Dedicated and intrinsic models of time
perception
Richard B. Ivry
1,2
and John E. Schlerf
2
1
Department of Psychology, University of California, Berkeley, CA 94720, USA
2
Helen Wills Neuroscience Institute, University of California, Berkeley, CA 94720, USA
Two general frameworks have been articulated to
describe how the passage of time is perceived. One
emphasizes that the judgment of the duration of a
stimulus depends on the operation of dedicated neural
mechanisms specialized for representing the temporal
relationships between events. Alternatively, the repres-
entation of duration could be ubiquitous, arising from
the intrinsic dynamics of nondedicated neural mechan-
isms. In such models, duration might be encoded
directly through the amount of activation of sensory
processes or as spatial patterns of activity in a network
of neurons. Although intrinsic models are neurally plaus-
ible, we highlight several issues that must be addressed
before we dispense with models of duration perception
that are based on dedicated processes.
Perceiving the passage of time
Cognition is dynamic, with our perceptions, actions and
comprehension of the world unfolding over time. A gener-
ation ago, research on timing was limited, emphasizing the
study of behaviors marked by temporal regularities [1].
More recently, a renaissance has taken hold in the study of
time perception, with researchers addressing a broad
range of temporal phenomena. Behavioral studies have
revealed a host of puzzling effects in which our perception
of time is far from veridical [2]. Neuroscientists have
described how activity in single neurons varies with time
and how this might relate to psychophysical judgments [3–
5]. Theorists have asked how the dynamics of neural net-
works might encode temporal patterns in a reliable man-
ner [6–11].
As has long been noted by philosophers and psycholo-
gists, we lack a sensory system devoted to the sense of time.
Nonetheless, many percepts, and our actions in response to
these percepts, are acutely dependent on the precise
representation of time. Of course the terms ‘time’ and
‘temporal processing’ encompass a broad range of phenom-
ena, including simultaneity, temporal order and the per-
ception of duration. In this review we focus on the last of
these, addressing how the nervous system encodes infor-
mation concerning the duration of events in the range of
hundreds of milliseconds, the units of time that are especi-
ally relevant for immediate perception and the actions we
produce in relation to these events. In particular we focus
on a fundamental question that has defined much of the
recent discussion: is our perception of the passage of
time the consequence of dedicated, clock-like neural
mechanisms? Or is duration coded in an accessible manner
as an intrinsic and ubiquitous property of neural activity?
Dedicated models of temporal processing
Dedicated models of time perception are, at their core,
modular. As vision scientists speak of dedicated mechan-
isms for color or motion perception, modular models of time
perception entail some sort of specialized mechanism that
represents the temporal relationship between events. The
pacemaker-counter model is one example of a modular
system [12]. These two components define a clock with
an interval specified by the accumulation of inputs from a
pacemaker. Spectral models of timing constitute a second
example of a modular process. The phasic interactions of a
bank of oscillators [8,13] or the exploitation of differential
activity patterns in a set of delay lines [14,15] can define
different intervals. In dedicated models these representa-
tions are viewed as specializations, unique to particular
neural structures, that provide a functional chronotopy
that is recruited across diverse task domains.
One motivation for dedicated models comes from the
observation that our sense of the passage of time appears to
transcend the sensory modality of a stimulus. We can
compare the duration of a tone to a light (although not
as well as we assume [16–18]) or metrically reproduce the
duration of a visual stimulus with a keypress. Such inter-
actions are less apparent in other perceptual domains; for
example, only rare individuals describe the color of a tone.
The facile manner with which we compare time across
different modalities suggests some sort of internal clock.
Behavioral data provide additional motivation. Individ-
ual differences in temporal acuity correlate between per-
ception and action [19]. Measures of variability or
dispersion are proportional to mean duration, and when
the tasks are appropriately matched this ratio is similar for
perception and action [20]. Based on the assumption that
this property arises from signal-dependent noise in a
common system, these results point towards a dedicated
system for timing.
A neural instantiation of a dedicated model is the
cerebellar timing hypothesis [21]. Patients with cerebellar
pathology are impaired on a range of tasks that require
precise timing, including perceptual tasks such as judging
the duration of brief tones [22,23] or categorizing speech
sounds that vary in the duration of a silent period [24]. The
timing hypothesis also provides a principled basis for
specifying the cerebellar contribution to sensorimotor
learning: this structure would be essential when learning
Review
1364-6613/$ – see front matter ß2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2008.04.002 Available online 6 June 2008 273
