TL;DR: Superstition never disappears because it is a by-product of an effective learning system. Two complementary explanations: chance coincidences get reinforced into rituals (Skinner 1948), and when the cost of missing a real causal link exceeds the cost of a false belief, the habit of broad causal attribution is retained (Foster and Kokko 2009). The evolutionary model is theoretical and has not been directly measured in humans.
Why does superstition still exist? Two complementary explanations
Superstition does not disappear because it is not a random error of the mind, but an understandable by-product of the very way we learn. Two lines of scientific explanation, rather than excluding each other, complement each other.
The first line comes from behavioural psychology. In a famous 1948 experiment, B. F. Skinner fed pigeons at fixed intervals, entirely independent of what the bird was doing, and found that they still developed repeated movements as if those movements called the food forth. Skinner treated this as a model for superstitious behaviour: a chance coincidence reinforced into a ritual[1].
The second line comes from evolutionary biology. Kevin R. Foster and Hanna Kokko (2009) built a mathematical model showing that when the cost of missing a real causal relationship is higher than the cost of falsely believing in one that does not exist, natural selection will favour strategies that assign causality broadly[2]. In other words, the habit of assigning causality is retained because it is often advantageous, even if it sometimes produces superstition.
Both lines answer the question "why does it exist", not the question "is it true". The fact that superstition is an understandable by-product of the learning machinery does not say that superstitious beliefs are true.
What was Skinner's pigeon experiment?
What did Skinner do?
In an experiment on operant conditioning, Skinner fed pigeons at regular intervals, regardless of what the bird was doing. He observed that the birds developed repeated, ritual-like movements: one turned its head toward a corner of the cage, another pecked its head at a particular spot, as if that behaviour were the cause of the food appearing[1].
Skinner described this phenomenon with the sentence: the bird behaves as if there were a causal relationship between its behaviour and the appearance of food, when in fact that relationship does not exist[2]. The important point in Skinner's approach is that he focused on behaviour (the external result) rather than on a mental representation of causality (the internal process)[2].
The mechanism: random reinforcement gives rise to ritual
The mechanism is very simple. When food appears at random, whatever behaviour the pigeon happens to be performing at that moment is incidentally "rewarded". Next time, the pigeon tends to repeat that behaviour. Over many repetitions, the behaviour is reinforced into a stable ritual, even though it has no real effect.
Limits when extrapolating to humans
It must be said clearly: Skinner's results on pigeons do not automatically carry over to humans. Later researchers (Staddon and Simmelhag 1971; Timberlake and Lucas 1985; Moore 2004) reinterpreted those behaviours as expressions of a food-seeking instinct, similar to Pavlov's dogs salivating, rather than as superstition in the exact sense Skinner meant[2]. This means the pigeon experiment is a useful analogy, not direct evidence about superstition in humans.
What is Foster and Kokko's evolutionary model?
The core idea
Foster and Kokko (2009) asked: under what conditions is a tendency to assign causality falsely advantageous from an evolutionary standpoint? They built a mathematical model based on optimality theory, in which an individual must decide whether to respond to an event, based on the probability it estimates[2].
The model shows that when an individual cannot fully separate events that are genuinely causally related from events that are merely coincidental, it is forced to respond to a mixed group containing both kinds, or not to respond at all. In that situation, natural selection may favour a strategy that lumps non-causal relationships together with causal ones[2].
Table: two kinds of error and the cost
| Type of error | Content | Cost | Evolutionary consequence |
|---|---|---|---|
| Type I (false positive) | Believing there is a causal relationship when in fact there is none | Wasting effort on a useless behaviour | Low if the behaviour is not costly |
| Type II (false negative) | Ignoring a real causal relationship | Missing a survival opportunity (food, avoiding a predator) | High if the miss is dangerous |
When the cost of a Type II error is higher than the cost of a Type I error, natural selection favours a strategy that makes more Type I errors. That is the condition under which superstition can be an adaptive trait[2].
Conditions for superstition to appear
According to the model, false responses become most common when the probability that two events are genuinely related is low to moderate. Very strong relationships are rarely responded to falsely, while very weak relationships are rarely favoured by natural selection. When the model is extended to many events at once, the authors identify the conditions under which natural selection may favour assigning causality to events that have never had a real causal relationship[2].
Foster and Kokko conclude that superstitious behaviours, or behaviours that look superstitious, are an inevitable feature of adaptive behaviour in all organisms, including humans[2].
Limits of the model
This is a theoretical model, not yet measured directly in humans. The model demonstrates that under certain conditions, natural selection may favour a broad habit of assigning causality, but it does not measure the extent or the specific form of superstition in humans. The authors also note that their model does not take cultural influence into account, although they acknowledge that culture plays an important role in shaping the specific content of superstition[2].
Why is superstition a by-product of a fast learning strategy?
Learning from little data
Jan Beck and Wolfgang Forstmeier (2007) offer another line of explanation, complementing Foster and Kokko's model. They argue that an effective learning strategy, drawing rules from very little data, inevitably produces a share of false beliefs such as superstition[3].
On this argument, superstition is a by-product of the capacity to learn, without any need to assume that it is itself advantageous. This is a familiar "by-product" argument in evolutionary biology: a trait exists not because it was directly selected for, but because it comes along with another trait that is advantageous.
Why does fast learning give rise to superstition?
If an organism learned only from complete and accurate data, it would be slow and would miss opportunities. In nature, recognising a rule from a few observations can be the difference between life and death. But precisely because it learns from little data, an organism easily draws rules from chance coincidences. Superstition is the price paid for speed of learning.
Note: this is a theoretical paper, not an empirical study measuring superstition directly in humans. Its result is an argument on evolutionary logic, not yet replicated by independent empirical studies.
Are people who see patterns in random sequences more likely to believe in the supernatural?
The study by van Prooijen and colleagues
Jan-Willem van Prooijen and colleagues (2018) published a series of five studies in the European Journal of Social Psychology, showing a link between illusory pattern perception and belief in conspiracy theories as well as supernatural phenomena[4].
In these studies, people who saw patterns in random coin-toss sequences or in abstract pictures tended to believe in conspiracy theories and the supernatural more. The experiments further suggest that priming participants to perceive patterns can increase that belief[4].
Scope of the data
The scope must be stated clearly: these studies were conducted abroad, mainly on English-speaking adults, not on Vietnamese people. The results show a correlation and an effect under experimental conditions, which does not mean that all Vietnamese people or all cultures share the same level. Drawing conclusions for Vietnamese people would require research on Vietnamese people[4].
Connection to the general mechanism
This result fits with both lines of explanation above. If our cognitive machinery is designed to find patterns, then finding patterns even where there are none is easy to understand. Those with a stronger tendency of this kind will believe in the supernatural more. This is a general mechanism of all people, not a peculiar trait of any particular group.
Why is superstition present in every culture?
Not because it is true, but because it is a by-product
The central question of this article is: if superstition is false, why is it not eliminated but present in every culture? The answer from the two lines of explanation above is: superstition exists because it is a by-product of an effective learning machinery, and because under certain conditions, a broad habit of assigning causality is more advantageous than missing a real causal relationship.
This explains why superstition is present in every culture, even though its specific content differs. The mechanism is shared, while the form is shaped by culture. Foster and Kokko acknowledge clearly that their model does not take culture into account, but culture plays an important role in shaping the specific content of superstition[2].
Distinguishing "exists" from "true"
Two things must be clearly distinguished. First: superstition exists and is widespread, and we can understand why it exists from an evolutionary and psychological standpoint. Second: its existence does not say that the content of superstition is true. The fact that a belief has an understandable evolutionary origin does not make that belief an accurate description of the world.
Real effects arising from belief
There is one point that must be stated correctly. Sometimes a belief changes behaviour, and behaviour changes outcomes. For example, a person who believes that a certain ritual helps them feel more confident may genuinely feel more confident, and therefore perform better. That effect is real, but it is not evidence that the ritual has a supernatural effect. It is the psychological effect of belief on behaviour. To speak correctly about it is to say that belief changes behaviour, behaviour changes outcomes, without calling it a case of the ritual guessing correctly, and without denying the effect.
Where is the evidence still lacking?
Foster and Kokko's evolutionary model is a mathematical theory, not yet measured directly in humans. It shows under what conditions natural selection may favour a broad habit of assigning causality, but it does not measure the extent or the specific form of superstition in humans. Beck and Forstmeier's argument is likewise a theoretical paper, not yet replicated by independent empirical studies.
The studies by van Prooijen and colleagues (2018) are empirical evidence for the link between illusory pattern perception and supernatural belief, but they were conducted abroad, mainly on English-speaking adults, not on Vietnamese people[4].
In addition, researchers after Skinner have reinterpreted the pigeon experiment as an expression of a food-seeking instinct, rather than as superstition[2]. This means that evidence from animals cannot be carried straight over to humans.
Summary table of the main sources
| Study | Author, year | Method | Scope of data | Main result |
|---|---|---|---|---|
| Pigeon experiment | Skinner, 1948 | Experiment on pigeons, fed at fixed intervals | Pigeons in the laboratory | Birds developed repeated, ritual-like movements[1] |
| Evolutionary model | Foster and Kokko, 2009 | Mathematical model of optimality theory | No empirical data on humans | When missing costs more than false belief, selection favours broad assignment of causality[2] |
| By-product of learning | Beck and Forstmeier, 2007 | Theoretical paper | No empirical data | Superstition is an inevitable by-product of learning from little data[3] |
| Illusory pattern perception | van Prooijen and colleagues, 2018 | Five studies, experimental and correlational | Adults, mainly English-speaking | Seeing patterns in random sequences predicts supernatural belief[4] |
