Using case specific experiments to evaluate fingermarks on knives given activity level propositions

(1)

Delft University of Technology

Using case specific experiments to evaluate fingermarks on knives given activity level

propositions

de Ronde, Anouk; Kokshoorn, Bas; de Puit, Marcel; de Poot, Christianne J.

DOI

10.1016/j.forsciint.2021.110710

Publication date

2021

Document Version

Final published version

Published in

Forensic Science International

Citation (APA)

de Ronde, A., Kokshoorn, B., de Puit, M., & de Poot, C. J. (2021). Using case specific experiments to

evaluate fingermarks on knives given activity level propositions. Forensic Science International, 320,

[110710]. https://doi.org/10.1016/j.forsciint.2021.110710

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.

(2)

Using

case

speci

ﬁc

experiments

to

evaluate

ﬁngermarks

on

knives

given

activity

level

propositions

Anouk

de

Ronde

a,b,c,

*

,

Bas

Kokshoorn

c

,

Marcel

de

Puit

c,d

,

Christianne

J. de

Poot

a,b,e

a

AmsterdamUniversityofAppliedSciences,Weesperzijde190,1097DZAmsterdam,TheNetherlands

b

VUUniversityAmsterdam,DeBoelelaan1105,1081HVAmsterdam,TheNetherlands

c

NetherlandsForensicInstitute,P.O.Box24044,2490AATheHague,TheNetherlands

d_Delft_University_of_Technology,_Van_der_Maasweg_9,₂₆₂₉_HZ,_Delft,_The_Netherlands

e_Police_Academy_of_the_Netherlands,_P.O._Box_348,₇₃₀₁_BB_Apeldoorn,_The_Netherlands

ARTICLE INFO

Articlehistory:

Received18May2020

Receivedinrevisedform24December2020

Accepted26January2021

Availableonline30January2021

Keywords: Evidenceinterpretation Bayesiannetworks Fingermarks Activitylevel ABSTRACT

Bayesiannetworkshaveshowntobeausefultoolfortheevaluationofforensicfindingsgivenactivity levelpropositions.Inthispaper,wedemonstratehowcasespecificexperimentscanbeusedtoassign probabilitiestothestatesofthenodesofaBayesiannetworkfortheevaluationoffingermarksgiven activitylevelpropositions.Thetransfer,persistenceandrecoveryoffingermarksonknivesisstudiedin experimentswhereaknifeiseitherusedtostabavictimortocutfood,representingtheactivitiesthat weredisputedinthecaseofthemurderofMeredithKercher.TwoBayesiannetworksareconstructed, exploringtheeffectofdifferentusesoftheexperimentaldatabyassigningtheprobabilitiesbasedonthe resultsoftheexperiments.TheevaluationofthefindingsusingtheBayesiannetworksdemonstratesthe potentialforfingermarksinaddressingactivitylevelpropositions.

1.Introduction

Evaluation of fingermarks given activity level propositions recently became a topic of interest [1,2]. The question which activityledtothedepositionofthefingermarksbecomesrelevant whenthesourceofthefingermarkisnotindispute.Researchbyde Ronde,Kokshoorn,dePootanddePuit[1]showedthatthereare multiple variables such as transfer, persistence, direction and pressure thatmay provideinformationwhen evaluating finger-marks given proposed activities that may have led to their deposition.Oneofthesevariablesisthelocationofthefingermarks onanobject.Basedonanexperimentwithpillowcases,deRonde, vanAken,dePuitanddePoot[2]haveshown thevalueofthe location of fingermarks with regards toassessing evidence for specificactivities.

Allvariablesthatinﬂuencetheinterpretationofevidencegiven activitylevelpropositionscanbecombinedinaBayesiannetwork toevaluateevidencewithregardstotherelevantactivitiesatstake [3].AstudybydeRonde,Kokshoorn,dePootanddePuit[1]has illustratedhowBayesiannetworkscanbeusedfortheevaluation of ﬁngermarks given activity level propositions by presenting

examples of Bayesian networks for a ﬁctitious balcony case example.However,inthatstudy,theassignmentofprobabilitiesto theconditionalprobabilitytablesofthenetworkswasleftoutof scope.

Thereareseveralsourcesofinformationthatcanbeusedto assign probabilities to the states of the nodes of a Bayesian network, mentioned in order of preference [4]. The forensic scientist mayperform case specific experiments and base the probabilities implemented in the network on these empirical data.Thisoptionispreferredsincetheseprobabilitieswillalign most closely with the circumstances of the case. Another possibility is to assignprobabilities based on studies reported inliteraturethatusedexperimentaldesignsthataresimilartothe case circumstances. If no empirical data are available the probabilities could be informed based on expertise by the forensicscientist.Thisoption,beingsubjectivetoalargerextent, isnotpreferredandputsaburdenonthescientisttosupporttheir probabilityassignment. Sourcesfor this could bea systematic review of resulting findings from similar cases, and/or expert elicitationfrommultiple experts. Wheneverdataare scarceor based on uncertain assumptions or sources, it is advisable to perform a sensitivity analysis to study the sensitivity of the likelihoodratiotoreasonablevariationsin theassigned proba-bilities. If data are not available, or the sensitivity analysis determinestheevaluationnottoberobust,itmaybedecidedthat thefindingsfromtheevaluationwillnotbereported.

*Correspondingauthorat:AmsterdamUniversityofAppliedSciences,

Wee-sperzijde190,1097DZAmsterdam,TheNetherlands.

E-mailaddress:a.de.ronde2@hva.nl(A.deRonde).

http://dx.doi.org/10.1016/j.forsciint.2021.110710

ContentslistsavailableatScienceDirect

Forensic

Science

International

(3)

Inthisstudy,casespecificexperimentsarecarriedoutforacase examplein ordertoshowhow thisinformation canbeusedto evaluatefingermarksgivenactivitylevelpropositions.Wewillfirst presentthecaseexampleandpresenttwoBayesiannetworksthat maybeusedfortheevaluationoffingermarksfoundonaknife given case relevantactivity level propositions.We describethe experimentsthatwereperformedandthedatagainedfromthose. We then demonstratehow theprobabilities in the conditional probabilitytablesoftheBayesiannetworkscanbeassignedbased ontheexperimentaldata.Finally,wewillshowhowthenetworks can be used to evaluate fingermarks given activity level propositions by calculating likelihood ratios for _fictitious case findings.

1.1.Caseexample–thedeathofMeredithKercher

Onthemorningofthe2thNovemberin2007,MeredithKercher wasfounddeadonthefloorofherbedroom.Itappearedthatshe wasstabbedinherneckandtorsoanditwasestablishedthatthese wounds were the cause of her death. Three suspects were identified: Rudy Guede,Kercher’s flat mate Amanda Knox and Amanda’sboyfriendRaffaeleSollecito.Allthreewereconvictedfor the murder of Meredith Kercher. Amanda Knox and Rafaele Sollecitowerelateracquitted[6].Forthiscaseexample,wewill focus ontheclaimsthattheprosecutionandthedefensemade withregardstotheknifethatwassubmittedasevidenceinthe caseagainstKnoxandSollecito.

There was noknife present onthe crime scene, raisingthe suspicion that the murder weaponwas removed. A knife was retrievedfromacutlerydrawerintheapartmentofSollecito.The knifewastestedforDNA,resultinginamatchingDNAproﬁleof AmandaKnoxonthehandleoftheknifeandamatchinglow-level DNAproﬁleofMeredithKercheronthebladeoftheknife.Theknife wastestednegativeforthepresenceofblood[5].Theprosecution claimed that the knife was themurder weapon,however; the defense denied this statementand claimedthat Knox usedthe knifeforcookinginSollecito’sapartment.

1.2.Objectives

To theauthors’ knowledge, no fingermark examination was carried outontheknifeand onlyDNAevidence presentonthe knifewasusedinthiscase.Forthispaper,weinvestigatewhatkind ofanalysiscouldbeperformedwhenfingermarkswereobtained fromtheknifeincaseslikethis.Incasefingermarkswerefoundon

theknife,thequestioninthiscasemayshiftfromsourcelevelto activitylevel;thesourceofthefingermarksontheknifewouldnot be disputed by the defense because the suspect provides an alternativeexplanationforthepresenceofherfingermarksonthe knife,namelycookingwiththeknife.Therefore,theactivityduring whichthemarksweredepositedisdisputedanditwouldbeof interesttoevaluatethefindingsgiventheactivitylevel proposi-tionsthatmaybeputforwardinthiscase.

2.Bayesiannetworkconstruction

In this section, we discuss the process of constructing a Bayesian networkto addressthequestion whetherthesuspect Amanda Knox (S) used the knife to stab the victim Meredith Kercher(V)orusedtheknifetocutfoodwhilecooking.Inthiscase, itisdisputedwhethertheknifewastheactualmurderweaponand thereforewecanformulatethefollowingpropositions,disputing theactivitythatiscarriedout:

Hp:SstabbedVwiththeknife.Sdidnotusetheknifetocut

food.

Hd:Vwasnotstabbedwiththeknife.Sonlyusedtheknifetocut

food.

AllnetworkswerebuiltusingthesoftwareHugin(version8.6) andthecorresponding.netﬁlescanbefoundinthesupplementary material.

Forthisstudy,severalassumptionshavebeenmade:

1.We assumed that the collected evidence represents one ﬁngermark grip on the knife, consisting of a collection of ﬁngermarksforwhichisassumedthattheyareleftinoneand thesameplacementofthehand.Thismeansthatanyhandling oftheknifepriortotheallegeduse(liketakingiffromadrawer orthedishwasher)isdisregarded.

2.Theassumptionismadethatthesourceoftheﬁngermarksis knowntobethesuspectandthatnooneelsetouchedtheknife. 3.TheknifeintheKerchercaseisa31cmlongknifewitha17,5cm steelbladeandablack,plasticizedhandle[7].Theknifeweused intheexperimentsisa22cmlongknifewitha11,5cm steel blade and a black plasticized handle. We assume that the patterns of ﬁngermarks on the knives resulting from the experimentsaresimilartothosethatwouldbeobtainedfroma slightlylargerknife.

4.Weassumethatthesizeofthehandofthesuspectisanaverage human hand.The assigned probabilities are based onhands fromvolunteersrangingfromsmalltolargesizehands.

(4)

5. Weassumethatthepurposeofthegripontheknifehandleisto usetheknifeasatool.Therearenumerouswaystoholdaknife. Tousetheknifeasatoolforstabbingorcuttingwouldmake some of those ways improbable. For instance, it would be improbablethat somebodywould holdtheknife withjust a ﬁngerandathumbonthehandletostaborcut.However,thisis notimpossible.Otherwaysmaysimplybeimpossibleduetothe intrinsic characteristics of both the knife and the hand, for exampleholdingtheknifewithjustathumb.Weassumethatall theimpossible,aswellasthehighlyimprobablewaystohold theknife areimpossible in the contextof the case. Wewill discussthisfurtherinSection4.

Ifanevaluationasdiscussedinthispaperwouldbeappliedtoa realcase,similarorotherassumptionsmayneedtobemade[8]. The relevance of the assumptions may be discussed with the mandatingauthoritytogetherwiththepropositionsbeingsetprior to the evaluation being carried out. Also, the impact of such assumptionsontheoutcomeoftheevaluationcanbeaddressedin thereport.

2.1.ConstructedBayesiannetwork

Basedontheshapeofaknife,itisexpectedthatfingermarks maybeobservedondifferentlocationsoftheknifewhencarrying out different activities. Three separate areas of the knife are thereforedistinguished:thehandleoftheknife,thebacksideof (thehandleof)theknifeandthebladeoftheknife.Fig.1showsthe constructedBayesian network,ofwhich twoversions(Bayesian network Iand Bayesian network II) are presented below, both showinga differentuseof theexperimentaldatabyadifferent definition of the states of nodes (4)–(12). Bayesian network I focussesonevaluatingthepresenceorabsenceoffingermarkson particular areas on the knife. Bayesian network II focusses on evaluatingtheareaoffrictionridgeskinthatwasleftonparticular areas of the knife. The networks are created following the procedure described by de Ronde, Kokshoorn, de Poot and de Puit[1],basedonthetemplatebyTaylor,Biedermann,Hicksand Champod[9].ThepresentednetworkinFig. 1hasastructurethatis differentfromthenetworkfortheevaluationoffingermarksgiven activitylevelpropositionsshowedinFigs.1,4and5presentedby de Ronde, Kokshoorn, de Poot and de Puit [1]. In that study, Bayesiannetworkswereconstructedfortheevaluationof finger-mark gripspresentona balconyrailing giventheactivitylevel propositionsthatthegripwasaresultofclimbingthebalconyor thatthegripwasaresultofleaningontherailing.Forthevariable locationinthebalconyexample,thebalconyrailingwasdivided intofourdifferentareasresultingintoregionswhichwerebigger thanthesizeofafingermarkgrip.Asaconsequence,fingermarks foundintheregionswereconsideredconditionallyindependent since the presence of a _fingermark grip in one region was considerednottoinfluencetheprobabilityforthepresenceofa fingermarkgripinanotherregion,giventheassumptionofasingle depositioneventthatwasmade.Forsmalleritems,suchasaknife, adivisionoftheitemintoregionsmayresultintoareasthatare possiblysmallerthanthesizeofafingermarkgripandassuchthe presence of a mark onthe handle of theknife mayaffect the probabilityofthepresenceofamarkonthebacksideortheblade oftheknife.Thiscausesconditionaldependenciesthatshouldbe taken into account, and therefore the nodes representing the transfer,persistenceandrecoverymechanismshavetobedefined foreachlocationregionandactivityseparately.Wesuggestthatfor items forwhich thelocation isdividedintoregions thatare of smallersizethanagrip,additionaldependencieshavetobetaken intoaccount andtheBayesiannetwork shouldbestructuredas describedinSections2.2and2.3.

2.2.BayesiannetworkI–locationoffingermarksontheknife The first Bayesian network is constructed to evaluate the presenceorabsenceoffingermarksontheknife.

2.2.1.Node(1)Propositions

The node (1) Propositions has two states, Hp and Hd,

representing the propositions of prosecution and of defense respectively.Weassignedanequalpriorprobabilityofp¼0:5to bothpropositions,asshowninTable1.

2.2.2.Nodes(2)Sstabbedthevictimwiththeknifeand(3)Scutfood withtheknife

Fromthe node(1) Propositions, two activitiesemerge: (2) S stabbedthevictimwiththeknifeand(3)Scutfoodwiththeknife, representedbythebluenodesinFig.1.Bothnodeshavethestates ‘true’and‘false’.Table2andTable3showtheprobabilitytablesfor thesenodes.Table2showsthatifHpistrue,thenode(2)Sstabbed

thevictimwiththeknifeistruewithprobabilityp¼1andfalsewith probabilityp¼0.IfHdistrue,(2)Sstabbedthevictimwiththeknife

is true withprobability p¼0and false with probability p¼1.

Table3showsthatforthenode(3)Scutfoodwiththeknife,the reversereasoningholds.

2.2.3.Nodes(4)(7)Marksonhandle,(5)(8)Marksonback,(6)(9) marksonblade

Nodes (4), (5), (6), (7), (8) and (9) represent the combined probabilityoftransfer,persistenceandrecoveryofthefingermarks toaparticularlocationoftheknifeasaconsequenceoftheactivity. Thisresultsinthenodes(4)Marksonhandle-stabbing,(5)Markson back-stabbingand(6)Marksonblade–stabbingforthetransfer, persistenceandrecoveryoffingermarkstoaparticularlocationon theknifefor thescenariostabbingand thenodes(7) Marks on handle-cutting,(8)Marksonback-cuttingand(9)Marksonblade– cuttingforthetransfer,persistenceandrecoveryoffingermarksto a particularlocationonthe knifeforthe scenariocuttingfood. These nodeseach have two states:‘fingermarksSpresent’and ‘fingermarksSabsent’.

The conditional dependencies between the three locations shouldbe considered.These dependenciesaremodelled in the Bayesian network byadding anarrow from node(4) Marks on handle–stabbingtonode(5)Marksonback–stabbing,andarrows fromnodes(4)and(5)tonode(6)Marksonblade–stabbing.The sameconnectionhasbeenmadebetweennodes(7),(8),and(9),as shown in Fig. 1. The probabilities assigned to the conditional probabilitytablesinthesenodesarebasedontheconductedknife experiment,andwillbediscussedinSection4.

Table1

Priorprobabilitytableforthenode(1)PropositionsinFig.1.

Propositions Probability

Hp:Sstabbedthevictimwiththeknife.Sdid

notusetheknifetocutfood.

0.5

Hd:Vwasnotstabbedwiththeknife.Sonly

usedtheknifetocutfood.

0.5

Table2

Conditionalprobabilitytableforthenode(2)Sstabbedthevictimwiththeknifein

Fig.1.

Propositions Hp Hd

Sstabbedthevictimwiththeknife:

True 1 0

(5)

2.2.4.Nodes(10)Findings–Marksonhandle,(11)Findings–Markson bladeand(12)Findings–Marksonback

Thenode(10)Findings–MarksonhandleinFig.1isasummary node,representingthepresenceorabsenceoffingermarksonthe handle oftheknife,withthetwo possiblestates ‘fingermarksS present’and‘fingermarks Sabsent’.Given thepropositionsand assumptionsthatweremade,wedonotconsidermarksbyother individuals. The nodes (11) Findings – Marks onblade and (12) Findings – Marks on back are similarly defined and represent respectivelythepresenceorabsenceoffingermarksonthebladeof theknifeandonthebacksideofthehandleoftheknife.

Table4 showstheconditionalprobabilitytablefornode(10) Findings–Marksonhandle.Ifeither(4)Marksonhandle-stabbing or(7)Marksonhandle-cuttingareinstate‘fingermarksSpresent’, thenode(10)Findings–Marksonhandleisinstate‘fingermarksS present’withprobabilityp¼1andinstate‘fingermarksSabsent’ withprobabilityp¼0.Theconditionalprobabilitytablesforthe nodes(11)Findings–Marksonbladeand(12)Findings–Markson backaresimilarlydefined.

2.3.BayesiannetworkII_–areaoffrictionridgeskinontheknife Thus far, we have dealt with the findings on the knife as presenceorabsenceoffingermarksonly.Itisuptothescientistto decidewhichlevelofdetailinthefindingswillbeconsideredin their evaluation. The choice will often be dictated by the observations made in the case (can certain details be deter-mined?),availabledataontransfer,persistenceandrecovery(do thedataprovidesufficientdetailtoassignprobabilities?),andthe contextual information in thecase (e.g. doesthe question that needsansweringrequireacertainlevelofdetail?)[10].

Fromtheknifeexperiment,weobservedthat aconsiderable differencebetweenthetwo activitiesstabbingand cuttingfood wasshowninobservingparticularareasoffrictionridgeskinon particular locations on the knife. We decided to add this information tonetworkII.Thehandthatleftthefingermarksis dividedintothreeareasoffrictionridgeskin:thepalm,thefingers andthethumb.Toeachtransfer,persistence,andrecoverynode representingthehandle,thebacksideandthebladeoftheknife,as wellasthethreefindingsnodes,wedefinedthestatesbasedonall possible combinations of the three areasof friction ridgeskin, leadingtothesevenstates:‘palm’,‘fingers’,‘thumb’,‘palm/fingers’, ‘palm/thumb’, ‘fingers/thumb’, ‘palm/fingers/thumb’ and ‘none’. An extra state‘undetermined’ is added toeach of these nodes representingfingermarksforwhichitisimpossibletodetermine whatareaofthehandleftthemark.

Whencombiningthe variableslocation andareaof friction ridgeskin, additional conditionaldependencies betweenthese variablesshouldbeconsidered.Forexample,ifathumbmarkis observed onthe backside of the knife, this will influence the probabilityofobservingparticularareasoffrictionridgeskinon thehandleandthebladeoftheknife,duetotheshapeoftheknife andtheshapeofahand.Sincethisdependencyexistsregardless oftheactivitythatiscarriedout,thesevariablesareconsideredto beconditionallydependentofeachotherandshouldbemodelled intheBayesiannetworkbyaddinganarrowbetweenthem[3]. ThisresultsinaBayesiannetworkthatissimilarlystructuredas BayesiannetworkIbutwiththestatesofnodes(4)_–(12)de_fined to include the area of friction ridge skin (thumb, palm, and fingers).

3.Knifeexperiment 3.1.Experimentalprotocol

Awithin-subjectsdesignwasusedinwhicheachparticipant conductedthesameexperimental tasks.Beforethestartof the experiment,informedconsentwasobtainedfromallparticipants, withwhichtheparticipantsgavepermissionfortheuseoftheir fingermarksforresearchpurposes.Atotalof24studentsofthe AmsterdamUniversityofAppliedSciences(7males,17females, allright-handeddonors)carriedouttwoseparatescenarios,each withthe useof a differentknife. Inthe firstexperiment, each participantwasaskedtopickupaknifefromthetableandtostab threetimes into a Styrofoam plateonwhich a silhouette of a personwasdrawn(Fig.2).Thefingermarksonthekniveswere directly visualized using cyanoacrylate fuming. In the second experiment,eachparticipantwasaskedtopickupaknifefroma tableand tocutapieceofgingerbreadintofourpieces(Fig.2), representing theactivity cutting food with a knife. Again, the fingermarks on the knives were directly visualized using cyanoacrylatefuming.

Thetypeofmaterialthatisbeingcutmayaffectthehandlingof theknife.Differentstructureortexture,orhardnessofthematerial mayaffecttheamountofforcebeingused(henceimpactonthe pressureassertedbytheindividualperformingthecuttingaswell asonthefrictionbetweenhandandkniferesultingfromthis)as wellasthepositioning ofthehand. Furtherwork is neededto exploretheimpactoftheseandothervariablesontheprobability oftransfer,persistence,andrecoveryofmarksfromfrictionridge skinonsurfaces.This,however,isoutsidethescopeofthecurrent study.

In this experiment, natural ﬁngermark samples were used, collected with minimal interference from the researchers to representthe conditions of thecase as closely as possibleand variables such as duration, pressure, temperature and time betweenwashinghands were notcontrolled. Between thetwo scenarios,aweektimespanwastaken.Theparticipantswerenot provided with instructions on how to handle the knife when carryingouttheactivities.

Table3

Conditionalprobabilitytableforthenode(3)ScutfoodwiththeknifeinFig.1.

Propositions Hp Hd

Scutfoodwiththeknife:

True 0 1

False 1 0

Table4

Conditionalprobabilitytableforthenode(10)Findings–MarksonhandleinFig.1.(*)denotesthefactthattheseprobabilitiesrepresentsituationswhichwillnotoccur

becausetheactivitiesstabbingandcuttingfoodarebothmutuallyexclusive(withinthecontextoftheexamplecase).

Marksonhandle-stabbing FMSpresent FMSabsent

Marksonhandle-cutting FMSpresent FMSabsent FMSpresent FMSabsent

Findings–Marksonhandle:

FingermarksSpresent 1* 1 1 0

(6)

3.2.Materials

Fortheknives,steakknivesofthemodelSNITTApurchasedat IKEAwereused(Fig.2).Theseare22cmlongkniveswitha11.5cm bladeandaplasticcoatedblackhandle.Thekniveswerecleaned withtheuseofacetone,followedbycleaningwithethanol(70%), rinsing with demi water and drying using Tork paper. For the stabbing scenarios, a Styrofoam platewas placed ona wooden standandcoveredwithplastic,onwhichasilhouetteofaperson withtheheightof1.78wasdrawn,asshowninFig.2.Aftereach stabbing scenario was carried out, the plastic was replaced to preventabiasforthenextparticipantofwheretostab.Afterthe stabbing the knives were put on a table and not covered or packaged.Theﬁngermarksweredirectlyvisualizedusing cyano-acrylatefuming(1,5g,120C)inaMVC3000fumingsystem(Foster andFreemanLTD)at80%humidity.Afterwards,thekniveswere directlyphotographedusingaNikonD60.Allexperimentswere ﬁlmedusingaLogitechC615HDwebcam.

3.3.Analysis

During the experiment, two knives were collectedfor each donor.Aftervisualization,thekniveswerephotographedbytaking fourpicturesofeachsideoftheknife.Fortheanalysis,picturesof theknivesandthevideofootageofthescenarioswerescoredbya singleresearcherusingaprede_finedsetofvariables.Duringthis analysis,thefinalgripthatwasusedfortheactivityofstabbingor cuttingfoodwas scored. Theresearcherscoredwhether finger-markswerevisualized(yes/no),whichhandtheyused(left/right), thedirectionoftheknife(overhand/underhand),therotationof the knife (cutting face of the knife pointing upwards or downwards) and whatareaof friction ridge skinonthe hands wereleftonwhichlocationontheknife.Forcodingthelocation and theareaoffrictionridgeskin,theknifewasdividedinto6 regions:side1,thetopsideoftheknifehandle(S1);side2,rotating theknife90degreesfromthetopsidetotherightsideoftheknife handle(S2);side3,thedownsideoftheknifehandle(S3);side4, rotatingtheknife90degreesfromthedownsidetotheleftsideof theknifehandle(S4);thebacksideoftheknife(back)andtheblade oftheknife(blade).RegionsS1-S4ontheknifehandleareshownin

Fig.3.

Foreachlocationontheknifewasdenotedwhatareaoffriction ridgeskinwasobservedinthevideofootage:palm,_fingers,thumb andallcombinationsthereof.Iftheareawasnottouched,thescore ‘none’wasgiven.Forthescoringprocedure,thegripusedduring theactivityobservedfromthevideofootages,wascomparedtothe picturesofthevisualizedfingermarksontheknivestodetermine whatareaofthehandleftthemarkspresentontheknife.Thefocus ofthisscoringwasnotonthequalityofthefingermarks,therefore notonlyidentifiablefingermarkswerescoredbutalsofingermarks that would possibly not be suitable for identification such as smearsorlowerscoringfingermarks[11].Tofingermarksforwhich itwasdifficulttodeterminewhatareaofthehandleftthemark,a score of NA was assigned. All video footages, pictures and the correspondingscoresweredoublecheckedbytheresearcherthat scoredthefiles.Thevideosthatwereinsomerespectuncleardue toforexamplemovementofthecamerawerediscussedwithan additionalresearcher. In case ofagreement,theareaof friction ridgeskinwasassigned,otherwiseascoreofNAwasassigned.This processshowedthatthecodingprocedurewasastraightforward processwithahighdegreeofintra-andintercoderreliability. 3.4.Results

Table5andTable6showtheobservationsfortheexperimentin which the participantsused theknife for stabbing and for the experimentinwhichtheparticipantsusedtheknifetocutfood, respectively.Thesetablesshowthatforeachscenarioandforeach donor, ﬁngermarks were visualized on the knife (column FM present=yes).

The video footages showed that there were two optional directionsforthegripasaresultofholdingtheknife.The_ﬁrstisto holdthe knife in an ‘overhand’position suchthat the wristis locatedhigherthantheelbowandtheknifeiscarriedatshoulder heightorhigher,resultinginagripinwhichthethumbisplaced nearthebacksideoftheknifehandle.Thesecondoptionistohold theknifeinan‘underhand’positionsuchthatthewristislocated loweror atequal height asthe elbowand theknife is heldat stomachheightorlower,resultinginagripinwhichthethumbis locatednearthebladesideoftheknifehandle.Table5showsthat 54%ofthedonorsthatcarriedoutthestabbingscenarioheldthe knifeintheoverhandposition.Table6showsthattheoverhand gripwasnotobservedforthecuttingfoodscenario.Thisseems logicalinviewof theactivity;cuttingfoodwiththeknifein an overhandpositioncanbeconsideredratheruncomfortable.The resultsconﬁrmourexpectationthatthedirectionofthegriponthe knifecanbedistinctivebetweentheactivitiesstabbingandcutting food.

During the experiment, we observed that two participants rotatedtheknifeduring thescenarioof stabbingsuchthat the cuttingfaceoftheknifepointedupwards(Table5).Thisrotation

Fig.2.Stabbingconstruction(left),steakknifeusedintheexperiments(right,up)

andgingerbreadusedforthecuttingscenario(right,down).

(7)

wasnotobservedforthescenarioofcuttingfood(Table6),dueto thefactthatitisimpossibletocutfoodwiththecuttingfaceofthe knife upwards.Therefore, the rotationof the knifecan alsobe consideredasadistinctivefeaturebetweentheactivitiesstabbing andcuttingfood.

Importanttonoteisthatthevariables_{‘direction’}and_{‘rotation’} of the knife as described here cannot be directly observed in casework and video footages were used in this experiment to observe these features. However, the variables location onthe knife and theareaof frictionridgeskinobserved ona speciﬁc locationindirectlyprovideinformationonthedirectionandthe rotation inwhich theknifewas held. For this reason, onlythe resultsfor thepresenceof theﬁngermarks, theareaoffriction ridgeskinandthelocationontheknifeinTable5andTable6were

used to assign probabilities to the states of the nodes of the Bayesiannetworks.

4.ProbabilityassignmentsandevaluationsusingtheBayesian networks

Fortheprobabilityassignmentstothestatesofthenodesofthe Bayesiannetworks,theprobabilityforstateiofnodekwithn_i;k observationscanbedeﬁnedas:

Pi;k¼  n_i;kþ1 I_þPI i¼1 n_i;k ð1Þ Table5

Resultingcountsforthescenariostabbing,inwhichDown=cuttingfaceofknifedownwards,Up=cuttingfaceofknifeupwards,P=Palm,F=FingersandT=Thumb.

Donor FMpresent Whichhand? Direction Rotation Side1 Side2 Side3 Side4 Backside Blade

Donor1 Yes Right Underhand Down P/F P/F F F/T None None

Donor2 Yes Right Overhand Down F P P/F F None None

Donor3 Yes Right Overhand Up P F F P T None

Donor4 Yes Right Overhand Down P P F F/T P None

Donor5 Yes Right Overhand Down F P P/F F T None

Donor6 Yes Right Underhand Down P/F/T P/F F F/T None None

Donor7 Yes Right Overhand Down F P P/F F/T T None

Donor8 Yes Right Overhand Down P/F P P F/T T None

Donor9 Yes Right Overhand Down F P P F/T T None

Donor10 Yes Right Overhand Down F P P F/T T None

Donor11 Yes Right Overhand Down F P P F/T None None

Donor12 Yes Right Overhand Down F P P F/T T F

Donor13 Yes Right Overhand Down F P P F/T None None

Donor14 Yes Right Underhand Down P P/F F P/F/T None None

Donor15 Yes Right Underhand Down P/T P F F/T None None

Donor16 Yes Right Underhand Down P/T P/F F P/F/T None None

Donor17 Yes Right Underhand Down P P/F F F/T None None

Donor18 Yes Right Underhand Down P/T P/F F P/F/T P None

Donor19 Yes Right Underhand Up P F F P/T P None

Donor20 Yes Right Underhand Down P/T P/F F F P None

Donor21 Yes Right Overhand Down NA NA NA NA NA None

Donor22 Yes Right Overhand Down F P P F T None

Donor23 Yes Right Underhand Down P P/F F F/T P None

Donor24 Yes Right Underhand Down P/F/T P/F F F/T P None

Table6

Resultingcountsforthescenariocuttingfood,inwhichDown=cuttingfaceofknifedownwards,Up=cuttingfaceofknifeupwards,P=Palm,F=FingersandT=Thumb.

Donor FMpresent Whichhand? Direction Rotation Side1 Side2 Side3 Side4 Backside Blade

Donor1 Yes Right Underhand Down P/F P/F F P/T P F

Donor2 Yes Right Underhand Down P/F F F P/T P F

Donor3 Yes Right Underhand Down P/F P/F F/T P/T P/F None

Donor4 Yes Right Underhand Down P P/F F P/T None F

Donor5 Yes Right Underhand Down P/F P/F F P/T P/F F

Donor6 Yes Right Underhand Down P/F P/F F T P/F None

Donor7 Yes Right Underhand Down P/F F F P/T P F

Donor8 Yes Right Underhand Down P/F P/F F P/F/T P/F F

Donor9 Yes Right Underhand Down P/F P/F F P/F/T P F

Donor10 Yes Right Underhand Down P/F P/F F P/T P/F None

Donor11 Yes Right Underhand Down P P/F F NA P None

Donor12 Yes Right Underhand Down P/F/T F F P/T P/F None

Donor13 Yes Right Underhand Down P/F F F P/T P/F None

Donor14 Yes Right Underhand Down P/F F F P/T P/F None

Donor15 Yes Right Underhand Down P/F F F P/T P/F F

Donor16 Yes Right Underhand Down P/F P/F F P/T P/F None

Donor17 Yes Right Underhand Down P/F F F P/T P/F F

Donor18 Yes Right Underhand Down NA P/F F/T P/T P/F F

Donor19 Yes Right Underhand Down P/F P/F F P/F/T F F

Donor20 Yes Right Underhand Down P/F P/F F P/T P None

Donor21 Yes Right Underhand Down P/F F F P/T NA None

Donor22 Yes Right Underhand Down P/F P/F F P/F/T P/F F

Donor23 Yes Right Underhand Down P/F/T P/F F P/F/T P/F F

(8)

whereIrepresentsthenumberofdifferentstatesfornodek[9,12]. NA observations were considered ‘ﬁngermarks present’ when assigningprobabilitiestothestatesoftheTPRnodesofnetworkI, andas‘undetermined’whenassigningprobabilitiestothestatesof thesenodesinnetworkII.Wehaveassumedthateach(technically possible) wayofholdingtheknifeisequallyprobable,and asa consequenceconsidereachdistributionoffrictionridgeskinmarks ontheknifeequallyprobable(apriori).Wehavethereforeassigned thesamepriorcountstoeachdeﬁnedstate.

However,somecombinationsoflocationsandareaoffriction ridge skin are impossible to realize in one grip due to the assumptionsofourstudy,theshapeoftheknifeortherestrictions inthemovementsofthehand.Forexample,asinglethumbcannot beplacedonthehandle,thebacksideandthebladeoftheknife sinceonlysinglegripsareevaluatedinthisstudy.Wedecidedto assign a probability of zero to these impossible combinations, denotedbythecolorgreyinthetablesinthismanuscriptandthe tablesinthesupplementarymaterialrepresentingtheconditional probabilitytablesforthenodes.

Totheauthors_’knowledge,theknifeinthecaseofthemurderof MeredithKercherwasnotexamined forfingermarks.Therefore, whenevaluatingfindingsusingthethreeBayesiannetworks,we willconsiderseveralfictitiousfindingsthatcouldbeobtainedina caselikethisandwewillcalculatetheweightoftheevidence.We note that the values which we calculate with the Bayesian networks in this section, are effectively posterior probabilities. Sincewehaveonlytwopropositionsinthepropositionnodes,and their assigned prior probabilities are equal, the ratio of the posteriorprobabilitiesequalsthelikelihoodratio.Hence,werefer totheratiooftheposteriorprobabilitiesaslikelihoodratios(LR) fromhereon.

4.1.BayesiannetworkI_–locationof_ﬁngermarksontheknife

Table7andTable8showtheconditionalprobabilitytablesfor thenodes(4)Marksonhandle-stabbingand(7)Marksonhandle -cuttinginBayesiannetworkIwithstates‘fingermarksSpresent’ and‘fingermarksSabsent’,inwhichtheprobabilitiesareassigned basedontheexperimentalresultsshowninTable5andTable6. Thetablesshowthatobservingfingermarksontheknifehandle doesnotprovideanyinformationontheactivitythatiscarriedout, sincetheprobabilitytoobservefingermarksontheknifehandleis equalgiventhetwopropositionsstabbingandcuttingfood.

Table9andTable10showtheconditionalprobabilitytablesfor thenodes(5)Marksonback–stabbingand(8)Marksonback– cutting in network I, respectively. The results show that the probabilitythatﬁngermarksarepresentonthebacksidegiventhat Sstabbedthevictimwiththeknifeandmarkswereobservedon the handle is considerably lower than the probability that ﬁngermarks are present on thebackside given that Scut food withtheknifeandmarkswereobservedonthehandle.

Theconditionalprobabilitytablesforthenodes(6)Markson blade -stabbing and(9) Marks onblade –cutting areshown in

Table11andTable12.Theseresultsshowthattheprobabilityto observefingermarksonthebladegiventhatthefingermarksended upontheknifethroughstabbingisverylowandforalmostall participants, fingermarks were absent on the blade. On the

contrary, the probability to observe fingermarks on the blade given that the fingermarks ended up on the knife through preparingfood arealmost equal ifmarksare alsoobserved on thehandleandthebacksideoftheknife.Ifmarksareonlyobserved onthehandle,theprobabilitytoobservefingermarksontheblade oftheknifeisslightlyhigherthantonotobservefingermarkson theblade.

4.2.NetworkI–explorationI

InstantiatingpropositionsHpandHdconsecutivelyinnetworkI

(supplementary material)showsthattheprobabilityforthepresence orabsenceoffingermarksontheknifehandleisequalgivenboth propositions,showing that thepresenceor absence offingermarkson theknifehandleindeeddoesnotprovideanyevidentialvalue.When evaluatingthefindingsthatfingermarksofSarepresentonallthree areas ofthe knife, the findingssupport the propositionthatthe suspectcut food with the knife. If thefingermarksofSareonlypresent ontheknifehandleandnotonthebacksideandthebladeoftheknife, thefindingssupportthepropositionthatthesuspectstabbedthe victimwiththeknife.Incaseweevaluatetheabsenceoffingermarks ontheknife,thefindingsdonotaddanyevidentialweightandresult inanLR of1.Thiscanbeexplainedbythefactthatthisfindingwasnot observedinourexperiment.

4.3.BayesiannetworkII–areaoffrictionridgeskinontheknife

Table13showstheconditionalprobabilitytableforthenode(4) Marksonhandle –stabbingand Table14shows theconditional probability tablefor the node(7) Marks on handle - cuttingin BayesiannetworkII.Theprobabilitiesareassignedbasedonthe experimentalresultsshowninTable5andTable6,forwhichthe observations in columns Side 1, Side 2, Side 3 and Side 4 are combinedtorepresenttheﬁndingsonthehandle.Theresultsshow thatforbothpropositions,theprobabilitytoobservethepalm,the ﬁngersandthethumbonthehandleisthehighest.Therefore,the areaoffrictionridgeskinobservedontheknifehandleprovides onlylittleinformationontheactivitythatiscarriedout.

Forthenodes(4)Marksonhandle–stabbingand(7)Markson handle–cuttingisdeterminedthatthestate‘thumb’isconsidered impossibletoachieveduetothefactthatplacingonlythethumb onthehandlewithoutthepalmorﬁngersmakesitimpossibleto even carry the knife. Thisstate is therefore removedfrom the optionalstates.

Theconditionalprobabilitytablesforthenodes(5)Markson back–stabbingand(8)Marksonback–cuttingcanbefoundinthe supplementary material. Since these nodes are conditionally dependentonthemarksobservedonthehandle,thereareagain multiplecombinationsoflocationsandareaoffrictionridgeskin whichareconsideredimpossiblegiventheallegedactivitiesand therefore received a probability of zero (denoted grey in the conditionalprobabilitytables).Theconditionalprobabilitytables forthenodes(6)Marksonblade–stabbingand(9)Marksonblade– cuttingcan alsobefoundin thesupplementary material.Since thesenodesareconditionallydependentonthenodes(4)(7)Marks onhandleand(5)(8)Marksonbackside,thelocationcombinations which were already consideredimpossible for these nodes are

Table7

Conditionalprobabilitytableforthenode(4)Marksonhandle-stabbinginnetwork

I.

Sstabbedthevictimwiththeknife True False

Marksonhandle-stabbing:

FingermarksSpresent 0.96 0

FingermarksSabsent 0.04 1

Table8

Conditionalprobabilitytableforthenode(7)Marksonhandle-cuttinginnetwork

I.

Scutfoodwiththeknife True False

Marksonhandle-cutting:

FingermarksSpresent 0.96 0

(9)

removedfromtheconditionalprobabilitytable.Newcombinations whichcanbeconsideredimpossibletoachievewiththedisputed knife and ahumanhand areagainmarked withgreycells and receiveaprobabilityofzero.

4.4.NetworkII–Exploration

Theexperimentalresultsshowedthatthumbswereonlyplaced on the backside of the knife in case theknife was held in an

overhand grip, which only occurred for the scenario in which participantsstabbedusingtheknife.Therefore,weareinterested intheevidentialvalueprovidedbythemodelforthisobservation. Therearefourstatesforthenode(12)Findings–Marksonbackthat incorporatethepresenceofathumbonthebacksideoftheknife: the states ‘thumb’, ‘palm/thumb’, ‘ﬁngers/thumb’ and ‘palm/ ﬁngers/thumb’. Instantiating one of these states for the node (12) Findings– Markson back providesa LR in supportfor the propositionthatthesuspectstabbedthevictimwiththeknife.The

Table9

Conditionalprobabilitytableforthenode(5)Marksonback-stabbinginnetworkI.

Marksonhandle-stabbing FMSpresent FMSabsent FMSpresent FMSabsent

Marksonback–stabbing:

FingermarksSpresent 0.62 0.5 0 0

FingermarksSabsent 0.38 0.5 1 1

Table10

Conditionalprobabilitytableforthenode(8)Marksonback-cuttinginnetworkI.

Marksonhandle-cutting FMSpresent FMSabsent FMSpresent FMSabsent

Marksonback–cutting:

FingermarksSpresent 0.92 0.5 0 0

FingermarksSabsent 0.08 0.5 1 1

Table11

Conditionalprobabilitytableforthenode(6)Marksonblade-stabbinginnetworkI.

Marksonhandle–stabbing FMSpresent FMSabsent FMSpresent FMSabsent

Marksonback–stabbing FMSpresent FMSabsent FMSpresent FMSabsent FMSpresent FMSabsent FMSpresent FMSabsent

Marksonblade–stabbing:

FMSpresent 0.118 0.091 0.5 0.5 0 0 0 0

FMSabsent 0.882 0.909 0.5 0.5 1 1 1 1

Table12

Conditionalprobabilitytableforthenode(9)Marksonblade-cuttinginnetworkI.

Marksonhandle–cutting FMSpresent FMSabsent FMSpresent FMSabsent

Marksonback–cutting FMSpresent FMSabsent FMSpresent FMSabsent FMSpresent FMSabsent FMSpresent FMSabsent

Marksonblade–cutting:

FMSpresent 0.52 0.667 0.5 0.5 0 0 0 0

FMSabsent 0.48 0.333 0.5 0.5 1 1 1 14

Table13

Conditionalprobabilitytableforthenode(4)Marksonhandle-stabbinginnetwork

II.

Marksonhandle-stabbing:

Palm 0,031 0 Fingers 0,031 0 Palm/Fingers 0,156 0 Palm/Thumb 0,031 0 Fingers/Thumb 0,031 0 Palm/Fingers/Thumb 0,625 0 Undetermined 0,063 0 None 0,031 1 Table14

Conditionalprobabilitytableforthenode(7)Marksonhandle-cuttinginnetwork

II.

Marksonhandle-cutting:

Palm 0,031 0 Fingers 0,031 0 Palm/Fingers 0,063 0 Palm/Thumb 0,031 0 Fingers/Thumb 0,031 0 Palm/Fingers/Thumb 0,75 0 Undetermined 0,063 0 None 0,031 1

(10)

resultsfromtheexperimentshowedthat13participantsplaced theirfingersonthebladeoftheknifewhilecuttingfood,whereas for the stabbing scenario,this was only one participant.When evaluatingthefindingthatfingerswereobservedontheblade,the findingssupportthepropositionthatthesuspectcutfoodwiththe knife. Thenetwork also showsthat in case nofingermarks are foundonthehandleoftheknife,theonlypossibilitytoholdthe knifeistoholdtheknifeatthebladewiththepalm/fingers,palm/ thumb, fingers/thumb or the palm/fingers/thumb. Additionally, whennofingermarksarefoundonthehandleoftheknife,theonly possiblefindingforthebackoftheknifeisthatnofingermarksare observed,sinceitisconsideredimpossibletoholdtheknifewhile onlytouchingthebackoftheknifeandnotthehandleoftheknife. 4.5.EvaluatingfictitiousfindingsintheMeredithKerchercase

In this section, we would like to explore the use of the constructedBayesiannetworkstoevaluatepossiblefindingsinthe MereditchKerchercase.Totheauthorsknowledge,nofingermark examinationwascarriedoutontheknifeintheMeredithKercher case,causingtheevaluationscarriedoutinthissectiontobesolely based on fictitious findings. Suppose that the knife that was retrievedfromtheapartmentofSollecitocontainedmarksofthe fingers,thepalmandthethumbonthehandle,marksofthefingers onthebladeoftheknifeandmarksofthepalmandthefingerson thebacksideoftheknife.Whenevaluatingthesefindingsusing network I, thestate‘present’ is instantiatedfor the nodes(10) Findings–Marksonthehandle,(11)Findings–Marksontheblade and(12)Findings–Marksontheback,shownbytheredbarsfor thesenodesinFig.4.ThisresultsinaLRof7insupportofthe proposition that the suspect used theknife tocut food.When evaluatingthesefindingsusingnetworkII,thestate‘palm/fingers/ thumb’isinstantiatedfornode(10)Findings–Marksonthehandle, thestate‘fingers’isinstantiatedfornode(11)Findings–Markson thebladeandthestate‘palm/fingers’isinstantiatedforthenode (12)Findings–Marksontheback,resultinginaLRof34infavorof Hd. This means that under the propositions stated and the

assumptions mentioned in Section2,theﬁndings are34 times moreprobableifthesuspectcutfoodwiththeknifethanthanif thesuspectusedtheknifeforstabbing.

Nowconsiderthat thefollowing fingermarkswereretrieved fromtheknife:marksofthefingersandthepalmonthehandle,no fingermarksonthebladeoftheknifeandamarkofthethumbon

thebacksideoftheknife.Whenevaluatingthesefindingsusing network I,thestate ‘present’is instantiated for thenodes (10) Findings–Marksonthehandleand(12)Findings–Marksontheback andthestate‘absent’isinstantiatedforthenode(11)Findings– Marksontheblade.TheresultingLRis1demonstratingthatwith networkIthefindingsareequallyprobablegivenboth proposi-tions.WhenevaluatingthesefindingsusingnetworkII,weobtaina LRof35insupportofHp,asshowninFig.5.

Onerequirementforaformalprobabilistic assessmentgiven activitylevelpropositionsisthattheoutcomeoftheevaluationis robust[13].Totestthis,asensitivityanalysiscanbeperformedto assess the impact of reasonable variations in the assigned probabilitiesontheresultingLR.Werefrainfromdoingsowith these ﬁctitious ﬁndings in the Meredith Kercher case. For an exampleoftheuseofsensitivityanalyseswerefertheinterested readertoSzkuta,Ballantyne,KokshoornandvanOorschot[14]. 5.Discussion

The purpose of this study was to demonstrate how data resultingfrom case specific experimentscan beused toassign probabilitiestothestatesofthenodesofaBayesiannetworkfor theevaluationoffingermarksgivenactivitylevelpropositions.For thispurpose,weconductedanexperimentinwhichaknifewas either used to stab a victim or to cut food, representing the activitiesthatweredisputedinthecaseofthemurderofMeredith Kercher.TwoBayesiannetworkswereconstructed:onetoevaluate thepresenceorabsenceoffingermarksonparticularlocationsof theknifeandonetoevaluatetheareaoffrictionridgeskinthatwas leftonparticularlocationsoftheknife.Probabilitieswereassigned basedontheempiricaldataresultingfromtheknifeexperiment andweexploredtheLRcalculatedwiththemodels.Wewouldlike toemphasize that the Bayesian networksare a result of many choicesmadeduringtheprocess.Forexample,thechoiceofhowto dividetheknifeintodifferentlocationsorhowtodividethehand into different areas directly influences the construction of the network. This is often a tradeoff between obtaining as much information as possible fromthe experimental dataversus the amountandqualityofthedatathatareavailabletoinformthe probabilityassignments.Forexample,basedonthecollecteddata fortheknifeexperiment,itcouldbequestionedwhetherafurther divisionoftheknifehandleintofourseparateareaswouldprovide moreinformation.However,whendefiningmorestatestoanode,

Fig.4. BayesiannetworkIforwhichthefindingsfingermarkspresentonthehandle,fingermarkspresentonthebladeandfingermarkspresentonthebacksideare

(11)

thenumberofobservationsforeachstatewilldecreasewhenusing the same sample size for the experiment. The conditional probability tables for network II already showed that dividing the knife into three areas caused many states for which no observations were available. Increasing the number of location nodeswhileassigningtheprobabilitiesbasedonthesamesample sizewillcausetheLRtobelessinformative(e.g.approaching1). Therefore, the design of the network always depends on data availabletoinformtheprobabilityassignments.Inthisstudy,we didnotfocusonthequalityofthefingermarkswithregardsto sourcelevelinformation,ameasurethatisnowadaysusedtoselect thefingermarks thatare collectedfroma crimescene[15]. For casesinwhich thedonorof thefingermarksremains unknown, networkIfocusingonthepresenceorabsenceoffingermarkson theknivescanverywellbeusedtoevaluatethefingermarksgiven activity level propositions since no source level information is required.Thismayforinstancebeusedincaseassessmentwhen therelevanceofaparticularknifetoacriminalactivityisdebated. For network II, comparison to reference fingerprints from the person of interest is usually required todeterminethe areaof frictionridgeskinthatleftthemarks.Inourexperiment,weused videofootagestogetherwithphotographstakenfromthe finger-marksontheknivestodeterminetheareaoffrictionridgeskinthat left the marks. Anadvantageof this choice isthat smears and fingermarksthat arenotsuitable foridentificationpurposesare alsotakenintoaccount.Adisadvantageofthischoiceisthatthese video footages are generally not available in casework, and therefore, theprobabilitytofindfingermarksand theabilityto assign the area of friction ridge skin to a mark based on this experimentareoverestimatedcomparedtocasework.Afurther study focusing on comparing the conducted approach to an approachfocusingonthequalityofthefingermarks(i.e.grading the fingermarks by using a scale as proposed by Sears, Bleay, BandeyandBowman[11]orBecue,Moret,ChampodandMargot [16]) is needed to point out the implications of the selected method.A limitationofourexperimentis thatalldonorswere

right-handed.For left-handeddonors,we expecta differencein areaoffrictionridgeskinthatwillenduponthedifferentsidesof theknifehandlesincethegripwouldmostprobablybeamirrored imageofaright-handedgrip.However,sincewehavetakenall sidesoftheknifehandletogetherbydividingtheknifeintothe three locations handle, backside and blade in the networks proposed, we do not expect much differences between right-handedandleft-handed donors.In casethedifference between right-handed and left-handed donors are a topic for further research,werecommendtodividetheknifehandleintosmaller areas(e.g.S1-S4)suchthattheinformationwhichareaoffriction ridge skin ended on which side of the handle may provide informationonthehandednessofthedonor.Thedatafromthe experimentspresentedheremustbecarefullyconsideredwhen usedincasework,tomakesuretheresultsarealsobeingapplicable tothecaseathand.Forexample,allassumptionsandevaluations describedinthispaperarebasedonthesteakknifeusedinthe experiment.Theresultsobtainedfromtheexperimentcouldalso beusedforknivesofsimilarsizeandshapeasthesteakknifeused inthisexperiment.However,ifthesizeortheshapeoftheknifeof interestchangestoacompletedifferentknifesuchasafoldable knifeoracleaver,theresultsmaynotbedirectlyapplicablesince the characteristics of the knife directly influence the possible combinationofgripsontheknife.Whenusingthedatapresented in this paper for evaluations in real casework, a carefull consideration of the characteristics of the knife, but also the activitiesatstake,theconductedexperimentandtheassumptions that were made is required. To be able to use the Bayesian networks for the evaluation of the findings, it is of great importance that all conditional dependencies between the variablesarecarefullyconsidered.Althoughthesedependencies mayresultinacomplexnetwork,ignoringdependenciesthatin reallifeexist mayresultinan overestimationof thelikelihood ratio.Forexample,ifthedependencybetweentheareaoffriction ridgeskinduetotheshapeofahandwasignoredinourresearch, thiswouldresultinanunjustifiedhigherlikelihoodratio.Onthe

(12)

other hand, an underestimation of the likelihood ratio is also possiblewhenprobabilitiesareassignedtocombinationsofareaof frictionridgeskinonparticularlocationsontheknifeforwhichis known they are impossible to achieve. If these combinations receivedaprobability,theyareconsideredfeasibleand combina-tionsthatareactuallyfeasiblereceivealowerprobability,resulting inanunderestimationofthelikelihoodratio.Therefore,wewould like tostress the importance of a careful consideration of the dependenciesbetweenvariablesandacarefulconsiderationofthe statesorcombinationsofstatesthatarenotfeasible.Additionally, assigning the prior probabilities to improbable combinations shouldalsobediscussedincourtsincethisalsodirectlyin_fluences thelikelihoodratio.Thelikelihoodratiovaluesresultingfromour calculations can be considered relatively low (0.01LR 50) resultinginaslightormoderatesupportforoneofthepropositions [13].Areasonforthisisthatourexperimentalsamplesizewas relativelysmall,i.e.24participantsforeachscenario.Duetothe numberofpossiblestatesforthenodes,thisresultsinmanystates whichstayunobservedinoursmallsamplesizewhiletheymay receiveobservationswhenusingalargersamplesize.Althoughthe rangeofLRsobtainedinthisstudymightbeconsideredrelatively low,thisdoesnotmeanthatanevaluationoffingermarksgiven activitylevelpropositionsisnotvaluable.Thisisbecausetheissue thatisbeingaddressedatactivitylevelisgenerallymuchcloserto the deliberations of the court than any source level issues. Depending onthesample size,thedatacollectionstrategy,the uniquenessofparticularobservationsforcertainactivitiesonthe objectofinterestandotherfactors,thelikelihoodratiovaluemay increase (or decrease) for other scenarios or other objects of interest.Furthermore,whencombiningtheresultsforfingermarks givenactivitylevelpropositionstogetherwithallotherevidence present in a case, this relatively‘low’LR valuemay still add a considerable value toa case and help a jury or judge in their decision. In this paper, we presented an approach to evaluate fingermarks given activity level propositions in cases like the Meredith Kercher case by using Bayesian networks and a case specificexperiment.Fromthecurrent trendswithinthefieldof forensicscience,afocusonquestioninghowandwhenevidence endeduponasurfaceisobserved[17].Inouropinion,thisnew focus ontheactivitythatledtothedepositionof tracesisalso relevantfor fingermarkevidence.TheuseofBayesiannetworks and case specific experiments to assign the probabilities to transfer, persistence, and recovery of friction ridge skinmarks shows great potential for the evaluation of fingermarks given activity level propositions in casework. With the use of this powerfulandtransparentmethod,ascientistisabletoassistthe courtinaddressingandevaluatingtheirfindingsgiventherelevant activitylevelquestionsinacase.

Funding

This work was supported by the RAAK-PRO funding of the FoundationInnovationAlliance(SIA–StichtingInnovatie Allian-tie),researchgrantno.2014-01-124PRO.

CRediTauthorshipcontributionstatement

AnoukdeRonde:Conceptualization,Formalanalysis, Investi-gation,Methodology,Projectadministration,Software, Visualiza-tion, Writing - original draft, Writing - review & editing. Bas Kokshoorn: Conceptualization, Formal analysis, Methodology, Software,Visualization,Writing-review&editing.MarceldePuit: Conceptualization, Funding acquisition, Methodology,

Supervision,Writing-review&editing.ChristianneJ.dePoot: Conceptualization, Funding acquisition, Methodology, Supervi-sion,Writing-review&editing.

DeclarationofCompetingInterest None.

Acknowledgements

WegratefullythankArameKeukerandKimvanderSalmfor their work on the design and data collection for the knife experiment. We thank Marc van Bochove for supervising and facilitatingtheexperimentandwewouldliketothankCaroline Gibbforherhelpduringthemethoddevelopmentstage. AppendixA.Supplementarydata

Supplementarymaterialrelatedtothisarticlecanbefound,inthe onlineversion,athttps://doi.org/10.1016/j.forsciint.2021.110710. References

[1]A. de Ronde, B. Kokshoorn,C.J. de Poot, M.de Puit, The evaluation of

ﬁngermarksgivenactivitylevelpropositions,ForensicSci.Int.302(2019).

[2]A.deRonde,M.vanAken,M.dePuit,C.dePoot,Astudyintoﬁngermarksat

activitylevelonpillowcases,ForensicSci.Int.295(2019)113–120.

[3]F.Taroni, A. Biedermann, S.Bozza, P.Garbolino, C.G.G.Aitken, Bayesian

NetworksforProbabilisticInferenceandDecisionAnalysisinForensicScience,

2nded.,JohnWiley&Sons,2014.

[4]D. Taylor, B. Kokshoorn,A. Biedermann, Evaluation of forensic genetics

ﬁndingsgivenactivitylevelpropositions:areview,ForensicSci.Int.Genet.36

(2018)34–49.

[5]P.Gill,AnalysisandimplicationsofthemiscarriagesofjusticeofAmandaKnox

andRaffaeleSollecito,ForensicSci.Int.Genet.23(2016)9–18.

[6]The SupremeCourt of CassationMotivation Report(the Marasca-bruno

MotivationsReport),(2015).http://www.amandaknoxcase.com/wp-content/

uploads/2015/09/Marasca-Bruno-Motivations-Report.pdf.

[7]S. Conti, C. Vecchiotti, The Conti-vecchiotti Report, (2011) . https://

knoxdnareport.ﬁles.wordpress.com/2011/07/translation-of-the-conti-vec-chiotti-report2.pdf.

[8]D. Taylor, B. Kokshoorn, T. Hicks, Structuring cases into propositions,

assumptions,andundisputedcaseinformation,ForensicSci.Int.Genet.44

(2020)102199.

[9]D.Taylor,A.Biedermann,T.Hicks,C.Champod,Atemplateforconstructing

Bayesiannetworksinforensicbiologycaseswhenconsideringactivitylevel

propositions,ForensicSci.Int.Genet.33(2018)136–146.

[10]B.Kokshoorn,L.H.Aarts,R.Ansell,E.Connolly,W.Drotz,A.D.Kloosterman,L.G.

McKenna, B. Szkuta, R.A. Van Oorschot, sharing data on DNA transfer,

persistence, prevalence and recovery: arguments for harmonizationand

standardization,ForensicSci.Int.Genet.37(2018)260–269.

[11]V.G.Sears,S.M.Bleay,H.L.Bandey,V.J.Bowman,Amethodologyforﬁnger

markresearch,Sci.Justice52(3)(2012)145–160.

[12]A.Biedermann,F.Taroni,S.Bozza,Implementingstatisticallearningmethods

throughBayesiannetworks.Part1:aguidetoBayesianparameterestimation

usingforensicsciencedata,ForensicSci.Int.193(1-3)(2009)63–71.

[13]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, European Network of Forensic

ScienceInstitutes(2015).

[14]B.Szkuta,K.N.Ballantyne,B.Kokshoorn,R.A.VanOorschot,Transferand

persistenceofnon-selfDNAonhandsovertime:usingempirical datato

evaluateDNAevidencegivenactivitylevelpropositions,ForensicSci.Int.

Genet.33(2018)84–97.

[15]S.Bitzer,O.Ribaux,N.Albertini,O.Delémont,Toanalyseatraceornot?

Evaluatingthedecision-makingprocessinthecriminalinvestigation,Forensic

Sci.Int.262(2016)1–10.

[16]A.Becue,S.Moret,C.Champod,P.Margot,Useofquantumdotsinaqueous

solutiontodetectbloodﬁngermarksonnon-poroussurfaces,ForensicSci.Int.

191(1-3)(2009)36–41.

[17]A.Biedermann,C.Champod,G.Jackson,P.Gill,D.Taylor,J.Butler,N.Morling,T.

Hicks,J.Vuille,F.Taroni,EvaluationofforensicDNAtraceswhenpropositions

ofinterestrelatetoactivities:analysisanddiscussionofrecurrentconcerns,