Every so often, I get random chest pain and immediately spiral – is this it? Am I having a heart attack? Then I remind myself that I’m 25, probably overcaffeinated, and maybe just a little stressed.
This type of irrational leap (thankfully) isn’t unique to just me. In fact, doctors are trained to avoid jumping to rare diagnoses simply because one symptom aligns. There’s even a saying that’s often taught during residency training:
“When you hear hoofbeats, think horses, not zebras.”
In other words, the most common explanation is usually the right one. For someone my age, chest pain is far more likely to be from anxiety or too much coffee, not a heart attack. So, maybe I should just chill and skip that third cup of coffee today…
The tendency to make irrational judgments in situations like these is related to the Inverse Base Rate Effect (IBRE), and there’s a whole area of research dedicated to understanding why we sometimes assume rare outcomes, even when we rationally know they’re unlikely.
Ideally, we should base our judgments on the “base rate,” which is the overall probability that something will occur; however, in practice, we don’t always do that. One well-established explanation for this type of irrational reasoning focuses on how we allocate our attention to specific features or symptoms.
For example, suppose someone learns the combination of symptoms A and B typically predicts Disease 1, which is relatively common, while symptoms A and C together usually predict Disease 2, which is much rarer. When people encounter symptoms A and C together, they often make a mistake and assume it’s Disease 1 instead of Disease 2 because they’ve generalized their learning about Symptom A, given its commonality. Once they realize their mistake, they begin to pay close attention to symptom C to avoid making the same mistake and misclassifying the disease again in the future.
Ironically, this increased attention to symptom C can lead to a whole new error in the future. Now, when symptoms C and B are observed together, people often jump to the conclusion that it must be Disease 2 (the rarer disease), because they were paying so much attention to Symptom C due to their prior error. In doing so, they ignore symptom B, which predicted a higher likelihood of Disease 1. This shift in attention, which causes people to irrationally select the rare outcome, is what defines the IBRE.
Two well-established models help explain how this attention allocation contributes to the IBRE: EXIT (Exemplar-based attention to distinctive InpuT) and NNARS (the Neural Network with Rapid Attentional Shifts). Interestingly, both models predict that distraction may reduce the IBRE (i.e., people’s tendency to behave irrationally) since it interferes with the attention-shifting process that might be driving the error in the first place.
In a recent study published in Psychonomic Bulletin & Review, researchers Lenard Dome and Andy J. Wills (pictured below) put this hypothesis about distraction (predicted by these two separate models) to the test.
To test this, the researchers first had participants undergo training where they learned how pairs of symptoms (e.g., skin rash, stuffy nose) predict a particular disease. Some of these pairings occurred more frequently, while others were much rarer. For example, as depicted in the figure below, symptoms A and B and D and E were common pairings, occurring 3 times more often than the rare pairings of A and C and D and F. After the presentation of each pair of symptoms, the participants guessed which disease they believed the symptoms belonged to. They received feedback, allowing them to learn which pairs of symptoms were associated with specific diseases.
To test the effect of distraction on the task, half of the participants were assigned to complete a distraction task during the training. At the beginning of each training trial, participants heard a list of six random single-digit numbers. After viewing the symptoms, guessing a diagnosis, and receiving feedback, they were given a probe (a number from the original list they heard at the beginning of the trial) and were asked to recall the number that immediately followed it in the list. The participants who completed this task were assigned to the “concurrent” condition, meaning they had an additional cognitive load that would reduce the amount of attention they could allocate to the task. The other participants in the “control” condition completed a different filler task instead, which didn’t involve distraction or affect their attention.
After the training phase, participants completed a test phase in which, on each trial, they were presented with one of the prior training symptom combinations or additional new symptom combinations and were asked to make a diagnosis (see the right side of the figure below for all new combinations). Participants in the concurrent condition also completed the distraction task on each trial. Unlike the training phase, in the test phase, participants didn’t receive any feedback on whether their diagnosis was correct.
Dome and Wills ran this experiment twice. The task remained largely unchanged the second time, with one key distinction: the introduction of time pressure into the participant’s decision-making. Differing from Experiment 1, in Experiment 2, participants were required to make their diagnosis choices and their number responses (from the distraction task) within only 5 seconds.
As expected, in Experiment 1, participants in the control condition exhibited the IBRE. They more frequently, irrationally predicted the rare disease in response to the B and C symptom combinations (top row, 5th column in the graph below). Evidence from participants in the concurrent condition did not provide definitive support for the IBRE effect.
In Experiment 2, when participants experienced time pressure throughout the task, those in the control condition exhibited the IBRE, similar to the results in Experiment 1. Interestingly, and in contrast to Experiment 1, participants in the concurrent condition didn’t really seem to show much irrational responding. There was actually strong evidence that the IBRE effect was absent in these participants (bottom row, 5th column in graph below).
These results aligned with the Dome and Wills’ hypothesis about distraction, with a few caveats related to time pressure. As summarized beautifully by the authors:
“Surprisingly, a lack of pressure can hinder rational generalization. People are less likely to show irrational generalization when under time constraints and extra task demands. Interestingly, this irrationality reduction was predicted by two mathematical models of attention in learning, where distractions that interfere with attentional processes diminished the irrational response bias.”
I found these results incredibly interesting and counterintuitive. Intuitively, you might expect having more cognitive resources to allocate to a task would make you better at it, but this didn’t seem to be the case for the IBRE effect. Irrational behavior may not always be due to not thinking hard enough; sometimes it might be from thinking too much!
Featured Psychonomic Society article
Dome, L., & Wills, A.J. (2025). Better generalization through distraction? Concurrent load reduces the size of the inverse base-rate effect. Psychonomic Bulletin & Review. https://doi.org/10.3758/s13423-025-02661-1
