Why Being Consonant With Your Personal Model Cannot Explain the Effect of Cultural Consonance

The central plaza of Ribeirão Preto, Brazil (photo by author)
The central plaza of Ribeirão Preto, Brazil (photo by author)

At the beginning of the semester and my class, “Culture, Mind, and Behavior,” I started thinking about this topic, because this class is devoted to cognitive culture theory, including the concept of cultural consonance. Cultural consonance is the degree to which people incorporate into their own beliefs and behaviors the cultural prototypes for belief and behavior encoded in shared cognitive models. In other words, it’s how closely people match up with the culture around them.Over a number of years of research, my colleagues and I have reliably identified shared cultural models in a number of domains (e.g., family life in Brazil), and we have found that higher cultural consonance–that is, people actually, for example, believing that their own family matches the prototypical Brazilian family–is associated with better health status (such as lower stress and depression, and lower blood pressure).

Whenever I lecture about or teach this material, inevitably there is somebody who raises some form of the following objection: well, what about people who reject the shared cultural model and follow their own personal model of how life is to be lived? Sometimes it’s a student who, I often suspect, is offended by the notion that he or she is not all that “special,” i.e., that what he or she thinks or does is actually a variant of what a lot of other people are thinking and doing, simply because they are all working from the same template. Other times it’s a more principled objection from anthropologists who are taken with the notion of personal agency and (mistakenly) think that I think that people are what the Brits charmingly call “cultural dopes,” i.e., we are all cookies stamped out by the cultural cookie cutter. Of course, the concept of cultural consonance is completely compatible with an agentic perspective, and in fact I’ve got at least one paper planned for the future to explore how that works in some detail. In any event, however it is framed, there are people who object to the notion of cultural consonance because they think that personal, individual models would trump the influence of cultural consonance were I to measure and incorporate them into my analyses.

At the outset, I would note that conceptualizing and measuring the concept of “personal consonance” or “individual consonance” is a much thornier issue than you might think. Remember that to measure something people have to be presented with the same stimuli, or, in our business, asked the same questions. If you are committed to the idea that people have individual or personal models, then logically each individual would have to be asked about their model only and how they are committed to it. How could that be turned into a comparable measurement from one person to the next? You might present an alternative to identifying each individual model that involves asking people questions like “I always try to do what is most important to me,” and having them respond on a Likert scale. Well, congratulations, you have just re-invented 1950s social psychology and the psychological construct of self-efficacy! Don’t get me wrong, I’m all in favor of incorporating a psychological construct like this into this research, and seeing how it might alter the effect of cultural consonance (it doesn’t). But my main point is that measuring something that people mean by personal or individual models is very, very difficult.

Let’s assume, however, that consonance with an individual model, or “IC” (for individual consonance) can in fact be measured, and hence both IC and cultural consonance (or “CC”) can be examined as influences on health outcomes (“HO”). What would happen? At this point I’m going to digress briefly to reiterate something one of my favorite bloggers, Paul Krugman, has written from time-to-time, and that is the importance of your theoretical model in thinking through a problem. It’s one thing to say, for example, well, somebody’s IC could be more important than their CC. This is on a par with saying, “well, I could learn to levitate.” Yes, maybe, but if your theoretical model of the world includes something called gravity, then you have to think more complexly about this levitation business. It’s the same in anthropology. Yes, somebody’s IC could trump their CC in influencing HO, but what is likely to happen given a particular theoretical model of how the world works?

Here is a theoretical model:  New slide

This theoretical model says that CC is associated with HO, and IC is associated with HO, and in some way, CC and IC are correlated. To simplify things, let’s assume that HO is measured in terms of positive health—like better self-reported health or positive affect—so that being culturally consonant and being individually consonant are associated with feeling better, so the association is in a positive direction.

This simple exercise clarifies things a bit. We already know that CC is associated with higher HO (this has been replicated many times by multiple investigators). The question is what happens when IC is introduced into the picture? Well, it depends, and it depends exclusively on how you conceptualize the correlation between CC and IC. What kind of operational or statistical model can we use here? I’ve already been talking about correlation, so that’s my operational model. When, in looking at the correlation between CC and HO, I take into account the correlation of IC with HO, and of IC with CC, what happens?

In somewhat more technical terms, what we want to do is to remove the effect of IC on CC, and remove the effect of IC on HO, and see what is left over in terms of the correlation of CC and HO. Or, put differently, we want to look at the correlations of the residuals among all of these variables.

That sounds like a statistical mouthful, but it actually can be understood very simply by looking at just the numerator of a partial correlation coefficient (which, I might add, is the same as the numerator of a partial regression coefficient; the only difference is in the denominator, or by what you are standardizing). Here’s the numerator, spelled out in prose:

[Correlation of HO and CC] – [(correlation of IC and CC) X (correlation of HO and IC)].

What this says is that if we want to control for, or otherwise get rid of, the influence that IC has in looking at the correlation of CC and HO, we have to subtract out the product of the correlation of IC and CC and the correlation of IC and HO. You don’t even have to be all that much of a statistical heavyweight to get this. Correlation = co-variation. To “purify” the co-variation of CC and HO, we have to get rid of the co-variation of HO and IC, and the co-variation of IC and CC (see the model above).

And, it depends exclusively on the correlation of IC and CC, and frankly has nothing much to do with the correlation of IC and HO. Think about the simplest case, where the correlation of IC and CC = 0. If you multiply the correlation of HO and IC by 0 (zero), that second term in the equation above becomes 0, and the correlation of CC and HO is completely independent of the correlation of IC and HO.

What happens if the correlation of IC and CC is positive? In this case, the second term in the equation would become negative, and the correlation of CC and HO would be reduced in proportion to the size of the IC/CC correlation (again, it doesn’t have much to do with the IC/HO correlation).

Finally, if the correlation of IC and CC is negative, the correlation between CC and HO would increase in proportion to the magnitude of the IC/CC correlation (remember that in this case the second term in the equation would be negative, and then you would be subtracting a negative number, which makes it positive).

To summarize: when the IC/CC correlation is zero, no effect. When the IC/CC correlation is negative, the CC/HO correlation goes up. Only when the IC/CC correlation is positive would the CC/HO correlation potentially go down, and then only in proportion to the magnitude of the IC/CC correlation.

This little exercise should help skeptics of cultural consonance re-think their critique. What do they really mean to say? If they are saying that an individual can have his or her own cognitive model of the world that sets them apart—and this is, I think, what most of them are trying to say—then it turns out that it does not alter the effect of cultural consonance. Another way of thinking about this is in terms of the scatter of data-points around a regression line in a scatterplot. If the plot is of the correlation of CC and HO, we will see that many people cluster around the regression line, i.e., as their cultural consonance goes up, their health outcome improves. The people farther from that regression line are the people who, perhaps, are adhering more closely to their individual model. If we take that into account, we are reducing the noise in the data, and the effect of cultural consonance on health outcomes becomes more clear (technically, the standard error of the regression coefficient will go down). Or, IC and CC simply become independent influences on health.

Alternately, skeptics might be saying that IC and CC are negatively correlated. This would be equivalent to saying that overall, people who adhere to their own individual models do so and explicitly respond in the opposite direction to questions about shared cultural models. Well, maybe, but think about this explicitly. In Brazil, for example, the cultural model of social support says, in part, that in response to many problems, you start by seeking help and assistance from your family and friends, and then you gradually seek help in less intimate relationships of work, church, and ultimately professional supports like doctors and lawyers and such. The “IC and CC are negatively correlated” position would argue that in describing how you follow your own model of social support, you also describe yourself in relation to the cultural model of social support in the opposite direction, i.e., you never ask family and friends for help, and you exclusively ask strangers and professionals for help. OK, if one or a few people answer like this, it’s just the wacky nature of humans. But for it to affect the correlation of CC and HO, you would have to have a set of people systematically responding in this way. Hmmm…and of course, even it you want to believe this, it is in this case that controlling for IC would cause the CC/HO correlation to go up.

The third position is that IC and CC are positively correlated. If this is so, it calls into question the whole notion of an individual model and individual consonance, since it (individual consonance) would turn out to be some version of the cultural model and cultural consonance. Actually, some skeptics fail to appreciate that, in one sense, cultural consonance is the personal, lived model, formed as people knit together a life for themselves in the context of the environment of shared meanings (the cultural model) and the various factors that enable them to act, or constrain them from acting, on those shared meanings. A super skeptic—like a full-blown psychological reductionist—could argue that the whole theoretical construct and measurement of cultural consonance is epiphenomenal to individuals, with individual models, making individual decisions, of how to live based on those models. They just turn out to be similar from one individual to the next, probably based on some neurocognitive module selected for in evolution, or some basic personality construct, or whatever. This seems highly implausible, however, given that the construct of cultural consonance is based on a well-developed, well-articulated cognitive culture theory, along with a well-understood measurement model. The full-blown psychological reductionist would have to argue that the theory and method to derive cultural consonance is a social scientific version of reading chicken entrails. They may believe that, but it seems pretty implausible.

So, to me the most plausible way that a consonance with individual models would work in this process is the independence of individual consonance and cultural consonance. If this is the case, bringing individual consonance into the theoretical model would not alter the influence of cultural consonance on health outcomes, except in the relatively trivial sense that controlling for individual consonance would increase the statistical significance of the coefficient assessing the effect of cultural consonance (because the standard error would go down). Where I’m skeptical is in regard to the measurement of individual consonance. I have difficulty envisioning a truly satisfying measurement.

In the final analysis, here, I return to the wisdom of Paul Krugman. Your theoretical model is of paramount importance. It is through your theoretical model that you can sort out the implications of various alternatives, just as I have done in this post. Models of every variety—theoretical scientific models, shared cultural models, personal models—are good to think with.