We make hundreds of decisions each day. Each of these decisions results from a complex interplay between large numbers of neurons, distributed across several regions of the brain. This interplay strongly depends on the brain’s internal state. Our lab would like to understand this state dependence.
Neuromodulation of cortical dynamics and decision computations
The main function of the brain is to select the most appropriate actions, given the sensory information gathered from the outside world and our knowledge about the world. This selection process is called decision-making. In species with complex brains (like humans) this process is remarkably flexible. Decision-relevant information can be stored in working memory and combined with variable amounts of new information, before an action ensues. The same information can be mapped onto one action in one context and into another action in another. The outcome of the selection can also remain internal, as a covert proposition, or belief, about the world. Finally, for reasons that are not well understood, decisions vary widely from one instant to the next, even if the environment is held constant.
Even simple sensorimotor decisions result from a complex interplay between large numbers of neurons distributed across many different specialized regions of the highest part of the brain, the cerebral cortex. The flexibility of human choice behavior suggests that this neural interplay is effectively sculpted “on the fly” in the healthy brain. We would like to understand how. Our current work focusses on the modulatory neurotransmitter systems of the brainstem that are known to govern the global state of the cerebral cortex. These brainstem systems are in a good position to orchestrate the flexibility of decision processes, for a number of reasons. They have widespread ascending projections through which they can “broadcast” the same signal to many cortical areas at once. They can tune cortical circuit parameters such as the balance between excitation and inhibition, the level of neural noise. And they can gate the molecular mechanisms underlying learning — synaptic plasticity — in the cerebral cortex. Intriguingly, evidence suggests that these same systems are also disturbed in many major neuropsychiatric disorders — such as schizophrenia, depression, or Parkinson’s disease.
Our goal is to pinpoint how specific neuromodulatory systems (e.g., the locus coeruleus norepinephrine system) remodel the cortical network dynamics underlying decisions. We try to tackle this challenging task simultaneously at different levels of analysis of human brain function: quantitative analysis of behavior, computational modeling of the underlying algorithmic and neural processes, multimodal neuroimaging (fMRI and MEG) measurements of these processes, and pharmacological intervention to pinpoint the role of the specific neurotransmitter systems involved. We hope that this integrative approach will not only make an important contribution to basic neuroscience, but also open up new directions for uncovering the mechanistic basis of neuropsychiatric disorders. Our current research proceeds along the following five main lines.
One remarkable aspect of decision-making is its variability from one decision to the next, even in the face of identical external conditions. We suspect that a large fraction of this spontaneous variability is due to ongoing changes in global cortical state caused by neuromodulatory systems. We test this idea by linking trial-to-trial fluctuations in cortical signals and people’s choice behavior to an established proxy of cortical state: the size of the pupil measured at constant light levels. We also characterize the spatio-temporal structure of spontaneous fluctuations in cortical network activity while tracking pupil size and manipulating the levels of certain neuromodulators.
A fundamental computation underlying decisions is the slow accumulation of evidence gathered from the environment. The timescale of this accumulation is at least one order of magnitude longer than the time constants of individual neurons. Remarkably, people can adapt the timescale of the accumulation to the characteristic timescale of the evidence. We aim to identify the underlying cortical mechanisms, at the level of individual areas and large-scale cortical networks.
Most experiments in neuroscience probe cognitive operations within discrete “trials”, which are, by design, statistically independent. By contrast, our natural environment and spontaneous activity in the brain predominantly changes on much slower timescales. We aim to link the intrinsic dynamics of cortical networks to a well-documented, but poorly understood, aspect of human decision-making: that decisions made in the past bias decisions in the present.
The human brain is equipped with powerful mechanisms that enable people to adapt their behavior to changing environmental contexts, and to learn new skills when the context remains stable. Neuromodulators seem to be enabling factors for the underlying synaptic plasticity mechanisms. Neuromodulatory systems also seem to broadcast uncertainty signals in the brain, which, in turn, might be key for controlling a decision-maker’s behavioural adjustment and learning. To explore these ideas, we track decision uncertainty, manipulate these neuromodulatory systems, and study the effect on learning on different timescales, along with the underlying changes in cortical network dynamics.