Why do we think some things are related when they aren’t?
Illusory Correlation
, explained.What is illusory correlation?
Illusory correlation, also known as illusory correlation bias, is the tendency to perceive a relationship between two variables when none actually exists. This cognitive bias often arises because the association aligns with our expectations or is amplified by the distinctiveness of certain events, making them more memorable and salient in our minds.
Where illusory correlation is seen
Consider the following hypothetical situation: Jane is an avid football fan and watches every game that her beloved “Guardians” play on live television. Jane always wears her tattered Guardians jersey while watching their games— the same one she has worn for years.
According to Jane, it is vital that she wear her “lucky jersey.” The success of her team depends on it. A few years earlier, Jane had noticed that when she didn’t wear her jersey, the Guardians lost. This phenomenon happened a few times before Jane’s superstitions were solidified. Now, she is certain: the success of her favorite football team is in some way related to her wearing this jersey.
Jane’s false perception that wearing her jersey at home is directly related to the performance of her favorite football team can be attributed to an illusory correlation.
Illusory correlation is a cognitive bias that is closely linked to memory and perception, as people tend to notice and remember patterns or associations that confirm their expectations or stand out in some way. Sometimes, our decisions hinge on the relationship between various phenomena—for example, when I do “Y,” I know “X” will occur. We see that certain events consistently occur at the same time as, or just after, other events, leading us to conclude that they are somehow related. We may not know why they are related, but we have enough evidence to suggest that their occurrence is linked.
Mis- and disinformation can also exacerbate illusory correlations and lead to their widespread acceptance. If the public learns about a flawed correlation between two events or factors from an individual with authority, they are more likely to accept it as truth. In 2006, J. D. Hayworth, one of America’s most outspoken Congressmen, published a book called Whatever it Takes: Illegal Immigration, Border Security, and the War on Terror, in which he claimed that inflows of people over the U.S.-Mexico border posed a “terrorist threat.” Juxtaposing undocumented immigration with terrorism produces an illusory correlation because the majority of migrants entering the U.S. from the South are people in search of better career opportunities, not terrorists with ulterior motives.18
We can’t always know if and why two things are causally related. Sometimes, it’s enough to know that they are somehow related. If we see this relationship occur many times, we become confident that the correlation will reliably recur in the future.
Related Biases
Individual effects
One of the most common manifestations of illusory correlation is the formation of stereotypes. People may associate particular traits, behaviors, or characteristics with specific groups without sufficient evidence to support such generalizations, leading to prejudice and judgment. For example, a person might believe that a certain group is more likely to engage in negative behaviors simply because those behaviors are more memorable or noticeable when they occur within that group. This, in turn, can lead to reduced empathy among individuals and unnecessary conflict.
Illusory correlation also plays a role in our decision-making processes. For instance, people may mistakenly believe that certain actions or rituals have a direct influence on outcomes, such as Jane wearing her "lucky" item to ensure her team’s success. This can lead to the development of superstitions or misconceptions about cause and effect. We often take risks based on our confidence in the correlation between different events or behaviors.1 For example, the time of day is correlated with how heavy or light the traffic will be. Based on our confidence in this correlation, we might choose not to take the risk of driving during rush hour to make it on time to an appointment. Because we often make risk calculations and decisions while noticing the occurrence and recurrence of correlations, it is important that they are accurate. Taking stock of illusory correlations jeopardizes this, causing us to make poor decisions and take unwise risks.
We are also influenced by illusory correlation when making judgments about rare or distinctive events. Since uncommon occurrences tend to stand out and be more memorable, individuals may overemphasize their connection to other events. Many people in Mexico City, for example, believe that September is ‘earthquake month’ simply because three major earthquakes have historically occurred on the exact same day in this month. The first devastating earthquake happened on 19 September 1985, claiming the lives of many and injuring others. 32 years later, on exactly the same day, another deadly 7.1 magnitude earthquake struck just hours after an earthquake drill to commemorate the 1985 event. On 19 September 2022, a further earthquake struck the western state of Michoacán, sending tremors through the capital once again. The mathematical probability of these three earthquakes striking on exactly the same day is between 0.000751% and 0.00000024%, leading people to believe that there’s a correlation between the month of September and the occurrence of large earthquakes.13 Scientists, however, have assured us that there's no empirical evidence that tectonic plates schedule their activity according to human calendars.
Systemic effects
Illusory correlation can have damaging implications and exacerbate systemic biases in several situations. Decisions made at an institutional level are usually informed by correlations drawn from data or observations, and false correlations can motivate biased institutional policy. For example, illusory correlations contribute to stereotypes and institutional racism. The public disproportionately notices violence perpetrated by minority groups and connects this violence with certain races or ethnicities despite no such correlation existing.2 Public officials and social debates have often connected immigrants to crime, and as a result, calls for harsh immigration restrictions and stricter deportation methods have stoked further racial tensions. However, a study released in 2013 found no evidence of a relationship between criminality and immigration in England and Wales.19 In fact, another report published in 2017 suggested that there could be a relationship between waves of immigration and falling crime rates.20
Within the workplace, illusory correlation can cause decision-making errors during the hiring process. A common assumption is that candidates with degrees from top-tier universities will automatically perform better than candidates from lesser-known institutions or alternative employment paths. If a hiring manager onboards many top-tier university candidates and they go on to perform well in the company, the strength of this correlation will grow, resulting in an illusory correlation between university and workplace performance. However, in reality, college performance does not always directly translate into job performance,15 meaning that employers blinded by illusory correlation in this area may be missing out on excellent talent simply because they don’t have the exact qualifications on paper.
Why it happens
There are two types of illusory correlations: expectancy-based and distinctiveness-based illusory correlations. The former occurs when we mistakenly see relationships due to our preexisting expectations surrounding them. The latter happens when a relationship is believed to exist between two variables because we focus too much on information that stands out.4
Both of these illusory correlations can be attributed to our brain’s use of “heuristics” or mental shortcuts. Evaluating evidence takes time and energy, and so our brain looks for such shortcuts to make the process more efficient.
The availability heuristic
This is our tendency to use information that comes to mind quickly and easily when making decisions about the future. As a result of the availability heuristic, variable pairings that come to mind easily (either because they appear, because they are quick to grasp, or because they seem likely), are seen as correlated.5 For example, ice cream and gluten intolerance mentioned together frequently, we might think they are correlated when they aren’t. This is because this pairing is more distinct than others, and will more readily come to mind when we look for correlations than pairings we haven’t seen before.
Confirmation bias
As another cognitive shortcut, confirmation bias, occurs when we notice, focus on, and give greater credence to evidence that fits with our existing beliefs. Confirmation bias has been linked to illusory correlation, as we look for relations that confirm our preexisting beliefs surrounding two variables. For example, if we believe flying is dangerous, we are more likely to expect correlations between increased flying and deaths related to transport.
Information processing
Martin Hilbert, a Professor at the University of California, proposes an information processing mechanism that explains how the brain’s limited ability to process and store information leads us to notice and believe in patterns that don’t exist. He suggests that the way we process information involves some "noise," meaning that when we recall objective facts from memory, they can get mixed up or distorted, resulting in subjective judgments that aren’t necessarily accurate.
Research shows that illusory correlations are subject to social amplification; that is, they can start as very small misperceptions but grow into widespread biases as they are shared among different people. In a study conducted in Portugal, researchers investigated illusory correlation using the serial reproduction paradigm, a method used in psychology to study how information changes as it is passed from person to person.14 Participants learned about members of certain groups, the behaviors that they performed, and the attributes that they possessed. They were then asked to recall this information and pass it on to other participants; any memory bias they produced was incorporated into the information they passed on to others. The illusory correlation between certain groups and certain attributes/behaviors was weak among the first participants, but this pattern grew stronger as information was communicated across participants in the chain. This finding is very important for understanding how incorrect information and stereotypes travel through different people, leading to widespread mis and disinformation about certain groups of people.
behavior change 101
Start your behavior change journey at the right place
Why it is important
Illusory correlations can cause us to make misguided and risky decisions. As mentioned above, when we make decisions surrounding a particular phenomenon, we look at its relation to other variables and make risk calculations based on how sure we are of those relations holding in the future. For instance, when a regulatory board looks at the dangers surrounding a new industrial chemical, they may look at whether rates of cancer have gone up in areas where it is most used. Based on the certainty of this correlation, they may or may not restrict this chemical. Illusory correlations can therefore impact the effectiveness of our decisions and how much risk we are willing to accept when making them, which can sometimes have disastrous consequences.
The spread of misinformation and disinformation related to illusory correlations can also have serious implications for public health and risk management during health crises. In 1998, British gastroenterologist Andrew Wakefield published a study in The Lancet suggesting a link between the MMR (measles, mumps, and rubella) vaccine and autism. Although the study only used a small sample of 12 children and flawed methodology, it was widely shared by the media and caused fear among the public. In the UK, MMR vaccination rates fell from 92% in 1996 to 80% by 2003, contributing to a resurgence of measles.17
Autism diagnoses often occur around the same time children receive vaccines (12–24 months old), leading parents to incorrectly infer a causal relationship. Since Wakefield’s paper, numerous studies have debunked the relationship between childhood vaccines and autism, however, myths and misinformation continue to circulate through word of mouth and on social media.16
Illusory correlation can also make us blind to correlations that really do exist. If we are focused on illusory correlations because we believe they are indeed true, we are less likely to look for other correlations that might actually be present. The illusory correlation that associates immigration with criminal activity, for example, can divert attention away from other important factors that contribute to higher crime, such as socioeconomic inequalities or funding cuts to public services. This can lead to missed opportunities and false conclusions about what should be prioritized in government policies and strategies.
How to avoid it
We should try to reduce illusory correlation where possible because of the damaging effects it can have across several contexts. A 2011 study found that illusory correlations can be reduced by understanding under which conditions our minds tend to misconceive relations. The researchers found that illusory correlations tend to occur “under conditions in which the participant is not personally involved.” In other words, we see false correlations in areas and circumstances in which we have little knowledge or personal experience.
As such, the authors concluded that “developing evidence-based educational programs should be effective in helping people detect and reduce their own illusions.”7 Because we are particularly susceptible to illusory correlations in unfamiliar areas due to our lack of experience in them, we can reduce the bias by becoming more informed in those areas.
"Illusions can be predicted (and reduced) if we understand the conditions in which our cognitive system tends to err in associating causes to effects and in assessing contingent relations."
– Psychology professor Helena Matute, et al.
How it all started
Humans have been linking events and strange behaviors to other phenomena for centuries. In the 1700s, people believed that a full moon could cause epilepsy or feverish temperatures due to a significant clustering of seizures around the full moon period.21 In 1842, England even introduced the Lunacy Act, which distinguished between behavior that was deemed normal two weeks before the full moon by abnormal two weeks after.22
Early observations by philosophers such as David Hume in the field of human reasoning noted that we have a tendency to perceive cause-and-effect relationships where none exist. One of Hume’s greatest contributions to what would later be known as illusory correlation was his assertion that experience cannot tell us much about causation. When observing two events, A and B, we claim that A causes B if they consistently occur together, meaning they are always linked. Whenever A is present, B is also observed, and we develop confidence that this pattern will persist. However, recognizing that the statement “A causes B” essentially translates to “we are psychologically convinced B follows A because of their consistent pairing” reveals a fragile concept of necessity (a condition or state in which something must occur or be true). This weak foundation for causal understanding contributes to the “Problem of Induction,” which questions whether we can reasonably justify any inductive conclusions about the world.
It wasn’t until 1967 that the term “illusory correlation” was coined by US psychologists Loren J. Chapman and Jean Chapman.8 They conducted various experiments in which subjects were told to draw a person according to their own diagnostic category. Chapman and Chapman found that there were semantic associations between diagnoses and certain features that were drawn. For instance, patients who worried about their intelligence would emphasize the head in their drawings.9
Illusory correlations were first empirically demonstrated by David Hamilton and Robert Gifford in 1976. They had participants read a series of sentences describing either desirable or undesirable behaviors, which were then attributed to the members of two different groups: locals and immigrants. Even though these behaviors were actually associated with both of these groups, participants overestimated the association between undesirable behaviors and the minority group (ie. immigrants). This illustrated the role of illusory correlation in stereotyping.10
How it affects product
In the world of product marketing, illusory correlations are rife. This is due, in part, to clever advertising that often links certain product characteristics and attributes together, leading us to believe that they are indeed related. Take, for instance, the cost of clothing and how long they will last. With the rise in fast fashion, we are often told that the more we pay for clothing, the longer it will last. However, research has shown that there’s not necessarily a correlation between price and quality. In fact, researchers at the University of Leeds’ School of Design in the UK found that the retail price of a new garment cannot be used as a reliable indicator to identify good or inferior durability.26 The durability of both high and low-priced garments ranged from excellent to very poor, debunking the belief that higher priced clothing is automatically better quality.
So how does this affect your average high street customer? If we possess an illusory correlation between price point and quality when we’re shopping for new clothes, electronics, or other products, we’re more likely to choose items based solely on how much they cost without taking the time to inspect their actual quality.
Our loyalty towards certain brands and products can also be influenced by illusory correlations and false beliefs about their performance and effects. For example, a skincare product that coincides with a user’s clear skin improvements may be perceived as the cause, even if other factors (like diet or a change in environment) were actually responsible. This may lead to people continuing to purchase the item based on the misattributed benefit when another product may actually be better suited to them.
Illusory correlation and AI
Like humans, artificial intelligence systems can sometimes identify patterns or relationships in data that do not actually exist. This issue can arise due to biases in the training data, algorithmic limitations, or how machine learning models interpret and process information. For example, if the dataset used to train an AI model contains spurious correlations or biased representations, the system may learn and reinforce these false patterns. Increasingly, AI is used in predictive policing to provide law enforcement teams with data about how to use resources efficiently and objectively, but some organizations have warned that these tools should be evaluated and regulated to avoid racial biases. Illusory correlations in these predictive models, they argue, can result in the disproportionate surveillance and policing of Black communities23, which, in turn, leads to more arrests and further reinforcement of the false correlation in the data.
One way to mitigate illusory correlation in AI is to develop Explainable AI (XAI). These models help to characterize model accuracy, fairness, transparency, and outcomes in AI-powered decision-making and can help identify and address false correlations.24 As AI technology advances, it becomes increasingly difficult for humans to understand and trace the algorithm's steps to reach a conclusion. This lack of transparency is often described as a "black box," where the inner workings of the model are opaque and hard to interpret. These black box models are built directly from the data, and even the engineers or data scientists responsible for designing them may be unable to fully explain how the algorithm processes information or arrives at a particular outcome.
As AI models develop, humans may actually be able to help detect and remove biases from their algorithms. A study conducted by researchers in France found that individuals who design and train artificial intelligence systems are well aware of the impact of their cognitive biases, including illusory correlation, on their productions.25 What is still not clear is how these biases are embodied or not in the work of these individuals who are responsible for designing AI systems. However, if AI designers are aware of their biases and how they impact their work, there is the potential for reducing their influence early on.
Example 1 - Looking for patterns in finance
Financial analysts and investors are often subject to illusory correlation. A 2011 paper by American business researchers looked into the certain stock price patterns and the future trends they are often associated with. More specifically, they examined the “head-and-shoulders” pattern (three price peaks, the highest of which is in the middle) that is said to predict a future downtrend in stock price.
In reality, the researchers claim that this pattern does not accurately predict any future price movements. The upshot is that “ the correlation between the signal [the head-and-shoulders pattern] and future U.S. equity price movements asserted by technical analysts does not exist.” They therefore claim it is an illusory correlation that impacts trading significantly, as “unusual trading upon the completion of a head-and-shoulders pattern averages over 40 percent of a day’s trading volume.”11 So, because investors often associate certain price patterns with future price trends, when no such association exists, illusory correlation causes significant sub-optimal decision-making in the financial world.
"…the human predilection to discover patterns, augmented by a strong desire to make money, leads some investors to believe in connections between price patterns and future price movements that do not truly exist."
- Professor of international business Carol L. Osler, et al.
Example 2 - CEO compensation and golf
In a unique study done by European business researchers in 2010, found that perceived golfing performance was positively related to CEO compensation. This is to say, CEOs who play golf well make more money than those who don’t— average nearly 15% more, they find.
This is because golf performance is taken to be a cue for corporate performance. People associate a CEO’s record on the golf course with how good they are at their job. However, the researchers also find that GEO golf performance is not positively correlated to the corporate performance. It is actually negatively correlated. So, the association between gold performance and corporate performance is actually an illusory correlation.
But why is golf so important? The researchers speculate that CEO remuneration decisions involve various tangible measures of past performance as well as observations of “soft” social skills such as golfing. Golf clubs are also settings in which other relevant business actors socialize. Being a good golfer therefore carries into the business world as a sort of “halo effect.”12
Summary
What illusory correlation is
Illusory correlation, or illusory correlation bias, occurs when we see a relationship between variables (events, actions, ideas, etc.) when they aren’t actually associated.
Why illusory correlation happens
There are two sorts of illusory correlations: expectancy-based and distinctiveness-based illusory correlations. The former occurs when we mistakenly see relationships due to our preexisting expectations surrounding them. The latter happens when a relationship is believed to exist between two variables due to focusing too much on information that stands out.
Both of these illusory correlations can be attributed to our brain’s use of “heuristics” or mental shortcuts. As a result of the availability heuristic, variable pairings that come to mind easily (either because they appear often or because they are quick to grasp), are seen as correlated. This is because this pairing is more distinct than others, and will more readily come to mind when we look for correlations than pairings we haven’t seen before. Confirmation bias has been linked to illusory correlation, as we look for relations that confirm our preexisting beliefs surrounding two variables.
Example #1 – Looking for patterns in finance
Financial analysts and investors are often subject to illusory correlation. A 2011 paper found that a certain stock price pattern is associated with a future downtrend in stock price. In reality, the researchers claim that this pattern does not accurately predict any future price movements. They therefore claim it is an illusory correlation that impacts trading significantly, as “unusual trading upon the completion of a head-and-shoulders pattern averages over 40 percent of a day’s trading volume.” Because investors often associate certain price patterns with future price trends, when no such association exists, illusory correlation causes significant sub-optimal decision making in the financial world.
Example #2 – CEO compensation and golf
A 2010 study found that CEOs who play golf well make more money than those who don’t— average nearly 15% more, they find. This is because golf performance is taken to be a cue for corporate performance, even though they find a negative correlation between the two. The association between golf performance and corporate performance is actually an illusory correlation. The researchers speculate that this is because CEO remuneration decisions involve observations of “soft” social skills such as golfing, and golf clubs are also settings in which other relevant business actors socialize.
How to avoid illusory correlation
A 2011 study found that illusory correlations can be reduced by understanding under what conditions our minds tend to misperceive relations. We tend to see false correlations in areas and circumstances that we have little knowledge or personal experience in. The researchers therefore concluded that “developing evidence‐based educational programmes should be effective in helping people detect and reduce their own illusions.”6 We may be able to combat the illusory correlations that we are particularly susceptible to in areas we aren’t familiar with by becoming more informed in those areas.
Related TDL articles
Why do we believe misinformation more easily when it’s repeated many times?
Similar to illusory correlation, humans are also susceptible to the illusory truth effect, which describes how when we hear the same false information many times, we believe it’s true. This article explores this bias and its impact on individuals and systems and provides some useful tips for avoiding it.
Why do we think we understand the world more than we actually do?
Here’s another illusion that humans often fall victim to: the illusion of explanatory depth, or IOED. We think we know more about the world around us than we actually do - we’ve all sat next to that person at dinner who is an expert in everything. In this article, we look at why we think we’re know-it-alls and how it impacts our everyday lives.