A functioning model of human time perception

Despite being a fundamental dimension of experience, how the human brain generates the perception of time remains unknown. Predominant models of human time perception propose the existence of oscillatory neural processes that continually track physical time - so called pacemakers - similar to the system clock of a computer. However, clear neural evidence for pacemakers at psychologically relevant timescales is lacking, raising the question of whether internal pacemakers are necessary for time perception. Here we show that clock-like pacemaker processes are not required for human time perception. We built an artificial neural system based on a feed-forward image classification network, functionally similar to human visual processing. In this system, input videos of natural scenes drive changes in activation within an image classification network and accumulation of salient changes in activations are used to estimate time. Estimates produced by this system match human reports made about the same videos, replicating key qualitative aspects such as report variability proportional to duration (scalar variability/Weber9s law) and response regression to the mean (Vierordt9s law). System-generated estimates also differentiate by scene type, such as walking around a busy city or sitting in a cafe, producing the same pattern of differences as human reports. Our results show how time perception can be derived from the operation of non-temporal perceptual classification processes, without any neural pacemaker, opening new opportunities for investigating the neural foundations of this central aspect of human experience.