Religion and Memory

In the paper I discussed the other day, Atran and Norenzayan argue that one of the most important factors in determining whether a religious narrative is successful is how memorable it is. Easily remembered narratives get passed on, while difficult to remember narratives are forgotten. Thus, successful religious narratives will likely exhibit features that make them memorable. Norenzayan and Atran hypothesize that the feature shared by successful religious narratives (and cultural narratives in general) is that "they correspond to a minimally counterintuitive (MCI) cognitive template that includes mostly intuitive concepts combined with a minority of counterintuitive ones1. In this post, I'll first talk a bit about what minimally counterintuitive means, and then describe two recently published studies in which Norenzayan and Atran support the hypothesis that minimally counterintuitive narratives are more memorable, and that successful narratives are minimally counterintuitive.

Intuition and Counterintuition

When Atran and Norenzayan refer to intuitions, they have something specific in mind. From a very young age, we humans exhibit what you might call an "intuitive ontology." For example, in their first years of life, infants develop the understanding that solid objects can't radically change their shape (from one kind of object to another), that they can't change in number (if one object goes behind a screen, and two come out, something's up), and that solid objects generally can't penetrate each other. These intuitions constitute what researchers sometimes call "folk physics," which also includes things like an early conception of causation. In addition to a folk physics, we also have a "folk biology," in which we have intuitions about biological kinds distinct from those we have about non-biological kinds; a folk psychology (sometimes called "theory of mind"), which includes our beliefs about what's going on in other people's heads, and develops over the first several years of life; and from a young age, we distinguish between animate and inanimate objects. These are the sorts of intuitions to which Atran and Norenzayan are referring.

In their theory paper (discussed at the link above), Atran and Norenzayan argue that one of the universal features of religious narratives is that they include counterintuitive beliefs. These beliefs violate intuitions in one or more of the "folk theories" mentioned above. They write2:

Invocation of supernatural agents implicates two cognitive aspects of religious belief: (1) activation of naturally selected conceptual modules, and (2) failed assignment to universal categories of ordinary ontology. Conceptual modules are activated by stimuli that fall into a few intuitive knowledge domains, including: folkmechanics [folk physics] (object boundaries and movements), folkbiology (species configurations and relationships), and folkpsychology (interactive and goal-directed behavior). (p. 720)

Examples they give include: "ghosts that walk through walls," "frogs that talk," and "mountains that are invisible to the human eye"3, or more familiarly, wafers and wine that are the body and blood of Christ. Ghosts walking through walls violates our folk physics, frogs talking violates our folk biology (nonhuman animals don't talk, unless they're parrots, and then we're totally fascinated by them), and invisible mountains our folk physics again. Bread and wine that are the body and blood of Christ violate our folk physics, our folk biology (biological kinds can't radically change their essence and change into something entirely different, like bread and wine), and our intuitions about animate and inanimate objects.

Schematic Memory

Now we get to memory. Contrary to common belief, long-term memory doesn't involve simply retrieving information holistically, and spitting it out the same way it came in. Instead, memory is largely reconstructive, in the sense that when we retrieve something from it, we do so piecemeal and put it together as we're doing so. The tools we use for reconstructing memories are called schemas. Instead of writing a new description of what schemas are, and what we know about how the affect reconstructive memory, I'm just going to quote myself (yes, I'm that lazy):

One way of thinking about schemas is as variables with relationships between them. Schemas will tend to be stored with default values for the variables, which serve as expectations For example, our restaurant schema contains a variable for the time when the server will bring the check, and the default value for that variable will be sometime between two other variables, which stand for the time we finish eating and the time we leave. Research has shown that how new information fits with our schemas will determine how well it is remembered, and how easy it is to comprehend and transmit. Information that is relevant to our existing schemas, i.e. information that fits into one of the variable slots, will be more easily remembered. There are two types of schema-relevant information. The first, schema-consistent information, is information that fits with our expectations. When the server brings us our check between our finishing and leaving, this is a schema-consistent instance. In these cases, we tend not to store particulars about the situation, but instead store the instance with the schemas default values. This allows us to reason about the situation using the other default values in the schema, but it also creates situations in which our memories are distorted. When reconstructing the situation from memory, later, we may include default values from the schema that were not present in the actual instance.

The second type of information that is easily remembered is schema-inconsistent information. This type involves values on the same dimensions as the schema's variables, but that differ significantly from the default values. For example, if our server brought us the check at the beginning of the meal, the information would fit into a slot in our schema (the time that the check is brought) but would be very different from the default value. When we encounter this type of information, it demands our attention, and we tend to store it in detail. Thus we will remember particulars about the instance, and are less likely to experience memory distortions. A third type of information, which is schema-irrelevant, does not line up with any of the slots in our schema. This type of information tends to be ignored and forgotten.

So, there are three types of information: schema-inconsistent, schema-consistent, and schema-irrelevant, and we tend remember the first type best and the last type worst. This led Atran and Norenzayan to predict that religious narratives that are easily remembered will include a lot of schema-relevant information, but in order to be remembered well, some of that information will be schema-inconsistent. In their terminology, successful religious narratives (which all contain some counterintuitive beliefs) will avoid radically counterintuitive beliefs, which don't connect to any of our intuitive schemas (from folk physics, folk biology, etc.), and would thus be treated as schema-irrelevant, and ignored. They will, however, contain minimally counterintuitive beliefs which connect to, but deviate from our intuitive schemas, and are thus treated as schema-inconsistent information, and easily remembered.

Two Studies

To test this aspect of their theory, Norenzayan et al.4 first constructed lists of intuitive and minimally counterintuitive phrases (see the table below for examples of both)


Each list varied in the proportion of intuitive and minimally counterintuitive phrases. Each participant first studied a list for five minutes, and then after a three minute delay they were asked to recall as many phrases as they could. Then, after a one week delay, they were given a "surprise recall task" for the phrases. Norenzayan and Atran predicted that, after the short delay (3 minutes), participants would show better recall for the intuitive phrases, but after the long delay (1 week), they would be able to recall more minimally counterintuitive phrases than intuitive ones. They also predicted that after the 1 week delay, memory for entire lists would be better when most of the phrases were minimally counterintuitive. All three predictions were confirmed, with higher recall scores for intuitive that minimally counterintuitive phrases after the three minute delay, and the reverse after the one week delay. Recall for entire lists was also better after one week for the lists that contained mostly minimally counterintuitive phrases. here are their graphs (from p. 539 and 540):


Their second study was a bit different. I'll let them describe it:

Our study focused on one of the most culturally important folktales in the Western tradition-- collected by the Brothers Grimm. We empirically assessed the cultural success of each folktale, and selected 42 tales for deeper analysis; 21 were relatively successful, and 21 were demonstrably less successful. Two trained raters, unaware of our hypotheses, read each tale and counted the number of counterintuitive elements in each folktale. In addition, 65 university students read these folktales and rated them on a number of characteristics, including memorability and ease of transmission. These procedures allowed us not only to test the hypothesis that MCI tales are more likely to be culturally successful, but also to examine whether perceived memorability mediates the relation between minimal counterintuitiveness and cultural success. (p. 542)

In this case, the prediction is that the successful folktales will contain some counterintuitive elements, but not too many, and that the stories with minimally counterintuitive elements would be rated as more memorable. And that's what they found. Most of the successful stories had 2-3 counterintuitive elements, while the uncsuccessful stories had either more than three or fewer than two counterintuitive elements. Stories with 2-3 counterintuitive elements were also rated as more memorable than those with either more or fewer. Using regression analysis, they were able to show that the relationship between minimally counterintuitive stories (stories with 2-3 counterintuitive elements) and the success of the stories was mediated by the stories' memorability, consistent with the hypothesis that minimally counterintuitive narratives are more successful because minimally counterintuitive information is more memorable.


One thing to take home from this theory, and the two studies that support it, in the context of the recent discussion of religion on this blog and elsewhere is that successful religious narratives, far from being "out there," are actually pretty mundane, though not so mundane that they're too consistent with our ontological intuitions and thus difficult to remember. Within Atran and Norenzayan's theory, it's not surprising that God concepts exhibit so many anthropomorphic features, for example. These features serve to both connect God concepts to our intuitions about agents, as well as to distinguish God concepts from those intuitions to make them more memorable. Given what we know about ontological intuition and memory, then, religious beliefs tend to be rather sane.

1Norenzayan, A., Atran, S., Faulkner, J., & Schaller, M. (2006). Memory and mystery: The cultural selection of minimally counterintuitive narratives. Cognitive Science, 30, 531-553.
2Atran, S., & Norenzayan, A. (2004). Religion's evolutionary landscape: Counterintuition, commitment, compassion, communion. Behavioral and Brain Sciences, 27, 713-770.
3Norenzayan et al. (2006).


More like this

Well, given your earlier post on memory and cursing, maybe we should liberally sprinkle our religious myths with some obscenities...

Praised be the holy Fucking Spaghetti Monster!

Colin, thanks. I wish I knew enough about Mormonism to apply them. I suspect that with Mormonism, the counterintuitiveness of a narrative can be greater, because it can piggyback on the cultural schemas that exist because of the prominence of an earlier, less counterintuitive religion (Christianity). But given how little I know about Mormonism, I could just be talkin' out my bum.

Kurt, lol... I thought about that, but chose not to say it.

The remembering of counterintuitive tales sits well with information theory. The entropy or information content is the sum of inverse log-likelihoods. Counterintuitive concepts therefore score high on entropy. The minimal constraint may have resulted simply from the 'memory cost' of storing that information.
Thanks for an interesting read!

Being a lay person I may be way off the mark on this but, It seems to me that the test results may simply speak to our tendency to recognize patterns and detect the anomolies. We tend to make note of the unusual and discard much of the rest. In light of that, that the intuitive statement would need to be notable or extreme in their own right, albeit still intuitive to combat this.

In other word I suspect intuitiveness is largly irrelevant. "Confused students" are a common sight, "Four-legged students" are not, the visualization is more memorable given it's oddity. Seems to me that if the MCI statement were "Legless student", which is entirely plausible, though still odd, you may get the same results.