Sometimes, people are no smarter in their decision-making than rats. At least, so it seemed from a paper shown to Arnon Lotem by his colleague and co-author Ido Erev. Lotem, a professor of zoology at Tel-Aviv University, wondered why this might be.

The paper reported on the 'certainty effect', whereby rats repeatedly faced with the option of receiving either a bigger reward infrequently or a lesser reward with certainty preferred the safer option, despite it being less profitable on average. In so doing, they behaved just as humans do when the pay-off odds of choosing two alternatives are described to them verbally.

However, Erev's group had found that if humans are faced with the same situation as the rats — that is, the pay-off probabilities are not explained to them — they behave differently. On a 'computerized money machine', they repeatedly gravitate towards the bigger pay-off, even though it is statistically less likely. This penchant is known as the 'reverse certainty effect'. Lotem and his colleagues set out to establish what lay behind this difference in behaviour. They discovered that both humans and other animals can exhibit certainty or reverse certainty, depending on the cues available to them.

Because humans can perceive the precise amount of a reward from reading numbers, whereas rats must rely on their senses to make estimates, Lotem and his team suspected that perceptual accuracy was the key to solving this paradox. They started by devising a series of experimental scenarios that manipulate the clarity of reward cues, then ran them in both honeybees and humans.

In one experiment, human subjects had to choose repeatedly between two unlabelled buttons on the money machine. One button always provided a small payout; the other supplied a larger one, but did not always pay out. No explanation was offered of the odds of the payout.

Watching people push buttons seemingly at random was enlightening, says Lotem. “In theory, people could have been very smart and pressed one button 20 times and pressed another button 20 times, and, like a scientist, figured out the average,” Lotem says. “But people don't behave like statisticians.” Instead, subjects tried both buttons and developed a preference for the button perceived as better 'most of the time' (see page 917).

However, when payouts were represented by a visual display of scattered dots rather than through clear numerical means — making it more difficult to tell which was better — people preferred the safer button. In other words, they exhibited the certainty effect, just like the rats in the study that first piqued Lotem's interest.

Meanwhile, honeybees rewarded with sugary solutions of varying concentration behaved remarkably similarly to humans, preferring the risky option when discrimination between rewards was easy, and the safe option when discrimination was difficult. “Honeybees can't count,” says Lotem. “But, if you give them a high concentration of sugar, they remember. If they get zero, they remember.” Like humans, the bees behaved as though they preferred the option perceived as better most of the time.

Lotem and his colleagues wondered how their findings might apply to real-world situations. “In the real world, rewards may frequently be ambiguous, and you never know when conditions are going to change,” says Lotem. “So perhaps the tendency to explore both options and to prefer the one perceived as better most of the time is a good strategy.”