Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

11. Use of emic units for time-use recall

Tufts University School of Nutrition, Medford, Massachusetts, USA


It is effective to use emic units of time, that is, units constructed from culturally familiar reference events such as work tasks, to collect time-use recall information when studying cultures that do not customarily use timepieces. The term emic refers to the definition of concepts according to the explanatory system of the culture that is the object of study. It is used in contrast to the term etic, which refers to definition according to the terms of the culture conducting the research (cf. Messer; Scrimshaw, this volume, for other uses of these terms). The case for the use of emic units rests on five arguments:

  1. People who do not have modern timepieces mentally represent their activities in terms of tasks or activities having relatively standardized times of performance; moreover, these modes of representation enable them to provide meaningful time-use recall information to an interviewer. This premise challenges the common statement, "they do not think in terms of time."
  2. It is possible either to use existing emic units or to construct meaningful units from local time reference points and to use these measurement units in survey research.
  3. There are purposes for which information from time-use recall is inherently more useful than information obtained using observational approaches.
  4. An emic approach to the recall process may enhance rather than detract from the value of time-use information, and may in fact independently shed light on aspects of resource allocation and on time-sensitive processes within a given culture.
  5. The amount of error inherent in measurement of timed events in general, and in the use of recall methods in particular, is so great that the imprecision introduced by using emic units does not substantially compromise precision of the data.


All languages have words for the times of day, but not all cultures represent time in the same way. The angle of the sun as an indicator of time appears to be one common human approach that does not require a formal sundial. Brearley (1919) made the distinction between natural sundials, which were naturally occurring rocks or trees whose shadows were used for making and keeping appointments and timing events, and artificial sundials specially constructed or placed for this purpose. He chronicles the gradual improvements in the technology of sundials through the ages. These timepieces were universally used until clocks and watches became widely available about 300 years ago.

Even today, natural sundials such as familiar trees are commonly used by villagers in Swaziland. During work designing the Swaziland Nutrition Survey (Government of Swaziland, 1983) for instance, one mother commented that her baby had no appetite, because it had not eaten since "the shadow was at this point" (Aphane, 1984). Another mother was heard scolding her child for lateness, saying, "You always come home from school when the shadow is here." Members of the same household commonly used the same nearby tree for keeping track of time (Aphane, 1984).

Swazi villagers were found to keep track of time in at least two additional ways. They had specific terms for 12 times of day: dawn, early morning, mid-morning, late morning, midday, early afternoon, mid-afternoon, late afternoon, sunset, early evening (just after sunset when people could still walk about freely), night (lasting until about 3 a.m.), and very early morning (before dawn). The use of the number 12 in time computation is ancient and tends to be widespread (Brearley, 1919). Swazi villagers appear to be more attuned to the angle of the rays of the sun than people accustomed to using timepieces. Moreover, these times consistently refer to the same sun angles. This hypothesis could be tested empirically. To the extent that this holds true, major bias would be introduced into estimates based on these time units only by seasonal changes in the timing of dawn and sunset.

Finally, villagers measure time by events and tasks. A specific time during the morning is "when boys take the cows out." If the village is on a bus route, the first, second, third, etc., buses passing through could be used as time references.


Field researchers have long used local events calendars with individualized protocols for measuring the passage of years and months to pinpoint ages and dates. The same approach can be used to measure time within daily or weekly activity cycles. This approach was taken in my work on a health and nutrition baseline survey conducted among the Dinka in Sudan (Zeitlin, 1977). We used emic time units to estimate the study population's walking distance from resources. This survey estimated distances from the household to the water source, the market, and to health services. The emic units were developed through discussions with informants in Abyei (Sudan) and by timing the sun. Informants were the town's two elementary-school teachers, who had grown up in rural homes but who possessed watches and were accustomed to measuring time in hours and minutes. As in the construction of a local events calendar, the assistance of interested individuals who were thoroughly familiar with both traditional and modern approaches to time computation was essential to the derivation of a normative schedule of times when familiar events occurred. Focus groups and semi-structured interviews with members of rural households provided additional information to fill in gaps in this schedule and helped describe its variability.

At the time of the survey, dawn to sunrise was timed to be exactly one hour. Milking time was claimed by the informants and the chief to be about 9 a.m. We accepted this and used it to represent 9 a.m. Since the nature of the information required by the survey was approximate, we did not validate this representation, although we easily could have. While we did not conduct validity checks on responses to these questions, the apparent speed and ease with which the respondents replied suggested that they were being questioned in terms that were familiar to them. This familiarity also appeared to facilitate rapport between the Arab Sudanese and American interviewers and the Dinka households they visited.

In another example, Marlett (1988) studied the manner in which the work burden of Gambian mothers influenced their choice of treatment for their children's diarrhoea. From her own experience living for two years in Gambian village compounds, she was able to estimate the average number of hours required for the mothers' basic work activities. By asking the mothers in a cross-sectional survey how many times per week they performed these activities and how many helpers they had per activity, she was able to estimate a total time burden per mother. The time required per activity was multiplied by the number of times the activity was performed, under the simplifying assumption that the tasks were not overlapping. In this case, Marlett did not abstract emic concepts of time from the work tasks in which they were embedded. Questions measuring time use were then developed. The survey respondents easily answered these questions because they understood that the intent of the study was to investigate the mothers' time/work burden.


In spite of certain limitations (see Johnson, this volume), recall methods have several major advantages over observational techniques in special circumstances. First, they require much less labour and less time to collect. Second, recalled activities are not biased by the presence of an outside observer while the activity is performed. Third, they can be collected in situations where privacy is valued too highly to permit the presence of an observer. This last feature is important in societies where the most basic of observational procedures for studying traditional households, the spot observation, is not feasible (see Johnson's paper in this volume for a description of this procedure). For example, accurate spot observations would have been impossible to collect in the Mexican squatter community outside Cuernevaca that was the site of a recent micro-behavioural study of infant feeding (Zeitlin et al., in preparation). Much of family life occurs inside oneor two-room cement blockhouses located inside walled compounds. Before an observer (even an acquaintance from the community itself) can gain entry, the household will rearrange itself.

Finally, recall captures the passage of time as people remember it, or, in other words, framed in the mental constructs that people use in planning and evaluating the use of time as a resource. Recall methods, therefore, have the advantage of eliciting information that is potentially more intuitively understandable to the study population than time sampling or other observational data and hence is more readily usable as the basis for planning and action. Expressing time-use information in emic terms should enhance its relevance to the life-styles of the study community. This feature should be particularly effective in applied research conducted for the purpose of designing development programmes.


Time tends to be quantified when the conscious and public measurement of time, as a commodity, yields social or economic benefits. Disputes over time allocation may need to arise in order to stimulate the quantification of time. Control of the labour of others is a major arena for time disputes. In the simplest case, a mother in a traditional agricultural society who sends one child for water and the other to collect firewood knows very well how long it should take to complete each task. The culprit who spends too much time playing along the way is also in trouble. The wife who does not have food ready for her husband is in trouble, as is the husband if he does not arrive home to eat it. It does not take a clock to tell that "Johnny is too long at the fair."

Developing an inventory of the types of time-use concerns that arise in a culture, and the types of events that are closely time-monitored, should yield insight into disputed resources and the intra-household decision-making processes for the allocation of these resources. More basically, time-use concerns reveal time-dependent agricultural, foodprocessing, biological, manufacturing, and other cycles.


Two types of error inherent in the measurement of timed events cause so much imprecision in the measurement process that any additional error caused by the use of emic units is small by comparison. The first type is error caused by imprecise definition of the study variables. The second will be termed fragmentation error.

Imprecise Definition of Variables

Basic problems of validity occur when interviewers, observers, and respondents mean different things by the words they use to represent their study variables. Such error commonly occurs (although it is preventable) when survey instruments are translated in an ad hoc manner from one language to another, without careful back-translation. It is possible to minimize this type of error when dealing with simple concepts such as numbers of children born, years of education, presence of a radio, or type of roofing material.

The words used to label events, however, are intrinsically much less precise than the words used to label objects. The label "housework," for example, is much less specific than the label "car" or even the label `'vehicle." The vehicle is a visible object. Housework consists of a sequence of movements that not only vary from individual to individual, but also may be automatic (and thus unconscious) to the persons performing them. Both familiar and unfamiliar activities may not consciously be recognized and reported as events either by the persons who perform them or by persons sent to observe them. A further problem is that activities commonly are classified according to function. Moreover, the functions of familiar activities may not be consciously recognizable or may be defined in general rather than specific terms.

At the micro-behavioural level, psychologists insist on using videotapes of behavioural sequences (Dehavenon, 1978) in order to operationalize the definitions of behaviours and to train observers to recognize and record them reliably. Most time-use studies are concerned with macro-level events and therefore do not aspire to this degree of precision. The danger is that definitions of events may be based on very specific microlevel assumptions that vary subjectively from individual to individual. Without descending to the micro-level to specify these assumptions, time-use research may be limited in its validity.

As an example of such problems of definition, a validation study of maternal time-use recall in the Boston area in the United States found that mothers' recall of time spent bottle-feeding their babies was particularly unreliable when compared to time recorded by an observer in the home (Zeitlin et al., 1988). A major reason for this apparent lack of reliability was that the time-use protocol had not specified whether time spent preparing bottles should be included in the record of time spent in bottle-feeding. Some mothers classified it in their memories as part of the feeding while others did not.

Another example of such lack of validity comes from a recent observational study of infant feeding in Bangladesh (Zeitlin and Guldan, 1987). Based on dietary-recall data, it has long been accepted in the infant-feeding literature that Bangladeshis delay the introduction of weaning foods to their infants until the children are between 1 and 2 years of age (Israel and Tighe, 1983). In the cross-sectional baseline survey conducted prior to the observational study cited above, mothers reported that they introduced rice to their infants on average at the age of 12 months (Zeitlin et al., 1985). Event observations conducted by local observers over a period of six months, however, revealed that infants received supplementary foods on average every three hours by 4-6 months of age, increasing to every two hours by 12 months and 1.5 hours by about 21 months. Rice feedings started on average once per day by a year, and increased to every four hours by 18 months and every three hours by 24 months. Discrepancy between reported and observed events in this case probably arose more from a difference in definition than from errors of recall. The mothers apparently did not consciously define placing small amounts of food into the child's mouth as "feeding" the baby, in part because there were cultural sanctions against feeding infants before a certain age (Israel and Tighe, 1983).

The same study (Zeitlin et al., 1985) investigated the speed with which the mothers cleaned up their infants after they defecated. In the baseline survey for this research (Zeitlin et al., 1985), 50 per cent of mothers had asserted that they cleaned their babies "when they saw it." The observational study, currently being analysed, which timed the delay in clean-up found, however, that only 5 per cent of mothers started to clean their infants within the first three minutes after the defecation occurred, and that median time from defecation to start of clean-up was six minutes. In this case, the mothers' and the interviewers' definitions of rapid action appeared to differ.

An example of lack of awareness of performing an activity comes from a cross-cultural study of infant development comparing Yoruba mothers in Nigeria with mothers in Scotland (Agiobu-Kemmer, 1981). As compared with Scottish mothers, the Yoruba mothers were observed systematically to stimulate their infant's motor development by such activities as repeatedly placing play objects just beyond the reach of the infant who was learning to crawl, for example, and praising the child when (s)he reached the object. When asked whether they tried to help their children to develop motor skills, however, these mothers denied that they did so and asserted that the child's development was up to God.

Errors of Fragmentation

The term fragmentation (Schlossman, 1986) refers to the degree to which an activity is fragmented into subunits of time. Housework, for example, can be done in a single long stretch, in several shorter stretches punctuated by breaks or other activities, or intermittently interspersed with other activities such as cooking, phone calls, etc. Two observers watching the same woman engage in housework may very likely report different numbers of house-cleaning episodes, and their reports might differ from the woman's own recalled activities. The reporting of any event that is made up of repeated actions with pauses between them is by definition vulnerable to such fragmentation error. Housework is an example of an activity for which the number of discrete episodes might be less important to the researcher than the total time spent on the task.

Yet imprecise reporting of fragmentation introduces error into counts of events where the number of events is of greater intrinsic interest, as for example numbers of breastfeedings, or number of snacks or meals in nutrition studies. Older infants commonly detach and reattach to the breast while nursing. If every reattachment is counted as a separate breast-feeding, a ten-minute feeding for a distractable baby may be recorded as ten feedings. Konner and Worthman (1980), in studying the !Kung, for example, reported that the 17 babies (ages 3-34 months) they observed for six hours each breast-fed on average every 14.8 minutes for an average of 1.92 minutes per feed. In the Bangladesh study noted above in which 185 babies, ages 4-27 months, were observed for eight hours each (Guldan et al., 1989), the duration per feeding ranged from 1 to 14 minutes and averaged 8.3 minutes, while the average number of detachments and reattachments from the breast per feeding was 2.3. The average frequency of feeding decreased from one per 52 minutes at 4-6 months to one per 73 minutes at 13-15 months and one every 2.5 hours at 22-24 months. If fragmentation had been computed differently (i.e. if each recorded reattachment had been counted as a separate feeding), the computed duration per feed would have been 3.6 minutes and the frequency would have gone down to one per 22.6 minutes at the younger age period.

In the Bangladesh study only detachments and reattachments that the observer could distinguish were recorded. Because the Bangladeshi mothers wore saris, the baby's mouth could be obscured from view during the feeding, in contrast to the !Kung motherinfant pairs. In observational studies of this nature, it would be possible to specify that a certain number of minutes between reattachments would differentiate between separate feedings. More generally, observational researchers should specify the time interval between recurrences of an activity that differentiates between recorded events (just as epidemiologists specify the number of symptom-free days needed to differentiate between illness episodes).

The two problems with such specification are that: (1) any given fixed interval is unlikely to capture the boundaries between events that are perceived by the various actors - two different mothers could easily count the same breast-feeding behaviour differently; and (2) the definition of episodes or events that describes the behaviour of some groups of individuals breaks down for others: as an example, the !Kung babies who are constantly held next to their mothers' uncovered breasts may nurse off and on in a continuous fashion, whereas Bangladeshi babies, whose mothers' breasts are covered by saris and who are frequently held by other family members, may exhibit less fragmented nursing behaviour.

It is precisely distinctions of this nature that the concept of fragmentation is designed to capture and study. The problem arises that policy-relevant research on time use tends not to be interested in this level of distinction. Adequately resolving the problem of fragmentation requires more expensive research procedures than applied researchers need to use to obtain the bulk of the information that they need. Therefore, major fragmentation errors persistently flaw their results.


Time-use recall information in emic units potentially provides insight into the manner in which cultures conceptualize the allocation of time-dependent resources as well as information regarding time use. Given the degree of imprecision inherent in most time-use research, it is concluded that researchers should not hesitate to obtain time-use recall data based on timed tasks and other emic units. Such data should ideally complement more objectively collected information.


Agiobu-Kemmer, J.S. 1981. Infant Development in Two Cultures: Nigeria and Britain. Unpublished doctoral dissertation. University of St. Andrews.

Aphane, J. 1984. Personal communication.

Brearley, H.C. 1919. Time Telling through the Ages. Doubleday, Page & Co., New York.

Dehavenon, A.L. 1978. Superordinate Behavior in Urban Homes: A Video Analysis of Request-compliance and Food Control Behavior in Two Black and Two White Families Living in New York City. Unpublished doctoral dissertation. Columbia University, New York.

Government of Swaziland. 1983. Swaziland National Nutrition Status Survey. Full report AID/ IHF, Chevy Chase, Maryland.

Guldan, G.S., M.F. Zeitlin, and J. Zeitlin. 1989. Breastfeeding Practices and Behaviours. In: M. Zeitlin, C. Super, A. Beiser, G. Guldan, N. Ahmed, and J. Zeitlin, A Behavioural Study of Positive Deviance in Young Child Nutrition and Health in Rural Bangladesh. Report to the Asia and Near East Bureau, US Agency for International Development, and the US Office of International Health, October 1989.

Israel, R., and J.N. Tighe, eds. 1983. Maternal and Infant Nutrition Reviews, Bangladesh. An Intemational Nutrition Communication Service Publication. Education Development Center, Newton, Mass.

Konner, M., and C. Worthman. 1980. Nursing Frequency, Gonadal Function, and Birth Spacing, among !Kung Hunter Gatherers. Science, 107: 788-791.

Marlett, M.J. 1988. The Effect of Time and Resource Availability on Mothers' Use of Oral Rehydration Therapy in the Gambia, West Africa. Unpublished doctoral dissertation. Tufts University School of Nutrition, Medford, Mass.

Schlossman, N.P. 1986. Work Resumption, Breast-feeding, and Time Allocation of Mothers in Boston, Massachusetts: The First Half-year Postpartum. Unpublished doctoral dissertation. Tufts University School of Nutrition, Medford, Mass.

Zeitlin, M.F. 1977. Report of a Nutrition Survey Conducted in Abyei District, S. Kordofan Province, Sudan, by the Nutrition Division of the Ministry of Health, Government of the Sudan, and by the Harvard Institute for Intemational Development, Cambridge, Mass., in November and December 1977.

Zeitlin, M.F., and G. Guldan. 1987. Bangladesh Infant Feeding Observations. Preliminary report submitted to the Office of Intemational Health and the Asia and Near East Bureau of the United States Agency for International Development, Washington, D.C., 7 September 1987.

Zeitlin, M.F., G. Guldan, and N. Ahmad. 1985. Sanitary Conditions of Crawling Infants in Rural Bangladesh. Report to the USAID Asia and Near East Bureau, Washington, D.C., November 1985.

Zeitlin, M.F., F.C. Johnson, and K. Houser. In preparation. Active Maternal Feeding and Nutritional Status of 8-20 Month Old Low Income Mexican Children.

Zeitlin, M.F., N.P. Schlossman, and A.S. Beiser. In preparation. Validation of a Telephone Time Use Recall Instrument.

Contents - Previous - Next