Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

Day-long Observations

Economists, anthropologists, and others have used day-long observations to study people's activities, in order to obtain accurate records of sequencing and duration of activities, particularly in populations whose sense of time does not lend itself to accurate recall of this kind. With this method, the observer may stay with the household during all waking hours. However, this much observation is exhausting for both the observer and the observed, and usually a compromise of 12 to 13 hours, covering most of the working day, is made. On the following day, household members relate how they spent the additional hours, prior and subsequent to the observer's presence, to complete the record. During the observation period the observer records all activities of the member(s) on whom the study is focused: e.g. the focal woman, if understanding her activities and their impact on food provisioning is the purpose of the observation, or children of particular ages, if the purpose of the study is to see how they acquire food through regular work, foraging, and stealth. Depending on the spatial and social layout of the residential units, proximity to the market-place, and so on, it is sometimes feasible to leave the focal household member performing a repetitive task and observe other members of the household who may be in the vicinity. Otherwise, one must rely on respondent reports of their activities in the interest of getting complete data on the focus of the study.

To record observations one of several procedures can be used. The most detailed and accurate device is a time record sheet, indicating starting and stopping times of all activities, interactions, and events. These records can be made using a timepiece, usually a watch. At the end of the day, for each focal individual, time spent on a standard list of activities can be summed for comparisons between households of different types, categorized by occupations, and ages of children. Alternatively, one can set up blocks of observation time during the day, record what activities take place during particular hours, and then compare activities across households and categories of individuals. For example, "morning" or "late-afternoon" activities can be compared for adult men and adult women.

Examples of studies using day-long observations are Peet's work on the economic value of Nepalese children (Nag, White, and Peet, 1978) and Evenson's research on children's household contributions in the Philippines (Evenson, Popkin, and Quizon, 1978). Peet collected work-input data for all children and adults in 45 to 50 households, once a month over a period of seven to ten months, depending on the household, as part of an economic survey of 674 households in a Nepalese village. During a month-and-a-half-long period when he could employ schoolchildren as field assistants, the data collections were extended to 106 households. Data were collected mainly by observation, in some cases supplemented by interview (Nag, White, and Peet, 1978). Results were reported in terms of 11 standard activity categories for each of eight age groups for each sex (table 6a). The time children spent on productive work and schooling were then compared, by age and sex, with results of a Javanese survey by Nag, White, and Peet based on data collected by the 24-hour-recall method (1978) (table 6b).

Table 6a. Average time input (in hours) per person per day in different work activities among females of various age groups in a Nepalese village

a. Includes collection of forest products, blacksmithing, and tailoring; excludes portions of handicrafts, agricultural products, and animal products which are sold.

Table 6b. Work inputs of children in various age-sex groups expressed as percentages of the average work inputs of males aged 15 years and over

Using a more systematic design, Evenson's survey team in the Philippines also studied the economic "values" of children, first with 24-hour-respondent recall, then with observations. During the latter phase of their study, which involved approximately 600 observation days, they covered 99 households at three times during the year to obtain an accurate record of the activities of all men, women, and children. To ensure that the presence of observers would not bias activities, they spent two days in each household for each day of data collection: one day to accustom members to the observer's presence and the second to record. Evenson had a large research team and, therefore, was able to cover a large sample. With a smaller team this type of method would have to be restricted to small samples.

A major limitation of the method is that it is exhausting for observer/recorders. In practice, one should schedule one day to rest and tally time observations for every two days observing per recorder. Despite this drawback, the results of such labour-time investments are rewarding. Quizon-King (1978), reporting on the data obtained by both recall and observation, shows that much more accurate records of labour-time allocation by household members can be collected by observation. Also, Quizon-King was able to stratify her sample according to father's occupation and numbers of children and compare labour-time allocation by different family members across households to discover significant differences in time budgets.

Nutritional Anthropology Studies Using Day-long Observations

While time-consuming and, therefore, necessarily limited in household coverage, the data produced in day-long observations are rich, diverse, and useful for many purposes. The method is the best for getting data on what children eat and how often and for what periods of time, while simultaneously collecting other food and activity data from households. For example, Messer (1981) was interested in documenting how different occupational characteristics of households affected women's and children's work and the focal woman's management of food for household members in a rapidly modernizing Mexican community. Households were, therefore, selected on the basis of the occupations of adults and the numbers and ages of children. While the original research plan had been to select households by occupation from a random survey sample, in fact, households were selected on the basis of occupation, child characteristics, and willingness to co-operate. Messer and a trained North American research assistant observed 20 households, four days each, two days in the wet (non-school) season, two days in the dry (school) season. Their observation periods were 12-13 hours long, from 7 a.m. to 7 p.m., and they concentrated on the focal woman in the household, but also recorded the activities of men and children around her (Appendix 1).

While the woman was performing repetitive tasks such as laundry, other members of the family were watched in more complete detail than was possible while focusing on the mother. Household arrangement of activities in a central room facing out onto a central courtyard made it possible to follow in most cases women's work, men's work, children's games, and, to a great extent, children's foraging, either in adjacent lots or in households of adjacent members of the "extended" family. Ordinarily the observer remained in the background.

Messer found that households that had carefully controlled their activities, such as child punishment, during participant observation, were capable of ignoring the observer, as she recorded unobtrusively throughout the day. Adults followed their natural routine, which occasionally included fits of anger and brisk thrashing of children. Though children initially explored the potential diversionary value of the observer, they were also able to ignore her presence, as she stoically ignored them. They continued chores or play; she continued her observation and recording. The only time the observer interrupted the normal routine of the household was at mealtimes when she weighed all food portions before consumption, and tallied all wastage per person afterwards. In addition, all food and other expenditures were tabulated for each observation day, either as they were purchased, or by brief questioning afterwards.

Day-long observations not only allow for the possibility of collecting complete information on food intake, they also may reveal bottlenecks in women's labour time, that is, when competing tasks prevent her from doing some of them well, if at all. Using this method, aspects of sanitation, illness, and healing that affect eating and the nutritional status of household members can be recorded. These variables would be missed by other methods. Day-long observations can also avoid some of the criticisms of less extensive participant observation cited by Johnson (1975), among others, that most anthropological observations are "impressionistic" - i.e. they do not provide a representative sample of the population and they may not get enough observations of activities at different times of the day or seasons of the year to fully represent activity patterns. By carefully selecting households on the basis of some systematic principles and setting up visits to cover the cycle of natural and cultural seasons, day-long observations can eliminate most impressionistic bias. The major source of bias remaining is that, since one observer cannot record everything, she obviously makes a number of choices about what to record. Day-long observations also avoid error created by informant impressions of the amount of time spent working by adults or playing by children, biases that may be introduced by the less time-consuming method of collecting 24-hour-recall data from respondents.

Recall Methods

Whereas observation methods must be used in cultures where people are not accustomed to calculating daily activities on an hourly basis, 24-hour recall may be a more efficient method to collect data where members are used to thinking within an hourly time-frame. With the serial recall method, a trained interviewer visits each household selected at random on a regular basis (every sixth day, every fifth day) so as to avoid always arriving on the same day of a cultural time-cycle. She asks a key household member or a group of household members to relate, sequentially. the time spent on all activities by each household individual for the previous twentyfour-hour period. Some interviewers prefer group interviews, since household members check and correct one another's activity claims and, therefore, provide more complete and reliable information. Ideally, the respondent can recall with some accuracy the starting and stopping times of the activities by citing coincidental events, such as the beginning and end of the school day, radio programmes, fixed mealtimes, and calls to prayer. The interviewer can then calculate both the hour and duration of activities of interest.

Alternatively, the interviewer can ask for "activity-specific recall," in which case the respondent is asked to produce a time-activity list indicating how much time is spent on each of a number of activities, classified by the respondent or suggested by the interviewer in advance. This method forces the respondent to calculate time budgets for herself or himself, and then for other members of the family. Comparing individual time budgets allows for some crosschecking for accuracy, based on the interviewer's composite picture of household functioning. In serial recall, as in day-long observa tion, recording may be by the diary method, in which the interviewer records times of starting and stopping of all activities, supplied by the interviewee, or by marking a time-sheet, already divided into blocks of time to indicate hours of the day. Cain (1977), in his study of fertility patterns in Bangladesh, used the latter method. The sample consisted of 120 households selected at random by cluster sampling, visited every 15 days "after work hours." Data sheets were made up in advance, by household, supplying an individual time column for all household members. He estimated one hour to complete each household budget form. The interviewer simply drew horizontal lines to mark starting and stopping of each activity, all household information having been pre-recorded (fig. 2). In this manner, trained research assistants could cover two households in the evening, after 5 p.m. Households were asked to list all activities, by type and duration, for the 24-hour period prior to the interview (5 p.m. to 5 p.m.). The sheets were checked for inconsistencies by comparing household members' columns, then brought to a centre in Dacca for coding. Activity times were calculated, summed, and coded for each household member and recorded on household worksheets. The data were then transferred to individual worksheets on which additional rounds of activity data were entered. The processing of each individual time-budget form took 3.5 person-hours to complete: one hour to collect the data, half an hour to field-edit the budget forms, one hour to code times and activities, and one hour to transcribe data to worksheets and to subtotal time segments.

Fig. 2. Diary method for serial recall (after Cain, 1977).

Ong (1982) also used a random sampling design and serial recall in her time-allocation survey of 40 households in a Malaysian village. She was interested in discovering varying patterns of work and leisure allocation by sex and groups in an economy dependent on income from both farms and wage work, with young women forming an important group of wageearners as factory-workers. She found that a trained research assistant could cover approximately four households per day. Households were visited once every seven to eight days to collect time-allocation data.

Respondents in Ong's study (1982) were very time-conscious, and she had no trouble collecting 24-hour-recall data in this Moslem society. She also tried to collect income and expenditure data, although she planned an economic-nutritional component to follow the timeallocation data collection, given the limits of respondent patience for interviews. Data were tallied and coded after each recall, and computer analysis was used to find total averages for each individual for each activity over a six-month period. Time-allocation patterns were later interpreted in relation to class and fertility patterns of households.

As a third example, Chapin (1974) used serial recall in a survey of recreation activities of households in the Washington, D.C. area. This method was chosen because the researchers wanted to cover a large number of households with limited interview-time input. Their purpose was to record a detailed plan of people's preferred ways for allocating time to leisure activities. Households were selected by a random design principle within the study region. They found the poorer segments of the population insufficiently skilled to keep diaries, so they used a detailed set of questions, and the respondents, primarily women, were able to produce elaborate timesheets within an hour's questioning. Interviewers were able to get a precise plan of the typical activities of each adult and child for weekdays and weekends and, with qualitative questioning, were also able to correct for irregularities in any particular day's events.

Recall, if it can be certified by observation checks not to overestimate or underestimate time allocations, is an excellent way to record what happens in the course of a day in households of different socio-economic status within a community. One can stratify household samples for the purpose of seeing how households and individuals differ, by class, in terms of how they spend their time, and the relationship of these different activity patterns to other aspects of behaviour, e.g. fertility and nutrition. Recall also makes efficient use of the interviewer's time. A trained interviewer can cover approximately four households per day (Ong, 1982) or two households per night (Cain, 1977), with minimal intrusion. A respondent may also verbally describe activities she did not have time to write down in a diary entry without interrupting her own activities.

Disadvantages are that adults may tend to underestimate, for example, the productive work of children (Quizon-King, 1978). Respondents may also fail to relate time spent doing several things at once; for example, mothers, without some prompting by an interviewer familiar with the culture, may recall very inaccurately time devoted to child care that was supplementary to some other primary activity, such as cooking. Having more than one person present at the interview, particularly older children or other adults also responsible for child care, may also help to overcome some of the limitations in recall that result from honest error or the desire to present one's time allocations in the best possible light. The latter may lead to overreporting of productive activities and underreporting of non-productive ones.

An additional disadvantage of the recall method is that it is often not feasible to collect other kinds of information concurrently. Eliciting a full record of all family members' activities and complete production/consumption data, plus full dietary recall statements, is painstaking, time-consuming, and beyond the patience of most respondents. Such information on food expenditures in money and time and food consumption data, however, may be collected separately and added to the activity data. For example, food interviews on the structure and content of meals and snacks, and the social division of labour in food acquisition and preparation, in combination with activity data, based on recall, could help one analyse significant differences in eating patterns. However, the information would have to be collected in two separate stages, and neither stage would be likely to give detailed information on snacking behaviour, which is most accurately collected by observation.

The Diary Method

As a final alternative for collecting 24-hour data, literate, time-conscious respondents can be asked to keep diaries, recording starting and stopping times for all activities. The method has been used by economists and urban planners to understand (a) labour and leisure time-use by European and American housewives, (b) urban activity patterns, including time spent commuting to work and time related to recreation activities, and (c) marketing (Szalai, 1972). Data from self-reporting have also been used by anthropologists to understand differing patterns of household-labour allocation in various food-producing societies. The method works best where respondents clearly understand the task involved and do not find too burdensome either the momentary interruption to record or end-of-the-day recall recording. If these tasks are too disruptive they will interfere with the respondents' daily performance and also jeopardize the study.

Diary-keeping should be restricted to studies where detailed data on one respondent are required, as where a mother's activities are followed, and other household members 'work, educational, and recreational schedules are incidental. One could not, for example, gather detailed data on children's activities by this method, although such information could be gleaned, on occasion, by the recall method. For example, a mother might remember what children were playing, where they ran for snacks, and what they ate, but not have time to write all this down during the day. Alternatively, diary-keeping can be combined with a daily recall strategy in which a respondent acts as the interviewer, asking herself to remember the sequence of activities at the end of a day. She would then present this information to the researcher in lieu of a separate interview.

To use the diary method. the researcher should first gain experience in the culture through observation and interviews aimed at defining the kind and duration of the activities of research interest. This preliminary work forms the basis for constructing an appropriate list of activities for the analysis and for sensitizing diary-keepers to relevant categories for their records. This preliminary list can be submitted to respondents for their comments and suggestions on how to improve and systematize the records and analysis.

Most sample notebooks use time sheets, ruled or sectioned off with the hours of the day, so that one can simply draw a line at the starting and stopping times of activities, and fill in the appropriate activity label. Alternatively, one can record times of stopping and starting activities, which can then easily be summed.

The main advantage of diary recording is that it potentially allows recording of data on a greater number of days for each household, since the initial hour or more of recording the primary data by the observer is avoided. However, this data must then be processed, so time constraints for the researchers should not be underestimated.

Experimental Designs

In addition to the time-data collection methods just described, psychologists have proposed several '`time-frame" methods for observing and measuring infants' and children's behaviours. These methods can be used to show how nutrient intake affects the cognitive, emotional, and physical development of infants and children. Such methods can also reveal if there is a discernible relationship between nutrient intake and behaviour, such as a child's responsiveness to persons and objects in the environment.

Instructive examples of the use of time-frame methodology are a series of studies by Chavez and Martinez (1979) on the effects of mother/child nutritional supplementation on child behavioural development. The investigators use both qualitative and quantitative reporting to classify actions within the mother-child-environment system. Observations were recorded in terms of 37 descriptive states of the child in relation to its environment, of the mother's actions toward the child, of the child's actions toward the mother, and of the father's, siblings', and other adults' relationships toward the focal child.

Classifications and records for each focal child were made after observation for 72 continuous hours in the household. These observations also enabled the observer to record all feedings. The first day was spent observing lactation and accustoming the child and other household members to the observer's presence. On the following days, the observer sat herself in a corner for periods of 11/2 hours in the morning and 11/2 hours in the afternoon, and pretended to read, while she observed the mother-child interactions. Every 30 seconds she would lift her head and note if the mother was talking to, smiling at, feeding, or otherwise interacting with the child. She then recorded behaviour frequencies as a proportion of total behaviours observed. Following the child's relationship with other members of the family was more difficult, since their presence in the household was less regular than that of the mother. To record these interactions, the observer recorded activities (interactions) at 10-minute intervals each hour, over a day of 12 hours. Thus, she captured activities at 8.30-8.40, 9.30-9.40, and so on.

To study play activities experimentally, the investigators constructed three-by-three-metre squares, with 30 cm fencing, and placed the child in the centre with the mother at one end and a pile of toys at the other. They then timed numbers of interactions of children with mother and toys to measure activity levels, maternal dependency, and also, to a point, aggression.

Results were analysed using Chi-square statistics, comparing two groups: well nourished and poorly nourished. Infants were followed at ages of 2, 8, 16, 24, 36, 52, 72, and 96 weeks. Graphing per cent times (y-axis) against weeks of age (x-axis) they compared time in the cradle; time sleeping; time playing (by proximity of the mother); time carried (by both arms or on the shoulder); time the mother spent smiling, talking, or playing with the child; time crying; time babbling or talking; time child spent smiling at the mother; rapidness of response (by infant and mother) to infant's dirtying; and accident prevention. The categories were primarily of behavioural interest, although data on sanitary conditions are also of interest for interpretations of health. Interactions of father and other family members were also graphed, and showed significantly greater interaction between a well-nourished child and its environment and significantly greater stimulus levels provided to the well-nourished compared to the malnourished child. Generalactivity-level behaviours were also calculated; in general, the well-nourished child was more active and spent more time exploring his or her environment. Finally, in the experimental set-up, observers graphed the positions of the child on a quadrant, showing initial position of infants and routes to mother and/or toys, and calculated "intensity" of a standard set of behaviours: cries, is upset, looks at toys, plays with toys, and talks. Comparing individual cases, these results again show the well-nourished child to be more active. One might question, however, whether participation in the supplementation programme, apart from the nutritional effect, did not somehow affect the activity levels of the better-nourished group, both children and environmental adults. (All data are taken from Chavez and Martinez, 1979).

This series of studies used ingenious methods of timed observation rather than time allocation to provide answers to questions about nutrient intake, child behavioural development, and household functions. The method is exhausting and time-consuming for the observer, and only a trained observer who has the confidence of the households observed would be able to carry out such a study. Interpretations must necessarily be based on observations of very small comparative groups (groups of 20 in the Chavez and Martinez study). Nevertheless, this combined sequencing of continuous hourly sweeps of activity-observation over the course of a day and intensive observation of mother and child pairs during morning and afternoon periods provides a method to measure activity and stimulus levels, which can then be correlated with levels of nutrient intake. If the investigators have the kinds of facilities that will allow them to follow experimental and control groups or a stratified population sample over a period of two years. and a well-trained observer staff, then such studies, which provide critical information on child behavioural development, are feasible.

For investigators without such resources, calculation of child activity levels, quality of activity, interaction levels with the household environment and qualities of interactions can best be extracted from day-long observation data, recorded either continuously or at short intervals. For studies focusing on children's behavioural developments, intensive observations with spots or spots with narratives have been employed by psychologists (e.g. Bloch, 1979).

Summary

Some of the preliminary questions that must be answered in designing either a time-allocation study or a time-frame study are: By what method should data be recorded? Should single individuals be followed or should social groups be scanned? Is it wiser to focus on certain target activities or social interactions or to rely on a random sample of behaviour to provide an accurate picture of all activities, including those one may be most interested in? Many of the studies described here have tried more than one method before settling on the one that will yield the best results, given their research interests and constraints (Quizon-King, 1978).

Sampling also presents difficulties. Among them is the claim by some researchers that the results of following single individuals, as opposed to periodic observations of their activities in the context of censusing the activity of a whole group, yield equivalent results; others dispute this. Altmann (1974) provides a thorough discussion of these issues in the context of primate behavioural studies.

In almost all of the time-allocation studies described here, with the exception of the studies by Erasmus (1955) and Johnson (1975), research assistants were relied upon, and it is appropriate to add a note on that subject here. To cover a large number of households by either observation or interview, the principal investigator will need research assistants. Appropriate personnel from the household's point of view are those whom households do not consider to be interfering and to whom they will allow ready access. Research assistants must also be able to provide reliable records.

Social scientists working in Islamic societies, particularly in South-East Asia, have found that the best research assistants are literate, with the equivalent of at least a secondary-school level of education, and come from the observed household's own section of the town or village. Closely related persons are in the best position to gain access to the detailed, intimate household time and monetary budget information, and are also in the best position to catch errors in reporting.

In contrast, studies in Mexico have used well-educated personnel from outside the community being observed. Excellent technical assistants have been of Mexican origin, with a secondary-school education plus additional training, or have been North Americans. Households were reluctant to allow local persons access to the intimate details of family life, eating patterns, and expenditures.

There is, then, no universal rule as to who will be accepted to record the highly confidential data on household functioning. In each case, the investigator will have to evaluate assistants on the basis of her or his own professional perception and knowledge of the culture. The acceptability of local versus non-local assistants and the acceptability of male or female personnel for certain kinds of research will have to be assessed.

Contents - Previous - Next