Crispness assessment of roasted almonds by an integrated approach to texture description - PDF document

Crispnessassessmentofroastedalmondsbyanintegratedapproachtotexturedescription:texture,acoustics,sensoryandstructureP.Varela,J.Chen,S.FiszmanandM.J.W.PoveyInstitutodeAgroquõmicayTecnologõadeAlimentos(CSIC)ApartadodeCorreos73,46100Burjassot,ValenciaSpainProcterDepartmentofFoodScience,UniversityofLeeds,LeedsLS29JT,UKReceived16May2006;Revised15December2006;Accepted22December2006Thisstudycombinespassiveacousticandmechanicalmeasuresofsensorycrispness.Weshowthattheacousticsignalisdominatedby‘bursts’ofsoundassociatedwithcrackfailureeventsintheproductwhichalsoreleasemeasurableamountsofelasticenergy.One-wayanalysisofvariance(ANOVA)andprincipalcomponentanalysis(PCA)wereperformedonthesensory,acoustical,mechanicalandcompositionalparameters.Weshowthatthischemometricapproachisapowerfulmethodfortheobjectiveanalysisoflarge,complexdatasetsinthecontextofhumansensorystudiesandtheobjectivemeasureofasensoryparameter;inthiscasecrispness.Wedemonstratethatsensorycrispnessinalmondsisanamalgamofacousticandmechanicaleffectsoccurringduringchewing.Weshowthatourmethodiscapableofpredictingthecrispnessofroastedalmonds.CopyrightJohnWiley&Sons,Ltd.crispness;texture;sound;sensory;microstructure;PCA;almonds1.INTRODUCTIONThedistinctionbetweenthetwomultidimensionalsensoryattributescrispnessandcrunchinessisnotyetfullyunder-stood;butweknowthattheyaredirectlyrelatedtothemechanicalandfracturepropertiesofsolidfoodmaterials,totheirmacroandmicrostructureandtothewaytheyareeaten[1,2].Differentapproacheshavebeenusedforcrispnessevaluation,sensorytests,instrumentaltextureandsoundanalyses,aswellasmicrostructuralobservationsinceSzczesniak[3,4]highlightedtheimportanceoffoodcrispnessforconsumers,andVickers,BourneandDrake[5,6]relatedtheperceptionofthisattributetotheauditorysensations.Studiesoncrispness/crunchinessassessmentcanbefoundintheliteratureoverthelast50years[7].Recently,someinterestingreviewshavebeenpublishedonthesubject[8–10].Thesesummarisethevariousmethodsusedandpinpointtwoprincipalapproachestakeninthestudyofcrispy/crunchytexturesusingacoustictechniques:themeasurementoftheperceptionofair-conductedsoundstoestablishitscontributiontothesensationofcrispness,andtherecordingofthesoundswhileperformingamechanicaltest.Theadvantageofthisapproachisthatallaspectsoffracturearewellcontrolled.Intherstcase,theseauthorsplayedpre-recordedbiteandchewsoundstosubjectstoassesscrispness,oraskedthemtocheworbiteandtoevaluatethesound,oracombinationofboth.Theresultsobtainedpermittedthedevelopmentofdenitionsofcrispy,crunchyandcracklyandthesuggestionofthepossibledifferencesbetweentheterms,whicharestillcontroversial.Theyalsoshowedthatsubjectscouldstillassesscrispnesswhenauditoryblockingwasappliedduringconsumptionofafood,suggestingthatcrispnesswastosomeextent,avibro-tactilesensation.Thecombinationofacousticrecord-ingwithmechanicaltestsresultsinamorecontrolledand *Correspondenceto:M.J.W.Povey,ProcterDepartmentoffoodScience,UniversityofLeeds,LeedsLS29JT,UK.E-mail:m.j.w.povey@food.leeds.ac.uk2007JohnWiley&Sons,Ltd. TextureAnalyserwaseffectiveindetectingtheacousticemissionofbiscuitsatrupture.Theyfoundanexcellentcorrespondencebetweentherecordedsoundanalysedasthenumberandmaximumamplitudeofsoundbursts,thesecondderivativeoftheforcecurveinabendingtestandthesensoryassessmentofcrispness.RecentlyCourcouxetal.al.usedacousticemissiononlyasanobjectivemeasureofcrispnessincerealakes,demonstratingthelimitationsofpreviousworkwithregardtodataanalysistechnique.Forexample,Tesch,Normand&Peleg[12]werenotabletoestablisharelationshipbetweenthemechanicalandacousticsignatureobtainedforcellularcerealfoodsbycompression,attributingittothedifferentsamplingratesofthemeasurementsortothedifferencebetweensamples.Almondshaveacrispy/crunchynature,sometimesenhancedbyroasting.Spainistheworld’ssecondalmondproducer[13].Itscultivationhasriseninthelastyears,particularlyinthesouth-easternregion,wheretheprincipalvarietyis,mostlyconsumedasaroastedorfriedsnack.Convectionoven,themostcommonroastingsystem,causeschangesinappearance,textureandavour,duetodehydration,browning,lipidoxidationanddiversestruc-turalchanges[14,15].Thedegradationofthestructureduringroastingcouldcausechangesintexturalpropertieslikecrispness,hardness,grittiness,porosity,fracturability,etc.,whichinturnwouldaffectthecharacteristicsofthenoiseemittedoneating.Almondswerechosenasstudymaterialbecausecrisp-ness/crunchinessisakeyfactorintheacceptabilityofroastedalmonds.LittleresearchhasbeendoneonalmondtextureinspiteofthefactthattheyareveryimportanttotheSpanishmarket.Inaddition,theconsiderabledifferencesbetweenalmondsandbiscuitsprovidedausefulcomparisonwithourpreviouswork[2].Theaimsofthisworkweretoassessthecrispnessofalmondswithdifferentdegreesofroastingbymeansofanintegratedapproach:thesimultaneousmeasurementofforce/displacementoncom-pressionoftheemittedsound,andtheirrelationtothestructure,microstructureandsensoryassessment.Inordertomeetthechallengeofthecomplexitiesofferedbyalmondsasanaturalandheterogeneousmaterial,weappliedchemo-metricmethodsofanalysingacousticandmechanicaldata,whichareofgeneralapplicabilitytocomplexheterogeneousmaterialsthatfailbycracking.2.MATERIALSANDMETHODS2.1.TestingmaterialsRaw,peeledalmonds(Prunusamygdaluspurchasedinalocalsupermarketweredividedintovebatches;fourwereroastedinaconvectionoven,at200C,for1.5,3.0,4.5and6.0min.Whentheyreachedambienttemperature,theywereplacedinairtightcontainers(rigidpolypropylene)andplacedinarefrigeratedchamberuntilAdditionally,thealmondswereselectedaccordingtotheirsizeandshape,toensuretheminimumvariationwithinabatch,withdimensions:length221mm,width141mm,height71mm,andweight1.20.1g.Thiswastoaidconsistencyinthemechanicaltesting.2.2.Instrumentalanalyses:textureandATA-XTplusTextureAnalyser(StableMicroSystems,Godalming,UK)wasusedforforce/displacementmeasure-mentwitha30kgloadcell,togetherwithaP/40cylinder(40-mmdiameter)usedforthecompressiontests.Thetestsettingswere:testspeed1mm/s,targetmode:strain,strain50%,trigger10g.AnAEDdescribedindetailelsewhere[2]wasusedforsoundrecordingtogetherwiththecorrespond-ingsoftware(TextureExponent32).ThegainoftheAEDwassetatone.ABruelandKjaerfree-eldmicrophone(diameter8-mm),calibratedusinganAcousticCalibratorType4231(94and114dBSPL-1000Hz)waspositionedat3cmdistanceandwithanangleof0tothesample(Figure1).Thisangleminimisedthedistortionthattheprobecreatedinthesoundeld.Ahighpasslterof1kHzremovedextraneousambientacousticandmechanicalnoise.Alowpassltersettheuppercalibratedandmeasuredfrequencyat16kHz.TheAEDoperatesbyintegratingallthefrequencieswithinthebandpassrangegeneratingavoltageproportionaltotheSPL.Theintegrationtimewassetat0.25ms.Thedataacquisitionratewas500pointspersecondforbothforceandacousticsignals.Alltestswereperformedinalaboratorywithnospecialsoundprooffacilities,witharelativehumidityofaround1%,andatemperatureof22C.Eachtestwasperformedonsixalmondsfromeachbatch(therawoneandthefourdifferentroastingtimes—sixreplicationsoneachsample).ForceversusdisplacementandSPLversusdisplacementwereplotted,theparametersextractedfromthecurvesandtheirunitsaredisplayedinTableI.2.3.SensoryanalysisApanelofeightassessorsoftheProcterDepartmentofFoodScienceoftheUniversityofLeeds,experiencedincrispnessassessmentwasused.Theassessorswereblindfoldedsothatnovisualinterferencewasinvolvedintheperceptionofcrispness,andtheywerefedwiththesamplestoavoidtactileinterference.Theyreceivedonealmondofeachdifferentroastingtimeandoneraw,followingabalancedcompleteblockexperimentaldesign.Foreachsample,theyhadtoquantifytheattributebyintroducingthewholealmondintotheirmouthandchewingthesample.Theyuseda9-boxscalelabelledontheleftwith‘nil’andontheright Figure1.Diagramoftheinstrumentalsetupforthetexture/acousticmeasurement.2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cem312P.Varelaetal. with‘high’toassessthelevel.Atthesametime,amicrophonewaspositionedinfrontoftheassessorswithinadistanceof31cmawayfromthemouthandacousticsignalswererecorded.Themicrophoneusedwasthesameasinthemechanicaltests.2.4.MacrostructuralandmicrostructuralMacro-structuralobservationphotographswereperformedwithanIntelComputerMicroscope(IntelSanDiego)usingthe10magnicationsetting.MicrostructuralanalysiswasperformedbyEnvironmentalScanningElectronMicroscopy(ESEM)ontherawand6min-roastedsamples.APhilipsXL30ESEM(OxfordInstruments,Oxford)wasused.Thesampleswerecom-pressedwiththetextureanalyserasinthetexture/soundtestsandthefracturedsurfacesobtainedwereobserved.Experimentalconditionswere30kVacceleratingvoltage,atemperatureof4C,andapressureoflessthan5.2torr.2.5.Compositionalanalyses:moistureandtotalfatcontentsMoisturewasdeterminedbyvacuumdryingat95constantweight,AOACmethod925.40.TotalfatcontentwasdeterminedbydirectextractionwithetherinaSoxhletextractor,AOACmethod948.22[16].Fourreplicationswereperformedforeachdetermination.2.6.StatisticalanalysesanddataacquisitionOne-wayanalysisofvariance(ANOVA)wasperformedonthesensory,instrumentalandcompositionalparameters;alsoprincipalcomponentanalysis(PCA)wasdonetoillustratethecorrelationamongthem.BothwereperformedwiththeuseoftheSPSS12package.TherotationmethodusedwasVarimaxwithKaisernormalization,andloadingsweretakenintoaccountiftheirabsolutevaluewashigherthan0.6(TableIV).ThederivativeanalysiswasperformedwithOriginLab7.ThedataanalysismethoddevelopedinChenetal.[2]wasfollowedbutwiththefollowingimportantdifferencesandcaveats.Thesecondderivativeoftheforcecurves(using11-pointSavitzky-Golaysmoothing)wasusedtostudythecorrelationwiththesoundplots.3.RESULTSANDDISCUSSION3.1.Analysisoftheforce/displacementandacousticplotsThetestspeed1mm/swasselectedbasedonprevioustests,reachingacompromiseofhavingreproducibleanddis-criminativeforcecurves,wherethefractureeventsarecontrolled,andalsowheresoundpeakscouldbedistin-guished.Figure2showstwoexamples(rawand6-minroastedsamples)oftheforceversusdisplacementoftheprobeandtheSPLversusdisplacementplots.Inbothcasesthecurvesaredisplayedtogetherasitappearedinthecontrollercomputerscreenwhenperformingthetest,thesignalsweresynchronized,allowingthecomparisoninrealtimeoftheforceandsounddata.Theforcecurvespresentedtworegions,therstpart,wheretheforceincreaseswiththedisplacement,andwhichisafunctionoftherigidity,shapeandsizeofthematerial,thesampleissubjectedtodeformationbutnomajorstructuralbreakdownoccurs,sothesampleisacoustically‘veryquiet’[1,2].Roastingcausesadecreaseintoughnessinthematerial:theforceandstrainofthematerialfailuredecreased,therstbreakdownisrecordedatabout200Nintherawsample,whileitoccurredatabout70Ninthe6-minroastedalmond,andatalowerdistanceoftheprobe[2].Thesecondpartofthecurveisconsideredtobeginatthepointwheretherststructuralbreakdownoccursandthesamplestartstobreak.Then,dependingonthebrittlenessofthematerial,asuddendecreaseoftheforcevaluesoccurs,thefracturepropagatesuptoaninhibitionofthefractureduetothepresenceofsomecrackstopper,andthenitstartsagainasthealmondisfurtherdeformed.Attheendofthetest,theforceishigh,asthesampleisbeingfurthercompressedand TableI.DeÞnitionoftheinstrumentalparametersextractedfromtheforce/displacementandsoundpressurelevel(SPL)curvesParameterDenitionAreaAreaoftheforcecurve(relatedtotheworkofcompression)N.secSlopeofthelinearpartoftheforcecurvetotherstforcebreakdownN/secNumberofforcepeaksNumberofpeaksofforce,usingathresholdvalueof0.05N—AveragegradientAveragegradientofallpositiveslopes(throughtopeak)N/secFitdistanceSectionsarecalculatedasalinebetweenthemidpointsoftheforcepeaksandtroughs.Thelengthofeachsectioniscalculatedusingsqr(isthechangeinforceinnewtonsandisthechangedistancemillimetres.ThelengthofeachsectionisthenaddedtogethertoformtheFitDistanceAveragedropoffTheaveragedropindatabetweenconsecutivepeaksandtroughsNLineardistanceThelengthofanimaginarylinejoiningallpointsintheselectedregion.Ahighlyjaggedline,withlotsofuctuationsinforceduetomanyfractureevents,hasalengthmuchlongerthanasmoothlineresultingfromthetestingofasimilar‘soft’productForceatfailureForcevalueoftherstforcebreakdown,usingathresholdvalueof0.05NNDistanceatfailureDistanceoftheprobeintherstforcebreakdownMNumberofsoundpeaksNumberofpeaksofthesoundplot,usingathresholdvalueof2.5dB—MaxSPLMaximumpeakintensityoftheSPLdB2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cemCrispnessassessmentofroastedalmonds313 thedensityincreased[1,17].Comparingbothforcecurves,itisclearthatthe6-minroastedalmondismorebrittle(Figure2(b)):theplotisveryjagged,therearemanysuddendropsinforce,reectingfracturesofrapidpropagation,andmanyacousticeventswererecordedwithinashortperiod.Intherawalmond(Figure2(a)),onlytwosmalldropsinforcecanbedetected,andamuchsmallernumberofacousticeventsweredetected.InbothexamplesshowedinFigure2,foreachmajorforcedropagroupofacousticeventsseemedtohappenandmanysoundeventsdidnotappeartobedirectlyrelatedtodropsinforce.However,theydonothavetobecorrelatedone-to-one,asthesoundemissionistheresultofasuddenreleaseofenergy,whiletheforcecurveisareectionoftheenergyappliedto,orreleasedfromthesample.Duringthedeformationofamaterial,stresswillbuildupinsideit,andthecrackwillstartattheweakestpointasthestressexceedstheyieldpoint,ifthespeedofthecrackpropagationishigh,theenergydissipatedfromthestructuralfailurewillspreadoutasaudibleshockwaves[2,17].Ingeneral,whenacrispmaterialisbroken,itishypothesisedthateachfractureeventcorrespondstoanacousticevent[2,18,19].Afeatureofouranalysistechniqueisthatthemechanicaldataareanalysedchemometricallytogetherwiththeacousticandsensorydata.Todothiswerstderivethesecondderivativeoftheforce-displacementcurve.Wehavefoundthatifthisisnotdone,correlationsbetweenacousticandmechanicaldataaremuchpoorer[2].Thisnecessaryderivatisationmeansthatthemaximummeasurablerateofcrackeventsis45crackeventspersecond.Thisisanunavoidablepricetobepaidifthemechanicalandacousticdataaretobecorrectlyanalysed.Ideally,wewouldliketoincreasethesamplingrateofthemechanicalsystemtotentimesabovethatofthemaximumfrequencydetectableinthehumanmouth.Thesecondsignicantdifferencetoourpreviouswork[2]istheuseofthelogarithmofthemodulusofthenegativevaluesofthesecondderivativeonly,rejectingthepositivevaluesfromtheanalysis.Therearetworeasonsfordoingthis.Firstly,itiswellknownthathumansensoryexperiencegenerallyrespondstostimuluslogarithmically.Thisiswhyacousticpressureisexpressedindecibelsratherthanaspureacousticpressure,sincetheaudibleintensityrangesfrom10to10Watts/m[20].Itisreasonabletoassumethatthisisalsothecasewithforcesinthemouth,whichourmechanicaltestsaremimicking.Secondly,thesecondderivativeoftheforcecurveisameasureoftherateofchangeofelasticenergyinthesampleasitisstressed.However,thesecondderivativereectsboththestorageandthereleaseofenergyanditisclearfromthemechanismsinvolvedinthecreationofcrispysensationsthatthecrackeventsthatgeneratethesesensationswillbecorrelatedintimewithenergyrelease,notenergystorage.Figure3(a)displaysthesecondderivativeoftheforcecurveandtheSPLplotforthe6-minroastedsample.Thecorrespondencebetweentheplotsisstillpoor;however,itseemsclearerjustforthebigcracksalreadyvisibleintheforcecurve.Manysoundpeakscannotbeexplainedbyapplyingthesecondderivativeanalysis.InFigure3(b)weshowthecorrespondenceoftheSPLplottothemechanicaldata,usingthenegativepartofthesecondderivativeoftheforcecurveonly,plottedonalogarithmicscale.IncontrasttoFigure3(a),weseethatnowmostofthesoundpeakshaveacorrespondingpeakinthesecondderivative.Moreover,notonlydoestheoccurrenceofthepeakscorrespondbutalsodoestheshapeofthecurves,ascanbeappreciatedintheenlargedzonepresentedinFigure3(c).Thisisasurprisingnding,asitsuggeststhatmechanicalandacousticdataarecorrelatedanddisplayrelatedinformationifadequatelyanalysed.Thenumberandintensityofacousticeventsregisteredbythedetectorwouldthereforedependonthedevelopmentofthefracture,inparticularontherateofchangeofenergyreleaseofthesamplebeingfractured.Vincent[1,18]statedthatthedisadvantageoftheacousticapproachwasthatthesignalcouldnotbetranslatedintomaterialssciencetermsortheeventscouldnotbeassociatedwithspecicfractureenergies.Onthecontrary,Luytenetal.[17,19]hypothesisedthatenergydissipativeprocesseswerecrucialforunderstandingthecrispy/crunchynatureofsolidfoods,buthithertotheyhavebeenneglectedinmostresearchstudies.Ourintegrativedataanalysisapproachisastepinthatdirection,relatingsoundemissionwithenergyrelease.Fromtheplotobtained(Figure3(b))thepeakswerecounted,usingathresholdof10,determinedbythederivationoftherstpartoftheforcecurve(secondderivative),beforetherstbreakdown,wheretherewasnoacousticeventdetected. Figure2.Force(blackline)andsoundpressurelevel(SPL;greyline)versusdistanceincompression:(a)rawalmond,(b)6-minroasted.2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cem314P.Varelaetal. 3.2.AnalysisofthemechanicalandacousticFromtheparameterscalculatedfromthecurves(TableII)itcanbeappreciatedthatthevalueoftherstforcebreakdown(forceatfailure)wassignicantlyloweredwithroastingand Figure3.Correspondenceofforce(blacklines)andSPLplots(greylines),exampleforthe6-minroastedsample,(a)usingthesecondderivativeoftheforcecurve,(b)usingthenegativevaluesofthesecondderivativeoftheforcecurve,(c)EnlargedsectionofFigure3(b),correspondenceoftheshapesoftheforceandSPLplots,usingthenegativevaluesofthesecondderivative. TableII.Meanvaluesoftheinstrumentalparametersextractedfromtheforce/distance,forcederivative/distanceandtheSPL/distanceplotsSampleArea(N.sec)(N/sec)NumberofforcepeaksAveragegradient(N/sec)distanceAveragedropoff(N)Lineardistance(N.sec)Forceatfailure(N)Distanceatfailure(m)NumberofpeaksofthederivativeNumberofsoundpeaksSPL(dB)Raw609.3198.11.5min789.5239.73.0min863.2157.74.5min817.7169.2105.26.0min605.3102.3Averagesofsixreplicates.Identicallettersindicatethatthereisnosignicantdifferenceat0.05.2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cemCrispnessassessmentofroastedalmonds315 itoccurredearlier,showingenhancedbrittlenessbecauseofthethermalprocessing.However,thevalueoftheslopeoftherstpartoftheforcecurve,whichcanberelatedtotherigidityofthesamples,waslowerintherawsample,butremainedunchangedwiththedifferentroastingtimes.Theworkrequiredtocompressthesamples(areaunderthecurve)increasedinitiallywiththeroastingtimebutstartedtodecreasefrom4.5-minroasting.Thisreectsthechangeofthemechanicalnatureofthealmonds,fromtough(strongandslightlydeformable)tobrittle(hardandweak).The‘averagegradient’and‘tdistance’didnotpresentsignicantdifferencesbetweenthesamples.The‘numberofforcepeaks’andtheparameters‘averagedropoff’and‘lineardistance’alsoshowedtheriseofbrittlenessandcrispnesswithroasting.Thevaluesreectedthefactthattheroastedsamplesundergoalargeramountoffractureevents,relatedtoamorecrispy/crunchytexture[1].Thenumberofpeaksofthesoundplotandthenumberofpeaksofthesecondderivativeplothadsimilartrends,increasingwithroasting,andinagreementwiththemechanicalobser-vations,becausethenumberofpeaksofthesecondderivativewasevenmorediscriminatingbetweensamples.ThemaximumoftheSPLissignicantlyhigherfortheroastedalmondsthantheraw.3.3.Structure,microstructureandcomposition,relationtothefractureWehavestatedthatthesoundemissionincompressiondependsonthemechanicsofthefracture,whichinturndependsontheparametersofthecompressionandontheshape,size,structureandmicrostructureofthesample.Theirregularpeaksobservedintheforcecurvesoftheroastedsamples,characteristicofacrispy/crunchyproduct,reectedmultiplefractureevents,andcouldbearesultofanon-homogeneousmorphology;microstructuralelementssuchaswateroroilphasedistributioncausingtheanisotropyofthestructure.However,theparticularimportanceoftheseelementsintheperceptionofcrispiness/crunchinessisstillunknown[1,17,21].Thus,studyoffoodstructureinrelationtofracturemechanicsisofgreatinterest.Figure4showsphotographsoftwoalmondsamplesfracturedafterthecompressiontest.Agreatdifferenceinthebreakagepatterncanbeseenbetweenthetwosamples.Therawsamplesubjectedtocompressionisdeformedintherststepofthecompressiontest,beingfracturedasaresultofthetensileforcesdevelopedandconcentratedinthetwoendsoftheshortaxis—becauseoftheshapeanddimensionsofthesample—andcausingthe‘opening’ofthesample.Thisbehaviourwasclearlyreectedintheforceplot,whichhasalongelasticrstpart,andtwonoticeablefractureevents,andinthenotveryjaggedacousticplot.Inthecaseofthe6-minroastedalmond,thematerialwasdrasticallychanged,itbecamemuchlessresistanttothecompression,theelasticdeformationzonewassmall,andthematerialcollapsedbecauseofthecompressionforcesonthesurface,fracturinginmanydirections.Therapidbreakdownandthemanyfractureswerereectedintheforceandsoundplots.Fractureinfoodmaterialsstartsininhomogeneitiesandwilltravelquicklyuntilacrackstopper,suchasanairpocket,emptyspaceorstarchgranuleinhibitsit.Inthesecases,heterogeneityisasimportanttocrispnessasbrittleness[1,17].ThemicrostructurewasobservedbyESEMinordertostudywhatcausedthechangeinthefracturebehaviourofthealmondsbecauseofroasting.ESEMisatechniquethatallowssamplestobeobservedinanelectronmicroscopewithminimalpreparation(nodrying,coatingorstaining)otherthanproducingasampleofthecorrectsize.Changesinbiologicaltissuescanbestudiedclosetotheirnativestateandatrelativelyhighresolution[22,23].Figures5(a)andbshowtheESEMphotographsofthefracturedsurfacesofthetwoextremealmondsamples(rawand6-minroasting).Therawsamplepresentedacleanerfracture,moreeven,andsomeofthecellsseemedtohavebeentakenawaytotheothersideofthecut.Thesefactssupporttheideapreviouslystated,thesamplefailedasaresultoftheresultanttensileforcesbecauseoftheshapeofthefooditemasmuchasthefailureofthematerialitself.Theroastedalmondshowedanunevenfracturedsurface,veryheterogeneous,withthepresenceofcellmaterialoutsidethecells.Pascual-Alberoetal.[14]observedthisphenomenonandattributedittoproteinagglomerationsorcoalescedlipidsthatexitedthecellswiththeheatingprocess.Inthepresentwork,basedonlyonESEMdata,wecannottellthedifferencebetweendifferentkindsofcellmaterial,butsomeextracellularmaterialappeared,increasingheterogeneity.Whereasthetotallipidcontentremainedconstantatallroastingtimes(TableIII),ifthisextracellularmaterialwas Figure4.Photographsofthealmondsamplesfracturedaftercompressinginthetextureanalyser,examplesof(a)rawand(b)6-minroasted.2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cem316P.Varelaetal. lipid,itwouldhavereorderedinsidethealmondandwasnotremovedintheroastingprocessasitreachedtheoutersurface(viatheovenrack,containers,manipulations).Theoutersurfacesoftherawand6-minroastedalmondswereobservedbeforefracturing(Figure5(c,d).Therawsamplepresentedaveryhomogeneoussurfacecomposedoflongcellsofuniformsize,orderedinaparallelpattern.Thiscellshapeandorganisationistypicaloftheouterepidermisbeneaththetesta(asthesamplesarepeeled);andithasbeenpreviouslyobservedbySEMinrawalmonds[14]andthesameorganisationbutwithmoreirregularlyoutlinedcellsinrawalmonds[24].Theroastingprocessalsohaditseffectsonthesurface:theescapeofwatervapourdamagedtheepidermis,andsomefailuresandperforationswerenoticed.Pascual-Alberoetal.[14]reportedthesamekindofdamageintheirworkonalmondsroastedat15045minwiththeuseofSEM.TheutilisationofESEMinthisworkpermittedamoredetailedobservationoftheouterepidermisinnon-manipulatedconditionsofthetissue.ESEMensuredthatthechangeswerecausedbytheroastingprocessandnotbythesamplepreparation.Ourobservationsshowedthatthecellslosttheirshapeandincreasedtheirvolume,theparallelpatternofthecellsdisrupted,withsomepartsofthestructureparticularlydamaged;presumablypermittingtheescapeofcytoplasmaticmaterialfromthecellsofbothepidermisandparenchyma.Asthesampleswereheated,thelowamountofwatercontainedinthecellsevaporatesandtendstoexitthemandreachtheexteriorofthealmond[14],thisissupportedbythecompositionaldata(TableIII).Thewatercontentofthesamplessignicantlydecreasedwithallroastingtimes.Inaddition,thetotalvolumeofporeslledwithairisincreasedbyroasting[15].Thesetwouidsexitthealmondcells,exertapressureandcausefailuresinthefoodmaterial.Then,addingupalltheeffects:theriseinheterogeneity,thedamagetotheindividualcellwalls,andthecreationoffracturesbecauseoftheeliminationofwater,explainsenhancedbrittlenessandcrispness,aswellasthefracturebehaviour.Thematerialbecamemuchlessresistanttothecompression,hadasmallelasticdeformation,andcollapsedasaresultofthecompressionforcesonthesurface,fracturingrapidlyandinmanydirections.3.4.SensoryanalysisSoundwasrecordedfortherst15swhiletheassessorschewedthealmonds.Figure6shows,asanexample,twosoundplots(rawand6-minroastedsample)foroneoftheassessors.Theincreasedjaggednessintheroastedalmond’splotisevident.TheSPLdecreasedwiththemasticationinbothcases,thesoundemissionalmostdisappearsattheendofthe15sofchewingtherawsample.InFigure7,thecorrespondenceofthemechanical,acousticandsensorydataisshownbyplottingthenumberofpeaksofthesoundrecordedbothinthesensorytestandinthecompressiontest,andthenumberofpeaksofthenegative Figure5.Microstructure(ESEM)ofthealmondsamples:fracturedsurfaces(a)rawand(b)6-minroasted;externalsurface(c)rawand(d)6-minroasted. TableIII.MeanvaluesofthemoistureandfatcontentsRaw1.5min3.0min4.5min6.0minMoisture(%)6.1Fat(%ondrybasis)57.9Identicallettersindicatethatthereisnosignicantdifferenceat2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cemCrispnessassessmentofroastedalmonds317 partofthesecondderivative.Theoccurrenceofthethreeparameterswasverysimilar.3.5.Correlations:principalcomponentFivecomponentswereextractedthattogetherexplain81.9%ofthevariance(TableIV).Figure8displaysthethree-dimensionalplotofthesamplesinthespacedeterminedbytherstthreeprincipalcomponents(PC),whichtogetherexplain65.7%ofthevariance.Theanalysiswasabletoseparatethesamplesintogroupsaccordingtotheparametersthataccountforthevarianceofeachoftheaxis.TherstPCexplains39.1%ofthevariance,anditiscomposedofacombinationofmechanical,acousticalandsensoryparameters.TherstPCshowedapositivecorrelationtothepanelsensorycrispness,thenumberofpeaksofsound,thenumberofpeaksofthederivative,andthenumberofpeaksofforce,ithasanegativecorrelationtotheforceatfailureandthedistanceatfailure.Thus,thefactorsalreadyexplainedbythestudyoftheforce,acousticsandderivativeplotsareconrmed,andasdiscussedbefore,inagreementwiththeexplanationsfoundintheobservationofstructure,microstructureandfracturebehaviour.Inourpreviouswork Figure7.Effectofroastingtimeonthenumberofacousticsignalpeaks:(a)sensorytextureevaluation,(b)instrumentaltextureevaluation,(c)secondderivativeofforcecurve. Figure8.PCA.3D-loadingplotofthesamplesintheÞrstthreeprincipalcomponents,togethertheyexplain65.7%ofthevariance. Figure6.Soundpressurelevel.Exampleofarecordingfromoneoftheassessorswhilechewingthesamples:(a)rawsampleand(b)6-minroastedsample.2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cem318P.Varelaetal. onbiscuits[2],therstprincipalcomponentwasdominatedbyacousticmeasureswhilstthesecondPCwasdominatedbymechanicalones.Incontrast,inalmonds,therstPCcontainsbothacousticandmechanicalmeasures.ThesecondPC(16.8%)includesthreeforceparameters(area,slopeandlineardistance)andthethirdPC(9.7%)containsonlydatafromtheacousticmeasurementsofthesensorypanel(panelnumberofsoundpeaksandpanelmaximumSPL).4.CONCLUSIONSThepresentworkintegratedthestudyofenergydissipativeprocessestotheanalysisofsoundemission,crucialforunderstandingthecrispy/crunchynatureofsolidfoods.Itisclearthatinalmondssensorycrispnessishighlycorrelatedwiththerateofemissionandsizeofacousticpeaksemittedbycracks,aspreviouslyobservedinbiscuits,andclearlydetectableinthenegativepeaksofthesecondderivativeoftheforce-displacementcurve.Thisprovidesaconsistentmechanisticexplanationofcrispness,whichclearlydifferentiatesmaterialswithtwoseparateobjectivemeasures;acousticandmechanical.Macro-andmicro-structuraldatasupportsourconclusions.TherstPCcontainsbothacousticandmechanicalmeasures.Weconcludethereforethatsensorycrispnessisevaluatedbothacousticallyandmechanicallyandthatbothmeasuresarenecessarytomosteffectivelydescribecrispness.Wehavealsoshownthatachemometricapproachinwhichcomplexacousticandmechanicaldataareappro-priatelycombinedstatisticallyprovidesaneffectivewayofobjectivelymeasuringthesensoryparameter‘crispness.’Thereisobviouspotentialforaneconomicandnon-invasiveobjectivemeasureofcrispnesswithpassiveacoustics.AcknowledgementsTheauthorsareindebtedtotheMinisteriodeEducacioCiencia(Spain)forthegrantawardedtoauthorPaulaVarela.WealsowouldliketothankStableMicroSystemsfortechnicalassistance.MalcolmPoveyalsowishestothankRoyGoodaereofManchesterUniversityforpointingouttheutilityofchemometricsinacousticanalysis.1.VincentJFV.Thequanticationofcrispness.J.Sci.Food:162–168.2.ChenJ,KarlssonC,PoveyM.AcousticEnvelopeDetec-torforcrispnessassessmentofbiscuits.J.TextureStudies:139–156.3.SzczesniakA,KleinD.Consumerawarenessoftextureandotherfoodattributes.FoodTechnol.:74–77.4.SzczesniakA.Themeaningoftexturalcharacteristics—J.TextureStudies:51–59.5.VickersZ,BourneMC.Apsycho-acoustictheoryofJ.FoodSci.:1158–1164.6.DrakeBK.Foodcrushingsounds.Anintroductorystudy.J.FoodSci.:233–241.7.SterlingC,SimoneMJ.Crispnessinalmonds.FoodRes.:276–281.8.DuizerL.Areviewofacousticresearchforstudyingthesensoryperceptionofcrisp,crunchyandcracklytex-TrendsFoodSci.Technol.:17–24.9.RodautG,DacremontC,VallesPamiesB,ColasB,LeMesteM.Crispness:acriticalreviewonsensoryandmaterialscienceapproaches.TrendsFoodSci.Technol.:217–227.10.LuytenA,PluterJJ,VanVlietT.Crispy/crunchycrustsofcellularsolidfoods:aliteraturereviewwithdiscussion.J.TextureStudies:445–492.11.CourcouxP,ChaunierL,DellaValleG,LourdinD,SemenouM.Pairedcomparisonsfortheevaluationofcrispnessofcerealakesbyuntrainedassessors:corre-lationwithdescriptiveanalysisandacousticmeasure-J.Chemom.:129–137.12.TeschR,NormandM,PelegM.Comparisonoftheacousticandmechanicalsignaturesoftwocellularcrun-chycerealfoodsatvariouswateractivitylevels.J.Sci.FoodAgr.:347–354. TableIV.PrincipalcomponentanalysisPC1(39.1%)PC2(16.8%)PC3(9.7%)PC4(8.6%)PC5(7.7%)Numberofforcepeaks0.1120.1640.067Averagegradient0.1780.102Fitdistance0.065Averagedropoff0.0740.126Lineardistance0.2410.2140.332Forceatfailure0.1740.0280.3030.230Distanceatfailure0.0190.0650.4300.161Numberofpeaksofthederivative0.2610.2890.1460.068Numberofsoundpeaks0.0360.3060.0750.043MaxSPL0.5340.559Panelsensorycrispness0.2480.256Panelnumberofsoundpeaks0.4480.0740.230PanelMaxSPL0.2100.273Rotatedcomponentmatrix.Extractionmethod:Principalcomponentanalysis.Rotationmethod:VarimaxwithKaisernormalization.Rotationconvergedin16iterations.Correlationsweretakenintoaccountiftheirabsolutevalueswere0.6(numbersinboldprint).2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cemCrispnessassessmentofroastedalmonds319 [13].FAO,FaostatStatisticsDatabase.LastupdatedFebruary2007.URLhttp://faostat.fao.org/site/567/14.Pascual-AlberoM,Perez-MunueraI,LluchMA.Cotyle-donstructureofraw,soakedandroastedalmond(nusamygdalusFoodSci.Technol.Int.:189–197.15.GouP,GuerreroL,ValeroA,ArnauJ,RomeroA.Physico-chemicalandsensorypropertychangesinalmondsofDesmayoLarguetavarietyduringtoasting.FoodSci.Technol.Int.:1–7.AssociationofOfÞcialAnalyticalChemists,OfÞcialMethodsofAnalysis.17thedition,AOAC,Gaithersburg,MD.17.LuytenH,PlijterJ,VanVlietT.Understandingthesensoryattributescrispyandcrunchy:anintegratedIn:Proceedingsofthe3rdInternationalSym-posiumonFoodRheologyandStructure,FischerP,MartiI,WindhabEJ(Eds.).Zurich,ISBN3-905609-19-3,18.VincentJFV.Thequanticationofcrispness.J.Sci.Food:162–168.19.VanVlietT,LuytenH.Fracturemechanicsofsolidfoods.Newphysicochemicaltechniquesforthecharacterizationofcomplexfoodsystems,EDickinson(ed.).Chapman&Hall:Cambridge,UK,1995;157–176.20.KinslerLE,FreyAR.Fundamentalsofacoustics,2ndedi-tion.JohnWiley&Sons,Inc.:NewYork,1962;108–127,21.LillfordP,vanVlietT,vandeVeldeF.Discussionsessiononsolidfoods.FoodHydrocol.:432–437.22.DonaldAM,BakerFS,SmithAC,WaldronKW.Fractureofplanttissuesandwallsasvisualizedbyenvironmentalscanningelectronmicroscopy.AnnalsBotany23.DangJMC,CopelandL.Studiesofthefracturesurfaceofricegrainsusingenvironmentalscanningelectronmicro-J.Sci.FoodAgr.:707–713.24.YoungCT,SchadelWE,PateeHE,SandersTH.Themicrostructureofalmond(Prunusdulcis(Mill.)D.A.Webbcv.‘Nonpareil’)cotyledon.LWT-FoodSci.Technol:317–322.2007JohnWiley&Sons,Ltd.J.Chemometrics:311–320DOI:10.1002/cem320P.Varelaetal.