Computation of Instinctive Decisions
Instinct and intuition are two fundamental components of behaviour. From the innate reaction of a prey to the sight of predator, to the departure from rationality in human economic decisions, information that is rapidly accessed without reaching consciousness can be used to influence behavioural choices.
Our goal is to understand at the cellular and circuit level how neurons in the brain compute instinctive decisions. We aim to produce a mechanistic model of how neurons integrate sensory information and past experience to generate instinctive behavioural choices, and we are particularly interested in determining the role of synaptic and ion channel mechanisms in the computation of decisions.
We hope that our research will advance our understanding of how choices are made, and provide a framework for investigating causes and therapies for maladaptive decision-making during mental illness states, such as anxiety and depression.
We currently use defensive behaviours in the mouse as a model system, and we are investigating how exposure to innately aversive threats is converted into behaviour, including in complex and dynamic environments. We emphasise the study of naturalistic behaviours and ethologically-relevant decisions, as we believe that this approach can reveal principles of behaviour and brain function behind the fast learning and decision-making processes that ensure survival.
Projects in the laboratory start at the behavioural level, with the development and application of quantitative assays. We are especially interested in making high-resolution measurements of a wide set of sensory and behavioural state variables in freely moving animals, so that we can model behavioural choices on a moment-by-moment basis. We record neuronal activity using miniature endoscopes and high density silicone probes, and when possible, we use in vivo whole-cell patch-clamp and two-photon microscopy. In collaboration with the Gatsby Computational Neuroscience Unit we develop algorithmic and data-driven models that aim to map behavioural computations onto neural activity dynamics and biophysical mechanisms.
These approaches are complemented by circuit level analysis using viral tracing tools, and a variety of high-resolution in vitro patch-clamp recording methods, together with molecular perturbations of ion channels using shRNA and CRISPR technologies.
We are actively pursuing two main questions:
How do neural circuits represent threat and compute the decision to engage in defence?
We focus on midbrain circuits, including the Superior Colliculus (SC) and the Periaqueductal Gray (PAG), and their neuronal activity during the decision-making process of starting defensive actions in response to visual and auditory and stimuli that innately threatening. We investigate information representation and transfer between these two networks, with a strong emphasis on identifying synaptic integration mechanisms that are key for the computation threat and escape initiation.
How is the choice between different defensive strategies computed?
We are investigating how threats of varying intensities and qualities are converted into different defensive behaviours, such as escape and freezing, as well as how these behaviours are gated and modified by learned knowledge about the environment. We aim to determine how cortically-encoded variables that control defensive behaviour interact with the midbrain escape circuits.