a critical review of the methodology for the study of secular … … · anthropology data bank....

Albanese TheStudyofSecularChangeUsingSkeletalData 139

A Critical Review of the Methodology for the Study of Secular Change Using Skeletal Data

JohnAlbanese

Secularchangesorseculartrendsarenon-geneticchangesthatoccurovermultiplegenerationsinapopulation.Thechangesarenotduetoevolutionbecause there are no corresponding changes inallelefrequenciesinthepopulation.Thesechangestypicallycorrespondwithchangesinlivingcondi-tionsandarereflectedinthepopulationmeansforgrowth anddevelopment, andnotnecessarily inany one individual. The most studied secularchangeischangeinstatureorheight.Improvementsin living conditions have resulted in measurableincreases or positive secular changes in staturefromparentstochildren.Themostobviouscaseshavebeenobservedwhencomparingfirstgenera-tion immigrants to their children. Conversely,decreases or negative secular changes in meanpopulationstaturefromparentstochildrenhavebeendocumentedwhenlivingconditionsworsendue tonaturaldisasters,wars,orprolongedeco-nomichardships.

Severalsourcesofdatacanbeusedforthestudyofsecularchangeinstatureandbodysizeinclud-inghistorical staturedata (Steckel1994) suchasrecruitment records (for example, Floud 1994;Mokyr and Gráda 1994) and stature measure-mentscollectedoncross-sectionalsamplesspecifi-callyforresearchonstature(forexample,Brauer1982; Tobias 1986). Skeletal data are underuti-lizedyetpotentiallyusefulsourcesofinformationforthestudyofsecularchangeinstatureaswellasbodyproportions.Witha fewexceptions (Angel1976;Jantz2001;JantzandMeadowsJantz2000;MeadowsJantzandJantz1999;OusleyandJantz1998; Tobias 1986; Tobias and Netscher 1977;Trotter and Gleser 1951) osteometric data havenotbeenusedextensivelyforthispurpose.Inallof these studies a similar methodology has beenused to investigate secular change using skeletaldata. The purpose of this paper is to critically

reviewthismethodologyandunderlyingtheoreti-cal conceptsusingempiricaldata (femur length)from three reference collections: the Robert J.Terry Anatomical Collection, the CoimbraIdentified Skeletal Collection and the ForensicAnthropologyDataBank.Theprimaryfocuswillbetoinvestigatetheeffectsofthecurrentstandardapproach of combining samples from differentreferencesources intoracialgroupsforthestudyof secular change. Because of the similaritiesamongthesourcesofdata,closecomparisonsaremadebetweentheresultsfromthecurrentstudyand Meadows Jantz and Jantz (1999), and to alesser extentwithTrotter andGleser (1951)andAngel(1976).

Investigation of Secular Change Using Osteometric Data

Longbonelengthdatacanbeusedtoassessstat-ure and secular change. Using in vivo staturemeasurements and long bone measurements,TrotterandGleser(1951)demonstratethattrendsderivedfromstaturedataandlongbonedataaresynchronized.However,themorerobustapproachistouselongbonelengthdatasincethederivedstature will result in additional sources of errorinherentinstatureestimationequationsandcor-rectionfortheeffectsofage(Tobias1986;TrotterandGleser1951).Longbonedatamaybesupe-rior to historical stature data for the followingreasons:1) Changesinlimbandbodyproportionscanbe

investigated along with absolute changes inany given skeletal element (Angel 1976;Meadows Jantz and Jantz 1999). Althoughstaturemeasurementshavebeencollected forcenturies as a biometric component of per-

140 OntarioArchaeologyNo.85-88/LondonChapterOASOccasionalPublicationNo.9

sonal identification in North America andelsewhere(Steckel1994),otheranthropomet-ricdataarerelativelyscarce(arareexceptionisGreinerandGordon1992).

2)Theuseofmodern,calibratedequipmentpro-duces greater precision of measurement(MeadowsJantzandJantz1999).

3)Accuracy related to the rounding of staturemeasurements in historical records is not asourceoferrorwithskeletaldata.Roundingtowhole inches, feetormeters andcertainages(evennumber,multiplesoffive,numbersend-inginzero)iscommoninmanydatasources,includingsomemodernstudies(Steckel1994).Althoughthisproblemcanbecorrectedstatis-tically(Steckel1994),roundingcanstillmask,ormagnify,modestpositiveornegativechang-esinstature.

4) It is possible to test and control for intra-observermeasurementerrorandinter-observermeasurement error if more than one personcollectsdata.

5) Sourcesofmeasurementerrorassociatedwithage, posture, or even wearing shoes duringmeasurementarenotanissuewithlongbonelengthdata (Brauer1982;Trotter andGleser1951).

The effects of age must be controlled in anyinvestigation of secular change involving livingstatureinacross-sectionalsample(Brauer1982).Forexample,Tobias(1986)foundthataftercor-recting for the effectsof ageona cross-sectionalsample,apositivesecularchangeof17mmoverthefirstfivedecadesofthe20thcenturywasactu-ally anegative secular changeof2mmover thesame period. Because the sample was cross-sec-tional, those individuals in the earliest birthcohorts were also the oldest and were affectedmost by age-related changes (compression ofjoints,posture,etc.).Inananalysisofuncorrecteddatafromacross-sectionalsampleitisimpossibleto separate age effects on stature from secularchangeinstature(TrotterandGleser1951).

With both skeletal and documented staturedata,samplinganddefiningtheparametersofthepopulation can be problematic. For example,some of the biases of historical military staturedataareobvious:onlymalesareincluded.Other

possiblebiasesinmilitarysamplesaremuchmorecomplex. The level of representativeness of themilitary sample of thepopulationwill vary overtime depending on approaches to recruiting,changes in entrance requirements, variation insocio-cultural and economic motivation for vol-unteering,andthephasinginandoutofmanda-toryservice(GreinerandGordon1992).Similarly,skeletalreferencecollectionsandcemeterysamplesarenotrandomsamplesofthepopulationsfromwhichtheyweredrawn(Albanese2003;HuntandAlbanese2005).Afurthercomplicationisthattheacquisitionandcataloguingofskeletons inrefer-encecollectionsusuallyoccursoverashortperiodproducing a collection with a relatively narrowrange of years of birth. Excavations of specificcemeteriestoformcollectionsmayresultinsimi-larbiases.Inseveralstudies,samplesfromdiffer-ent sources have been combined to extend therangeofyearsofbirththatarecoveredbyanyonesample. For example, Angel (1976) used datafrommanysourcestoinvestigatesecularchangeinthe entire skeleton in American “Blacks” and“Whites”intwoperiods.Thefirstperiod,referredtoasColonial-AmericanCivilWar(1675-1879),was composed of samples from over two dozenfamily plots or small burial grounds in over adozen states including New Mexico, Texas,Montana,NewYork,VirginiaandMaryland.Thesecondperiodreferredtoasmodernmiddleclass(withyearsofdeathbetween1950and1975),wascomposedofasmallnumberofbequeathedindi-vidualsfromtheTerryCollection,and163foren-siccaseswheredeathwasaccidental.

In a more recent study, Meadows Jantz andJantz(1999)usedskeletaldatatoinvestigatesecu-larchangeinbonelengthsandlimbproportionsusing six long bones: humerus, radius, ulna,femur,tibia,andfibula.FollowingAngel’s(1976)approach,theycombineddatafromfourdifferentsources (inoverlappingchronologicalorder): theHuntington Collection, the Terry Collection,World War II Casualties, and the ForensicAnthropologyDataBank(FDB).Thedatafromthesesourceswerecombinedintotworacialcate-gories,“Black”and“White,”coveringarangeofyearsofbirthfrom1800to1979.Theyfoundthatsecular change was significantly greater in males


thanfemales,andgreaterinthelowerlimbbonesthan the upper limb bones. Differences weregreater in “Whites” than “Blacks” but not at astatisticallysignificantlevel.

Regardlessofwhetherskeletalorhistoricalstat-ure data are used, the approach to sample con-structionanddefiningtheparametersofthepop-ulationarecriticalandaffectinferencesofsecularchange. There are several methodological prob-lems that can produce misleading results whensamples from different sources, which may notrepresent the samepopulation, are grouped intoracialcategories.Amalgamating these individualsintoasinglesamplecanconfoundinterpretationsofsecularchangebecauseitisnotclearifthedif-ferent sourcesofdata thatare sampled representthe same population. There are several possibleexplanationsorinterpretationsifsecularchangeisobservedinacombinedsample.Apparentchangethroughtimecanbecausedby:1)secularchangedueto improvingorworseningconditions if thedifferent samples represent the samepopulation,2)geneticvariation indifferentpopulations thatjusthappentobefromdifferenttimeperiods,3)differences in livingconditionsbetweentwodif-ferent populations that are coincidently separateintime,and4)anycombinationoftheabove.

IncontrasttoAngel(1976)andMeadowsJantzandJantz(1999),TrotterandGleser(1951)useda different approach. They investigated secularchange using two samples in two periods: theTerryCollection (yearsofbirth1840-1909)andWorldWarIIcasualties(1900-1924).Unlikelaterinvestigationsofsecularchange,theseauthorsdidnot combine the samples but instead comparedthepatterninthedifferentsamples:

no attempthasbeenmade todelineate acontinuouscurveforeitherstatureorbonelength based on data from the TerryCollection and from military personnel...the two groups are not comparable in many respects.Nevertheless,thetrendspresentedby separate curves for the two sources ofdatamaybecomparedandareseentodif-fer[TrotterandGleser1951:437;empha-sisadded].

Combiningsamplesfromdifferentsourcescanconfoundinterpretations.

Dividingsamplesfromreferencecollectionsby“race” can produce equally misleading results.Overwhelming evidence indicates that the bio-logical race concept is not applicable to Homosapiens and that racial categories are not usefulresearchtoolsforthestudyofphenotypicorgeno-typic variation (Armelagos and Goodman 1998;BrownandArmelagos2001;Cartmill1998;Keitaand Kittles 1997; Lieberman 2001; Relethford2001,2002;Templeton1998).Classificationintoracialgroupshasvarieddependingonthenumberofcategories in the racial scheme,andthe socialcriteriausedtosegregate“races,”whichhavealsovariedover time and space inbothpopular andacademiccircles(ArmelagosandGoodman1998;KeitaandKittles1997):“individualscanchangetheir racebygettingonaplaneand flying fromNew York to Salvador or Port-au-Prince... whatchanges isnot theirphysical appearancebut thefolktaxonomiesbywhichtheyareclassified”(Fish1999:198). In a recent meta-analysis of genetic,proteinandenzymevariationwithinandbetweenracialgroups,BrownandArmelagos(2001)dem-onstratethatconsideringsamplesinracialcatego-riesdoesnotcontrolforgeneticdifferences.Evenstudieswhoseaimwasto“maximizetheamountof variance accounted for by race” (for exampleHartmann et al. 1994; Nei and Roychouldhury1982; Ryman et al. 1983) found clear evidencesupportingthenegativeconclusionsregardingraceand genetics established over 30 years ago byLivingstone(1962)andLewontin(1972).Usingacombinationofgeneticevidenceandcraniometricdata, Relethford (1994, 2001, 2002) reaffirmedthese conclusions: 1) there is much more intra-race variation than inter-race variation; 2) raceaccounts for only about 6-13% of genetic andcraniometricvariation;3)thereisnoconcordanceof human (genetic and craniometric) variationwith racial categories, continental origin or skinpigmentation.

Inthecontextofanydiscussionregardingskel-etal reference collections, it is important to con-siderthattheterms“White,”“Black,”“Negroid,”“Caucasoid,”etc.weretermsappliedbythepeoplewhowerecollectingskeletons(ordata)attheend


ofthe19thcenturyandthroughoutthe20thcen-tury based on changing social criteria and per-ceived phenotypic criteria.The phenotypic traitsaredescribedasperceivedbecauseracialcategoriescramcontinuousvariationintodiscretecategories.Usingamorebenignandlesspoliticallychargedexample, what measured stature can separate“short”peoplefrom“tall”people?

Itisproblematictoassumethatthereisconsis-tencyinthecriteriausedbyphysicalanthropolo-gistsoveracenturytoassignanygivenindividualto a specific racial group. It cannot be assumedthat the criteria used by Huntington to classifysomeone in 1895 are the same criteria used byTerryin1935,orTrotterin1955.Evenwithinanyonecollection thereare likelymany inconsisten-cies in the criteria used for racial designation.Popularandacademicconceptsofracechangedagreatdealover thesixandahalfdecadesduringwhich the Terry Collection was created. Racialdesignation is even more complex in the FDBbecauseracialdesignationisbasedonselfreport-ing on ante-mortem documents. For theHuntington collection, the nationality ofEuropean individuals was documented whenknown. After it was moved to the SmithsonianInstitution,Hrdlička(1934)begandifferentiatingbetween“Whites”bornintheU.S.A.andrecentEuropeanimmigrantstotheU.S.A.

WhenTerrywasdesignatingrace,hewastryingto categorize continuous human variation (e.g.Terry 1932); racial designation in the FDB wasbasedonanindividual’sperceivedplaceinasoci-ety; and racial designation in the HuntingtonCollectionwasascribedbyHrdlička(butcurrentresearchersarenot limitedbyhisclassifications).The racial termsused todescribe individualsdonotnecessarilyhavethesamesocialconnotationsineachcollectionorevenwithinanyonecollec-tionanditisnotpossibletoreconstructhowanyofthecollectorsmayhaveusedthesetermswhenclassifying any one individual. For the majorAmerican anatomical collections (including theTerry and Huntington Collections), St. Hoymeand Işcan (1989:61) describe the racial designa-tionas“socialorlegal,notbiological,assessments,basedonlocalcustom.”Evenifthesamecriteriafor classification were used, skin colour is not a

proxy for either genetic or skeletal variation(Relethford2002).

Methodologicalproblemsassociatedwithusingracialcategoriesforthestudyofsecularchangeorhuman variation in general are illustrated in thefollowing analysis. Quotation marks are usedaround racial terms when referring to thecategorization of individuals in the variousreferencecollectionstodrawattentiontothepointthatthesetermshavedifferentmeaningsineachofthe sources of data and for different individualswithin each source of data. Unless otherwisestated,thetermpopulationreferstothestatisticaldefinition of population and not its biologicalmeaning. A biological population is a group ofinterbreedingindividualsthatisrelativelyisolatedfrom other similar groups in a species (Molnar2002), whereas statistically, “a population alwaysmeansthetotalityofindividualobservationsaboutwhichinferencesaretobemade,existinganywhereintheworldoratleastwithinadefinitelyspecifiedsamplingareainspaceandtime”(SokelandRohlf1973:7).Thedistinctionbetweenabiologicalandstatistical population is made because there isoverwhelming genetic and phenotypic evidencethat racial groups do not represent biologicalpopulations.

Materials and Methods

Sources of Skeletal DataThereissomeoverlapamongthesamplesourcesusedbyMeadowsJantzandJantz (1999),Angel(1976), Trotter and Gleser (1951) and in thisinvestigation.The four inquiriesdrawdata fromtheTerryCollection and all but theTrotter andGleser study use data from the FDB. For thisstudy,datawerealsocollectedfromtheCoimbraCollection, a cemetery reference collection fromPortugal.SeeTable1fordetailsregardingsamplesizesusedinthecurrentstudy.OusleyandJantz(1998) and Meadows Jantz and Jantz (1999)describetheFDBinsomedetail.ThesinglelargestcontributortothedatabaseisLawrenceAngel(n= 182 cases) and the data bank undoubtedlyincludes many of the individuals used by Angel(1976)inhisinvestigationofsecularchange.TheFDBhasdata from individuals fromalmost the


entire U.S.A. with a bias towards southeastern,northeastern,andsouthwesternstates.

TheTerryCollectionwascollectedbyRobertJ.TerryandMildredTrotteratthemedicalschoolatWashington University in St. Louis, Missouri,from the end of the second decade of the 20thcenturyuntil1967.CollectionpracticeswereverydifferentunderTerryandTrotter.Terry’sinterestinnormalhumanskeletalvariationresultedinalargesampleofindividualsbeingincludedinthecollec-tionwithnogrosspathologicalconditions.AfterTerry’s retirement in 1941, Trotter concentratedonincludingalmostexclusivelyyounger“White”femalesinordertobalancethedemographicpro-file of the collection. In the collection, years ofbirthrangefrom1828to1943withthemajorityfalling between 1850 and 1920 (see Hunt andAlbanese 2005 for more information about theTerryCollection).

TheTerryCollectionwasderivedfromanatomyschoolcadavers.MostwereunclaimedbodiesfromvarioushospitalsandinstitutionsinSt.LouisandtheStateofMissouriwhodidnothavethemeanstopayfortheirownburial.Thisfactalonestrong-lysuggeststhatmostindividualsinthecollectionwereoflowsocio-economicstatus,atleastatthetimeofdeath.Onlyarelativelysmallnumberwhodiedafter1955, about10%of the entire collec-tion,werepeoplewhobequeathedtheirbodiesforscientificresearch.Angel(1976)drewhissample(19femalesand6males)forthestudyofsecularchange from this small segment of the TerryCollectioninanefforttosamplewhathedescribedasmodernmiddleclassAmericans.Moredetaileddataonplaceofbirthareavailableforaseriesof107individualsintheTerryCollectionwhodiedbetweenJuly,1926,andMarch,1928.Althoughsome of the individuals in this sub-sample wereEuropean immigrants or born outside the state,the single most common birthplace is Missouri.Sixty percent of the sub-sample were born inMissouriorthesemicircleofstatestoitssouthandeast including Arkansas, Louisiana, Tennessee,Mississippi,Alabama,Georgia,andKentucky.

TheCoimbraCollection(n=505),iscuratedattheMuseumofAnthropologyattheUniversityofCoimbra, Coimbra, Portugal. This collectionconsists of individuals who died between 1904

and 1936 and who were excavated from thecommon burial ground at the Cemitério deConchada in the city of Coimbra (see Cunha1995;Rocha1995formoreinformationaboutthecollection).Thetypeofburialisstrongevidenceoflow socio-economic status, and the availableinformation on cause of death and occupationconfirm this assessment (Cunha 1995).Informationregardingnativityisavailablefor501of505individualsintheCoimbraCollection.ThesinglemostfrequentplaceofbirthistheDistrictof Coimbra, and 68% of the individuals in thecollectionwerebornintheDistrictofCoimbraorthesurroundingdistrictsthatshareaborderwithCoimbraincludingAvieroandViseutothenorth,GuardaandCasteloBrancototheeast,andLeiriatothesouth.

Thesethreesampleshaveoverlappingrangesofyearsofbirth that cover aperiod from the thirddecadeofthe19thcenturytothelastdecadeofthe20thcentury.Therangeofyearsofbirthsampledin my study roughly corresponds with previousresearch:1841to1977.However,MeadowsJantzandJantz(1999)usedatafromAmericanWorldWarIIcasualtiesfromthePacificTheatercollectedbyTrotterduringrepatriationoftheremains.Thelattersourceisreflectedinthelargesamplesize(n> 460 for the femur) for “White” males for the1910-1919 and 1920-1929 cohorts (MeadowsJantzandJantz1999:Table1).AportionofthesemilitarydatawereoriginallyusedbyTrotterandGleser(1951)intheirstudyofsecularchange.

A seconddifference existsbetween the earliestpartoftherangeofyearsofbirth.TheCoimbraCollectionwasusedinmystudy.TheHuntingtonCollectionwasusedbyMeadowsJantzandJantzandwasamassedbyGeorgeS.Huntingtonattheendofthe19thcenturyandbeginningofthe20thcenturyattheCollegeofPhysiciansandSurgeons,NewYork and transferred after his death to theSmithsonian Institution’s National Museum ofNatural History in 1927 (Hunt and Albanese2005).TheHuntingtonandCoimbraCollectionsaresimilarinclasscompositionandtemporalrep-resentation.TheHuntingdonCollectionismadeupofpeopleoflowersocio-economicclasseswhoimmigrated from Europe at the end of the 19thcentury(MeadowsJantzandJantz1999:59)and


the Coimbra Collection consists of Portuguesepeople of lower socio-economic classes whoremained in Europe from about the same timeperiod.Thereare,however, someobviousdiffer-encesbetweentheseCollections.TheHuntingtonCollection was derived from cadavers that wereused for anatomical instruction, whereas theCoimbra Collection was excavated from a com-monburialground.Neithercollectionisnecessar-ily representative of Europeans (geopolitical orracially). Most importantly neither should becombinedwith“Whites”fromtheTerryCollectionortheFDBforthestudyofsecularchange.

Data Collection and Sample SizeAlthough other researchers have used combina-tionsoflongbonelengths,cranialmeasurementsand various indices, in this study only femurlength data are used. If it can be demonstratedthattherearemethodologicalproblemswithcom-bining samples fromdifferent sources into racialgroups using femur length measurements, thenthe same should be true regardless of which variable is used.Thegoalofthispaperistohighlightprob-lems associated with sample construction andpopulation parameters rather than to investigatesecularchangeinanygivenvariable.

Maximum femur length was collected by theauthor fromboth theTerry (n=327)and fromthe Coimbra (n = 237) Collections, while datafromtheFDB(n=317)werecollectedbydiffer-ent contributors to the FDB.The samples weredividedintofive25-yearbirthcohortsinordertomaximizethesamplesizeswithineachcohort.Thedateofbirthisnotreadilyavailableformostofthe

individualsintheTerryandCoimbraCollections.Yearofbirthwascalculatedbysubtractingtheageatdeathfromtheyearofdeathforeachindividu-al.Individualswhoseyearofdeathorageatdeathwereinanywaysuspectwhereexcludedfromtheanalysis (see Hunt and Albanese 2005 for moreinformation on assessing the accuracy of thesedata for the Terry Collection). Individuals wereincluded in the FDB sample only if they werepositively identified. For the samples from allthree data sources, ages are between 18 and 80years. Details on the composition of the threesamplesbysexand“race”(asdescribedbythecol-lectors) for each birth cohort are available inTable1.

Intra-observer and Inter-Observer Measurement ErrorFemur length data were re-collected for a sub-sample of 66 individuals from the CoimbraCollection(n=13malesand13female)andtheTerryCollection(n=20malesand20females)toassess intra-observer measurement error. Percentintra-observer measurement error was calculatedusingthefollowingequation:

Percentintra-observererrorwasdeterminedforeach of the 66 individuals. Mean percent intra-observermeasurementerroris0.06%witharangefrom0to0.46%.

Some of the cases in the FDB (n = 135) areTerryCollectionindividualswhowerebornafter1898. Eight of these individuals from the FDB

%Intra-observerError=AbsoluteValue(Measurement1-Measurement2x100Measurement1

Coimbra Collectiona Terry Collectionb Forensic Data Bankb Total

BirthCohort F M WF WM BF BM WF WM BF BM WF WM BF BM

<1875 45 46 11 20 9 14 56 66 9 14

1875-1899 59 58 20 20 44 41 1 1 1 80 79 45 41

1900-1924 15 14 39 13 48 34 11 33 18 32 65 60 66 66

1925-1949 3 7 2 2 19 52 8 19 22 59 10 21

1950+ 40 52 10 18 40 52 10 18

Total 119 118 73 60 103 91 71 138 37 69 263 316 140 160

Table 1.Composition of the samples from three sources by sex and “race” for each birth cohort.

aUsingcurrentsocialcriteriainPortugal,alltheindividualsintheCoimbraCollectionsamplewouldbeconsidered“White.”b“Race”wasassignedatthetimeeachindividualwasincludedinthecollection.Note:F=female,M=male,WF=“White”female,WM=“White”male,BF=“Black”female,BM=“Black”male.


overlapwith the intra-observer error samplecol-lectedbytheauthor,andtherefore,itispossibletoassess the level of inter-observer measurementerror. Following a similar approach used for theassessment of intra-observer error, the percentinter-observer errorwas calculatedusing the fol-lowingequation:

Aswiththeintra-observererror,themeaninter-observer measurement error is also 0.06%. Therange of inter-observer error is 0 to 0.23%.Although the sample size for the assessment ofinter-observermeasurement is small, it follows apatternsimilartotheintra-observermeasurementerrorandsuggestsoverall that theeffectof thesekindsofmeasurementerrorsisnegligible.NoneoftheindividualsfromtheTerryCollectionthatarerepresented in the FDB and collected by thisauthorwereincludedintheFDBsampleusedinthisstudy.

Statistical ApproachOne-way analysis of variance (ANOVA) proce-durewasusedinthisstudybecauseposthoctestscan be used to assess significant differencesbetween specific birth cohorts and sub-sampleswithin each birth cohort. Statistical tests usingone-wayANOVAwereconducted inthreesepa-rate phases using SPSS 9.0 for Windows (SPSS1998). First, the samples from the three datasources (the Coimbra Collection, the TerryCollectionandtheFDB)werecombinedbyprevi-ouslydefined“race”inordertolookforsignificantdifferencesinmeanfemurlengthovertimesepa-rately for “Black”males, “White”males, “Black”femalesand“White”females.Thisfirstphasewasanattempttoreproducethepreviousapproachesthathaveusedskeletaldataforthestudyofsecularchange(forexample,Angel1976;MeadowsJantzandJantz1999;OusleyandJantz1998). In thesecondphaseoftesting,datafromeachcollectionwere analyzed separately in order to test for sig-nificant differences in mean femur length overtimewithineachcollection.Forexample,istherea significant secular change in femur length inCoimbra Collection males? In the third phase,birth cohort was held constant and variation

within each cohort was compared in order toassesswhetheritisappropriatetodividethesam-plesby “race.”For example, are there significantdifferences in mean femur length among TerryCollection“Black”males,“White”males,“Black”females and “White” females, and Coimbrafemales and Coimbra males in the 1875-1899cohort?

Inall threephasesTukey’sHSD(honestlysig-nificantdifference)testwasusedpost hoc.Anum-ber of different post hoc tests can be used withone-wayANOVA.Tukey’sHSDtestwasselectedforseveralreasons.First,asthenameimpliesthetestisneithertooconservative(aswiththeScheffeor Bonferroni tests) nor too liberal (as with theLST,leastsignificanttest) inassessingsignificantdifferenceswhencomparedtootherposthoctests.Second,Tukey’sHSDisbothamultiplecompari-sontest(pairwisecomparisonsaremadebetweenmeans to identify significant differences) and arangetest(similarmeansaregroupedintohomo-geneoussubsets).Thus,whenbirthcohortisheldconstant, it ispossibletoassesswhetherracialorother more relevant criteria should be used todefinesub-samples,andit ispossibletoassess inwhichcohortstherearesignificantchangesinlongbone lengths. InTables 2, 3, and 4, the signifi-cancevaluesatthebottomofeachcolumnindi-cate that there are no significant differencesbetweenthesub-samplemeanslistedinthatcol-umn. These significance values should not beconfusedwith theoverallFandpvalues for theone-wayANOVAanalysis.

Results

Phase IThe mean femur lengths for each sub-samplespanningfivebirthcohortsarepresentedgraphi-cally in Figure 1a.There seems to be a positivesecularchangeinallfoursub-samples,althoughitisslightinsomecases.Separateone-wayANOVAtests for each sexand“race” sub-sample seemtosuggestdifferentpatternsofchangeovertimeforeach “race.” For “White” females, the two birthcohortsbefore1900aresignificantlysmallerthanthe three cohorts after 1900 (F = 14.168, p <

%Intra-observerError=AbsoluteValue(Measurement1-FDBMeasurementx100Measurement1


0.0001). Similarly, for “White” males, the twobirthcohortsbefore1900aresignificantlysmallerthanthethreecohortsafter1900(F=21.860,p<0.0001).“Blacks”seemtofollowadifferentpat-ternthan“Whites.”Thereisnosignificantchangein mean femur length over time for “Black”females(F=1.361,p=0.251)or“Black”males(F= 0.493, p = 0.741). Homogeneous subsets foreach sex and “race” sub-sample are presented inTable2.

Phase IIThepatternof secularchange is similar foreachsex and “race” sub-samplewithin each source ofdata.Thereisnosignificantsecularchangeinanyofthesesub-sampleswhenexaminedseparately.InFigure2a,themeansfor“White”femalesineach

birthcohortareplottedalongwiththemeansfor“White” females separately by data source. Thesignificant positive secular change in “White”females described above disappears when thesample is considered separately by data source.There areno significant changes inmean femurlengthinCoimbraCollectionfemales(F=0.052,p=0.949),TerryCollection“White”females(F=0.166,p=0.919),orFDB“White”females(F=0.481, p = 0.620). In Figure 2b, the means for“White” males in each birth cohort are plottedalong with the means for “White” males sepa-ratelybydatasource.Aswiththe“White”females,thesignificantpositivesecularchangeinthecom-binedsampleof“White”malesdisappearswhenthe samples are considered separately by datasource.Therearenosignificantchangesinmean

“White” Females “Black” Females

Cohort n subset 1 subset 2 Cohort n subset 1

(mean) (mean) (mean)

<1875 56 412 <1875 9 432

1875-1899 80 413 1875-1899 45 439

1900-1924 65 429 1900-1924 66 441

1925-1949 22 436 1925-1949 10 448

1950+ 40 439 1950+ 10 443

sig. 1.000 0.263 sig. 0.483

“White” Males “Black” Males

<1875 66 447 <1875 14 482

1875-1899 79 451 1875-1899 41 478

1900-1924 60 467 1900-1924 66 481

1925-1949 59 476 1925-1949 21 487

1950+ 52 474 1950+ 18 483

sig. 0.915 0.156 sig. 0.714

Table 2. Homogeneous subsets (using Tukey HSD post hoc) of mean femur length for sex and “race” groups divided into five birth cohorts. All means are in mm.

Females Males

Group n subset 1 subset 2 Group n subset 1 subset 2

(mean) (mean) (mean) (mean)

Coimbra 15 407 Coimbra 14 450

Te"White" 39 433 Te"White" 13 470 470

Te"Black" 48 439 Te"Black" 33 473

FDB"White" 11 445 FDB"White"

34 474

FDB"Black" 18 446 FDB"Black" 32 489

sig. 1.000 0.444 sig. 0.089 0.138

Table 3. Homogeneous sub-sets (using Tukey HSD posthoc) of mean femur length for each sex in the 1900-1924 birth cohort. All means are in mm. Te = Terry Collection; FDB = Forensic Anthropology Data Bank.


femur length inCoimbraCollectionmales (F=1.040,p=0.357),TerryCollection“White”males(F=1.119,p=0.349),orFDB“White”males(F=0.144,p=0.866).Asimilarbreakdownbydatasourcefor“Black”femalesandmalesareplottedinFigure2candFigure2d,respectively.Thelackofsignificantdifferences inthecombinedsample isalsoseenwhenthesamplesareseparatedbydatasource.Therearenosignificantsecularchangesinmean femur length of Terry Collection “Black”females (F = 0.587, p = 0.625) or males (F =0.273,p=0.844)norarethechangessignificantforFDB“Black”females(F=0.428,p=0.655)ormales(F=0.381,p=0.684).

Phase IIIWhenbirthcohortisheldconstantandthefemurlengthmeansarecompared, therearenosignifi-cant differences between racial categories in anygivenbirthcohort.Table3includesthehomoge-neous subsets of sub-samples for the1900-1924cohortinwhichallthreesourcesofdataoverlap.The Coimbra females are significantly differentfromthe“White”and“Black”femalesinthebirthcohort(F=7.198,p<0.0001)andtherearenosignificant differences between “Black” and“White”femalesfromtheTerryCollectionortheFDB. The pattern is very similar for males.Coimbra males are significantly different fromTerry Collection “Black” males, FDB “Black”

<1875 1875-1899

Group n subset 1 subset 2 subset 3 subset 4 Group n subset 1 subset 2 subset 3

(mean) (mean) (mean) (mean) (mean) (mean) (mean) (mean) (mean)

CoFe 45 408 CoFe 59 407

Te“White”Fe 11 430 430 Te“White”Fe 20 429

Te“Black”Fe 9 432 Te“Black”Fe 44 438

CoMa 46 445 445 CoMa 58 442

Te“White”Ma 20 464 464 Te“White”Ma 20 463

Te“Black”Ma 14 482 Te“Black”Ma 41 478

sig. 0.075 0.385 0.154 0.256 sig. 1.000 0.197 0.109

1925-1949 1950+

Group n subset 1 subset 2 Group n subset 1 subset 2

FDB“White”Fe 19 437 FDB“White”Fe 40 439

FDB“Black”Fe 8 445 FDB“Black”Fe 10 443

FDB“White”Ma 52 476 FDB“White”Ma 52 474

FDB“Black”Ma 19 488 FDB“Black”Ma 18 483

sig. 0.695 0.407 sig. 0.958 0.608

*Allmeansareinmm.Te=TerryCollection,Co=CoimbraCollection,Fe=female,Ma=Male

Table 4. Homogeneous subsets (using Tukey HSD posthoc) of mean femur length for four birth cohorts.*

Figure 1. Mean femur lengths for each sex and “race” group by birth cohort with Coimbra Collection data (a) and without Coimbra Collection data (b). See Table 1 for sample sizes.


males andFDB“White”males (F=5.655, p<0.0001), and there are no significant differencesbetween “Blacks” and “Whites” from the TerryCollectionortheFDB.HomogeneoussubsetsinTable 3 group Terry Collection “White” maleswithCoimbraCollectionmalesandwithallothermales from the Terry Collection and the FDB.Stated another way, Terry Collection “White”malesarenotsignificantlydifferentfromCoimbramales nor from all other males from the TerryCollectionandtheCoimbracollection.

Resultsfortheotherfourbirthcohortsarepre-sentedinTable4.Thepatternisconsistentforallbirthcohortsexceptfortheslightvariationintheearliestcohort. Intheearliestbirthcohort, thereare significant differences between Coimbrafemales andallother sexand“race” sub-samplesexceptTerryCollection“White”females;CoimbraCollection males are significantly smaller thanTerry Collection “Black” males but not TerryCollection“White”males(apatternsimilartothe1900-1924cohort)(F=32.235,p<0.0001);andthere are no significant differences between“Blacks”and“Whites”of either sex in theTerryCollection. In the 1875-1899 cohort, there are

significant differences between Coimbra femalesandallothersexand“race”sub-samples;CoimbraCollection males are so small that they are notsignificantly different than Terry Collectionfemales regardless of “race” (F = 53.951, p <0.0001); and there are no significant differencesbetween“Blacks”and“Whites”ofeithersexintheTerry Collection. In the two most recent birthcohorts (1925-1949 and 1950+), there are nosignificant differences between “Blacks and“Whites.” Males from the Terry Collection andtheFDBaregroupedtogetherregardlessof“race”and the females from these two sources aregroupedtogetherregardlessof“race.”

Discussion

Despite the differences in the sources of data(Coimbra Collection instead of HuntingtonCollection and lack of World War II casualtydata),theresultsfromthefirstphaseofanalysisaresimilartotheresultsofMeadowsJantzandJantz(1999:Figure1d).Bothstudiesshowoverallposi-tive changes in “Blacks” and “Whites” of both

Figure 2. Femur length means of samples combined by “race” and separately for each source of data: (a) “White” females, (b) “White” males, (c) “Black” females, (d) “Black” males. See Table 1 for sample sizes.


sexesalthoughtheyarenotalwayslargeincreases(aswith“Black”males)orsignificantinthecur-rent study.The greatest similarities occur in thebirth cohorts starting at 1900 in both studiesexceptfor“White”femalesintheMeadowsJantzand Jantz (1999) study. With the exception ofthese“White”females,bothstudiesshowincreas-es in femur length beginning at about 1900(Figure1a).Larger25-year cohorts in this studymakeitappearasiftheincreaseisbeginningear-lier when the data are presented graphically(Figure 1a). Despite the lack of World War IIcasualtydata,thepatternfor“White”malesovertheentirerangeofyearsofbirthincludedinthisstudy closely resembles the pattern described byMeadowsJantzandJantz(1999).Arelativelyflatorslightlynegativetrendchangessuddenlyintoalarge positive trend around 1900. By the fourthdecadeofthe20thcenturyfemurlengthlevelsoffandthenseemstodecreaseslightly.

TheresultsfromphaseIIofthestatisticalanaly-sisclearlyshowthatthesignificantpositivesecularchangeseenin“Whites”andthepositive,butnotsignificant, changes in “Blacks” are the result ofcombiningsamplesthatrepresentdifferentgroupswithdifferentmeanfemurlengthsthatarecoinci-dently separated in time (Figure 2). It is a fairassumption that any significant positive secularchange in the combined sample should alsoappear in each of the different sources of datawhen analyzed separately. For example, the sig-nificantpositivechange in“White” femalesovertime should appear at least as a positive, if notsignificant,changewhenthedatafromthediffer-entdatasourcesareanalyzedseparately.Figure2adoesnotreflectthispattern.Althoughthefemurlengthdoesincreaseslightly(3mm)fortheTerryCollection“White”femalesoverthe85yearscov-ered by the first three birth cohorts, the femurlengthinCoimbraCollectionfemalesdecreasesby1mmoverasimilarperiodandfemurlengthofthe FDB “White” females decreases by 6 mmfrom the 1900-1925 cohort to the most recentcohort.Inallthreesourcesofdata,testinginphaseII shows that slight negative and slight positivechangesarenot significant.The significantposi-tive trend is clearly only an average that resultsfromthecombinationofdatathatrepresentdif-

ferent populations rather than any real secularchange in the combined sample.Thepattern inFigures2band2candthelackofsignificanceofchangeovertimeforanysub-samplewithinadatasource suggest that a similar pattern exists for“White”malesand“Black” females.Thepatternfor“Black”malesinFigure2dappearstobediffer-ent. The two negative, though not significant,trendsintheTerryCollection“Black”malesandFDB“Black”malesresultinafalsecyclicaltrendin mean femur length over time when the datafromthesetwosourcesarecombined.Notsurpris-ingly, the flat lines forTerryCollection“White”and“Black”malesmirrortheresultsofTrotterandGleser (1951) who also used samples from theTerryCollection.

As noted above, there is an increase in femurlengtharound1900inall sub-samplesdescribedin this study and the Meadows Jantz and Jantz(1999) study except for their “White” femalesample.Afteranalyzingthesourcesofdatasepa-rately, it isclearthat inmystudytheincreaseinfemur length occurs in all sex and “race” sub-samples because the 1900-1924 cohort is thetransitionalcohortwherethemainsourcesofdatachange from the Coimbra Collection andTerryCollectionaroundtheturnofthe19thcenturytotheTerryCollection and theFDBat thebegin-ningofthe20thcentury.Similarly,theshiftinthesource of data to almost exclusively the FDB inthe1925-1949birth cohort, and exclusively theFDBinthemostrecentbirthcohort,resultsinanapparenthalt in the increase in femur lengthorevenaslightdecreaseinfemurlength(Figure1a).The samepattern isvisible in threeof four sub-samples in theMeadows Jantz and Jantz (1999)study, which suggests that the changes in thesourceofdataarehavingasimilareffectontheirresults. In the first cohort and then the fourthcohortofthe20thcentury,thesourceofdatashiftsas well.The shift in data source is different formales and females. For “White” and “Black”females, the data source shifts from the TerryCollectionattheendofthe19thcenturytotheTerryCollectionandFDBinthefirstcohortofthe20thcentury,andthentoFDBdataexclusivelyinthefourthcohortofthe20thcentury.Althoughthesourceofdatachangesinthesamecohortsfor


“White”females,thecohortslikelyfollowadiffer-entpatternbecauseofstochasticvariationresult-ing from small sample sizes in the narrower10-yearcohortsusedbyMeadowsJantzandJantz(1999). Their sample size for “White” femalesfluctuatesfrom4to25foreachbirthcohortfrom1840-49to1920-29.

Formales,thedatasourcesshiftfromtheTerryCollectiontotheFDBandWorldWarIIcasual-ties,whichisobviousbecauseofthesuddenjumpinsamplesizeinthe1910-1919and1920-1929cohorts for “White” males (Meadows Jantz andJantz1999:Table1).Thesamplesizefor“Black”males does not increase substantially over thisperiodbecausetherearesignificantlyfewer“Black”WorldWarIIcasualtiesinthesamplecollectedbyTrotter and Gleser (1951:429) and used byMeadowsJantzandJantz(1999).Thesuddenhaltintheincreaseoffemurlengthinthe1930-1939cohort for “White” and “Black” males in theMeadows Jantz and Jantz (1999) study corre-spondswithasecondmajorshiftintheirsourceofdatafromoverwhelminglyWorldWarIIcasualtydatatoexclusivelyFDBdata.Inallbutonecase(“White”femalesintheMeadowsJantzandJantz[1999] studywhere sample size is very small) inboth studies, the large increasesanddecreases infemur length, whether significant or not, corre-spondtochangesinthesourcesofdata.IfTrotterandGleser (1951)hadalsocombined theirdatafromtheTerryCollectionandWorldWarIIcasu-alties, they would have found a similar suddenincreaseinfemurlengthafter1900.

Thefirstphaseofanalysisseemstosuggestthatthere aredifferentpatterns in secular change for“Blacks”and“Whites.”“White”malesandfemalesseemtoshowasignificantsecularchangewhereas,theapparentpositivechangein“Black”malesandfemales is not significant. In Figure 1b, theCoimbraCollectionindividualsareexcludedfromthe “White” male and female samples, and as aresult, the change over time in “Whites” is verysimilar tothe“Black”pattern.Theapparentdif-ferences between “Black” and “White” femalesand“Black”and“White”malesdisappear intheearliest two cohorts (compare Figure 1a withFigure1b).Aswith“Black”males,thedifferenceinmeanfemurlengthovertimeforthiscombina-

tionof“White”malesthatexcludestheCoimbraCollectionsampleisnotsignificant(F=2.038,p=0.091).Similarly,thedifferenceisnotsignificantfor“White”females(F=0.829,p=0.509).Thecombination of data that included the CoimbraCollection in the “White” sub-sample is statisti-callysignificantbecausethemalesandfemalesinthe Coimbra Collection are much smaller andcoincidently have earlier average birth years. InseveralbirthcohortstheCoimbrafemalesaresig-nificantlyshorterthanboththeTerryCollectionfemalesandtheFDBfemales,andCoimbramalesaresignificantlyshorterthantheTerryCollectionand FDB males. In fact, Coimbra males are sosmall that they are consistently and significantlygroupedwithTerryCollectionandFDBfemales(Table4).Thus,thepatternofsecularchangemayseem to be different for “Blacks” and “Whites”dependingonwhichcombinationsofsamplesareused.

Thedifferentsourcesofdata,therefore,makeitimpossible to interpret the cause of the greatersecular change in “White” males. It could be agreater improvement in living conditions for“White” males, a sex difference in response toimprovedlivingconditions,orastatisticalanom-alyresultingfromtheinclusionofalargenumberof“White”malesfromdifferentdatasources.AsMeadowsJantzandJantz(1999)note,itiscriticalthat males and females in a sample come fromsimilarsourcesifsexdifferencesinsecularchangeare being investigated. However, it is absolutelyessential that all the males and all the femalescombined into one sample from different datasources (FDB, Terry Collection, CoimbraCollection,HuntingtonCollection)representthesamegeopolitical, socio-economic andbiologicalpopulationwheninvestigatingsecularchange.

Combiningdatafromdifferentpopulationswillonly produce confounded or misleading results.For example, an analysis that combines theCoimbra Collection sample with the TerryCollection “Whites” provides no basis for infer-encesaboutsecularchangeina“White”popula-tion.Asidefromthepossiblegeneticdifferencesinthe two populations represented in these twosamples,theCoimbraandTerryCollectionssim-ply have too many micro and macro socio-eco-


nomic and geopolitical differences. The TerryCollection(regardlessof“race”)andtheCoimbraCollection are samples of two populations fromMissouri (and surrounding states) and Coimbra(and surrounding districts), respectively, whichhave different biocultural histories.Their differ-encesdonotjustresultfromdifferencesinbirthcohorts.Similarly,combining“Whites”fromtheTerryCollection(mostlyMissouriandneighbour-ing states) with “Whites” from the HuntingtonCollection (mostly first generation Europeanimmigrants)whowherebornonaverageacoupleofdecadesearlierdoesnotrevealanythingaboutsecularchange ina“White”population.Itdoes,however, show that the poorer immigrants whowere eventually included in the HuntingtonCollectionaresmallerthanthepoorerresidentsofMissouriandsurroundingstateswhowereeventu-allyincludedintheTerryCollection.

The clear skeletal differences between theCoimbra and the Terry Collection, when birthcohort isheld constant,do strongly suggest thatthesetwosamplesshouldnotbecombinedforthestudyof secularchange.Similarly, thepattern infemurlengthin“Whites”and“Blacks”(Figure1aandFigure1b)withmajor changesoccurring incohortsthataretransitionalfromonedatasourcetothenextsuggestthattheTerryCollectionandthe FDB may represent different populations.Furthermore, there are many other non-skeletaldifferencesbetweenthesetwosourcesofdatathatmustbeconsidered.TheFDBisaselectivesampleofmostoftheU.S.A.whereastheTerryCollectionis a selective sample of a cluster of about a halfdozenstates.Asidefromsocio-economicandgeo-politicalfactors,complicatedissuesrelatedtohowthecollectionswereamassedmustalsobeconsid-ered. It has been stated by some authors (forexample,Angel1976;Ericksen1982;OusleyandJantz1998)thattheTerryCollectionmaynotberepresentativeofthegreaterAmericanpopulationbecauseitwasderivedfromacadaversample.Thesourcesandmagnitudeofbiasesareactuallymuchmorecomplex.Forexample,changesincollectingpractices by Trotter after Terry’s retirement cor-rectedsomebiasesinthecollectionwhileincreas-ing others (Hunt and Albanese 2005). Thesechanges were due to Trotter’s efforts to correct

whatshesawasashortcominginthecollection,andamassivechangeinpopularviewsofhumandissection and anatomical instruction in theU.S.A. following the Second World War. TheapproachusedbyTerry(andthenTrotterfortheTerryCollection)forcreatingcollectionswasverydifferentthanthatusedfortheFDB(whichalsoincludes a very selective sampling of the TerryCollection).Thespecificbiasesofeachsourceofdatawillhaveadifferent(thoughnotnecessarilynegative) impact on different research questionsdependingontherelativeimportanceofage,yearofbirth,andcauseofdeath.

This study shows that controlling for “race”when constructing a sample or defining theparameters of a population only further con-founds the analysis of secular change.There aremethodologicalcostsforusingracialcategoriesinthis manner. As a methodological tool, the raceconceptisnotausefulcriterionfordefiningsub-samples or populations for the study of secularchange using reference collections. When thefemur length data are presented graphically by“race,”itseemsasif“Black”malesareconsistentlylarger than “White” males and that “Black”females are consistently larger than “White”femalesinanygivencohort.However,theresultspresented inTable4areveryclear. Inanygivencohort,therearenosignificantdifferencesbetween“Blacks”and“Whites”intheTerryCollectionandtheFDB,andthereisnoevidencefromthefemurlengthdatatosuggestthat“Whites”and“Blacks”intheTerryandFDBcollectionsshouldbeana-lyzed separately. Males are consistently groupedtogetherregardlessof“race”andfemalesarecon-stantly grouped together regardless of “race.”Furthermore, this approach is problematic forinvestigating the detrimental effects of racism.Combiningsamplesbyraceinthismannerrevealslittleabouttheeffectsofracismandrelatedsocio-economic disparity on skeletal growth. It is notsafe to assume that the same criteria to describerace were used by Huntington,Terry,Trotter oranyoftheself-reportingintheFDB.Thus,ifthesocialcriteriaforclassificationinanygiven“race”havevariedovertime,thencombiningdatafromdiverse sources that represent different popula-tions where samples were segregated based on


inconsistent social criteria reveals nothing aboutthe impact of racism on living conditions andsecularchange.

Theresultsofthisstudyindicatethatdivisionofsamplesby“race”forthestudyofsecularchangeinbone size isnotmethodologically sound, andthatthereisnoevidencetosuggestthatmalesandfemales should be combined for an analysis ofsecular change. Beyond obvious sexual dimor-phism, there is some evidence that males andfemales do react differently to changes in livingconditions(Brauer1982;Greulich1951;Greulichetal1953;Stini1975,1982;Stinson1985;Tobias1972, 1975). It is possible that these sex differ-enceswillappearinananalysisofsecularchange,although25yearbirthcohortsmaybetoobroadtodetectdifferencesrelatedtospecific,acuteinci-dentsofimprovementorworseningofconditions.However, one essential reason why males andfemalesmustbeanalyzedseparatelyisthatequalnumbersofmalesandfemalesarenotalwaysavail-ableforeachbirthcohortwhenusingdatafromreference collections. Although not all males arelarger than all females, males are significantlylarger than females in femur length and manyother variables. Differences in the proportion ofmales and females fromcohort to cohortwouldresultinwhatmightappeartobesecularchange.

Conclusion

Thereareseveralbenefitstousingskeletaldataforthe study of secular change in stature and limbproportions.Withskeletaldata,differenttypesofmeasurement error can be controlled or mini-mizedanditispossibletoinvestigatechangesinabsolutesize,andbodyandlimbproportionsovertime for both sexes. In this study, femur lengthdatahavebeenusedforillustrativepurposes;how-ever,issuesrelatedtohowsamplesareconstructedforthestudyofsecularchangearerelevantregard-less of which variables are used.This study hasshownthatsomecautionmustbeexercised,andoffersthreemajorconclusions:

1)Combiningdata fromdifferent sourceswith-outconfirmingthatthesamplesrepresentthe

samebioculturalgroupcanproduceverymis-leading results. The focus in this study hasbeen on skeletal data to highlight the differ-encesbetweensourcesofdata;however,othercriteria must be considered before samplesfrom different sources are combined. Socio-economicandgeopoliticalcriteria,andbiasesinthecollectionprocesswhenthedatasourceswereamassed(whichcanamplifyorminimizethe effects of these criteria), must be consid-eredbeforesamplesarecombinedforthestudyof secular change.Combiningdata fromdif-ferent sources without considering these fac-tors will result in other sources of variationbeing attributed to secular change. However,for some research it is very advantageous tocombine data from multiple sources. Forexample,thedifferencesbetweentheCoimbraCollectionandtheTerryCollectioncanactu-allybeexploitedtoconstructareferencesam-pletodevelopmetricsexdeterminationmeth-ods that can be applied successfully to bothsmall and large individuals (Albanese 2001a,2001b,2002,2003).

2)Dividing samples into racial groupsdoesnotcontrol for differences between the samplesources.Thereissomeevidence(documentednationality, place of birth, bequeathed versusnotbequeathed,etc.)thattheTerryCollectionmay not represent a single population, butthereisnoevidencethatdividingsamplesfromthe same data source by “race” in any waycontrolsforthispossiblelackofhomogeneity.There are methodological problems with theuseof the raceconcept todefine samples forthestudyofhumanvariation.Raceisneitherabiologicalconceptthatappliestohumansnorafixedsocialcategory.Variationthroughtimeinthesocialcriteriausedforracialdesignationraises problems even when investigating theeffectsofracismonsecularchangeusingrefer-enceskeletalcollections.

3) Itisessentialthatmalesandfemalesbeconsid-ered separately when investigating secularchange for biological reasons and for reasonsrelated to possible variation over time in thesexproportionsinthestudysample.


Acknowledgements. ThisresearchwassupportedbyfundingfromtheSocialScienceandHumanitiesResearchCouncilofCanada.IwouldliketothanktheMuseumofAnthropologyattheUniversityofCoimbra in Portugal for allowing access to theCoimbra Collection and the curators of theForensicAnthropologyDataBankattheUniversityofTennessee,Knoxville,fordata.Iwouldalsoliketothanktheorganizersofthissessionandvolumeforinvitingmetoparticipateinhonouringagreatanthropologist. And of course, I would like tothankMichaelSpence.HisinfluenceonmycareerbeganwhenIwasanundergraduatestudentattheUniversity ofWestern Ontario and continues tothisday.

References Cited

Albanese,John2001aMeasurementerrorandMetricMethodsfor

SexDetermination.Paperpresented at the29th Annual Meeting of the CanadianAssociation of Physical Anthropology,Winnipeg,Manitoba.

2001b ANewMetricMethodforSexDeterminationUsingthePelvisandFemur.American Journal of Physical Anthropology2001Suppl32:31.

2002A New Metric Sex Determination MethodUsing the Clavicle and Humerus. Paperpresentedatthe30thAnnualMeetingoftheCanadian Association for PhysicalAnthropologyinOttawa,Ontario,October24-26.

2003ANewMetricMethodforSexDeterminationUsingtheHipboneandTheFemur.Journal of Forensic Science48:263-273.

Angel,Lawrence1976Colonial to Modern Skeletal Change in the

U.S.A. American Journal of Physical Anthropology45:723-736.

Armelagos,GeorgeJ.,andAlanH.Goodman1998Race,Racism,andAnthropology.In Building

a New Biocultural Synthesis: Political-economic Perspectives on Human Biology,edited by A. H. Goodman and T. L.Leatherman, pp. 359-378. University ofMichiganPress,AnnArbor.

Brauer,GerhardW.1982Size SexualDimorphism andSecularTrend:

Indicators of Subclinical Malnutrition? InSexual Dimorphism in Homo Sapiens: A Question of Size, edited by R. L. Hall, pp.245-259.Praeger,NewYork.

Brown,RyanA.,andGeorgeJ.Armelagos2001ApportionmentofRacialDiversity:AReview.

Evolutionary Anthropology10:34-40.Cartmill,Matthew

1998TheStatus of theRaceConcept inPhysicalAnthropology. American Anthropology 100:651-660.

Cunha,Eugenia1995Testing Identification Records: Evidence

from the Coimbra Identified SkeletalCollections (Nineteenth and TwentiethCenturies). In Grave Reflections: Portraying the Past through Cemetery Studies,editedbyS.R.SaundersandA.Herringpp.179-198.CanadianScholarsPress,Toronto.

Ericksen,MaryF.1982How“Representative”istheTerryCollection?

Evidence from the Proximal Femur.American Journal of Physical Anthropology59:345-350.

Fish,JeffersonM.1999Why Psychologists Should Learn Some

Anthropology. In The Biological Basis of Human Behavior: A Critical Review, editedby R. W. Sussman, pp. 197-198. PrenticeHall,NewJersey.

Floud,Roderick.1994The Heights of Europeans Since 1975: A

New Source for European EconomicHistory. In Stature, Living Standards and Economic Development,editedbyJ.Komlos,pp. 25-38. University of Chicago Press,Chicago.

Greiner,ThomasM.,andClaireC.Gordon1992Secular Trends in 22 Body Dimensions in

Four Racial/Cultural Groups of AmericanMales.Human Biology4:235-246.

Greulich,W.W.1951 The Growth and Developmental Status of

Guamanian School Children in 1947.American Journal of Physical Anthropology 9:55-70.

Greulich,W.W.,C.S.CrimsonandM.L.Turner1953 The Physical Growth and Development of

Children Who Survived the AtomicBombing of Hiroshima and Nagasaki.NagasakiJournal of Pediatrics 43:121-145.

Hartmann,J.,R.Keister,B.Houlihan,L.ThompsonandR.Baldwin

1994 DiversityofEthnicandRacialVNTRRFLPFixed-bin Frequency Distributions.American Journal of Human Genetics55:1268-1278.

Hunt,DavidR.,andJohnAlbanese2005TheHistoryandDemographicComposition


of the Robert J. Terry AnatomicalCollection. American Journal of Physical Anthropology 127:406-417.

Hrdlička,Alěs1934 Contributions to the Study of the Femur:

TheCristaAsperaandthePilaster.American Journal of Physical Anthropology19:17-37.

JantzRichardL.2001 Cranial Changes in Americans:1850-1975.

Journal of Forensic Science46:784-787.JantzRichardL.andLeeMeadowsJantz

2000 SecularChangesinCraniofacialMorphology.American Journal of Human Biology 12:327-338.

Keita,S.O.Y.andR.A.Kittles1997 The Persistence of RacialThinking and the

Myth of Racial Divergence. American Anthropology99:534-544.

Lewontin,Richard.1972 The Apportionment of Human Diversity.

Evolutionary Biology6:381-398.Lieberman,Leonard

2001 How“Caucasoids”GotSuchBigCraniaandWhy They Shrank. Current Anthropology42:69-95.

Livingstone,FrankB.1962 On the Non-existence of Human Races.

Current Anthropology 3:279-181.MeadowsJantz,LeeandRichardL.Jantz

1999 Secular Change in Long Bone Length andProportionintheUnitedStates,1800-1970.American Journal of Physical Anthropology 110:57-67.

Mokyr,JoelandCormacÓ.Gráda1994 The Heights of the British and the Irish c.

1800-1815:Evidence fromRecruits to theEast India Company’s Army. In Stature, Living Standards and Economic Development,editedbyJ.Komlos,pp.39-59.UniversityofChicagoPress,Chicago

Molnar,Sephen2002Human Variation: Races, Types, and Ethnic

Groups.FifthEdition.PrenticeHall,UpperSaddleRiver,newJersey.

Nei,M.,andA.K.Roychoudhury1982 EvolutionaryRrelationshipsandEvolutionof

Human Races. Evolutionary Biology 14:1-59.

Ousley,StephenD.,andRichardL.Jantz1998 The Forensic Data Bank: Documenting

Skeletal Trends in the United States. InForensic Anthropology.SecondEdition,edit-ed by K. J. Reichs, pp. 441-458. CCThomas,Springfield.

Relethford,JohnH.1994Craniometric Variation Among Modern

Human Populations. American Journal of Physical Anthropology95:53-62.

2001 Global Analysis of Regional Differences inCranio-Metric Diversity and PopulationSubstructure.Human Biology73:629-636.

2002 Apportionment of Global Human GeneticDiversityBasedonCraniometricsandSkinColor. American Journal of Physical Anthropology118:393-398.

Rocha,MariaA.1995 Les Collections Ostéologiques Humaines

Identifiées du Musée Anthropologigue del’Université de Coimbra. Anthropology Portugal13:7-38.

Ryman,N.,R.Chakraborty,andM.Nei1983 Differences in the Relative Distribution of

Human Genetic Diversity betweenElectrophoretic and Red and White CellAntigens.Human Heredity 33:93-102.

St.Hoyme,LucileE.,andMehmetYaşarIşcan1989 Determinationofsexandrace:accuracyand

assumptions. In Reconstruction of Life from the Skeleton,editedbyM.Y.IsçanandK.A.Kennedy, pp. 53-93. Alan R. Liss, NewYork.

Sokal,R.R.,andF.J.Rohlf1973 Introduction to Biostatistics. San Francisco:

WHFreemanandCompany.SPSS

1998SPSS 9.0 for Windows.SPSSInc,Chicago.Steckel,RichardH.

1994 Heights and Health in the United States,1710-1950.InStature, Living Standards and Economic Development,editedbyJ.Komlos,pp. 153-172. University of Chicago Press,Chicago.

Stini,W.A.1975 Adaptive Strategies of Human Populations

Under Nutritional Stress. In Biosocial Interrelations in Population Adaptation,edit-edbyE.S.Watts,F.E.JohnstonandG.W.Lasker,pp.19-41.Mouton,TheHague.

1982 Sexual Dimorphism and Nutrient Reserves.In Sexual Dimorphism in Homo Sapiens: A Question of Size, edited by R. L. Hall, pp.391-419.Praeger,NewYork..

Stinson,S.1985 SexDifferencesinEnvironmentalSensitivity

DuringGrowthandDevelopment.Yearbook of Physical Anthropology28:123-147.

Templeton,A.R.1998 Human Races: A Genetic and Evolutionary


Perspective. American Anthropology 100:632-650.

TerryRobertJ.1932The Clavicle of the American Negro.

American Journal of Physical Anthropology3:351-379.

Tobias,PhilipV.1972 Growth and Stature in Southern African

Populations. In Human Biology of Environmental Change, editedbyD. J.M.Vorster, pp. 96-104. Inter-nationalBiologicalProgram,London.

1975 Anthropometry Among DisadvantagedPeoples: Studies in Southern Africa. InBiosocial Interrelations in Population Adaptation, edited by E. S. Watts, F. E.Johnston and G. W. Lasker, pp. 287-305.Mouton,TheHague.

1986 War and Negative Secular Trend of SouthAfrican Blacks with Observations on theRelativeSensitivityofCadavericandNon-Cadaveric Populations to Secular Effects.Proceedings of the 5th Congress of the European Anthropological Association1:451-461.

Tobias,PhilipV.,andD.Netscher1977 Reversal of the Usual Secular Trend, as

EvidencedbySouthAfricanNegroCraniaandFemora.Sonderab-duck aus Mitteilungen der Anthropologischen Gessellschaft in Wien 103:233-239

Trotter,MildredandGoldineC.Gleser1951 Trends in Stature of American Whites and

Negroes Born between 1840 and 1924.American Journal of Physical Anthropology 9:427-440.

JohnAlbanese,DepartmentofSociology,AnthropologyandCriminology,UniversityofWindsor,Ontario

a critical review of the methodology for the study of secular … … · anthropology data bank....

Documents