We have two questions to consider in this section. How come you and others in your community feel the same about so many things? And what does consensus imply in terms of measuring shared meanings and feelings?

Emergence of Cultural Standards

Your sentiment about an object may be the result of private encounters with the object. Additionally your sentiment is shaped by interactions with others - individuals in your social groups, or strangers in public places or the mass media. Your encounters with others pull your sentiment toward a cultural standard.

Let's use child as an example.

Imagine you are individual A in the diagram below. Your own private experiences with children are one source of your sentiment about child. You express your sentiment in public actions toward children and in talking about children with your associates. Your public acts influence others' social acts, as indicated by the curved arrows branching away from "A's Acts." However, your associates are not passive: they too have private experiences with children which they express in their public behavior and talk, thereby influencing their associates' social acts, including yours. Interacting and talking together changes your sentiment toward child to be like the sentiments of your associates, and their sentiments become more like yours. A shared sentiment toward child emerges. That shared sentiment affects your private experiences with children in the future as you try to experience children in a way that confirms your sentiment about them. So even your private experiences will tend to support the shared sentiment toward child, and will provide little basis for further change in the group's shared sentiment.

Sharing means that A's sentiments are almost the same as B's sentiments, and B's are almost the same as C's, etc.  This consensus puts a cultural jacket around individual sentiments.

Intersubjectivity is one important benefit of cultural consensus. When with others from the same culture as you are from, you evoke shared sentiments as you talk about your personal experiences, and to a large extent your audience feels the same as you do about your experiences as you describe them.

Another important consequence of cultural consensus is that we can measure sentiments efficiently.

Some Principles of Sampling

The error in estimating a statistic from small samples depends on how variable the population is, as demonstrated in the following example.

Suppose that we want to know the average height of the population shown below in aqua, titled "Eight different values in the population." Individuals in the population are shown as distributed horizontally in terms of their heights; the population mean is 3.25 measurement units of height. (You can imagine that the population consists of one set of quintuplets, one set of quadruplets, one set of triplets, three sets of twins, and two other individuals.)

From the random sample shown in red, we would estimate the average height as 4.5 units of height. The random sample shown in green gives an estimate of 6. Obviously, estimating the average height of the population from a sample of four could be very misleading.

Now consider the population shown in aqua in the lower part of the diagram, titled "One value in the population." (You might imagine this population to consist of 20 biological clones.) The population mean is 5 measurement units of height. The random sample shown in red has a mean of 5. The random sample shown in green also has a mean of 5. Every sample of four will have a mean of 5. In fact, a sample of four is extravagant in this case. We confidently can estimate the average height of the homogeneous population from a single case.

Socially sharing a sentiment makes individuals homogeneous with regard to the sentiment. Then any individual represents the group with regard to the sentiment. Maybe individuals' private experiences make each diverge a little from the cultural norm at a given point of time, but averaging over just a few individuals gets rid of effects of unique private experiences. Thus, just a few  informants provide a good basis for assessing any aspect of culture, including a shared sentiment. Call this the ethnographic simplification.

Assessing shared characteristics contrasts with the procedure required to assess purely personal characteristics in a population.

For example, age is a personal characteristic. An individual's age does not influence anyone else's age, so we cannot take any one person's age as an indicator of other individuals' ages or of the population mean age. To estimate accurately the average age in the population, we have to measure the age of every individual in the population - or at least of individuals in a representative sample - and compute the average. Assessing unshared characteristics like age requires censuses or sample surveys, and these are much more costly than studies using the ethnographic simplification.

The role of people and questions is reversed in studies using the ethnographic simplification, as compared to survey studies. In traditional surveys we have this model:

In studies using the ethnographic simplification we have this model:

Typically a culture has just one position with regard to a specific condition, and everyone gives the same answer to a relevant question.

For example: Is a father always a male? would be answered yes by anyone from traditional U.S. culture.

In principle, only a single informant would be needed to ascertain a cultural condition, if every informant knew the culture perfectly and answered every question thoughtfully. However, we have to assume that a given informant might get distracted or be inadequately socialized with regard to some cultural conditions. So, responses from one informant are not perfect indicators of cultural conditions. We need to improve validity by combining answers from multiple informants.

Yet the sampling issue involved in combining multiple informants is completely different than the issue that arises in survey analysis. Number of informants needed is a question of how many indicators we need to identify the unique cultural position in the presence of measurement error, not how many instances from a population we need to represent the population's variation.

Consensus and Informant Accuracy

Anthropologists (Romney, Weller, and Batchelder, 1986, Culture as concensus: A theory of culture and informant accuracy. American Anthropologist 88: 313-338) developed a statistical model of error in informant's answers to questions about culture. They used their model to compute the minimal number of informants needed to infer cultural positions from true-false questions (p. 326).

The table shows that a half-dozen expert informants can provide a very clear picture of how true-false conditions are set within a culture. True-false questions may seem a simplistic way of approaching the problem of culture definition, but:

Classic Measurement Theory

Classic measurement theory can be brought to bear on the problem, once we understand that informants serve as multiple indicators of a cultural position, just as items on a questionnaire serve as multiple indicators of an individual's position.

Summing, or averaging, across different indicators of the same thing produces a more valid measurement than is provided by any one of the indicators. In the following diagram each indicator - Measure 1, 2, 3, and 4 - correlates .707 with the Factor (which is the ideal variable measured without error), but the Score, which is the sum of the indicators, correlates .89 with the Factor. (Based on Heise and Bohrnstedt: Validity, invalidity, and reliability; pp. 104-29 in E. Borgatta and G. Bohrnstedt (eds.), Sociological Methodology: 1970, San Francisco: Jossey-Bass.)

The usual application of the classic measurement model assumes that we have a population of people, each with a distinctive true score on the factor. We assess each person on different indicators of the factor, sum or average the results for each person in order to estimate the person's true score, and thereby get a score that shows each person's relative position on the factor. We improve measurement either by designing new indicators that correlate better with the factor, or by adding additional indicators of the same quality as the ones we have.

Applying the classic measurement model to the study of cultural sentiments works like this. Within a particular society we have a population of cultural sentiments, each with a distinctive value on a dimension like evaluation, potency, or activity. We assess each cultural sentiment in terms of the responses of different informants, average the results for each sentiment, and thereby get a score that shows each cultural sentiment's relative position on the dimension. We can improve measurement either by choosing new informants with better internalization of cultural sentiments, or by adding additional informants comparable to the ones we already have.

Different Voices

Suppose people lack consensus in their sentiments regarding an issue? In the extreme case sentiments about the issue would be unpredictable from one person to another, one could not identify a shared affective meaning characteristic of culture, and the ethnographic simplification would not apply. Maybe this kind of situation does occur with some new and controversial issues, but such a situation won't last because of the sociality of humans.

Groups of people begin developing a shared perspective and shared sentiments as soon as individuals begin talking to one another about an issue. Not every individual will be absorbed into the group position, but those individuals who disagree too much to adopt the group norm gravitate to other groups providing them with better affective resonance. In this way pockets of consensus - or subcultures - emerge in different interest groups. Consequently persistent diversity in sentiments implies not anarchic individuality but the existence of subcultures.

Subcultural diversity complicates the measurement problem. The researcher has to identify each subculture and then assemble a set of expert informants to assess unique sentiments within that subculture. Thus, more informants are needed than when there is overall consensus regarding an issue. However, we still do not need to treat every individual in the population as the only valid informant regarding self - not as long as the sentiments of each individual are predictable from sentiments of the individual's associates.