chapter 4

An extensive literature review of quantitative and process-oriented studies of TEK was carried out in order to assess the current state of the field and the relevance of this prior work for development of the VITEK. Special attention was given to surveying the principal field and analytical methods used for its study as well as the major findings and conclusions regarding general trends of TEK change or continuity and the factors driving them. A primary goal was to inventory and evaluate the range of methodological options that have been tried and tested previously, their strengths and limitations, and to identify those which are best suited for a quantitatively measured indicator that strives to be both locally appropriate and globally applicable. A careful review of the scope of available data and information tells us what variables should potentially be taken into account and also serves to evaluate what gaps and needs should be addressed by the VITEK. We also wanted to explore the feasibility of developing the indicator from these secondary sources.

4.1. Ethnographic and Theoretical Research Methods

Prior to the mid-1980's, virtually all studies of TEK phenomena were focused on either ethnographic descriptive reports or on ethnobiological theory-building. This phase is characterized by long-term fieldwork, detailed and comprehensive documentation of local environmental knowledge and practice, intimate contact and participant observation of the study community's daily life, in-depth and open-ended interviewing of a few key informants, and mostly qualitative forms of data analysis. The descriptive studies depict particular groups' classification, use and manipulation of biological species and groupings, and is best exemplified by ethnobotanical surveys or monographs of the agricultural cycle. These studies have highlighted the empirical perspicacity, logical organization and ecological rationality of knowledge and use patterns as well as the holistic integration of these with other aspects of the cultural lifestyle. The theoretical investigations are oriented toward the comparative analysis of the linguistic, cognitive and biological properties of ethnobiological classification systems among diverse (mostly rural, nonwestern and nonliterate) groups among whom descriptive studies have been carried out with the purpose of identifying universal or evolutionary principles of the human apprehension of the biological world. This body of work has revealed that folk peoples who live in close contact with the land typically recognize and classify a large portion of local biodiversity in similar ways (to each other and to scientists), a shared trait that points to a panhuman cognitive ethnobiological faculty. At the same time, the culturally programmed development of this capacity was found to be affected by utilitarian factors and economic orientation. Normal individuals of urban industrialized societies, for example, display more attenuated and structurally simplified inventories, which may indicate that the ability to develop this innate capacity is to some extent determined by environment and lifestyle.

4.2. Quantitative Approaches

In recent years, there has been a growing trend to use quantitative research approaches, especially in ethnobotanical investigations. The quantitative revolution in ethnobotany and other TEK-related subjects has sought to transform this field into a more rigorous science based on the principles of precision, replication, comparability, prediction, and deduction (Johnson 1978; Phillips and Gentry 1993a). The case studies are distinguished by innovative research methods and their direct applications for conservation and development issues. The research designs rely heavily on formal and replicable data collection procedures, such as structured interviews administered to a population sample, and quantitative analysis and hypothesis testing of the results. Important statistical measures produced through these studies include: a) the percentages of local biodiversity that people name and use b) the cultural importance values and use preferences of different biological taxa (e.g. species, families, biotopes), c) the monetary valuation of wild resources and habitats for local communities, d) the actual use frequencies or harvest intensities of certain resources, and e) the perceptual, morphological, biological, and ecological characteristics of useful plants. Such information is useful for understanding the extent to which people are dependent upon locally available natural resources, establishing conservation targets, assessing the ecological impact of use levels on the survival of natural populations, and promoting sustainable development options. Furthermore, the identification of species with high importance value provide useful ecological or socioeconomic indicators for monitoring the effects of conservation and development programs.

Another application of ethnobotanical quantification is found in the field of medical treatment. Measures of informant-based consensus and fidelity of use have been employed to rate the local use value of medicinal plant species as a way of identifying the most empirically effective remedies. These candidate species can then be tested for bioactivity and therapeutic effect. This kind of study has two potential applications: a) to identify effective and inexpensive herbal remedies for treating local health problems and b) to discover and develop new pharmaceuticals that may be industrially manufactured.

In other quantitative studies, variations in knowledge and use habits according to ethnic groups, age, gender, education, wealth, occupation and other social variables have been measured. In some cases, the social distribution of knowledge has been examined to infer processes of change of the knowledge system by showing how different groups and subgroups are affected by surrounding changes in the socioeconomic or biophysical environment. Hypotheses regarding knowledge change and its causes have been tested by correlating knowledge differentials (within and between groups) with dynamic social and environmental variables, many of which are themselves indicators of change. Such investigations are especially relevant for the present objective of indicator development because they afford a more disaggregated, microcontextual and dynamic view of the corpus of knowledge.

4.3. Intracultural Variation

The attention given to TEK variation across space and time draws theoretical inspiration from recent views of culture as a partially shared and socially patterned information pool and as a dynamic network of distinctively positioned perceptions and actions that are constantly being produced and reproduced through ongoing experiences, movements and interactions with other such networks. The variable, dynamic perspective of TEK effectively supercedes former depictions of it as a collectively homogeneous, static and self-contained phenomenon. The justification for a more fine-grained and partitioned concept of the knowledge system follows from the recognition that work roles, social networks and interaction contexts are sufficiently differentiated so as to produce distinct subcultures and subgroups. According to this viewpoint, the social distribution of knowledge is closely bound to social organization which is patterned by age, gender, kinship, marriage, residence, education, occupation, socioeconomic class, ethnic affiliation, trade, religion, and other statuses. These variables constrain and determine people's customary activities, spatial mobility, and access and control of resources, and in that sense directly affect their contact and familiarity with environmental components. Moreover, social relationships constitute pathways for the exchange and flow of information between individuals and groups. One of the key findings in this regard has been that the pattern of social distribution of knowledge within a community closely reflects the pattern of knowledge transmission. Thus synchronic variation can be used to infer ongoing processes of diachronic change, including ontogenetic as well as phylogenetic development. A focus on dynamic properties therefore requires a shift of the analytical locus from the group to the individual. The determination of individual knowledge variation depends essentially on a two-step procedure: (1) the intersubjective sampling of a person's knowledge of a relevant domain of TEK, usually through a structured questionnaire or interview technique administered on an individual basis to a sample of persons, and (2) the measurement of interinformant patterns of similarity and difference, often through some statistical operation (e.g. cultural consensus analysis, principal components analysis).

4.3.1. Knowledge Sampling

The goal of observing and measuring interinformant cognitive variation imposes the constraint of data comparability on the method chosen which is usually achieved through some process of quantitative data collection and manipulation. In order to comply with the criteria of comparability and replicability, the vast majority of quantitative-based studies of TEK variation and change rely on a controlled data collection format, especially structured or semi-structured interviews. The structured interview type entails the same set of questions or response stimuli being administered to each and every informant in the sample and may take the form of a written questionnaire, for literate populations, or an oral interview schedule, for preliterate populations. This method is transparently quantitative in the sense that the verbatim responses can be submitted directly to statistical analysis without further coding or data manipulation. The semi-structured interview follows an interview guide, that is, an established list of questions and topics that need to be covered, but actual administration of the interview is more flexible and open-ended, allowing the respondent more freedom to answer in their own words and permitting the interviewer to follow up interesting lines of inquiry with spontaneous questions. In other words, it may begin with highly structured queries but is actually intended to elicit more free-ranging, informant-initiated responses. Such responses must then be categorized through an interpretive process controlled by the researcher (i.e. according to some predetermined classification scheme) in order to produce a comparable data set. In consideration of the pros and cons of the two interview formats, we note that the former has the advantage of simple and straightforward coding, which confers a higher level of comparability and thus superior analytical capability, but has the disadvantages of limiting the amount and quality of the information provided, imposing an artificial communicative context onto the data interaction, and restricting the analysis to the categories deemed relevant by the researcher. The latter has the advantages of eliciting more culturally appropriate (i.e. emically valid) information as well as supplying leads for new questions that otherwise might be completely unanticipated by the researcher, but has the disadvantage of requiring a more complicated, time-consuming, and opaque data coding process. While the traditional anthropological field methods of participant observation and informal or unstructured interviewing are not conducive to quantitative treatment, they are still often used alongside more formal methods to get a better sense of locally appropriate topics and questions, build rapport and reduce intrusiveness, enhance communication between researcher and research participants, match people's statements about their behavior with observed behavior, explore rare or unmeasureable knowledge domains, and gain a more holistic understanding of people's daily life and its influence on their knowledge pattern.

4.3.1.1. Structured Interview Formats

The most common form of structured interviewing is the questionnaire. The questionnaire has been a primary methodological tool in social science research for many years. In anthropology, this method is associated mostly with ethnoscientists, who developed formal interviewing procedures in order to eliminate observer bias and extraneous contextual noise, and therefore elicit only culturally acceptable emic-type information. The formal or controlled elicitation procedure, modeled after structural linguistics methodology and adhering to a tightly controlled query-response framework for collecting data, rests on the use of standardized question frames ideally posed in the native vernacular. An example of a class inclusion frame would be: "What is the name of a kind of _____?" An example of a use frame would be: "How do they serve us, _____?". The question frame is the basic technique used by researchers of folk biology to elicit plant and animal taxonomies and activity contexts. The kinds of questions that appear most frequently in the controlled elicitation procedure may be classified as dichotomous (yes/no, true/false), multiple-choice, or fill-in-the-blank.

Besides questionnaires, local ecological knowledge can also be revealed through "analytical tools" involving controlled experimental cognitive exercises or games. The basic objective of these tools is to probe the underlying, sometimes unconscious, cognitive organization and content of selected cultural domains. In some cases, the informant may be asked to explain the reasons for his or her choices. Using these mental exercises, the researcher is able to infer the range and extent of an informant's knowledge as well as the cognitive criteria and biases that shape his or her view of the domain. The most common techniques of this kind include: free listing, paired comparison, triad test, pile sort, rank order test, and psychological projection test.

Free Listing

Free listing is an open-ended interview technique in which the interviewer asks the respondent to name all of the things that he or she can think of on the spot corresponding to a specified cultural domain. The relative position and frequency of mention of a term is thought to reflect its saliency in the culture being studied. In TEK studies, this technique has been most commonly used as an adjunct to the question frame for eliciting the inventory of terms comprising native taxonomies of plants and animals. In that case, the free list query is administered successively and exhaustively at different levels of inclusiveness (e.g. animals, fish, catfish, etc.) to chart the hierarchical structure and composition of the taxonomic system. It can also be used to determine the taxa employed for different use categories (e.g. medicinal plants, diarrhea remedies) or the use values (e.g. food, medicine, construction, etc.) per species. One criticism of free listing is that people tend to mention abundant, conspicuous and easily accessible taxa as well as taxa providing goods during the season of investigation, and thus these factors may bias the overall results.

Paired Comparison

Paired comparison (or pairwise comparison) involves presenting the respondent with two objects or categories and asking him or her to judge which of each pair is preferred or better according to a given criterion (e.g. firewood) or has a greater amount of some property (e.g. soil fertility). The procedure is carried out successively for all pairs in a given set and the results are tallied to determine the preferred option or to produce a ranking of the array. A paired choice matrix can be constructed to help with this type of analysis. This technique is especially useful for ranking the members of a class that is too large to order easily. In TEK studies, it has been employed to rate the relative importance value of plant species, preferences in relation to use categories, and perceptions of activities threatening wild edible plants. An interesting variation on the paired comparison method has been developed and used to investigate ecological relationships between plant and animal species.

Triads Test

The triads test entails presenting informants with three things and asking them to choose "which one doesn't fit" or "which two go together." The informant may be asked to explain the reasons for his or her choices to explore what criteria for grouping they consider important or the exercise may be directed in relation to some predefined criteria. This is done for all triplet combinations from a set of objects or concepts and the results are analyzed to map degrees of similarity/difference among them. The method is widely used in cognitive research as it sheds light on the internal organization of a cultural domain and its criterial attributes. TEK researchers have made use of triads tests to investigate covert groupings of biotaxa, the utilitarian and agronomic significance of weeds, pests and crop varieties, agronomic and gastronomic criteria for the recognition and classification of plants and arthropods, cognitive prototypes in medicinal plant selection, and residential decision-making. The administration of numerous triads tests can be a lengthy and tedious process for researcher and respondent alike, and therefore for complicated issues or very large cultural domains redundancy-cutting procedures are recommended.

Pile Sort

The pile sort, like the triad test, is geared toward organizing objects or categories into groups and thus revealing relationships of similarity/contrast among them. In this case, the informants are handed a set of objects - which may be cards with names written on them, drawings, photos or even biological specimens - and asked to sort them into groups according to whatever criterion they choose. The exercise is repeated successively on each pile of the previous sort until they cannot subdivide the piles any further (or the method can be applied from lower- to higher-inclusive grouping). At each sorting level, the informant may be asked to describe a word or phrase that explains the reason for making each pile. The successive pile sort is considered to be especially useful for making interinformant comparisons. In ethnobiological studies, pile sorts have been used to explore the taxonomic relationships among organisms, covert (i.e. unnamed) categories, and the intercultural correspondence of taxonomic knowledge. The method has also been applied in studies of the agroecological knowledge of crops, soils and fertilizer and the folkecology of fish.

Rank Order Test

The rank order test involves presenting informants with a list of taxa or categories and asking them to rank them in order according to relative importance or some other scale. Sometimes this procedure is performed using a Likert scale format in which respondents are asked to specify their level of agreement to a statement about the item in question. By administering this technique to a sample of respondents, the average rankings can be calculated per social group or subgroup and then compared between them. Ranking scales produce ordinal data directly, are fast and easy to administer, and are considered easy to understand by informants. The principal use of this method in TEK studies is to rate the importance, use or preference value of biotaxa based on local people's own assessments.

Psychological Projection Tests

Psychological projection tests refer to devices used to infer personality and character orientations, thoughts and values, that may be unconscious and thus not susceptible to direct verbalization by the informant. The best known projective devices are the Rorschach Inkblot Test and Thematic Apperception Test (TAT) and they have been used a good deal in culture-and-personality research. In the TEK field, experiments have been conducted to induce one's informants to project (i.e. reveal by indirect means) their ethnoenvironmental understandings through natural element (e.g. tree) drawings, biocultural mapmaking, storytelling, and the interpretation of pictorial representations of environmental scenes. The drawing of so-called "mental maps" by different individuals or focus groups is a common technique used in participatory mapping or ethnocartographic projects for it reveals those landscape features which they consider to be most important. Nazarea employs a modified TAT, involving photographs depicting local landscapes and human-environment interactions, to investigate sustainability and quality of life judgments by different age, gender, and ethnic groups among farmers-fisherfolk in the Phillippines. This class of techniques can provide insight into perceptual biases as well as values and factual knowledge. However, the coding of responses can be somewhat difficult and subjective, and therefore it may not be easily adapted to quantitative and comparative analysis.

4.3.1.2. Sample Selection

Another key constraint that must be taken into account to capture intracultural knowledge variation is data representativity. Interviewing the entire population of a community or larger unit is usually impractical, inefficient, and in some circumstances (e.g. when the sociodemographic composition is highly disproportionate or momentous events that change people's opinions occur before the study is completed) can be less reliable than a careful representative sample. The random (or probability) sample, whether simple (i.e. every person has an equal chance of being selected) or systematic (e.g. every nth person or household from a census list is chosen), is generally considered to be the most reliable type because randomization reduces biases and allows for the extension of results to the entire sampling population. However, the proper use of this technique requires a prior idea of the range of variation of the test variables in order to select an adequate sample size. Unfortunately, this critical information is usually unavailable, unless a prior study had already been conducted, and a pilot study of knowledge variation prior to the full study may not be efficiently possible. In consequence, one can find sampling intensities ranging from 1-100% across the studies reviewed here. Another problem with the probability sample is that it is often the case that some people chosen for the study refuse to participate, are absent during the research period, or do not answer all questions, which may invalidate the assumptions of probabilistic-statistical inference. Therefore in human subjects research, it is often impossible to achieve the ideal sample and flexible adjustments to the sampling design must be made.

The stratified sample attempts to capture all the relevant subgroups of the population and is preferred whenever it is likely that they may be underrepresented by random sampling. This requires a good grasp of all of the independent variables on which to stratify and therefore depends on a fairly detailed population census. Given the typically small size and dispersed distribution of most rural folk populations among whom studies have been carried out, the stratified sample seems to be the preferred choice in studies of TEK variation and change. The relevant social variables used to design such samples have been: age, sex, schooling, occupation, residential history, languages spoken, local group, and kinship group. In most cases, the range of variables selected for stratification is not really justified by reference to a prior socioeconomic survey but instead is simply assumed or dictated by the research problem.

Non-probability sampling may be more effective for some research situations, for example when time and resources are limited or one needs to study a cultural domain characterized by specialist knowledge or experts who are much more knowledgeable than the average person. One of these is the purposive sampling technique, also called judgment sampling, which involves the deliberate selection of informants due to their recognized expertise or other qualifications that they possess. The experts chosen should be reliable (i.e. consistent across the community), competent (i.e. qualified through reputation or demonstration), and willing and able to communicate their wisdom. The danger of this method is that the researcher exercises uninformed judgment about which informants are duly qualified at the risk of data quality. In order to ensure that the purposive sample is indeed representative and accurate, it is advisable that the selection process count on prior familiarity with the population and culture, consultation with the community leaders or contact persons, or even a survey of community members who are asked individually to name the most appropriate informants and then those people who are most frequently mentioned are chosen. A similar approach is used in snowball sampling, which involves asking an informant to suggest another informant and so on. Purposive sampling can be used to generate key informants, focus groups, surveys, questionnaires and all of the data instruments discussed above. In TEK studies it has been employed in studies of specific skills, knowledge or practices, comparisons between practices, and case studies of natural resource perception and management. The nonrandom sample is frequently used in ethnomedical investigations which focus on specialist or expert healer knowledge. In distributional studies, the sampling of expert knowledge during a preliminary phase serves to establish a baseline of information from which a sample of questions and the answer key can be drawn for interviewing the general population during a later phase.

Opportunistic sampling is the least reliable form of sampling but it may be the only option when time and resources are severely limited, it is impossible to obtain a socioeconomic profile of the study population, the population is highly transient, and/or local cooperation is very uneven. In any case, the sampling procedure should be openly described in order to permit confident comparison.

Whatever the sampling method chosen, the question of adequate sample size or intensity should be addressed but very few of the studies that we reviewed attempt to do so. One way of dealing with this problem in the context of taxonomic richness and use studies is to plot accumulation curves which show the rate of addition of new species or uses in relation to increased sampling effort. Such curves tell us whether the asymptote has been reached, which is the point at which the collection of new data becomes less likely and hence the point at which the sample size is considered to be adequate. The method is extremely useful for comparative research because comparisons between two or more studies are more reliable if we have some idea of what percentage of the total stock of knowledge is represented in each study. It remains to be seen, however, how accumulation curves can be adapted to the distributional study format where the main concern is not to inventory the entire range of species or uses but instead to compare knowledge differences between subgroups of the same population.

4.3.1.3. Domain Sampling

The concept of representative sampling should also apply to the knowledge system itself. TEK is not a monolithic entity but instead is composed of a number of distinct cultural domains and subdomains of associated meanings and practices. This means that not only should it be expected that knowledge per se will vary across individuals and groups but also that types of knowledge will vary as such and therefore this parameter of variation should be taken into account when drawing up the sample. However, determining the locally relevant universe of knowledge domains on which to base the sample looms as a formidable task and from a quantitative methodological standpoint poses a very problematic issue.

Some domain distinctions ostensibly have universal significance (e.g. plant versus animal taxonomies) while others display wide intercultural relevance (e.g. edible versus medicinal resources), have a more restricted distribution (e.g. basketry), or appear to be culture-specific. Although cultural domains are ideally supposed to correspond to emic semantic constructs and therefore have psychological reality for native actors, very few of the recent wave of TEK studies actually attempt to uphold this epistemological standard. Thus it is exceedingly common to find case studies focused on use value without any mention of how the notion of usefulness in a generic sense is locally conceived or the classification of uses into broadly defined categories (e.g. edible, medicinal, construction, handicrafts, firewood, etc.) without having demonstrated that such categories are locally understood. The bias toward researcher-determined categories of significance in quantitative research appears to stem from the need to reduce data complexity, facilitate statistical analysis, and permit comparison of differences within and between groups. But given the recent emphasis on rigorous scientific methodology, it seems rather curious that the question of representative sampling of the knowledge system has received little attention. In any case, it is a problem that deserves serious consideration for the field of TEK research in general and for the development of the VITEK in particular

The principal ethnoecological knowledge domains which have been tested for variation include: biotaxonomic classification, cultural uses or significance of selected species, ecology (e.g. habitat, behavior, interspecies interaction, reproductive ecology), human impact and sustainability, and survival skills (e.g. subsistence, shelter, mating, childcare). The content focus has therefore been trained on essentially practical knowledge that is thought to be important for the material and/or social functioning and reproduction of the cultural group: the ability to discriminate among the multitude of biological species found locally, the identification of useful properties, ecological information that aid in locating, exploiting and managing natural resources, and technological know-how. By contrast, we observe in Atran's recent work on the variation and transmission of cultural models of the environment among Maya and Ladino inhabitants of the Petén region of Guatemala that some researchers are now beginning to look at the spiritual, moral, and affective dimensions of TEK in a dynamic, pluralistic context.

4.3.1.4. Stimulus Materials

Ethnoscientific methods for controlled interviewing have traditionally relied on exclusively verbal frames or prompts for data elicitation. But this may lead to erroneous results if there are synonyms or disagreement about the correct names for certain taxa, and it may be inadequate for analyzing variations in multilingual contexts unless interlanguage correspondences can be established. For instance, some researchers have noted confusion in their data caused by the inconsistent application of vernacular names to certain taxa as well as the lack of perfect match between folk taxa and botanical species.

One way of gaining greater control and precision over the query-response elicitation process is to make use of sensory stimulus materials depicting natural objects or events. In ethnobiological studies, pictorial images (photographs, drawings) and biological specimens (fresh plant collections, herbarium voucher specimens, stuffed animals) are commonly used prompts. However, these may not be appropriate in some contexts. For example, some researchers have reported difficulties of some informants (especially older individuals or groups unfamiliar with such media) to identify taxa from pictures alone. Botanical specimens consisting of the leaf and fruit or flower may not be useful because some people detect other morphological and ecological characteristics (e.g. stem habit, bark, habitat) or nonvisual cues (e.g. smell, taste) to identify the plant. The "walk in the woods" interview technique provides a more realistic and appropriate context for ethnobotanical interviews as it features living plants in their natural habitat but may lack the control needed for systematic comparison. For more standardized and uniform interviewing, plant trails and plots offer an attractive alternative and are becoming increasingly popular in quantitative studies. The advantage of transects is that they are easier to set up and are more efficient in the sense of being able to encompass a wide range of plant species growing in distinct habitats and therefore a greater portion of the culturally relevant local flora can be included in a single, continuous interview. The advantage of doing interviews in precisely measured quadrats is that quantitative vegetation surveys can be accomplished at the same time as folk botanical knowledge is measured and thus the relationships between phyto-ecological variables (e.g. density, frequency, dominance) and knowledge levels can be tested. Furthermore, the results are more conducive to rigorous comparison across different study sites and habitat types. The use of transects or plots is probably not practical for animal identification and we know of no studies in which it is used as such although this method is commonly employed for conducting faunal censuses and behavioral studies. However, some researchers report that informants have less problem identifying animal species from pictures or drawings with the exception of certain bird species.

4.3.2. Measurements of Similarity and Difference

After the raw data is collected, it must be coded and converted into a quantitative, which is to say numerical, form for the purpose of measuring the patterns of similarity/difference of responses among the sample of informants. Statistical analyses can then be performed in order to reveal nonobvious patterns in the data and to test for the significance of relationships between the measured variables. Both the quantitative measurements and the statistical operations used for the study of TEK variation differ a great deal across the studies reviewed here and depend on the field methods used, the type(s) of data collected, and the research questions being asked. Before describing the different techniques, we should make clear that quantitative measurements relevant to the study of knowledge distribution apply to two primary dimensions: (1) items and (2) people. Each one of these can be analyzed at two further levels: (a) individuals and (b) classes. The resulting matrix of data levels thus follows as: (1a) individual objects within a class, (1b) significance classes, (2a) individual informants, and (2b) classes of informants. Examples of these include: (1a) measures of different uses, citations, or importance value per species; (1b) measures of diversity, equitability and consensus per use class or per taxonomic family; (2a) numbers of species known or used and measures of knowledge level per informant; and (2b) measures of consensus, variation and average knowledge score per social group or subgroup (e.g. ethnic group, community, age, gender, etc.). Typically these measures are used in combination and interactively such that the distribution of knowledge about taxa, uses and preferences across informants is used to measure the cultural significance of the taxa while the diversity and importance values of taxa serve as input for measuring interinformant differences in knowledge.

The degree or amount of knowledge exhibited by informants has been measured at different numerical levels (e.g. ordinal, interval, and ratio scales) and these depend on the type of quantitative manipulation that is used. Three basic types of measurements can be observed and are described below: counting, ranking, and scoring. In order to calculate these measurements, the researcher must have an a priori concept of what constitutes a correct answer. The problem of validation is an important one because most researchers usually enter a research situation without any prior idea of right versus wrong information and often encounter considerable variance in responses across informants, including obvious mistakes. Our review revealed that the standard of truth judgment used in quantitative research varies according to the types of data collected as well as the types of measurements being calculated. In some cases, it is simply assumed that any and all supplied answers are valid and true, such as when simple counts are made of the number of taxa or uses known by the informants. In other cases, the answers supplied by informants are scored as correct or incorrect in reference to an answer key constructed on the basis of information given by local experts, focus groups, long-term experience in the community, or even scientific textbooks. A third approach that seems to be gaining ground, judging from the number of recent studies that use it, is to perform statistical analysis of the consensus and variation in response data and calculate relative cultural significance or importance values. These values are then used to calculate individual competence or expertise.

4.3.2.1. Counting

The most elegantly simple and extensively used measures are simple counts. Counting refers to the numerical addition of observed objects or events falling in a certain category, such as ethnobiological inventories, answers to questions about the names, habits and uses of taxonomic groups, or the number of activities or use events recorded per unit of time. Comparison between groups or samples of respondents is facilitated by converting the counts into means or percentages, such the average number of biotaxa named in freelisting interviews or the average proportion of known vs. unknown elements in identification tasks.

In theoretical ethnobiological research, counts have been employed in several ways: (a) to establish the modal size and range of ethnobiological inventories in order to make inferences about the inherent limits of the pan-human cognitive faculty for biological classification, (b) to explore the structural and substantive organization of the taxonomic system, (c) to determine the degree of correspondence between folk and scientific classification, (d) to compare the plant and animal taxonomic inventories of different groups in order to assess the impact of subsistence mode (e.g. agriculture vs. foraging) or environment (e.g. rural vs. urban) on the development of ethnobiological classification capacity, and (e) to assess the relationship between lexical retention and cultural importance.

In ethnographic studies, the counting of interview response data has served as a tool for enhancing the descriptive precision of patterns of knowledge or use. The data produced by free listing are often measured by way of simple counts. Counting the number of items listed can provide rough measures of: a) the prominence, importance or consensus of different items for local people, based on their comparative frequencies of mention across the entire sample of respondents, and b) the expertise or depth of knowledge held by different persons, based on comparison of the number of items listed per individual. Another research situation in which counts have been done is in interviews about knowledge and usage of plants found in measured transects or plots. Informants are walked through the transect or plot and asked to identify and name the plant types and state their uses. The (mean) percentage of plants known or used can be compared across individuals, groups or ecological zones. Some researchers have counted the number of plant-animal interactions cited by informants during structured queries and then compared the counts to infer differences in knowledge of ecological relationships. Counting offers the easiest, simplest, and fastest way of converting qualitative (binary, nominal) data into a quantitative form but also yields the least valuable data for statistical and comparative analysis. For example, in studies of plant use value all of the plant use citations are simply added up without any weight assigned for relative importance. In view of these limitations, counts are often made as a first step in the data analysis and then serve as input for more complex calculations.

4.3.2.2. Ranking

The ranking of data items on an ordinal scale represents another form of quantitative measurement that goes beyond the mere counting of taxonomic inventories or use citations by assigning relative weights to different taxa. The most common application of data ranking in TEK research is to measure the relative preference, importance or conservation value of individual biological species for certain use categories. The ranking of species by cultural use or significance is potentially important for management policies because it affords insight into how and why species are valued, and how these values respond to changes in the surrounding sociocultural and biophysical environments. Two basic forms of ranking can be observed: informant-generated and researcher-generated. The former type involves administering the rank order test or Likert scale interview to a sample of informants for a local inventory of plant taxa. The array of rankings are then converted to a score based on average ranking across the sample of informants. Lawrence et al. note several advantages with the informant-ranking method: a) it is relatively rapid, b) it allows all taxa to be compared by the respondents themselves (avoids researcher bias), c) it is easy for respondents to understand, d) it permits comparison of the values of different social groups, e) it is much less time-consuming than full inventories. A primary disadvantage of ranking is that the exercise becomes more difficult to perform with larger inventories of items.

Researcher-generated rankings have mostly involved the transformation of interval-scale values to an ordinal scale in order to perform nonparametrical statistical analyses due to absence of normality in the data distribution. In some studies of medicinal plant uses, researchers have first calculated use values based on the frequency of citation among a sample of informants and then converted these into an overall ranking to identify promising species for bioassays. Ross develops numerical indices of plant-animal interactions (e.g. extent to which individual plants help animals and extent to which animals benefit from all plants) and then arranges these into rankings to facilitate statistical analysis.

4.3.2.3. Scoring

Scoring represents a more refined level of measurement in the sense that the data is enumerated on interval or ratio scales. This usually involves more elaborate data collection and manipulation procedures. Scoring techniques have become especially popular in quantitative ethnobotanical research concerned with developing indices of the cultural significance or importance value of different plant species. A large and growing variety of methods have been used to produce such indices in recent years. These may be differentiated according to whether the scores are assigned entirely by the researcher or are based on judgments supplied by the informant.

Some researchers have assigned scores to distinguish between minor and major uses. For example, Prance et al. used this technique to quantify the cultural importance of forests for different Amazonian groups. Turner developed a Cultural Significance Index of plants used by Salish groups (Canada), based on a 5-point scale for three variables (frequency, intensity, and exclusivity of use) and subsequent integration of these points into a single overall score. This measure was subsequently employed with modifications in the original formula in other cultural contexts by Stoffle et al., Todt and Hannon, and Silva et al.. The advantages of research-assigned scoring are that it is relatively efficient, produces a highly differentiated data set, and can be used to reanalyze data from published sources. The main criticism of this approach is that it is biased toward the criteria imposed by the researcher and therefore may not reflect accurately cultural insiders' notions of significance.

Informant-based scoring techniques were explicitly designed to overcome researcher-biased measures of cultural importance. This approach relies on some form of consensus analysis of the pattern of interinformant agreement/disagreement in regards to a given data set. Different methods and formulas for calculating consensus were applied to investigate the medical efficacy of herbal remedies in different cultural contexts. Trotter and Logan measured consensus by way of an Informant Agreement Ratio (IAR), formulated as the total number of cases of an ailment in a sample minus the number of separate remedies cited for the ailment, divided by the total cases of the ailment minus 1. Friedman et al. computed efficacy as a function of rank order priority (ROP), calculated as the product of fidelity level (FL = ratio between the number of informants who gave the use of a species for the same treatment and the total number of informants who mentioned the plant for any use) and relative popularity level (RPL = ratio of the number of diseases treated by a particular plant and the number of informants). Johns et al. used a log-linear model to calculate the interaction effect for each remedy cited in a sample as a measure of its degree of confirmation.

Probably the most well known and influential consensus-based method used in quantitative TEK research is the informant-indexing technique pioneered by Phillips and Gentry. This method quantifies the use value of a plant species based on the overall average frequencies with which a group of informants state particular uses of particular species throughout a series of walking interviews in natural vegetation settings. Interviews may be repeated with the same informants in order to discriminate consistent/inconsistent information at the individual level. The values obtained per species were then utilized to analyze knowledge variation by age among the study population (mestizos of the Peruvian Amazon). The overall plant use knowledge for each informant was measured as the standardized ratio between the summed use value recorded for him and the summed use value recorded for the entire group of informants. In another paper, Phillips et al. adapted this method to rate the use value of different forest types, based on the summed use values of all species censused per forest type. The informant-indexing use-value measure has subsequently been utilized or adapted by a sizeable number of ethnobotanical researchers.

Another prominent variant of interinformant consensus measurement that has been used in quantitative TEK research is the cultural consensus model (CCM) innovated by Romney and associates. Since its inception two decades ago, Romney's method has been utilized in a wide range of investigations of cultural phenomena, including research on different aspects of TEK variation and change. In addition to ethnobotanical topics, the applications include studies of ethnozoological knowledge (e.g. animal habits and management), plant-animal interactions, agroecological knowledge and practices, and cultural notions of illness and curing. Rather than being focused on the valuation of biological resources and resource zones, it is more directly concerned with the structure and distribution of knowledge in a cultural group.

CCM combines mathematical with psychometric techniques, based on factor analysis, and is designed to measure patterns of interinformant agreement/disagreement about selected culturally shared domains. The method requires obtaining a single factor solution (expressed by a first eigenvalue three times greater than the second eigenvalue), which indicates that a group consensus model exists. Having established that consensus configures the domain, it permits: (a) determination of the correct (i.e. consensual) answers (when such answers are unknown beforehand) and (b) rating of the individual knowledge levels, expressed in terms of competence scores. Similar to the Phillips and Gentry technique described above, it measures individual knowledge as a function of the degree to which an individual's answers concord with the "correct" answers derived from the group. However, there are also substantial differences. The Romney consensus method should be considered more rigorous than that of Phillips and Gentry in the following ways: (a) it starts with a matrix comparison of the paired responses of all informants across all questions, (b) it requires that all informants included in the analysis be administered the same set of questions, (c) it distinguishes between right and wrong answers, (d) it accounts for the probability of guessing the right answer, (e) it is able to handle different data formats (true-false, multiple choice, and fill-in-the-blank type questions), and (e) it gives greater weight to the answers provided by presumed cultural experts (i.e. those displaying higher competence score). A valid criticism of CCM is that its applicability is limited to the quantification of consensual based knowledge and of individual departures from that standard, and thus it is clearly inappropriate for analyzing the evolution of culturally valid specialist type knowledge. Moreover, it is based on a relatively complicated computation process and therefore requires access to a computer and a pertinent computer program (e.g. Anthropac) to perform. A more general criticism of all consensus-derived measures of use-based knowledge is that they fail to distinguish potential from actual, and present from past, uses, thus inflating current real use levels.

Apart from consensus measures, there are some other, less commonly used, techniques on which to base the calculation of TEK scores that should be mentioned here. The "matching with expert" test entails preliminary elicitation of response items from one or more locally recognized expert consultant(s). Individual scores are then determined by proportional agreement with the expert(s) over a structured set of questions. The "matching with science" test uses scientist information as the authority of truth. Individual scores are calculated as the percentage of correct answers given by the respondent. These methods are much less time- and labor-intensive than consensus-based analysis, but not without their limitations. The expert matching test assumes that one or a few individuals are effectively omniscient or all-knowing standardbearers of the knowledge system, a premise that may not be compatible with the distributional model of cultural knowledge described earlier (section 4.3). The science matching test (dubiously) assumes that traditional knowledge should correlate perfectly with scientific knowledge.

4.3.2.4. Classification and Ordination

Statistical classification and ordination techniques are commonly used to study multivariate ecological relationships (e.g. species distributions in relation to environmental gradients) but in a few cases they have also been used to measure and visualize patterns of interinformant similarity/dissimilarity in regards to their ecological knowledge and practices. To carry out these types of analyses, a distance or similarity matrix is set up showing the pairwise distance (e.g. Euclidean distance, percentage dissimilarity) or similarity (e.g. correlation, covariance) between all pairs of sample objects (in this case, the objects refer to informants). In classification, the analysis seeks to partition the set of heterogenous objects into relatively homogenous groups based on the calculation of class centers, density, variance, number of members, and distinctiveness of delimitation. The main form of classification used in TEK-related research is hierarchical cluster analysis, which has been used to compare ethnobiological taxonomies within and between different cultural groups.

The basic objective of ordination is to reduce multidimensional data sets to a lower number of dimensions to facilitate analysis. It does this by screening out redundant variables and retaining those characteristics of the data set that contribute most to its variance (i.e. compressing many measurements into a fewer number). The statistical procedures for most types of ordination are computation-intensive, especially for large data sets, and hence one of the biggest limitations with this technique is the need for computers. The principal types of ordination applied to TEK data are principal components analysis (PCA) and nonmetrical multidimensional scaling (NMDS).

In PCA, a multi-dimensional set of observations, usually consisting of correlated variables, is converted into a smaller linear-combination of uncorrelated variables. PCA is generally undertaken to identify patterns of grouping (i.e. similarity) of compressed data points as well as deviations (i.e. dissimilarity) of data points. Höft et al. propose that PCA applications for ethnobotanical research include: (a) revealing whether certain groups of people value the same species in the same ways, (b) spotting individuals who respond differently from the majority, and (c) grouping species according to the use values assigned by people. When this type of analysis is performed on people (instead of item or attribute variables), it can be used to calculate the amount of shared knowledge as well as the degree of correlation of each subject's responses with those of the group, i.e. individual scores. In that sense PCA is very similar to CCM but the former uses a correlation matrix for the analysis while the latter is measured with the proportion of identical answers (adjusted to account for guessing or chance) between each pair of informants or with covariance (i.e. correlation of pairwise variances). In some of the studies reviewed here, PCA was employed as a supplement to CCM to test whether the data really conforms to a single underlying model of group consensus (i.e. a single-factor solution).

NMDS arranges multidimensional objects in a low-dimensional space suitable for visualization in two- or three-dimensional graphic displays. The objects are reproduced in the graph to reflect the observed distances among them with the least amount of stress (i.e. deviation from real distance). Besides providing the researcher with a visual display of the clustering and dispersal of the objects in relation to each other, the objective here is to detect meaningful underlying dimensions that allow the researcher to explain observed similarities or dissimilarities between the objects. NMDS has several advantages over linear ordination: (a) it makes no assumptions of linearity, (b) it can be used with ordinal-scale data, and (c) the underlying dimensions of contrast come from the respondents' own judgments or answers. This technique has been used to study intracultural and intercultural variation of concepts of illness, specialist versus nonspecialist ethnomedical knowledge, cross-cultural ethnobiological taxonomic classifications, agricultural management style in relation to acculturation, and comparison of knowledge about yellow-fin tuna between folk fishermen and fishery scientists. Unlike factor-based analyses (e.g. CCM and PCA), NMDS is especially useful for comparing specialist and nonspecialist knowledge because it shows data clusters or orderings on the basis of one-to-one or pairwise comparisons rather than one-to-all comparisons.

4.3.2.5. Diversity Indices

Diversity indices are quantitative tools that have been widely used in ecological research to measure species diversity and evenness in a community . Begossi proposes that ecological diversity indices have several potential applications for the study of the relationship between human populations and biodiversity: (a) to evaluate the breadth and intensity of resources used by particular groups, (b) to facilitate comparisons between different groups in different environments, (c) to assess the adequacy of sampling effort, and (d) to determine the land area needs of a population based on the data of resources used. The literature review revealed that diversity indices have primarily been used in TEK research to characterize plant use inventories, the sole exception being Hardesty's measure of niche width which is focused on the total breadth and evenness of dietary resources.

The most commonly used indices include: Species richness, Simpson's diversity index, and the Shannon-Wiener diversity index. Species richness (S) refers to the total number of different species present without taking into account proportions or distributions in an area. This is the principal measure of diversity taken to calculate the famous species-area curve (i.e. increase of species per area sampled). A variation on the species-area curve, plotting the number of useful folk taxa or use types against the number of interviews or informants interviewed, has been utilized to evaluate the completeness of the sample and is considered important for quality control when comparing inventories across different groups. Simple counts of the inventories of useful plants is a common statistic reported in many ethnobotanical studies and these have been used to proxy the richness of knowledge and use habits among different groups. S is sometimes calculated per individual and then plotted in the form of rank abundance curves to visualize the equitable/inequitable distribution of knowledge within a community.

The Simpson diversity index (D) represents the probability that two randomly selected individuals in a habitat belong to the same species. It takes into account the number of species present as well as the abundance of each species. The Shannon-Wiener diversity index (H') measures the uncertainty of predicting to what species an individual, chosen at random from a sample of S species and N individuals, will belong. Similar to the Simpson index, this measurement takes into account both species richness and proportions. When applying these measures to ethnobotanical data, the number of interview citations per taxa is substituted for species abundance. Both of these measures have been used, sometimes side by side within the same study, to compare the diversity of knowledge among different cultural groups as well as the diversity of plants used for different use categories. Begossi et al. compared H' values between gender and age subgroups in a Caiçara community (Brazil) and was able to identify which subgroups had greater knowledge of diversity. In the same study, the authors also calculated rarefaction curves, a complementary measure to Shannon, to control for different sample size. The rarefaction method involves rarefying the samples to make them comparable by taking random subsamples of individuals of equal size from the total. Hardesty uses the Shannon-Wiener formula for calculating niche width according to the criteria of caloric significance, protein contribution, biotopic variation, and seasonal variation.

Similarity indices, such as Jaccard's index and Sorenson's coefficient, represent another set of ecological indices which have been utilized, though more rarely, in ethnobotanical research. The general purpose of these in ecological research is to measure beta (rather than alpha) diversity. Campos and Ehringhaus calculated the similarity of cultural uses of palms among different communities and the use specificity of species with the Jaccard index. Lozada et al. also used Jaccard to measure the overlap of edible and medicinal species between sexes in a single community. In plot-based ethnobotanical inventories, the Sorenson coefficient of floristic similarity was calculated to establish floristic similarity/divergence between communities whose knowledge was being compared.

4.4. From Variation to Process

The pattern(s) of TEK variation in space can be directly linked to the process(es) of TEK change over time through another set of methods and measures. Similar to the previous phases of inquiry, we find a great variety of methods being used to reveal the dynamic states and movements of knowledge in relation to historical shifts in the surrounding human and biophysical landscape. In addition to documenting the empirical trends of knowledge in different groups and localities, this body of work also addresses the major question of why knowledge changes or persists and what are the main environmental factors driving change or persistence. There are two basic approaches for documenting cultural knowledge change/continuity through time: a) longitudinal studies and b) cross-sectional studies.

4.4.1. Longitudinal Studies

Longitudinal studies entail the collection and comparison of time-series data, which may be before and after key events or interventions, at periodic intervals, or restudies of previously studied communities. Given the relatively short time constraints on most research projects as well as the lack of similar baseline or prior data sets in most places, it is not surprising that we are able to find few studies of this kind (thus far anyway). The only truly diachronic TEK studies that we are aware of are Zarger and Stepp's restudy of ethnobotanical acquisition among Mayan children in the community of Majosik' (Chiapas, Mexico), studied 30 years beforehand by Stross, and van Etten's inventory of maize varieties among highland Guatemalan farmers which he then compared to inventories conducted in the same region by two separate researchers more than 60 years before his fieldwork. Surprisingly, both studies found remarkable persistence, rather than erosion, of knowledge despite several decades of progressive socioeconomic and ecological transformations in the study communities. These findings highlight the fact that TEK persistence rests upon an active process of intergenerational transmission that is not always destroyed or debilitated by changes in the surrounding environment (a point that I will return to later). But for most researchers, the only option has been to rely on indirect methods for inferring the transmission process.

4.4.2. Cross-Sectional Studies

Cross-sectional (or transversal) studies refer to data collecting operations at one point in time. However, several creative techniques have been devised for inferring diachronic processes of TEK change or persistence from synchronic data. The basic strategy employed in this case is to study and measure the socially and spatially patterned variation of knowledge and practices within and between groups. A second step is to examine the correlations of such variations with other social change indicator variables. A third step is to interpret the observed relationships by reference to the larger historical-ecological context. The overwhelming majority of empirical studies of TEK change and process conform to this type.

Age stands out as the social variable most commonly used to chart the slope of knowledge change in dynamic contexts, probably because it is universal and directly indexed to time (see below). In studies of the ontogenetic development of knowledge acquisition, the array of individually recorded and scored knowledge types and amounts are indexed according to the person's age and the resulting knowledge-by-age trend lines are read as site-specific models of the age-dependent learning process. This technique is widely used in developmental research on folk biological classification, naive biological reasoning in children. Age-on-knowledge curves are also frequently employed to infer the diachronic evolution of TEK amount and content in groups over time. In that case, the difference(s) between older versus younger people's knowledge levels is measured and thought to approximate the degree of change (whether loss or increase) that has occurred in that time interval. It should be noted, however, that using age alone as the measuring stick for change can be problematic given that the normal learning curve leading from novice (e.g. child) to expert (e.g. adult) is also time- and age-dependent. For this reason, some researchers have cautioned that the knowledge-on-age trendline in a stable (i.e. nonerosional) situation should reflect gradual increments of change whereas trendlines displaying sharp breaks or noticeable tips are indicative of irreversible change (e.g. erosion) over time. Another way of overcoming the explanatory limitations of age is to include additional change indicator variables in the analysis. Because age overlaps and interacts with other social variables (e.g. education, occupation, language(s) spoken), it can also be used as a control variable to isolate the impact of these on patterns of knowledge variation. An alternative approach is to confine the sample population to adults beginning with the age level at which intellectual maturity is thought to be reached. Accordingly, the difference between knowledge levels of younger versus older generation adults is represented as an approximate measure of the direction and rate of knowledge change. The resulting trend lines over time can be used to identify the most endangered types of knowledge and even to project the estimated extinction dates if present trends continue.

Besides age, we were able to observe a number of other social variables that are treated as proxies or indicators of acculturation. Some of these are obviously direct markers of modernization and are easily measured on an individual basis, like years of school education completed, literacy, fluency in the national language, amount of income obtained from cash copping or wage labor, material possessions, and travel experience. Others are more indirect or qualitatively described change indicators that are assumed to affect entire communities, such as availability of modern services (e.g. schools, health clinics, communications and transportation, nontraditional housing, indoor plumbing), changes in settlement pattern, introduction of exogenous technology, economic orientation (e.g. commercial vs. subsistence activity), proximity to roads or urban areas, contacts with outsiders, habitat degradation, and nontraditional beliefs and values. Rocha attempts to measure traditional/modern attitudes in a systematic way by conducting structured interviews on people's views of community life and regional development and their personal roles in it as well as proper gender roles. Interestingly enough, he finds a strong positive correlation between traditionalism in such attitudes and conservation of traditional agro-ecological knowledge among Peruvian peasants. In some cases, the validity of some of the measures purported to proxy modernization/marginality may be questionable (e.g. housing type, straight-line distance to large towns). Moreover, few studies really attempt to consider a wide range of these indicator variables at the same time and investigate the correlations or interactions between them. In our opinion, the most convincing explanations of the impact(s) of environmental variables on TEK change are those that integrate careful measurement of specific variables along with a sensitive and more textured (i.e. qualitative) understanding of the overall context (social, historical, ecological) at the local as well as higher levels.

Another technique that has been used to adduce change from synchronic data is to gauge the discrepancy between oral statements of potential use and observed behaviors of actual use, based on the logic that usage, or practice-based knowledge, should be impacted before theoretical, or conceptually-based, knowledge but that decline of the former will foreshadow decline of the latter. Byg and Balslev, for example, found a credible gap between individuals' knowledge and uses of Dypsis fibrosa (Arecaceae) as reported in interviews versus the actual use habits by the same individuals as observed by the researchers among villagers in eastern Madagascar. The correct interpretation of this discrepancy, however, still requires a sensitive appreciation of the local dynamics and motives of knowledge change and the contextual factors driving them. For example, Albuquerque, interprets the knowledge of an abundant number of little-used medicinal plant species among rural dwellers of caatinga forest in northeastern Brazil as a strategy of diversification of therapeutic options, rather than passive erosion of knowledge, in response to acculturation pressures. Gazzaneo et al. report that cultivated and exotic species are the preponderate medicinal plants used by inhabitants of a sugarcane refinery land grant in northeastern Brazil due to extensive vegetational alteration and prohibitions on the exploitation of protected forest reserves. Paying attention to informants'statements about which uses are ancestral but no longer current may provide useful insights in this regard.

4.4.3. Statistical Tests of Significance

In the literature review we came across a large number of different statistical procedures and tests that are used to explore the covariate and causal relationships between knowledge or practices and the surrounding environmental variables. The main ones include correlation, contingency tables, t-tests, analysis of variance, and regression, each one of which has variant forms. The conceptual and mathematical description of all of these would add considerable length and take us beyond the scope and purpose of the present review. It will have to suffice to simply advise the interested reader to consult a standard textbook on probabilistic-statistical methods for the social sciences and to point out that the operations chosen depend on the type of data collected and questions being asked. Principal data factors determining the choice of method include whether the data are measured as continuous or discrete values and whether the distribution encountered is normal (i.e. data points are distributed more or less symmetrically around the mean) or not. Some of the authors explicitly ran tests to determine normality or employed normalization techniques to transform their data in order to be able to use the more powerful parametric tests.

The methods used for testing differences between groups are of particular interest here because irreversible, as opposed to stochastic, change should be reflected at the collective level and because most historical and ecological processes affect entire local populations, or subgroups within populations (e.g. age groups, genders, socioeconomic class), rather than just certain individuals. In order to establish that significant differences between groups exist, it may be necessary to analyze patterns of knowledge variation within them as well. Weller and Baer, for example, discuss various statistical procedures for analyzing within- and between-sample agreement and show how the proportion of shared vs. unique understandings about disease concepts among different cultural groups is directly related to the historical depth of their exposure to the disease in question. In several studies that were reviewed, the authors measured patterns of residual agreement (i.e. observed agreement minus predicted agreement) to explore within- vs. between-group differences. When within-group residual agreement was greater than between-group residual agreement, or if residual agreement is correlated with total observed agreement, this was interpreted as reliable evidence of group differences. The main techniques for testing the alternative hypothesis that groups are different were one-way analysis of variance and t-tests. Where socioeconomic variables could be measured on a continuous scale, regression was the preponderate method of choice. In that case, the test structure was set up with the latter as the explanatory variables and the knowledge score as the dependent variables. To our knowledge, no one has attempted yet to investigate the possible causal or positive feedback effects of TEK shift on other types of sociocultural change. To do so would probably require a nonlinear analytical model as well as more complex statistical procedures than those currently in use (e.g. computer simulation).

4.5. Structure and Process of TEK Transmission

In addition to measuring the change/continuity of knowledge content and distribution over time, there is yet another approach to studying the processual dimensions of TEK that deserves mention. This approach focuses on the social organization, mechanics, and contexts of intergenerational knowledge transmission, which has come to be recognized as a crucial component for long-term preservation. Several studies have explored the social relationships and channels through which knowledge is passed from individual to individual, and from generation to generation. Different data collecting styles have been used to explore these relationships. The studies by Messer, Ruddle and Chesterfield, Murphy, and Wilbert relied on the traditional anthropological techniques of participant observation and in-depth, informal interviewing of key informants. The results of this approach are richly detailed, normative, qualitative descriptions of the key relationships (e.g. grandmother-granddaughter, father-son, grandfather-grandson, female coresidents) that regulate the flow of different kinds of ethnoecological information (e.g. plant uses, food procurement tasks, phytotherapies) in diverse ethnographic settings. By contrast, Hewlitt and Cavalli-Sforza and Ohmagari and Berkes used a more rapid appraisal, quantitative survey approach in their studies of Aka hunter-gatherers (Central African Republic) and Cree women (Canada) respectively. The method used there was to administer a highly structured interview schedule to a sample of respondents, asking them if and when (at what age) they had learned an inventory of traditional bush technical and social skills and from whom they had learned it. Analyzing the results statistically, they were able to provide a precise quantitative profile of the knowledge transmission process among the respective groups, specifying the specific category of knowledge, the age when acquired, and who taught it. A third approach modeled the social networks through which ecological information is exchanged. The method consisted of the following steps: (a) careful selection of a sample of respondents (all mature adults, evenly divided by sex, no one immediately related by kinship or marriage to anyone else in the sample), (b) elicitation of ranked lists of socially significant others, of most frequent interlocutors about forest matters, and of recognized forest experts, and (c) follow-up interviews with the highest and lowest ranked individuals appearing in the lists referred to above, eliciting their respective social interaction and forest expert networks.

Descriptions of the specific mechanisms by which knowledge is passed along or picked up provide unique insights into why TEK is maintained in some places and not in others. Good descriptions of the various "informal" (i.e. extrascholastic) learning modalities by which TEK is acquired can be found in the ethnographic accounts authored by Messer, Ruddle and Chesterfield, Borofsky, Katz, and Murphy. These studies highlight careful, qualitative, detail-driven, and culturally-sensitive observation of the various forms of knowledge acquisition that appear in normal social and daily intercourse, a feat usually achieved through long term fieldwork and intimate contact with the study community (the main exception being Ruddle and Chesterfield whose fieldwork was relatively brief). The reports by Katz on the children of rural Sudanese herders and farmers is especially insightful in this regard. She reports that most environmental learning takes place by accompanying and assisting elders in their agricultural work activities (i.e. peripheral participation) and describes a "pattern of integrated learning," characteristic of other non-industrialized settings as well, consisting of "observation, instruction, imitation, instruction and correction, and supervised or guided practice [that] unfold in succession (1986:46)." A less nuanced but more quantitative approach to documenting the prevailing local pattern of learning is exemplified in the Ohmagari and Berkes study. These authors included a multiple choice question about the stages or processes of the learning complex (hands-on experience, learned by observing only, and not learned), based loosely on the typology proposed by Ruddle and Chesterfield, in their structured questionnaire. Obviously this is a quicker and more direct way to answer the question about local learning styles, and the quantitative method allows for statistical analysis, but it is also probably less reliable and in any case should be confirmed by the researcher's own observations. Finally, Zent examines the impact of acculturation on traditional patterns of ethnobotanical knowledge transmission in relation to observed variation in knowledge content between age and community subgroups of the Jotï Amerindians of Venezuela. This study combined structured interviews and statistical analysis of the results along with quantitative time allocation and informal participant observation methods.

4.6. Summary Evaluation of Methods Review

The foregoing review reveals an enormous variety of quantitative methods and measures employed to study the variation and change of TEK in different settings. It was emphasized that the best method to use depends to a large extent on the particular research problem and type of data being collected. A large number of the studies reviewed here in fact make use of a combination of various techniques. For purposes of the VITEK, many different combinations could be selected in theory. However, we consider that this choice needs to be limited in order to satisfy the goal of global applicability. We believe that the best approach for achieving widespread acceptance and sustained implementation is by engaging and empowering local community members to eventually carry out the data collection and analytical process on their own, which is to say a participatory, collaborative approach. As mentioned in section 1.3, the optimal use of the indicator and the assessment test on which it is based would be at regular time intervals (e.g. every 5 or 10 years) to produce true time series measurements. The ideal scenario for making this happen would be by giving the participating communities the training and capacity to manage the assessment on a self-sustaining basis (and use the results for their own purposes in addition to the global assessment). Unfortunately, many of the methods and calculations reviewed above are too complicated and/or laborious to expect that local people would be willing or capable of doing this over the long term without specialized and perhaps recurrent instruction. For these reasons, the best methods from the standpoint of the VITEK will have to be the most simple ones. Further justification of this approach is presented in section 6, which outlines the recommended methodological protocol.