Understanding how nervous systems sense, learn and adapt

Neural Computation

We study the principles of neural computation: how nervous systems transform sensory information into perception, action and adaptive behaviour. Our research integrates computational modelling with experimental neuroscience to uncover the principles governing nervous-system function across multiple levels of biological organisation—from primary sensory neurons and spinal circuits to brainstem nuclei, distributed brain networks and behaviour. We are interested in how computations emerge from biological mechanisms, and how those computations enable an organism to respond flexibly to its environment and its own changing needs.

We investigate both ascending sensory pathways and the descending systems through which the brain regulates sensory processing and behaviour. Sensory signals are not simply relayed from the periphery to higher-order brain regions. Their influence depends on the state of the system, previous experience and ongoing action. By studying these interactions, we seek to understand how local neural dynamics contribute to coordinated function across the nervous system, and how control at one level changes the processing of information at another.

Computational models give us a way to connect these levels of explanation. They turn ideas about neural mechanisms into quantitative predictions that can be tested against observations. We use modelling to examine how the properties of cells and circuits shape population dynamics, how those dynamics relate to sensory processing and sensorimotor control, and how behaviour changes through learning and adaptation. The aim is to identify explanations that account for both the organisation of a system and the flexibility of its responses, rather than describing each experimental observation in isolation.

Our work combines human studies with data from animal models generated through external collaborators. Experimental work in animal models is conducted by those collaborators; within NOX Lab, we integrate the resulting evidence with human experiments and computational modelling. This approach connects measurements that capture different aspects of nervous-system function. It also allows us to ask which computational principles generalise across species, while remaining attentive to differences in anatomy, behaviour and experimental context. Models provide a shared framework for making those comparisons explicit and for identifying where further evidence is needed.

In parallel, we develop computational methods for analysing neural and behavioural data across multiple experimental modalities. These methods help characterise complex dynamics across spatial and temporal scales and connect neural measurements with behaviour. Integrating modalities is more than combining datasets: it requires understanding what each measurement reveals, what remains unobserved and how uncertainty affects an interpretation. Our analytical work supports this process by making relationships within and between datasets quantitatively testable, and by helping distinguish competing accounts of the underlying mechanisms.

Across this programme, sensorimotor control, sensory processing, learning, adaptation, pain and homeostasis provide interconnected scientific questions. How does experience alter the way sensory information guides action? How do neural systems remain responsive while maintaining physiological stability? When does an adaptive response become persistent or poorly matched to its context? We address these questions through an exchange between theory, analysis and experiment. The broader goal is a mechanistic understanding of how nervous systems sense, learn, adapt and regulate behaviour, with computational principles that connect biological detail to the function of the whole organism.

Neurotechnology

Neurotechnology increasingly depends on computational models capable of interpreting complex neural dynamics. We develop computational approaches that enable adaptive technologies for measuring, predicting and influencing nervous-system function. Our starting point is the biological system: what can be inferred about its state, how that state changes over time, and how an intervention might alter its dynamics.

This connects fundamental neuroscience with computational neuroengineering. A neural signal becomes useful for adaptive technology when it can be interpreted in relation to a mechanism or function. We develop models and analytical methods that support this interpretation, including computational biomarkers that describe relevant aspects of nervous-system dynamics. These measures are intended to link observations to testable accounts of the processes they reflect, rather than treating a predictive signal as a complete explanation.

Closed-loop systems make this relationship between measurement and intervention particularly important. They use ongoing observations to update an intervention as the system changes. Computational models can help determine which signals are informative, what changes an intervention is expected to produce, and how feedback should shape subsequent decisions. We are interested in how such systems can accommodate variability between individuals and within the same individual over time.

EPIONE: theory informing adaptive neurotechnology

Professor Flavia Mancini is the theory lead for EPIONE—Effective Pain Interventions with Neural Engineering. NOX Lab leads the theoretical and computational framework within the programme, developing models of pain regulation and neural dynamics to inform adaptive sensing and intervention. This leadership places computational theory at the centre of the connection between biological mechanisms and engineering design.

EPIONE is an interdisciplinary programme developing engineering approaches to chronic pain. Its research brings together sensing, interventions that modulate brain activity, computational models and adaptive control. Our contribution asks how models of the nervous system can guide the interpretation of measurements and the design of interventions, providing a quantitative basis for investigating when and how a system should respond.

Pain is a concrete setting in which to develop these ideas, while the computational questions extend across neurotechnology. How can a model remain useful when observations are incomplete? Which aspects of neural dynamics need to be represented to guide an intervention? How can a system adapt without losing a clear account of why it acts? These questions connect our technology research to the broader aim of understanding nervous-system regulation.

Learn more about EPIONE →

Sensorimotor & Defensive Systems

Nervous systems continually coordinate information about the body and the environment with the demands of action. We study how sensory processing, sensorimotor control and defensive regulation work together to support adaptive behaviour. This includes how organisms respond to changing conditions, learn from experience and maintain physiological stability while pursuing different goals.

Sensorimotor function is an active process. Movements change the sensory information available to the nervous system, while incoming signals guide subsequent actions. Understanding this interaction requires considering both the processing of sensory input and the mechanisms that regulate its influence on behaviour. We investigate these relationships across biological scales, connecting the activity of sensory neurons and circuits with the organisation of behaviour.

Defensive systems are an important part of this broader programme. They coordinate responses to potentially harmful events and help protect the body. Their function depends on context: a response that is useful in one situation may be unnecessary or disruptive in another. We ask how nervous systems regulate this balance, how experience changes protective behaviour, and how defensive responses interact with ongoing sensorimotor demands.

Pain provides one model for studying these principles. It allows us to investigate how sensory evidence, prior experience and physiological state shape perception and action. Persistent pain raises questions about how regulatory processes become sustained or poorly matched to current conditions. These questions sit within a wider investigation of adaptive regulation; they do not define the full scope of our sensory and sensorimotor research.

Homeostasis is another central thread. Organisms must respond to external events while regulating their internal state. We are interested in how these demands are coordinated, and how neural dynamics support both stability and flexibility. Learning and adaptation connect these processes over time, allowing behaviour to change as the environment, the body and the consequences of action change.

By linking these biological questions to computational theory, we seek to explain how sensory information is selected, transformed and used to regulate behaviour. Our aim is to understand the shared principles that organise perception, action and physiological regulation, as well as the mechanisms that give each system its particular function.