A triads test is an elicitation procedure used in cognitive anthropology to investigate local perception of relationship and contrast among items under a domain in our case, among different varieties. Informants are presented with three sets of stimuli: either cards with names (i.e., of varieties) written on them or actual specimens. The task is for each informant to examine the sets, or triads, and to pair two on the basis of perceived similarity and isolate one on the basis of contrast.
1. Assemble a list of all varieties that the people plant or have planted, are consuming or have consumed. These will constitute the "stimuli" for the triads test. Names written on cards have the advantage of not influencing, through tangible physical attributes the way informants think about the varieties. On the other hand, actual specimens have the advantage of refreshing memories and also of being applicable even for non-literate informants.
2. Randomize the combinations such that the items have equal chances of being included in the triads and there is no particular order by which those related and those to be isolated are arranged in the triads. This can be done by computer, using the software "Anthropac" by Borgatti (1989).
3. It is important to give very clear instructions and to stress at the outset that there are no correct or wrong answers, that what is asked for is the informants' perspective on the matter.
4. Each informant should be given ample time to think and ideally should be isolated from others while doing the triads exercise.
The common way of analyzing the results of the triads test is to group together items that are frequently lumped together by informants and to figure out why the clustering is such. In the form used in the memory banking project, however, the investigator does not stop with the question "which one does not belong?" but proceeds to ask why do you think so?" The reasons given by informants are considered because these reveal the indigenous criteria for discriminating among varieties. By assessing the relative frequency in which each criterion is used (See Memory Banking: The Conservation of Cultural and Genetic Diversity...), it is possible to infer which ones are most relevant to local farmers.
3.2. Sorting and ranking
Sorting and ranking complement triads tests in eliciting local discrimination criteria. In addition, they give a much clearer and more direct indication of local systems of classification and prioritization.
1. In sorting, the informant is presented with an array of items: either names written on cards or actual specimens.
2. The informant is asked to sort the items (or names) into as many piles as he or she wants and to "label" each pile. This is the same as asking why he or she sorted in this manner.
3. The researcher can also request the informant to sort the items into smaller and smaller number of piles each time--in other words, to sequentially collapse piles. In this manner, one can see how informants join categories hierarchially and it is then possible to construct indigenous "taxonomic trees" pertaining to the domain under investigation.
4. Based on a narrower array of items, the informant is then asked to rank the items by whatever parameter he or she finds meaningful. The informant is later asked what parameter he or she used.
5. The researcher may subsequently focus on parameters he or she is interested in (e.g., productivity, palatability, storability, salability) and ask the informant to sort and rank based on that particular parameter.
Allowing the informants freedom to decide how many piles to make and what parameters to use as a basis for both sorting and ranking enables the researcher to get closer to the indigenous classification system and evaluation criteria (Fig. 6). This kind of information can give plant breeders and agronomists a better understanding of how farmers (and consumers) evaluate the varieties that they are cultivating or will consider cultivating.
Different informants sometimes give different responses to one question or may even give different answers to the same question on different occasions. This, according to our experience, is particularly true for varietal names and attributes. The apparent lack of consensus may reflect either a patterned distribution of indigenous knowledge -- something which in itself is interesting but beyond the scope of this protocol -- or a disagreement among informants as an artifact of the method of elicitation. If the latter is suspected to be the case, then verification studies have to be conducted to ascertain names and attributes of local varieties.