The ability to sense elapsed time is central to complex cognitive functions, informing every day the decisions we take and the actions we make. Yet, how neural populations distributed across the brain keep track of time remains unclear. Here, we set out to investigate large-scale neural population activity associated with timing behaviour. We developed a self-paced behavioural paradigm in which head-fixed mice learn to initiate reaching movements towards a spout dispensing drops of water reward. Mice start each trial themselves by holding onto a bar; the reward is then only dispensed if a certain amount of waiting time (target delay) has elapsed before they initiate the reach. We observed that mice can learn the target delay through trial-and-error, successfully adjusting their behaviour to optimize performance. We used multiple neuropixel probes to simultaneously record neural activity from several brain regions, including the striatum, nucleus accumbens, hippocampus, thalamus, parietal, and prefrontal cortices, while mice learned and performed this task. Our preliminary analyses indicate that population activity preceding self-paced reaches contains information that allows elapsed trial time to be decoded, with neurons across all recorded regions contributing to this time representation. By analyzing cells with significant decoding contribution, we further observed that time keeping relates to a limited set of temporal firing patterns. Embedding population firing into a low-dimensional space allowed unsupervised retrieval of these distinct neuronal tuning curves. This ongoing work suggests a principle of organisation for time tracking by neuronal populations in goal-directed self-paced behaviour.