The Australian Research Council Centre of Excellence for Integrative Brain Function (CIBF) included neuroscientists, engineers, psychologists and computer scientists, and covered a multitude of topics ranging from ion channel function in single neurons to sensory and cognitive processing within neural networks. The overarching aim of the research was to increase our understanding of how the brain integrates information across multiple levels. Key to this process was the development of new experimental paradigms, instruments and computational tools. The new insights obtained have also provided a basic-science foundation for the development of approaches to repair and restore function to the damaged brain.

In order to understand integrative brain function, it is critical to understand how the brain processes information. Information processing in the brain relies on spiking activity in single neurons, which requires the movement of charged ions through ion channels in neuronal membranes. CIBF researchers have investigated ion flow through small conductance calcium-activated potassium channels which contribute to the after hyperpolarization that follows spike activity in many neuronal cell types. To further gain insight into information processing in the brain, an understanding of how populations of neurons encode information in their patterns of spiking activity is essential. Methods for extracting variables that quantitatively describe how sensory information is encoded, such as for estimating receptive fields, modelling neural population dynamics and inferring low dimensional latent structure from neuronal populations, are crucial. Centre researchers have developed analytical approaches to make inferences from population recordings in multiple brain areas, such as dimensionality reduction and changes in correlated variability.

The brain integrates and processes a massive amount of sensory information to guide behaviour crucial for survival. Recent discoveries have provided evidence that the brain continually predicts incoming sensory events based on past experience in order to respond to unexpected events in a fast and efficient manner. Centre researchers have explored how the visual system codes visual stimuli based on biologically plausible models to investigate how complex experimental phenomena, such as the shapes of receptive fields and contrast invariance of orientation tuning, can be implemented in primary visual cortex by sparse coding. Other studies have compared the representations of space and motion in the visual and auditory cortex, and examined how single neurons in these two areas encode the direction of motion.

A key objective of the Centre’s research was to support the development of new technologies and approaches to build better models of integrative brain function. CIBF researchers have identified how neural activity across different cortical areas can boost synchronized oscillations between brain areas. A nonlinear, hierarchical model that provides unique insights into the brain architecture underlying the representation and appraisal of perceptual belief and precision have also been developed. These approaches can be used to gain a greater understanding of functional encoding within sub-nuclei during memory formation and may prove advantageous for studying the cellular basis of addiction as well as pathological memory models. An investigation of gene expression patterns associated with hub connectivity in neural networks has presented evidence that some expression patterns are conserved across species and scales. Together, these studies provide new models of brain networks which aids our understanding of how the brain integrates information across multiple brain regions.

Research Programs

The Centre’s scientific aims were to reveal how the brain integrates information in large-scale networks to yield complex behaviour, and develop neural technologies and models to understand the complexity of dynamic brain processes.

Sensory Decision Making

Our understanding of the way that neurons in the brain process sensory information was advanced. We discovered new pathways critical for processing sensory information as well as a deeper understanding of how different brain states impact on the activity of neuronal populations during sensory decision-making behaviour. We characterized how prediction changes neuronal activity in the primary sensory cortex and in turn affects the amount of information individual neurons carry about the sensory input.

The superior colliculus was identified as a key brain region involved in attention, as well as the way that brain states and attention modulate the efficiency of sensory processing. We established new technologies for recording large populations of neurons using multi-channel Neuropixels probes as well as novel methods for manipulating neurons using optogenetics. The research provided a clearer picture of how neurons and neural systems process information, which has in turn advanced our understanding of how our brains work.

Decision Making and Memory Formation

Our research evaluated decision making during fear conditioning and extinction, in a brain network that engages the hippocampus, amygdala and prefrontal cortex.  We established the nature and activity of circuits that transmit information between these three regions and how information is processed within these regions. Entirely new circuits in the prefrontal cortex were discovered with inter-neuronal connections that were not previously known. A new type of interneuron called the chandelier neuron was revealed to form a small population with wide-spread connections. The research brings to light how neural networks are engaged during decision making, and memory formation and consolidation.

In fear extinction, neurons in the hippocampus project to the medial prefrontal cortex that drive a particular set of interneurons that then inhibit cells that send projections to the amygdala. Characterisation of this novel neural circuit opens avenues to investigate how danger is evaluated in the environment. Memories of events are formed and consolidated by network oscillations. The circuits and neurons that initiate these oscillations and maintain their structure were discovered, as well as the circuits that drive the particular frequencies and how they terminate. The research revealed new insights into how the nervous system processes information and how memories are formed and updated. The findings provide crucial foundational knowledge upon which to build innovative ways to treat disorders such as addiction, which are in part due to problems with memory formation and retrieval.

Visual Cortical Processing

The study of receptive fields in the visual cortex is a long-standing and complex endeavour. A non-linear input model of visual cortical processing was developed and implemented. The model is a sophisticated neural network that incorporates concepts linked to deep neural networks and predictive coding. Our development of the non-linear input model has revolutionised understanding of visual receptive fields in the visual cortex. Prior to this research, our understanding of visual computations was based on theoretical models that were compared with data. The use of data-driven modelling to reveal how neural processing occurs promises to have a major impact on the field of visual processing in the next decade.

We initiated the use of a range of new multi-electrode recording devices using new materials to make the devices functional. Novel technologies were developed to analyse hundreds of times more data than was previously possible. The research produced an unanticipated technology development that utilised flexible carbon-based recording materials instead of metal electrodes. The devices are in development for medical grade equipment for use in human brain implants, as carbon-based electrodes have the potential to survive and function in the brain for a lifetime. The development of new carbon-based, high-density electrode recording devices is expected to revolutionise the ability to communicate with the human brain.

Biophysical Neuron Models

Biophysically detailed neuron models revealed that tonic inhibition can increase excitability in somatostatin interneurons, whereas until now tonic inhibition was thought to reduce excitability of all neurons. The discovery has significant implications for the role of gap junctions in controlling cortical network activity. By combining realistic neuron models with patch clamp measurements featuring Dynamic Clamp analysis, the impacts of mutations in two genes (SCN1A and SCN2A coding for sodium channels Nav1.1 and Nav1.2) were elucidated. The dynamic action potential clamp (DAPC) system was shown to provide clear benefits over conventional voltage clamp for rapid and definitive prediction of neuronal channelopathies. The model also provides insight into the impact of tonic inhibition upon neural activity, and suggests a mechanism through which GABA, the predominant inhibitory neurotransmitter of the mammalian brain, may modulate the excitability of neurons in a selective manner.

Brain Dynamics

Brain models of attention and prediction using decision and control functions based on brain physiology has placed brain dynamics in a form compatible with engineering control systems. The modelling research focused on the formulation of a theory of brain dynamics and connectivity in terms of eigenmodes and their resonances to simultaneously analyze brain activity, structure and function. Analytical approaches for linear and nonlinear brain waves were developed using state tracking and brain modelling algorithms. Links between evoked brain responses and the underlying dynamics, especially in the context of gain-mediated attention, were revealed. The research has applications to vision and brain stimulation methodologies, as well as brain disorders such as epileptic seizures and Parkinson’s disease.

Novel quantitative approaches to interrelate brain phenomena and measurements across brain scales led to the discovery of hemodynamic waves in the cortex. Furthermore, brain modes were discovered to be locked into remarkably stable fixed orientations by the brain’s convolutions. Modelling of the clearance of waste products from the brain was revealed as a critical factor to understand arousal dynamics which contribute to cognition. Models of brain connectivity shifted the field away from phenomenology toward physically based approaches. Phenomenological methods are prone to artefact, whereas methods to obtain eigenmodes from data are of immense practical benefit to compactly describe and monitor brain states and connectivity.

Brain Computation

Studies of perceptual learning and the underlying connectivity of distributed neural architectures responsible for evoked brain responses were undertaken. Dynamic causal models of event-related potentials provide a non-invasive method to investigate experience-dependent coupling among brain regions. Computational models were applied to investigate the challenging question of rapid stimulus appraisal and the functional role of subcortical pathways to the amygdala. A longstanding debate in the field of affective neuroscience was resolved by demonstrating that a subcortical pathway to the amygdala bypasses cortical pathways to provide rapid information about visual and auditory objects.

Our research extended foundational connectivity models into disorders of consciousness. Empirical tests of predictive coding models were applied in altered states of cognition. Foundational work using a dynamic causal modelling examined key questions about the specific roles of forward and backward connections in the brain and their relation to exogenous and endogenous brain process. A specific network disruption associated with perturbed consciousness in coma was identified, whereby a top-down connection between frontal and temporal cortices was discovered to be disrupted in vegetative patients but preserved in patients in a minimally conscious state and in healthy controls.

Selective attention

A central research goal was to understand how the human brain filters and integrates information from the principal sensory modalities – vision, audition, somatosensation and olfaction – and how these processes regulate perceptual and cognitive processes. We examined the role of neural oscillations in the regulation of sensory and attentional processes. Multivariate analysis approaches were used to characterise how simple feature information, including orientation, motion and spatial location, was affected by changes in attentional state. We discovered that both the frequency and phase of endogenous neural activity, particularly in the alpha and gamma bands, affects attentional allocation and detection thresholds in attentional cueing and multiple-object tracking tasks, and that these oscillations either predict or correlate with perceptual awareness. We developed a novel paradigm for use with neuroimaging approaches whereby multiple, concurrent sensory events were counter-phased or flickered at unique frequencies. The signatures of the stimuli were recovered individually via Fourier analysis to simultaneously track the activity patterns from multiple attended and ignored stimuli.

The influence of attentional load on visual and auditory attention was examined using neural signatures of attentional tracking during approach and avoidance behaviour, to determine how the brain represents attended and ignored sensory events that have been masked from awareness. A brain computer interface system acquired neural activity with ~300 millisecond delay to provide a feedback signal to participants. The approach enabled examination of the extent to which people are able to modulate their own brain activity to achieve specified task goals via neurofeedback, and whether such strategies can be harnessed to drive implicit and explicit learning of attentional control.

Predictive coding of perceptual features

According to predictive coding theory, neural responses to sensory stimuli arise from an active Bayesian inference mechanism, which generates predictions (or ‘beliefs’) about the world and compares these with incoming sensory inputs. If the evidence matches the prediction, the model is retained. If there is a mismatch – also known as a prediction error – the model is updated. Research was undertaken aimed at understanding how the brain establishes and tests models of the world against current sensory inputs, and whether such processes are governed by or determine attentional selection. Studies examined how the statistical properties of sequential stimuli generate surprise responses in the brain, whether these responses are influenced by attention and task-relevance, and whether neural responses to elementary visual features are encoded differently for unpredicted and predicted stimuli.

To understand how predictive coding alters activity at the level of individual neurons, neurophysiological data were recorded from visual areas of the brain in awake mice using two-photon calcium imaging. We examined when and how neuronal responses are altered by changes in the predictability of simple stimuli. The work used computational modelling to relate patterns of neural activity at the level of single neurons with wider brain networks, across tasks and species in an integrative framework, to understand predictive coding in the mammalian brain.

Attentional amplification of neural activity

An enduring question is whether and to what extent focused attention is required to accumulate statistical information about sequence regularities. Our research found that even under high concurrent attentional load the human brain tracks the statistical properties of ignored auditory events and shows an enhanced prediction-error response to oddball events. Bayesian model selection was used to determine the effects of attention on mismatch responses in an auditory selective listening task, which revealed that attention amplifies neural responses equally to predicted and unpredicted events. Multivariate forward-encoding models determined that representations of elementary visual features, such as colour, orientation and motion, are encoded differently for predicted and unpredicted visual stimuli. Significantly, an approach to measure feature-specific tuning of neural responses to elementary visual features based on forward-encoding modelling revealed that the amplification of a visual feature increased with prediction violation, whereas there was no effect on the sharpness of the visual feature.

Perceptual decision making

Virtually every aspect of waking life requires a decision. What are the brain processes that underlie decision making in humans? A range of approaches was used to investigate the neural and behavioural correlates of simple and complex perceptual decisions in the visual modality. For studies of simple decision making transcranial random noise stimulation was used to test a key prediction from stochastic resonance theory, which postulates that the addition of an optimal amount of external or internal noise to a sub-threshold stimulus can push the stimulus above threshold and thus improve perceptual performance. Consistent with stochastic resonance theory, an enhancement of simple perceptual decisions when transcranial random noise stimulation was delivered in conjunction with subthreshold motion signals was discovered. A model of the behavioural data under a hierarchical drift diffusion framework revealed that the addition of neural noise specifically increased the drift rate of evidence accumulation.

Bayes’ theorem postulates that the probability of a hypothesis given an observation is a weighted sum of the probability of the hypothesis on its own and the probability of the observation given the hypothesis. Bayes’ rule has been used successfully to model different aspects of human cognition, from shape constancy to spatial navigation. A Bayesian approach was used to understand complex perceptual decision making, which entails the integration of sensory information from two or more sensory sources. In one approach, participants viewed two or more sequential stimuli embedded in dynamic noise and judged the average of the stimuli, which required participants to weight and integrate sensory evidence from two or more different signals. A number of key properties of the sensory events that contributed to the final decision, both in terms of behaviour and corresponding neural activity, were discovered consistent with Bayesian theory.

Brain Maps

The primate cortex is separated into dozens of visual areas that form a mosaic of individual visual maps. Topographic brain maps have been classified according to whether they represent visual information as a mirror image or a non-mirror image of the object being viewed. The classification is used to determine the transition between different areas in the visual cortex. Research to model how topographic maps are formed during brain development was undertaken to investigate the interaction between two maps as they develop in adjacent brain areas. Some configurations of brain areas were discovered, suggesting a previously unknown, ‘twisted’ type of map that combined regions that represent images as both mirror images and non-mirror images. Advanced electrophysiological techniques showed that this third type of map within a single brain area exists in the primate brain.

This line of work demonstrates that the formation of two adjacent brain areas can create new types of organization that would not be predicted by modelling the formation of each area independently. To capture the full complexity of the human brain it is necessary for models to incorporate multiple areas and take into account the fact that they develop at different times. Understanding order in which brain areas develop across the visual cortex is important for understanding differences in the organisation of the cortex between species, including humans.


Scientists use brain atlases to navigate seamlessly between the central nervous system of humans and experimental animals to test hypotheses inspired by human considerations and observations. Atlases have traditionally been developed using post-mortem brain tissue. Our research used in-vivo MRI images to construct a high-resolution atlas of the living human brain and the rat brain. Brain regions in MRI post-mortem brainstem images were delineated to identify the boundaries of more than 300 different structures across approximately 60 levels. The level of delineation was highly significant as all structures identified in previous histological atlases were revealed and the greatest number of delineations achieved compared with any previous brain atlas or tool. Notably, a nucleus that was previously unknown (endorestiform) was identified as well as four homologies between the human and rat brain. A comprehensive set of atlases of the adult human brain and spinal cord of humans and experimental animal models in rat, mouse, rhesus monkey, and marmoset were completed.

Lightsheet Microscope

A custom lightsheet imaging microscope that provides high resolution images of specific brain regions was built, compatible with many brain clearing techniques to generate numerous data sets. The design and construction of our microscopes for dynamic imaging included the control software and analysis tools to process the images into opto-physiology data sets. The microscope was used to identify novel morphological changes in a model of Dravet Syndrome that were consistent with known cognitive and behavioural co-morbidities in this neurodevelopmental disease model. The instrument was used to establish a novel methodology to clear, label, image and automatically register images to a mouse brain atlas. Statistical analyses of regional differences identified a time point prior to the onset of seizures in an epileptic encephalopathy mouse model of Dravet syndrome. The lightsheet microscope findings demonstrated that the glass-brain imaging methodology provides a valuable adjunct tool for anatomical analysis of neurodevelopmental disease models.

Neural microchips

Existing neural microchips either rely of the use of sub-threshold circuit designs to achieve low power consumption at the expense of manufacturing yield, or come with the requirement for batteries or wires to satisfy high power requirements. A novel low-power neural implantable in-vivo recording microchip and associated external transmitter/receiver hardware was built to record Local Field Potentials and Action Potentials. Low-power techniques for signal amplification and analogue-to-digital conversion driven from a wireless power supply were developed. The microchip design enables volume production and channel scalability of low-power wireless recording implants, and facilitates closed-loop implant operation for self-optimising systems such as bionic eye systems where significant patient training is not required. A manufacturable design which achieves a calculated balance of performance and power consumption was demonstrated.


The Centre of Excellence for Integrative Brain Function research program studied the relationship between brain activity and behaviour at multiple spatial and temporal scales – from single cell electrical and biochemical activity to patterns of activity in large scale circuits and networks. In doing so the Centre contributed to the development of an integrated model of how the brain processes information and thereby advanced our understanding of how the brain interacts with the world.