and target-absent lineups

ABSTRACT

A robust finding from the eyewitness literature is that children are as accurate as adults on target-present lineups from the age of five years, whereas they continue to make an erroneous false positive identification from a target-absent lineup up until around fourteen years of age (Pozzulo, J. D., & Lindsay, R. C. L. (1998). Identification accuracy of children versus adults: a meta-analysis. Law and Human Behavior, 22(5), 549–570). The current study explores whether the same pattern occurs when voices are used instead of faces. A total of 334 participants from six age groups (6–7-year-olds, 8–9-year-olds, 10–11-year-olds, 12–13-year-olds, 14–15-year-olds and adults) listened to a 30 second audio clip of an unfamiliar voice and were then presented with either a six person target-present or target-absent voice lineup. Overall, participants were more accurate with target-present than target-absent lineups. Moreover, performance on target-present lineups showed adult-like levels of attainment by 8–9 years of age. In contrast, performance on target-absent lineups was extremely poor, with all age groups tending to make a false identification. Confidence was higher when participants made correct rather than incorrect decisions for both types of lineup and this did not change with increasing age. Given these results, both child and adult earwitness evidence needs to be treated with considerable caution.     

Overhearing the Planning of A Crime: Do Adults Outperform Children As Earwitnesses?

This study examined the reliability of earwitnesses using an ecologically realistic experimental set-up. A total of 282 participants, distributed over three age-groups (7–9 vs. 11–13 year olds vs. adults), were exposed to an unfamiliar voice for 40 seconds. After a two week delay, they were presented with a 7-voice lineup. Half of the participants were exposed to a target-present lineup (TP), and the other half to a target-absent lineup (TA). For both types of lineups the participants performed poorly. In the TP-condition only the 11–13-year olds (with 27% correct identifications) performed above chance level. Furthermore, in the TA-condition all age-groups showed a high willingness to make an identification (overall mean = 53%). For both groups of children, voice identification co-varied significantly with speaking rate and pitch level, as did pitch variation for the youngest children. Neither factor correlated significantly with the adults’ identifications.     

Interviewing to Detect Deception: When to Disclose the Evidence?

Research shows that there are few objective cues to deception. However, it may be possible to create such cues by strategic interviewing techniques. Strategic Use of Evidence (SUE) is one such technique. The basic premise of the SUE technique is that liars and truth tellers employ different counter-interrogation strategies, and that the evidence against the suspect can be used to exploit these differences in strategies. This study examined the effect of the timing of evidence disclosure (early vs. late vs. gradual) on verbal cues to deception. We predicted that late disclosure would be most effective in differentiating between liars and truth-tellers, and that cues to deception in the gradual disclosure condition would progressively disappear due to the suspects’ realization that evidence against them exists. That is, we expected that liars in the gradual presentation condition would become more consistent with the evidence over time. A sample of 86 undergraduate students went through a mock-terrorism paradigm (half innocent, half guilty), and were subsequently interviewed using one of three disclosure strategies: early, gradual, and late disclosure. We measured statement-evidence inconsistencies as cues to deception . Results supported our predictions in that cues to deception were most pronounced in the late disclosure condition. Contrary to our expectations, the results suggested that presenting the evidence gradually may put innocent suspects at a higher risk of misclassification as they seem to adopt a strategy that is more similar to guilty suspects.     

Strategic Use of Evidence During Police Interviews: When Training to Detect Deception Works

Research on deception detection in legal contexts has neglected the question of how the use of evidence can affect deception detection accuracy. In this study, police trainees (N=82) either were or were not trained in strategically using the evidence when interviewing lying or truth telling mock suspects (N=82). The trainees’ strategies as well as liars’ and truth tellers’ counter-strategies were analyzed. Trained interviewers applied different strategies than did untrained. As a consequence of this, liars interviewed by trained interviewers were more inconsistent with the evidence compared to liars interviewed by untrained interviewers. Trained interviewers created and utilized the statement-evidence consistency cue, and obtained a considerably higher deception detection accuracy rate (85.4%) than untrained interviewers (56.1%).     

Using sketch drawing to induce inconsistency in liars

Purpose. Consistency as a cue to detecting deception was tested in two experiments using sketch drawing and verbal reports in repeated interviews. Liars were expected to be less consistent than truth‐tellers. Methods. In Expt 1, 80 undergraduate students reported truthfully or deceptively about an alleged lunch date – they sketched the layout of the restaurant and then answered spatial questions about objects in the restaurant. Ratings were given for the consistency between sketches and verbal reports. In Expt 2, 34 undergraduate students reported truthfully or deceptively about completing a series of unrelated tasks – they answered spatial questions about objects in a room and then sketched the layout of the room. Proportions were calculated for the consistency between verbal reports and sketches. Results. Expt 1. Liars were rated as less consistent than truth‐tellers. Up to 80% of truth‐tellers and 70% of liars could be correctly classified. Expt 2. Liars were less consistent than truth‐tellers on consistency proportions. Up to 100% of truth‐tellers and 77% of liars could be correctly classified. Conclusions. Using sketches to induce inconsistency may be a reliable, resource efficient way to help investigators detect deception.     

Lie Detection Accuracy—the Role of Age and the Use of Emotions as a Reliable Cue

Literature surrounding the accuracy of deception detection has produced inconsistent findings, and the majority of investigations have been based upon low-stakes lies. Although recent research has suggested that high-stakes situations may produce reliable cues to deception, it remains unclear whether knowledge of these cues actually improves the detection of lies. In an online experiment, we assessed participant’s ability to detect lies in 22 public appeals for help with missing or murdered relatives (N = 196). Participants were randomly allocated to either the cue condition (presented with previously identified cues to deception) or no cue condition (instructed to make judgement on instinct), before being presented with the video footage. Participants were asked to indicate whether the appealer is lying or telling the truth, how confident they are in their judgement and if they were familiar with the case. At the end of the experiment, participants wrote qualitative responses on the cues that they used during lie detection. Although cue knowledge and confidence did not significantly predict accuracy scores, there was a positive relationship between accuracy and age. Participants who used emotion-based cues were significantly better at detecting deception. The findings are discussed with reference to the existence of reliable cues.     

Helping to sort the liars from the truth‐tellers: The gradual revelation of information during investigative interviews

Research examining detection of verbal deception reveals that lay observers generally perform at chance. Yet, in the criminal justice system, laypersons that have not undergone specialist investigative training are frequently called upon to make veracity judgements (e.g., solicitors; magistrates; juries). We sought to improve performance by manipulating the timing of information revelation during investigative interviews. A total of 151 participants played an interactive computer game as either a truth‐teller or a deceiver, and were interviewed afterwards. Game information known to the interviewer was revealed either early, at the end of the interview, or gradually throughout. Subsequently, 30 laypersons individually viewed a random selection of interviews (five deceivers and five truth‐tellers from each condition), and made veracity and confidence judgements. Veracity judgements were most accurate in the gradual condition, p < .001, η² = .97 (above chance), and observers were more confident in those judgements, p < .001, η² = .99. Deceptive interviewees reported the gradual interviews to be the most cognitively demanding, p < .001; η² = .24. Our findings suggest that the detection of verbal deception by non‐expert observers can be enhanced by employing interview techniques that maximize deceivers' cognitive load, while allowing truth‐tellers the opportunity to respond to evidence incrementally.     

Human Factors Effecting Forensic Decision Making: Workplace Stress and Well‐being

Over the past decade, there has been a growing openness about the importance of human factors in forensic work. However, most of it focused on cognitive bias, and neglected issues of workplace wellness and stress. Forensic scientists work in a dynamic environment that includes common workplace pressures such as workload volume, tight deadlines, lack of advancement, number of working hours, low salary, technology distractions, and fluctuating priorities. However, in addition, forensic scientists also encounter a number of industry‐specific pressures, such as technique criticism, repeated exposure to crime scenes or horrific case details, access to funding, working in an adversarial legal system, and zero tolerance for “errors”. Thus, stress is an important human factor to mitigate for overall error management, productivity and decision quality (not to mention the well‐being of the examiners themselves). Techniques such as mindfulness can become powerful tools to enhance work and decision quality.     

Clusters of Nonverbal Behaviors Differ According to Type of Question and Veracity in Investigative Interviews in a Mock Crime Context

Evaluating truthfulness and detecting deception is a capstone skill of criminal justice professionals, and researchers have long examined nonverbal cues to aid in such determinations. This paper examines the notion that testing clusters of nonverbal behaviors is a more fruitful way of making such determinations than single, specific behaviors. Participants from four ethnic groups participated in a mock crime and either told the truth or lied in an investigative interview. Fourteen nonverbal behaviors of the interviewees were coded from the interviews; differences in the behaviors were tested according to type of question and veracity condition. Different types of questions produced different nonverbal reactions. Clusters of nonverbal behaviors differentiated truth tellers from liars, and the specific clusters were moderated by question. Accuracy rates ranged from 62.6 to 72.5% and were above deception detection accuracy rates for humans and random data. These findings have implications for practitioners as well as future research and theory.     

The devil is in the detail: deception and consistency over repeated interviews

ABSTRACT

Research indicates that truthful statements typically contain more details than fabricated statements, and that truth tellers are no more consistent than liars over multiple interviews. In this experiment, we examine the impact of (i) multiple interviewers and (ii) reverse order recall on liars’ and truth tellers’ consistency and amount of reported detail over repeated recall attempts. Participants either took part in a mock crime (lying condition) or an innocent event (truth telling condition) which they were subsequently interviewed about in two separate interview phases. Truth tellers provided more details overall, and more reminiscent details than liars. There were no differences between veracity groups for the number of omissions made or repetitions reported. Despite the popular belief that inconsistency is a cue to deception, we found little support for the notion that consistency (or lack of consistency) offers a diagnostic cue to deception. We found little evidence that switching interviewer or recalling in reverse order induced inconsistencies in liars. In fact, due to the number of reminiscent details in truth tellers’ accounts, our findings suggest that accounts provided by liars tend to be slightly more consistent than those provided by truth tellers. Materials for this paper can be found at osf.io/hgvmk/.     

Behavioral Cues to Deception vs. Topic Incriminating Potential in Criminal Confessions

Coding statements of criminal suspects facilitated tests of four hypotheses about differences between behavioral cues to deception and the incriminating potential (IP) of the topic. Information from criminal investigations corroborated the veracity of 337 brief utterances from 28 videotaped confessions. A four-point rating of topic IP measured the degree of potential threat per utterance. Cues discriminating true vs. false comprised word/phrase repeats, speech disfluency spikes, nonverbal overdone, and protracted headshaking. Non-lexical sounds discriminated true vs. false inthe reverse direction. Cues that distinguished IP only comprised speech speed, gesticulation amount, nonverbal animation level, soft weak vocal and “I (or we) just” qualifier. Adding “I don't know” to an answer discriminated both IP and true vs. false. The results supported hypothesis about differentiating deception cues from incriminating potential cues in high-stakes interviews, and suggested that extensive research on distinctions between stress-related cues and cues to deception would improve deception detection.