Hendrikje Nienborg (left) and Bruce Cumming.

Our brains decipher a wealth of sights, sounds and other sensory information to allow us to make sense of our environment. But interpreting sensory inputs isn't always straightforward, as anyone who has stared at the infamous 'young girl–old woman' illusion knows. Hendrikje Nienborg and Bruce Cumming, working at the National Institutes of Health in Bethesda, Maryland, have now discovered that the brain 'tampers' with the signals it receives to favour one interpretation over another.

Signals from sensory organs activate sensory neurons, which in turn relay those signals to the brain areas that decode them and act upon the information provided. Researchers have known for more than a decade that the activity of sensory neurons varies not only in response to a particular stimulus, but also according to how the brain ultimately interprets that stimulus. “Imagine you are walking in thick fog, looking for a friend wearing a green leather jacket,” says Nienborg, who was a postdoc in Cumming's lab before moving to the Salk Institute for Biological Studies in La Jolla, California. “When a blob appears in front of you, you have to decide whether or not the blob is a green jacket.” The activity of the sensory neurons will vary depending on whether the decision is 'yes' or 'no'.

The widely accepted explanation for such variation is that sensory neurons have a role in decision-making. To test this idea, Nienborg recorded the electrical activity of individual sensory neurons in two monkeys as the animals performed a simple task. Each monkey was shown a series of dot patterns on a computer screen, and had to decide whether the centre of a circular pattern was protruding or receding.

The work, which provided a detailed description of the sensory neurons' activities during the course of each trial, was painstaking. “One downside of the technique is that it requires a lot of data, so we had to do many, many trials for each neuron,” says Nienborg. Although individual trials took only a couple of seconds, the monkeys performed between 800 and 900 trials for each neuron studied. And the data generated were not always of sufficient quality. “You could find out at the end of a week of recording all day, every day that you did not have any data,” says Cumming. “On the other hand, the next week you could have recordings from five neurons.”

In the end, they were able to analyse data from 76 neurons. But the results were not consistent with the conclusion that sensory neurons have a direct effect on decision-making (see page 89). Instead, Nienborg and Cumming concluded that brain areas involved in decision-making are sending signals to sensory neurons, altering their activity. “What we have shown is that the brain changes the sensory input,” says Cumming. “In a way, the brain is tampering with the data.” To explain the findings in terms of the fog analogy, Nienborg adds: “What this means is that if we expect to see a green jacket we are more likely to see a green jacket”.

From an evolutionary perspective, this might indicate that, when faced with uncertain sensory information, it is better to commit to one interpretation or another than to hesitate. If you aren't sure where a predator is coming from, “it might be better to make the decision to run left and get it wrong 50% of the time than to just stand there and get killed every time”, says Cumming.