Delft University of Technology
The evaluation of fingermarks given activity level propositions
de Ronde, Anouk; Kokshoorn, Bas; de Poot, Christianne J.; de Puit, Marcel
DOI
10.1016/j.forsciint.2019.109904
Publication date
2019
Document Version
Final published version
Published in
Forensic Science International
Citation (APA)
de Ronde, A., Kokshoorn, B., de Poot, C. J., & de Puit, M. (2019). The evaluation of fingermarks given
activity level propositions. Forensic Science International, 302, [109904].
https://doi.org/10.1016/j.forsciint.2019.109904
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy
Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
Original
Research
Paper
The
evaluation
of
fingermarks
given
activity
level
propositions
Anouk
de
Ronde
a,b,c,*
,
Bas
Kokshoorn
c,
Christianne
J.
de
Poot
a,b,d,
Marcel
de
Puit
c,eaAmsterdamUniversityofAppliedSciences,Weesperzijde190,1097DZAmsterdam,TheNetherlands bVUUniversityAmsterdam,DeBoelelaan1105,1081HVAmsterdam,TheNetherlands
c
NetherlandsForensicInstitute,LaanvanYpenburg6,2497GBTheHague,TheNetherlands
d
PoliceAcademyoftheNetherlands,P.O.Box348,7301BBApeldoorn,TheNetherlands
e
DelftUniversityofTechnology,VanderMaasweg9,2629HZDelft,TheNetherlands
ARTICLE INFO Articlehistory: Received3May2019
Receivedinrevisedform2July2019 Accepted9July2019
Availableonline30July2019 Keywords: Touchtraces Evidenceinterpretation Activity Bayesiannetwork ABSTRACT
Fingermarksarehighlyrelevantincriminalinvestigationsforindividualizationpurposes.Insomecases, thequestionincourtchangesfrom‘Whoisthesourceofthefingermarks?’to‘Howdidthefingermark enduponthesurface?’.Inthispaper,weexploretheevaluationoffingermarksgivenactivitylevel propositions byusingBayesian networks. Thevariables that provide informationon activitylevel questionsforfingermarksareidentifiedandtheircurrentstateofknowledgewithregardstofingermarks isdiscussed.Weidentifiedthevariablestransfer,persistency,recovery,backgroundfingermarks,location ofthefingermarks,directionofthefingermarks,theareaoffrictionridgeskinthatleftthemarkand pressuredistortionsasvariablesthatmayprovideinformationonhowafingermarkendedupona surface.Usingthreecaseexamples,weshowhowBayesiannetworkscanbeusedfortheevaluationof fingermarksgivenactivitylevelpropositions.
©2019PublishedbyElsevierB.V.
1.Introduction
Fingermarksplayanimportantroleinforensicscience.Based
ontheassumptionthateveryindividualholdsauniquepatternof
frictionridgeskin ontheirhands, this pattern canbe usedfor
identification.Bydeterminingthesourceofthefingermark,alink
betweenthedonorandacrimescenecanbeestablished.Thereisa
wealthofresearchonthevisualizationof latentfingerprintsin
ordertoenhancethefrictionridge patternforindividualization
purposes[1,2].Whilethistypeofresearchisveryvaluableforthe
individualizationofthesourceofatrace,thefingermarkitselfmay
notunequivocallybeattributedtoacriminalactivity.
An important question that often comes up in court cases
regardingforensicevidenceistodeterminehoworwhenatrace
was deposited.Consider the following case example;a woman
calls thepolicetoreport thatthere hasbeena burglaryinher
apartment.Thepolicefindfourfingermarksontherailingofthe
balcony, which leads to the assumption that the perpetrator
enteredtheapartmentviathebalcony.Throughadatabasesearch,
amatchisfoundwithasuspect,whoisanacquaintanceof the
woman. The suspect claims that, instead of an unauthorized
intrusionviathebalcony,hevisitedthewomanaweekearlierand
smokedacigaretteonthebalconywhileleaningontherailing.In
cases likethis, thequestionatstake changesfrom ‘Whois the
sourceofthefingermarks?’to‘Whatactivityledtothedeposition
ofthefingermarks?’,whichrequiresadifferentassessmentofthe
findings.
When investigating forensic evidence, a forensic scientist
formulates a set of propositions, usually representing the
prosecution and thedefense propositions.Cook, Evett, Jackson,
Jones and Lambert [3] propose three classes of propositions:
sourcelevel,activitylevelandoffencelevelpropositions.Inthe
balconycaseexample,theinvestigationshifts fromdetermining
thesourceofthefingermarkstoaddressingtheactivitythattook
place.IntheforensicexpertisefieldsofDNA,fibres,glass,paintand
gunshotresidues,evaluationoftheevidencegivenactivitylevel
propositions is already being studied [4]. However, for
finger-marks,thistopicisnotyetexplored.
Therearemanyvariablesthatmayprovideinformationonhow
a fingermark was deposited on a surface. In the balcony case
example,wherethequestionnowiswhetherthesuspectclimbed
thebalconyorthesuspectsmokedacigaretteonthebalconyand
leaned on the railing, variables such as the location of
thefingermarks,andthedirectionofthefingermarksmayprovide
information on the activity that took place. In general, the
interpretation of evidence at activity level requires
more contextual information [3]. When multiple variables
* Correspondingauthorat:AmsterdamUniversityofAppliedSciences, Weesperzijde190,1097DZAmsterdam,TheNetherlands.
E-mailaddresses:a.de.ronde2@hva.nl,a.de.ronde@nfi.nl(A.deRonde). http://dx.doi.org/10.1016/j.forsciint.2019.109904
0379-0738/©2019PublishedbyElsevierB.V.
ContentslistsavailableatScienceDirect
Forensic
Science
International
influencetheinterpretationoftheevidence,itcanbedifficultto
taketheirdependenciesintoaccountina directcalculationofa
likelihoodratio[5].
A methodthat iscommonlyusedforcaseswhereadditional
factorsplayaroleisaBayesiannetwork.ABayesiannetworkisa
graphicalrepresentationofamathematicalmodelwhichcanbe
usedtoevaluatethefindings,particularlyifthereisadependency
betweenrelevant variables[4]. A Bayesian network consistsof
nodes,directedarcsandprobabilityassignmentsofthenodes.It
can for instance be used to computea likelihood ratio of the
evidence given the prosecution proposition and the defense
proposition,basedonallvariablesthatareconsideredrelevantin
theinterpretationoftheevidence.ThismakesBayesiannetworks
anappropriatemethodtoevaluateevidencegivenpropositionsat
activitylevelwithinthefieldofforensicscience.AlthoughBayesian
networks have been proposed to interpret fingermarks given
sourcelevelpropositions[6],theyhavenotbeenusedtoevaluate
fingermarksgivenactivitylevelpropositions.
In thispaper,we describeaframeworkfortheevaluationof
fingermarks given activity level propositions using Bayesian
networks.Wediscussthevariablesthatprovideinformation on
fingermarksatactivitylevel,followedbythreecaseexamplesfor
whichBayesiannetworksarecreated.Weultimatelyelaborateon
possibledirectionsforfurtherresearchonthistopicsuchthatthe
proposedframeworkcouldbeoptimallyappliedincasework.
2.Relevantvariables
Inthissection,weexplorethevariablesthatprovide
informa-tiononfingermarkswithregardstoactivitylevelpropositions.We
donotdiscussvariablesrelatedtosourcelevelpropositionssince
determiningthedonorofafingermarkisconsideredoutsidethe
scopeofthisstudy.Furthermore,weassumedthatifafingermark
ispresent,thedonoractuallytouchedtheitem.1Touchingasurface
canbeseen as anactivity in itself,and therefore activity level
propositionsmaydisputewhetherthesurfaceisactuallytouched
orthefingermarkisaresultofforgery[1].Anotherdisputemay
focusonthecircumstancesofhowthefingermarkisrecovered,for
instancewhenthereareissueswiththechainofcustody[7].These
types of propositions are considered outside thescope of this
paper by assuming the surface is actually touched when a
fingermarkispresent.
Wedividedtherelevanteventsthatprovideinformationonthe
activitythatledtodepositionofthefingermarksintwogroupsof
variables:‘fingermarkformationprocess’,and‘mannerof
deposi-tion’. The group ‘fingermark formation process’ represents the
factorsthat relatetotherequirementsof fingermarkformation,
visualizationandrecovery.Thevariablesidentifiedinthisgroup
arethetransfer,persistenceandrecoveryoffingermarksandthe
backgroundlevelsoffingermarksalreadypresentonanitem.The
group‘mannerofdeposition’representsthefactorsthatrelateto
howthedonordepositedthefingermark.Thevariablesidentified
inthisgrouparethepositionofthehandduringplacement,the
locationofthefingermarks,areaoffrictionridgeskinthatleftthe
mark,thedirectionofthefingermarksandthepressureappliedto
thesurfaceduringdeposition.
2.1.Fingermarkformationprocess
2.1.1.Transfer
Aconsequenceofanactivitymaybethetransferofmaterialtoa
surfacebyafinger,creatingafingermark.Untilnow,researchon
thetransferoffingermarksfocusedmostlyonthecompositionof
the residue for the purpose of enhancing the quality of the
fingermarkforindividualizationatsourcelevel[8].However,the
guidelines of theENFSI [9] show that transfer is animportant
variable to consider when looking at the scientific findings in
relationtoactivities.
Fingermarks have advantages over other types of forensic
evidence.Fingermarksareconsideredtobeaproofofcontactdue
toadirecttransferoftheridgedetailtoasurface.Furthermore,
fingermarkscannottransferindirectlyviasurfacesorindividuals
unlessgreateffortismade[10].Secondaryorfurthertransferof
fingermarksisgenerallynottakenintoaccount(pleasenotethe
exception of fingermarks on tape [11]). These are important
advantagesoverDNA,sinceDNAcantransferindirectlyandeven
retransfer from onelocation toanother [12]. Althoughindirect
transferisgenerallynotapplicabletofingermarks,transferisstill
animportantvariabletoconsidersincetheprobabilityoftransfer
ofafingermarkmaydifferbetweenactivities.
The transfer of fingermarks depends on several factors: the
nature of the surface, the deposition conditions and donor
characteristics [8,13,14]. The deposition conditions such as
pressure and duration of contact may vary between activities,
and this may result in different transfer probabilities. If the
pressureofthehandonthesurfaceishigher,theprobabilityof
transfermightbehigher[13].Thepropositionsoftheprosecution
andthedefensemaysuggestdifferentlevelsofpressureneededto
conduct the proposed activities, leading to the assignment of
different transfer probabilities. This is also true for other
depositionconditions,whichmaketheobservedtransfer(orthe
absencethereof)more orlessprobablegivendifferent
proposi-tions.However,thedevelopmentandrecoveryoffingermarksona
surface depend on more than the mechanisms of transfer;
variables such as persistence and recovery also influence the
probabilityofrecoveringfingermarks.
2.1.2.Persistence
Afingermarkmaynotberecoveredinthesameconditionasit
was deposited. This is due to degradation, the process during
which the initial composition of a fingermark changes after
deposition[8].Degradationwilloccurfromthetimethe
finger-markhasbeendeposited,tothesubsequentevidencerecoveryand
mayaffectthepersistenceofafingermark.Thedegradationofa
fingermarkisinfluencedbythe‘triangleofinteraction’,consisting
of the fingermark composition, the nature of the surface and
environmentalconditions[2].Forthenatureofthesurfaceitis
knownthatfingermarkcompoundsmaybeabsorbedbysurfacesof
porousmaterial,whereastheystayonthesurfaceofnon-porous
materials.Thissurfaceinteractionmayinfluencethedegradation
ofthefingermarks[15].Furthermore,environmentalfactorslike
temperature,light, humidity and air circulationhave shown to
influencethedegradationoffingermarksovertime[14].
It isgenerallynotexpectedthat thenatureof thesurfaceis
disputedbetweenactivitylevelpropositionssincethesamesetof
fingermarks on the same item is questioned under both
propositions(unlessthere isanissuewiththechain-of-custody
[7]).However,environmentalconditionsmayvarybetweenapair
of activity level propositions for fingermarks, for example, if
propositionsdisputethemomentwhenthefingermarkisleftand
thus thetime interval betweenthemoment of depositionand
recovery. During that time interval, the fingermarks could be
subjectedtodifferentenvironmentalconditions.Inthatcase,the
factorpersistenceplaysasignificantrole.
2.1.3.Recovery
Aftertransfertoandpersistenceonasurface,thefingermark
mustbedetectedandrecoveredfromthecrimescene.Thisprocess
1
Onacrimescene,fingermarkscanbefoundonitemsandfixedsurfaces.Inthis article,weusethetermitemforboth,unlessfurtherspecified.
isdescribedbythevariablerecovery.Fingermarkscanbelatent,
meaning that they must be visualized with the use of an
enhancementtechnique.Severalfactorsinfluencethesuccessrate
ofthedetectionofafingermark. Thesensitivityoftheavailable
methodstovisualize fingermarksvaries [16],meaning thatnot
every technique has the same success rate. Furthermore, an
incorrect choice of technique, an incorrect application of a
technique or applying multiple techniques in thewrong order
can result in lower success rates of finding a fingermark [17].
Anotherfactorinfluencingtherecoveryprobabilityistargetingof
the correct location. Fingermarks could be missed by a wrong
selectionoflocationstosampleonthecrimescene,resultingina
different probability torecover fingermarks. Other factors that
impactontheprobabilityofrecoveryarethelevelofbackground
marksthat arealready present, and the criteria established to
determinewhetherafingermarkissuitableforindividualization.
Forexample,ifpartial fingermarksarepresent, thesewillmost
likelynot berecoveredif theyarenotof valueforcomparison.
However,whenthequestioniswhetherthesuspectworegloves,
thepresenceofthesepartialfingermarksmayverywellinfluence
the interpretation at activitylevel. As a result, the probability
to recover fingermarks may vary between the activity level
propositionsatstake.
2.1.4.Combinationoftransfer,persistenceandrecovery
Allthreevariablestransfer,persistenceandrecoveryinfluence
the probability of the findings separately, but they cannot be
clearly separated. If no fingermark is recovered, it does not
automatically mean that the fingermark was not present
(transfer).Thefingermark couldhavebeen degradedsuchthat
visualizationwasnotpossible(persistence),thechosen
enhance-ment technique could have been unsuccessful(recovery) or it
maybetheresultofa combinationof thesefactors.Therefore,
thesevariablesareoftentakentogetherandasingleprobabilityis
assignedtothefindings.
2.1.5.Backgroundfingermarks
Thereareoftenalreadyfingermarkspresentonitemsthatare
unrelated to the activities at stake. This means that the
fingermarkscouldhavealreadybeenpresentontheitembefore
the alleged activity took place or may have ended up onthe
surfaceafter theallegedactivitiestookplace.Fingermarksthat
aretransferredtothesurfacebyactionsunrelatedtotheactivities
atstakeareconsideredasbackgroundfingermarks.Consider,for
example,thattheissueiswhetherasuspectstabbedthevictim
withaknifeorthatanunknownpersonstabbedthevictimwith
theknife.Saywefindfingermarksofthesuspectonthehandle,as
wellassomefingermarksofoneormoreunknownindividuals.
Nowthe weight of theevidence given these two propositions
woulddependontherelationthatthesuspecthaswiththeitem
(e.g. could he have handled the knife prior to or after the
incident?),butalsoontheprobabilitythatwefindbackground
fingermarksonthehandleofthisspecificknife.Iftheknifewas
cleanedrecently,thatprobabilitymaybelowandtherecoveryof
fingermarksofanunknownindividualmaysupportthesuspect's
proposition.However,ifwehaveahighexpectationofrecovering
backgroundfingermarks(forinstancebecausetheknifeisnota
personalitemandwasincommonuse)theobservedfingermarks
of unknown individual(s) may be neutral towards the two
propositions. The probability thatthese unknown fingermarks
belongtobackground levelsoffingermarksontheitemshould
thereforebetakenintoconsideration.Duringinvestigation,itis
therefore important to consider the general activities that
occurred priorto orafter the allegedactivities that mayhave
resultedinfingermarksontheitem.
2.2.Mannerofdeposition
2.2.1.Positionofthehandandfingersduringdeposition
Thewayinwhichthefingermarksaredepositedonasurface
depends on the positioning of the hand and fingers during
deposition.Thepositionofthehandandfingersonanitemmay
differbetweenactivities,whichisdeterminedbythepurposeof
the activity, theanatomy of the humanbody and thephysical
characteristicsoftheitem.
The anatomy of the human body causes restrictions in
movements ofthelimbs.Duetotheserestrictions,thepossible
positions of the hand and fingers on an item are limited. The
physicalcharacteristicsoftheitemalsoinfluencethepositionof
thehandandfingersonanitem.Thesecharacteristicsincludesize,
weight,shape,structure,typeofmaterial,itsfunctionetc.Consider
thatsomeonegraspsaknifeforstabbing:heorshemostlikely
grabstheknifeatthehandleduetotheshapeandstructureofthe
knife. The physical characteristics of the handle of the knife
influence thepositioning of the hand and fingers, as may the
purposeoftheactivity:cuttingapieceofbreadversusstabbing
mayforinstanceaffectthewaytheknifeisheld.
Sincethemovements,thephysicalcharacteristicsoftheitem
and the goal of the activity may differ between activities,the
positionofthehandandfingersprovidesinformation thatmay
assistinevaluatingthefindingsgivenactivitylevelpropositions.
Sinceitcanbedifficulttodescribethepositionofthehandand
fingersdirectly,wedescribethepositionofthehandandfingers
duringdepositionthroughfourvariables:locationofthe
finger-marks,directionofthefingermarks,partofthehandthatleftthe
fingermark,andpressure.
2.2.2.Locationofthefingermarks
The position of the hand and fingers on an item during
depositioninfluencesthelocationofthefingermarksontheitem.
deRonde,vanAken,dePuitanddePoot[18]designeda model
that can be used to analyzethe location of fingermarks on
2-dimensionalitemsgivendifferentactivities.Withtheuseofthis
model,pillowcasescouldbeseparatedinthetwoactivityclasses
smotheringandchanging,basedonthelocationofthefingermarks
onthepillowcases.Thisshowsthatthelocationoffingermarkson
anitemprovidesinformationontheactivitythatthedonorcarried
out,andisthereforeanimportantvariabletotakeintoaccount.
2.2.3.Directionofafingermark
Whentouchingasurface,thehandandfingersarepositionedin
a certain direction. This direction varies between different
activitiesandassuchmaybedistinctiveforparticularactivities.
Inthebalconycaseexample,thefingermarkdirectionasaresultof
climbing the balcony may be different from the fingermark
directionasaresultofleaningontherailing.Thevariabledirection
is used bycrime scene officers tomake inferences during the
investigationphaseonacrimescene.Anexampleofthisisthat
fingermarks found pointing inwards on theinside of a broken
windowframeareoftenconsideredtoberelatedtotheactivityof
climbingthroughawindowduringaburglary.However,thereare
nostudiesthatreportonthedirectionoffingermarksinrelationto
activities. Theprobability tofindacertain fingermarkdirection
underthedifferentpropositionsmayprovideinformationonthe
activitylevel.
2.2.4.Areaoffrictionridgeskin
Differentactivitiesrequiretheuseofdifferentpartsofthehand
andthereforetheareaoffrictionridgeskinthatleftafingermark
mayprovideinformationontheactivity.Considerthebalconycase
example: itmaybemore probabletorecovera complete palm
the suspect simply touched the railing while standing on the
balcony.Theareaoffrictionridgeskinthatleftthemarkcanbe
determinedwhenthedonorofthefingermarkisknown.Incases
where a suspect or a corresponding reference print is absent,
determiningtheareathatlefttheprintmaybedifficult.
Although recent research has focused on determining
whetheritwasa left-hand or a right-handthatdeposited an
individualfingerprint [19–21], assigning a specific finger to a
fingermark is still a topic for further research. Nevertheless,
forensic examiners are trained to nominate corresponding
fingerstofingermarks based onthe size,patterntype, shape,
etc.Thisinformationmightbeveryvaluablefortheevaluation
offingermarksgivenactivitylevelpropositions.Ifa likelihood
ratio can be determined on whether a recovered fingermark
comesfrom a specific finger, or comes from another area of
friction ridge skin, this information can be used in the
evaluationof thefindings.
2.2.5.Pressure
Whenfrictionridgeskintouchesasurface,theshapeoftheskin
changesasaresultofthepressureappliedonthesurfaceandthe
pliabilityoftheskin.Maceo[22]identifiestwotypesofpressureof
afingeronasurface:verticalpressureandhorizontalpressure.An
increasedverticalpressureresultsinmorepointsofcontactwith
the surface, causing a broader fingermark [23]. Furthermore,
verticalpressureaffectsthewidthoftheridgesandthefurrowsina
fingermark[24].Asaresult,thesizeofafingermarkandthewidth
oftheridgesinafingermarkmayprovideinformationaboutthe
verticalpressureapplied.However,weexpectthatitwillbevery
difficulttodeterminetheverticalpressureappliedtoasurfaceby
justlookingatthefingermark,sincethesizeofafingermark,the
widthoftheridgesandtheconditionoftheskinvariesgreatly
betweendonors.
Pressure inthe horizontalplanecausesdeformationof the
skinthat mayresult ina distortion of thefingermarks in the
formofsmearsorswipes[22].Thispressuredistortionisoften
directional,andthedistortionseldommovesintwodirections
[22,24].Studyingthesedirectionaldistortions ina fingermark
canbeof greatervalueforthe interpretationat activitylevel.
Theprobabilityofdetectingapressuredistortioninaparticular
directionmaybedifferentfortwoactivitiesandthis
informa-tioncanbeusedintheassessment.Anotherpossibilityisthat
someactivitiesmayalwaysresultindistortedfingermarks.Ifthe
probability to obtain a distorted fingermark differs for two
activities,thisinformationmaybeofgreatvaluefortheactivity
levelinterpretation.
3.Bayesiannetworkconstruction
Withthevariablesidentified,weshowtheimplementationof
theseinaBayesiannetwork.Inthispaper,wefocusonfingermark
gripspresent onanitem. Bya grip, werefer toa collectionof
fingermarksforwhichitisassumedtheyareleftinoneandthe
sameplacementofthehand.Thismeanstheconsideredmarkscan
varyfromonefingermarktoacompletehandmark,althoughthey
originatefromone andthesamehandand bedeposited atthe
same time. In this paper, we assume that the source of the
fingermarks is identified or unknown. Recent literature on
fingermarksatsourcelevelfocusonamoreprobabilisticapproach
to present the evidential strength of a match [1,25]. The
implementationofthis probabilisticsourcelevelinformationin
Bayesiannetworksisconsideredoutsidethescopeofthispaper;
wereferthereadertoTaroni,Biedermann,Bozza,Garbolinoand
Aitken[4].
WebuiltthreedifferentBayesiannetworks,eachbased ona
versionofthebalconycaseexampledescribedintheintroduction
ofthispaper.Inthefirstcaseexample,onegripisrecoveredonthe
railing and it is questioned whether the suspect climbed the
balconyorleanedonthebalcony.Thesecondcaseexamplefocuses
onthequestionofwhetherthesuspectclimbedthebalcony or
someoneelseclimbedthebalcony.Inthefinalcaseexample,the
implementationof multiple gripsis discussedfor thequestion
whetherthesuspectclimbedthebalconyorsomeoneelseclimbed
the balcony. All three networks werebuilt using the software
Hugin (version 8.6)2 and can be found in the supplementary
material.Forthepurposeofillustration,weaddedsomefictional
probabilities in the network for the first case example. The
probabilitiesusedinthisexamplearesolelybasedoninformed
judgement of the authors, and are not based on any scientific
experimentsorpublisheddata.
Becausethepurposeofthispaperistoshowtheconstructionof
Bayesian networksfor theevaluationof fingermarks atactivity
level,wedonotelaborateonhowthevariablescanbeobjectively
measured, nor do we aim to assign exact probabilities to the
network.Themainfocuswillbeontheconsiderationsaforensic
scientisthastomakewhencreatingaBayesiannetworktoevaluate
fingermarksgivenactivitylevelpropositions.Inthediscussion,we
willelaborateonhowprobabilitiescanbeassignedtothenodes
andweproposetopicsforfurtherresearchthatwillgivesubstance
totheseprobabilityestimations.
Fig.1.Bayesiannetworkfortheevaluationoffingermarksatactivitylevelincaseexample1.
2
3.1.Caseexample1:Natureoftheactivitydisputed
3.1.1.Backgroundinformation
Consider the balcony case example we described in the
introduction.Thepolicefoundagripoffingermarksontherailing
ofthebalcony,whichleadstotheassumptionthattheperpetrator
enteredtheapartmentviathebalcony.Thesuspect,foundthrough
adatabasesearch,claimsthathisfingerprintsarenotleftonthe
balcony due toan unauthorized intrusion via thebalcony, but
duringalegalvisittothewomanwhenleaningontherailingwhile
smokingacigarette.Thedisputeofthedefense isaimedatthe
natureoftheactivity[26],resultinginthefollowingactivitylevel
propositions:
Hp:Sclimbedthebalconyanddidnotleanontherailing.
Hd:Sleanedontherailinganddidnotclimbthebalcony.
FollowingtheprocessdescribedbyTaylor,Biedermann,Hicks
andChampod[27],weconstructedtheBayesiannetworkshownin
Fig. 1, using the same colouring scheme. Sections 3.1.2–3.1.7
describethenodes,thedependenciesandtheconsiderationsfor
thestatesofeachnode.Weconstructedthisnetworktoevaluatea
positiveresult,e.g.afingermarkfoundonasurface.Ifnomarksare
recovered,theproposedBayesiannetworkwouldonlyconsistof
nodes [1] to [5], since determining the findings [6] to [12] is
impossible.
3.1.2.Node[1]propositions
TheblacknodePropositionsinFig. 1representsthemainactivity
level propositions. This node has two states, Hp and Hd,
representingrespectivelythepropositionoftheprosecutionand
the defense. Assignment of the prior probabilities is generally
outsidethedomainoftheforensicscientist.Forthepurposeofthis
example, we have assigned equal prior probabilities to each
proposition(Table1).
3.1.3.Nodes[2]Sclimbedthebalconyand[3]Sleanedontherailing
Thepropositionalnodeimpliestwoactivitynodes:Sclimbedthe
balconyand S leaned on the railing, denotedblue in Fig.1.We
definedthestates‘true’and‘false’tobothnodes.Theprobabilities
ofthestatesofnodeSclimbedthebalcony(Table2)andnodeS
leanedontherailing(Table3)areconditionedonthestatesofnode
propositions.Table2showsthatgiventhatHpistrue,thenodeS
climbed balcony is true with probability p=1 and false with
probabilityp=0.IfHdistrue,thenodeSclimbedthebalconyistrue
with probability p=0 and false with probability p=1. For the
probabilitytableofnodeSleanedontherailingshowninTable3,
thereverseholds.
3.1.4.Nodes[4]FingermarksSthroughclimbingand[5]FingermarksS
throughleaning
Asa resultoftheactivities climbing orleaning,fingermarks
endedupontherailing.InFig.1,themechanismsbywhichthe
activitiesleadtothefindingsarerepresentedbytheyellownodes
FingermarksSthroughclimbingandFingermarksSthroughleaning,
both with states ‘true’ and ‘false’. Within these nodes, the
combined probabilities of transfer, persistence and recoveryof
the fingermarks as a result of the proposed activities are
considered.
Table4 showstheconditionalprobabilitytableforthenode
FingermarksSthroughclimbing.Thisnodedependsontheactivity
nodeSclimbedthebalcony.GiventhatSclimbedthebalconyistrue,
Padenotestheprobabilitytoobtainfingermarksgiventheactivity
climbing. This incorporates the probabilities for transfer, the
persistenceandtherecoveryoffingermarksontherailingthrough
climbing. From the fact that the states of nodes are mutually
exclusiveandexhaustivefollowsthattheprobabilitythatthereis
no transfer, persistence and recovery of fingermarks through
climbing is equal to 1Pa. The probability table for the node
Fingermarksthroughleaningisconstructedinanequalmanner.
3.1.5.Node[6]direction
Oneaspectwecanobservefromtherecoveredfingermarksis
theirdirection.Thenodeforthisvariableisshownbythecolour
red in Fig. 1. Before the direction of the fingermarks can be
determined,thetransfer,persistenceandrecoveryofthe
finger-markshadtobesuccessful,whichmeansthatthenodeDirectionin
thenetworkisdependentontheprobabilitytoobtainfingermarks
undertheallegedactivities.ThisisshowninFig.1bydrawingan
arrowfromFingermarksthroughclimbingandFingermarksthrough
leaningtothenodeDirection.
There aremultiple options todefine thestates of thenode
Direction;theoretically,everyanglecouldbeaseparatestate.Inour
caseexample,wechosetodefinetwostatesforthedirectionofthe
fingermarks:thefingermarksarepointinginwards(tothehouse)
andthefingermarksarepointingoutwards(awayfromthehouse).
TheconditionalprobabilitytableofthenodeDirectionisshownin
Table 5. Assume that fingermarks through climbing is true and
Table1
Priorprobabilitytableforthenode[1]PropositionsinFig.1.
Propositions Probability
Hp:Sclimbedthebalconyanddidnotleanontherailing. 0.5
Hd:Sleanedontherailinganddidnotclimbthebalcony. 0.5
Table2
Conditionalprobabilitytableforthenode[2]SclimbedthebalconyinFig.1.
Propositions Hp Hd
Sclimbedthebalcony:
True 1 0
False 0 1
Table3
Conditionalprobabilitytableforthenode[3]SleanedontherailinginFig.1.
Propositions Hp Hd
Sleanedontherailing:
True 0 1
False 1 0
Table4
Conditionalprobabilitytableforthenode[4]FingermarksSthroughclimbingin Fig.1.
Sclimbedthebalcony True False
Fingermarksthroughclimbing:
True Pa 0
False 1Pa 1
Table5
Conditionalprobabilitytableforthenode[6]DirectioninFig.1.
Fingermarksthroughclimbing True False
Fingermarksthroughleaning True False True False
Directionoffingermarks:
Inwards * Pc1 Pd1 *
Outwards * 1Pc1 1Pd1 *
(*)denotesthefactthattheseprobabilitiesrepresentsituationswhichwillnot occurbecausetheactivitiesclimbingandleaningaremutuallyexclusiveinour example, and the network is not constructed to evaluate the absence of fingermarks.
fingermarksthroughleaningisfalse,theprobabilitytofindinward
pointingfingermarksisdenotedbyPc1.
3.1.6.Node[7]location
SimilartothenodeDirection,thenodeLocationisdependenton
thenodesFingermarksthroughclimbingandFingermarksthrough
leaning,asshownbythearrowsbetweenthesenodesandthenode
LocationinFig.1.Inourcaseexample,weassumethatthereisno
directdependencybetweenthevariableLocationandthevariable
Direction.Theprobabilitytofindthefingermarksonaparticular
locationontherailingdoesnotdirectlydependonwhetherthe
fingermarksareplacedinwardsoroutwardsandviceversa;they
bothdirectlydependontheactivitythatiscarriedout.
Fig. 2 shows the top view of the balcony. During the
investigation,itwasdeterminedthattheonlywaytoclimbthe
balconyisviathedrainpipelocatedontheleftsideofthebalcony.
ForthestatesofthenodeLocation,wedecidedtodividetherailing
intofourareas:theleftbeam,themiddle/leftbeam(withplanter),
themiddle/right beamandthe rightbeam,asshown in Fig.3.
Again,therearemanywaystochoosethepossiblestates.Forthis
scenario,we considerdividingthe railinginto thesefourareas
appropriategiventhestructureandsetupofthebalcony.Theleft
sideisscreenedoffbythedoorwhenopen,theplantershieldsthe
railingandthefoursurfaceareasareapproximatelyequal.
TheprobabilitytableforthenodeLocationisshowninTable6.
Sincetherearefourpossiblestates,wedenotedtheprobabilitiesof
thestatesleft,left/middle,right/middleandrightincase
Finger-marksthroughclimbingistrueandFingermarksthroughleaningis
falsewithPe1;Pe2;Pe3and1 Pe1þPe2þPe3
.Theprobabilitiesin
caseFingermarksthroughclimbingisfalseandFingermarksthrough
leaningistruearedenotedwithPf1;Pf2;Pf3and1ðPf1þPf2þPf3Þ.
3.1.7.Node[8]areaoffrictionridgeskinwithsub-nodes[9]which
hand,[10]palm,[11]fingersand[12]thumb
Giventhatitisknownthatthesuspectleftthefingermarkson
the railing, the corresponding area of the hand that left the
fingermarkscanbedetermined.ThenodeAreaoffrictionridgeskin
withitssub-nodesWhichhand,Palm,FingersandThumbareusedto
incorporatethevariableareaof frictionridge skinthat leftthe
fingermarks,asdiscussedinSection2.2.4.
Inourcaseexample,wechosetodividethehandthatleftthe
fingermark(s)inthreeareas:thepalm,thefingersandthethumb.
WithinthenodesPalm,FingersandThumb,thepartofthehandthat
leftthemarkscanbespecified.Eachnodehastwopossiblestates:
‘true’and‘false’.Whetherthemarkscamefromtherightorleft
hand can be specified within the node Which hand, also with
possiblestates‘true’and‘false’.Allthesenodesareconnectedto
thesummarynodeAreaoffrictionridgeskin,thatcombinesallthe
information provided in the previous nodes. In this node, the
probabilityofallpossiblecombinationsofthestatesofthenodes
Whichhand,Palm,FingersandThumbissummarized.
In some cases, differentiation between each finger or even
betweenspecificareasonthehandmaybemoreappropriatesince
theprobabilityofoccurrenceofcertainareasmaydifferbetween
theallegedactivities.Adirectresultofdefiningsmallerareason
thehandisthatthenumberofstatesforthenodeAreaoffriction
ridgeskin increasessubstantially,sinceeach combinationofthe
specifiedareasforeachhandshouldbeassignedaprobability.For
example, dividingthehand intosix regions (fivefingers and a
palm)andaccountingforthepossibilitythattheleftortheright
hand is used, already results in 126 combinations. Assigning
probabilities toall theseseparate combinations may becomea
difficult task. Since in our case example, we expected the
probabilities to observe fingermarks of a specific finger to
differentiatebetweenclimbingandleaning,wechoosethethree
states‘palm’,‘fingers’and‘thumb’.Table7showstheprobability
tableforthenodeAreaoffrictionridgeskin.Fromthistable,wecan
observethatadifferentiationof3areasofthehandresultsin14
possiblestatestowhichprobabilitieshavetobeassigned,varying
from the probability to observe only the left-hand palm, to
observingthecombinationoftheright-hands’fingers,palmand
thumb.Wedidnottakeintoaccountcombinationsoftherightand
the left hand, since we limited our network to one grip of
fingermarksforwhichitisassumedthefingermarksaredeposited
byonehand.
3.2.Caseexample2:Actorthatcarriedouttheactivitydisputed
3.2.1.Backgroundinformation
Considerthesamescenarioasdescribedincaseexample1,but
insteadofclaimingthattheclimbingdidnottakeplace,thesuspect
claimsthatsomeoneelsemusthaveclimbedthebalcony.Hestates
thathevisitedtheapartmentaweekearlieroninvitationbythe
womanandsmokedacigaretteonthebalconywhileleaningonthe
railing.ThewomanconfirmstheinformationthatSvisitedaweek
earlier.Thedisputeofthedefenseisnowaimedattheactorofthe
activity[26],resultinginthefollowingactivitylevelpropositions
(definedassuchinnode[1]PropositionsintheBayesiannetwork
showninFig.4):
Hp:SclimbedthebalconyandSleanedontherailing.
Hd:UclimbedthebalconyandSleanedontherailing.
Thepolicestillfoundonlyonegripoffingermarks.However,
this situation is different from case example 1 since if the
fingermark grip belongs toS, the probabilitythat there are no
fingermarksfoundofanunknownindividualhavetobetakeninto
account.ThisresultedintheBayesiannetworkshowninFig.4.
3.2.2.Nodes[2]Uclimbedthebalcony,[3]Sclimbedthebalconyand
[4]Sleanedontherailing
Thepropositionsnowimplythreeactivities,whicharedefined
withthenodesUclimbedthebalcony,SclimbedthebalconyandS
leanedontherailing,eachwithstates‘true’and‘false’.Tables8–10
show the probability tables for these nodes. For example, in
Table8,giventhatHp:SclimbedthebalconyandSleanedonthe
railing istrue,the probabilityforthestate ‘true’of thenodeU
climbedthebalconyis0andtheprobabilityforthestate‘false’is1.
3.2.3.Nodes[6]FingermarksUthroughclimbing,[7]FingermarksS
throughclimbingand[8]FingermarksSthroughleaning
Thethreedifferentactivitieseachimplyadifferentprocessby
which fingermarksweredeposited andpersistedontherailing,
representedbythenodesFingermarksUthroughclimbing,
Finger-marksSthroughclimbingandFingermarksSthroughleaning.These
nodeshavethestates‘true’and‘false’andtheirprobabilitytables
Fig.2. Topviewofthebalconyinscenario1.
Fig.3.Thefourdifferentareasrepresentingthestatesofthenode‘Location’in Fig.1.
are similar to the probability table for the node Fingermarks
throughclimbingincaseexample1,showninTable4.
3.2.4.Node[5]backgroundfingermarksU
Incaseexample2,thereisanothermechanism possiblethat
needsto beconsidered: fingermarks of one or more unknown
personscouldalreadyhavebeenpresentpriortotheactivitiesthat
havetaken place.Thisis denoted bytherootnodeBackground
fingermarksU,denotedbythecolourgreyinFig.4,withstates‘true’
and ‘false’. Within this node, we consider the probability of
observing backgroundfingermarks ontherailingthat arenota
resultofthedisputedactivities.Incasenounknownfingermarks
werefoundbesidesthefingermarksofS,thebackgroundnodewill
beinstate‘false’withaprobabilityp=1.
3.2.5.Nodes[9]marksofUpresentand[10]marksofSpresent
Thissectionstillfocusesononegripoffingermarksdeposited
duringone handplacement,there areonlytwo optionsforthe
Table6
Conditionalprobabilitytableforthenode[7]LocationinFig.1.
Fingermarksthroughclimbing True False
Fingermarksthroughleaning True False True False
Locationoffingermarks:
Left * Pe1 Pf1 *
Middle/left * Pe2 Pf2 *
Middle/right * Pe3 Pf3 *
Right * 1ðPe1þPe2þPe3Þ 1ðPf1þPf2þPf3Þ *
(*)denotesthefactthattheseprobabilitiesrepresentsituationswhichwillnotoccurbecausetheactivitiesclimbingandleaningaremutuallyexclusiveinourexample,and thenetworkisnotconstructedtoevaluatetheabsenceoffingermarks.
Table7
Conditionalprobabilitytableforthenode[8]AreaoffrictionridgeskininFig.1.
Fingermarksthroughclimbing True False
Fingermarksthroughleaning True False True False
Areaoffrictionridgeskin:
Left–Palm * Pg1 Ph1 *
Left–Fingers * Pg2 Ph2 *
Left–Thumb * Pg3 Ph3 *
Left–Palm–Fingers * Pg4 Ph4 *
Left–Palm–Thumb * Pg5 Ph5 *
Left–Fingers–Thumb * Pg6 Ph6 *
Left–Palm–Fingers-Thumb * Pg7 Ph7 *
Right–Palm * Pg8 Ph8 *
Right–Fingers * Pg9 Ph9 *
Right–Thumb * Pg10 Ph10 *
Right–Palm–Fingers * Pg11 Ph11 *
Right–Palm–Thumb * Pg12 Ph12 *
Right–Fingers–Thumb * Pg13 Ph13 *
Right–Palm–Fingers-Thumb * 1ðPg1þþPg13Þ 1ðPh1þþPh13Þ *
(*)denotesthefactthattheseprobabilitiesrepresentsituationswhichwillnotoccurbecausetheactivitiesclimbingandleaningaremutuallyexclusiveinourexample,and thenetworkisnotconstructedtoevaluatetheabsenceoffingermarks.
sourceofthefingermarks:thefingermarksarefromanunknown
personUorthefingermarksarefromS,denotedbythefindings
nodesMarksofUpresentandMarksofSpresent.Bothnodeshave
states‘true’and‘false’.Thearrowbetweenthesenodesrepresents
thedependencybetweenthem:ifMarksofSpresentistrue,Marks
ofUpresentcannotbetrue.
TheprobabilitytablesforthenodesMarksofSpresentandMarks
ofUpresentareshowninTables11and12.ThenodeMarksofS
presentdependsonthetwonodesFingermarksSthroughclimbing
andFingermarksSthroughleaning.Table11showsthatifoneof
thesenodesisinstate‘true’,theprobabilitythattherearemarksof
Spresentis1.Ifbothofthesenodesareinstate‘false’,thereisa
probabilityof0thattherearemarksofSpresent.ThenodeMarksof
Upresentdependsonthreenodes:FingermarksUthroughclimbing,
BackgroundfingermarksUandMarksofSpresent.Table12shows
thatifthenodeMarksofSpresentistrue,theprobabilitythatthere
aremarksofUpresentisfalse.Thisisbecausewefocusononegrip
offingermarksleftduringoneplacement.
3.2.6.Findingnodes[11]to[17]
ThenodesDirection,Location,andAreaoffrictionridgeskinare
definedthesamewayasdescribedinpreviousSections3.1.5–3.1.7,
withanadditionalarrowfromthenodesBackgroundfingermarksU
andFingermarksUthroughclimbing.ThenodesWhichhand,Palm,
FingersandThumbaredefinedexactlythesamewayasdescribedin
Section 3.1.7.An exampleof the probabilitytable for thenode
DirectioninFig.4isshowninTable13.
3.3.Caseexample3:Multiplegrips
3.3.1.Backgroundinformation
Oftenthereismorethanonegripoffingermarksfoundonan
item. Supposethat in additiontothe firstgrip, anothergripis
foundontherailing.Again,thesuspectclaimsthathevisitedthe
apartmentaweekearlierandleanedontherailingofthebalcony
and this information is again confirmed by the woman. The
propositionsbroughtforwardbytheprosecutionandthedefense
arethesameasusedforcaseexample2:
Hp:SclimbedthebalconyandSleanedontherailing.
Hd:UclimbedthebalconyandSleanedontherailing.
Now the Bayesian network should account for two grips,
resultingintheBayesiannetworkshowninFig.5.
3.3.2.Structureofthenetwork
TheBayesiannetworkinFig.5consistsoffour‘modules’.The
networkstartswithapropositionnodePropositions[1],followed
by the nodes describing the alleged activities: U climbed the
balcony,[3]Sclimbedthebalconyand[4]Sleanedontherailing.
Table9
Conditionalprobabilitytableforthenode[3]SclimbedtherailinginFig.4.
Propositions Hp Hd
Sclimbedthebalcony:
True 1 0
False 0 1
Table10
Conditionalprobabilitytableforthenode[4]SleanedontherailinginFig.4.
Propositions Hp Hd
Sleanedontherailing:
True 1 1
False 0 0
Table11
Conditionalprobabilitytableforthenode[10]MarksofSpresentinFig.4.
FingermarksSthroughclimbing True False
FingermarksSthroughleaning True False True False
MarksofSpresent:
True 1 1 1 0
False 0 0 0 1
Table8
Conditionalprobabilitytableforthenode[2]UclimbedtherailinginFig.4.
Propositions Hp Hd
Uclimbedthebalcony:
True 0 1
False 1 0
Table12
Conditionalprobabilitytableforthenode[9]MarksofUpresentinFig.4.
FingermarkUthroughclimbing True False
BackgroundfingermarksU True False True False
MarksofSpresent True False True False True False True False
MarksofUpresent
True * * * 1 * 1 0 0
False * * * 0 * 0 1 1
(*)denotesthefactthattheseprobabilitiesrepresentsituationswhichwillnotoccurbecausetheactivitiesclimbingandleaningaremutuallyexclusiveinourexample,and thenetworkisnotconstructedtoevaluatetheabsenceoffingermarks.
Table13
Conditionalprobabilitytableforthenode[11]DirectioninFig.4.
BackgroundfingermarksU True False
FMUthroughclimbing True False True False
FMSthroughclimbing True False True False True False True False
FMSthroughleaning True False True False True False True False True False True False True False True False
Direction:
Inwards * * * * * * * Pi1 * * * Pj1 * Pk1 Pl2 *
Outwards * * * * * * * 1Pi1 * * * 1Pj1 * 1Pk1 1Pl2 *
(*)denotesthefactthattheseprobabilitiesrepresentsituationswhichwillnotoccurbecausetheactivitiesclimbingandleaningaremutuallyexclusiveinourexample,and thenetworkisnotconstructedtoevaluatetheabsenceoffingermarks.
Thesenodeshavethesamesetupasincaseexample2.Belowthese
nodesaretwonearlyidenticalmodulesthatrepresenttwodistinct
fingermarkgrips.Thefirstgripoffingermarksisdescribedbythe
nodesontheleft-handsideofthenetwork,indicatedby(1).The
secondgripoffingermarksisdescribedbythenodesindicatedby
(2).Betweenthesetwosub-networksisamoduleconsistingoffour
greennodesthatdescribedependenciesbetweenthetwotraces.
Weconsider conditional dependenciesbetweenthe two traces
basedonthelocationofthemarks,thedirectionofthemarksand
whetherornotthetwomarkswereleftbythesamehandsincethe
findingsmaybedependentonthesefactors.Weconsiderthem
conditionallyindependentfromthepropositions.Wechosethese
dependenciessince weconsiderthattheprobabilityofthetwo
marksbeingfromthesamedonorishigherwhentheyarefoundat
thesame location,havethesamedirectionandare leftbytwo
differenthands,thanifeitherlocationordirectiondiffer(where
locations within reach of both arms still have an increased
probabilityforthefingermarksbeingfromthesamesource).
Ifthetwogripsaredepositedduringthesameactivity(holding
therailingwithbothhandswhileclimbingorleaningontherail
withbothhands),therearetwooptionalsituations:thedeposition
ofthetwomarksisstrictlyconstrainedintime,e.g.theymusthave
beenplacedattheexactsamemomentduringthesameactivityor
thedepositionof thetwomarksislessconstrained intimeand
multipleinteractionsbetweenhands and therailingtook place
duringthesameactivity.Tobothsituations,itappliesthatifthe
two fingermark grips are found in close proximity, this will
influence the probability that they were left by the same
individual,regardlessoftheactivitiesdefinedinthepropositions
thatledtotheirdeposition.
Ifweassumethetwomarksarestrictlyconstrainedintimeand
wereleftthroughthesameactivity,giventhecasecircumstances,
thereisahighprobabilitythattheywillhavethesamedirection,
sinceitisunlikelytoplaceonehandinwardsandonehandoutwards
whencarryingoutthesameactivityinthesamemomentintime.
Furthermore,ifthetwomarkswereleftthroughthesameactivityat
thesametime,theycannothavebeenleftbythesamehand.
However,sinceboththeactivitiesleaningandclimbingarea
dynamic process, it is unlikely that this assumption holds. If
multipleinteractionsbetweenhandsandrailingmayhavetaken
place,itisnotunlikelytofindmultiplemarksofthesamehand
closetogether.Also,dependingonhowstrictorbroadtheactivities
aredefinedindynamicsandtime,itmaybeconsideredequally
probabletofindthemarkshavingthesamedirectionoradifferent
direction.Withaverybroaddefinitionandmultipleinteractions
withtherailingover extendedperiodsof time,only locationis
expectedtobeadependentfactorbetweenthetwomarks.
We haveaddedfournodesto thenetworkthatmodel these
dependencies.Node [31] Same direction? models whether both
markshavethesamedirectionornot(respectivelystate‘true’or
‘false’),andisdependentofthedirectionnodesforthetwoseparate
grips.Ifthedirectionofbothgripsisequal,thenodeSamedirection?
isinstatetruewithaprobabilityp=1.Otherwise,thenodeSame
direction?isinstatefalsewithaprobabilityp=1.Node[32]Same
location?modelswhetherbothmarkshavethesamelocation.The
statesofthisnodeconsistofallpossiblecombinationsofthestates
forthenodesLocation(1)andLocation(2),whichresultsin ten
combinations.IfLocation(1)isleftandLocation(2)isleft,thenode
Same location? is in state ‘left-left’ with a probability of p=1
Choosingfortwopossiblestates‘true’and‘false’isalsoapossibility.
However,in this case theproximityof two consecutive beams
cannotbetakenintoaccountin thenode[34]Samesource.The
dependencybetweentwohandsismodelledwithintheNode[33]
Samehand?,withstates‘true’and‘false’.IfWhichhand(1)andWhich
hand(2)arebothleft,thenodeSamehand?istruewithaprobability
ofp=1Thenode[34]Samesource?containsaprobabilitytablethat
holdstheprobabilitiesforthefingermarksbeingfromthesame
donorbasedontheirrespectivelocations,directionandleftorright
handsetting.Additionally,node[23]MarksofSpresent(2)isnow
dependentonthenode[34]Samesourceandnode[20]MarksofS
present(1)(inadditiontonodes[11]and[12]).
Thisnetworkcouldbeextendedtoanetworkthatallowsforthe
evaluationofmorethantwogripsoffingermarks,byconcatenating
multiplesub-networksinthesameway.Whenconstructingsucha
network,possiblenewdependenciesbetweenvariablesdescribing
different grips should be considered. A combined network
accounting for multiple gripsmakes a complete analysis of all
thefingermarkspresentonanitempossible.
4.Discussionandconclusion
Inthispaper,wehavedescribedaframeworkfortheevaluation
of fingermarksgivenactivitylevel propositionswiththe useof
Bayesiannetworks.Weprovidedanoverviewofthecurrentstate
of knowledge of the variables that provide information on
fingermarks given activity level propositions, followed by an
implementationof thesevariables ina Bayesian network using
threecaseexamples.Theresultingnetworksenablesthe
evalua-tion of (multiple) fingermark grips present on an item given
propositionsthatdisputetheactivitythatwascarriedoutorgiven
propositionsthatdisputetheactorthatcarriedouttheactivity.
TheBayesiannetworksproposedinthispapercouldfunctionas
basic networks for the evaluation of fingermarks, with the
possibilitytobemodifiedaccordingtospecificcasecircumstances.
Furthermore,partsofthenetworkmayfunctionasbuildingblocks
tocreatenewnetworksforitemsotherthanabalconyrailing,to
evaluate fingermark grips given activity level propositions.
AnotheradvantageofusingofBayesiannetworksisthatitmakes
theprocessofevaluationofthefindingsexplicit.Thenetworkcan
beusedasatooltodiscusstheselectedvariables,thedependencies
between them and the probabilities used, resulting in open
discussionsincourt.
Theprinciplesdiscussedinthispaperaremeanttobeusedasa
guidelinetohelpforensicscientistsmakewell-consideredchoices
dependingonthecaseathand.Theproposedlistofvariablesisa
recommendation: it depends onthe case circumstances which
variablesmaybeimportanttoconsider.Thechoiceofthestatesof
the variables also depends on the case circumstances, the
possibilities toobjectively measurethe possible states and the
feasibility of assigningprobabilities tothestates. These factors
needtobecarefullyconsideredwhenselectingthestatesofthe
nodes.Similarly,weproposeddependenciesbetweenthevariables
basedonourcaseexample,whichshouldbereconsideredwhen
applyingtheframeworktoadifferentcaseexample.
The final step to complete a Bayesian network is to assign
probabilitiestothenodes[28].AccordingtoTaylor,Kokshoornand
Biedermann[29],aforensicscientisthasanumberofoptionstodo
this(mentionedinorderofpreference):performexperimentsby
simulating the case circumstances, use values reported in
literature from studies using similar case circumstances and
outline the differences when reporting, consider a range of
reasonable values and examine the sensitivity of the LR (see
[30]),assignvaluesbasedontheexpert'sexperienceorknowledge,
or not carry out an evaluation. For fingermarks, the current
situationisthatevaluationsoffingermarksgivenactivitylevelare
notcarriedoutbyforensicexperts.Thisleavestheevaluationof
fingermarks given activity level propositions up to the court
although the forensic scientist has the specialized knowledge
regarding the variables that is required to properly assign
probabilities[29].
Inthefieldofforensicbiology,anincreasingbodyofliteratureis
transfer,persistenceandrecoveryofDNAinrelationtoactivities
(seefor example[31,32]).These studiesinvolve experimentsin
whichparticipantscarriedoutactivitiesthatresultedintouching
surfacesoritems,andfactorsliketransferandpersistencewere
evaluated in relation tothe activities performed. The study of
fingermarks in time and space would benefit from similar
experimentaldesigns.Experimentsintoprobabilitiesoftransfer,
persistence,recovery,direction,locationoffingermarks,orwhat
fingers are used when carrying out different activities with a
particularitemwouldhelpforensicscientiststoassign
probabili-tiestothesevariablesincases withsimilarcase circumstances.
Althoughtheobtainedprobabilitiesmaynotalways bedirectly
applicable to other cases, the experimental data may still
contributetoascientificknowledgebase[29]andmaycontribute
toabetterunderstandingofthegeneralmechanismsoffingermark
dynamics.
Other recommendations for further research are designing
methodstoobjectivelymeasureaspecificvariable.Forexample,
thereisnomethodavailabletoobjectivelymeasurethedirectionof
afingermarkonasurface.Anotherexampleisthevariabletransfer:
howdowemeasurethetransferofafingermarktoasurfaceasa
resultofanactivity? Nowadays,fingermarkscanbescored(for
example by the CAST scale [14]) to compare the quality for
individualizationpurposes.However,thequantityoffingermarks
transferredtoasurfacemayalsoprovideinformationonactivity
level. These examples show that for some variables describing
fingermarks at activity level, a clear definition or method to
measure the variable is required before the variables can be
describedbycasespecificexperiments.
Withthis paper, we want toinitiate a discussionabout the
evaluationoffingermarksgivenactivitylevelpropositions.Until
now,thistopichasbarelybeentouchedupon,possiblybecausethe
necessityisnotacknowledged.However,anevaluationof
finger-marksgivensourcelevelpropositionsdoesnotalwaysamountto
theactivity[9].Inthesecases,anevaluationofthefingermarks
givenactivitylevelpropositionscouldaffectthestrengthof the
evidencewithinthecasecircumstances.Wehopethispaperwill
leadtonewperspectivesonthistopicandstimulatesopportunities
forfurtherresearch.
Author'scontribution
Anouk deRonde: Conceptualization,Methodology, Software,
Formalanalysis,Investigation,Writing–Originaldraft,Writing–
Review and Editing, Visualization, Project administration. Bas
Kokshoorn: Conceptualization, Methodology, Software, Formal
analysis,Writing–ReviewandEditing,Visualization.Christianne
dePoot:Conceptualization, Methodology,Writing–Reviewand
Editing, Supervision, Funding Acquisition. Marcel de Puit:
Conceptualization, Methodology,Writing – Reviewand Editing,
Supervision,FundingAcquisition.
Conflictsofinterest
Nonedeclared.
Funding
This work was supportedby the RAAK-PRO funding of the
Foundation Innovation Alliance (SIA – Stichting Innovatie
Alliantie),researchgrantno.2014-01-124PRO.
Acknowledgment
WewouldliketothankCarolineGibbforhercommentsonan
earlierversionofthismanuscript.
AppendixA.Supplementarydata
Supplementary data associated with this article can be found, in the
onlineversion,athttps://doi.org/10.1016/j.forsciint.2019.109904.
References
[1]C.Champod,C.Lennard,P.Margot,M.Stoilovic,FingerprintsandOtherRidge SkinImpressions,2nded.,CRCPress,2016.
[2]S.M.Blaey,R.S.Croxton,M.dePuit,FingerprintDevelopmentTechniques. TheoryandApplication,JohnWiley&SonsLtd,2018.
[3]R. Cook, I.W. Evett, G. Jackson, P.J. Jones, J.A. Lambert, A hierarchy of propositions:decidingwhichleveltoaddressincasework,Sci.Just.38(1998) 231–239.
[4]F. Taroni,A. Biedermann, S.Bozza, P. Garbolino,C.G.G. Aitken, Bayesian NetworksforProbabilisticInferenceandDecisionAnalysisinForensicScience, 2nded.,JohnWiley&Sons,2014.
[5]C.G.G.Aitken,A.J.Gammerman,Probabilisticreasoninginevidential assess-ment,J.ForensicSci.Soc.29(1989)303–316.
[6]R.Haraksim,D.Meuwly,G.Doekhie,P.Vergeer,M.J.Sjerps,Assignmentofthe evidentialvalueofafingermarkgeneralpatternusingaBayesianNetwork, Lect.NotesInform.212(2013)99–109.
[7]T.Mollet,C.Mollet,BloodyLiesToo.MoreLiesExposedintheIngeLotzMurder Case,PiquetPublishers,CapeTown,2015.
[8]A.Girod,R.Ramotowski,C.Weyermann,Compositionoffingermarkresidue:a qualitativeandquantitativereview,ForensicSci.Int.223(2012)10–24. [9]S.Willis,L.McKenna,S.McDermott,G.O’Donell,A.Barrett,B.Rasmusson,A.
Nordgaard, C. Berger, M. Sjerps, J. Lucena-Molina, ENFSI Guideline for Evaluative Reporting in Forensic Science,EuropeanNetwork of Forensic ScienceInstitutes(2015).
[10]S.Bleay,Stillmakingamark?Fingerprintsinthe21stcentury,Sci.Justice54 (2014)1–2.
[11]R.Wieten,J.DeZoete,B.Blankers,B.Kokshoorn,Theinterpretationoftraces foundonadhesivetapes,LawProbab.Risk14(2015)305–322.
[12]R.A.H.vanOorschot,B.Szkuta,G.E.Meakin,B.Kokshoorn,M.Goray,DNA transferinforensicscience:areview,ForensicSci.Int.Genet.38(2019) 140–166.
[13]O.P.Jasuja,M.Toofany,G.Singh,G.S.Sodhi,Dynamicsoflatentfingerprints:the effect of physical factors on quality of ninhydrin developed prints—a preliminarystudy,Sci.Justice49(2009)8–11.
[14]V.G.Sears,S.M.Bleay,H.L.Bandey,V.J.Bowman,Amethodologyforfinger markresearch,Sci.Justice52(2012)145–160.
[15]K.Bobev,Fingerprintsandfactorsaffectingtheircondition,J.ForensicIdentif. 45(1995)176–183.
[16]A.Becue,S.Moret,C.Champod,P.Margot,Useofstainstodetectfingermarks, Biotech.Histochem.86(2011)140–160.
[17]C.Lennard,Fingerprintdetection:currentcapabilities,Aust.J.ForensicSci.39 (2007)55–71.
[18]A.deRonde,M.vanAken,M.dePuit,C.dePoot,Astudyintofingermarksat activitylevelonpillowcases,ForensicSci.Int.295(2019)113–120. [19]S.Subhashree,Determinationofsexandhandfromindividualfingerprints,Int.
J.ForensicSci.2(2017).
[20]I.Singh,P.K.Chattopadhyay,R.K.Garg,Determinationofthehandfromsingle digitfingerprint:astudyofwhorls,ForensicSci.Int.152(2005)205–208. [21]N.Kapoor,A.Badiye,Ananalysisofwhorlpatternsfordeterminationofhand,J.
ForensicLegalMed.32(2015)42–46.
[22]A.V.Maceo, Qualitativeassessment ofskindeformation: apilot study,J. ForensicIdentif.59(2009)390–440.
[23]B.Scruton,B.W.Robins,B.H.Blott,Thedepositionoffingerprintfilms,J.Phys. D:Appl.Phys.8(1975)714–723.
[24]D.R.Ashbaugh,Quantitative-QualitativeFrictionRidgeAnalysis:An Introduc-tiontoBasicandAdvancedRidgeology,CRCPress,1999.
[25]A.deJongh,A.R.Lubach,S.L.LieKwie,I.Alberink,Measuringtherarityof fingerprintspatternsinthedutchpopulationusinganextendedclassification set,J.ForensicSci.(2018).
[26]B. Kokshoorn,B.J.Blankers,J. DeZoete,C.E.H.Berger,ActivitylevelDNA evidenceevaluation:onpropositionsaddressingtheactorortheactivity, ForensicSci.Int.278(2017)115–124.
[27]D.Taylor,A.Biedermann,T.Hicks,C.Champod,Atemplateforconstructing Bayesiannetworksinforensicbiologycaseswhenconsideringactivitylevel propositions,ForensicSci.Int.Genet.33(2018)136–146.
[28]F.Taroni,A.Biedermann,P.Garbolino,C.G.G.Aitken,Ageneralapproachto Bayesiannetworksfortheinterpretationofevidence,ForensicSci.Int.139 (2004)5–16.
[29]D. Taylor, B. Kokshoorn, A. Biedermann, Evaluation of forensic genetics findingsgivenactivitylevelpropositions:areview,ForensicSci.Int.Genet.36 (2018)34–49.
[30]A.Biedermann,F.Taroni,Bayesiannetworksandprobabilisticreasoningabout scientificevidence whenthere isa lackofdata,ForensicSci. Int.157(2006)163–167. [31]M.Goray,S.Fowler,B.Szkuta,R.A.H.VanOorschot,Shedderstatus—ananalysis ofself andnon-selfDNAin multiplehandprints depositedbythe same individualsovertime,ForensicSci.Int.Genet.23(2016)190–196. [32]A.K.Buckingham,M.L.Harvey,R.A.H.vanOorschot,Theoriginofunknown