Disclaimer: This dissertation has been written by a student and is not an example of our professional work, which you can see examples of here.

Any opinions, findings, conclusions, or recommendations expressed in this dissertation are those of the authors and do not necessarily reflect the views of UKDiss.com.

Sex Differences in Deception Detection Accuracy

10128 words (41 pages) Dissertation

9th Dec 2019 Dissertation Reference this

Tags: Medical

Introduction

Deception occurs on a daily basis, with people admitting to telling a lie in one out of every five of their social interactions (DePaulo, Kashy, Kirkendol, Wyer & Epstein, 1996). Despite the common occurrence of lies humans are relatively poor lie detectors (Bond & DePaulo, 2006; Geiselman, 2012). In everyday life, people tend to take statements at face value and assume others are mostly being truthful. They are less inclined to start questioning and assuming everyone is deceitful (Bond & DePaulo, 2006).  Law professionals, such as police officers, are required to make critical decisions about the veracity of the information they receive on a regular occurrence. However this is not an easy task for many people. Much of the general research into deception detection is contradictory and does not give a clear answer on how to distinguish a deceiver from a truth teller.

Often it is believed that there are clear stereotypical signs of a liar such as gaze aversion and increased fidgeting (Bogaard, Meijer, Vrij, & Merckelbach, 2016; Sporer, & Schwandt, 2007). However, the research has shown that in actual fact, these stereotypical signs of lying are not always reliable ways to spot a liar. Often people tend to believe and rely on non-verbal behaviours to make a decision on whether they believe someone to be lying or not (Vrij, 2008). Despite people still relying upon non-verbal behaviours to come to a verdict on deception, the research has shown that the non-verbal cues to deception are mostly unreliable (DePaulo, Lindsay, Malone, Muhlenbruck, Charlton & Cooper, 2003).  DePaulo et al., (2003) conducted a meta-analysis of 116 studies and in these studies they found a potential 158 verbal and nonverbal cues to deception. However, out of these 158 behaviours they found that there was no clear link or only a weak link to deception. Yet, the findings showed that there were some faint verbal cues which could be picked up on, such as more negative impressions, lairs tell less compelling stories, fewer imperfections and less unusual aspects in their stories. The method of splitting non-verbal behaviours and verbal behaviours has been criticised by Bull (2006) because he believes that this is out dated. Also that understanding the relationship between verbal and non-verbal behaviours is more useful when detecting deception than looking at them individually.

DePaulo et al.’s meta-analysis found 158 cues to deception which they state that they can confidently link some to deceit. However, the evidence shows that despite them feeling confident on some of the cues, in fact many of the cues were weakly linked to deception. Furthermore, many of the behaviours which are linked to deception are behaviours which can be attributed to other emotions and feelings. This idea is supported by Vrij and Granhag (2012), who have criticised the use of cues in deception detection due to their ambiguity. They argue that many of the common cues which are displayed by liars are also signs of other psychological states, such as anxiety or stress. It is likely that if a truthful person is under pressure in a high stakes situation they will show signs of stress, anxiety or nervousness which may be misconstrued as a deceptive behaviour (Vrij & Granhag, 2012).

When it comes to detecting lies, some evidence has shown that humans struggle to achieve high accuracy rates.  Research has found that participants on average tend to get accuracy rates that are only marginally higher than chance level (Bond & DePaulo, 2006; Aamodt & Custer, 2006).  With some studies finding accuracy rates as low as 45.3% (Peace, Porter & Almon, 2012), and this study will be discussed at a later point.  In most experiments, including the current one, the observers are required to classify senders as either lying or telling the truth. Their classification of the senders is then an indication of their overall accuracy at detecting deception. As their can only be two out comes (lie or truth) and there is often half truthful messages from senders and half deceitful messages then this means that if an observer was to purely guess all their answers they are likely to achieve 50% accuracy rates purely by chance anyway (Bond & DePaulo, 2006: Geiselman, 2012; Aamodt & Custer, 2006, Ekman & O’Sullivan 1991; Vrij, 2000). It may be inferred from the evidence that has found results of around this 50% mark that this is purely due to chance rather than the participants’ deception detection ability.

Bond and DePaulo (2006) carried out a meta-analysis and found that across 292 samples the average accuracy rate was 54% which is only slightly greater than the chance level. Despite this accuracy rate being a mere 4% above chance level it was significantly above chance, though this may have been due to a small effect size. In this meta- analysis the highest result found in any sample was 73% and the lowest accuracy was 31% (Bond & DePaulo, 2006). Similarly, an early meta-analysis by Kraut (1980) found accuracy rates of just 57%, later Krauts (1980) study was replicated by Vrij (2000) with newer and more studies which produced a  similar result of 56.6% average accuracy rate.

It has been suggested by others that the low accuracy rates which have previously been found (Bond & DePaulo, 2006; Krauts, 1980; Vrij, 2000, Peace, Porter & Almon, 2012) are due to using low stakes stimuli. Low stakes stimuli are the material which may be used in deception research which tends to lack any serious consequences for the sender involved in the stimuli. Furthermore it is unlikely that low stakes stimuli will involve any highly emotional content (Vrij, Fisher, Mann & Leal, 2006). Low stakes lies are often used in lab settings and are not a true representation of real deceitful situations and they tend to lack ecological validity. Based on the criticism of low stakes lies there has been research which has used high stakes stimuli and this has found higher accuracy rates (Wright-Whelan, Wagstaff &Wheatcroft, 2015; Vrij, Mann, Robbins & Robinson, 2006; Vrij & Mann, 2001). It is thought that when the stakes are high and the liar is under more pressure they are more motivated to have their lie believed by their observers (Wright-Whelan, Wagstaff &Wheatcroft, 2014). Whereas when the stakes are lower and stimuli is fabricated the liar has no pressure on them to lie so therefore may seem calmer or not show clear signs that they are being deceitful (Wright-Whelan et al., 2014). 

It can be cognitively demanding for someone trying to control and maintain their behaviour when they are telling a lie, especially if the stakes are high. It is often the case where cues are clearer to the observer because the liar may slip up with some aspect of their nonverbal behaviour or verbal cues (Carlucci, Compo, Zimmerman, 2013). Thus, it has been suggested that high stakes stimuli has the potential to increase the participants lie detection accuracy (Carlucci et al., 2013).  Furthermore, high accuracy rates have been found by Vrij et al. (2006) with an average of 72%, Vrij and Mann (2001) with an average of 64% and Wright-Whelan et al. (2015) with an average of 68%. All of these studies have used real life high stakes, emotional stimuli in their research and have produced results which are higher than the average accuracy rates previously found by those who have used low stakes stimuli (Kraut, 1980; Bond & DePaulo, 2006; Vrij, 2000; Aamodt & Custer, 2006). 

Despite the higher accuracy results found with the real life emotional stimuli, it has been found that using just highly emotional stimuli which is not high stakes only produced accuracy rates of 47.6% (Peace & Sinclair, 2012). They argue that the reason for this low result was due to people getting distracted by the emotional nature of the stimuli. They conclude that the emotional aspect can overpower people’s ability to detect deception even if the cues are clearer. Then again, this study used emotional content but it was not high stakes. The emotional statements were fabricated and therefore the sender did not feel there were any consequences if they were not believable.  However, if Peace and Sinclair’s (2012) conclusion of why they produced such low results in their study is correct then it could be assumed that other studies which use emotional stimuli should also produce low accuracy rates. This is not the case though, the studies which have used high stakes stimuli which are also emotional (Vrij et al., 2006; Vrij and Mann, 2001; Wright-Whelan et al., 2015). Therefore, it is likely that the reason that Peace and Sinclair’s (2012) study produced low accuracy rates due to them using fabricated emotional content and results tend to be lower in research using low stakes stimuli.

In the same way, it has been found that lies are easier to tell from truths if the sender is motivated to lie (Bond & DePaulo, 2006). A sender is someone who is telling the lie which the participants are trying to detect. A sender who is motivated is someone who has more of a motive to be believed. For example someone who is on a murder trial trying to convince others they are innocent. Although it may be assumed that all liars are motivated to lie, this may not be the case. If the stakes are low and there are little consequences to the persons lie then they may care less if they are found out or not. Thus are classed as ‘unmotivated’ senders.

Bond and DePaulo (2006) hypothesise that it is someone’s fear of being disbelieved that is what makes it easier to identify their deceit. This is in line with the research which suggests that when high stakes stimuli is used then the senders are easier to detect if being deceitful. There is evidence to suggest that lies were easier to tell from truths in ‘motivated’ senders compared with ‘unmotivated senders. However, the meta-analysis from Bond and DePaulo (2006) also found that people who were motivated to have their lie believed were seen as less truthful. It was concluded that the reason for this is because if someone has motivation to be believed this could possibly make them resemble a ‘stereotypical’ liar. The classification for ‘motivation’ in this study though was defined by senders being rewarded for success in their lie. So the lies used in this meta-analysis are fabricated and using lower stakes stimuli.

Ekman (1985) proposed a theory that liars who feel guiltier about their lies should show more cues of guilt. However, guilt cues have not been clearly determined but they may include signs of sadness, using a lower pitched voice with softer speech, and looking downwards. Yet, it has been noted that some liars may also feel emotions of excitement at the challenge of lying and getting away with their lies. This makes it more difficult to distinguish between lies and truths because the cues and emotions people display are so varied and subjective to the individual displaying them. An important conclusion from this theory is that the role of emotions in deception is a key factor which can be used to distinguishing between liars and truth tellers. Thus, it could be assumed that if someone could understand the emotions that a person is experiencing then this would make it more likely to be able to distinguish between liars and truth tellers.

Zuckerman, DePaulo, and Rosenthal (1981) developed the multi-factor model which looked at emotional reactions when telling lies. For example, they predicted that liars show signs of fear, guilt and even delight when deceiving others, this would show cues such as lack of eye contact and increased physiological arousal. The increased physiological arousal would include things such as increased blinking, pupil dilation, more speech errors and disturbances. Zuckerman et al. (1981) also stated that the reason for a liar to be unable to control all aspects of their behaviour is because telling a lie is more cognitively demanding than telling the truth. Therefore they are less likely to monitor their behaviour because they are concentrating on the lie they are telling, this can lead to speech errors, lack of illustration in the body or facial expression leakage (Ekman, 2003). Emotional leakage often occurs when someone is displaying a fake emotional expression. It is when a very brief expression is shown on the face which shows the true feelings of the sender and is inconsistent with the statement which is being spoken (Ekman, 2003).

The idea of emotions being important in deception has been touched upon by the theories by Ekman (1985) and Zuckerman et al. (1991), however these theories lack experimental validity and contradict each other in some aspects of the theory. Since these theories were proposed there has been studies conducted which look into the relationship between emotional intelligence and deception detection (Wojciechowski, Stolarski & Matthews, 2014; Porter, ten Brinke, Baker & Wallace, 2011; Ekman, 2003). Emotional intelligence has been defined as the ability to accurately express, appraise and perceive emotions, including being able to understand emotion and regulate their own emotions (Mayer & Salovey, 1997). Hence, high emotional intelligence it is associated with understanding facial expressions and emotions others are displaying. The influence of emotional intelligence on detecting deception will be discussed next.

It is thought that in order to deceive others, liars must be able to mask the emotion they actually feel and stimulate a different emotion which they wish to portray (Ekman & Friesman, 1971). If the verbal message from someone matches with the non-verbal signals being demonstrated by them (in particular, facial expressions) then it is likely they will seem more believable and be more successful in their deceit (Wojciechowski et al., 2014). However, when someone is displaying fake emotional facial expressions they are susceptible to emotional leakage (Porter et al., 2011). This is when a brief facial expression is shown on the liars face which is inconsistent with their statement, these tend to last less than a second (Ekman, 2003). This idea of emotional leakage has been linked to deception detection because it is thought that someone who is good at understanding other people’s emotions is likely to then be able to detect these facial expression inconsistencies and then be able to tell when someone is lying (Porter & ten Brinke, 2008).

Research has shown that people who tend to have a higher emotional intelligence are better at detecting facial inconsistencies (Wojciechowski et al., 2014). Wojciechowski and colleagues carried out research into emotional intelligence and processing facial emotions using a face decoding test (FDT). The FDT is a computer test which has been developed to measure individual effectiveness on reasoning based facial expressions. Participants are showed facial expressions, followed by a sentence. They are then asked to decide whether they think the person who showed the facial expression could have honestly said the sentence that followed.  The results showed that people who were scored higher on emotional intelligence also scored highly on the FDT. They concluded that the people with higher emotional intelligence were able to identify facial inconsistencies and this has been linked to better accuracy when detecting deception. However, the FDT is designed to test all facial expressions, not just faces when someone is lying and this study did not look at real examples of people lying. Instead, the study makes an conclusion on based on previous theories and ideas that the likelihood of people who are good at detecting facial inconsistencies are also good at detecting deception when this may not actually be the case. But this was not made clear in this study and they attribute their findings to also linking emotional intelligence with deception detection.

The findings from Wojciechowski et al. (2014) have given evidence that emotional intelligence does seem to impact on people’s ability to detect facial inconsistencies and potentially detecting deception. This is due to the link between facial inconsistencies and deception. However, contradicting research has suggested that having a higher emotional intelligence can make someone a poorer lie detector (Baker, ten Brinke & Porter, 2013). Baker et al. (2013) concluded that the reason individuals with higher emotional intelligence were poorer lie detectors was because they were sympathising with the emotional pleas from both liars and truth tellers. Someone who scored particularly high on emotionality may be in a more emotional state and this made them more gullible to deceptive pleaders. Furthermore Baker et al. (2013) suggest that emotionally intelligent people are less likely to be analytical of the pleader they are viewing because they are unable to suppress their emotion processing.

The reason for the difference between Baker et al.’s (2013) research and Wojciechowski et al.’s (2014) research may be the methodology used. Baker et al. (2013) used videos of real emotional pleas for family members, which included some senders who are being deceitful. Whereas Wojciechowski et al. (2014) used stimuli which were computer generated prototypes of certain facial expressions. So this type of stimuli is not highly emotive and thus very unlikely that any participants would become gullible and sympathetic towards the faces unlike what occurred in Baker et al.’s (2013) study. Thus it may be that in situations where the stimuli is real and relatable for the participants then this may actually lower the deception detection abilities if those who would usually be able to accurately detect emotions.

Emotional leakage is thought to be the reason how people with high emotional intelligence are able to spot a liar. If there is to be any emotional leakage when someone is being deceptive then it is more prominent if there is a highly intense emotion (Porter, ten Brinke & Wallace, 2012). However, despite this Porter et al. (2012) found that participants were unable to discriminate between sincere and insincere emotions, with this they only performed at chance level. Furthermore, the results concluded that emotional intelligence was no influence to whether someone was a better lie detector but instead that someone who has been trained to identify emotional cues will be a better lie detector (Porter et al., 2012; Shaw, Porter & ten Brinke, 2011).

The emotional intelligence and deception detection literature often overlaps with gender differences in detecting deception. This is because in many cases it has been found that on average females tend to have a higher emotional intelligence than males do (Cabello, Sorrel, Fernández-Pinto, Extremera, & Fernández-Berrocal, 2016; Wojciechowski et al., 2014; Goldenberg, Matheson & Mantler, 2006; Farrelly & Austin, 2007).  Women have been found to be more accurate than males when detecting deception due to their higher emotional intelligence (Wojciechowski et al., 2014; Hoffman, Kessler, Eppel, Rukavina & Traue, 2010) and this is thought to be because they are able to identify small changes in facial expressions. Females pay more attention to non-verbal behaviours and affective cues (Donges, Kersting & Suslow, 2012; Hurd and Noller, 1988) which enables them to pick up on any facial expression leakage which occurs when someone is telling a lie (Porter et al., 2011).

Furthermore, there has been research to suggest that females are superior to males when detecting deceit from their male partners (McCornack & Parks, 1990) and when carrying out ‘mind reading’ tasks when observing their partners (Thomas & Fletcher, 2003). However, when it comes to strangers’ deceit female superiority of detecting deception seems to disappear (DePaulo, Wetzel, Weylin Sternglanz & Wilson, 2003). This is contradictory to what was found by Wojciechowski et al. (2014) and Hoffman et al. (2010) because these both found that females were better at detecting deception than males and both studies used strangers, not their partners.

Moreover, females may be limited when it comes to strangers (DePaulo et al., 2003b) but there was no gender differences between male and female when the expressions were highly expressive (Hoffman et al., 2010).  The evidence showed that females performed better than males when shown subtle emotional expressions not highly expressive. Although, when it comes to deception detection it is more likely that people will have to detect subtle emotional expressions, due to emotional leakage (Ekman, 2003) rather than detecting highly expressive facial emotions. Therefore the results from Hoffman et al.’s study (2010) would support the idea that females are better at detecting emotions and therefore detecting deception. This is especially because the facial emotions they are noticing are subtle ones.

Contrary to the research presented so far, there has been evidence to suggest that males can achieve higher accuracy rates of detecting deception than females (Mann, Vrij & Bull, 2004). However the methodology of this study may be what produced these significant results. In the Mann et al. (2004) study there were only 24 females compared with 75 males and the senders used for the stimuli included only 2 females compared with 12 males. This uneven number of male and female participants and senders may have affected the results which made the evidence seem as though males are much better at detecting deception than females. There is not much evidence to show that males are better at detecting lies than females. Although the evidence in the Mann et al. (2004) study found that there males are better at detecting deception, it may not be an accurate representation of females because there was so few used in the study.

Therefore the evidence for females being slightly better at detecting deception than males are seems to be stronger than the evidence to suggest that males are better than females at deception detection.  Further evidence for females comes from Barnacz, Amati, Fenton, Johnson and Keenan (2009); females were significantly above chance at detecting female senders but were significantly below chance when it was detecting a male sender. This finding is contrasting to what has previously been suggested by Hurd and Noller (1988) and Thomas and Fletcher (2003) which has found that females are better at detecting their male partners. The difference in these results though may be due to the fact that females are not as good at detecting males if they are not their partners.

It is often reported that people tend to be over confident in their ability to detect lies (Elaad, 2003; Meissner & Kassin, 2002); but the literature has found that higher confidence is not linked with improved accuracy (DePaulo & Pfeifer, 1986; Aamodt & Custer, 2006). DePaulo and Pfeifer (1986) carried out research into law enforcement officials and students and they found that the law officials believed that they were better than the students and had more confidence in themselves. However this did not improve their accuracy as they were no more accurate than the students. Contrasting to what Elaad (2003) and Meissner and Kassin (2002) reported, Mann et al. (2004) found that people tended to rate their confidence as low as 50%. Interestingly though Mann et al., found accuracy results which were significantly lower than their confidence levels. This suggests that the people in this study had little confidence in themselves, but it seems that they were justified in doing so seen as their accuracy rated fell below 50%.

Furthermore, evidence has shown that confidence has been positively correlated with emotional intelligence scores on the TEIQue-SF (Baker et al. 2013). This shows that those who have higher emotional intelligence tend to be more confident in themselves when detecting deceit. This however does not show that when they are confident that they are correct. Therefore it would be more useful to have evidence which would distinguish between confidence when participants are correct and when they are incorrect. Confidence can also be a more sensitive measure of accuracy, with confidence levels and with emotional intelligence. It is useful to understand whether a participants accuracy scores overall can predict their confidence when they are incorrect and correct. Therefore, the current study will carry investigate this to understand whether accuracy can predict confidence levels. Also whether there is a relationship between emotional intelligence with confidence levels with participants are correct and incorrect.

Currently the literature is contrasting and not come to a definitive answer regarding deception detection. This includes, whether emotional intelligence does have an impact on deception detection ability, and whether this helps or hinders. This also includes gender differences in detecting deception, currently there a body of literature which would suggest that females are better at detecting deception than males (Thomas & Fletcher, 2003; Wojciechowski et al., 2014; Hoffman et al., 2010; McCornack & Parks, 1990; Donges et al., 2012; Hurd and Noller, 1988). However, there is research to suggest that this is limited to just their partners and in low stakes conditions, therefore it would be useful to explore this further. Additionally, the influence of confidence does not seem to be explored much with its relationship with emotional intelligence, this is a gap in the research which would be interesting to explore further.

This study aims to contribute to the literature by investigating whether there is an interaction between senders, observers, and emotional intelligence. It will investigate this by calculating emotional intelligence scores by using an emotional intelligence questionnaire. Participants’ accuracy of deception detection will then be calculated for both male senders and female senders by watching real life high stakes videos of missing person pleas which were broadcast on the news. Furthermore, it will investigate whether there is a correlation between emotional intelligence scores and confidence scores. This will be explored by running a simple linear regression for the emotional intelligence scores with both confidence scores when participant was correct and confidence scores when participant was incorrect. The following predictions were made:

 Hypothesis 1) Based on the previous findings (Cabello et al., 2016; Wojciechowski et al., 2014; Goldenberg et al., 2006; Farrelly & Austin, 2007) it would be hypothesised that females will have a higher emotional intelligence than males.

 Hypothesis 2) Based on the assumption of hypothesis 1 and the research which suggests females have higher accuracy rates due to their higher emotional intelligence (Wojciechowski et al., 2014; Hoffman et al., 2010), it would be predicted that females will produce higher lie detection accuracy rates than males.

 Hypothesis 3) It will be hypothesised that there will be a relationship between emotional intelligence and confidence scores, both when participant is correct and when they are incorrect. A higher emotional intelligence score will predict a higher confidence score when correct.  A low emotional intelligence score will predict a low confidence score when incorrect.

Method

Participants

The initial sample in this research consisted of 81 participants (including 44 females and 37males) and they ranged in age from 18 years to 77 years. Participants were recruited voluntarily using a link posted on several social media sites, including Facebook, Twitter, LinkedIn and Instagram. To each of these sites a small amount of information about the survey provided with the link to the survey to which they could click on to find out further information and decide if they wanted to take part. Participants will also be referred to as ‘observers’ throughout the results and discussion sections of this report.

Materials

This study was hosted on Survey Gizmo and this is where the study was designed and where the participants took part in study. Participants were required to complete The Trait Emotional Intelligence Questionnaire – Short form (TEIQue-SF) (Petrides, 2009) to measure emotional intelligence. It is a self-report measure consisting of 30 items which have a Likert style response which ranges from 1, completely disagree, to 7, completely agree. The TIEQue-SF is a reliable measure of trait emotional intelligence (Petrides & Furnham, 2003). The TEIQue-SF comes from the full 150 item TEIQue and it measures four factors. These are well-being, self-control, emotionality and sociability. Cronbach’s alphas for the four factors are 0.79, 0.72, 0.68 and 0.59, respectively. Of the four factors, emotional intelligence was of most interest to this study and this factor measured empathy, emotional perception, and expression. Sociability factor measures social awareness and emotional management.  Self-control measures impulse control and emotion regulation. Well-being factor measures self-esteem, optimism and happiness.

This study consisted of 16 short video clips, each lasting under two minutes. The video clips were British or American news interviews and police interviews with the relatives of missing or murdered people, each of the clips are available to the general public. Each clip has one main person speaking (this person will be referred to as the ‘sender’ throughout the method, results and discussion sections of this report) and this person is either being interviewed or appealing for their missing or murdered relative. In eight of the 16 videos the sender in the video is being truthful, and in the other eight videos the sender is telling a high stakes lie; in this study if someone is classed as lying then they was ultimately convicted of having full or some involvement of the murder of the person which they are appealing for. By being defined as truthful this means the sender in the video was not convicted of any crimes relating to the murder or kidnapping of the individual. Eight of the clips have male senders and eight of the clips have female senders. Five females are telling the truth with three females lying. Five males are telling lies and three males are telling the truth.

The participants were recruited using Facebook, Twitter, Instagram and LinkedIn using a scripted post with the URL to Survey Gizmo (Appendix A). Upon data collection the results were analysed using IMB SPSS Statistics 23.

Procedure

Following ethical approval from the University of Central Lancashire participants were recruited voluntarily using the social media sites Facebook, Twitter, LinkedIn and Instagram. The participants were provided with some information about the study and the link which directed them to Survey Gizmo where the survey was hosted.  Once the participants have clicked on the link they are directed to the information sheet and consent form (Appendix B), after reading this and giving consent participants were asked two demographic questions. They were asked to provide their age and gender, following this they were required to complete the TEIQ-SF. Next the participants were shown 16 video clips individually and after each clip they were asked three follow up questions; ‘Do you think the person in this video is lying or telling the truth?’ and ‘What cues from the person in the video made you come to this decision? List up to three cues or answer with ‘Don’t know’ if you just have a feeling about it.’ Another follow up question was ‘How confident are you about this decision?’, this was on a scale from 1 ‘Not confident at all’ to 6 ‘Absolutely sure’. When the survey had been completed the participants were shown a debrief form (Appendix C) where they were provided with contacts if they felt negatively affected by the study.

Results

Design and Analysis

An independent sample t-test was used to investigate the difference between the emotional intelligence scores for female observers and for male observers. A 2x2x2 mixed design was employed to investigate whether there was an interaction between emotional intelligence, gender of observer and gender of sender. The between participants measures were emotional intelligence (with two factors, high or low) and gender of observer (male or female). The within participants measures was the gender of sender (male and female).  A simple linear regression was run to investigate whether emotional intelligence scores could predict the average accuracy scores of the participants. A second simple linear regression was run to investigate whether average emotional intelligence scores could predict the average confidence scores when the participant was correct. A third simple linear regression was also run to investigate whether average emotional intelligence scores could predict the average confidence scores when the participant was incorrect. In each of the three regressions emotional intelligence scores were the dependent variables and accuracy, confidence when correct and confidence when incorrect were the predictors.

Preparation of data

Out of the 81 participants there was no missing data for any of the questions across the 16 the videos shown. First, for each participant their average scores were calculated; for their overall accuracy, accuracy on videos where senders were truthful and for videos where the sender was lying. The overall average for all participants across both genders of senders, was 73.62%, this was calculated using the raw correct scores out of the 16 videos and changing it into a percentage. Two male participants got a score of 100% (identified 16 videos correctly out of 16); the lowest score was a percentage of 31.25% (identified 5 videos correctly out of 16).

A more accurate way of assessing the responses is to use signal detection theory, this is to identify the number of hits and misses on the signal (a lie). A hit is defined as the proportion of correct classifications of a lie. D’ is a measure of lie detection ability, which measures the ability to distinguish between a lie and a truth, and the bigger the d’ value, the more able someone is to distinguish a lie from a truth. The raw accuracy scores for each participant were then converted into d prime values to give an accurate value of hits and misses. The d prime values were also calculated for when there was a male sender and a value was calculated for when there was a female sender for each participant. Then, for each participant their average confidence scores were calculated for the videos where they were correct and calculated for when the participant was incorrect.

The emotional intelligence scores were also calculated for each participant. For the SPSS data file the emotional intelligence scores were coded either low or high; any score below 5.1 was coded as ‘low’ and any score which was 5.1 and higher was coded as ‘high’. The emotional intelligence scores were split by using an almost equal number of participants who scored below 5.1 to number of participants who scored above 5.1.The means and standard deviations for accuracy scores are displayed in Table 1 and Table 2. The emotional intelligence scores were only split for the ANOVA analysis. For the three regression tests and the independent samples t-test the raw emotional intelligence scores for each participant were used.

Table 1: Means (standard deviations) for accuracy for females and males with low and high emotional intelligence, with accuracy when there was a female sender using the d prime value

Emotional intelligence

Male observers

Female observers

Total

       

Low emotional intelligence

0.19(0.75)

0.11(0.64)

0.14(0.68)

High emotional intelligence

0.25(0.85)

-0.10(0.72)

0.06(0.80)

Total

0.22(0.79)

0.01(0.68)

0.10(0.74)

Table 2: Means (standard deviations) for accuracy for females and males with low and high emotional intelligence, with accuracy when there was a male sender using the d prime values

Emotional intelligence

Male observers

Female observers

Total

       

Low emotional intelligence

0.54(0.81)

0.60(0.60)

0.57(0.68)

High emotional intelligence

0.16(0.76)

0.31(0.93)

0.24(0.85)

Total

0.35(0.80)

0.46(0.77)

0.41(0.78)

An independent samples t-test was conducted to compare emotional intelligence scores for females and males. There was not a significant difference in the scores for females emotional intelligence scores (M=4.98, SD=0.66) and males emotional intelligence scores (M=5.10, SD=0.76) (t(79)=,p=0.55). The output for this analysis can be seen in appendix D.

ANOVA

The 2x2x2 mixed ANOVA for gender of observer, emotional intelligence and gender of sender. In this test the gender of observer and emotional intelligence were between participants measures. Gender of sender was a within participants measure. The dependant variable of this ANOVA was the d’ accuracy scores. The ANOVA found that there was a main effect for gender of sender (F(1,77)=7.77, p=0.007), with male senders having average higher scores than female senders. There was no main effects forgender of observer (F(1,77)=0.19, p=0.664) or emotional intelligence (F(1,77)=2.19, p=0.14). There was no interaction for gender of sender with gender of observer (F(1,77)=2.35, p=0.13) or with gender of sender with emotional intelligence (F(1,77)=1.52, p=0.22). Additionally there was no interaction between gender of observer with emotional intelligence (F(1,77)=0.12, p=0.73). Finally, there was no interaction between gender of sender, gender of observer and emotional intelligence (F(1,77)=0.79, p=0.37). The output for the ANOVA analysis can be seen in appendix E.

REGRESSION

A simple linear regression was run to predict accuracy scores based on emotional intelligence. The regression was not significant (F(1,79)=0.761, p=0.386) with an R2 of 0.10. The participant’s predicted accuracy score is equal to 5.035 + -0.054 when the emotional intelligence scores are measured. Participants’ accuracy scores changed by -0.054 for each emotional intelligence score (SPSS output for this analysis can be seen in appendix F).

A simple linear regression was calculated to predict confidence when correct based on emotional intelligence. The regression was not significant (F(1,79)=0.046, p=0.83) with an R2 of 0.001. The participants predicted confidence scores when correct is equal to 3.39 + -0.025 when the emotional intelligence scores are measured. Participants’ confidence rating changed by -0.025 for each emotional intelligence score (SPSS output for this analysis can be seen in appendix G). 

A simple linear regression was calculated to predict confidence when incorrect based on emotional intelligence. The regression was not significant (F(1,79)=0.162, p=0.689) with an R2 of 0.002. The participants predicted confidence scores when incorrect is equal to 3.028 + – 0.053 when the emotional intelligence scores are measured. Participants’ confidence rating changed by -0.053 for each emotional intelligence score (SPSS output for this analysis can be seen in appendix H).

Discussion

The aims of this study were to investigate whether there were sex differences in deception detection accuracy, based on both the gender of senders and the gender of observers. The study also aimed to investigate whether emotional intelligence was an influencing factor on participants’ deception detection ability. The results of the ANOVA showed that there was a significant main effect for gender of sender. This showed that males achieved a higher d prime score than females. However there were no other significant main effects found and no significant interactions between any of the three variables of the ANOVA. Additionally, there was no significant difference found between males and females accuracy and emotional intelligence raw scores on a t-test. The results of the regression were not significant for emotional intelligence with confidence scores when correct or when the regression was ran with emotion intelligence and confidence scores when incorrect. Neither was there a significant result for accuracy and emotional intelligence. Based on the results of this study none of the three hypotheses can be accepted. The overall means found for accuracy over the 16 videos for all 81 participants produced percentage of 73.62%.

The current study found that when there was a female sender in the videos the d’ prime score was lower (0.10) than it was when there was a male sender (0.41). This would indicate that it was harder to distinguish between female liars and truth tellers than male liars and truth tellers. The d’ value accounts for bias in the accuracy scores. So the low d’ score for females indicates that participants rated female senders as lying when they were telling the truth and lying more often than with male senders. These results would indicate that either the participants find it more difficult to distinguish between females who are lying and who are telling the truth, and people chose to class them as lying due to the nature of the study. Or the results could indicate that females give the impression they are lying even when they are not.

The results indicate that it is easier to distinguish between a male liar and truth teller. Therefore, this may mean that males show clearer cues when they are lying which enables people to clearly distinguish between when someone is lying and when someone is telling the truth. Whereas with females, these cues may not be as clear and therefore participants find it difficult to tell when someone is lying because their behaviour is not much different to the behaviour of a female truth teller. Thus is could be argued that because it is more difficult to distinguish between female liars and truth tellers then it may be more likely that females are harder to detect when lying. A significant result of gender of sender was not hypothesised in this research but it has shown an important factor that there is differences between the way males and females lie. Or it shows that participants perceive male and female senders differently and this impacts on their decision.

It was hypothesised that females would have a higher emotional intelligence than males. This hypothesis cannot be accepted because the results of the ANOVA showed no significant main effect of emotional intelligence. Furthermore there were no significant interactions with emotional intelligence and the regressions ran also produced insignificant results. These results of this study show evidence that males and females score similar on the emotional intelligence test.

The hypothesis of this study was based upon previous research into emotional intelligence which has suggested that females do on average score higher than males do (Cabello et al., 2016; Wojciechowski et al., 2014; Goldenberg et al., 2006; Farrelly & Austin, 2007). The emotional intelligence questionnaire used has yielded significant gender differences in a similar study which used it (Baker et al., 2013). One reason for the difference in results may be that Baker et al. (2013) used a sample consisting of 83 females with only 32 males, whereas, in the current study the ratio of females to males was a lot smaller than this. It is possible that the uneven number of males and females produced a significant result for females having a higher emotional intelligence. When there is more even numbers of males to females it may be that this produces results which are not biased by a larger proportion of females.

An explanation for the lack of significant emotional intelligence finding may be due to the questionnaire used (TEIQue-SF). The TEIQue-SF is based on trait emotional intelligence theory, which uses the idea of emotional intelligence as a personality trait (Petrides, Pita, & Kokkinaki, 2007). This scale used has four factors which are measured; wellbeing, self-control, emotionality, sociability. Although these factors reliably measure emotional intelligence as a whole trait it may be that a limitation that it does not specifically focus on factors of emotional intelligence which have been reported to directly impact deception detection abilities. These abilities include reading facial expressions, understanding non-verbal behaviours (Ekman, 1985; Zuckerman et al., 1981; Ekman, 2003). The information highlighted about the TEIQue-SF is a limitation of this study because it may be that this particular emotional intelligence score is not measuring the factors of emotional intelligence which are most relevant to deception detection. A way to overcome this limitation would be to develop an emotional intelligence questionnaire which is made up of mostly the factors and traits which relate to deception detection abilities; such as emotional perception, empathy, reading expression.

The TEIQue-SF has been found to be reliable for males and females on all four of the factors it consists of.  However, in this current study there was no difference between males and females found like what has been found in other evidence (Baker et al., 2013). Although it is worth noting that Baker et al. (2013) only reported evidence of there being gender differences in the emotionality factor of the questionnaire. The current study however did not measure the differences between individual factors. It may be the case that the females may score higher on the individual factors (such as emotionality) but not on other factors of the emotional intelligence questionnaire. Thus, statistics from each factor may have highlighted some significant differences between emotional intelligence and enabled a further understanding of the factors which may or may not be different in males and females.

The second hypothesis which was made was developed based on the first hypothesis. This stated that females would be better at detecting deception than males, due to their higher emotional intelligence. However the results show that neither of the hypotheses was could be accepted, therefore it can be assumed from this evidence that emotional intelligence is not related to accuracy in any way. The results would suggest that males and females are equally as accurate on average at detecting deception and that emotional intelligence does not play a role in influencing deception detection ability. Research has previously shown that emotional intelligence does play a role in deception detection ability, whether this is by it negatively impacting (Porter et al., 2012) or positively impacting (Wojchiechowski et al., 2014) ability. Yet, the findings would suggest that emotional intelligence cannot explain why there was no difference between males and females accuracy rates.

As the results seem to show that emotional intelligence does not influence deception detection then it may be that there are some other factors which can help or hinder people with their deception detection abilities. Participants overall scored relatively high on this study and two participants achieved 100% accuracy rate, which would suggest that these individuals have a great ability to detect deception. Previously it has been found that there are ’wizards’ who are superior at detecting deception (O’Sullivan & Ekman, 2004). These people were classed as ‘wizards’ because they performed much higher than others on deception detection tasks. In a sample of 1200 people only 29 of these were classed as a deception ‘wizard’. The characteristics which made them a genius at detecting deception were that they were able to focus intensely on their task and they were extremely devoted to performing well (O’Sullivan & Ekman, 2004).  It may be that in the sample in this current study there were people who are ‘wizards’ of deception and have an innate ability to detect deception along with having the characteristics found by O’Sullivan & Ekman (2004).

Then again, this theory of deception ‘wizards’ has been criticised by Bond and Uysal (2007) because the statistical power in O’Sullivan and Ekmans’s (2004) study was not sufficient and did not seem to be reliable enough. Even if this is the case the idea of a small minority of people being very good at detecting deception due to personal characteristics is one that may be an explanation. It may be that personality traits such as emotional intelligence do not play a role in deception detection ability.  Instead it may be that it is a rare trait that few people possess which means they are able focus and have a further understanding of whether someone is lying or not.

  However there is a chance that these two participants who achieved 100% just happened achieve that high by a combination of genuinely detecting deception and lucky guesses on the ones they did not know. Based on this there is a limitation to this study is that these participants who achieved 100% are not able to be followed up due to anonymity. It would have been interesting to investigate further whether these individuals can achieve full accuracy or very high accuracy scores again. Also it would be interesting to understand whether this was a fluke or whether they actually have some special ability which enables them to tell when someone is lying to them. If it is an ability then understanding what it is that enables them to understand deception more than your average human. These questions which need answering are aims for future research in this area.

Furthermore it was hypothesised that emotional intelligence would be a predictor for confidence levels for when participants were both correct and incorrect. Again this hypothesis could not be accepted as no significant result was found in either of the two confidence regressions. This hypothesis was developed because it was thought that someone with a high emotional intelligence would have a higher accuracy rate of detecting deception and this would mean that when these people were right they would have higher confidence levels but when they were wrong they would have low confidence levels. Nor was a significant result found for emotional intelligence scores with accuracy. This is further support for the evidence that emotional intelligence does not play a role in any aspect of deception detection ability as no relationship has been found throughout the results of this study. Confidence in this study was used as a more sensitive measure for accuracy. It may be that the reason there was a lack of significant results with both confidence and accuracy regressions may be because there is no evidence to suggest that emotion intelligence is related to deception detection.

There was no significant interaction identified between gender of observer and gender of sender. These findings are in line with some of the literature which has found no gender differences in deception detection (Aamodt &Custer, 2006). The results would show that neither males nor females are advanced in detecting either specifically males or females. Their ability is equal with both genders; this is the case for when participants achieved low and high accuracy. Since there was no significant finding for gender of observer then it is likely that because neither gender is better at detecting deception overall then neither are better at specifically detecting males or females either.

There have been some limitations to this study stated previously but another issue which may have affected this research are methodological issues. This includes the use of an online questionnaire to gather data. As the research was not carried out under the supervision of researchers or in controlled settings it means that it cannot be known if all of the answers were honest and answered by just one person. Also the research was an online survey which was carried out by the participant without any time constraints. Therefore there is a chance that the participant was able to research any of the videos to find out answers or able to ask others in order to achieve a higher accuracy rate. Furthermore, there was no limit on the number of times participants could watch a video which could mean that the participants could watch the video over and over to come to a decision rather than an instant decision. In most real life settings a person is likely to have to have an instant feeling about someone, they cannot re-watch their behaviour. This therefore, may have led to an increase in accuracy in this current study. 

In this study an online survey was chose instead of an experiment where the researcher was present; this was a compromise which was made in order to increase the number of participants which took part. With an online survey it is possible to get a greater number of participants. This is important because this will increase the amount of data in the sample and increase the strength of the results. To gain a large number of participants in an experiment where the researcher is present with each participant would take considerable amounts of time which were not able to deal with in the time constraints of this current study. However, it would be beneficial for future research to use an experiment which the researcher is present and but is still able to obtain a large number of participants so these limitations can be overcome.

The videos used in this study were publically available and possibly shown on news programs. Therefore there is a chance that participants may have seen some of the videos before and know whether the sender in the video was lying or telling the truth. This would mean that if someone did already know they were not making a decision based on what they saw at the time but they are answering the questions based on a fact they already know. This is a limitation to this study because it means that the accuracy scores may not be completely based on participants’ initial judgments or their deception detection ability. A way to overcome this limitation would have been to add another question after each video to ask if participants already knew the outcome of the video. Any sets of data where participants already knew the outcome could then be discarded as it is not necessary to understanding participants’ deception detection ability.

The only significant finding from this research is that lies from males are easier to detect than females. This can have future implications for the research field; the evidence presented tends to suggest that females are easier to detect when lying (DePaulo, 1992; Millar & Millar, 1997; Zuckerman et al., 1981) Therefore, it would be useful to research this further, and to try and understand what it is that makes it harder for participants to distinguish between a female liar and a female truth teller. Significant findings from such research may be an addition to the field of deception detection because it may help understand what signs of deceit from males are that are different to those of females.

Finally, the overall accuracy descriptive statistic for the overall accuracy rate found a percentage accuracy rate which was high relative to some of the other research in this area. Researchers have found results as low as 54% (Bond & DePaulo, 2006) and 47.6% (Peace & Sinclair, 2012). Yet this current study found an overall accuracy rate of 73.62%; higher results like such have also been found in research which uses high stakes stimuli like the current study did (Vrij et al., 2006; Wright-Whelan et al., 2015; Vrij & Mann, 2001). The reason for this high accuracy rate may be due to the use of high stakes stimuli which may have made cues easier to detect in the senders. Thus, an implication from this evidence may be that under high stakes people are easier to detect and observers are better lie detectors. However in this study it was not analysed whether this result was significantly greater than the chance level of 50%. Furthermore, although 73% is a relatively high accuracy score it could be argued that in forensic, police or legal settings that this may not be a high enough accuracy rate when the stakes are so high.

A score of 73.2% is high and would show that on average participants were getting 12 out of the 16 videos correct. Although other previous studies have attributed this to the use of high stakes stimuli, there may be other reasons for the high accuracy rate. It may be that the task was too easy and people were able to get more questions right than if they would if the task had been more difficult. Alternatively , participants could achieve 50% just if they were to guess as the answers could only be either ‘lie’ or ‘telling the truth’, so it is a possibility that people were able to achieve high rates by sheer luck and educated guesses. This may potentially be a factor in why no effects for emotional intelligence were found. If people are guessing rather than demonstrating their actual ability to detect deceit then this has the power to shift the results so if there was any possibility of emotional intelligence impacting deception detection then this was not shown in the evidence.

The implications of this study will add a further insight to the vast body of literature currently into deception detection. This evidence from this study implies that sex of the observer has no influence over accuracy of deception detection. Furthermore, this study implied that emotional intelligence has no influence over accuracy of deception detection, adding to the contrasting findings already in this area of research. Thus the implications of this study are that they are supportive of the idea that there is no clear answer, trait or characteristic which can make some better at detecting deception.. Neither emotional intelligence nor being a particular gender has the power to make someone superior to others with detecting deceit. 

Conclusion

The evidence from this study has shown that there is a significant difference between male and female senders. The difference is that it is easier to distinguish a male liar from a male truth teller, whereas it is not as easy to distinguish this difference in female senders. From the other evidence provided in this research it can be conclude that emotional intelligence does not seem to impact on accuracy of detection deception which was originally hypothesised. Although none of the hypotheses could be accepted the evidence from this study has pointed in a different direction. This is that there must be some other underlying trait which makes people good at deception detection because emotional intelligence and sex of the observer do not have a relationship with detection accuracy.  Therefore the implications of this study would suggest that other factors must be focused on in future in order to understand further deception detection. Any future research which has been proposed is linked to understanding further the significant finding of this research and expanding upon this research further.

Cite This Work

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.

Related Services

View all

DMCA / Removal Request

If you are the original writer of this dissertation and no longer wish to have your work published on the UKDiss.com website then please: