27
Measures of Divorce Frequency in Simple Societies Author(s): J. A. Barnes Source: The Journal of the Royal Anthropological Institute of Great Britain and Ireland, Vol. 79, No. 1/2 (1949), pp. 37-62 Published by: Royal Anthropological Institute of Great Britain and Ireland Stable URL: http://www.jstor.org/stable/2844502 . Accessed: 08/08/2013 05:37 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. .  Royal Anthropological Institute of Great B ritain and Ireland is collaborating with JSTOR to digitize, preserve and extend access to The Journal of the Royal Anthropological Institute of Great Britain and Ireland. http://www.jstor.org

Measure of Divorce Frequency in Simple Societies

Embed Size (px)

Citation preview

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 1/27

Measures of Divorce Frequency in Simple Societies

Author(s): J. A. BarnesSource: The Journal of the Royal Anthropological Institute of Great Britain and Ireland, Vol79, No. 1/2 (1949), pp. 37-62Published by: Royal Anthropological Institute of Great Britain and Ireland

Stable URL: http://www.jstor.org/stable/2844502 .

Accessed: 08/08/2013 05:37

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of 

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

.

 Royal Anthropological Institute of Great Britain and Ireland is collaborating with JSTOR to digitize, preserv

and extend access to The Journal of the Royal Anthropological Institute of Great Britain and Ireland.

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 2/27

37

MEASURES OF DIVORCE FREQUENCY IN SIMPLE SOCIETIES

By J. A. BARNES,M.A. UniversityCollege London

I. THE NEED FOR MEASURES

As the study of kinshipdevelops, o it becomesnecessary ndpossible o give ncreasing recisionothedescriptiveerms sed nthat tudy. Thus, uchwords s exogamy, amily, insman,ndlineagehaveacquired n the last hundredyearsprecisereferentswhich, if not always employed,can be statedunambiguouslywhen necessary n any particularanalysis. In thispaperwe shall consider ne itemin the studyof kinship,divorcefrequency.Afterexamininghedegree fprecision ithwhich hishas

beenreportedn somerecent tudies,we shalloutlinenew methods of computation, llustrating hesemethods- ithpreviously npublished ata collectedamongthe Fort JamesonNgoniduring1946-49.1

In many ocietiesmarriages an be terminatedydivorce, nd in anydiscussion fmarriaget is clearthatone importantactors thepresence r absenceof hepossibilityfdivorce. However,n a few ecentstudies n whichmarriages discussed, here s nospecific iscussion fdivorce Childs,1949; Earthy,1933; Green, 947; Kluckhohn ndLeighton, 946;Kuper,1947). From uchnegative videncewemaybe tempted o infer hat divorcedoes not occur n

thesocieties oncerned,ut this s an unsatisfactorybasisfornference,nd nsome nstanceshe nferencewould efalse. Theres,perhaps, reaterustificationfornot specificallytating hatdivorce s impossiblewhendescribing Catholiccommunitye.g.,Miner,1939).

In thenextgroup fstudies,marriage nd divorceare discussed, ut althoughwe may nfer rom omeof themthat divorce s not veryfrequent,n noneof them s thereany explicit ndication f divorcefrequencyElwin, 1939, p. 284; Fei, 1939,p. 49;Field, 1940,pp. 40ff; Fiirer-Haimendorf,943,pp.140f; 1945, pp. 135f; Goodwin, 1942,p. 343; Krige &Krige, 1943, p. 158; La Barre, 1948,p. 132; Leach,1940, p. 21; Mcllwraith, 1948, i, p. 418; Mair,1934, pp. 97ff; Meek, 1937, pp. 276, 279ff; Nadel,1942, pp. 152ff; Paulme, 1940, pp. 410ff; Smith,

1These data were collectedwhile was on the staff f theRhodes-Livingstone nstitute. I am much indebtedto mycolleagues, Dr. J. C. Mitchell,ProfessorGluckmanand Dr.E. Colson,for timulus nd assistance n preparing hispaper.

1940, P. 167 Wilson, 1942, pp. 64ff). In someinstancest is fairly learthat data are not availablefor estimating ivorcefrequency, hile n others tseems s though he matter ouldhave been enquireintobutthatthiswasnot done. In onestudyn thigroup Culwick& Culwick, 935,pp. 329ff) ivorcfrequencys indicated n a separatepublication.

Manywriters ave been content o indicate ivorcfrequency-byomedescriptive hrase. Thus,divorcis said to be rare Harrison, 937,p. 401), or unusua(Stayt,1931, . 152), rfairly requentPowdermake

1933, p. 228), or frequent Brown& Hutt, 1935p. 83 ; Elkin, 1940, p. 213 ; Nadel, 1947, pp. 125,226, 291, 339, 434; Opler, 1940, p. 198), or common(Bowers, 1950, p. 80; Landes, 1937, p. 80; 1938p. 85), or not uncommon (Spicer, 1949, p. 50).Sometimes we may infer the frequency of divorcefrom tatementsthat marriage s brittle Joffe, 940p. 305; Opler, 1940, p. 151), or is rarely dissolved(Firth, 1936, pp. 132, 571), or tendsto be a lastinunion Wagner, 949,pp. 440ff),r that themajoritof marriages ndure Herskovits, 938, i, p. 349).Williamson 1939, pp. 189ff) ays that divorce seasy and appearsto implythat it is also frequenwhile Hsu (1949,p. 105) says that barrenmarriagofteneadstodivorce, ut doesnot ayhowfrequentmarriages re barren. Otherwritersndicatethain certain ircumstancesivorce s lessfrequent haninwothersBarton,1949,pp. 50,54; Hulstaert, 938p. 379; Wagley, 949,p. 44),orthattheres a loweproportionf divorces Hunter,1936, pp. 212, 220484).

While it is difficult o arrange these descriptivphrases n an ascending rderofdivorcefrequencit is clearthatfor omepurposes hey readequateThus, ffor xamplewe wish to distinguishetweethe frequency f divorceamongthe Zulu and thNyasalandYao, or in the Irish Republic and th

State of Nevada, it may be sufficiento say thadivorce s rare in one society nd frequent n thother. If, however,we wish to comparesocietiewhich re not in thisrespect o very differentromone another, escriptionnthese ermss not enougand we needmeasures ffrequency hich re morpreciseand more objective. Two societies n bothofwhichdivorce s said to be commonmay, n fact

(2499)

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 3/27

38 J. A. BARNES

have verydifferentivorce requencies,nd the samesocietymay be described n conflicting ays bydifferentbservers. For example, he Fort JamesonNgoni considerthat divorce s now rife in theirsociety,while the Administrationonsiders hat itis not so very frequent. The Ngoni standardistaken from ts ownpast,whiletheAdministration's

standard s based on the neighbouringmatrilihealtribes.2In order o givegreater recisiono their ccounts,

a number of writershave published numericalindicationsof divorce frequency.Unfortunately,therehas been considerable ariation n theway inwhich hese ndiceshave been calculated, o that itis sometimes ifficulto compare ata from ifferentsocieties. The measuresused are of four kinds.Firstly,here rethose eferringo thepresentmaritalstate of the populationand, in particular, o thenumberof people in a populationwho have beendivorced nd whohave not yetremarried.Secondly,

we have thosereferringo the cumulativemaritalexperience f a population, ndicatinghow manytimes persons have been marriedand divorced.These twokinds ofmeasure re discussedbelowinsection I. Measures in the third group indicatewhat proportion f marriages nd in divorce, ndtheseare discussed n sectionV, while he astgroupcomprisesdivorce rates, showingthe number ofdivorcesgrantedwithin giventime.

Divorce rates are often the only indicationsofdivorce frequency vailable for modernsocieties,but have not been favouredby writers n simplesocieties. The term " rate" has, however,oftenbeenused whenwhat s correctlymeant s ratio or

incidence r cumulativexperience. Culwick1935,pp. 188, 195) correctlypeaksofan annualratefordivorces rantedn courtwhich, estates,s calculatedper 1000marriages.His data suggest hat hisratesmay, perhaps,be based on 1000married ouples.

In mostmodern ocieties llmarriagesnddivorcesare registeredwith some centralorganization,nddemographic ools are designedto make the bestuse of information lready collected. Hence thecommonestmeasures re rates relating he numberof divorces ranted uring yearor a decadeto thenumber f marriages akingplace during he sameperiod, rto the mean number fmarried ouples n

thepopulation. Theserates redifficultodeterminein simple societieswhere the total population susually unknownand where the registration fmarriagesnddivorcess at bestonlypartial. Evenin modern ocieties hese rates are not verysatis-factoryfor detectingchanges in divorce habits.

2 Cf. the discussion on whether or not divorce is rareamong he SouthernBantu in Gluckman(1950, p. 204ff).

In studying imple societiesthe fieldworker hausually o collecthis owndata as well s to analyse tOurtaskis thereforeo devisetechniqueswhich reasy toemploy n thefield s well s being nalyticalfruitful. t might e possible n a simple ociety oobserve a restricted opulation nd recordall thmarriagesnd divorces ccurring ithint during h

period f observation,ndthus o construct divorcrate,but the small ize of thepopulation hat can behandled n thisway makes t an inefficientay oworking. Fewmarriagesnddivorces anbeactuallobserved n the field,and to overcomethis disadvantage we need to collect information boumarriages nd divorceswhichhave takenplace inthepast. In a rapidly hangingocietyhisprocedurhas its hazards,but at the same time t enablesusto givean indication fthemagnitudef thechangetakingplace. This kindof-informationan be usedto calculate rates of a differentind, as will bedemonstratedn sectionVIII.

In societies wheredivorce s impossible r nonexistent, simple statement o that effectwoulseem to be all we need (e.g.,Wisdom,1940,p. 302)If,however, ivorce s a possibilitynd yet doesnooccur,an indicationof the size of the populatiowithinwhichno divorces re foundenables us tojudgetherarity fdivorceLevin, 1947,pp. 2, 5, 86;Spicer,1940, pp. 76ff).

Thesevariousmeasures o notall deal with xactlthe ame ocialphenomena,ndtheparticularmeasurwe utilize n analytical nd comparativeworkwildependon thepurposewe have in mind and on theavailabilityof the information. or instance,wemay be concernedwith the stabilityof only firs

marriagesnd need an indexwhich mits econd ndhigher rdermarriages.We may be concerned iththerelative fficiencyfdeath nddivorce s methodofendingmarriagend will thennotneed an indexfromwhich the effectsof mortalityhave beeneliminated. However, n general,we cannotmakespecialenquiries monga variety fpeopleson themattersn whichwe are interested nd in any, omparative tudywe have to relyon such nformatias is to be found n tribalmonographsnd generaarticles. It is thereforemportanthatthere houldbe some degreeof standardizationn the way inwhich uthorspresentdata on divorce nd cognat

topics o that valid comparisonsan be made. It isalso important hat certainbasic materialshouldalways be. published n any generalaccount of asocietywhetherrnot theauthorhas anyparticulacomparative nquiry n mind. If, forexample,weare describing amba society, t may not appearto be of any significanceo knowthat 119 out of194marriages nded in divorce. Yet this factcanbecomeofinterestwhen we discoverhowthese 194

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 4/27

Mectsure8fDivorce requencynSimple ocieties 39

marriageswere collectedand what happened to thoseof them that did not end in divorce, for we can thencompare this proportion with similar measurescollected in the same way fromother societies. Wemay then be able not only to make some generalobservations about all the societies compared butalso to add to our understandingof Lamba societyitself.

The more preciselywe measure divorce frequency,the more necessary it is to know exactly what wemean by marriage and divorce. If we vary ourdefinitions, e are likely to vary our frequency. Forexample, Fortes (1949, p. 85) specificallyexcludesfrom his statistics short-lived marriages in whichno bridewealthwas paid and no childrenborn. Hadhe not made this exclusion,his measures of Tallensidivorce frequencywould have been much greater.Bridewealthis paid in only 4 per cent. of marriagesamong the Fort JamesonNgoni, and had I restrictedmydivorce calculationsto bridewealthmarriages,mymeasures of frequency would be very different rom

those I have calculated on the basis of all marriages.In some- ocieties it may be difficult o distinguishdivorce from separation The problem of decidingwhen a divorcetakes place is a difficultnd importanttheoretical problem,and to discuss it would take usfar beyond the bounds of this paper. Nor can wediscuss here n any systematicmannerthe sociologicalsignificanceof the differences etween one societyand another as shown by the various measures ofdivorce frequency. For purposes of this paper, weassume we know what is a divorce, and we shall ingeneral confineourselves to a considerationof thedifferencesetween one measure and another.

One society varies -fromanother not only infrequency of divorce but also in the facility withwhichdivorcecan be obtained. Divorce is sometimessaid to be " rare and difficult or " frequent andeasy," and in analysis it is importantto keep theseaspects separate, even though they are plainly inter-related. " Difficult and " easy " describe theobstacles, in particularthe legal obstacles, in the wayof divorce and do not automatically indicate itsfrequency. n somestudies nwhichdivorcefrequencyis not explicitly ndicated,we are told that it is easyto obtain divorce and we may be tempted to inferthat divorce is therefore requent. This inference sliable to be false, for Granqvist (1931, pp. 164ff;1935, pp. 257ff) has described a society in whichdivorce is easy and yet infrequent. W& may expectto find occasionally a society in which divorce isdifficult nd yet frequent. It might be possible toconstruct some index of the facility with whichdivorce can be obtained in any society, but we arehiere oncernedonlywith the measurement f divorcefrequency.

II. PRESENT MARITAL STATUS AND CUMULATIVEMARITAL EXPERIENCE

The presentmarital status ofa populationindicatesthe proportion who at a particular moment areunwedded (i.e., have never been married),marriedwidowed, and divorced. Information of this kindis usually easy to obtain, but tells us little about the

frequencyof divorce. If divorce is followedswiftlby remarriage, then, however frequentdivorce maybe, at any one timetherewill be fewdivorcedpersonin the population. The status of divorced personis a transitional one between one marriage and thenext, and few,may be in it, though many passthrough it. A second complication is that in apolygynous society a man who divorces his wifmay never appear as divorced, for he may be stilmarried to a second wife,although his firstwife wilbecome a divorced woman.

Since divorce is a possibilityonly aftermarriageit seems preferable o relate the number of divorcedpersonsto the wedded population (i.e., all those whohave ever been married) rather than to the totapopulation which ncludesunweddedpeople. Recentstudies in which this proportion is indicated alldistinguishbetween men and women, and there issometimes n analysis by age as well. Some recentlpublishedfigures re set out in Table I, togetherwith

TABLE I. PresentMqrital Status: Percentageof WeddePersons Who reDivorced

Numberof Numberof PercentageSociety Wedded Pros Pretgociety Persons n Persons Divorced

_ Sample Divorced

MenFort JamesonNgoni (a) ... 116 7 6*0Lamba (b) ... 133 3 2-3Palestinian

Arabs (d) ... 105 0 0

WomenFort Jameson

Ngoni (a) ... 223 15 6*4Lamba (b) ... 195 22 11.3Nuer (c) ... ... 32 2 6-25Palestinian

Arabs (d) . . 148 0 0

Yao (e) ... ... 219 29 13 3

Sources(a) Barnes, 1951b, Tables III and IV.(b) Mitchell& Barnes, 1950, p. 46, Table XIX.(c) Extracted fromEvans-Pritchard,1945, pp. 31ff.(d) Extracted from Granqvist, 1931, pp. 157ff; 1935

p. 269. Betrothedpersonsexcluded.(e) Mitchell,1949, p. 297, Table III.

(2499) c 4

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 5/27

40 J. A. BARNES

material derivedfromraw data published by Evans-Pritchard (1945, pp. 31ff) and Granqvist (1931,pp. 164ff). Culwick & Culwick (1938, p. 377,Table IV), Forde (1941, p. 78, Table 6), and Fortes(1943, p. 102,Table I) give information bout presentmarital status formen, but do not distinguish etweenwidowersand those whose wives have been divorced.

Redfield 1941, p. 382), using officially ollected data,relates the number of divorced persons to the totalpopulation.

The cumulativemarital experience of a populationindicates the number of times to date that themembers of the population have been married,divorced, and left as widowers and widows. The

experiences f men and women are likely to diffin thismatter n an open community,nd everywhethere re differencesith espect o age. Alteratioin the age composition f a samplepopulation ffethe average number of marriages and divorceexperienced y each sex. We may sometimesbeable to infer rom n age-specificnalysis hatchang

musthave takenplace. For example, fwe find hamen now in their thirtieshave on the averagexperiencedwiceas manydivorces s mennow intheirforties, hen,other hingsbeing equal, we caninfer hat the expectation f divorce t ages belowforty as risen n the ast twenty earsor so.

Some-writersave indicated hemeannumber

TABLE II. Cumnulative arital Experience: Divorce8per head of Wedded opulationand Percentage f WeddedPopulationDivorced t leastOnce

Numberof Mean number I PercentageofNumber of divorces of divorces Number of wedded

Society weddedpersonsexperienced

per head of personsdivorcedpopulationin sample by them wedded at least once divorced tthem -

| population - _ least once

Weddedmen, ll agesBemba (a) ... ... ... 0 2 19Bemba (b) ... ... ... 0 53 35Fort JamesonNgoni (g) . . 104 51 0-49 41 39*4Lamba (h) ... ... ... 125 51 0*41Palestinian Airabs k) 105 6 0*06 6 5*7

Weddedmen, ver 5years ldFortJamesonNgoni (g) ... 48 29 0*60 21 44Lamba (h) ... ... ... 53 32 0*60Tallensi (m) ... ... ... 20 5 0-25

Weddedwomen, ll agesBemba (e) *... ... 048 43Bemba (d) ... ... ... 48Bemba (e) ... ... ... 57Bemba (f) ... ... ... 16Fort JamesonNgoni (g) ... 210 62 0*30 48 23Lamba (h) ... ... ... 177 64 0-36Ntier (j) ... ... ... 32 3 0-09 2 6-25Yak6 (n) ... ... ... 313 121 39Yao (p) ... ... 219 116 053 l

Sources(a) Brelsford, ited in Richards, 1940, p. 120, as Sample A.(b) Richards, oc.cit.,Sample B.(c) Ibid, Sample C.(d) Ibid, modernvillage (1).(e) Ibid, moderni illage (2).(f) Ibid, old-fashioned illage 3).(g) Barnes, 1951b,Tables IX and X.(h) Mitchell& Barnes, 1950, p. 48, Table XX.(j) Extracted from vans-Pritchard, 945, pp. 31ff.(k) Extracted fromGranqvist, 1931, pp. 157ff. Betrothedmen excluded.(n) Fortes, 1949, p. 85. Bridewealthmarriages nly.(n) Forde, 1941, pp. 75f,Table 5.(p) Mitchell.1949, pp. 297f,Tables III and IV.

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 6/27

Measures of Divorce Frequency n Simple Societies 4

marriagesnd divorces er head,but it is preferableto indicate the number of persons who haveexperienced0, 1, 2, . . . marriagesand divorces,sothat standard deviations can be calculated andmedian ndmodalvaluesusedas a check nthemean.

Recently ublishedmeasures fcumulative ivorceexperience,ogether ithmaterial erived rom ata

givenbyEvans-Pritchardnd Granqvist,reset outin Table II. Material given by Richards (1940,pp. 99, 101, 120) ppears o relate o theproportionfmarriagesndingn divorce ather han o cumulativedivorce experience. Siegel (1942, p. 61) presentscumulative ata, but does not distinguishbetweenwiveswhodied andthosewhowere ivorced. Du Bois(1944, p. 97) givesdata referringartly to presentmarital state and partly to cumulative divorceexperience,o that,unfortunately,e cannot rriveat a clearpicture f either. Harris 1940,pp. 98f)provides figuresfor the number of marriagesexperiencedymenandwomen, ut notthe number

ofdivorces.III. MARRIAGE SAMPLES

In these first wo computations, resentmaritalstateand cumulativemarital xperience,ursampleconsists f individuals f varying xperience. It isrelativelyasytoselectndividualsna manner hich,whilenotrandomnanystrict ense,doesstand omechanceofeliminatings muchbias in sampling s ispossibleunderfield onditionsn primitiveocieties.The rest of the measures iscussed n thispaper areconstructedfrom samples of marriages, not ofindividuals,ndmarriagesreharder oselectwithoutbias than people. One commonly sed method sthroughhecollection fgenealogies. The effective-ness of this methodappears to vary, for n somesocietieswheremarriages re not very stable, thechance hat a marriagewillbe remembereds greaterif it lasted a longtimeand gave riseto many ivechildrenhan f twas childlessnd of hort uration.Hence,any probingntothe unrecordedastby thismethodwill tend to give a sample which, t leastin theearlieryears, s biased in favourof long andfertilemarriages. Theseare ess ikely ohaveendedin divorce than are short, nfertile nes, and thefrequencyfdivorce, herefore,ppears to be lowerthan t actuallywas. Thisphenomenon f selective

forgetting,orone aspectof which have suggestedthe name " structuralamnesia (Barnes, 1947,pp. 52-53), s ofwidespread ccurrencend can haveimportant ociologicalsignificancecf. Kuczynski,1944, pp. 25-26). In this context,however, t ismerely hazard. On the otherhand, f divorce srare, t is quite possiblethat marriages nding ndivorce are rememberedfor their very rarity.Schapera (1940, p. 294) has made enquiries nto

divorce by the genealogicalmethod among thTswana,wheredivorce s rare.

An alternativemethod is to examine all thmarriages ver experienced y a selectedbody omen and to classifyhese by theway inwhich hewereterminated,ow ong they asted, nd whethor not they are still extant. The most seriou

objection to this procedure s that the universampled s notrelated n any simpleway to the totanumber f marriageswhichhave occurredwithingiven time in a society. In modernsocieties wexamine ll themarriageshat have taken place anddivide up this universeby isolating he marriagtaking lacein eachyear. Mostof the men nvolvein these marriagesn recentyearsare stillalive, buas we goback n time o a largernd arger roportiof the menmarryingn each year has died. In aclosed simplesociety, f we could examineall thmarriages xperienced y all the men we wouldbdealingwith lmost ll the marriages hichhadtake

place in the society n recentyears but owingtomortality ewouldonly nclude smaller roportiofmarriagesrom achofthe earlier ears. Withoufull nformationn male mortalitynd on the ageat whichmen have marriedwe cannot accuraterelate our findingso those frommodern ocietyIf we make the assumption hat our closed simpsociety s also static,withno changes akingplacewe can overlook hisdispersal foursample hrougthe yearsand then make a directcomparison, uthis ssumptions often ifficulto defend. We mayabandontheassumptionnd stillclaimfairly nougthat our samplereflectswhat has been happeninwithin he astgenerationr so and comparet witother amplesderivedn the sameway,but we musrealisethat some differences ay be due solelytodifferentortalityrtodifferencesnageofmarriag

A further bjection s that if our society s noclosed our sample includessome marriageswhicwouldnot be included n the corresponding odersample. Our informantsmay have marriedwhiltheywere n other egionsr whileworkings migranlabourers. Their marriagesdid not occur withitheterritoryssociatedwiththesociety tudied,yetheydo-form artof the total marital xperiencethe men n the society.

This method of examiningall the marriag

experienced ot by all themen n a societybut bya selectedbodyofthem s, however,he commonemethoddopted ndis alsothe methodmost ikely oyield ccurate, s distinct romunbiased,raw data;we must thereforexamine ts implications.Anysampledrawnfrommen shouldbe checked gainsa sampledrawnfrom similarly elected body ofwomen. Womenmaybemore, r ess,reticentbouttheirmaritalhistories hanmen and theremay be

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 7/27

42 J. A. BARNES

differencesn theirhistories as actually experienced.Thus, if men move about the country more thanwomen and contract marriages while woiking in thetowns, their marital histories may show a largeraverage numberofmarriages nd divorces than thoseofwomen. In an increasing rdecreasingpopulation,with men ma.king heir firstmarriage at a later age

than women,the numberofmarriages, nd hence thenumberof divorces, per head in each age group willdiffer or men and women (Hajnal, 1950b, p. 317).Women may conceal their divorces more carefullythan men, as among the Tallensi and Plateau Tonga.

If men and women can both be interviewed nd alltheirmarriages collected,then in general some ofthemarriages experienced by the men will also havebeen experienced by some of the women. Thus, ifwe take all the men and women in some residentialunit,for xample, a village,we shall learnabout thosemarriages linking each married couple living in thevillagefrom hewomenas well as from heirhusbands.

The testimony f a woman may not be identical withthat of her husband, and in general the historiesofman and wife will overlap in only one oftheirseveralmarriages. It is not often that a man marries thesame woman twice. The simplestway of indicatingthis overlap between the experiences of male andfemale informants s to divide the material collectedinto three parts, as follows

(1) Marriages for which both partners wereinformants.

(2) Marriages for which the only informantswere men.

(3) Marriages for which the only informants

were women.By comparing the accounts of part 1 given by

men and by women we can gain some idea of therelative reliability of their information, and bycareful nquiry we can hope to guess at what actuallyhappened in these marriages. Armed with thisestimate of the reliabilityof men as informants,wecan look at part 2, and can estimate the reliabilityand possible bias of the informationthere. Aftersimilarly weighingthe evidence of part 3, we are ina better position to decide whether or not anydifferencesheremay be between parts 2 and 3 aredue to differencesn actual experience or merelyto

differencesn reportingthat experience.Even with completely reliable. informantsand acompletely losed and homogenouscommunity, art 1is likely to differfromthe other two parts. Thisarises fromthe kind of way in which we choose ourgroup of informants. Often the most convenientmethodof sampling s to use all thewedded (i.e., ever-married) inhabitants of a village as informants.In this way an enquiry into divorce can be con-

veniently combined with a large number of otheenquiries into such topics as kinship, co-residenceeconomic co-operation, and the like. Using thimethod, a marriage in which, at the time of thenquiry, the partners are living in different illageis twice as likely to be included in the sample as amarriage n whichtheyare living n the same village

Suppose, for example, there are 300 villages in thsocietywe are studyingand we decide to make ourenquiry in one of these villages selected arbitrarilyIf a man X and his wife,or formerwife,Y are bothliving in P village, then if we select P village weinclude the X-Y marriage in our sample, and if weselect any othervillage we exclude it. Hence, theris one chance in three hundred that their marriagbe chosen. On the otherhand, suppose that Z livein P village, and his wife, or formerwife, T, livein Q village. Their marriage will be included inour sample if we choose either P or Q village, andhence they have two chances in three hundred of

being selected. In this way our sample, whichevevillage we select, contains a higher proportion ofmarriages ike Z-T than it would had we drawn oursample from he whole ofthe societystudied. Henceour calculations derivedfromthe sample are biasedin favour of marriages like Z-T and against thoslike X-Y. If there was no significant differenbetween these two kinds of marriagethis would notmatter. If men and women take spouses indiscriminately romwithin and from outside their ownvillages and also maintain separate households aftemarriage,then theremay be no bias in the sampleBut in most societies this is not so. Most of themarriagesin which both spouses are to be found in

the same village are extant, although we mayoccasionallyfind woman living n the same villageas a man fromwhom she has been divorced. Mostof the marriages in which spouses are found indifferent illages have been terminated in divorcealthoughwe may also find a fewmarriages n whichfor some reason or other, such as the demands ofco-wives,or the exigencies of employment,husbandand wifeare living apart. In general, then,the biasin our sample in favour of marriageswith spatiallyseparated spouses is a bias in favour of marriageending in divorce.-

Bias ofthiskindis greatestwhen the samplingunit

is completely xogamous. If there s village exogamyand husband and wife live together,there is only asmall chance that a woman will be living n the samevillage as her divorced husband. If she has subsequently married another man in her formehusband's village,orif he has married notherwomanin hers, the former couple may be found in onevillage. Most divorced couples are found livingapart, and under these conditions almost all extant

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 8/27

Measures of Divorce, requency n Sl,mple ocieties 43

marriages have a single chance of selection andalmost all marriages ending in divorce a doublechance. Bias will be least when the sampling unitis endogamous, for all marriages in which bothpartners re alive thenhave equal chances of selection,whetherthey are extant or have ended in divorce.It seems probable that in general the smaller the

residentialunit the greater heproportion f marriagescontracted with spouses outside the unit, so that thesmaller our sample the greater the danger of thiskind of bias.

This bias applies to the whole sample of allmarriages xperiencedby one or twoof the nhabitantsof a residential unit such as a village. It does notapply equally to the three parts of this sample, asdefined bove. Part 1 consists ofmarriages n whichbothpartners ive in the village and therefore onsistslargely of extant marriages. Marriages ended indivorce appear in part 1 only if former spousescontinue to live in the same village after divorce or

iftheyhappen to have moved independently o someotherone. The bulk of marriages ending in divorcefall in parts 2 and 3.

A simple correction for this bias which suggestsitself is to weight the parts of the sample so thatmarriages in part 1 have twice the weight of thosein parts 2 and 3. Unfortunately, hereare objectionsto this.

In the first lace, some of the marriages n parts 2and 3 may involve spouses or former pouses wholive outside the area being studied. For example,if T lives not in Q village but in some other society,the chance that hermarriage with Z is in the sampleis reduced to one in three hundred, for she herself

cannot be an informantwhatever village we select.Therefore,her marriage ought to receive the sameweight as those in part 1. The same argumentapplies ifX or Z is a labour migrant. If the societyconcerned is relatively isolated, we may, perhaps,ignore this complication. If it lacks clear socialboundaries, either because it is stateless or becausemany of its members are temporary migrants, or ifit is one in which there is frequentmarriage withforeigners,we must take this complication intoaccount.

A second objection s that this suggested weightingoverlooksthe effect fmortality. In general, part 1

does not contain any marriages ended by death.Yet in our example, even if T did live in a villagewhich we can select, she cannot be an informant fshe is now dead. By this reckoning ll marriages nwhich one spouse is dead ought to be weightedas ifthey were in part 1. If only marriages ended bydeath were involved, we could easily isolate themin parts 2 and 3 and weightthemaccordingly. Evenif our attention is restrictedto divorce and we are

not directly nterested n mortality,we still cannotneglectmarriages nded by death for, s will be shownlater, we need to correctour first stimatesof divorcerates by allowing for mortality. We have also toconsiderthe effect f death for another reason. Insome marriagesended in divorce one spouse is aliveand the other dead. These marriages appear in

parts 2 and 3 but, like marriages ended by deaththeyhave onlya single chance of selectionand shouldtherefore e weighted s thoughthey were n part 1.

Ouj revisedweightingprocedure s then as followsWe subdividepart 2 of oursample ntoparts2a and 2bMarriagesin which the non-informingpouse is deador livingoutside the societystudied fall into part 2athose in which the non-informingpouse is alive andwould have been interviewed fwe had selected someother village fall into part 2b. Part 3 is subdividedin the same way. We then give parts 1, 2a and 3aa double weight in our final sample and parts 2band 3b a single weight.

It will be seen that correctionby weighting s noa simple matter. To apply this correctionwe musenquire about the whereabouts of our informantformer pouses and ascertain whetherthey are aliveor dead. If this can be done easily and accuratelythe way is open for applying this correction. If iis impossibleto get reliable information bout formespouses it is, perhaps, safer to use an unweightesample, givingdetails of how it is divided among itthree parts, rather than to attempt a correctiononan unsatisfactorybasis.

Samples based on the experience of men aloncontain the equivalent of parts 1 and 2, and thosbased on women's experiencesofparts 1 and 3. For

conveniencewe may refer o samples of this kind asmen's and women's samples. If we add togethemen's and women's samples drawn from the sameresidential unit without making any allowance fooverlap, we then have a sample made up of parttwice and parts 2 and 3 once. This is merely ouoriginal sample corrected by giving part 1 doublweight.

An illustrationof the different alues obtained fodivorce ratios by changing the basis of sampling igiven in sectionV.

The sample village should be selected carefully othat it may be representativeof the society bein

studied. Since usually among simple societies verlittleis knownin advance about the population, andsinceit is usually quite impracticableto selectvillageat random, this is not easy. What can be done ito compare the village selected with a large numbeof other villages in respect of criteriawhich do norequire lengthy nvestigation. Thus, we may not bable to repeat a divorce enquiry in many villages inorderto see ifour first nquirywas typical. But w

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 9/27

44 J. A. BARNES

can compare the village in which, for better or worse,the detailed enquiry was made with many others forsize, history, kinship structure, apparent wealth,and the like, and so verify hat in those respects oursample village is typical of its area. In the absenceof any better nformationwe then make the assump-tion that our divorce data are also typical fora larger

population. We can increase the size of our sampleby adding together data from several villages,selecting these so as to cover any variation in typethat we may be aware of.

Even if the sample village is found not to betypical it seems worth while pursuing enquiries,provided *that the fact of atypicality is alwaysemphasized. If enquiries have reached an advancedstage it may not be worth while shifting rom well-tried and, well-known set of informants, howeveratypical they may be, to a typical but untried andunknown set elsewhere. It is surely better in ourpresent state of knowledge to have good informationabout atypical

communitieshan

meagre informationabout typical ones.

IV. DIVORCE RATIOS

Marriages can' be classifiedby age and status ofspouses, year of commencement,duration, termina-tion, and so on. The simplestmeasures of divorcefrequency elatedto marriages, nd notto individuals,take into account only method of termination andignoreall other factors. The usual kind of statementmet with is that such-and-such a percentage ofmarriages ended in divorce. It has been pointedout elsewhere (Mitchell & Barnes, 1950, p. 16) that

there is often; ome ambiguity in these statementsand it is not always clearwhichofthe following hreeratios are intended.

A. The number of marriages ended in divorceexpressed as a percentageof all marriages.

B. The number of marriages ended in divorceexpressedas a percentage of all completed marriages.

C. The number of marriages ended in divorceexpressed as a percentage of all marriages exceptthose that have ended by death.

Ratio B is a simple measure of the relative effectof divorce and mortality in terminating marriage.

Hence differencesn the value of this index may bedue solely to differencesn mortalitybetween twosocieties and not at all to differences n divorcehabits. The other two indices, A azid C, havemeaning only in relation to a sample drawn from alive population, for t is onlyfrom hem that we collectextantmarriages. Therefore, heseratios are affectedby mortality nd migration s they show themselvesin the age composition of the population, the sex

ratio, and so on. Hence, inA and C we are comparingterminations by divorce with a rather complicatedset of data. Ifwe turnfromA and C and concentratour attention on B, using only completed marriagesour sample contains marriages that on the whole areless recent than those involved in A and C. Inaddition, if our marriageshave been collected from

living people in the manner discussed in section III,ratio B will be biased in favour of short-livemarriages. The longer a marriage has persistedthe shorterthe period whichwill remain after t hasbeen terminated during which one or both of thespouses will be still alive and able to give us information. To give an extreme case, suppose, hypothetically, that in a certain society most coupleslive togetherto a ripe old age and then die of griewithin a few days of each other. Any analysis ofmarriage n this societybased-only on the completedmarriagesexperienced by living people would excludethe bulk of these long-livedmarriages nd would deal

mainly withthe

marriagesof

couples who haddeviated from the common pattern. Yet, if weattemptto overcome these shortcomings y includingmarriages experienced by dead people we encountethe bias due to selectiveforgettingmentioned earlierIn modem societies where calculations are based onall registeredmarriages,deaths and divorces and arenot restricted to marriages experienced by livingpeople, these objections to ratio B do not apply.

Ratio C is less directlyaffectedby mortality hanA and it seemslikelythat C willprove to be the mossatisfactory of these indices for frequency comparisons. Since figures ormodern ocieties are usuallygiven in the formof ratio B, it is desirable that data

should be published for differentocieties sufficienfor calculating both B and C. Ratio A can bederived fromB and C by the formula

1/B + 1/C =1 -+ 1/ADivorce ratiosare available for everalsocietiesand

are set out in Table III.Given the information et out in Tables I, II and

III, we are in a positionto determinewhetherdivorcis, in some unqualified sense, more frequent n onesocietythan in another. Thus, in termsof each ofthe measures used, divorce is less frequent amongthe Palestinian Arabs than among the Fort JamesonNgoni. Where the differencebetween societies is

not so marked,we cannot easily arrive at any conclusion. Thus, for example, Ratios A, B and Cof Table III are all lowerfor he Fort JamesonNgonthan for-the Lamba. Can we therefore ay thatdivorce is less frequent amnong he Fort JamesonNgoni than among the Lamba ? The informatioavailable from Tables I and - I suggests that thiunqualified question cannot be answered, for weobserve that as compared with the Fort Jameson

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 10/27

Measures fDivorce requencyn Simple ocieties 45

TABLE III. DivorceRatio8: Pereentage8fMarriage8 ndingn Divorce

Percentageof marriages nding n divorcecalculatedby variousmethods Number ofdivorces

Society in sample

A B C Method not stated

Bemba (a) ... ... 20-5 43Bemba (b) ... ... 35 0 33Bemba (c) ... ... ... ... ... 44 0 43England and Wales, 1938-39 (d) ... . 2 *Fort JamesonNgoni (e) ... ... .. 285 558 36* 115Kgatla (f) ... ... ... ... 3 0 9Kgatla (g) ... ... . .. 3 5 14Lamba (h) ... ... . ... 331 61P3 41 8 119Muria j) 2... ... ... ... ... 295 59Ngwaketse g) ... ... ...' 5*7 17Ngwato (g) ... ... ... ... 9.3 61Nuer (k) ... ... ... . 9-4 25 13-0 3Palestinian Arabs (in) ... ... 4 3 7 7 87 11Tlokwa (g) ... ... ... . ... 3-8 8

United States, 1926 (p) ... ... 26United States, 1928 (q) ... ... . 18Yako (r) ... ... ... ... ... 57 66Yao (s) .... ... ... ... 34 6 1;82 41 3 11,6

Sources(a) Brelsford, ited in Richards, 1940, p. 120, as Sample A, male informants. The number of divorces n all the samples

of Brelsford nd Richards has been inferred rom he size of the sample.(b) Richards, loc. cit., Sample B, male informants.(e) Ibid, Sample C, female nformants.(d) Hajnal, 1950a, p. 181.(e) Barnes, 1951b,Table XI.(f) Schapera, 1940, p. 294.(g) Schapera,citedby Gluckman n Mitchell& Barnes, 1950, p. 16. Method communicated rally.(h) Mitchell& Barnes, 1950, p. 47.

(j) Elwin, 1947, pp. 616f,634f. He distinguishes arious categoriesofmarriage, hepercentages nding n divorcerarnginfrom 3 to 11L7.(k) Extracted fromEvans-Pritchard, 1945, pp. 31ff. Leviriticmarriage treated as merely continuation f earlier egal

marriage.(in)Extracted fromGranqvist,1931, pp. 157ff, sing ive male informants nly. Betrothals excluded.(p) Willcox,1940, p. 348, Table 176.(q) Cahen,1932, p. 18. Cahen and Willcoxuse different ethods nd neither s identicalwith that described n thispaper.(r) Forde, 1941, as citedby Gluckman n Mitchell& Barnes, 1950, p. 15.(8) Mitchell,1949, p. 298, Table IV, female nformants.

TABLE IV. FortJamesonNgoni and Lamba DivorceRatios

r_____________________ SocietyRatio _2 P

Lamba FortJamesonNgoni

A33-1% (119/360) 28.5% (115/403) I 183 -1<P< *2B 61P3% (119/194) 55 8% (115/206) 1 79 1 1 <P< 2C 41*08% 119/285) 36.9% (115/312) 1P50 2<P< -3

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 11/27

J. A. BARNES

Ngoni, the Lamba have a lower percentage ofdivorced men, but a higher percentage of divorcedwomen; their men have fewer divorces per head,but theirwomen have more. Further,we can examinein the usual way the possibility that the differencesbetween Fort Jameson Ngoni and Lamba in Table IIImay have arisen by chance, in a statistical sense.

The calculations forthis are set out in Table IV.From Table IV it is clearthat the differencesetweenthe Fort Jameson Ngoni and Lamba divorce ratios,as shown by these samples are not statisticallysignificant. We would require further nformationbefore we could attach any importance to the factthat the Lamba ratios are slightly higher than theNgoni.

V. SAMPLESDRAWN FROMTHE EXPERIENCEOF ONE SEX: AN ILLUSTRATION

We can use ratios B and C to illustrate the argu-ments put forward in section III for separating

marriages for which both partners are informantsfrom hoseforwhichonly one partner s an informant.Our data referto the Fort Jameson Ngoni.

Information was collected concerning all themarriages experienced by wedded (i.e., ever-married)men and women n six villages. One village, yielding107 informants, as twenty-fivemiles away and in adifferent ounty3 from another village yielding 55informants. The remaining four villages, yielding173 informantsbetween them, were close togetherbut in a third county and twentymiles away fromthe other two. These fourvillages were those mostaccessible frommy base camp, built at a site chosen

by the Administration. The othertwo villageswereselected for me to live in by the county chiefs con-cerned. I was able to compare these six villageswith many others and in so far as size, ethnic com-position, wealth, historyand kinship structure areconcerned, they are fairly typical Ngoni villages.These six villages constituteabout one per cent. ofthe total Fort Jameson Ngoni population.

I endeavoured to include all wedded people asinformants ut failed to do so. A few people werealways away from home when I called on them, as,for example, a brickmaker who worked long hoursevery day at a Mission and who went offdrinkingevery week-end. 'A few other people who wereinterviewedwere excludedfrom he sample sincetheirhistories s givento me wereeithergrossly runcated,internally nconsistent, rpatentlyuntrue. I amendedthe histories related by yet a few other inforniantsin the lightof information eceivedfrom heirneigh-

3The area controlled by a minor chief I have called acounty. See Barnes (1950a, p. 199.)

bours and kinsfolk. Whenever possible, I returnedto the informants concerned and checked theemendations with them. Even after this processI occasionallyhad reliable nformationn one or moremarriages xperienced y an informant utinadequateinformation n othersknownor thought o have beenexperiencedby him (or her). In these circumstances

I rejected his entire history, excluding from mysample even his adequately recorded marriagesexcept where his spouses were also informantsSome of these rejected informants old me of theirmore respectablemarriages and concealed marriagesthat had ended unhappily. Had I included,theirpartial histories I would have introduced bias infavour of long and happy marriages.

In all 117 men and 222 women were interviewedand theirmarital histories tilizedto form hesample.They constituted86 per cent. of theweddedmenand88 per cent. ofthe wedded womenwhowere normallyresidentin these six villages during 1946-49. The

preponderance of women over men is due to theabsence of more men than women as migrantlabourers n the towns. The men and womenin thesample had experienced 403 marriages. Both manand wifewere interviewed n 111 marriages,while inthe remaining 292 marriages only one partnerwasinterviewed. In the course of this and otherenquiries nformationwas acquired about manyothermarriages, of both living and dead people. Forreasons discussed in section III, this additionainformation as notused at all inmakingcalculations

All marriages collected in this way were dividedinto four categories,E, D, H and W, as follows:-

E. Marriagesextant at the end of the enquiry(1948 in one village and 1949 in the otherfive, including marriages in which thespouseswere separated but not divorced)

D. Marriagesended in divorce.

H. Marriages ended by death of the man.

W. Marriagesended by death of the woman.

The sample was also divided into three parts, asdescribedabove in section III. Divorce ratios werecalculated foreach part separatelyand for the whole(unweighted)sample.

Had I conductedmy enquiries throughmen only

I would have collected only parts 1 and 2 of thisample. Divorce ratios were calculated for thestwopartstakentogether, nd for he similarwomen'sample formedby adding parts 1 and 3. The distributionof marriagesin four categories and threparts, and divorce ratios B and C forthese differesamples, are set out in Table V.

The distribution fmarriages n the threeparts imuch as might have been anticipated. The larg

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 12/27

MIeasuresfDivorce requencyn Simple ocieties 47

TABLE V. FortJamesonNgoni DivorceRatios,distinguishingetween ale and Female Experience

Marriages n each category Divorce ratios

Basis of calculation _ _ _ __

E D H W Total1 B C

1. Marriages for which bothpartnerswereinformants 103 7 1 (a) 0 111 87-5% ( 7/8 ) 6 4% ( 7/110)

2. Marriages for which menweretheonlyinformants... 22 42 0 26 90 61 8 % ( 42/68 ) 65 6% ( 42/64

3. Marriages orwhichwomen |weretheonlyinformants.. 72 66 63 1 (b) 202 50-8% ( 66/130) 47 8% ( 66/138)

Total ... ... ... 197 115 64 27 403 55.8% (115/206) 36.9% (115/312)

Allmarriagesforwhichmenwere nformants ... ... 125 49 1 26 201 64.5% ( 49/ 76) 28 2% ( 49/174)

All marriages for whichwomen were nformants .. 175 73 64 1 313 53O0% ( 73/138) 25-8% ( 73/248)

Notes.-(a) After had collected hemaritalhistoriesofman andwife nthismarriage, nd before had finished heenquirythe man died.

(b) I interviewed hiswomanbefore he-died. I also interviewed erhusband,but since I could not get adequateinformation n some of his othermarriageshe was not used as an informantn thissample.

number of extant mnarriagesor which women werethe only informantswas due to labour migration,where men had gone away and lefttheirwives behind.Some of the extant marriagesforwhich men were

the only informantswere polygynousmarriages inwhich the couples lived in separate villages.The divorce ratios for part 1 differwidely from

the rest, as could be expected. Ratio B, the simplemeasure of the relative frequency of death anddivorce as a method of terminatingmarriage, doesnot vary widely between the other five possiblesamples,with a minimumvalue of 5(0 8 per cent. forpart 3 and a maximum of64 5 per cent. for hemen'ssample. Since part 3 containsall, except one, of themarriages erminatedby death ofthe man, the lowervalue of B for part 3 is a reflection f the fact thatmarriages re more ikely to end by death of the man

than,of the woman. Ratio C shows a greaterrangeof variation, from 25 8 per cent. for the women'ssample to 65 6 per cent. forpart 2. The unweightedsample gives a higherC than either the men's or thewomen's sample for, as we have discussed earlier,an unweighted ample is likelyto be biased in favourof marriages ended in divorce. We may, perhaps,regardthe figure f 36 9 percent.for he unweightedsample as an upper limitofthe truevalue of C.

Unfortunately, did not systematically collectinformationbout thewhereaboutsofformerpouses,although I did ascertain that in many instancesinformantsknew neither where their former pouses

were living nor whether they were alive or dead.A division of parts 2 and 3 into 2a, 2b, 3a and 3b,therefore annot be undertaken. I have used theunweightedsample, with a certain amount of mis-giving, in all the subsequent calculations in thispaper.

VI. MARRIAGE URATIONTABLE

A more elaborate method of analysis depends onknowing how long each marriage lasts. With thisadditional informationwe can calculate the proba-bility that a marriagewill end in divorce or deathwithin a given time and hence compute the mean

expectation of married life at the beginning ofmarriage. We canprovide fuller ictureofmarriageand divorce than is possible by means of a simpleratio. Calculations of this general kind have beenmade for the United States (Cahen, 1932; Jacobson,1950) and England and Wales (Hajnal, 1950a), butI donot know ofany previousdata for imple ocieties.The methods of calculation used in modernsocietiesare not applicable to simple societieslacking divorce

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 13/27

48 J. A. BARNES

registration,but comparable calculations can bemade using a sample of marriagesarrived at by oneof the procedures discussed in section III. Thecriticisms evelled in that section at this kind ofsample continueto apply, but theredoes not appearto be any better basis available. The method ofcalculation is explained in detail as it does not seem

to have been applied to this kind ofmaterialbefore.Using the sample of marriages from the FortJameson Ngoni described in section V, I estimatedthe year in whicheach marriage began and, forcom-pletedmarriages, he year in which t ended. A fewinformants ould state the year themselvesand fora fewothers twas possibletoget accurate nformationfromMission records. For the rest, estimateswerebased on the number of years elapsed since theevent; its relation n timeto the life ofthe informantand to dateable historicalevents; and by referenceto the presentage of childrenborn of the marriage.In this way marriages were classified in terms oftheir

duration,that

is, how long theylasted or, withextant marriages, how long they had lasted so far.The longest duration encounteredwas 53 years, theshortestone week.

These marriages were already divided into fourcategories,E, D, H and W, defined arlier n connec-tion with divorce ratios. The numberof marriageswith the same duration and in -the same categorywas computed for each year from0 to 53, and foreach of the four categories. The resulting distribu-tion is set out in coluimns , 5, 8 and 11 ofTable VI.In this table the duration of the marriages n anyparticular cell is referredto in column 1 as thenominal value of the cell.

Estimates of the beginningand end of marriageswere made only in years and not all the marriagesgrouped together n one cell of the table lasted forexactly the same lengthof time. For example, wemay consider the cell with nominal value 10 yearsin column 3, in whichwe have 6 marriages. Theseincludemarriageswhose lengthshave been estimatedin two differentways, directly and indirectly. Insome instances I was able to ascertain directlythatthe marriage lasted ten years, to the nearest year,and I then made an estimate as best I could of thecalendar years nwhich hemarriage egan and ended.In other nstances proceeded ndirectly. I was able

to establish that a marriagebegan in such and sucha year,and ended in such and such anotheryear,butI could not tell whether these events had occurredat thebeginning,middle or end oftheyears concerned.The arithmeticaldifference etween the dates wasten years and therefore included the marriage inthe 10-year ell. In fact, helength f such marriagemay have been anything between just over nineyears to just under eleven:years. This variation in

possible actual length applies at every year in thetable, so that those marriages ncludedin the 11-yearcells may have lasted anythingbetween just overten years and just under twelve years. Each cellhas thus a coverage in realityof two years, one yearon eitherside of its nominal value, and there is anoverlap of a year on either side in the coverage of

successive cells. For purposes of calculation wemake the assumptionthat the coverage of each cellis only one year, half a year ont ither side of itsnominal value, thereby eliminatingoverlap. Thus,for example, we assume that all marriages in the10-year cells lasted for between nine years and ahalf and ten years and a half. This assumption isjustified for those marriages whose lengths wereestimated directlyand does not introduceany greaterrorwhen applied to indirectlyderived durationsWe make the further ssumption that in each cellmarriagedurations re distributed niformlyhroughout the period of a year coveredby the cell.

With these assumptions made, the 1-year cellsinclude all marriageswhich lasted between six andeighteenmonths. The 0-year cells have a span ofonly six monthsas they include all marriageswhosedurations lie between nothingand six months. Thedurations of all marriages in the 0-year cells wereestimated directly.

Columns 3, 5, 8 and 11 of Table VI contain theraw data of the sample. All subsequent operationsdescribed in this paper consist in manipulatingthefiguresn these columns.

Althoughall the extant marriageswill eventuallyend in divorce or death, we cannot say of any

particular marriage which way it will end. If weuse only completedmarriageswe introduce the biasin favour of short-livedmarriagesmentionedearlier.Therefore,we utilize at each stage of calculation thefact that after a given number of years such andsuch a numberofmarriageshave not yet ended. Tofacilitatethis,we abandon the customary procedureby which we would begin by examining successiveyears of marriage duration and consideringwhatevents take place during each year. Instead, webegin by taking points of time so many years afterthe beginningof marriageand considerwhat is theend resulton oursample ofmarriages fall the events

that have occurredbetweenthebeginning fmarriageand the given point of time. Ini other words weexamine what is the end result of so many years'exposureto the risk ofdivorceand death.

We can best explain this process by consideringfor xample, what happens to our sample duringthefirstten years and a half since the beginning ofmarriage. We divide all the marriages n our sampleinto five classes, as follows:

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 14/27

Measures of DivorceFrequency n Simple Societies 49

P. Marriageswhich asted at least tenyearsanda half.

Q. Marriageswhich ended in divorce withintenyearsand a half.

R. Marriages which ended in death of the manwithinten years and a half.

S. Marriages which ended in death of the

woman withinten years and a half.T. Extant marriageswhichhave not yet lasted

ten years and a half.

If we could return to the fieldseveral years hencewe would then be able to reclassify ll the marriagesnow in class T and distributethem among the firstfour classes. For purposes of calculation we makethe assumptionthat these marriages n class T willeventuallybe distributed mong the first our classes

so as to preserve their relative size. Such anassumption ignoresany trendthere may be, for theclass T marriages are, on the whole, more recentthan those in other classes. It also overl6oks anyunevenness in the distributionof extant marriagesamong the different ear cells, for ideally this dis-tribution hould be the same as the marriage urvivaltable that we shall eventuallyderivefromthe table.This assumption,however,providesus witha simpleway ofconstructing first pproximation o a marriagedurationtable.

Applying this assumption to the point of time,ten years and, a half after the beginningof marriagewe have that the proportion f marriageswhich endin divorce by this time is given by the ratio of thenumber of marriages in class Q to the sum of thenumbersin classes P, Q, R and S.

TABLE VI. MarriageDurationTable

Column

1 2 3 4 5 6 7 8 9 10 11 12 13

Nominal Assumedvalue of coverage E U-2,E D ED 6/04x H LH 100 x W 2W 100 xcell in of cell in 100 9/4 W W 12/4years years

0 0 - j 16 387 9 9 2-3 0 0 0 0 0 0

1 -1 9 378 10 19 5.0 0 0 0 1 1 0*32 1 21 19 359 8 27 7-5 1 1 0-3 4 5 1-43 2j-31 13 346 11 38 11*0 2 3 0?9 0 5 1-4

1 31- 41 12 334 13 51 15*3 6 9 2-7 0 5 1-55 4- 51 11 323 13 64 19*8 2 11 3*4 1 6 1.96 51 61 5 318 10 74 23*3' 2 13 4*1 2 8 2-5

7 61- 7 5 313 2 76 24*4 5 18 5*8 2 10 3*28 7 81 5 308 10 86 27*9 7 25 8*1 1 11 3169 81- 9 11 297 2 88 29*7 4 29 9-8 0 11 3-7

10 91-10 6 291 5 93 32*0 4 33 11*3 3 14 4-8

11 lO -llj 4 287 4 97 33.9 2 35 12*2 0 14 4*912 11f-121 4 283 5 1UF2 36*0 2 37 13'1 0 14 4-913 12J-131 5 278 1 103 37*1 3 40 14*4 1 15 5-4

14 13j-141 4 274 2 105 38*4 0 40 14*6 2 17 6*215 144-15J 4 270 0 105 39*0 0 40 14 8 3 20 7 416 15j-161 1 269 3 108 40*3 2 42 15*6 0 20 7-4

17 16j-171 7 262 2 110 42-0 0 42 16*1 0 20 7 618 174-181 0 262 0 110 42*0 2 44 16-8 0 20 7*61 9 18j-191 2 260 1 111 42*8 0 44 16*9 1 21 8 1

20 19j20j 4 256 0 111 43*2 4 48 18*8 0 21 8*2

21 20Q-221i 2 254 0 111 43*8 0 48 18-9 1 22 8*722 21j-22i4 3 251 0 111 44*1 2 50 19*9 0 22 8*823 22J-23,1 3 248 1 112 45*2 0 50 20*2 0 22 8'9

(2499) D

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 15/27

50 J. A. BARNES

TABLE VI. MarriageDuration Table-contd.

Column

1 2 3 4 5 6 7 8 9 10 11 12 13

Nominal Assumed Ivalue of coverage EE UOZE D I D 100 XI H ZH 100 W W100 X

cell in ofcell in E -~ D 6/4 H H 9/4 WW 12/4years years - _ - _ - _ I _ _

24 231-24 5 243 0 112 46*2 3 53 21*8 0 22 9.125 241-251 3 240 0 112 46*8 0 53 22*1 0 22 9 226 251-261 2 238 0 112 47*1 0 53 22*3 0 22 9-3

27 26J-27 1 3 235 1 113 48-2 0 53 22*6 0 22 9*428 271-28 1 1 234 1 114 48*8 1 54 2311 1 23 9-829 281-291 4 230 0 114 49 6 2 56 24 4 0 23 10 0

30 29-1301 0 230 0 114 49 6 1 57 24*8 1 24 10*4

31 30j-1j 1 2 228 0 114 50*0 3 60 26*4 0 24 10532 31i-321 2 226 1 115 51-0 0 60 26-6 0 24 10-633 32J-331 0 j 226 0 115 51*0 0 60 26*6 0 24 10 6

34 331-34j 0 226 0 115 51*0 0 60 26*6 0 24 10*635 341-35j 3 223 0 115 51*7 0. 60 27*0 0 24 10*836 35j-36j 3 220 0 115 52-4. 1 61 27*8 0 24 10.9

37 361-371 1 219 0 115 52*6 0 61 27*9 1 25 11*438 371-381 2 217 0 115 53-1 0 61 28-2 0 25 11-539 381 391 2 215 0 115 53*6 0 61 28*4 1 26 12*1

40 399-401 0 215 0 115 53*6 0 61 28*4 0 26 12*1

41 401-41 1 214 0 115 53*8 0 61 28-5 0 26 12-242 411 421 1 213 0 115 54*0 0 61 1 287 0 26 12-243

421-4311 212 0 115 54 3 0 61 28:8 0 26 12-3

44 431 441 3 209 0 115 55.1 1 62 29*2 1 27 12-945 44F-451 0 209 0 115 55*1 0 62 29*2 0 27 12*946 451 461 0 209 0 115 551 1 63 30*2 0 27 12*9

47 46j-471 0 209 0 115 55.1 0 63 30*2 0 27 12*948 47j-48j1 0 209 0 115 5511 0 63 30*2 0 27 12*949 481 491 2 207 0 115 55-6 0 ! 63 30*4 0 27 13-1

50 49-1-50 1 206 0 115 5581 0 63 30*6 0 27 13*1

51 501511 0 206 0 115 55*8 0 I 63 30*6 0 27 13*152 511 521 0 206 0 115 55-8 0 63 30 6 0 27 13d153 52j-531 0 206 0 115 55*8 1 64 31*1 0 27 13*1

Total 197 115 5456 2298 3 64 2498 1082 0 27 1076 463 4

LegendE (Extant) Marriages tillextant.D (Divorce) Marriages nded in divorce.H (Husband) Marriages nded bydeath ofhusband.W (Wife) Marriages ndedby death ofwife.U (Universe) Total numberof marriages n sample (403).

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 16/27

Measures of Divorce Frequency n Simple Societies 51

The numbers of marriages n classes Q, R and Sare obtained for each year by adding successivetermsof columns 5, 8 and 11 of Table VI, and theresultsare set out in columns6, 9 and 12. The sumof classes P, Q, R and S is obtained by subtractingthose in class T fromthe total numberof marriagesin the sample. These sums are obtained for each

year by subtractingsuccessive terms of column 3

from hetotal, and the results re setout in column4.The proportionof marriageswhich end in divorceby a given time after the beginningof marriage isfoundby dividingthe appropriateterm n column 6by the corresponding erm in column 4. The resultis shown as a percentage in column 7. Thus, forexample, n the 10-year ellswe have that93marriagesout of 291, or 32 per cent.,end in divorce withintenyearsand a halfofthebeginning fmarriage. Corre-sponding percentagesfor termination f marriagebydeath ofman or ofwoman withinthe same time areprovided in columns 10 and 13. These percentagesindicate what, in fact, happened in our sample.

Making the assumption as to typicality discussedearlier,we may regardthem as indicatingthe proba-bility t thebeginning fmarriage hat any particularmarriagewould end in a certainway within a giventime.

By this method we ignore any changes that mayhave taken place in the last fifty ears or so, forwemixtogethermarriageswhichbegan at differentimeswithin that period. Our probabilities are specificfor duration only. Thus, a divorce in 1915 in amarriage begun in 1910 is treated identicallywith adivorce in 1947 in a marriage begun in 1942. Weignore the possibilitythat divorce was in any case

more,likely n 1947 than in 1915, whateverwas thedurationofmarriage. We also ignorethe possibilitythat one marriagewas childless and the other not,that one was a firstmarriageand the othera third,that one concerned headman and the other a schoolteacher, and so on. If our sample were larger itwould, however, be relatively easy to isolate thesedifferentategoriesof marriageand then to proceedwitheach category n the same way as we have doneforthe whole sample. Thus, forexample, we couldprepare separate tables for all the marriages enteredinto in each decade, for the year in which eachmarriage began is known. With only 403 marriagesto manipulate these refinements re, unfortunately,not practicable.

In column 5 of Table VI it will be seen that nomarriageended in divorce aftermore than 32 yearsof married life. Yet in column 7 the probabilitythat a marriagewill end in divorcewithin giventimecontinuesto increase slowlybut steadily beyond thethirty-thirdear. Here the limitations four methodof calculation become apparent, since this rise does

not reflectany social fact but is merely artificialAs we deal with successively onger periods of timeelapsed since marriage,so the completed marriagein our effectivesample begin to outnumber theextant ones. Hence our basis of calculationbecomesmore and more like that used earlier in calculatingdivorce ratio B, consisting of only completedmarriages. We have observed that ratio B is biasedin favourof short-livedmarriagesand indicates morefrequent divorce than do other indices. This samebias is gradually introducedinto Table VI until inthe last line of the table a result is presentedwhichis identical with that given by ratio B. Both fromratio B in Table V (total sample) and from the lastline of Table VI, column 7, we have that theprobability that a marriage will eventually end indivorce is 55 8 per cent. Yet we have also fromTable VI, column 7, that the probability that amarriage will end in divorce in 321 years is only51 per cent. This adds weight to our criticism ofratio B as a measure of divorce frequency when

it is based only on marriages experienced by livingpeople.

VII. PROBABILITY o0 DIVORCE WITHIN A

GIVENTIMEThe figures f Table VI provide a basis for making

a variety of furthercalculations. First we maycompare the probability in different ocieties ofdivorcewithina givennumberofyears elapsed sincethe beginningof marriage. There is information nthis for he UnitedStates in 1928 Cahen, 1932, p. 118;Nimkoff,1947, p. 633) and for England and Wales

(Hajnal, 1950a, p. 181). Probabilities in thesecountrieshave been calculated forintegralnumberof years aftermarriage,while the probabilitiesgivenin Table VI, column7, relate to mid-points fyearsAs a preliminarywe have therefore o calculate theprobabilities at integral numbers of years for theNgoni. We assume that these are the arithmetimeansofsuccessivepairsofprobabilities t mid-pointofyears. Thus, at the 9- and 10-yQar ells we havethat the probabilityof divorce taking place within91 years is 29 7 per cent. and within10- years 32 0per cent. The arithmeticmean ofthese is 30 85 percent., taken to be the probabilityof divorce within10 yearsofmarriage. The calculation ofthese meansis set out in columns 1, 2 and 3 of Table VII. Thisaveragingprocesshas thedesirableeffect fsmoothingourprobability urve. In viewofthebias introducedinto the later years of Table VI, the calculation hasnot been carriedbeyond the thirty-firstear.

The resultantprobabilitiesfromTable VII, column3, are repeated in Table VIII, together with com-parable figures relating to the United States and

(2499) D 2

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 17/27

52 J. A. BARNES

TABLE VII. Co&putationTable for Shifting atum Point8

Column 1 2 3 4 5 6

Years Probability of divorce Rate of divorceper annum at a Yearselapsed within a given time giventime elapsedsince _ _ _ since

beginning - - beginningofmarriage Data from Sums f Differences I Ofmarriage

Table VI, pairs of New data between puirsof NeW ratescol. 7 termS successive paterms

probabilities

0 0 0 0 01

I (2.3 x 2=)4-6

2-3 4-0 11 7-3 3-65 2-7 1

5-0 5-2 2-62 12-5 6-25 2-5 221 7-5 6-0 3-0 2I

3 18-5 9-25 3-5 33j 11.0 7-8 3.9 314 26-3 13-15 4-3 441 15-3 8-8 4.4 4I5 35-1 17-55 4-5 55j 19-8 8-0 4-0 516 43 1 21 55 3-5 , 66j 23-3 4-6 2-3 67 47.7 23 85 1.1 77j 24-4 4-6 2-3 78 52-3 26 15 3-5 881 27 9 5-3 2 65 819 57-6 28-8 1-8 99j 29-7 4 1 205 91

10 61 7 30 85 2 3 1010 32-0 4-2 2 1 101

11 65-9 32-95 1.9 11III 33.9 4-0 2-0 11I12 69 9 34.95 2-1 1212j 36 0 3-2 1 6 12413 73-1 36 55 1-1 13131 37 1 2 4 1 2 13i14 75.5 37.75 1 3 1414j 38-4 1-9 0-95 14115 77.4 38-7 0-6 1515 39-0 1-9 0 95 1516 79.3 39-65 1 3 16161 40 3 3 0 1-5 16117 82-3 41 15 1*7 1717j 42 0 1*7 0*85 17118 84-0 42-0 0 1818 42*0 0-8 0 4 18119 84-8 42:4 08 1919j 42-8 1'2 0-6 19j20 86 0 43 0 0 4 20201 43*2 1.0 0-5 20121 0*6 2121j 43*8 211

29j 49-6 29230 99-2 49-6 30830 | 49|6 30i

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 18/27

Measures of Divorce Frequency n Simple Societies 53

England and Wales.4 The same results re presentedgraphically n Diagram I.

TABLE VIII. Probability at Beginning of Marriage thatwithin givenTime a Marriagewill have ended n Divorce

SocietyTime in yearselapsed sincebeginning f Fort United EnglandMarriage Jameson States and Wales,

Ngoni 1928 1938-39

0 0 0 0

1 3 65 0-7 0o012 6 25 2-0 0*023 9*25 3*4

4 13*15 5 05 17 55 6*4 0-26 21 55 7.7

7 23 85 8.88 26-15 9.99 28-8 10*8

10 30 85 11-6 0 9

11 32 9512 34 9513 36*55

14- 37 7515 38 7 14-516 39 65

17 41 1518 42-0

19 42*4

20 43 0 16 1 2 0

30 49 6 17 -9 2 6

From this diagram it will readily be seen that,although we are accustomed to think of divorce inthe United States as " common" or " frequent,"it is comparativelyrare as compared with divorceamong the Ngoni. This is so despite the lowermortality n the States by whichcouples are exposedto the risk of divorce for ongerthan theywould be

were theyto die as quickly as do the Ngoni.

4 I do not discuss the manner in which I have adaptedthe data on these modern ocietiesto suitmy purpose,sincethis does not seem relevant to the themeof this paper. Ihave merelychosen the most easily accessible sources andmanipulatedtheirdata arbitrarilyo obtainmeasures coulduse forcomparison.

VIII. DURATION-SPECIFIC IVORCE RATES

A further omputationbrings out other differenceInstead of considering he cumulative effect f yearsof exposure to divorce risk, we can examine theprobabilitythat divorce will occur in any given yearof married ife. We begin by subtracting uccessiveterms in Table VI, column 7. This gives us the

number of divorces which take place during theperiod covered by each cell per original cohort of100 marriages, Thus, for example, we have fromcolumn 7 that out oT 100 marriages 29*7 willhaveended in divorce within 91 years and 32-0 within10 years. Therefore2*3 marriages (the differencbetween these two figures)will end in divorce duringthe period 91l-10- years after the beginning ofmarriage. Since the coverage -ofeach cell is 'one

TABLE IX. DivorceRatesSpecific orDurationofMarriageexpressed s Divorcesper annum per 100OriginalMarriage

Society

MarriageDuration Fort United Englandin Years Jameson States, and Wales,

Ngoni 1928 1938-39

4 0 0-7 0.01

2.6 1.3 0.0121 3*0 1*4

31 3.9 1-6 0*06

4-1 4.4 1*4

51. 4 0 1*3

61 2*3 1*1

71 2.3 1.1 0.14

8I 2X65 0*9

91 2.05 0.8

101 2.1

11L 2.0

121 1.6 0 6

13i 1.2

141 0.95

15 0 11

15' 0.95

1611.5

171 0.85 0*3

181 0.4

191 0 6

201 0.5

25 102 0!06

(2499) D3

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 19/27

54 J. A. BARNES

year, we may regardthese figures s indicatingratesper annum per 100 original marriages at whichdivorces occur at different urations of marriage.The 0-yearcell covers only six months,so we haveto double the figure btained by subtraction o obtaina rate per annum. We can compare the annualrates so obtained with those relating to the UnitedStates and England and Wales, but first he sameaveragingproceduremust be carried out as withtheprobabilities of divorce. We may regard our ratesas operating at the mid-pointsof cells, that is, at1, 1, 2, 3, . . . years aftermarriage. For example,since 2-3 marriages end in divorce between 91 and101 years afterthe beginning fmarriage,we assumethat the rate per annum at 10 years' duration s 2 3divorcesper 100 originalmarriages. For direct com-parison we need to know the rates at 1, 1, 2 1years. We make a similarassumption as before hatthese newrates are the arithmeticmeans ofsuccessive

pairs of old rates and calculate accordingly. Thiscalculation is set -out in columns 4, 5 and 6 ofTable VII. The rate at half a year after marriageis calculated by adding two-thirdsof the rate at aquarter ofa year and one-third f the rate at one year.

The rates of Table VII, column 6, are repeated inTable IX, together with the United States andEngland and Wales rates. The same informationsshown graphically n Diagram II.

The similarity between the Ngoni and UnitedStates curves is noticeable. In both the rate risesto a maximum round about the fifth ear of marriageand thereafter lowly declines. The modal year isthe fifth or the Ngoni and the fourth or the UnitedStates. In England and Wales the second quin-quennium is modal.

These curves do not indicate at what duration therisk of divorce is greatest. They merely ndicate atwhat durations most divorces occur. This is so

TABLE X. Calculation of Divorce Rates related to Surviving Marriages

Colunn 1 2 3 4

Years elapsed Percentage of Rate of divorce Rate of divorce

sie beginning marriages terminated Percentage of related to 100 related to 100

of marriage by divorce or marriages surviving original survivingof marriagedeath marriages marriages

22 497 7 4 0 4*1

1i 5-3 94-7 2-6 2-7521 9f92 90*8 3*0 3-331 13-3 86-7 3.9 4.5

41 19*5 80-5 4-4 5.551- 25*1 74.9 4*0 5-3561 29-9 70-1 2-3 3-3

71 33-4 66-6 2-3 3-58 iff 39-6 60 4 2 65 4.4gi 43-2 56-8 2-05 3.5

10 4811 51-9 2-1 4-05

11 1 5154 49 0 2-0 4-1121 54 0 46-0 1-6 3-513 56 9 43 1 1 2 2 8

142 59*2 40-8 0 95 2.3151 61*2 38 8 0 95 2-45161 63-3 36-7 1*5 4-1

172 65-7 34*3 0-85 2-5181- 66-4 33-6 0414219i 67-8 32-2 0-6 19

20i 70-2 29-8 0*557

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 20/27

Measures of Divorce Frequency n Simple Societies 55

because the rates we have calculated are related toan original cohort of so many marriages. Thedifference etween these two kinds of measurementm&y be illustrated by a hypothetical example.Suppose that the fiftieth ear of marriage s the mosthazardous of all, so that very few couples enteringtheir fiftieth ear of married ife togethermanage to

survive o the end oftheyearwithout eekingdivorce.Even under these conditions t is unlikelythat mostof the divorcesgrantedwould be to couples in theirfiftieth ear, for the hazards of death and divorcethat lie in the firstforty-nine ears of married lifewouldensure hat fewcouplesever reach thebeginningof their fiftieth ear. Therefore, t any particulartime couples passing through heir fiftieth earwouldconstitute only a small part of the total weddedpopulation, and the divorces granted to them,evenif they were all divorced,would account for only asmall part of all the divorces granted. To gain anidea of the rate at whichmarriages hat have reachedany particular duration are then terminated bydivorce,we must referthe divorces taking place ineach year of duration to the number of marriagessurviving o that duration.

This can be done by dividing the number ofdivorces occurring n each period of a year by thenumber of marriages survivingto the mid-pointofthat period. This calculation is set out in Table X.Column 1 is formed y adding together orrespondingtermsfrom columns 7, 10 and 13 of Table VI andgives the percentage of marriages terminated bydivorceor death withina giventime from he begin-ning of marriage. These percentagesare subtractedfrom 100 to give, in column 2, the percentage of

marriagessurviving. Column 3 repeats the divorcerates per annum already obtained in Table VII,column6. In Table X, column 3 divided by column2 and multiplied by 100 gives column 4, which isthenthe requiredrate of divorce relatedto survivingmarriages.

These rates may be comparedw-ith imilarlycom-puted rates for the United States (Cahen, 1932,p. 120) and England and Wales (Hajnal, 1950a,pp. 181, 185). The comparativerates are set out inTable XI and showngraphically n Diagram III. 5

Unlike all the preceding measures of divorcefrequencywe have discussed, this one is free from

the effects fmortality, xcept inso far as

mortalityaffects he compositionof our sample of marriages.It is thereforewell suited for comparing divorcefrequency in societies with different mortality

rMorerecent data indicatethat in 1948 the rates fortheUnited States were considerablycloser the Fort JamesonNgoni than in 1928. Cf. Jacobson (1950, pp. 238, 240, 243,Charts2, 3 and 5).

TABLE XI. DivorceRatesSpecific orDurationofMarriage,expressed s Divorcesperannumper 100Marriagessurviving

to thatDuration

Society

MarriageDuration Fort United England

in Years Jameson States, and Wales,Ngoni 1928 1938-39

4-1 0-71 0*01

2*75 1'23 0.0121 3*313731 4- 5 1P38 0-06

41 5*513551 5-35 1P186j 3*3 1.09

7i 335 1P07 0.1581 444 0*94

91 3-

0 76101 4*05 0 83

1i 441 0*64121 3 5 0 56131 2 8 0*51

14j 2*3 0-52

15 0.13

15j 2*45 0 *50

161 4 1 0*44

17j 2*5 0 4318j 1.2 0 41

191 1;9 0-36

20j 1*7 0*37

25 0*08

The general effectof relatingdivorces to survivinratherthan originalmarriages s to increase the rateof divorce at later years of marriage.6

The generalsimilarityn shape betweenthe Ngonand United States curves is maintainedwith a peakaround about the fifthyear. The fifthyear ofmarriageamong the Ngoni (the fourth n the United

6 The United States rates in Table XI are not all highethanthe correspondingates n Table IX. Those n Table XIare takendirectly romCahen. For Table IX I have piecetogether s best could nformationrom ahen andNimkofCahen makes allowance for the effect f migration. Thimakes thebasis ofcomparison etween he twosets offigurrather exiguous. Unfortunately, hen drafting his paperI overlooked the work of Monahan (1940).

(2499) D 4

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 21/27

56 J. A. BARNES

States) is seen to be the timewhen the riskof divorceis greatestas well as the time at which most divorcesoccur. The Ngoni curve behaves differently romthe othertwo during he firstyear ofmarriage. Thedivorce rate is higher during the first year thanduring the second. This reflects real differencendivorce procedure in the three societies. In occi-

dental societies divorce is usually sought onlyafterconsiderabledeliberation and the legal processlasts a long time. Most people enter into marriagehoping that it will last for life and relinquish thishope only gradually. The divorce rate during thefirstyear of marriage is thereforerelatively low.Amongthe Fort Jameson Ngoni a bride,even thoughmarriage payments may have been made for her,may run away fromher husband after only a fewdays and within a short while will be divorced:Despite its shortduration, this is a propermarriageand is quite distinct.fromtemporaryconcubinage.Hence the divorcerateduring he first earof marriageis relautively igh among the Ngoni. The graphicalmethod of presentationadopted here gives a clearindication of the numerical importance of theseshort-livedmarriages.

This difference etween the Ngoni and occidentalsocieties demonstrates he point made earlier of howimportant t is to state clearlywhat is meant in anyparticular ociety by marriage nd divorce; Apparentdifferences ay be due as much to varyingdefinitionsof these termsas to real differencesn habit.

IX. MEAN EXPECTATIONS AND MEDIAN

DURATIONS

Some of the, nformation ontained n the foregoingtables -can be summarized by calculating the meanexpectation of married life at the beginning ofmarriage. We can consider various categories ofmarriage and obtain a mean expectation for eachcategory.

Our first calculation is based on the figures ncolumns 7, 10 and -13of Table VI, utilizingall themarriages in our sample, including those that areextant. The sums of corresponding erms of thesethree columns, some ofwhichhave been worked outin Table IX, column 1, are the percentages ofmarriageswhichwill have been terminatedby death

or divorce within a given time. Subtracted from100, they generate a marriage survival curve, thefirst ermsof which are set out in Table IX, column 2.By integratingunder this curve we can deduce themean number of years of married life expected atthe beginning of marriage.

The area under this curve, allowing for the factthat thecoverageofthe 0-yearcell s onlysixmonths,whereas subsequent cells each cover a year, is found

by inspectionof the totals of columns 7, 10 and 13of Table VI to be

[100 + 100 - (2.3 + 0 + 0)]/4+ [100 - (2.3 + 0F 0) + 100 - (5.0 + 0 +

0 3)]/2 + ..

...+ [100 -(55.8 + 30.6 + 13.1) + 100 (558+ 31-1 + 13.1)]/2

54 terms

100 X 54 + (2* )/4 - (2298 3 + 1082.0+ 463 6)

100 x 15-549This is the aggregatenumberof years of married ifelived by couples in 100 marriages. The meanexpectation of married life at the beginning ofmarriage s therefore 5 5 years. It shouldbe noted,however,that this figure s an under-estimate, or tdoes not allow for the fact that the couples in theextant marriages of our sample will live additional

yearspfmarried ifebefore heirmarriages re ended,It mightbe thoughtthat we could correctforthiserrorby considering nly completedmarriages. Thiswe do by using the totals of columns 6, 9 and 11 ofTable VI. The area underthe correspondingurvivalcurve is found to be

206 X 54 + 9/4 (5456 + 2498 + 1076)

i.e.,206 x 10X28

The mean expectationof married ife t the beginninof marriage, consideringonly completed marriagein our sample, is therefore10-3 years. This figuris less,notmore,than ourfirst alculated expectation

The difference f over fiveyears between these twomean expectations s a good illustration f the extentto which calculations based only on completedmarriages experienced by living people are biasedin favourof short-livedmarriages.

Next we consider only those marriageswhich endin divorce. Using column 6 alone, the area underthe corresponding urvival curve is foundto be

115 x 54 + 9/4 5456

115x 6 58Thus, if we knew at the beginning f a marriagethat

it was destined o end n divorce, hemeanexpectationofmarried ifewould be 6*6 years.Similar expectations, consideringonly those mar

riages ending by death, are 15* years for thosending by death of the man, 14 1 years for thosending by death of the woman, and 14* years fothese two categories taken together.

These results are compared in Table XII withsimilar expectations for the United States (Cahen

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 22/27

Measures ofDivorce Frequency n Simple Societies 57

1932, pp. 124f) and England and Wales (Hajnal,1950a, pp. 181f, 185).

TABLE XII. Mean Expectation f MarriedLife atBeginningfMarriage

Mean expectation n years

Basis of calculation Fort United EnglandJameson States, and Wales,Ngoni 1928 1938-39

All marriages ... ... 15-5 20-4 35 4Completedmarriages

only ... ... ... 10*3Marriages ended in

divorce ... ... 6 6 7 14 3Marriages ended by

man's death ... ... 15*0Marriages ended by

woman's death ... 14*1

All marriagesended bydeath ... ... ... 147 23 360

From this table we see that, although divorce ismuch more frequent among the Ngoni, the meandurationof marriagesendingin divorce is much thesame as in the United States. The mean expectationfor each category of marriage is lower among theNgoni than in the United States and lower therethan in England and Wales. In all three societiesmarriages ending in divorce are, on the average,shorterthan marriages ended by death. Termina-tions n divorceare characteristic f the early yearsof

marriage and terminationsby death of the lateryears. This contrast is less marked among theNgoni than in the other two societies,- s mightbeexpectedfrom he highermortality mongtheNgoni.

A further summary of some of the foregoinginformation an be made by calculating the medianduration of different ategories of marriage. Weconsider the same marriage categories as used formeans. Table X, column 2, shows the percentageof all marriagessurvivingaftera given time,whilesums of corresponding ermsof columhs 6, 9 and 12of Table VI give the numberof marriagesnow com-pleted, out of a total of 206, that were terminated

-within given time. Taken separately,these threecolumns ndicate thenumberofmarriages erminatedwithin given time n each smallermarriagecategory.The resulting median durations are set out inTable XIII, together with comparable mediandurations for England and Wales (Hajnal, 1950a,pp. 181f, 185).

For the Ngoni the mean exceeds the median ineach categoryof marriage, ndicatingthat the mean

TABLE XIII. Median Duration ofMarriage

Median duration inyears

Basis of calculationFort England

Jameson and Wales,Ngoni 1938-39

All marriages ... ... ... 11 30-39Completed marriages only ... 7Marriages ended in divorce ... 5 10-19Marriages ended by man's

death ... ... ... ... 10Marriages ended by woman's

death ... ... ... ... 10All marriages ended by death ... 10 38

expectations of married life are greater than theywould otherwise be because of a few couples whosemarriages ast a long time. For England and Walesthe only clear comparison of means vnd medianswhich the data permit is of marriages ending bydeath. Here the mean is less than the medianindicatingthat the mean expectation of married ifis less than it would otherwisebe due to a few coupleswhose marriagesare ended by death comparativelearly.

X. COLLECTING INFORMATION

Our discussion so far has been mainly about howwe can manipulate our data. We may now outlinthe procedure to be followed in the field which

should ensurethat we obtain the rightkind of data.We assume that we sample by

(1) delimiting segmentof the society studiedsuch as a villageor clan or politicalsector,,and hen

(2) examining he life historiesofall the relevanindividuals belongingto the chosen segment.

We should indicate how the segment was chosenhow it was delimited,and its size, both absolutelyand in relationto the whole society. It is desirablthat the kind ofsegment electedshouldbe such thateach individual in the society belongs to one, andonlyto one, segment,but this is not always possible

It is also desirable that the segments hould not varywidely in size; if they do, two (or more) sampleshould be made, one froma large segment and onefrom small.

Often we lack information n individuals who, inthese terms,should formpart of the sample. Evenif we can findnothingelse about them, we shouldendeavour to findout how many there are of theseIn tabulations showing he distribution f ndividual

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 23/27

58 J. A. BARNES

into differentategories, s in a marital status table,we can then include these individuals in an"unknown " category. Where our units of tabula-tion are not individuals but events they haveexperienced, uch as marriage nd divorce,we cannotdo this. We can, howeVer, ndicatethe completenessof our sample by showing how many informantswe

used out of the total number possible, as well asproviding an " unknown" category for those eventsknown to have been experienced by our informantsbut forwhich furthernformations lacking.

The minimum mount of information eeded fromeach individual, man or woman, adult or child, inthe sample to enable us to make the computationsdiscussed in this paper is as follows

1. Name.2. Sex.3. Date of birth.4. Marital history howingfor each marriage:

(a) date ofcommencement,

(b) date oftermina-tion,f not stillextant,(c) whether erminated y deathordivorce,and

(d) name of spouse.In some societieswe may be able to determine hesedates accuratelv while, in others, we may have toestimate n decades. The morepreciseour raw datathe greater the' precision of our measures, but themethod employed remains essentially the samewhether we are workingto the nearest year or thenearest fiveyears.

It will be seen that the amount of informationrequired is not great. The collection of these data

can be easily combined with other systematicenquiries into marriage and fertility, inship affilia-tion, labour migration and the like.

XI. CONCLUSION

Weare nothereconcernedwitb the fulldemographicanalysis of marriage termination by death, but itshould be noted that Table VI can be used in thisconnection s well. For example, columns 10 and 12ofthat table give the- robability hat a marriagewillend by the death of one spouse or the otherwithinagiventime. At every year beyond 31 years termina-tion smore ikelyto have taken place by the death of

the man than of the woman. Among the FortJamesonNgoni, as in most societies,men are usuallyolder than theirwives and this may be part of theexplanation. There are nearly twice as many womenas men in the sample and whereas some of the menwho used to live in the sample villages and whosewives have died are now labourmigrantsn thetowns,comparatively few widows have become migrants.The figures s they stand appear to show that the

additional hazards of childbearing do not maketermination y death ofthewomanas likelyas mightbe expected. Any investigationof this point wouldhave to allow for the unusual sex ratio and fordifferentialates ofmigrationfor men and women.

13y calling upon further aw data, such as age ofspouses at marriage,whetherthey were related to

one another,whether heyweremarried as Christianor pagans,howmanychildren hey had,whatrelativestheylived with,and the likp, t is possible to analysedivorce frequencyfurther. In our own society weoften emphasize, and probably over-emphasiz(cf. Jacobson, 1950, pp. 242ff),how divorce is morelikely n a barren hanin a fertilemarriage. It wouldbe well to know to what extent this is true of othersocieties where the rearing of children, and theirsocial importance, is regarded differentlyrom ourown (cf. Hsu, 1949, p. 105). Frequencies and ratescan be calculated specificnot only forduration butalso forman's age,-or forreligion,or social status.The particular line of enquiry followedwill dependon the problem n hand.

Divorce is not one " thing" and cannot be describedby a single ndex or a single adjective. It is a socialprocess which has many aspects, some of which wecan measure and some we cannot. Divorce in twosocietiesmay be similar n one respect and differenin another, s we have seenby contrasting heUnitedStates and the Fort Jameson Ngoni. Some aspectsare related to each other ogicallyand we can defininadvance the ogicalconnection etween heirvariousmeasures, as, for example, we have noted ratio Bas the end-pointof the probabilitycurve shown inpart in Table VII, columns 1 and 3. We may

assume a priori that there are other connectionsinreality betwecn aspects of divorce which may nowappear independent. These connections re propertienot of our analytical tools but rather of the humansocietieswe examine, and we may hope to discoverbhem by further nquiry in the field. For example,is it true that as divorce becomes morefrequent hetime of greatest risk shiftsfromthe second to thefirstquinquennium of marriage,as is suggested bythe data from our three societies? In other wordsif divorce is more likelydoes it come sooner? Thiskind of question can best be answered for simplesocietieswiththehelp oftools such as those described

inthispaper. Wemaythengoon to considerwhetheornotwe are ustifiedntreating s similarphenomenthe comparativelyrare terminationsof marriage nsay, Zulu and Nuer societies, arisingfrom conditionof extreme marital stress, and the comparativelfrequent erminationsn, say, Yao and Lozi societieswhichappear to arise oftenfrom rivial precipitatincauses and which may have quite different ocialfunctions.

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 24/27

Measures of Divorce Frequency n Simple Societies 59

I am well aware I have constructed a ratherpretentious superstructureon raw data which arequite unsuited to uphold it. My plea must be that,

if we draw attention to the uses to which data ofthis kind can be put, workers n the field may beencouraged to collect more, and to collect better.

DIAGRAM 1. Probability t beginning f marriage hatwithin giventime marriagewill have ended n divorce

..r

O < = -r : | 5V8f17DJetATESjf92.4 j =

Z

-U

. 1 2 3 4 5 6 7 9 ? 1 20 SO3

Time in years elapsed smnce belnrmncy f makrria&g

s __ ~~~~~~~~~~~~~~~~~~~~~~~~__3

DIAGRAM 2. Divorcerates, pecific orduration fmarriage, xpresseds divorces er annum. er I 00 originalmarrivages

L ~ ~~~~foer2wsctoovlv/

iC- -7 - _ I _ _ __ _ __ 1\1 11

AS 1928

0 A

o 1 2 3 4 5 6 7 3 9 to 15 20

JAarria5mduration ngr

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 25/27

60 J. A. BARNES

DiAGRAM 3. Divorcerates, pecificor duration fmarriage, xpressed s divorces er annumper 100marriagesurvivingothaduration

Foer JAsa ONI I-

0 - - - - -0

P.5

UN)TE'D JTiEs /928

Mari-ia?ge durabion in yeare.

ReferencesBRNss, J. A. ... ... ... ... 1947 " The Collection of Genealogies." Rhodes-Livingstone J., 5.

- ... ... ... ... 1951a "The Fort Jameson Ngoni," in Colson, E., & Gluckman,H. M., eds.,Seven Tribes ofBritish CentralAfrica. London: Oxford UniversityPress.

- -- ... ... ... ... 1951b Marriagein a Changing ociety. Capetown: OxfordUniversity ressBARTON, R. F. ... ... ... 1949 The Kalingas. Chicago: University of Chicago Press.BOWERS, A. W ... ... ... 1950 Mandan Social and CeremonialOrganization. Chicago: University f

Chicago Press.BROWN, G.,G., and HUTT, A. McD. B. 1935 Anthropologyn Action. London: Oxford University Press.CAHEN, A . ... ... ... 1932 Statistical Analysis of American Divorce. New York: Columbia

UniversityPress.CHILDS, G. M. ... ... ... 1949 Umbundu Kinship and Character. London: Oxford University Press.CULWICK, A. T ... ... ... 1935 " A Method of studying Changes in Primitive Marriage," J. Roy

anthrop. nst., 65.

CULWICK,A. T., and CuLWIcK, G. M. 1935 Ubena of theRivers. London: Allen and Unwin.... ... ... ... 1938-39 "A Study of Population in Ulanga, Tanganyika Territory." SociolRev., 30, 31.

Du Bois, C. ... ... ... ... 1944 The People of Alor. Minneapolis: University of Minnesota Press.EARTHY, E. D . ... 1933 Valenge Women. London: Oxford University Press.ELKIN, H. ... ... ... 1940 "The Northern Arapaho of Wyoming, "pp. 207-258, in Linton, R.,

ed., 1940.ELWIN, V. ... ... ... ... 1939 The Balsga. London: Murray.

-... ... ...1947 The Muria and their Ghotul. Bombay: Oxford University Press.EVANS-PRITCHARD, E. E. ... ... 1945 Some Aspects fMarriageand the amily among heNuer. Livingstone:

Rhodes-Livingstone nstitute.

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 26/27

Measures of Divorce Frequency in Simple Societies 61

References

FBI, H-T. ... . .. ... 1939 Peasant Life in China. London: Kegan Paul.FIELD, M. J. ... ... ... ... 1940 Social Organization f theGa People. London: CrownAgents for the

Colonies.FIRTH, R.... ... ... ... ... 1936 We, the Tikopia. London: Allen and Unwin.FORDE, C. D. ... ... ... ... 1941 Marriage and the Family among the 'Yako in South-Eastern Nigeria.

London: Lund Humphries.FORTES, M. ... ... ... ... 1943 " A Note on Fertility among the Tallensi of the Gold Coast." Sociol.

Rev., 35.- -- ... ... ... ... 1949 The Web of Kinship among heTallensi. London: OxfordUniversity

Press.FURER-HAIMENDORF, C. VON ... ... 1943 The Chenchus. London: Macmillan.

-...I ... ... 1945 The Reddis of theBison Hills. London: Macmillan.GLUCKMAN, H. M. ... ... . . 1950 " Kinshipand Marriage mongtheLozi ofNorthernRhodesia and the

Zulu ofNatal." In A. R. Radcliffe-Brown and D. Forde, eds., AfricanSystems of Kinship and Marriage. London: Oxford UniversityPress.

GOODWIN, G. ... ... ... ... 1942 The Social Organization of the Western,Apache. Chicago: Universityof Chicago Press.

GRANQVIST, H. ... 1931 and " Marriage- Conditions in a Palestine Village." Societas Scientiarum1935 Fennica, CommentationesumanarumLitterarum, , 6. Helsingfors.

GREEN, M. M. ... ... ... ... 1947 Ibo Village Affairs. London: Sidgwick and Jackson.HAJNAL, J. ... ... ... ... 1950a " Rates of Dissolution of Marriages in England and Wales, 1938-39."

In Reports nd SelectedPapers oftheStatisticsCommittee. (Papers of

the Royal Commission on Population.) London: H.M.S.O.... ... ... ... 1950b " Births, Marriages and Reproductivity, England and Wales1938-1947." Ibid.

HARRIS, J. S. ... - ... ... 1940 " The White Knife Shoshoni of Nevada." In Linton, R., 1940.HARRISON, T. ... ... ... ... 1937 Savage Civilization. London: Gollanez.HERSKOVITS, M. J. ... ... ... 1938 Dahomey. New York: Augustin. (2 vols.)Hsu, F. L. K. ... ... ... ... 1949 Under the Ancestor's Shadow. London: Routledge.HULSTAERT, P. G. ... ... ... 1938 Le Marnage des Nkundo. Bruxelles: van Campenhout.HUNTER, M. ... ... ... ... 1936 Reaction to Conquest. London: Oxford University Press.JACOBSON, P. H. ... ... ... 1950 " Differentials in Divorce by Duration ofMarriage and Size ofFamily."

Amer. sociol. Rev., 15.JOFFE, N. F. ... ... ... 1940 " The Fox of Iowa." in Linton, R., ed., 1940.KLuciKHOHN, C., and LEIGHTON, D. 1946 TheNavaho. Cambridge, Mass.: Harvard University Press.KRIGE, E. J., and KRIGE, J. D. ... 1943 The Realm of a Rain-Queen. London: Oxford University Press.KucZYNSKI, R. R. ... ... ... 1944 "Population Movements: the Contribution of Demography to the

Study of Social Problems." Rhodes-Livingstone J., 2.

KUPER, H. ... ... ... ... 1947 An African Aristocracy. London: Oxford University Press.LA BARRE, W. ... ... ... ... 1948 "The Aymara Indians of the Lake Titicaca Plateau." Amer.Anthrop.,1, 1, 2 (Mem. No. 68).

LANDES, R. ... ... ... ... 1937 Ojibwa Sociology. New York: Columbia University Press.... ... ... ... 1938 The Ojibwa Woman. New York: Columbia University Press.

LEACH, E. R. ... ... ... ... 1940 Social and Economic Organization of the Rowanduz Kurds. London:Lund Humphries.

LEVIN, R. ... ... ... ... 1947 Marriage in Langa Native Location. Capetown: University of CapeTown.

LINTON, R., ed...... ... ... ... 1940 Acculturation in Seven American Indian Tribes. New York: AppletonCentury.

MCILWRAITH, T. T. ... ... ... 1948 The Bella Coola Indians. Toronto: University of Toronto Press.MAIR, L. P. ... ... .... ... 1934 An African People in the TwentiethCentury. London: Routledge.MEEK, C. K. ... ... ... ... 1937 Law and Authority n a Nigerian Tribe. London: Oxford University

Press.MINER, H. ... ... ... ... 1939 St. Denis. Chicago: University of Chicago Press.MITCHELL, J. C. ... ... ... 1949 "An Estimate ofFertility n some Yao Hamlets in Liwonde District

of Southern Nyasaland." Africa, 19.MITCIELL, J. C., and BARNES, J. A. 1950 The Lamba Village. Capetown: University of Cape Town.MONAHAN,T. P.... ... ... ... 1940 " The Changing Probability of Divorce." Amer. sociol. Rev., 5.NADEL, S. F. ... ... ... ... 1942 A Black Byzantium. London: Oxford University Press.

........ ... ... ... 1947 The Nuba. London: Oxford University Press.NIMKOFF, M. F.... ... ... ... 1947 Marriage and theFamily. Boston: Houghton Miffin.OPLER, M. K. ... ... ... ... 1940 " The Southern Ute of Colorado." in Linton, R., ed., 1940.PAuLME, D. ... ... ... ... 1940 Organization Soocale des Dogon. Paris: Domat-Montchrestien.POWDERMAKER, H. ... ... ... 1933 Life in Lesu. London: Williams and Norgate.REDFIELD, R. ... ... ... ... 1941 The Folk Cuxlture f Yuhcatan. Chicago: University of Chicago Press.

This content downloaded from 192.134.152.160 on Thu, 8 Aug 2013 05:37:33 AMAll use subject to JSTOR Terms and Conditions

8/22/2019 Measure of Divorce Frequency in Simple Societies

http://slidepdf.com/reader/full/measure-of-divorce-frequency-in-simple-societies 27/27

62 J. A. BARNES

References

RIcHARDS, A. I. ... ... ... 1940 Bemba Marriage and Present Economic Conditions. Livingstone.Rhodes-Livingstone Inst.

SCHAPERA, I. ... ... ... ... 1940 Married Life in an African Tribe. London: Faber.SIEGEL, M. ... ,.. ... ... 1942 " Effects fCultureContactontheForm of heFamily in a Guatemalan

Village." J. Roy. anthrop. nst., 72.SMITH, M. W. ... ... ... ... 1940 The Puyallup-Nisqually. New York: Columbia University Press.SPICER, E. H. ... ... ... ... 1940 Pascua, a Yaqui village in Arizona. Chicago University of Chicago

Press.SPICER, R. B. ... ... ... ... 1949 "People on the Desert." Part I, in A. Joseph, R. B. Spicer, and

J. Chesky,The DesertPeople. Chicago; University fChicagoPressSTAYT, H. U. ... ... ... ... 1931 The Bavenda. London: Oxford University Press.WAGLEY, C. ... ... ... ... 1949 "The Social and Religious Life of a Guatemalan Village." Amer.

Anthrop., 1 (Mem. 71).WAGNER, G. ... ... ... ... 1949 The Bantu of North Kavirondo. London: Oxford UniversityPress

(Vol. 1).WHITING, J. W. M. ... ... ... 1941 Becoming Kwoma. New Haven: Yale UniversityPress.WILLCOX, W. F.... ... ... ... 1940 Studies in AmericanDemography. Ithaca: CornellUniversityPressWILLIAMSON, R. W. ... ... ... 1939 Essays in Polynesian Ethnology. Ed. R. Piddington. Cambridge:

Cambridge University Press.WILSON, G. ... ... ... ... 1942 An Essay on the Economics of Detribalization in Northern Rhodesia.

Livingstone: Rhodes-Livingstone Inst. (Vol. 2.)WISDOM, C. ... ... ... 1940 The Chorti Indians of Guatemala. Chicago: University of Chicago

Press.