Humans have a remarkable ability to see patterns where none exist. In the hot-hand phenomenon we perceive streaks of wins or losses where the data, in truth, are random. New research shows that monkeys are subject to the same bias, which might suggest that the bias is evolutionarily adaptive, and maybe even rational.
The moniker comes from the observations of basketball games. Both players and fans believe that shots are more likely to be successful following successful shots than following misses. That's despite the fact that numerous studies have verified that each individual shot is statistically independent from each other attempt.
Here's another way to think of it: the hot-hand phenomenon is the opposite of the gambler's fallacy. The gambler's fallacy is the belief that if something happens often during one period of time, it will become less likely to happen in the future. If you flip a coin twenty-five times, and twenty-five times it comes up heads, you might catch yourself thinking that you're "due" for a tails. But the truth is that each coin flip is statistically independent. You have even odds of catching a heads or tails on each flip, but the results of one flip do not predict the results of the subsequent flip.
The hot-hand phenomenon isn't unique to watching sports. The cognitive bias has been seen in industrial and pre-industrial cultures, in the West and in the East, and so on. It's everywhere. Humans, it turns out, are simply extraordinarily bad at understanding randomness. We can't easily distinguish random from ordered patterns, our ability to generate random choices is laughable (when asked to "pick a random number," your decisions are actually predictable), and we see order where none truly exists.
At first, given how random the universe appears to be, the bias towards seeing order might be confusing. But University of Rochester cognitive scientist Tommy C. Blanchard and colleagues argue, in a recent issue of the Journal of Experimental Psychology: Animal Learning and Cognition that order, rather than randomness, was probably the norm for most of the natural resources – food and tools – that humans would have encountered through the course of our evolution.
"Natural resources that primates forage for, such as specific plants and animals, rarely distribute themselves in a purely random manner in their natural environment because individual organisms are not independent of one another," he writes, adding, "aggregation offers considerable benefits such as a common habitat, mating and parenting, or the benefits of group foraging." After all, where you find one juicy wildebeest, or a tasty fig tree, you're likely to find others. Blanchard hypothesizes that "our evolved psychology is adapted to assume such aggregate resource distributions as the default."
He goes even farther, to argue that the bias might actually be adaptive. Knowing where food has been found in the past will probably help you find food in the future. When aggregations don't exist, and the distribution of food sources is indeed random, the hot hand bias doesn't actually decrease accuracy, because then success occurs according to random chance. A strategy that is usually beneficial and sometimes neutral, but rarely detrimental, would seem to be a useful one.
Of course, this could all be just a fancy "just-so" story, a tale about an evolutionary environment that could have given rise to a particular sort of cognitive trait which makes a good deal of intuitive, logical sense, but which can't be proven. While it's impossible to assess the cognitive abilities of early humans, we can do the next best thing by looking at non-human primates. If monkeys and humans show the same biases – and if those biases emerge the same ways in the same situations, and break down in the same ways – then it's reasonable to infer that they are in fact the same, having emerged in a common ancestor, and were preserved in each of our independent primate lineages.
To see if the hot-hands bias exists in juvenile rhesus monkeys, Blanchard created something he calls the "correlated outcomes task." The monkeys were put in front of a computer screen and trained to play a game. Two images were shown on the screen, and the monkey chose one of them by gazing towards it. If the monkey chose the rewarded side (left or right), he was given some juice or water.
In two conditions, the rewards were delivered according to a clear pattern. Either the reward was consistently located on one side of the screen (right or left), or the reward consistently alternated between the two sides (left, right, left, right). In a third condition, the reward was completely randomized.
As expected, the monkeys quickly learned to make smart decisions in the two scenarios with obvious patterns. But when the rewards were randomized, the monkeys made their choices as if they expected "streaks" to continue. As with coin flips or basketball shots, the monkeys expected prior rewards to predict future ones.
And they showed that bias consistently, day after day, across thousands of trials. "They had lots and lots of opportunities to get over this bias, to learn and change, and yet they continued to show the same tendency," said Blanchard to the University of Rochester Newscenter.
Using a very simple, intuitive task, Blanchard and his colleagues showed that monkeys' performance was more optimal when rewards were presented in clumps than randomly, and that even when faced with a random distribution of rewards, they behaved as if they expected to find a pattern.
Whether or not Blanchard's evolutionary explanation for the phenomenon is true – that it arises from unevenly distributed food sources faced by our primate ancestors – his data do suggest that the bias is not uniquely human. Our flawed decision-making skills are shared with other primates. Better understanding the nature of these sorts of biases could lead to a more nuanced understanding of problematic gambling and other compulsive behaviors in humans and ultimately to better therapeutic interventions.
They also underscore the importance of determining the environmental context in which minds evolve. Cognitive mechanisms evolve as a direct consequence of the social and physical environment in which animals live, and describing those mechanisms absent their contexts only reveals half a picture.
But these findings also reflect more profoundly on what it means to be human. "Biases in our decision-making mechanisms, like this bias towards belief in winning and losing streaks, say something really deep about what sorts of creatures we are," Blanchard argues. "We often like to think we make decisions based only on the information we're conscious of. But we're not always aware of why we make certain decisions or believe certain things."
[Journal of Experimental Psychology: Animal Learning and Cognition; read the whole paper (PDF)]