Attention-deficit hyperactivity disorder (ADHD), one of the most common psychiatric disorders, is characterised by unstable response patterns across multiple cognitive domains. However, the neural mechanisms that explain these characteristic features remain unclear. Using a computational multilevel approach, we propose that ADHD is caused by impaired gain modulation in systems that generate this phenotypic increased behavioural variability. Using Marr's three levels of analysis as a heuristic framework, we focus on this variable behaviour, detail how it can be explained algorithmically, and how it might be implemented at a neural level through catecholamine influences on corticostriatal loops. This computational, multilevel, approach to ADHD provides a framework for bridging gaps between descriptions of neuronal activity and behaviour, and provides testable predictions about impaired mechanisms.

Survival and reproduction entail the selection of adaptive behavioural repertoires. This selection manifests as phylogenetically acquired activities that depend on evolved nervous system circuitries. Lorenz and Tinbergen already postulated that heritable behaviours and their reliable performance are specified by genetically determined programs. Here we compare the functional anatomy of the insect central complex and vertebrate basal ganglia to illustrate their role in mediating selection and maintenance of adaptive behaviours. Comparative analyses reveal that central complex and basal ganglia circuitries share comparable lineage relationships within clusters of functionally integrated neurons. These clusters are specified by genetic mechanisms that link birth time and order to their neuronal identities and functions. Their subsequent connections and associated functions are characterized by similar mechanisms that implement dimensionality reduction and transition through attractor states, whereby spatially organized parallel-projecting loops integrate and convey sensorimotor representations that select and maintain behavioural activity. In both taxa, these neural systems are modulated by dopamine signalling that also mediates memory-like processes. The multiplicity of similarities between central complex and basal ganglia suggests evolutionarily conserved computational mechanisms for action selection. We speculate that these may have originated from ancestral ground pattern circuitries present in the brain of the last common ancestor of insects and vertebrates.

Reinforcement (trial-and-error) learning in animals is driven by a multitude of processes. Most animals have evolved several sophisticated systems of 'extrinsic motivations' (EMs) that guide them to acquire behaviours allowing them to maintain their bodies, defend against threat, and reproduce. Animals have also evolved various systems of 'intrinsic motivations' (IMs) that allow them to acquire actions in the absence of extrinsic rewards. These actions are used later to pursue such rewards when they become available. Intrinsic motivation has been studied in Psychology for many decades and its biological substrate is now being elucidated by neuroscientists. In the last two decades, investigators in computational modelling, robotics and machine learning have proposed various mechanisms that capture certain aspects of IMs. However, we still lack models of IMs that attempt to integrate all key aspects of intrinsically motivated learning and behaviour while taking into account the relevant neurobiological constraints. This paper proposes a bio-constrained system-level model that contributes a major step towards this integration. The model focusses on three processes related to IMs and on the neural mechanisms underlying them: (a) the acquisition of action-outcome associations (internal models of the agent-environment interaction) driven by phasic dopamine signals caused by sudden, unexpected changes in the environment; (b) the transient focussing of visual gaze and actions on salient portions of the environment; (c) the subsequent recall of actions to pursue extrinsic rewards based on goal-directed reactivation of the representations of their outcomes. The tests of the model, including a series of selective lesions, show how the focussing processes lead to a faster learning of action-outcome associations, and how these associations can be recruited for accomplishing goal-directed behaviours. The model, together with the background knowl-1 edge reviewed in the paper, represents a framework that can be used to guide the design and interpretation of empirical experiments on IMs, and to computationally validate and further develop theories on them.