How monkeys and humans make novel choices
By Dr Alessandro Bongioanni, Department of Experimental Psychology
One thing that makes our brain so fascinating is the staggering range of behaviours it allows. We are not just good at doing things in environments we know well (such as shopping at our usual grocery store), but we are surprisingly successful at navigating novel environments (such as scrolling this blog, making new friends, finding a job, etc.).
When I was an adolescent, I went for the first time to the only Thai restaurant in my city. Every single dish in the menu was new for me. I knew oysters, I knew beef, but I had never tried fried beef in oyster sauce before. Still, I was able to make up my mind and speculate that I would probably like the fried beef more than the chicken in coconut soup. This simple anecdote illustrates something bigger: all human progress required people to leave the comfort of a familiar situation and explore new places or new ideas. Today, with social and technological changes accelerating, we are continuously exposed to novel objects and situations and we generally cope very well with them.
If the ability to make adaptive choices in novel situations is so crucial for our human exploits, surely this is a unique ability of our species? Well, no. Animals too are able to adapt to new environments and to express meaningful preferences among objects or situations that they have never encountered before, based on similarities with what they know.
In the last few decades, scientists have understood a great deal about the neural mechanisms for learning the value of things, and for making decisions based on these values. However, by necessity, most neuroscience research is done with animals, and often with monkeys who are our closest relatives. Findings in monkeys allow to get an understanding of what may happen in the human brain as well, in the case of fundamental behaviours that both species display.
To study decisions in monkeys, scientists use very simplified and abstract tasks. All the irrelevant features of the real world are stripped down, to ensure that the neural activity measured is uniquely due to the key decision process. Many classic studies, for example, required binary choices between coloured images on a screen: initially, the blue stimulus gives a larger reward (in the form of fruit juice) than the red stimulus; in the next stage, the stimulus values reverse. Because this is all so abstract and unnatural for the animals, they may need months of training before data can be collected.
With this classic approach researchers discovered that the value estimate of each option is tracked by neurons in the most anterior part of the brain, just above the eyes. We gathered a detailed understanding of how this neural circuit represents the value of each individual item, and how it allows comparing options in order to make choices. But have we learned anything about the way our brain makes choices in our daily lives?
It turns out that if we put a person into an MRI scanner and ask her to make choices based on preferences, a distinct area usually lights up: it is still in the anterior part of the brain, but in the midline, where the two hemispheres face each other. So which one is the key brain area? Because MRI scanners cannot see individual neurons but estimate activity indirectly, for the past decade animal researchers and human researchers have thought that limitations to each others’ techniques have led their colleagues to focus on the wrong areas. But reaching a firm conclusion has been difficult.
A new explanation was put forward by a group led by Matthew Rushworth, professor at Oxford University and fellow of the Royal Society. He thought that different neural circuits may be involved in familiar and novel decision-making. Familiar items have been experienced repeatedly over a long time period and therefore the consequences of a familiar choice may be easily predictable. If, for example, I go to my favourite restaurant and I order my favourite dish, I can feel in advance how the experience is going to unfold. In contrast, novel decisions require constructing “on the fly” a somewhat hypothetical expectation of how the new options may feel like. This is what I did when I went to the Thai restaurant for the first time and I had to choose between fried beef with oyster sauce and chicken in coconut soup.
An experiment run by myself (Dr Alessandro Bongioanni), Dr Miriam Klein-Flügge and others in the Rushworth team, used MRI scans of monkeys to prove that indeed we can identify a specific brain circuit for novel choice; interestingly, its location in the brain corresponds with the one found in human fMRI studies, not previous monkey studies.
This discovery was accomplished thanks to a clever experimental design allowing monkeys to express preferences among items that they had not encountered before: stimuli were made of coloured dots, where colour represented the amount of juice at stake, and dot number the probability of receiving it. Some combinations of dot number and colour were highly familiar to the monkeys, others were new. Neural recordings revealed that it was the different degree of familiarity with the task items that had caused the puzzling mismatch between brain areas activated in humans and monkeys.
The fact that monkeys could solve the novel task quite easily opens new avenues for the study of decision neuroscience in primates: they are more intelligent than they may appear when faced with abstract lab tasks, and if the task is designed carefully, it is possible to elicit and study more complex behaviours than it was previously thought possible, such as choosing among novel options. This also bridges a gap between our knowledge of the human and the animal brains: making novel choices is not a unique human ability, but instead it is rooted in a neural circuit that is already present in our primate cousins.
We did not just find out where in the brain subjective value is computed, we also attempted to find out how this happens. We found that the monkeys mentally located each item offered in the “space” of all possible items (here, dot and colour combinations) and did so using a grid code. It is as if I “placed” the value of a given dish at a location in the space of all possible dishes, where one axis may represent texture, another axis flavour, and then temperature, price, etc.
We know that the brain’s “GPS system” for spatial navigation employs a grid code, the discovery of which led to the 2014 Nobel Prize in Physiology or Medicine to professors O’Keefe, Moser and Moser. Now, our new results suggest that the brain may use the same neural code employed to represent physical position in space, also to represent the value of a novel item in the space determined by the item’s features. This is interesting because it suggests that the mechanisms used by the brain to encode information are evolutionarily preserved and transferred across domains.
As Prof. Rushworth points out: 'It is intriguing to think that mechanisms for finding one’s way in physical space also underlie our ability to navigating an abstract space of choice possibilities.'
In order to validate this result, the same team also tested what happens if one disrupts specifically this small region of the brain. In order to do that, we relied on a newly developed ultrasound technique that permits stimulating deep in the brain without any invasive surgery; such stimulation can reversibly alter brain activity for one or two hours. Ultrasound provided the final proof: when the neural circuit for novel choice was targeted, the animals’ behaviour changed, as if they were no longer able to integrate the different features of a given novel option into a single mental representation.
While ultrasound has been around for a long time in medicine, it is only in the last couple of years that it has been used to modify brain functioning in this way. Jérôme Sallet, who developed the experimental setup in Oxford, says: 'Ultrasound neurotechnologies are offering new possibilities to identify the causal roles of brain areas. It is non-invasive and could be coupled with other techniques used to record brain activity. Beyond pure research, ultrasound stimulation might prove to be also a valuable tool to help patients with pharmaco-resistant conditions such as depression.'
This is because it avoids some limitations of current techniques such as TMS, which cannot reach deep into the brain or stimulate with precision. Other techniques such as DBS are precise but require surgery. Ultrasound may have the ideal combination of depth and precision without the need for surgery.
We never stop learning novel things about the brain, and if you think about it, this is the brain learning novel things about itself!
Read the full paper, which was published recently in Nature.