Time Perception

Abstract

Time perception is the estimation of duration, a sense with no dedicated receptor organ yet one that governs speech, music, and movement. Its central puzzle is that a mind with no clock nonetheless times intervals, and the dominant account posits an internal clock: a pacemaker emits pulses that a switch gates into an accumulator, so more pulses mean a longer felt duration. This framework explains the signature regularity of interval timing, the scalar property, whereby timing error grows in proportion to the interval timed, and predicts how attention, arousal, dopamine, and emotion distort felt time. Competing accounts hold that time is not read from a dedicated clock at all but from the evolving state of ordinary neural circuits. Three interactive demonstrations model the scalar property, the pacemaker-accumulator clock, and the temporal bisection task that maps subjective duration.

Keywords: time perception, interval timing, internal clock, scalar timing

Time perception refers to the set of processes by which durations in the range of milliseconds to minutes are estimated, compared, and anticipated, in the absence of any sensory organ dedicated to time. Unlike light or sound, time is not transduced; it is constructed, inferred from the changing state of the nervous system as events unfold (Wittmann, 2013). The field's organizing achievement is to have made this construction measurable: by asking observers to reproduce, discriminate, or categorize intervals, researchers recover lawful functions that constrain any theory of the underlying mechanism (Grondin, 2010). This article moves from the behavioural signature of interval timing, through the internal-clock model that dominates the field and the tasks that test it, to the roles of attention, dopamine, and emotion, the neural circuits implicated, and the challenge from accounts that dispense with a clock altogether.

What Time Perception Is

Time perception is unusual among the perceptual capacities because it has no proximal stimulus. Vision begins with light on the retina and audition with pressure at the eardrum, but there is no organ that transduces duration; the interval between two events is not a physical energy that impinges on a receptor (Wittmann, 2013). What the nervous system has instead is change, and duration must be recovered from how much the system has changed between one event and another. This makes time perception a constructive process throughout, and it is why the same physical interval can feel long or short depending on the observer's state, task, and expectations.

The field concentrates on interval timing in the range from a few hundred milliseconds to a few minutes, the scale at which timing is deliberate and cognitively penetrable, as distinct from the circadian clock that governs the daily cycle and from the microsecond computations of sound localization (Buhusi & Meck, 2005). Within this window the discipline studies a small set of capacities: estimating how long an interval lasted, reproducing it, discriminating two intervals, and anticipating when a recurring event will arrive. These capacities are measured with a handful of standard tasks whose lawful results, reviewed below, any adequate theory of timing must reproduce (Grondin, 2010).

The Scalar Property

The most robust fact about interval timing is the scalar property, the timing analogue of Weber's law. When observers estimate a duration, the variability of their estimates is not constant but grows in proportion to the interval being timed, so that the standard deviation of estimates divided by the mean, the coefficient of variation, stays roughly constant across a wide range of durations (Gibbon, 1977). Timing a two-second interval and timing a twenty-second interval yield estimates with the same relative spread; the longer interval is simply timed with proportionally more absolute error. This constancy of the coefficient of variation is the quantitative fingerprint of scalar timing, and it is what a plot of error against duration must show for the scalar property to hold.

A striking consequence follows when timing distributions are rescaled. If the response distributions for several different target durations are plotted on a normalized axis, each time divided by its own target, the curves superimpose almost exactly, collapsing onto a single function (Wearden & Lejeune, 2008). This superimposition is a stronger claim than constant relative error, because it says the whole shape of the timing distribution scales with duration, not merely its width. The scalar property holds across species, across the seconds-to-minutes range, and across reproduction and discrimination tasks, which is why it serves as the primary benchmark for models of the clock. The first demonstration lets a reader set a target duration and a Weber fraction and watch both the raw timing distributions widen with duration and the normalized distributions superimpose.

The Internal Clock

The dominant explanation of interval timing is the internal-clock model, whose modern form is scalar expectancy theory. The proposal, which grew from early work modelling temporal discrimination as the reading of an internal clock, is that a pacemaker emits pulses at a roughly regular rate; when an interval to be timed begins, a switch closes and lets the pulses flow into an accumulator; when the interval ends the switch opens, and the number of pulses collected is the raw measure of elapsed time (Treisman, 1963). More pulses mean a longer perceived duration, so anything that speeds the pacemaker or holds the switch closed longer inflates the estimate, and anything that slows the pacemaker or leaks pulses shortens it. This clock-counter architecture is the engine behind most predictions in the field.

Scalar expectancy theory extends the bare clock into a three-stage information-processing model, adding memory and decision to the clock itself (Gibbon et al., 1984). The accumulated count is transferred to a reference memory that stores the pulse counts associated with remembered durations; a comparator then judges the current count against a sample from that memory and issues a response when the two are close enough. Locating the scalar variability in the memory stage, as a constant proportional distortion in the stored counts, is what allows the theory to reproduce the scalar property while keeping the pacemaker itself simple. Figure 1 sets out the stages, and a comprehensive review places this architecture at the centre of the field's account of timing (Buhusi & Meck, 2005). The second demonstration exposes the clock directly, letting a reader change the pacemaker rate and read off how the accumulated count and the subjective duration depart from real time.

Measuring Subjective Time

Because subjective duration cannot be observed directly, it is inferred from choice, and the temporal bisection task is the paradigm case. Observers first learn two anchor durations, a short standard and a long standard, and are then presented with intermediate probe durations and asked whether each is closer to the short or the long anchor. The proportion of long responses rises smoothly from the short anchor to the long one, tracing a psychometric function, and the duration at which the observer is equally likely to answer short or long, the bisection point, marks the subjective midpoint of the interval range (Wearden & Lejeune, 2008).

The location of that midpoint is theoretically informative. If subjective time were linear in physical time, the bisection point would fall at the arithmetic mean of the two anchors; in practice it falls close to their geometric mean, the point that is equidistant from the anchors on a ratio scale (Gibbon, 1977). A bisection point at the geometric mean is exactly what scalar timing predicts, because when variability is proportional to duration the relevant comparisons are ratios rather than differences, and the geometric mean is the ratio-scale centre. The bisection task thus does double duty: it measures the subjective midpoint and it tests the scalar account through the predicted location of that midpoint. The third demonstration runs a bisection procedure, letting a reader set the anchors and slope and watch the psychometric function and its bisection point emerge at the geometric mean.

Prospective and Retrospective Timing

A duration can be judged in two fundamentally different ways, and which one applies depends on whether the observer knew in advance that time would be relevant. In prospective timing the observer is told beforehand to attend to duration and times the interval as it passes; in retrospective timing the observer is asked only afterward how long something lasted, having formed no deliberate estimate at the time (Zakay & Block, 1997). The two draw on different resources and produce different results. Prospective estimates depend on attention allocated to timing and are the province of the internal clock, whereas retrospective estimates are reconstructed from memory, and depend on how much change or how many events the interval is remembered to have contained.

The dissociation is diagnostic. Filling an interval with a demanding task shortens prospective estimates, because attention is drawn away from timing and the switch admits fewer pulses, but the same manipulation can lengthen retrospective estimates, because a busier interval leaves a richer memory trace (Zakay & Block, 1997). This opposite response to a single manipulation is strong evidence that prospective and retrospective judgements are produced by distinct processes, a clock-like mechanism in the first case and a memory-based reconstruction in the second. Most laboratory work on the internal clock is therefore prospective by design, holding memory constant so that the clock can be studied in isolation.

Attention and the Felt Passage of Time

Prospective timing competes with other tasks for a limited pool of attention, and this competition is built directly into the clock model as the gating switch. When attention is diverted from timing to a concurrent task, the switch flickers or opens, fewer pulses reach the accumulator, and the interval is underestimated; when attention is fully devoted to timing, the switch stays closed and the count runs high (Coull et al., 2004). This is the mechanism behind the everyday observation that a watched interval feels long and an absorbing one feels short: absorption steals attention from the clock, starving the accumulator.

The attentional modulation has a measurable neural signature. When observers divide attention between timing a stimulus and judging another of its features, greater attention to time produces greater activity in a network including the supplementary motor area and the striatum, and the degree of that activity tracks how much the interval is subjectively lengthened (Coull et al., 2004). The attentional-gate elaboration of the clock model formalizes this by inserting an attention-controlled gate between the pacemaker and the accumulator, so that the flow of pulses is scaled by attention to time (Zakay & Block, 1997). Attention thus does not merely accompany timing; it sets the effective gain of the clock.

Dopamine and Clock Speed

If the pacemaker has a physical rate, then pharmacology should be able to change it, and the neurotransmitter most consistently implicated is dopamine. Dopaminergic drugs shift timing in the direction predicted if they alter pacemaker rate: dopamine agonists such as stimulants make animals and people behave as though an interval is longer than it is, consistent with a sped-up clock, while dopamine antagonists such as neuroleptics produce the opposite underestimation, consistent with a slowed clock (Meck, 1996). The effect is on the clock's rate rather than on memory or motor output, because it appears immediately and scales with the interval, which is the signature of a change in pulse rate.

The dopaminergic account also helps to carve timing into subsystems. Evidence that different drugs selectively affect timing in different ranges suggests that sub-second and supra-second timing are not one mechanism: pharmacological manipulations of dopamine influence timing in the seconds range, while timing in the hundreds-of-milliseconds range appears to depend on other systems (Rammsayer, 1999). This dissociation matters for theory, because it implies that the single-clock story is at best an account of cognitively controlled interval timing in the seconds-to-minutes window, and that faster timing is handled elsewhere, a theme the debate over dedicated versus intrinsic mechanisms takes up directly.

The Neural Basis

No single brain structure is the clock; interval timing depends on a distributed circuit centred on the basal ganglia and their cortical partners. Functional imaging shows that timing an interval recruits the striatum, the supplementary motor area, and the prefrontal cortex, with the basal ganglia engaged early as an interval begins and prefrontal and parietal regions supporting the working memory and decision demands of the task (Rao et al., 2001). The prominence of the cortico-striatal system in these studies motivated the leading neural model of the clock, in which timing is read from the coincident firing of cortical oscillators detected by striatal neurons (Matell & Meck, 2004).

The cerebellum contributes as well, but on a different timescale, and this division of labour is one of the field's more durable findings. The cerebellum is implicated primarily in the timing of brief, sub-second intervals, especially those tied to discrete movements, whereas the basal ganglia and cortex dominate longer, supra-second interval timing (Ivry & Spencer, 2004). This maps onto the pharmacological dissociation of the previous section: the sub-second cerebellar system and the supra-second cortico-striatal system are doubly dissociable across lesions, drugs, and tasks. Timing, on this evidence, is not one faculty but a family of mechanisms recruited according to the duration and the task.

Dedicated Clocks versus Intrinsic Timing

The internal-clock tradition assumes a dedicated timing mechanism, a system whose function is to measure time and whose output other processes read. A rival tradition denies this, holding that timing is intrinsic to the ordinary business of neural computation. On the intrinsic view there is no clock to consult; duration is encoded in the changing state of a recurrent network, which follows a trajectory through its space of possible states, so that the point reached along that trajectory itself carries information about how much time has elapsed (Karmarkar & Buonomano, 2007). Any circuit with the right dynamics can tell time as a by-product of processing, without a pacemaker or an accumulator.

The two views make different commitments and fit different data. Intrinsic, state-dependent models are natural for the sub-second, sensory timing that pervades perception and motor control, where a dedicated central clock seems extravagant, and they predict that timing should be local, modality-specific, and tied to the recent history of the circuit (Karmarkar & Buonomano, 2007). Dedicated pacemaker-accumulator and cortico-striatal models remain better suited to the longer, more cognitive intervals that survive across modalities and support explicit judgement (Matell & Meck, 2004). The contemporary position is not that one is right but that timing spans mechanisms: intrinsic circuit dynamics for the fast and sensory, dedicated cortico-striatal machinery for the slow and cognitive, with the boundary between them an active question.

Emotion and the Distortion of Duration

Felt duration is famously labile, and emotion is among its most powerful distorters. Emotional states bias time estimation systematically, and within the clock model they do so through two routes: arousal, which changes pacemaker rate, and attention, which changes how many pulses the switch admits (Droit-Volet & Meck, 2007). High-arousal states speed the clock and lengthen perceived duration, so that a brief threatening stimulus is judged to last longer than a neutral one of the same length, while states that capture attention away from time can shorten it. The direction of the distortion depends on which route dominates, which is why fear and other high-arousal emotions reliably dilate time.

These effects are not incidental curiosities but a test of the clock model's explanatory reach. That a single framework predicts both the arousal-driven dilation of time under threat and the attention-driven compression of time during absorbing activity, using the same two parameters that govern the pacemaker and the switch, is a substantial point in the model's favour (Droit-Volet & Meck, 2007). Emotional distortions of time also connect timing to the body, since the same arousal systems that drive them regulate heart rate and interoception, supporting the view that the felt passage of time is grounded partly in bodily signals rather than in an abstract clock alone (Wittmann, 2013).

Criticisms and Open Questions

The internal clock is a model, not a discovered organ, and its literal reading has drawn sustained criticism. No pacemaker neuron emitting timing pulses has been found, and the scalar property that the model was built to explain is not universal: careful measurement reveals systematic violations, with the coefficient of variation changing across the timed range rather than staying flat, which a strict scalar clock cannot accommodate (Wearden & Lejeune, 2008). These deviations suggest that the clean scalar signature is an approximation that holds over a limited range, and that the true mechanism generates it only under certain conditions.

A deeper criticism is that the pacemaker-accumulator architecture may be a useful abstraction rather than a description of the machinery, with the same behaviour produced by network dynamics that contain nothing resembling a pulse counter (Karmarkar & Buonomano, 2007). The field's own reviews acknowledge that no consensus mechanism has emerged and that timing is almost certainly distributed across multiple systems tuned to different scales (Buhusi & Meck, 2005). The honest summary is that time perception is well characterized behaviourally and still contested mechanistically: the scalar property, the attentional and dopaminergic modulations, and the cortico-striatal and cerebellar contributions are robust, but whether they are best explained by a dedicated clock, by intrinsic neural dynamics, or by some combination remains genuinely open (Grondin, 2010).

Worked Example

The second demonstration implements the pacemaker-accumulator clock with a reference rate of 10 pulses per second, the rate at which subjective time equals real time. Over a real interval of 4 seconds a normal clock accumulates 10 times 4, or 40 pulses, which read against the reference of 10 pulses per second yields a subjective duration of 4.00 seconds, so the estimate is accurate. If arousal or a stimulant speeds the pacemaker to 12 pulses per second, the same 4-second interval accumulates 48 pulses, which read against the unchanged reference gives 48 divided by 10, or 4.80 seconds, an overestimate of 20 percent. If a dopamine antagonist slows the pacemaker to 8 pulses per second, the interval accumulates 32 pulses and is judged as 3.20 seconds, an underestimate of 20 percent. The proportional distortion is the fingerprint of a change in clock rate.

The first demonstration shows the scalar property with a coefficient of variation fixed at 0.10. A target of 2 seconds then carries a standard deviation of 0.10 times 2, or 0.20 seconds; a target of 4 seconds carries 0.40 seconds; a target of 8 seconds carries 0.80 seconds. The absolute error grows with duration, but dividing each standard deviation by its target returns 0.10 in every case, and rescaling each distribution by its own target superimposes the three curves onto one. The third demonstration sets bisection anchors of 2 and 8 seconds. Their arithmetic mean is 5.00 seconds, but scalar timing predicts the bisection point at their geometric mean, the square root of 2 times 8, which is 4.00 seconds. At a probe of 4.00 seconds the modeled probability of a long response is 0.500, the defining property of the bisection point; at 3 seconds it is about 0.30 and at 5 seconds about 0.66, so the subjective midpoint sits well below the arithmetic mean, exactly as a ratio-based clock requires.

Discussion

Time perception occupies an unusual place among the senses precisely because there is nothing to sense. Every other perceptual chapter begins with a receptor and a stimulus; this one begins with their absence and must explain how a lawful percept arises anyway. The field's answer has been to treat time as inferred from neural change, to measure that inference with reproduction, discrimination, and bisection, and to organize the results around the scalar property and an internal clock. That programme has been remarkably productive: it predicts how attention, arousal, dopamine, and emotion distort felt duration, it identifies a cortico-striatal circuit for supra-second timing and a cerebellar one for sub-second timing, and it supplies a common currency in which pharmacology, imaging, and behaviour can be compared. Table 1 sets the principal paradigms side by side, with what each measures and the signature result it yields.

What remains unsettled is the mechanism. The internal clock is a model of impressive reach whose literal components have never been found, and the intrinsic-timing alternative reproduces much of the same behaviour without them. The likeliest resolution is that both capture part of the truth: sensory and motor timing in the sub-second range emerges from the intrinsic dynamics of local circuits, while cognitive interval timing in the seconds-to-minutes range is served by a dedicated cortico-striatal system that behaves, to a good approximation, like a pacemaker and an accumulator. Read this way, time perception is less a single sense than a coalition of mechanisms that the brain recruits according to the duration it must measure and the use to which the measurement will be put.

Glossary

Accumulator.

The store in the internal-clock model that collects pacemaker pulses during an interval; its running total is the raw measure of elapsed time.

Attentional gate.

A proposed attention-controlled valve between pacemaker and accumulator that scales the flow of pulses by how much attention is devoted to timing.

Bisection point.

The probe duration at which an observer is equally likely to call an interval short or long; the measured subjective midpoint of the range.

Coefficient of variation.

The standard deviation of timing estimates divided by their mean; its constancy across durations is the quantitative marker of scalar timing.

Dedicated timing.

The view that time is measured by a specialized mechanism whose function is timing and whose output other processes read.

Interval timing.

The estimation of durations from roughly hundreds of milliseconds to minutes, the range in which timing is deliberate and cognitively penetrable.

Intrinsic timing.

The view that duration is encoded in the evolving state of ordinary neural circuits, with no pacemaker or accumulator dedicated to time.

Pacemaker.

The hypothesized source of regular pulses in the internal clock; its rate sets the speed of subjective time.

Prospective timing.

Duration judgement made when the observer knows in advance that time is relevant and attends to it as it passes; driven by the internal clock.

Reference memory.

The long-term store in scalar expectancy theory that holds the pulse counts associated with remembered durations for comparison with the current count.

Retrospective timing.

Duration judgement made only after the fact, reconstructed from memory of the interval's contents rather than from a clock.

Scalar expectancy theory.

The dominant formal model of interval timing, comprising a pacemaker-accumulator clock, a reference memory, and a comparator, and built to reproduce the scalar property.

Scalar property.

The finding that timing variability grows in proportion to the interval timed, so that the coefficient of variation stays roughly constant; the timing analogue of Weber's law.

Striatal beat-frequency model.

A neural model in which striatal neurons detect the coincident firing of cortical oscillators, reading elapsed time from their beat pattern.

Temporal bisection.

A task in which observers classify probe durations as closer to a short or a long learned anchor, used to map subjective duration and locate the bisection point.

Key Researchers

John Gibbon. Mathematical psychologist at Columbia University and the New York State Psychiatric Institute; formulated scalar expectancy theory and established the scalar property as the central law of interval timing. Wikipedia) - Obituary

Warren H. Meck. Professor of Psychology and Neuroscience at Duke University; developed the neuropharmacology of timing, the dopaminergic clock, and the cortico-striatal beat-frequency model. ORCID - Memorial

John H. Wearden. Emeritus Professor of Psychology at Keele University; tested scalar timing in humans and documented both its conformity and its systematic violations. Faculty Page - Google Scholar

Dean V. Buonomano. Professor of Neurobiology and Psychology at the University of California, Los Angeles; developed state-dependent network models in which timing is intrinsic to neural dynamics rather than read from a dedicated clock. ORCID - Faculty Page - Google Scholar - Wikipedia

Sylvie Droit-Volet. Professor of Psychology at Universite Clermont Auvergne; established how emotion and arousal distort felt duration and how timing develops in childhood. ORCID - Faculty Page

Catalin V. Buhusi. Professor in the Neuroscience Program at Utah State University; co-authored the influential synthesis of the functional and neural mechanisms of interval timing. Faculty Page - Google Scholar

Frequently Asked Questions

What is time perception?
It is the subjective experience and estimation of duration, built by inference from neural activity because there is no sense organ that transduces time (Wittmann, 2013).

What is the scalar property of timing?
It is the finding that the variability of timing estimates grows in proportion to the interval timed, so the coefficient of variation stays roughly constant; it is the timing analogue of Weber's law (Gibbon, 1977).

What is the internal clock model?
It proposes that a pacemaker emits pulses that a switch gates into an accumulator during an interval, with the accumulated count read as the perceived duration (Treisman, 1963).

Why does an interval feel shorter during an absorbing activity?
A demanding concurrent task draws attention from timing, so the switch admits fewer pulses to the accumulator and the interval is underestimated in prospective judgements (Coull et al., 2004).

How does dopamine affect time perception?
Dopamine agonists speed the clock and lengthen perceived duration, while antagonists slow it and shorten perceived duration, consistent with a change in pacemaker rate (Meck, 1996).

What is the difference between prospective and retrospective timing?
Prospective timing is judged in advance using attention and the clock, whereas retrospective timing is reconstructed afterward from memory of the interval's contents (Zakay & Block, 1997).

Which brain regions support interval timing?
A cortico-striatal network including the basal ganglia, supplementary motor area, and prefrontal cortex supports supra-second timing, while the cerebellum times brief sub-second intervals (Ivry & Spencer, 2004).

Is there really a clock in the brain?
No pacemaker neuron has been found, and intrinsic-timing accounts reproduce much of the behaviour from ordinary network dynamics, so whether a dedicated clock exists remains debated (Karmarkar & Buonomano, 2007).

References