Flashcards in Polygraph Week Deck (72):
A successful or unsuccessful deliberate attempt, without forewarning, to create in another a belief which the communicator considers to be untrue (Vrij, 2008)
Main Types of Lies (Vrij, 2008)
Lying to partner about where you are when you are having an affair
Exaggerating a point such as how much experience you have at a job interview
‘Forgetting’ to mention something such as alcohol at customs
Use of lies (general statement)
The ability to deceive is useful in everyday social situations.
Diary study (DePaulo et al., 1996)
Participants reported all lies from interactions 10 minutes or more in length
Diary study (DePaulo et al., 1996)
Students told 2 lies per day on average
Non-Students told 1 lie per day
34% of all interactions contained at least 1 lie
Diary study (DePaulo et al., 1996)
25% other oriented
67% outright lies
60% psychological reasons
40% materialistic reasons
Low stake lies (DePaulo et al. 1996) example
I told her she looked good when she didn’t
High stake lies (DePaulo et al. 1996) definition
Most often originate from behaviour which the target of the lie would perceive as bad
High stake lies (DePaulo et al., 1996) reasons
20% to avoid punishment or blame (highest stress)
32% Instrumental lies (material gain, personal pleasure) – (lowest stress).
29% Psychological (Protect self, hurt others)
10% Protect others
Kraut (1980) - detecting deception
Looked at 10 studies which had tested lie/truth detection accuracy of observers
57% lies and truths were correctly detected
Vrij (2008) - detecting deception
Compared 39 studies conducted after 1980
56.6% lies and truths correctly detected.
Bond & DePaulo (2006) - detecting deception
Compared accuracy using 206 studies between 1941 and 2005.
Overall accuracy = 53.98%
No difference between Experts and Non-Experts
Detection main approaches
Lying has been associated with increased Fear, Guilt & Excitement compared to truth telling (Ekman, 2001)
Assumption: increases in verbal and non-verbal behaviours which are linked to emotions provide cues to deception
E.g. lack of eye contact due to guilt.
Detection main approaches
Attempted behavioural control
Assumption 1: Liars are aware people may monitor their behaviour and so attempt to ‘act’ like a truth teller.
Assumption 2: Some cues (e.g. speech rate, facial expressions) are very hard to control consciously. Attempts to do so may give the liar away through ‘unnatural’ behaviour.
Detection main approaches
Assumption 1: In some situations telling a lie takes more mental effort than telling the truth.
Assumption 2: Some cues (e.g. speech rate, movements) may increase/decrease when a person experiences high cognitive load.
Result = mental effort may impair motor responses, leading to showing no physical movements during an interview.
(Non-verbal) Vocal cues
Examples of (Non-Verbal) Vocal cues: (Vrij, 2008)
Speech hesitations (umm)
Changes in voice pitch
Speech rate (words per min)
Frequency of pauses
Non-verbal and verbal cues
All three main approaches account for verbal and non-verbal cues to deception
Time between end of question and start of answer
How well do vocal (NV) cues fare?
People think more hesitations, speech errors, higher pitched voice and more pauses are indicators.
Actual indicators are higher pitched voice, longer latency period and longer pauses.
(Non-Verbal) Visual cues
Self-Adaptors (playing with hair)
Illustrators (arm movements to enhance speech)
Hand and Finger movements
How well do Visual (NV) cues fare?
People think poor gaze, self adaptors, body movements, blinking and position shifts suggest deception.
Actual indicators are less hand/finger, left/foot movements and less illustrators.
Self-references (me, I)
Immediacy (direct/clear not indirect/vague)
Consistency in answers
Contradictions in answers
How well do Verbal cues fare?
Pople think less self-references, less immediacy, fewer plausible answers, less consistency and more contradictions suggest deception.
Actual indicators are more negative statements, less immediacy, shorter responses and less plausible answers.
Non-verbal cue evaluation
Not much of what people believe about non-verbal deceptive behaviours matches research.
Difficult to spot in real-time
Slow/complex coding (reliability issues)
Relate more to cognitive effort than arousal
Need to be considered intra-personally
Verbal cue evaluation
Verbal cues offer slightly better diagnostic ability but:
Subject to individual differences
Need also to be considered Intra-personally
Not easily coded in real time
Coders can be subjective
What does a polygraph measure?
Rate of breathing
Changes in blood pressure
Electro dermal activity (palmar sweating)
Very accurate and sensitive to changes.
No set standard to which variables to measure.
Is the polygraph a lie detector?
It tells that there is a physiological reaction taking place but cannot tell what a person is thinking.
Concern based approach
Polygraph uses it.
Assumes liars will be more concerned than truth tellers so will show greater physiological changes
Barland (1988) - polygraph
Polygraph plays major role in police investigations.
Helps prosecutors decide whether to accuse and charge someone.
A number of books claim that polygraphs "helped" clear the innocent and ‘solved’ crimes.
BPS 2004 - slides
Procedures are not standardised to the extent in which they are satisfactory in psychometric terms.
No easy way to check the practices used by individual polygraphers.
Some aspects (e.g. misleading the examinee) might be contrary to current British Law.
Four types of polygraph test
Relevant/Irrelevant Test (Larson)
Comparison/Control Question Test (Raskin)
Directed Lie Test (Raskin)
Guilty Knowledge/Concealed Information Test (Lykken)
Relevant/Irrelevant Test method
Irrelevant: Do you live in the UK?
Relevant: On March 12, did you shoot John Smith?
Assumed that liars will show a larger physiological response to the relevant question.
It was the first most widely used method of testing.
Control Question/Directed Lie Test (CQT/DLT) Phases
Phase 1: Questions discussed with the examinee.
Phase 2: Demo the effectiveness of test
Phase 3: Proper test: 3 relevant and 3 control questions run 3 times.
Phase 4: Interpretation of results (global or numerical approach)
Control Question/Directed Lie Test (CQT/DLT) general idea
Compare responses to relevant questions to control questions.
Control questions designed to control for effect of threatening questions.
CQT/DLT expected results
Innocent people react to stronger comparison questions, and guilty examinees will react stronger to relevant questions.
Innocents fear control questions more because they arouse concern about subject's past truthfulness whereas relevant questions aren't applicable to them.
Global vs. Numerical Interpretation
Global: Looking at charts and inspecting whether a stronger response is seen for relevant questions.
Numerical: Rating responses to each physiological measure on a numerical scoring system.
Both subjective as numerical isn't standardised and global is open to interpretation.
Criticisms concerning CQT/DLT
Testing is not standardised as examiner has to formulate new control questions for each test (CQT).
Examiner misleads examinee to create arousal on control questions (CQT).
Scoring is not standardised so is subjective (Both).
No evidence supporting assumption that innocents should show a stronger response to control questions (National Research Council, 2003).
Patrick & Iacono, (1991)
Examined over 200 Polygraph field tests.
Used scoring sheets of primary examiner who observed test and independent examiners who used only polygraph output.
Original examiners: 90% accuracy.
Independent examiners: 55% accuracy.
Conclusion - Original examiners not solely using polygraph output to judge.
Innocent suspect dilemma
Polygraphs often conducted when no other evidence is available
Means innocent suspects cannot prove their innocence via other means.
They realise this will become more difficult if they fail test but refusing test may lead to suspicion.
Types of error
False positive - false accusation of an innocent
False negative - saying guilty suspect is innocent
False positives more serious in investigations (opposite for catching and terrorists).
Types of study
Lab and field
Participants lie or tell the truth about a staged event
Certainty about who lies and who tells the truth (ground truth).
Not realistic – Lies are about trivial issues and ‘liar’ has been told to lie by experimenter.
Polygraph tests in real cases examined
Realistic (Ecologically valid).
Uncertainty about who lies and who tells the truth as polygraph tests are only carried out when there is no evidence. (Ground truth unknown).
Information provided by direct observation (empirical) as opposed to information provided by inference.
Checking the results of machine for accuracy against the real world.
Ginton et al. (1982) - field study method
Policemen - paper pencil test
Scored own tests (7 cheated).
Paper chemically treated so that changed answers could be detected
All told suspected of cheating and offered polygraph.
All agreed but one ‘guilty’ didn’t show.
Ginton et al. (1982) - field study results
3 guilty confessed before taking the test.
Final sample 2 Guilty, 13 Innocent
Both guilty detected, but 2 innocent rated guilty.
11 out of 13 correct sounds impressive but observer without polygraph also scored 11 out of 13
Independent rater using only polygraph = 7 out of 13
Sampling Bias in Field studies?
Difficult for examiners to receive feedback that contradicts their verdicts.
Guilty verdicts lead to interrogations which lead to confessions or denials.
Confessions (including false) are considered as support.
Denials (likely from innocents) are considered as “no evidence”.
Neither prove the examiner wrong.
Confessions by the ‘real perpetrator’ (only way to prove examiner wrong) are unlikely to occur
Pressing foot down
Thinking scary thoughts
Floyd 'Buzz' Fay countermeasurse
Floyd falsely convicted of murder based upon polygraph evidence.
Studied polygraph in prison.
Trained 27 prisoners, who admitted to him they were guilty, in countermeasures.
23 out of 27 passed the polygraph test
Guilty Knowledge Test (GKT)
Based on orienting theory
Presents suspects with sets of similar items, one of which will be true.
Knowledge is supposedly only one a guilty suspect could have e.g. which murder weapon.
Theoretically more sound and less disputed than the CQT.
75% of psychologists considered the GKT to be based on science whereas only 33% said so for CQT.
Field study results CQT and GKT
CQT: 85% accurate for guilty, 70% accurate for not guilty.
GKT: 60% accurate for guilty, 95% accurate for not guilty.
GKT and CQT error vulnerability
CQT vulnerable to false positives because innocents are aroused when answering relevant questions.
GKT vulnerable to false negatives because guilty suspects do not know what the correct response item was.
Both vulnerable to countermeasures so field study results may not reflect genuine accuracy.
Limited applicability: Can’t use it whenthere are no case facts known and a suspect admits having guilty knowledge but denies guilt.
Don't know that a guilty suspect knows the correct answer?
Can't ensure that an innocents does not know the correct answer
Innocents sometimes know correct answer because it was accidentally revealed in their police interview or in the media.
GKT expected outcome
Guilty suspects will have increased physiological arousal to the correct answer e.g. "she was shot".
Innocent suspects should have the same arousal for all answers.
Where did the idea of a polygraph for lie detection come from?
Throughout history it has been assumed that lying is accompanied by a change in the body’s physiological activity.
The polygraph accurately measures various physiological changes.
Therefore it is claimed that the polygraph can detect deception.
What constitutes a good psychometric measure?
Validity and reliability
CQT control questions
Control questions deal with acts that indirectly relate to the crime but do not refer to the crime in question.
The suspect has to lie as they are told that admitting the control questionwould conclude they are the type of person that would do the crime.
The questions are formed based on what the examiner believes the examinee has done earlier in life.
DLT control questions
DLT addresses the standardisation issue of CQT control questions.
The control questions are standardised in all situations “Did you ever tell a lie?”
People are instructed to lie on control questions rather than coerced into it by the examiner telling them that admitting it would conclude they were probably guilty.
The examinees are specifically told to respond negatively and untruthfully to control questions.
National Research Council (2003) GKT Criticisms
Doesn’t consider that the reactions to familiar stimuli is continuous rather than dichotomous.
Innocent examinees might have an unregistered gun similar to the murder weapon. The innocent examinee will still react to the gun and therefore be concluded as guilty.
Validity of polygraphs (lab tests)
(Raskin and Honts, 2002).
Validity – the extent to which the procedure measures what it claims to measure.
Lab studies use mock crimes.
Favourable results but guilty participants have little incentive to beat the test and innocent ones won’t be concerned about relevant questions so results are likely inaccurate.
Validity of polygraph tests (field tests)
Quality of field studies is low.
Hard to discover ground truth (whether a suspect is actually innocent or guilty).
Confessions accepted as ground truth but this is problematic.
A guilty suspect who is found innocent by the polygraph won’t confess.
A lack of confession means the study won’t be included in a field trial and therefore the incorrect polygraph decision won’t be noted (publication bias).
Suppressing physiological reactions to relevant questions/items (mental countermeasures e.g. meditation)
Reducing overall anxiety (use of drugs prior to exam)
Increasing physiological reactions to control/neutral questions (inflicting pain e.g. muscle tension)
Number of people that can be expected to be guilty in criminal investigations.
If 1/1000 employees in an organisation were spies then catching the spy with a polygraph is near impossible.
Polygraphs falsely accuse some innocent employees (say 30%).
Means 300 people will fail the polygraph test and so impossible to identify spy.
Polygraph uses in clinical settings
Increasing medication compliance
Treating sex offenders
Exploring mental states e.g. delusions.
Grubin (2000) polygraph in sex offenders
Studied 50 offenders and asked them to take two polygraph tests six months apart.
Polygraph increases the disclosure of information such as type of offenses, deviant fantasies/behaviours, number of victims etc.
High drop out rate (18% left in final sample).
BPS (2004) issues with polygraphs and clinical settings
Not scientifically established as a treatment so unsure of effectiveness.
High refusal/drop out rates.
Could be perceived as coercive rather than voluntary e.g. sex offenders may be worried that it would make them look bad if they didn’t partake.
ERP Lie Detection
P300 response may be useful in line with the GKT test.
Has similar accuracy when used alone but when used together it is possible that they may increase accuracy.
fMRI Lie Detection
Expensive and time consuming.
Brain activity more active in deception are also more active in non-deception tasks and therefore false positives are common.
More research needed.