Zambezia (1995), XXII (ii).THE POPULATION DYNAMICS OF THE TEMBOMVURAPEOPLE OF ZAMBEZI VALLEY, NORTHERN ZIMBABWE:SOME METHODS OF COLLECTING AND ANALYSINGBIRTH HISTORY DATA1RAVAI MARINDO-RANGANAIDepartment of Sociology, University of ZimbabweAbstractCollecting quantitative demographic data from illiterate populations continuesto be a major drawback in the study of demographic phenomena. Theproblems of illiteracy and lack of numeracy are more acute in Africa wherethe majority of the rural populations are not well educated. This articledescribes a method of collecting quantitative data using simple techniques ofdrawing and modelling. Using a small semi-nomadic population of theTembomvura people of Zambezi Valley, northern Zimbabwe, the studyconcludes that it is possible to collect birth history data from illiteratepopulations. A combination of the traditional questionnaire approach andthe 'participatory' methods proposed in this article provide a more reliablemethod for collecting birth history data. With large populations aiming fornationally representative samples, the method would be very time consumingand costly. The approach is more useful for studying small populations orsmall samples. The major strength is that it gives the respondents a chance toexpress their knowledge in simple numeric form.DATA DEFICIENCIES CONTINUE to be a major drawback in the study ofdemographic levels and trends in Sub-Saharan Africa. The need forquantifiable data, usually in large sample sizes, has led to the use of theprecoded quantitative questionnaire as the main method of collectingdemographic data. However, most rural populations in Africa from whomdemographic data are collected are illiterate and have little if any cultureof numeracy. In Zimbabwe, 25% of women aged 15 years and above and14% of men aged 15 years and over reported themselves as illiterate(those who have not completed Grade Three) in the 1992 census (CSO,1995). These rates of illiteracy were believed to be underestimates sincethe respondents were not actually asked to prove their literacy. What can1 This study was part of a Ph.D. in Medical Demography done at the London School ofHygiene and Tropical Medicine under the supervision of Basia Zaba. Funding for fieldwork forthis study was provided by Population Council, New York.177178 THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEbe done to make the demographic questionnaire more usable with illiteratepopulations?One approach has been the development of indirect estimationtechniques which rely on data collected from simple questions. Forexample, indirect estimation of adult mortality can be done from simplequestions on the survival of mother or father (Blacker, 1977, p. 28) andfertility and childhood mortality can be estimated indirectly from data onchildren ever born and children dead (Brass, 1968, p. 68; 1971, p. 24; andUN, 1983, p. 53). These indirect approaches have to some extent alleviateddata deficiencies but their major weakness is the retrospective nature ofthe estimates.In spite of the simplicity of data required for indirect estimates, theremay be need to adjust and/or impute the data because they are eitherinaccurate or coverage is incomplete. Such adjustments and imputationsintroduce their own errors (Meekers, 1991, p. 249).CURRENT FERTILITY AND THE NEED FOR THE BIRTH HISTORYQUESTIONNAIREIn studying current changes in fertility, the birth history questionnaireremains our major source of demographic data. In Sub-Saharan Africa, therecent interest in whether fertility has declined has highlighted the needfor reliable sources of current fertility, since the debate focusses onrecent changes in fertility which cannot be obtained from indirectestimates.The birth history questionnaire has undergone some refinementsever since the World Fertility Survey (1978) but has retained most of itsbad characteristics, which are an excessive length and a purely quantitativeapproach that is not user friendly when information is obtained fromilliterate populations (see both 1988 and 1992 Zimbabwe Demographicand Health Surveys' birth histories). The challenge that faces demographersis how to make the birth history questionnaire more applicable to peoplewith little or no numeracy. This study attempts to provide some suggestionsto meet this challenge, using a case study of the Tembomvura people ofZambezi Valley, Zimbabwe.DESCRIPTION AND OBJECTIVES OF THE STUDYThis article discusses a simple procedure to refine a birth historyquestionnaire in an attempt to make it more usable with illiteratepopulations. An attempt is made to shift away from the traditional codedquestionnaire as the centre of the data gathering process by focussing onmethods referred to either as 'microapproaches' (Caldwell, Hill and Hull,1988, p. 43) or 'participatory' (Chambers, 1992, p. 15).RAVAIMARINDO-RANGANAI 179Using data from the Tembomvura people, the study examines anddiscusses how period fertility data was collected through the use of acombination of methods which involve group interviews, participatorymodelling and diagramming, dramatisation, and a demographicquestionnaire. Focussing on the small nomadic population where all thepopulation was functionally illiterate, the study aims to show how localpeople can present their own concepts through the use of simple methodslike dramatisation, building models or drawing.The study also discusses and shows how the period data was adjustedand analysed to give estimates of current fertility, that indicated somerecent changes which had affected the population.The study provides some alternative methods of demographic datacollection, relying on techniques already present in many rural populations,which include drawing, the use of local materials like seeds to buildmodels, dramatisation and focus groups. The study does not totally doaway with the birth history questionnaire, but attempts to make it moreusable by taking account of the context within which data is collected,focussing on the creation of rapport and allowing local people to expresstheir knowledge in their own way. The birth history questionnaire becomespart of a cumulative process of information gathering rather than somethingimposed on the population.BRIEF BACKGROUND OF THE STUDY POPULATIONThis study is based on a small semi-nomadic population of the Tembomvurapeople. Commonly known in Zimbabwe as the Dema (a name which theydeplore) the Tembomvura reside in Chapoto Ward on the Western side ofthe Mwanzamutanda River close to the border of Mozambique, Zambiaand Zimbabwe.The Tembomvura community has a total population of 7802 peopleand of these 235 are in the reproductive ages 15 to 49 years. Their livelihoodis based on a combination of clandestine hunting in Chewore and Dandegame reserves, selling their labour to the neighbouring Chikunda people,minimum agriculture, gathering the tubers Disocorea bulbiferra (manyanya),Tacca leontapetaloid.es (bepe), and Boscia angustofolia (mupama) as wellas fishing and honey extraction. According to their oral history they werehunters and gatherers until 1974, when they were first resettled by theRhodesian government.In 1978, the Tembomvura people, together with their neighbours theChikunda, were moved from Chapoto Ward and resettled in a security2 This figure is based on a census done by the researcher during the period of research inJuly 1992.180 THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEFigure 1CHAPOTO WARD AND ITS PROXIMITY TO CHEWORE AND DANDE GAMERESERVESROADTARMAC ROADRIVERAPROX SCALE Um» 1?kmHARARERAVA! MARINDO-RANGANAI 181village in Mashumbi Pools. Many Tembomvura people escaped from thevillage and went back to their original home in Karemwa within Cheworegame reserve until 1980. From then onwards, more effort was made toresettle the Tembomvura and to encourage them to engage in agriculture.These resettlement efforts have been met with resistence by theTembomvura, who regard agriculture as inferior to hunting and try theirbest to continue with a nomadic lifestyle.In the 1990s current efforts at sustainable utilisation of wildliferesources through the activities of CAMPFIRE (Communal AreasManagement Programmes for Indigenous Resources) led to the introductionof control by mostly the Chikunda people (a neighbouring agriculturalcommunity residing in Chapoto Ward) which prevented the Tembomvurafrom hunting in the game reserve. Gathering for tubers and honey in thegame reserves was also controlled and became defined as 'loitering in thegame reserve with intent to commit a crime'.The socio-political dynamics of the CAMPFIRE programme at villagelevel have led to the exclusion of Tembomvura people, and this hascreated ill-feeling between the Tembomvura, wildlife officials and the localsafari operators (Marindo-Ranganai and Zaba, 1995). In addition theTembomvura were experiencing a drought which was at its height in 1992when the survey was done. With their traditional life under threat due tothe hunting ban and gathering being policed, they began to express fearsthat their community was facing demographic extinction due to fewerbabies being born and mortality being very high because of food shortages.DATA SOURCES AND METHODOLOGYData for the study is based on a demographic, socio-cultural and healthsurvey which was carried out among the Tembomvura from July 1992 toNovember 1992.3 The data gathering process used was hierarchical,whereby a combination of qualitative methods using participatory modelsand diagrams, group interviews and a demographic questionnaire were allcombined to provide detailed information on current fertility. The datagathering approach is shown in Figure 2 below.BRIEF DISCUSSION OF DATA GATHERING METHODSParticipatory modelling and diagrammingParticipatory models and diagrams are methods of gathering data fromlocal populations by using models and diagrams. By using materials like3 Data collection took five months from the beginning of July to the end of November 1992.All demographic data was collected in a cumulative process and since data was entered in thefield, cross-checking and recalling was also done during this period. The first fertility interviewstook one month but cross-checking and re-interviews continued until the end of the study.182 THE POPUUTION DYNAMICS OF THE TEMBOMVURA PEOPLEFigure 2QUANTITATIVE AND QUALITATIVE METHODS USED IN DATA COLLECTIONQuantitative Data CollectionHousehold Questionnaireon births in the household.All heads of householdsinterviewed. Also infor-mation on births in the lastyear.Women's individual ques-tionnaire. All women aged15 and above were inter-viewed.Completing birth historyquestionnaire. Only wo-men who had a birth in thelast 10 years were inter-viewed on date of birth,season of birth, etc.Qualitative Data CollectionDetails of people's views on theirfertility, changes in marriagepatterns, number of births, changesin population size. Groups of menand women interviewed.Participatory modelling anddiagramming of children everborn, dramatisation usingother children around.Calendar of local events,collection of information onseasons, birth practices. Onlywomen were involved ingroups.Plausible estimate of fertility: how do the demographicestimates relate to the qualitative information?RAVAI MARINDO-RANGANAI 183leaves, mud and stones, local people are assisted in making models oftheir environment as they see it. To facilitate the spirit of participation,researchers are encouraged to let the local people 'hold the stick' indrawing their diagrams and to give the locals a chance to express theirknowledge. The process of handing over part of the research to locals is anecessary step towards participation (Marindo-Ranganai, 1994).In this study, participatory modelling was used to create a local mapof people's dwellings and to model the whole population using seedsarranging them according to their relative ages. Seeds were placed on theground indicating how many people were in which age group. In additionwomen used the seeds of Ziziuphus mauritania (masawu) and Tamarindusindica (musika) to model their fertility histories. Each woman with thehelp of a spouse or her older children or friends arranged the fruits on theground with the Ziziuphus mauritania representing female children andthe Tamarindus indica indicating male children. Through this modellingprocess, the woman created her own birth history based on local materials.Dead children were represented by stones.Once a woman had made a model of her own fertility, she was askedto indicate the relative ages of the children.In some cases, women were given coloured powders to make somepaint, which they used to draw their children and produce a birth history.Using different colours they indicated first the different sexes, then therelative ages of the children. Paper diagrams were less popular thanmodels although ground diagrams were popular with some of the women.Diagramming was found to be useful for the estimation of relative agesbecause it was easier to show age differences than on models.Although the participatory nature of the methods have been questioned(Rifkin, 1994, p. 22), they did give women a chance to express theirknowledge in numerical form.During the course of the survey, groups of children usually followedthe researcher around as she talked to the women. Some of these childrenwere used to dramatise the fertility histories of the women, standing infront of a particular woman while she pointed out and arranged themaccording to their ages and sexes as if they were her children. Deadchildren were represented by stones because most of the living childrendid not want to dramatise the part of a dead child.To cross-check the birth history models created mostly by women,focus group discussions were held with fathers and they were askedinformation about their children, the children's names and ages. Some ofthe models were shown to the men and they were asked to comment orcontribute to the women's presentations.A demographic questionnaire was used at the final stage of the studyafter the participatory modelling and diagramming as well as group184 THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEinterviews. A birth history questionnaire was completed for each womanand details of each birth, place of birth, date of birth, season of birth andif the child died, details of death place, cause and age were all collected.Because rapport had been created and almost all women were known, itwas possible to cross-check the birth history by diagramming data andgroup information.The combination of the methods was found to be extremely useful inthe collection of birth history information. The major criticism providedin the post-mortem group discussion was that there was too muchrepetition Š that the participatory models, the group interviews and thedemographic questionnaire all asked the same questions. It was thenexplained to the respondents that the idea was to cross-check theinformation and the respondents suggested that perhaps the researchershould learn a bit of trust.The smallness of the population made the application of the methodseasier and by the end of the study all the women were known personallyby the researcher. The demographic questionnaire was completed as partof a cumulative process and within some form of cultural context. Analysisof the birth history data is discussed in the next section.DATA ANALYSISA detailed birth history for each of the women, with information on thenumber of births, the year of birth (based on a historical calendar) theseason of birth, the current development stage of each child (for thoseunder the age of one), the sex of each child and on dead children Š thesex, age, year and season of death of each child and the cause of death asperceived by both parents was collected. In addition, from each household,information on whether a birth occurred in the last 12 months was collectedand used for providing the fertility pattern of the Tembomvura (BLY).The data was organised into a child file tabulated by year of birth andalso by year of survey and year of death for the children who died. Awoman's file was also created where the births were tabulated by her age.These combination tables provided the basic data for fertility estimation.The data from qualitative and quantitative approaches from each womanwere put into a database and the different files for each woman cross-checked for consistency. Because of the size of the population, dataanalysis and fertility estimates were not done using existing packages likeSPSS but by designing spreadsheets for manipulation of the data.RESULTSTable 1 shows the period fertility data of births by calendar year. Bearingin mind that the births were collected only from women who are currentlyRAVAI MARINDORANGANAI 185in the reproductive ages (15-49), the births were adjusted for truncationeffects. The adjustment is discussed below.Table 1BIRTHS BY CALENDAR YEAR ADJUSTED FOR WHOLE YEAR ANDTRUNCATION EFFECTS, TEMBOMVURA SURVEY, 1992.Age Group15-1920-2425-2930-3435-3940-4445-49TotalBLY569741133923586420289193373132990169854235892141113642528828101274144870121699315086041013850408505910106040840271374336830056134129SOME MINOR ADJUSTMENTS TO THE BIRTH DATAAccording to household reports, 33 children were born in the year beforethe survey. Using birth history reports, we estimate that 28+29/4 = 35children were born in this time if we assume that lU of all births in 1991occurred in the last three months of the year. This indicates that under-reporting of births last year by the household head is in the order of 6% ifthe assumption of an even spread of births throughout the year is justified.The births history data collected by calendar year had to be adjustedbecause of the following reasons:1. The births in 1992 represented only3/4 of the births expected for thewhole year because the survey was carried out in September. The1992 births were adjusted to represent the whole year by multiplyingby 4/s.2. Birth history data was collected from women aged 15-49 in 1992. Thebirth history covered 1983-1992, which means that data was truncatedas we go back in time since the women who were aged 40-49 years in1983 would be over 50 in 1992 and these did not provide birth historyinformation during the survey.3. The existing births were based only on live women. Using informationon female mortality, the births could be adjusted for maternal mortality.CORRECTING FOR TRUNCATIONTo correct for truncation in the birth history data, the only completeinformation on births available to us is through the births in the last year,186 THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEwhich suggests that '/33 = 3% of births occurred to mothers aged 45-49 inthe year before the survey. This is rather high as most fertility modelsshow that for a non-contracepting population, the ASFR for women aged45-49 is between 1% and 3% of the total (Bongaarts and Porter, 1983, p. 48)and we expect about half of this as a fraction of the total births, due to theage structure of the mothers being weighted towards younger ages.On the other hand, the proportion of all births to mothers aged 40-44reported from BLY, also 3% is probably too low as typically ASFR ofwomen aged 40-44 account for between 5% and 12% of reported TFRs andthis is reduced less when allowing for mother's age structure in calculatingthe proportion of births contributed by these mothers, as the 40-44 agegroup is larger than the 45-49 age group.It would be possible to estimate the proportion of births to womenaged 40-44 from births history data from 1987 to 1992 but we would havefirst to allow for the missing births to 45-49 year olds, as well as reassigningthe births tabulated by current age of mother to the age of the mother atbirth of child.USING STABLE POPULATION MODELSIt is preferable to use stable population models, especially since thenumbers involved are very small. In a population growing at 3% perannum with late child-bearing (i.e. no contraception) we would expectapproximately 2% of births occurring to women over 45 and 10% towomen over 40. We then adjust the births by a factor which lies between2% and 10% for the 40-45 age group in 1992.ADJUSTING FOR MATERNAL MORTALITYThe adjustment for truncation only corrects the reported total births dueto not having interviewed women aged 50-59 who could have bornechildren in the 10 years prior to the survey. It does not adjust for births tomothers who may have died in the interval. Assuming an annual mortalityrate in the 15-50 age group of 0,01 (which roughly corresponds to a 45I15 of0,70, obtained for this population Š Marindo-Ranganai, 1995, p. 196) wecan conclude that births n years ago should be multiplied additionally by(l,01)n, giving the total adjustment factors showing in the last row inTable 2.Figure 3 shows the distribution of births by calendar year. The graphsindicate that the number of births increased steadily from 1983 to 1987. In1988 there was a decline to just above the 1986 figure, but 1989 saw arecovery in the birth rate. After 1989, the births fell to lower levels than forany year since 1984. A three-year moving average smoothens the trend,but the sharp fall in the late 1980s is still visible.Table 2CALCULATION OF ADJUSTMENT FACTORS FOR TRUNCATIONYears prior to survey 123456789 10Last child-bearing age reported 49 48 47 46 45 44 43 42 41 40 >>Proportions of births lost (p) 0,004 0,008 0,012 0,016 0,020 0,036 0,052 0,068 0,084 0,100 >zAdjustment factor for truncation §l/(l-p) 1,004 1,008 1,012 1,016 1,020 1,037 1,055 1,073 1,092 1,111 i>Adjustment factor for >maternal mortality 1,010 1,020 1,030 1,041 1,051 1,062 1,072 1,083 1,094 1,105 >Total adjustment 1,014 1,028 1,042 1,058 1,072 1,101 1,131 1,162 1,195 1,228188 THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEFigure 3 (a)BIRTHS BY CALENDAR YEAR ADJUSTED FOR WHOLE YEAR ANDTRUNCATION EFFECTS6050 -in«-Ł 40 "30 ->jQE3Z20 -10 -831|-»-REPORTEDŁ i i-* ADJUSTEDi I\I1848586 87 88 89Calendar Year9091Figure 3 (b)ADJUSTED BIRTHS WITH THREE-YEAR MOVING AVERAGES[ŁADJUSTED -»-3yr MOVING AVERAGES83848586 87 88 89Calendar Year909192RAVAI MARINDO-RANGANAI 189SOME EXPLANATIONS OF THE LONG TERM PLAUSIBILITY OF RATES OFINCREASEWe can make some rough estimates of the rates of increase and decreaseto check their long term plausibility. The slope calculated for the years1983 to 1988 (which suggest an upward trend) indicates an increase ofnine births over five years, an increase of 1,8 births per year. The meanbirths per year for this period is 46 so that an increase of 1,8 is equivalentto an annual increase of nearly 4%. This is a high rate of increase, but justabout feasible for a rapidly growing population. For the period 1989-91,the three-year moving averages decline steadily at six births per year. Themean annual births for this period is 43 giving an annual rate of decline of14%. This is a massive reduction, unlikely to be sustained for long periods,and in a community with little evidence of voluntary fertility control,probably represents an involuntary response to a crisis situation, such asdocumented by Dyson (1991) in the South Asia famine.It is thought unlikely that the under-reporting of events alone wouldproduce such a pattern of births, as the observed pattern would implythat reporting would be very incomplete in the most recent years, improvesteadily going back over the last five years and then deteriorate slightlyover the preceding five-year interval whereas experience in most fertilitysurveys is one of decreasing completeness going back over time. Also it isunlikely that misdating of events could be responsible: this would requirethat children aged 0-3 were consistently reported as 4-6. Sincedevelopmental milestones are easily noticed at these ages, and all but thedead children were seen personally by the researcher, this is not a likelyexplanation.CALCULATING THE ASFRIn order to try to smooth some of the fluctuations which are due to smallnumbers and minor misdating of events, ASFR were calculated for periods0-4 and 5-9 years before the survey. Assuming that the crisis that led tothe dramatic reduction of births began somewhere between 1987 and1989, the 1983-87 figures can be interpreted as pre-crisis rates and 1988-92 figures as crisis rates.The rates were computed allowing for the age of mother at the time ofbirth in the standard five-year age groups to define numerators of therates and the women years of exposure in the age group as denominators.The births and numbers at risk were back-dated over the five-year period(0-4) or (5-9) years to allow for the movement of the women within andbetween five-year age groups.190THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEa) Calculation of the person years of exposureThe duration of exposure within each age group varies according to theexact age of the women. For example, a woman aged 20 years at thesurvey, will have spent four of the last five years in the 15-19 age group,and one year in the 20-24 age group, whereas a 16-year old would havespent two years in the age group 15-19 and three years under age 15.Figure 4 shows an example of the calculation of person years ofexposure in the period 0-4 years before the survey for women aged 15 to25 showing the movement of a group of women from one age group goingback into the past by one year, two years, three years.Figure 4CALCULATION OF THE PERSON YEARS OF EXPOSURE IN THE PERIOD 0-4YEARS BEFORE THE SURVEYSingleages15161718192021222324Number ofwomen ateach age81110116102957Years before the survey081110116102957111101161029572101161029573116102957! 461029571Person yearsspent in theage group15-19 in theperiod 0-4years beforesurvey = 203For calculating births in the period 0-4 years before the survey, theprocedure is essentially the same. The number of births are tabulated bythe single year ages of the mother as recorded in 1992 and then steppedbackwards to calculate the age of the mother and the number of birthsRAVAI MARINDO-RANGANAI191that the women had one year before the survey, two years before thesurvey e.t.c. An example is shown in Figure 5.Figure 5CALCULATION OF BIRTHS IN THE PERIOD 0-4 YEARS BEFORE THE SURVEYAge of mother insingle years15161718192021222324Births in the period 0-4 years before the survey0002011222110222310101200001201123000112252340010131023Births tomothers aged15-19 in theperiod 0-4years beforethe surveytotal = 34The results of the births divided by the person years of exposure forthe period 0-4 years, 5-9 years and the last 12 months are shown in Table3 below. Because of the shifting backwards of both births and ages ofwomen, these births are tabulated by age of mother at the time of thesurvey.Figure 6 shows the plots of the age specific rates by age. Theadjustments made to the total births to allow for age truncation andmaternal mortality are not incorporated into the age specific fertility ratescalculated here, as the mortality correction would not affect the rates(numerator and denominator would be increased by the same amount).Agegr15-1920-2425-2930-3435-3940-4445-49TotalEst of TFRBirthsin last12months0569741133Table 3BIRTHS BY WOMEN YEARS OF EXPOSURE,Birthsfor 88-9213436473323131*188Mean age at child-bearingStandard deviationWomenyears ofexposure20313717715714010759*980Birthsfor 83-87630525435171na*195TEMBOMVURA SURVEYWomenyears ofexposure13717715714010759*na*777ASFRfromBLY0,1090,1820,2810,2120,1480,0420,0535,1329,27,83ASFRfor88-920,1670,2630,2660,2100,1640,1210,017*6.04*29,07,9ASFRfor83-870,2190,2940,3440,2500,1590,0176,42*27,06,55CDroTHE POPUUVTION C2oCOoH3138C3D«* PEOPLEŁEstimates affected by incomplete or missing data due to truncation.RAVAI MARINDO-RANGANAI193Figure 6AGE SPECIFIC FERTILITY RATES BY AGE OF MOTHER AT BIRTH OF CHILD15-1920-71,?5-?9 30-34 35-39Age Group(.0-4445-Ł ASFR(BtY) Ł ASFR1S8-9?) -fr- ASFR (83-87)FERTILITY LEVELS FROM ASFRIn terms of level, the TFR estimates made from the .births in the last yearsuggest that fertility was 5,1 children per woman whereas that from thebirths in the last five years suggest a TFR of 6,0. From the birth historyinformation on births in the period 5-9 years, the TFR is around 6,4 birthsper woman. The TFR from BLY appears rather low and this suggests someunder-reporting and/or reference period errors.The TFR for both five years before the survey and 10 years before thesurvey are affected by truncation: the extent of truncation error is about1% for the former and 6% for the latter estimates. Allowing for thetruncation, the birth history data would give ASFR estimates of 6,1 and 6,8.These fertility rates are higher for purely nomadic populations (Howell,1975, p. 18) but are comparable to estimates made for semi-nomadicpopulations using indirect procedures (Blurton-Jones et al., 1992).It is possible that our birth history data may be under-reported, evenwhen corrected for truncation error. However, the trend of falling fertilityappears to be supported by other information obtained from thepopulation. There is evidence from qualitative data that the populationexperienced a reduction in the number of births due to the drought and194 THE POPULATION DYNAMICS OF THE TEMBOMVURA PEOPLEfood shortage. Furthermore, from qualitative information, there areindications that Tembomvura people used to have high fertility; colloquialevidence from the neighbouring Chikunda points to the fact that theTembomvura women are very prolific and historically had a tendency tohave large families. This suggests higher fertility in the past.There is evidence from mortality data that childhood and adultmortality increased during the crisis period (Marindo-Ranganai, 1995,p. 253) and this suggests that fertility could have declined through anincrease in intra-uterine mortality which usually increases during famine(Riley et at, 1993, p. 57) and also through increasing maternal mortality. Itappears as if the drought and famine might have led to a reduction infertility. Whether this decline can be sustained over a long period can notbe ascertained from the data.CONCLUSIONLocal communities, even illiterate ones, can provide quantitative data onthe number of their children if methods are used which allow them toexpress their knowledge. The creation of rapport during modelling,diagramming and dramatisation of fertility histories allowed for cross-checking of quantitative information because the respondents and theirchildren were known by the researcher by the time the demographicquestionnaire was completed. By allowing respondents to provide data intheir own way, we may get more reliable data.Although the data may not be perfect, the study offers some ideas onhow current fertility information can be obtained, and how the birthhistory may be refined to make it more usable with illiterate populations.This contribution may be useful to many African demographers who arefrustrated by rural people's lack of numeracy. The study provides a startingpoint from which other ideas may be generated in an effort to collectbetter demographic data from most of the rural populations in Africa.With larger populations however, applications of models and diagramshave to be fine-tuned because they require a lot more time and involvementthan precoded questionnaires.ReferencesBLACKER, J. G. C, (1977) 'The estimation of adult mortality in Africa fromdata on orphanhood' Population Studies, XXXI, (i), 21-32.BLURTON-JONES, N. G.; SMITH, L. C; O'CONNELL, J. F.; HAWKES, K.; AND KAMUZORA,C, (1992) 'Demography of the Hadza, An increasing and high densitypopulation of Savanna foragers' American Journal of PhysicalAnthropology, LXXXIX, 159-181.RAVAI MARINDO-RANGANAI 195BRASS, W.; ANSELY, J.; COALE, F.; HEISEL AND F. LORIMER (1968) The Demographyof Tropical Africa (Princeton, Princeton University Press).BRASS, W., (1971) 'Methods for Estimating Fertility and Mortality fromLimited and Defective Data', Based on Seminars held 16-24 September,CELADA.BONGAARTS, J., AND PORTER, R. G., (1983) Fertility, Biology and Behaviour. AnAnalysis of the Proximate Determinants (New York, Academic Press).CALDWELL, J. C; HILL, A. G.; AND HULL, V. H. (1988) Microapproaches toDemographic Research (London, Kegan Hall International), 44-49.CENTRAL STATISTICAL OFFICE (1995) Women and Men in Zimbabwe: Facts andFigures (Harare, Govt. Printer).CHAMBERS, R. (1992) 'Rapid appraisal: Rapid, relaxed and participatory',Institute of Development Studies Discussion Paper 311 (Brighton,University of Sussex).DYSON, T. (1991) 'On the demography of the South Asian famines part 1',Population Studies, XXXV, (v), 5-25.HOWELL, N., (1975) Demography of the Dobe IKung (New York, AcademicPress).INSTITUTE OF RESOURCE DEVELOPMENT AND CENTRAL STATISTICAL OFFICE (1988)Zimbabwe Demographic and Health Survey Report (Harare, Govt.Printer).MARINDO-RANGANAI, R. C, (1995) 'The Population Dynamics of theTembomvura People of Zambezi Valley, Northern Zimbabwe' (London,University of London, Ph.D thesis).Š (Feb. 1994) 'Diagrams in demographic data collection: Examples fromthe Tembomvura, Zimbabwe', RRA Notes, (xxii), 53-61.MARINDO-RANGANAI, R. C, AND ZABA, B., (1995) 'Animal Conservation andHuman Survival: A Case Study From Zambezi Valley, Zimbabwe'(Harare, University of Zimbabwe, CASS Occasional Paper).MEEKERS, D., (1991) 'The effect of imputation procedures on first birthintervals: Evidence from five African countries' Demography, XXVIII,(ii), 249-260.RIFKIN, S. B., (1994) 'Participation and health in research', in Korrie deKoning (ed.), Proceeding of the International Symposium on ParticipatoryResearch in Health Promotion (Liverpool), 21-25.RILEY, P. A.; SAMUELSON, J. L.; AND S. L. HUFFMAN (1993) 'The relationshipbetween age at menarche and fertility in undernourished adolescents',in Ronald Gray (ed.), Biomedical and Demographic Determinants ofReproduction (Oxford, Clarendon Press).UNITED NATIONS (1983) Manual X: Indirect Techniques for DemographicEstimation (New York, Department of Economic Affairs).