You are on page 1of 14

1

Rough Notes on a Preliminary Analysis of the


Entire Published Corpus of Indus Weights
RichardSproat
rws@xoba.com
August2013
This is an evolving document. Please consult with me before quoting.
The Data
TherehasbeensomeconfusingdiscussionrecentlyontheWebabouttheaccuracyofthe
IndusValleyweightsystem.Oneexampleofsuchathreadcanbefoundhere:
http://groups.yahoo.com/neo/groups/IndoEurasian_research/search/messages?query=indus%
20weight%20standards.Dependinguponwhatextremeonewantstotake,theIndusValley
eitherhadnostandardsatall,orthestandardswereveryaccurate,andfurthermoreextended
overbothalargegeographicalareaandoverhundredsorthousandsofyears.
Theclaimsfromthetraditionalarcheologicalliterature(e.g.Hemmy,inMarshall,1938)isthatthe
weightvalueswereaccuratetowithinabout23%measuredastheratioofthestandard
deviationtothemean,andthatthiswasmuchmoreaccuratethancontemporaneous
civilizations(e.g.Mesopotamia,Egypt),whichusuallyhadaccuraciesofabout5%.Furthermore,
theHarappanstandardseemedtoholdovertheentirecivilization,withthesamestandards
beingusedatthevariouscitiesofHarappa,MohenjoDaro,ChanhuDaroandsoforth.Butallof
theseclaimsweremadeonthebasisofasubsetofthedatathatwenowhaveatourdisposal.
Furthermore,recentworkontheweightsystemoftheIndusValley(Vahia&Yadav,2007
Fournet,2011),hasbeenbasedsolelyononelist,whichcomesfromTableXofMackay1938
(seebelow).
Thepurposeofthepresentworkisthreefold:
1. Toperformapublicservicebyprovidinganopensourcedistributionoftheentire
publishedcorpusofIndusweights
2. Toassesstheaccuracyofthesystembasedonthatcorpus
3. Hopefullytoacquiresimilardatafromothercontemporaneouscivilizationssothatwecan
seehowthesystemscompare.
AnotherreasonforbeinginterestedinthisrelatestothesocalledIndusscript.Inoneversionof
hispaper(http://www.scribd.com/doc/156175428/WSRFournetEntropyIndusSigns),Fournet
2
makesthebizarreclaimthatitishardtobelievethatawellorganizedsocietywithprecise
weightscouldnothavesomekindofwritingsystematthesametime.Andthenagain:Inother
wordsifwedismissthecorpusofIndusSignsasbeingawritingsystemwearethen
facedwiththeparadoxofawellorganizedsocietywithplannedurbanizationandpreciseweights
butnowritingsystematthesametime.Isthatreallypossible? Theproblemwiththesekinds
1
ofIcantimagineargumentisthattheyusuallyhavemoretosayabouttheimpoverishmentof
imaginationoftheproponent,thanaboutthephenomenonunderconsideration.Inanycase,itis
unclearwhypreciseweightswouldnecessitatewritingasopposedto,forexample,mereskillat
handiwork.Butinanycasewhateveronemaybelieveabouttheconnection,itisusefultosee
whatthefactsreallyare.
Tothatend,Ihavecollectedallofthepublishedarchaeologicalmaterialthatgivestablesof
individualweights.Thesourceswereasfollows:
Marshall,J.1931.TheIndusCivilization.VolumeII,ChapterXXIX,"SystemofWeightsat
MohenjoDaro",AppendicesI,II.London:ArthurProbsthain.AppendixIcontains121
weightsfromMohenjoDaroAppendixIIcontains39weightsfromHarappa.
Mackay,E.J.H.1938.FurtherExcavationsatMohenjoDaro.VolumeI,ChapterXVII,
"SystemofWeights",byA.S.Hemmy,AppendixI,TableX.NewDelhi:Governmentof
IndiaPress.577weightsfromMohenjoDaro.
Vats,M.S.1940.ExcavationsatHarapp.VolumeI,pp360365.(Republished1974)
BhartiyaPublishingHouse.199weightsfromHarappa.
Mackay,E.J.H.1938.ChanhuDaroExcavations.ChapterXV,"ObjectsofScientific
Interest:WeightsatChanhudaro",byA.S.Hemmy,TableITheCubeWeightsin
Boston",byA.R.Hall.NewHaven:AmericanOrientalSociety.123weightsfrom
ChanhuDaro.
Rao,S.R.1985.Lothal:AHarappanPortTown,19551962.pp560564,TableXIX.New
Delhi:ArchaeologicalSurveyofIndia.27weightsfromLothal.
HarappaArchaeologicalResearchProject,19862001,CourtesyRichardHMeadow.68
weightsfromHarappa.
Inthefewcaseswhereitwasfeasibletodosobasedoncatalognumberse.g.insome
casesinHallstabulationoftheweightsfromChanhuDaroIeliminatedduplicates.
Notethereforethattheremaywellstillbeduplicatespecimensrepresented,whichwouldtendto
1
NotethatFournethas,ontheotherhand,disavowedinemailanyconnectionbetweentheprecisionofthe
weightsandtheissueofliteracy.
3
spuriouslyraisetheprecision.
Therawdataforalloftheseweightscanbefoundhere:
http://rws.xoba.com/indus_weights/weights.xml
Analysis of the Precision of the System
Forapreliminaryanalysisofthedata,Iconsideredhowprecisetheweightsare,wherewe
defineprecise(ashavepreviousauthorssuchasHemmyandHall)astheratioofthestandard
deviationoftheweightvaluestothemeanofthevalues.Claimsintheliteraturehavebeenthat
thisprecisionisaslowas0.02,muchlowerthanthe0.05moretypicalfor3rdmilleniumBCE
civilizations.
Aprioriwedonotknowwhattheweightstandard(s)was(were).Andmoreimportantly,wedo
notknowwhichweightspecimenswereintendedbythemakertobeinstancesofwhichweight
standard.Forexample,ifweassumethatthebasicunitweightisaround13.7 grams,whatare
2
wetodowithaspecimenthatweighsinat12.3grams?
Soinordertomakethefewestassumptions,Iperformedthefollowinganalysis,focusingfornow
onspecimensthatweighin20gorless:theseareinanycasethemostcommonweightsinthe
collection.Iconsideredcentervaluesrangingfrom1.0to20.0in0.1gintervals.ThenI
consideredthecollectionofweightsthatwaswithinawindowand1ofthatvalue.Thusfor
exampleforacentervalueof10.0,thewindowwouldincludeanyweightsthatarewithinabout
6.7gand13.3g.Thewindowofplusorminusisofcoursearbitrary,butsincetheweightsat
thelowerendofthescalehavebeenarguedtobeinvaluesofmultiplesof2ofeachother,a
windowofoneithersidewouldnot,inatrulyperfectweightsystem,includespecimensthat
shouldproperlybelongtothenexthigherorlowerweight.Finally,Ithenshrunkthewindowin
incrementwidthsof0.1g.Thusonegetsfewerandfewerweightsinthesetfallinginthe
window,andconcomitantlyhigherprecision.Thenforeachsetwecomputethestandard
deviationdividedbythemean(/).
Ifoneplotsallthesevaluesof/,onewouldexpecttoseethefollowingbehavior.Firstofall,the
valuesof/shouldbecomeminimalasthecentervalueapproachesthetargetstandard
weightsofroughly13.7,13.7/2and13.7/4.Second,obviously,astheincrementwidthofthe
windowdecreases,theratio/shoulddecrease.
Icomputedthesestatisticsforthefollowingsubsetsoftheweights.Firstofall,Iremovedfrom
considerationanyspecimenthatwaslistedasdoubtfullyaweight(onlythuslistedinMackays
datafromChanhuDaro).ThenIconsideredthefollowing:
2
Or13.6grams,seebelow.Inanyeventsomewherearoundthesevalues.
4
Allremainingweights(n=1119)
Onlyweightslistedasperfecti.e.notdamaged.(OnlyavailableforMackayfor
MohenjoDaroandChanhuDaroandRaoforLothal.)(n=108)
AllweightsfromMohenjoDaro(n=698)
AllweightsfromHarappa(n=306)
AllweightsfromChanhuDaro(n=88)
(TherearetoofewweightsintheLothaldatatobeworthconsideringbyitself.)
Theplotscanbefoundbelow,andalsohere:http://rws.xoba.com/indus_weights/plot.pdf
5
6
7
8
Intheplots,thehorizontalaxisrepresentstheweightcentervalue,theverticalaxis/.The
coloroftheplottingcharacterrepresentstheratioofthecentervaluetothesizeofthewindow,
asfollows:
red:ratio>10(i.e.thesmallestrelativewindowsizes)
orange:ratio>8
purple:ratio>6
green:ratio>4
blue:otherwise
Sofora13.7gcentervalue,redwillrepresentcaseswherethewindowisbetween13.71.37,
orangecaseswherethevalueisbetween13.71.71,purplecaseswhereitisbetween
13.72.28,greenwhereitisbetween13.73.43,andblueforrangesgreaterthanthatupto
9
13.74.52.
Finally,verticallinesareplottedforthenominalstandardsof13.7,6.85and3.43grams.
Ascanbeseen,theminimaforthe/areindeedroughlyatthenominalstandardvalues,which
isconsistentwiththosevaluesbeingtheonesthattheweightmakerswereaimingfor.And,
obviously,asthesizeofthewindowisincreased,theprecisiondecreasesfromsomewhere
around/=0.02to/>0.06.Ifanything,theimprecisionseemstobecomeworseifone
considersonlytheperfectweights.OntheotherhandthesetfromChanhuDaroseemstobe
ratherbettersincetherearemoredotsclusteredclosetothelowerendsofthe/range.
Partoftheproblemwiththeseanalysesis,asnotedabove,thatweknowthereareduplicate
weights:TableXisagoodsourceofsuchduplicatesbutevenbeyondthatinthevariousreports
onMohenjoDaro,HarappaandChanhuDaro,weexpectthatsomeoftheweightshavebeen
duplicated.Itishardtotellfromthepublishedmaterialwhichonestheymightbe,anddepending
onhowmanyofthemthereare,theirpresencemaytendtoincreasetheapparentaccuracyof
thesystem.Toseethepossibleeffects,weplotbelowthesystemwithallduplicatesremoved
(yielding595weights),whichshouldgiveanideaoftheworsttrueaccuracythesystemmight
have.
Comparingtheplotbelowwiththeplotforallweightsabove,thereisnotalotofdifference,
thoughonecanseeperhapsaslightincreaseof/for,say,the6.85gramweightforthe
smallestwindowsizesfromabout0.03toabout0.04.
10
Itisworthmentioningatthispointthattherearenodifferencesinstandardsbetweenthethree
mainsitesofMohenjoDaro,HarappaorChanhuDaro.Performingasimilarvariablewindowing
tothatdescribedabove,butnowjustconsideringwindowsaroundthenominal13.7g,6.85gand
3.43gtargets,attestofpairwisecomparisonsofthesitesrevealednosignificantdifferences
betweenthemeans.Soatleastitdoesseemtobetruethatthesystemisfairlyconsistent
acrossdifferentsites.
Alloftheaboveanalysismayseemratherobvious,butitisworthbearinginmindthatifaweight
systemhadtheaccuraciesthatamodernsystemhas,thenalloftheweightspecimensfora
giventargetwouldbewellwithinthenarrowest0.1gwindow,andexpandingthewindowaround
thetargetwouldnotyieldadifferentresult.Thefactthatadifferentresultisfoundinthecaseof
theIndusweightsisconsistentwiththeideathatstandardswerecrudehardlysurprisingfora
3rdmilleniumBCEcivilization.Howcrudeorprecisetheyweredependsofcourseuponhow
11
liberalyouareaboutthrowingoutfromconsiderationspecimensthatdonotfitwithinagiven
window.
The Weight Types
Theplotbelowshowstheresultoftakingallreportedweightvaluesfornondoubtfulweights,
roundingthevaluetothenearesttenthofagramforvalueslessthan50andtoonegramfor
largervalues,andcountingthenumberofweightsineachbin.Theresultisshowninaloglinear
plot,wherethehorizontalaxisistheweightvalueandtheverticalaxisthenumberofweight
specimenswiththatvalue.Theweightvaluesofthepeaksareindicatedintheplot.
Theratiosofeachpeakvaluetothepreviousonesareasinthetablebelowforthefirstsix
weightsforthethreehighestweightsforwhichwehavemuchevidence,thecomparisonis
12
giventoaputativebasevalue.
Previous Current Ratio
0.9 1.8 2.00
1.8 3.4 1.89
3.4 6.8 2.00
6.8 13.6 2.00
13.6 27.3 2.01
27.3 54 2
Base Current Ratio
13.6 136 10
27.3 275 10
13.6 1375 101
Aswithpreviousanalyses(e.g.Hemmyswork),thebasisofthesystemseemstobebinaryfor
thelowestvaluedweightsthesimplestsortofsystemfromthepointofviewofweight
manufacture.Andconsistentwithpreviouswork,theheavierweightsseemtobebasedona
decimalsystem.
Thereisinanycasenoevidencehereofadvancedmathematicalabilities,assomehave
claimed.Theevidenceformuchbeyondabinarysystemforlighterweightsisweak:thereis
someevidenceintheformofthreepeakweightvaluesforadecimalsystem,twobasedon
multiplesof10ofthebaseof13.613.7grams,andonepossiblybasedonthedoubleofthat,the
27.3gramweight.Butnoneofthisevidence,exceptforthe136gweight,isparticularlyrobust.
13
Noneoftheseresultsarenew.Whatisnewisthefreshperspectivethatlookingatallthe
availabledatacanbringintermsofthesoberingrealizationthatwiththeexceptionoftherather
simplebinarysystemforthelightestweights,thereisntreallyahugeamountofevidenceto
supportsweepingclaimsaboutthesophisticationofthesystem.
Further Analyses
Moreanalyses(hopefully)tofollow,including:
Asimilaranalysisofanother3rdmilleniumcivilization(assumingonecanfindthedata).
Amoresophisticatedwayofcomputingthewindowsthatmightgiveafairer
representationofwhichweightsmightreasonablybeconsideredoutliers.
References
1. Bosak,Jon.2012.Canonicalgrainweightsasakeytoancientsystemsofweightsand
measures.http://www.academia.edu/Documents/in/Ancient_Weights_and_Measures
2. Fournet,Arnaud.2011.DeterminingStatisticalSignaturesforUndecipheredScriptsand
Corpora:thecaseoftheIndusValleySigns.
3. Fournet,Arnaud.2012.TheissueofastandardizedIndusValleyweightsystem.
4. HarappaArchaeologicalResearchProject,19862001,CourtesyRichardHMeadow.68
weightsfromHarappa.
5. Hemmy,A.S.1937.AnanalysisofthePetriecollectionofEgyptianweights.TheJournal
ofEgyptianArchaeology.23(1),3956.
6. Kenoyer,J.Mark.2010.MeasuringtheHarappanworld:insightsintotheIndusorderand
cosmology.InMorley,I.andRenfrew,C.TheArchaeologyofMeasurement:
ComprehendingHeaven,EarthandTimeinAncientSocieties.Cambridge.Cambridge
UniversityPress.
7. Mackay,E.J.H.1938.FurtherExcavationsatMohenjoDaro.VolumeI,ChapterXVII,
"SystemofWeights",byA.S.Hemmy,AppendixI,TableX.NewDelhi:Governmentof
IndiaPress.577weightsfromMohenjoDaro.
8. Mackay,E.J.H.1938.ChanhuDaroExcavations.ChapterXV,"ObjectsofScientific
Interest:WeightsatChanhudaro",byA.S.Hemmy,TableITheCubeWeightsin
14
Boston",byA.R.Hall.NewHaven:AmericanOrientalSociety.123weightsfrom
ChanhuDaro.
9. Marshall,J.1931.TheIndusCivilization.VolumeII,ChapterXXIX,"SystemofWeightsat
MohenjoDaro",AppendicesI,II.London:ArthurProbsthain.AppendixIcontains121
weightsfromMohenjoDaroAppendixIIcontains39weightsfromHarappa.
10. Miller,Heather.2013.WeightyMatters:EvidenceforUnityandRegionalDiversityfrom
theIndusCivilizationWeights.Chapter9in,Abraham,SA,Gullapalli,P.,Raczek,Tand
Rizvi,U.(eds).ConnectionsandComplexity:NewApproachestotheArchaeologyof
SouthAsia.LeftCoastPress.WalnutCreek,CA.
11. Petruso,KarlM.1981.EarlyWeightsandWeighinginEgyptandtheIndusValley.M
Bulletin(MuseumofFineArts,Boston).Vol.79.4451
12. Rao,S.R.1985.Lothal:AHarappanPortTown,19551962.pp560564,TableXIX.New
Delhi:ArchaeologicalSurveyofIndia.27weightsfromLothal.
13. Vahia,M.N.andYadav,Nisha.2007.HarappanWeights.Puratatva,37,213217
14. Vats,M.S.1940.ExcavationsatHarapp.VolumeI,pp360365.(Republished1974)
BhartiyaPublishingHouse.199weightsfromHarappa.

You might also like