Vous êtes sur la page 1sur 16

American Psychology-Law Society (AP-LS)

The Admissibility of Polygraph Evidence in Court: Some Empirical Findings

Author(s): Ann Cavoukian and Ronald J. Heslegrave
Source: Law and Human Behavior, Vol. 4, No. 1/2 (1980), pp. 117-131
Published by: Springer
Stable URL: http://www.jstor.org/stable/1393498
Accessed: 10-02-2016 10:29 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Springer and American Psychology-Law Society (AP-LS) are collaborating with JSTOR to digitize, preserve and extend access
to Law and Human Behavior.


This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions
Law and Human Behavior, Vol. 4, Nos. 1/2, 1980

The Admissibility of Polygraph Evidence

in Court
Some EmpiricalFindings*
Ann Cavoukiant and Ronald J. Heslegravet

Polygraph evidence is presently inadmissible in Canada and many jurisdictions of the United States. One of
the major reasons for its exclusion lies in the belief (held by members of the judiciary) that jurors would
accept such evidence without question due to its technical/scientific nature. The question of such blind
acceptance was examined in two experiments on the influence of polygraph evidence on people's judgements
of guilt. A second question that was also raised was whether a caution on the limitations of the polygraph
would be effective in reducing people's weighting of such evidence. Although polygraph evidence was ex-
pected to exert some influence over judgements of guilt, it was not expected to be so great as to result in
"blind acceptance." The results of both experiments supported this hypothesis. The inclusion of a caution
was also effective in reducing the influence of such evidence. The implications of these findings are discussed
in the context of the need to reexamine the admissibility of polygraph evidence in a court of law.

Withinthe UnitedStatestheadmissionof polygraph or "liedetection"evidenceinto
thetrialprocessvariesacrossjurisdictions(Abrams,1977).In Canada,however,such
evidenceis inadmissibleat present(Phillionvs.theQueen,1978).Someof thereasons
for its inadmissibility
arelegalin naturerequiring legalcounterargument, andthus
arenotwithintheboundsof empiricalinvestigation. Forexample,onelegalissuehas
been primarilyconcernedwith the failureof polygraphevidenceto satisfythe

*During the course of this research, both authors were supported by Doctoral Fellowships from the Social
Sciences and Humanities Research Council of Canada. In addition, this research was supported in part by
a grant-in-aid of research to the second author from the Centre of Criminology, University of Toronto,
from funds made available by the Solicitor General of Canada. The authors wish to thank Anthony N.
Doob for his helpful comments during the early stages of this research, and Marco Katic and Julian
Roberts for their help in the collection of the data.
tDepartment of Psychology, University of Toronto, Canada.

? 1980 PlenumPublishingCorporation

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

necessary requirements of expert testimony-most notably that such evidence not in-
vade the domain of matters considered to be exclusively up to the trier of fact (R. vs.
Phillion, 1973). This point is not, however, accessible to the tools of the empiricist.
Rather, it could be argued that other experts also give opinion evidence on issues that
the jury must decide (e.g., R. vs. Wong, 1977). For example, in a charge of threatening
where the threat is conveyed by a letter, it is admissible for a handwriting expert to
testify as an expert that, in his opinion, the accused was responsible for writing the
There are, however, arguments for the exclusion of polygraph evidence that may
be examined empirically. The principal argument open to investigation concerns the
accuracy of the polygraph technique. It has been successfully argued in some jurisdic-
tions that polygraph techniques are not sufficiently accurate to warrant acceptance as
evidence. This accuracy issue has not only been predominant in the courts but has also
been the theme of most psychophysiological research dealing with the polygraph as an
instrument for detecting deception. A wealth of information is, therefore, now
available on this issue. Even though there is considerable controversy over the "true"
accuracy rate that is attainable by skilled scientists using laboratory paradigms
(Abrams, 1975; Lykken, 1974, 1978, 1979; Podlesny & Raskin, 1977, 1978; Raskin,
1978; Raskin & Podlesny, 1979), it can be estimated that the polygraph, through ob-
jective quantification, can detect deception with accuracy rates between 64% (Hor-
vath, 1977) and 96% (Raskin & Hare, 1978) against chance rates of 50%. Clearly,
these detection rates define the technique as a legitimate phenomenon. Whether these
rates of accuracy are sufficient to warrant the inclusion of polygraphers' testimony as
evidence is a question that has both legal and empirical relevance. For the judiciary
the question is whether these accuracy rates are adequate; for the researcher the ques-
tion is whether this level of accuracy is at least as good as, if not better than, the ac-
curacy attributed to other types of evidence. For example, one can investigate whether
polygraph evidence is as accurate at detecting underlying psychological processes as
subjective and projective tests, such as the Rorschach or Thematic Apperception Test.
The latter are presently admissible pieces of evidence that may be introduced in sup-
port of expert testimony.
Little work has been done on this issue of relative accuracy. However, in the only
study comparing polygraph evidence to other forms of evidence which are presently
admissible in court, the polygraph fared very well. The accuracy of the polygraph as
an investigative tool was compared with the accuracy of several other methods of
identification: fingerprints, handwriting analysis, and eyewitness identification
(Widacki & Horvath, 1978). The polygraph was found to be the most accurate
method of those examined. As the authors state, "the number of correctly resolved
cases (those in which the perpetrator and the three innocent subjects were correctly
identified) was the greatest for the polygraph examiner, followed, in order, by the
handwriting expert, the eyewitnesses, and the fingerprint expert" (p. 598). This finding
should not be particularly surprising. The criteria for admissibility do not appear to
rest upon empirical support for accuracy. As Silverberg (1979) notes, "many other
forms of evidence that have traditionally been accorded uncritical judicial approba-
tion are far less reliable [than the polygraph]. For example, the eyewitness account ...
has a degree of unreliability horrifying to anyone who still clings to the notion of ob-
jective certainty in the judicial process" (p. 6).

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

The studiesin the presentarticledeal with an empiricalissue to whichvery little

attentionhas been directed.This issue concernsthe effects of polygraphevidenceon
jury decisions.Even if the problemof the degreeof accuracyis resolvedin favor of
polygraphevidence(as it has been in some jurisdictions),an independentissue con-
cernsthe effectsof the inclusionof suchevidence.If polygraphevidencehas negligible
effects on jurors, then the importanceof the accuracyissue becomes severelycur-
tailed. If, on the otherhand, polygraphevidencehas a significanteffect, then the ac-
curacyissue remainscritical.In addition,researchmustthenbe undertakento demar-
cate the limitations of this effect, i.e., under what circumstancesdoes polygraph
evidencesignificantlyaffect decisions?The initial goal of the studiesto be reported
herewas to discoverwhetherpolygraphevidencesignificantlyaffectedthejudgements
of individualpeople. The individualjudgEmentsof jurors prior to deliberationhave
beenfoundto be one of the best predictorsof finaljurydecisions(Doob & Cavoukian,
1977, p. 201; Kalven& Zeisel, 1966, p. 488).
Although most would expect that polygraphevidence has an effect on jury
decisions, this assumptionmust be supportedprior to acceptance.If we assume,
however,that supportis forthcoming,a more criticalissue ariseswhichhas led to the
inadmissibilityof polygraphevidencein Canadaand otherjurisdictions(e.g., see R.
vs. Phillion,1973).The problemas expressedby some membersof thejudiciaryis that
"thejury, by reasonof the technicalityof the evidence,might be temptedto blindly
accept the witness'(polygraphexpert)opinion"(R. vs. Phillion, 1973, p. 210) and
thereforeplace undueweighton this type of evidence.A secondgoal of these studies,
then, is to examinewhetherjurors blindlyaccept such evidenceand whetherundue
weight is in fact placed on polygraphevidenceby jurors. The only researchon this
issue has been of the quasiexperimental variety(Barnett,1973;Carlson,Passano,&
Jamuzzo, 1977; Forkosoch, 1939; Koffler, 1957) and has yielded equivocalresults'
therebydemonstratingthe need for an experimentalapproachto the problem.
The final goal of these studiesis to examineand documentany ameliorativein-
fluence on the effect of polygraph evidence that may occur due to cautionary
statementsconcerningthe accuracyof polygraphtests. Sincejurisdictionswhichhave
ruled polygraphevidence inadmissiblewould be likely to include such cautionary
statementsif the admissibilitystatusof suchevidencewerereversed,it is importantto
estimatethe debilitativeeffect suchstatementswouldhave on polygraphevidence.In
addition,if such evidencewere found to be undulyweightedby jurors, cautionary
statementson the accuracyof the polygraphcould effectivelyreducethe impact of
such evidence.In the event that polygraphevidencedid not carry undue weight, it
wouldbe expectedthat cautionarystatementswouldhaveless influence;how powerful
a debilitativeeffect such statementsactuallyhave is a separateempiricalquestion.


The first experimentwas designedto obtainsome preliminaryexperimentaldata

with respectto the threegoals of the researchprogramlistedabove.Therewerethree

'The results of these studies are considered to be equivocal primarily because of their lack of proper control
conditions as well as the fact that the results obtained fell in opposite directions with respect to the
weighting and influence accorded to polygraph evidence.

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

differentconditionsin this experiment.The firstcondition(Basic)consistedof a sum-

mary of the majorpointsof evidencein an actualcase and the judge'sinstructionsas
to the laws pertainingto the case. Since the major precedent-settingcase involving
polygraphevidencein Canadawas R. vs. Phillion(1973, 1975;Phillionvs. the Queen,
1978),it was decidedto use this case as the basis for the summary.R. vs. Phillionis
also an appropriatecase to summarizein another respect, i.e., it did not contain
polygraphevidence. In this case polygraphevidencewas deemed inadmissible-a
judgementthat was latersupportedby the SupremeCourtof Canada(Phillionvs. the
Queen, 1978). This condition,then, providesa baselinelevel for the degreeof guilt
that the summaryof the actual case engenders.
The secondcondition(Polygraph)was designedto examinethe firsttwo goals of
this research.This conditionconsistedof the identicalsummaryof the case usedin the
Basicconditionplusthe inclusionof polygraphevidenceshowingthe accusedto be in-
nocent.2Since the only differencebetweenthe Basic and Polygraphconditionsis the
inclusionof the polygraphevidencein the latter,contrastingtheseconditionsprovides
data on the firsttwo goals of this research:the questionof whetherthe admissibilityof
polygraphevidencehas a negligibleor significanteffectis directlyanswered.Also, if it
can be demonstratedthatthe presentationof suchevidenceresultsin a significantshift
towardsjudging the accused as less guilty, the extent of the shift along the in-
nocent-guiltycontinuumwill enablethe determinationof whetherthis evidenceis in
fact "blindly"accepted.3
The thirdcondition(Judge'sCaution)was designedwith respectto the last goal.
This condition was identical to the Polygraph condition with one exception: a
cautionaryinstructionwas givenby thejudge that polygraphtests wereabout80%ac-
curate,a fact that they shouldbearin mindand be cautiousaboutwhenweighingthe
evidence.Since the only differencebetweenthe Polygraphand Judge'sCautioncon-
ditions was the inclusionof the cautionarystatementby the judge, contrastingthese
two conditionsprovidesevidenceas to whetheran appropriatecautionby thejudgeis
sufficientto effectively reduce the impact of the polygraphevidenceproviding,of
course, that such an impact is significant.

Subjects. The subjects were 150 visitors to the Ontario Science Centre who
volunteeredto participatein the experiment.Subjectswereof both sexes and over 18
years of age.4

2This experiment is not a complete factorial design since the polygraph evidence only supports the accused
by showing him to be innocent. The reason for this restriction is that in jurisdictions where such evidence is
inadmissible, it is likely that the admission of such evidence would occur only if it were introduced by the
defense to benefit the accused. This is especially likely since it can be demonstrated that errors in the out-
come of polygraph tests are more likely to find the innocent guilty than the guilty innocent (Horvath,
1977). In fact, this was the condition of admission of polygraph evidence in the only Canadian case where it
was admitted. The judge's opinion was that "polygraph evidence led by the Crown as evidence of guilt, not
of innocence, should be excluded as highly prejudicial and less reliable" (R. vs. Wong, 1977, p. 6).
3Although we refer to shifts in perceived guilt towards "more guilt" and "less guilt" on an innocent-guilty
continuum, this continuum does not refer to the amount of guilt but rather refers to the probability that the
guilt category of the innocent-guilty dichotomy is correct.
4Since the subjects used in these experiments were drawn from visitors to a science center, it is possible that
such a group would tend to be middle class and fairly well educated. It may be the case that less well-

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

Procedure. Subjects were individually randomly assigned to one of the three con-
ditions. Each subject was given a booklet containing a description of the case of R. vs.
Phillion (1973) and a questionnaire. The variations in the three conditions were in the
defense that was presented for the accused. In the Basic condition, the defense con-
sisted of the testimony of two expert witnesses: a psychologist and a psychiatrist.
Their evidence was directed at showing that the accused had a deranged personality,
and that he had a tendency to invent and attest to circumstances which had never
happened. These witnesses supported the defense's contention that the accused had
been lying when he had confessed to the police. They testified that their conclusions
had been reached on the basis of several psychological tests which showed the accused
to be psychologically unstable.
In the Polygraph condition, subjects received the identical information contained
in the Basic condition. In addition, the testimony of a polygraph expert was included.
His testimony supported the testimony of the other experts since he testified that, in
his opinion, the accused had not killed the victim. The polygraph expert testified that
his opinion was based on the results of a polygraph test where the accused had
answered truthfully when he responded "no" to whether he had killed the victim.
In the Judge's Caution condition subjects received the identical information to
the Polygraph condition except that they were also given a cautionary instruction
from the trial judge with respect to the limitations of polygraph evidence. The judge
drew the jury's attention to the fact that polygraph tests had about an 80% accuracy
rate, or that polygraphers were correct in their findings about 80 times out of 100. The
judge then instructed the jury to be cautious and bear this in mind when weighing the
polygraph evidence.
The 80% accuracy level used by the judge was deemed reasonable by the authors.
Although it has been stated above that accuracy estimates range from 64% (Horvath,
1977) to 96% (Raskin & Hare, 1978) against chance rates of 50%, the high levels of
reported accuracy are probably overly optimistic. Accuracy estimates are affected by
many factors, one of which is that accuracy is directly related to the experience and
training of the polygrapher (Horvath & Reid, 1971) and most have limited training
(Lykken, 1974). Lykken (1974, 1978, 1979) has also raised several other problems
with the determination of accuracy estimates. For instance, a polygrapher's decision is
often a clinical judgement based on the entire knowledge from the examination (e.g.,
the demeanor of the subject) and the facts of the case rather than an objective,
numerical assessment of the polygraph charts. Accuracy estimates are thus inflated
due to increased knowledge on which to make a decision. Finally, accuracy estimates
are often inflated since corrections for unconfirmed and inconclusive cases, as well as
baseline levels of deception, are often lacking. In view of these problems, an 80% ac-
curacy level seemed reasonable since it was within the accepted range of accuracy, but
short of the high levels reported. In addition, since the polygraph examination in this
case showed the defendent to be honest, the level chosen was above the arithmetic
average of the estimates due to the fact that errors occur less often in the false negative

educated people would have been more influenced by the polygraph than those in this sample. This
remains, however, only a speculation since no demographic information was available. Nevertheless, the
authors felt that such a group would be more representative of those called to jury duty than a sample of
college students.

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

direction(Horvath, 1977;Lykken, 1979). In any case, the level chosenmay have an

effect on the outcomeregardlessof how closely it resemblesthe true state of affairs,
thus the stated level of accuracyshould be a subjectof enquiryin furtherresearch.
At the end of the writtenmaterialin all three conditions,the subjectsread the
judge'schargeto thejuryin whichhe summarizedthe evidenceandinstructedthemon
the laws pertainingto the case. He also instructedthem on the standardof proof re-
quired in a criminal case, giving the traditional instructionon proof beyond a
DependentMeasures.After readingthe case, all subjectswere asked to answer
the followingtwo questions.Theywerefirstaskedto indicatehow likelythey thought
it was that the accusedwas guiltyof committingthe murder(on a 7-pointscale where
1 = definitely guilty, 3 = probably guilty, 5 = probably not guilty, and 7 = definitely
not guilty).Theywerethen askedwhatverdictthey, as membersof a jury,wouldgive.
All subjectswere then thankedand given a debriefingexplanation.

Results and Discussion

Figure 1 shows the mean perceived-guiltratings, where a value of 1 indicates
definitely guilty and 7 indicates definitely not guilty, and the standard errors
associatedwith each groupmean. An overallanalysisof varianceindicatedthat the
differencesamong the groupswere highly significant[F(2,147) = 6.22, p < .01].
Effectsof PolygraphEvidence.Onepurposeof this studywas to evaluatewhether
the admissionof polygraphevidenceshowingthe accusedto be innocentwouldresult
in a significantshift along a perceived-guiltscale towardsless guilt. Therefore,the
Basicand Polygraphconditionswerecontrasted.As can be seen in Figure1, the inclu-
sion of polygraphevidencesupportingthe innocenceof the accusedresultedin the ac-
cused being perceivedas significantlyless guilty [F(1,147)= 11.62,p < .001]. The
dichotomousverdictdata shownin Table 1 supportthese resultsin that only 48%of



3t 4-


b_l 2-
r'",Q~~CAUT 3-ION

Fig. 1. Mean perceived-guilt ratings and standard errors for each group (7-point scale ranging from
1 = definitely guilty to 7 = definitely not guilty; the greater the numerical rating, the less guilt perceived).

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

the subjects in the Basic condition acquitted the accused whereas 72% of the subjects
in the Polygraph condition voted for acquittal [x2(1) = 5.04, p < .05]. It would
therefore seem clear from both the perceived-guilt ratings as well as the resulting ver-
dict decisions that the inclusion of polygraph evidence had the effect of making the ac-
cused appear significantly less guilty. This point is critical since it defines a basic dis-
tinction between those jurisdictions that allow polygraph evidence and those that do
not, thus leaving open the possibility of altered verdicts in those jurisdictions that do
not allow such evidence.
Although there is a significant effect of including polygraph evidence, it is the ab-
solute mean shift along the scale that is important in the judgement of whether the
subjects "blindly" accepted the polygrapher's testimony. If subjects blindly accepted
the polygrapher's testimony of the innocence of the accused, especially since it cor-
roborated the testimony of the psychologist and psychiatrist, it may be expected that a
mean judgement of 6 or greater would occur. The actual result was significantly less
than 6. As in any study, there are sampling errors involved; however, there is less than
1 chance in 1000 that the actual value of perceived guilt should even be 5 (probably not
guilty) or greater (standard error = .185).5 Therefore, it would seem that the subjects
did weigh the polygraph evidence rather than blindly accepting it. Judgements of
whether such evidence has an "undue influence" are largely subjective in nature,
though the second experiment attempted to provide a more objective answer.
Effects ofJudge's Caution. In the Introduction it was stated that in the event that
polygraph evidence produced an effect, it would then be of interest to see whether a
cautionary statement by the judge could mitigate this effect. In Figure 1 it can be seen
that the Judge's Caution condition resulted in the effect of the polygraph evidence be-
ing reduced somewhat, i.e., relative to the Polygraph condition the Judge's Caution
resulted in the accused being perceived as slightly more guilty. Although the results
were in the predicted direction, the difference between the Judge's Caution and
Polygraph conditions was not significant. As in the case of the Polygraph condition,
the Judge's Caution condition resulted in the accused being perceived as significantly
less guilty than in the Basic condition [F(1,147) = 6.21, p < .02]. This result indicates
that despite the judge's cautionary instruction, the polygraph evidence resulted in the
accused being perceived as significantly less guilty.
The distribution of verdicts over the three conditions was found to vary
significantly [X2(2) = 6.00, p < .05] as presented in Table 1. In the Judge's Caution
condition 60% of the subjects acquitted the accused. It is interesting to note that this
result was not only in the predicted direction but fell midway between the Basic condi-

Table 1. Distribution of Verdicts in Experiment I

(Frequencies and Percentages)

Basic Polygraph Judge's Caution

N % N % N %

Guilty 26 52 14 28 20 40
Not guilty 24 48 36 72 30 60

5These probability levels can be calculated using the standard error in Figure 1 for the polygraph group.

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

tion and the Polygraphcondition.Thejudge'scautionaryinstructionreducedthe in-

fluenceof polygraphevidenceby half. These resultscould be regardedas desirable
since they suggestthat peoplehad, in accordancewith the judge'scaution,weighted
the polygraphevidencewithgreaterreservationand as a resultgave feweracquittals.
If the resultsof both dependentmeasuresare takentogether,it wouldappearthat the
judge's caution reduced the effect associated with the inclusion of the polygraph

The first experimentprovidedthe initial experimentallyproduceddata on the
effects of consideringpolygraphevidenceas probative.It also providedinformation
on the degreeto whichsuch effects could be attenuatedby a cautionaryinstruction
fromthejudgeconcerningthe accuracyof polygraphtechniques.However,this single
demonstrationcannot, by itself, necessarilybe considereda convincingor reliable
demonstrationof a phenomenon.Therefore,the first goal of the second experiment
was to attemptto replicatethe findingsof the firstexperimentin orderto demonstrate
the reliabilityof those effects.To accomplishthis goal the threeconditionsusedin the
first study(i.e., Basic, Polygraph,and Judge'sCaution)werereplicatedin this study.
However,it was believedthat the generalityof similarfindingsin this studywouldbe
enhancedif a differentbasic case descriptionwas used in orderto vary the baseline
perceived-guiltlevels. Thusa secondcase was used.The case chosenwas R. vs. Wong
(1977)-the only otherreportedCanadiancase in whichpolygraphevidencehad been
an issue.6
Two new issues were also examinedin this study.The first issue centeredon the
effect of the cautionaryinstructionby the judge in the first experimentas well as the
artificialnatureof sucha circumstance:if a polygrapher'stestimonywereacceptedin
a case as expertopinionevidence,it is unlikelythat a judge's caution would be the
only cautionarystatementmade.Quiteoftenopposingtestimonyis broughtforthwith
respect to the adequacy of the particulartest and the examiner'sopinion. This
testimonywouldservethe dual purposeof providinga cautionarystatementfrom an
expert as well as relievingthe judge from making such a statementwhich could,
perhaps,later constitutegroundsfor appeal.Therefore,a fourth(Expert'sCaution)
conditionwas addedto this experimentwhichwas identicalto the Polygraphcondi-
tion except that experttestimonyreplacedthe judge's caution. Althoughthis state-
mentwas strongerthanthe statementusedby thejudge,boththejudgeandthe expert
referredto an 80%accuracyrate.
The second issue concerns the weight which jurors attach to the polygraph
evidence.The first study supportsthe contentionthat peopledo not "blindly"accept
polygraphevidence,and the replicationof the same conditionsin this study may
providefurthersupportfor this assertion.However,this study attemptedto objec-
tivelyinvestigatewhetherpolygraphevidencehas an "undueinfluence."Sinceany im-

6Incontrastto R. vs. Phillion(1973),the polygraphevidencewas ruledadmissiblein this case. The Basic

conditiondid not, however,includethis evidenceandonly provideda descriptionof the remainingfactsin
the case. The polygraphevidencewas addedin the Polygraphcondition,as in the first study.

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

portantpiece of evidencewouldbe expectedto shift the degreeof perceivedguilt, it is

impossibleto knowhow muchof a shiftwouldbe consideredan undueinfluence;only
subjectiveappraisalsare possible. For example, in ExperimentI the inclusionof
polygraphevidenceshiftedthe perceivedguilt 1.04unitsalong a 7-pointscaletowards
less guilt. Subjectivejudgementscan be madewithregardto whetherthis shift is "too
much," but such judgements will have great variation and different criteria.
Therefore,to judge "undueinfluence"there must be an estimateof how much of a
shift would be expectedif an additionaladmissiblepiece of evidencewas included.
With respectto this questionof "undueinfluence,"the resultsof ExperimentI are
confoundedsince the differencebetweenthe Basic and Polygraphconditionsis the
summationof two types of influenceswhich can be regardedas quantitativeand
qualitative.The quantitativeinfluencecan be interpretedas the effect of an additional
piece of evidenceshowingthe accusedto be innocent,whilethe qualitativeinfluence
relatesto the fact that this evidencewas polygraphicin nature.It is not possibleto
partitionhow muchof the 1.04shiftfromthe Basicto the Polygraphconditionwas at-
tributableto the quantitativeand qualitativeaspects. In an attempt to answerthis
questionin the contextof the R. vs. Wong(1977)case description,a fifth (Alternative
Evidence)conditionwas addedwhichwasidenticalto the Basicconditionbut included
anotherpiece of evidencesupportingthe accused.It was predictedthat this evidence
wouldalso resultin the accusedbeing perceivedas less guilty and the degreeof this
shift would pertainto the questionof "undueinfluence."

Subjects. The subjects were 250 visitors to the Ontario Science Centre who
volunteeredto participatein the experiment.Subjectswereof both sexes and over 18
years of age.
Procedure.Subjectswere individuallyrandomlyassignedto one of the five con-
ditions. Each subjectwas given a booklet containinga short descriptionof R. vs.
Wong (1977) and a questionnaire.Althoughthe case was differentfrom that used in
ExperimentI, both involved charges of murder,and the additional evidence in-
troducedin the variousconditionsattestedto the innocenceof the accused.As in the
first study, variationswere introducedin the defensepresentedfor the accused.The
Basic, Polygraph,and Judge'sCautionconditionswere meant to replicate,using a
differentcase, the essentialsof those conditionsin ExperimentI.
The Basicconditioninvolveda descriptionof the evidencepresentedin the case.
The facts presentedto the subjectswere similar to those in the Wong case. Four
Chineseyouthsattemptedto entera ChineseNew Year'spartyto whichthey werenot
invited.Whenthe youths were refusedadmittanceto the partyoutsidethe house, a
fight ensued. Five men were injuredin the fight, one of whom died on route to the
hospital.An eyewitnesstestifiedthat as he was arrivingat the party at about 11:15
p.m., he saw the fight and subsequentstabbing.Followingthe stabbing,the assailant
ran past his car with the weapon.The witnessidentifiedthe accusedas the assailant
and statedthat he had seen him on previousoccasionsand was thus ableto recognize
him. The accusedtestifiedthat on the way to the partyhe discoveredthat his friends
had weaponson them and so had decidedto remainin the car. He deniedattempting
to enter the party and denied stabbingthe victim.
In the Polygraphcondition,subjectsreceivedthe identicalinformationcontained

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

in the Basic condition.In addition,a police polygraphertestifiedthat, in his opinion,

the accused answeredtruthfullywhen he responded"no" to the questions "On
January23, 1976,did you cut someonewitha knive?"and "OnJanuary23, 1976,did
you stab anotherman?".Further,a recognizedexpertin lie detectiontestifiedthat he
had reviewedthe polygraphchartsand was also of the opinionthat the accusedhad
respondedto the questionshonestly.
In the Judge'sCautioncondition,subjectsweregiventhe same informationcon-
tainedin the Polygraphconditionplus a cautionfrom the judge that was identicalto
the first study. Specifically, the judge drew the jury's attention to the fact that
polygraphtests had about an 80%accuracyrate, or that polygrapherswere correct
about 80 times out of 100. Thejudgethen instructedthejury to be cautiousand bear
this in mind when weighingthe polygraphevidence.
The Expert'sCautionconditionwas identicalto the Polygraphconditionwithad-
ditionaltestimonyfrom a recognizedexpertin lie detectioncautioningthe jury about
the accuracyof polygraphtests. He testifiedthat "underthe bestconditionspolygraph
tests are only about80%accurate,that is to say they correctlyidentifythe truthabout
80 times out of 100.Also, the 80%accuracyfigureis basedon laboratorystudiesand
thereis virtuallyno informationon how accuratepolygraphexaminationsare in real-
life situationsthoughit is likelythat theyareevenless accurate."He also saidthatthe
results should be treatedwith skepticism.
The finalAlternativeEvidenceconditionwas identicalto the Basicconditionex-
cept that anotherwitnesstestifiedin supportof the accused'stestimony.One of the
four youthsinvolvedin the fight testifiedthat the accusedhad stayedin the car when
the otherstriedto enterthe party.The rationalefor this additionalpiece of evidence
was that it corroboratedthe accused'stestimonyin a similarmannerto the Polygraph
evidence.However,since eyewitnessidentificationhas been shown to exert a great
deal of influencewhen introducedas evidence(Loftus, 1975), one of the youths in-
volved in the fight (whose credibilitywas somewhat suspect) was chosen as the
eyewitnessin order to reducethe otherwisepowerfuleffects of such testimony.
At the end of all five conditions,the subjectsreadthejudge'schargeto thejuryin
whichhe summarizedthe evidenceand instructedthem on the laws pertainingto the
case. He also instructedthemon the standardof proofrequiredin a criminalcase, giv-
ing the traditionalinstructionson "reasonabledoubt."
DependentMeasures.After readingthe case, all subjectswere asked to answer
severalquestions.Theywerefirstaskedto indicatehow likelytheythoughtit was that
the accused was guilty of committing the crime (on a 9-point scale where
1 = definitely guilty, 3 = probably
guilty, 5 = uncertain, 7 = probably not guilty,
and 9 = definitelynot guilty).Theywerethen askedwhatverdictthey, as membersof
a jury, wouldgive. Finallythey wereaskedhow confidentthey werein theirjudgments
(on a 9-point scale runningfrom 1 = very confidentto 9 = not at all confident).
Threegeneralquestionswerealso asked;subjectswereinstructedto disregardthe
case descriptionwhen answeringthese questions.Subjectswere asked how familiar
they werewith polygraphs(1 = veryfamiliarto 9 = not at all familiar);how accurate
they thoughtpolygraphtests were(1 = very accurateto 9 = not at all accurate);and
whetherthey thoughtpolygraphtests weremore, or less, accuratethan psychological
tests (1 = polygraph tests much more accurate, 5 = about the same, and 9 = psy-
chologicaltests much more accurate).

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

Results and Discussion

The firstgoal of the secondexperimentwas to replicatethe effectsof Experiment
I withthe use of a differentcase, whichmay haveresultedin a differentbaselinelevel
of perceivedguilt, in orderto determinewhetherthose effectswouldgeneralize.The
resultsof this experimentclearlyreplicatethe firststudy,as will be demonstrated.The
overallresultsand standarderrorsassociatedwith each groupmean are presentedin
Figure 2. An analysis of variancerevealedsignificantdifferencesamong the mean
perceived-guiltratings for each group [F(4,245)= 3.47, p <.01].
Effectsof PolygraphEvidence.In orderto specifythe effects of the inclusionof
polygraphevidence, the Basic and Polygraph conditions were compared on the
perceived-guiltrating scale. The results from this study replicatedthose resultsob-
tained in ExperimentI and revealedthat the accusedwas perceivedas significantly
less guilty when polygraph evidence attesting to his innocence was included
[F(1,245)= 4.37, p <.05]. The verdictdata, however,cannot be comparedfor these
groupsdue to the anomalousresponsesof the Basic group:these subjectsresponded
with a surprisinglylow convictionrateconsideringtheirmeanperceived-guiltratings.
Specifically,all subjectswho respondedwith a perceived-guiltrating of 1 or 2 con-
victedthe accused.However,excludingthe Basicgroup,81%(25 out of 31) of the sub-
jects who respondedwith a ratingof 3, whichwas labeled"probablyguilty,"also con-
victedthe accused.For the Basic group,only 36%(4 out of 11) of the subjectsrating
the accusedas probablyguilty convicted,whichis a significantlylow convictionrate
[x2(l) = 5.52, p < .05].
The questionas to whethersubjects"blindly"acceptedthe polygraphevidence
also arisesin this experiment.Figure2 showsthe meanandstandarderrorsassociated
with the Polygraphcondition(standarderror= .261). If, as in ExperimentI, it is es-
timatedthat "blindacceptance"shouldresultin a mean of at least 8 on this 9-point
scale, it can be seen that the resultis significantlybelow 8. In fact, the resultis also
significantlybelow 7 which was labeled "probablynot guilty" (p <.001). It would
seem that subjectsdid not blindlyaccept such evidence.Althoughboth experiments

() 7-

0 5-
iLJ -I-
Fig. 2. Mean perceived-guilt ratings and standard errors for each group (9-point scale ranging from
1 = definitely guilty to 9 = definitely not guilty; the greater the numerical rating, the less the perceived

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

showthat the inclusionof polygraphevidenceshiftedthe level of perceivedguilt, the

effect was not potentenoughto move the level of perceivedguilt significantlyclose to
the value of probablynot guilty.
Effects of Judge'sCaution.Figure2 showsthe effect of the cautionarystatement
by the judge. Althoughthe effect of the judge's caution was in the predicteddirec-
tion-showing the accusedto be slightlymoreguilty-the magnitudeof the reduction
fromthe Polygraphconditionwas small. As was the case in the first experiment,the
differencebetween the Polygraphand Judge's Caution conditions did not reach
significance(F <1). The verdictdata in Table 2 also showedthat the judge'scaution
did not significantlyincreasethe numberof guilty verdictsrelativeto the Polygraph
condition[X2(1) = 1.86,p > .10]. These resultsindicatethat althoughthe cautionary
instructionfrom the judge did not have a significanteffect, its influencewas in the
predicteddirectionin both experiments,i.e., towardsmore guilt.
Effects of AlternativeExpert Testimony.As stated in the Introduction,rather
than the judge merely cautioning the jurors about possible errors involved in
polygraphevidence,it is more likely that an alternativepolygraphexpertwould be
called to testify about the inadequacyof the polygraphevidence.Therefore,the Ex-
pert's Caution condition was included to estimate the mitigating effect of such
testimonyon polygraphevidence.
On the perceived-guiltscale in Figure2 it can be seen that the experttestimony
completelyeliminatedthe effect of the polygraphevidence.It was foundthat alter-
nativepolygraphexperttestimonysignificantlyshiftedthe level of perceivedguilt in
the direction towards more guilt when compared to the Polygraph condition
[F(1,245)= 7.39, p <.01]; the Expert'sCautionand Basic conditionsdid not differ
significantly(F <1). In addition,whenthe Expert'sCautionconditionwas compared
to the effects of the Judge'sCaution,it was found that the testimonyof the expert
reducedthe effect of polygraphevidencesignificantlymore than the judge's caution
[F(1,245)= 5.05,p <.01]. The verdictdata also supportedtheseresults:comparedto
the Polygraphcondition,the Expert'sCautionconditionresultedin a significantly
greater numberof convictions[x2(1) = 4.16, p < .05], as shown in Table 2.
It mustbe added,however,that the contentof the two cautionswas not identical:
the expert'scautionwas purposelysomewhatstrongerthanthat givenby thejudge,as
noted earlier. The expert added two additional comments in criticism of the
polygraph.Althoughboth the judge and expertused the same degreeof accuracyin
theircautions,the expertwent on to say that this 80%figurewas basedon laboratory
studiesandthat therewas practicallyno informationon the accuracyof polygraphsin
real-lifesituations,thoughsuch evidencewould likely be even less accurate.The ex-
pert also addedthat, in his opinion,the resultsof the polygraphtest shouldbe treated

Table 2. Distribution of Verdicts in Experiment II (Frequencies and Percentages)

Judge's Expert's Alternative

Basic Polygraph Caution Caution Evidence
N % N % N % N % N %

Guilty 7 14 5 10 11 22 14 28 12 24
Not guilty 43 86 45 90 39 78 36 72 38 76

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

with a great deal of skepticism.The differencein the effects of the two cautionsmay
not then be solely attributableto differencesin the source.The additionalcomments
in the expert'scautionmay have contributedto the great successof the expertin re-
ducingthe influenceof the polygraph.
Does PolygraphEvidenceUndulyInfluenceJurors?The other issue in this ex-
periment was to determine whether polygraph evidence unduly influencedjury
decisions. Data have been providedto demonstratethat jurors do not appear to
"blindly"acceptsuchevidence;however,it was statedthat the appropriatemethodof
determiningwhether such evidence had an "undue" effect was to compare the
polygraphevidenceto a qualitativelydifferenttype of evidence,so that the amountof
evidencein favor of the accusedwould be equivalent.However,this other piece of
evidencein the AlternativeEvidencecondition(namely,the corroborationof the ac-
cused's testimony by another witness) must at least minimally have moved the
perceived-guiltrating and verdictsin the same directionas the polygraphevidence,
that is towardsless guilt and more acquittals.The AlternativeEvidenceconditionin
this experimentfailed on both countsand so failedto providean appropriatecontrol
condition.The questionof undueinfluencethus remainsopen to empiricalinvestiga-
tion. Subjectively,however,shifts of 1.04on a 7-pointscale and .80 on a 9-pointscale
do not seem to constitutean "undueinfluence."
were first asked how confident they were in their judgements. There were no
differencesamongthe groupswith respectto this questionandthe meanratingof con-
fidence was 3.98 on this 9-point scale. However, if subjectsrespondedto the con-
fidencequestionin a consistentmannerwith their perceived-guiltrating,they should
have been more confidentin theirjudgementsthe furtherthey moved towardsboth
ends of the scale. To examine whetherthis internal consistencywas achieved, a
quadraticfunction was fitted to the mean confidencerating at each level of the
perceived-guiltscale. This result was highly significant[F(1,241)= 58.00, p <.001]
indicatingthat subjectsweremoreconfidentin theirratingsas theirperceptionsof the
accusedbecame more polarizedtowardsguilt or innocence.
Subjectswerealso asked threegeneralquestionsrelatedto polygraphtests:how
familiarthey were with polygraphtests, how accuratethey thoughtpolygraphtests
were, and how accuratethey thought polygraphtests were relative to other psy-
chologicaltests used by psychologistsand psychiatrists.The mid-pointof the scale on
the last questionwas labeled "aboutthe same." In terms of group responses,there
were no differencesamong the groups on any of the questions.Subjectsresponded
that they weresomewhatfamiliarwith polygraphs(mean = 5.44), thoughtpolygraph
tests were somewhataccurate(mean = 4.62), and felt that polygraphtests had about
the same accuracyas otherpsychologicaltests (mean = 4.98). It is interestingto note
fromthese resultsthat subjectsdid not thinkthat polygraphtests wereveryaccurate,
despitethe fact that three of the groupswere told the accuracyrate was about 80%.
They also thoughtthat polygraphtests were about the same as psychologicaltests in
terms of accuracy. This suggests that polygraph tests are generally not blindly
acceptedor even carry a great deal of weight relativeto other psychologicaltests,
which are presentlyadmissiblein court.
Using trendanalysisto interrelatethese dependentvariables,it was determined
that all significant interrelationshipsinvolved significant linear components

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

alone-all involvingthe questionof polygraphaccuracy.In summary,the more ac-

curatesubjectsviewedpolygraphtests:(a) the moretheythoughtpolygraphtests were
betterthanpsychologicaltests [F(1,241)= 65.06,p <.001]; (b) the morefamiliarthey
werewithpolygraphtests [F(1,241)= 8.78,p <.01]; and (c) the morethey viewedthe
accusedas innocent [F(1,241)= 8.91, p <.01]. Taken together,these resultswould
suggestthat in generalmost subjectsviewedpolygraphsas only mildly accurateand
about the same in accuracyas psychologicaltests. However,as they came to regard
the polygraphas more accurate (possibly because of increased familiarity),they
weightedit morestronglythanpsychologicaltests, andalteredtheirperceptionsof the
guilt of the accused(in the directionof greaterinnocence).

As stated in the Introduction,there were three questionswhich these studies
attemptedto address:(a) does polygraphevidenceaffectjury decisions,(b) if so, do
jurors "blindly"accept such evidence,and (c) if jurors are cautionedabout the ac-
curacyof polygraphtests, does this cautionreducethe effect of polygraphevidence?
The results of both experimentsprovideconsistentanswersto these questions.
Both experimentsrevealedthat the inclusionof polygraphevidenceshowingthe
accusedto be innocentresultedin peopleperceivingthe accusedas beingsignificantly
less guilty in terms of both perceived-guiltratingsand verdicts.Similarresultswere
obtained even though differentcase descriptionswere used, and the accused was
perceivedas being differentiallyguilty in the two cases. These resultssuggest that
significant differences could arise between jurisdictionson the basis of the ad-
missibilityor inadmissibilityof such evidence.Thus, such evidencecould, depending
on how guilty the jurorsperceivedthe accusedto be, raise a reasonabledoubtwhere
none existed before.
With respectto the questionof whethersuch evidencewas "blindly"accepted,
neither study found evidence for such blind acceptance.The shifts in the level of
perceivedguilt with the inclusionof polygraphevidencewerestatisticallysignificant
but not overwhelming.In fact, in neither study did the perceptionof guilt shift
significantlyclose to the point on the scale labeled"probablynot guilty."Thesefind-
ings are in line with the conclusionarrivedat by Tarlow(1975), wherehe statedthat
"the concernfor the 'overwhelmingimpact'of the polygraphis greatly exaggerated
and totally unjustified"(p. 968). In addition,it was demonstratedthat peopledid not
considerthe polygraphto be very high in accuracynor did they considerit to be
superiorto other types of psychologicaltests. These results suggest that the ad-
missibilityof polygraphevidenceshouldnot fostergreatconcernsince peopledid not
appearto have blindly acceptedsuch evidence,even when it was presentedwithout
Finally, even though polygraphevidencewas not blindly acceptedand did not
appearto have had a great influence,the effect of includingpolygraphevidencewas
diminishedby includingcautionarystatementsaboutthe accuracyof suchevidence.A
cautionfromthe judgeproduceda small reductionin the effect of polygraphevidence
in both studies.A strongercautionfrom the testimonyof a polygraphexpertin the
second experimentcompletelyeliminatedthe effect of the polygraphevidence.Al-
though these specific results were likely related to the level of perceivedguilt at-

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions

tributed to the accused, the results clearly demonstratedthat alternativeexpert

testimonywas influentialin reducingthe effect of polygraphevidence.
Of course, furtherresearchin this area is necessaryto establishsome of the
limitingparametersaffectingthese conclusions.If, however,the presentfindingsare
replicated,thenjurisdictionsin whichpolygraphevidenceis presentlyexcludedshould
reconsiderthe question of admissibilitywith specific reference to some of the
parametersobservedin these studies. Althoughthe inclusionof such evidencemay
serveto raise a reasonabledoubt, undercertaincircumstances,such evidencewould
not, in the opinion of the authors,raise an unreasonabledoubt.

Abrams, S. A Polygraph Handbook For Attorneys. Toronto: D.C. Heath and Co., 1977.
Abrams, S. A response to Lykken on the polygraph. American Psychologist, 1975, 30, 709-711.
Barnett, F.J. How do a jury view polygraph examination results?Polygraph, 1973, 2,275-277.
Carlson, S.C., Passano, M.S., and Jamuzzo, J.A. The effect of lie detector evidence on jury deliberations:
An empirical study. Journal of Police Science and Administration, 1977, 5, 148-154.
Doob, A.N., and Cavoukian, A. The effect of the revoking of bail: R. v. Demeter. Criminal Law Quarterly,
1977, 19, 196-203.
Forkosoch, M.D. The lie detector and the courts. New York University Law Quarterly Review, 1939, 16,
Horvath, F.S. The effect of selected variables on interpretation of polygraph records. Journal of Applied
Psychology, 1977, 62, 127-136.
Horvath, F.S., and Reid, J.E. The reliability of polygraph examiner diagnosis of truth and deception.
Journal of Criminal Law, Criminology, and Police Science, 1971, 62, 276-281.
Kalven, H., and Zeisel, H. The American Jury. Boston: Little Brown and Co., 1966.
Koffler, J.H. The lie-detector: A critical appraisal of the technique as a potential undermining factor in the
judicial process. New York Law Forum, 1957, 3, 123-158.
Loftus, E. Reconstructing memory: The incredible eyewitness. Jurimetrics Journal, 1975, 15, 188.
Lykken, D.T. Psychology and the lie detector industry. American Psychologist, 1974,29,725-739.
Lykken, D.T. The psychopath and the lie detector. Psychophysiology, 1978, 15, 137-142.
Lykken, D.T. The detection of deception. Psychological Bulletin, 1979, 86, 47-53.
Phillion vs the Queen. Canadian Criminal Cases, (2nd edition), 1978, 33,542.
Podlesny, J.A. and Raskin, D.C. Physiological measures and the detection of deception. Psychological
Bulletin, 1977, 84, 782-799.
Podlesny, J.A., and Raskin, D.C. Effectiveness of techniques and physiological measures in the detection of
deception. Psychophysiology, 1978, 15, 344-359.
Raskin, D.C. Scientific assessment of the accuracy of detection of deception: A reply to Lykken.
Psychophysiology, 1978, 15, 143-147.
Raskin, D.C., and Hare, R.D. Psychopathy and detection of deception in a prison population.
Psychophysiology, 1978, 15, 126-136.
Raskin, D.C., and Podlesny, J.A. Truth and deception: A reply to Lykken. Psychological Bulletin, 1979,
86, 54-59.
Regina vs Phillion. Canadian Criminal Cases, (2nd edition), 1973, 10, 562.
Regina vs Phillion. Canadian Criminal Cases, (2nd edition), 1975, 20, 191.
Regina vs Wong. Canadian Criminal Cases, (2nd edition), 1977, 33, 515.
Silverberg, B.A. A few kind words in defense of the polygraph. Canadian Lawyer, 1979, 6, 6.
Tarlow, B. Admissibility of polygraph evidence in 1975: An aid in determining credibility in a perjury-
plagued system. Hastings Law Journal, 1975, 26, 917-974.
Widacki, J., and Horvath, F.S. An experimental investigation of the relative validity and utility of the
polygraph technique and three other common methods of criminal investigation. Journal of Forensic
Sciences, 1978, 23, 596-601.

This content downloaded from on Wed, 10 Feb 2016 10:29:45 UTC
All use subject to JSTOR Terms and Conditions