The Neuronal Umwelt: spikes as actions in a multi-scale ecology of adaptive agents

We propose an alternative paradigm that treats the individual neuron as an active, teleological agent operating within its own distinct umwelt. Rather than conceptualizing neurons as passive input-output devices optimized exclusively for information processing, we frame them as living cellular agents embedded within a complex socio-ecological environment composed of other neurons, glia, trophic constraints, an ionized interstitial fluid environment, and constraining large-scale network dynamics. Driven by the fundamental biological imperative of structural permanence, neurons engage in coalition-building, functional specialization, and competition for finite metabolic and neurotrophic resources. Developmental periods are interpreted as critical windows during which neurons attempt to consolidate stable positions of influence and control within evolving circuits, while circuit membership simultaneously enables and constrains individual neuronal function. By leveraging Perceptual Control Theory (PCT) and optimal feedback control, we model the neuron as an active controller that uses action potentials as actions to steer its network environment toward preferred, survival-conducive sensory states - sensed via distributed dendritic branches. In this view, spikes are not merely reactive outputs but interventions that modify downstream dynamics and shape the neuron’s future inputs through closed feedback loops. This framework naturally accounts for intrinsic spontaneous activity and departs from traditional feedforward McCulloch-Pitts-Rosenblatt abstractions by positioning neurons as adaptive feedback controllers capable of learning predictive models of their relationships across highly distributed network structures. Central to this perspective is the neuron’s unique capacity to integrate spatio-temporally distant information through long-range projections, thereby dramatically extending its effective sensory world relative to other cell types. Furthermore, we extend this notion of agency hierarchically by viewing dendritic branches as semi-autonomous computational subunits capable of localized plasticity regulation independent of global somatic activity. Local dendritic spikes, compartmentalized biochemical dynamics, and branch-specific plasticity mechanisms suggest nested layers of adaptive feedback control spanning dendrites, neurons, assemblies, and large-scale circuits. Within this framework, morphological growth, synaptic stabilization, and predictive modeling form a self-reinforcing feedback loop through which neurons progressively expand their capacity to sense, predict, and influence network dynamics. We argue that many phenomena traditionally interpreted through purely computational principles - such as efficient coding, irregular spiking, stochastic resonance, or sparse representations - may instead emerge from negotiated equilibria among locally self-preserving cellular agents operating under ecological and developmental constraints. Ultimately, this perspective reframes the nervous system as a multi-scale ecology of interacting adaptive agents whose collective dynamics give rise to cognition and behaviour in service of organismal-level survival goals.