The image of time: A voxel-wise meta-analysis
Introduction
Numerous neuroimaging studies investigating the neural correlates of time perception in the range of milliseconds to minutes have been published in the last decade. The results of these studies have led to a number of different theories about the roles of particular neural structures in time perception. For example, some investigators have emphasized the role of the basal ganglia in temporal processing (e.g., Buhusi and Meck, 2005), whereas others suggest a preferential role of the right prefrontal cortex (Lewis and Miall, 2006a, Lewis and Miall, 2006b). Still other accounts have assigned essential functions to the cerebellum (Ivry et al., 2002), supplementary motor area (SMA; Macar et al., 2006), right inferior parietal lobe (IPL; Bueti and Walsh, 2009), and insular cortex (Craig, 2009). Substantial differences in the nature of the tasks employed, the duration of the stimuli to be timed, the nature of the response and the baseline or control condition complicate comparisons between these studies and may account for the lack of consistency. In order to fully characterize the manner in which brain activity relates to temporal perception, quantitative methods may be necessary to search for concordance among neuroimaging studies. Here we present a quantitative, voxel-wise meta-analysis of neuroimaging data designed to determine regions of concordance in the current literature on time perception.
Several factors may contribute to the variability in imaging studies reported to date. One factor may be the wide variety of tasks that have been employed. A dimension on which timing tasks vary is the degree to which they engage “motor” or “perceptual” timing routines. For motor timing tasks, the interval to be timed is defined by a motor response. The most commonly used task among motor timing studies is paced finger tapping, in which subjects are typically required to tap a response key in time with an isochronous metronome at a given rate (synchronization), and then continue tapping at the same rate without any external stimulation (continuation). In some studies, no continuation phase is employed, while in others the subject is required to tap at the midpoint between sets of pacing stimuli (syncopation). Another motor timing task is interval production/reproduction, in which subjects produce an interval from memory, by indicating in a stopwatch-like fashion when the interval begins and ends. A common variation in production/reproduction is the type of memory utilized, as investigators may require subject to produce an interval that was experienced recently (reproduction), or produce a prescribed duration (i.e., “produce 7 s”). Perceptual timing tasks, in contrast, require subjects to make judgments about temporal intervals. A task frequently employed in neuroimaging studies of perceptual timing is temporal discrimination, in which subjects compare the duration of successive stimuli by indicating whether the second stimulus is longer than or differs from the first.
A second factor that may contribute to variability is the duration of the stimulus to be timed. Neuroimaging studies have utilized test intervals ranging from 200 ms to 24 s. A number of distinct lines of evidence suggest that mechanisms and brain regions recruited in timing tasks differ as a function of the stimulus duration (Mauk and Buonomano, 2004, Buhusi and Meck, 2005, Lewis and Miall, 2003b). Evidence from neuropsychological studies in humans and investigations in animals suggest that timing of intervals above and below 1 s relies on different procedures (Harrington and Haaland, 1999, Buhusi and Meck, 2005, Gibbon et al., 1997). Lewis and Miall (2003b) suggested that sub-second intervals are embedded in motor action plans, while supra-second intervals require greater cognitive control. Additionally, Mauk and Buonomano (2004) suggest that the timing of sub-second stimuli may rely on sensory processes, wherein state-dependent networks may encode changes in duration (Karmarkar and Buonomano, 2007). Rammsayer (1997) proposed that timing of different durations may be dependent on different neurotransmitter systems; he reported, for example, that haloperidol, a non-specific D2 dopamine-receptor antagonist, altered timing performance in humans across both sub-second and supra-second ranges, whereas remoxipride, a D2 dopamine-receptor antagonist which primarily acts on mesolimbocortical projections (Gerlach and Casey, 1990), altered only supra-second performance.
Several reviews to date have attempted to characterize activations across the corpus of neuroimaging studies of timing. Macar and colleagues (2002) divided neuroimaging studies into rhythmic and perceptual categories; only activations resulting from control task subtractions were considered. The authors concluded that the basal ganglia (caudate and putamen), SMA, cerebellum, dorsolateral prefrontal cortex (DLPFC, Brodmann area [BA] 9/46), anterior cingulate and right IPL (BA 40) were often active across all timing tasks, suggesting that these regions formed the core network of time perception in the brain. However, in addition to studies employing positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) techniques, the authors also included results from studies measuring surface Laplacians, electroencephalography (EEG) and scalp current densities. One potential weakness of the latter studies is that EEG and related techniques do not readily interrogate sub-cortical structures and may thus underestimate the role that these structures play in timing.
Lewis and Miall (2003b) reported a label-based meta-analysis of neuroimaging of timing. The authors considered three factors in partitioning the datasets for their analysis: interval duration (sub-second or supra-second intervals), nature of the response (motor vs. non-motor), and whether the motor response was continuous or discontinuous. The results taken from each study were from the least restrictive contrasts, the majority of which were [task – rest]. The results of their meta-analysis suggest different neural networks are recruited, depending on the task constraints. Sub-second tasks commonly activated the bilateral SMA, left sensorimotor cortex, basal ganglia and thalamus, as well as the right cerebellum, superior temporal gyrus and lateral premotor cortex. Supra-second tasks frequently activated the left cerebellum, as well as the bilateral prefrontal and parietal cortices, with the right DLPFC demonstrating more common activation than any other structure. Among all studies, the SMA and right cerebellum were the most commonly activated structures.
Two recent label-based reviews have also attempted to characterize common activations across timing tasks. Meck, Penney and Pouthas (2008) reviewed a number of neuroimaging studies, concluding that the basal ganglia serve as a core timer across sub-second and supra-second timing tasks. Penney and Vaitilingam (2008) reviewed a larger corpus of neuroimaging timing studies; their review also divided experiments into sub-second and supra-second categories. The authors concluded that sub-second timing tasks most commonly activated the cerebellum, DLPFC, inferior frontal gyrus (IFG), superior temporal gyrus, premotor cortex, insula (BA 13), cingulate, pre-SMA, SMA, basal ganglia and thalamus, with the cerebellum being the most frequently activated. Supra-second timing tasks commonly activated the IFG, DLPFC, supramarginal gyrus (SMG), superior temporal gyrus, premotor cortex, cingulate, pre-SMA, middle temporal gyrus, basal ganglia and thalamus, with the IFG being the most frequently activated. However, this review included both activations subtracted from control tasks and from rest; as subtractions from rest do not control for other cognitive processes, activations in these studies may be related to a variety of cognitive operations in addition to timing procedures. Furthermore, the authors did not include any imaging studies employing paced finger tapping in their review.
Although these reviews have generated important and provocative findings, inconsistency across the studies persists, perhaps reflecting the fact that different studies were included, and discrepant procedures were employed to identify sites and extents of activation. Finally, although label-based reviews and meta-analyses may provide an overview of neuroimaging results, they do not provide any quantitative measure of the probability that a structure will be activated (Laird et al., 2005b).
With the dramatic increase in functional neuroimaging studies, the need has arisen for quantitative methods to evaluate cross-study results. Activation likelihood estimation (ALE) is a powerful, well-validated technique for conducting meta-analyses that circumvents a number of problems inherent in previous techniques. Developed concurrently yet independently by Turkeltaub and others (2002) and Chein and others (2002), ALE is a quantitative technique for conducting a voxel-wise analysis of cross-study data.
The ALE technique involves several steps. First, data is pooled across the set of selected studies; the data consists of the activation peaks reported in each study. Next, each activation peak is modeled as a 3D Gaussian probability distribution in the brain; the width of the Gaussian models the uncertainty in localization. The overlap among these distributions is used to estimate the probability that at least one of the peaks from the literature should have fallen within a given voxel (ALE score). A permutation test is then conducted by computing voxel-wise ALE scores for randomly generated lists of activation peaks. The ALE scores from the literature analysis are compared to those from the random sets of foci to calculate the statistical significance of each ALE score. Since the inception of ALE, a number of improvements have been made to increase the flexibility and utility of the algorithm. Many of these improvements were borrowed from techniques used for the analysis of functional neuroimaging data, and include the introduction of false-discovery rate (FDR) thresholds and cluster analysis (Laird et al., 2005a).
While the ALE technique is a quantitative technique for characterizing large sets of neuroimaging data, several weaknesses also exist. First, the size and shape of activation clusters is not considered; the ALE algorithm determines the spread of activation as a user-defined full-width at half-maximum (FWHM) Gaussian. Second, the number of subjects tested in each study is not considered. Finally, the ALE technique does not consider the relative contribution of each study to each ALE value; because of this limitation, particularly strong data from a single study may identify voxels that are judged to be “significant” despite the fact that they are not activated in the vast majority of studies included in the meta-analysis. We attempt to correct for some of these weaknesses with a new masking technique, described in more detail below.
Section snippets
Activation likelihood estimation
The ALE algorithm models the probability that a particular focus will be located in a given voxel, using 3-dimensional Gaussian distributions. The probability that a focus should have been identified at a given location is calculated aswhere d is the Euclidean distance from the center of the voxel to the focus and σ is the standard deviation of a Gaussian distribution. The probability value, which ranges between 0 and 1, is multiplied by 8 mm3, to give the cumulative
Sub-second motor timing
High areas of concordance for tasks involving sub-second intervals, as evaluated by a motor task, were found in the bilateral SMA, left middle frontal gyrus (BA 6) and precentral gyrus, right putamen and lateral globus pallidus, left thalamus, claustrum and posterior cerebellum. Additional clusters were found in the right IFG (BA 47), left middle frontal gyrus (BA 10), right posterior cerebellum, right middle frontal gyrus (BA 6), substantia nigra, IPL (BA 40), left superior temporal gyrus,
Discussion
The results of our meta-analysis revealed a number of distinct regions associated with the processing of temporal intervals. Furthermore, the likelihood that particular structures were activated depended on the duration of the interval to be timed and whether the task required predominantly motor or perceptual processing. Only the right IFG and bilateral SMA demonstrated significant activation likelihood across all conditions. These results support a number of conclusions. First, partially
Conclusions
The present study provides the first voxel-wise meta-analysis of the neuroimaging literature on time perception. Although our findings reinforce a number a previous observations about the roles of brain structures such as the right IFG and the SMA, they provide a more nuanced perspective on the contributions of structures such as the cerebellum and basal ganglia to timing. Our results suggest that different neural regions are recruited between sub-second and supra-second intervals, with
Acknowledgement
The authors would like to thank Dr. Matthew Matell for his helpful comments on the data presented in this manuscript.
References (130)
- et al.
The contribution of the insula to motor aspects of speech production: a review and a hypothesis
Brain Lang.
(2004) - et al.
Distributed neural systems for temporal production: a functional MRI study
Brain Res. Bull.
(2003) - et al.
The neuroanatomical substrate of sound duration discrimination
Neuropsocyhologia
(2002) - et al.
Topographic segregation of corticostriatal projections from the posterior parietal subdivisions in the macaque monkey
Neuroscience
(1991) - et al.
Functional heterogeneity within Broca's area during verbal working memory
Physiol. Behav.
(2002) - et al.
Cognitive timing: neuropsychology and anatomic basis
Brain Res.
(2009) - et al.
Basal ganglia and supplementary motor area subtend duration perception: an fMRI study
Neuroimage
(2003) - et al.
Remoxipride, a new selective D2 antagonist, and haloperidol in cebus monkeys
Prog. Neuropsychopharmacol. Biol. Psychiatry
(1990) - et al.
Toward a neurobiology of temporal cognition: advances and challenges
Curr. Opin. Neurobiol.
(1997) - et al.
The cognitive functions of the caudate nucleus
Prog. Neurobiol.
(2008)
Neural representation of interval encoding and decision making
Cog. Brain Res.
Phonological processing during language production: fMRI evidence for a shared production-comprehension network
Brain Res. Cogn. Brain Res.
Frontal–striatal circuitry activated by human peak-interval timing in the supra-seconds range
Cog. Brain Res.
Neural systems supporting timing and chronometric counting: an fMRI study
Brain Res. Cogn. Brain Res.
Cortical activations during paced finger-tapping applying visual and auditory pacing
Brain Res. Cogn. Brain Res.
Functional MRI reveals the existence of modality and coordination-dependent timing networks
Neuroimage
A parametric fMRI investigation of context effects in sensorimotor timing and coordination
Neuropsychologia
Cortical involvement in temporal reproduction: evidence for differential roles of the hemispheres
Neuropsychologia
Timing in the absence of clocks: encoding time in neural network states
Neuron
Impaired reproduction of second but not millisecond time intervals in Parkinson's disease
Neuropsychologia
The basic pattern of activation in motor and sensory temporal tasks: positron emission tomography data
Neurosci. Lett.
Brain activation patterns during measurement of sub- and supra-second intervals
Neuropsychologia
Distinct systems for automatic and cognitively controlled time measurement: evidence from neuroimaging
Curr. Opin. Neurobiol.
Remembering the time: a continuous clock
Trends Cogn. Sci.
A right hemispheric prefrontal system for cognitive time measurement
Behav. Proc.
Brain activity correlates differentially with increasing temporal complexity of rhythms during initialization, synchronization, and continuation phases of paced finger tapping
Neuropsychologia
Time perception: manipulation of task difficulty dissociates clock functions from other cognitive demands
Neuropsychologia
Timing functions of the supplementary motor area: an event-related fMRI study
Brain Res. Cogn. Brain Res.
Dissociable contributions of the prefrontal and neocerebellar cortex to time perception
Cog. Brain Res.
Brain activation induced by estimation of duration: a PET study
Neuroimage
Phonological agraphia following a focal anterior insulo-opercular infarction
Neuropsychologia
Cortico–striatal circuits and interval timing: coincidence detection of oscillatory processes
Brain Res. Cog. Brain Res.
Cortico-striatal representation of time in animals and humans
Curr. Opin. Neurobiol.
Perceptual timing in cerebellar degeneration
Neuropsychologia
Representation of time intervals in the right posterior parietal cortex: implications for a mental time line
Neuroimage
Parallel organization of functionally segregated circuits linking basal ganglia and cortex
Effector-independent voluntary timing: behavioural and neuroimaging evidence
Eur. J. Neurosci.
Visual feedback about time estimation is related to a right hemisphere activation measured by PET
Exp. Brain Res.
The parietal cortex and the representation of time, space, number and other magnitudes
Phil. Trans. R Soc. B
Sensory and association cortex in time perception
J. Cog. Neurosci.
Different brain circuits underlie motor and perceptual representations of temporal intervals
J. Cog. Neurosci.
What makes us tick? Functional and neural mechanisms of interval timing
Nat. Rev. Neurosci.
Effects of divided attention on timing in patients with lesions of the cerebellum or frontal lobe
Neuropsychology
Functional anatomy of the attentional modulation of time estimation
Science
Timing, storage, and comparison of stimulus duration engage discrete anatomical components of a perceptual timing network
J. Cog. Neurosci.
Emotional moments across time: a possible neural basis for time perception in the anterior insula
Phil. Trans. R Soc. B
Neural systems supporting interoceptive awareness
Nat. Neurosci.
Role of the cerebellum in externally paced rhythmic finger movements
J. Neurophysiol.
Semantic encoding and retrieval in the left inferior prefrontal cortex: a functional MRI study of task difficulty and process specificity
J. Neurosci.
Cited by (448)
Time perception in stimulant-dependent participants undergoing repetitive transcranial magnetic stimulation
2024, Behavioural Brain Research