Martin Wiener

Minimizing Within-Experiment and Within-Group Effects in Activation Likelihood Estimation Meta-Analyses

Activation Likelihood Estimation (ALE) is an objective, quantitative technique for coordinate‐based meta‐analysis (CBMA) of neuroimaging results that has been validated for a variety of uses. Stepwise modifications have improved ALEs theoretical and statistical rigor since its introduction. Here, we evaluate two avenues to further optimize ALE. First, we demonstrate that the maximum contribution of an experiment makes to an ALE map is related to the number of foci it reports and their proximity. We present a modified ALE algorithm that eliminates these within‐experiment effects. However, we show that these effects only account for 2–3% of cumulative ALE values, and removing them has little impact on thresholded ALE maps. Next, we present an alternate organizational approach to datasets that prevents subject groups with multiple experiments in a dataset from influencing ALE values more than others. This modification decreases cumulative ALE values by 7–9%, changes the relative magnitude of some clusters, and reduces cluster extents. Overall, differences between results of the standard approach and these new methods were small. This finding validates previous ALE reports against concerns that they were driven by within‐experiment or within‐group effects. We suggest that the modified ALE algorithm is theoretically advantageous compared with the current algorithm, and that the alternate organization of datasets is the most conservative approach for typical ALE analyses and other CBMA methods. Combining the two modifications minimizes both within‐experiment and within‐group effects, optimizing the degree to which ALE values represent concordance of findings across independent reports. Hum Brain Mapp, 2012.

The image of time: A voxel-wise meta-analysis

Although there has been an explosion of interest in the neural correlates of time perception during the past decade, substantial disagreement persists regarding the structures that are relevant to interval timing. We addressed this important issue by conducting a comprehensive, voxel-wise meta-analysis using the activation likelihood estimation algorithm; this procedure models each stereotactic coordinate as a 3D Gaussian distribution, then tests the likelihood of activation across all voxels in the brain (Turkeltaub et al., 2002). We included 446 sets of activation foci across 41 studies of timing that report whole-brain analyses. We divided the data set along two dimensions: stimulus duration (sub- vs. supra-second) and nature of response (motor vs. perceptual). Our meta-analyses revealed dissociable neural networks for the processing of duration with motor or perceptual components. Sub-second timing tasks showed a higher propensity to recruit sub-cortical networks, such as the basal ganglia and cerebellum, whereas supra-second timing tasks were more likely to activate cortical structures, such as the SMA and prefrontal cortex. We also detected a differential pattern of activation likelihood in basal ganglia structures, depending on the interval and task design. Finally, a conjunction analysis revealed the SMA and right inferior frontal gyrus as the only structures with significant voxels across all timing conditions. These results suggest that the processing of temporal information is mediated by a distributed network that can be differentially engaged depending on the task requirements.

Double dissociation of dopamine genes and timing in humans

A number of lines of evidence implicate dopamine in timing [Rammsayer, T. H. Neuropharmacological approaches to human timing. In S. Grondin (Ed.), Psychology of time (pp. 295–320). Bingley, UK: Emerald, 2008; Meck, W. H. Neuropharmacology of timing and time perception. Brain Research, Cognitive Brain Research, 3, 227–242, 1996]. Two human genetic polymorphisms are known to modulate dopaminergic activity. DRD2/ANKK1-Taq1a is a D2 receptor polymorphism associated with decreased D2 density in the striatum [Jönsson, E. G., Nothen, M. M., Grunhage, F., Farde, L., Nakashima, Y., Propping, P., et al. Polymorphisms in the dopamine D2 receptor gene and their relationships to striatal dopamine receptor density of healthy volunteers. Molecular Psychiatry, 4, 290–296, 1999]; COMT Val158Met is a functional polymorphism associated with increased activity of the COMT enzyme such that catabolism of synaptic dopamine is greater in pFC [Meyer-Lindenberg, A., Kohn, P. D., Kolachana, B., Kippenhan, S., McInerney-Leo, A., Nussbaum, R., et al. Midbrain dopamine and prefrontal function in humans: Interaction and modulation by COMT genotype. Nature Neuroscience, 8, 594–596, 2005]. To investigate the role of dopamine in timing, we genotyped 65 individuals for DRD2/ANKK1-Taq1a, COMT Val158Met, and a third polymorphism, BDNF Val66Met, a functional polymorphism affecting the expression of brain-derived neurotrophic factor [Egan, M. F., Kojima, M., Callicott, J. H., Goldberg, T. E., Kolachana, B. S., Bertolino, A., et al. The BDNF val66met polymorphism affects activity-dependent secretion of BDNF and human memory and hippocampal function. Cell, 112, 257–269, 2003]. Subjects were tested on a temporal discrimination task with sub- and supra-second intervals (500- and 2000-msec standards) as well as a spontaneous motor tempo task. We found a double dissociation for temporal discrimination: the DRD2/ANKK1-Taq1a polymorphism (A1+ allele) was associated with significantly greater variability for the 500-msec duration only, whereas the COMT Val158Met polymorphism (Val/Val homozygotes) was associated with significantly greater variability for the 2000-msec duration only. No differences were detected for the BDNF Vall66Met variant. Additionally, the DRD2/ANKK1-Taq1a polymorphism was associated with a significantly slower preferred motor tempo. These data provide a potential biological basis for the distinctions between sub- and supra-second timing and suggest that BG are integral for the former whereas pFC is implicated in the latter.

Implicit timing activates the left inferior parietal cortex

Coull and Nobre (2008) suggested that tasks that employ temporal cues might be divided on the basis of whether these cues are explicitly or implicitly processed. Furthermore, they suggested that implicit timing preferentially engages the left cerebral hemisphere. We tested this hypothesis by conducting a quantitative meta-analysis of eleven neuroimaging studies of implicit timing using the activation-likelihood estimation (ALE) algorithm (Turkeltaub, Eden, Jones, & Zeffiro, 2002). Our analysis revealed a single but robust cluster of activation-likelihood in the left inferior parietal cortex (supramarginal gyrus). This result is in accord with the hypothesis that the left hemisphere subserves implicit timing mechanisms. Furthermore, in conjunction with a previously reported meta-analysis of explicit timing tasks, our data support the claim that implicit and explicit timing are supported by at least partially distinct neural structures.

Interval timing disruptions in subjects with cerebellar lesions

The cerebellum has long been implicated in time perception, particularly in the subsecond range. The current set of studies examines the role of the cerebellum in suprasecond timing, using analysis of behavioral data in subjects with cerebellar lesions. Eleven cerebellar lesion subjects and 17 controls were tested on temporal estimation, reproduction and production, for times ranging from 2 to 12s. Cerebellar patients overproduced times on both the reproduction and production tasks; the effect was greatest at the shortest duration. A subset of patients also underestimated intervals. Cerebellar patients were significantly more variable on the estimation and reproduction tasks. No significant differences between normal and cerebellar patients were found on temporal discrimination tasks with either sub- or suprasecond times. Patients with damage to the lateral superior hemispheres or the dentate nuclei showed more significant impairments than those with damage elsewhere in the cerebellum, and patients with damage to the left cerebellum had more significant differences from controls than those with damage to the right. These data suggest that damage to the middle-to-superior lobules or the left hemisphere is especially detrimental to timing suprasecond intervals. We suggest that this region be considered part of a network of brain structures including the DLPFC that is crucial for interval timing.

Multiple mechanisms for temporal processing

Many models suggest that time perception is mediated by a unitary mechanism. For example, scalar expectancy theory (SET), the dominant model of timing for the past 30 years, suggests that temporal processing is mediated by a centralized clock-counter module in which elapsed time is measured by the summation of pacemaker pulses (Gibbon et al., 1984). A number of alternative, neurally plausible models have been proposed with clock processes that incorporate either the pacemaker-counter elements of SET, or other neural dynamics such as decay processes or state-dependent network activity (Staddon and Higa, 1999; Karmarkar and Buonomano, 2007; Simen et al., 2011a,b). While these models differ in the mechanisms utilized for the temporal control of behavior, they all suggest that timing is accomplished by a single, amodal process. Support for the hypothesis that timing is mediated by a single mechanism comes from several sources. A number of studies demonstrate that performance is independent of whether the task utilizes motor or “perceptual” temporal representations (Ivry and Hazeltine, 1995; Meegan et al., 2000). Additionally, although an effect of interval duration has been postulated for over a hundred years, such an effect has not been consistently identified; Lewis and Miall (2009), for example, failed to identify a fundamental change in timing performance or “break-point” using stimuli ranging from 68 ms to 16.7 min. We suggest the alternative hypothesis that timing functions are mediated by multiple, overlapping neural systems, which may be flexibly engaged depending on the task requirements. These systems may function independently of one another and may be adaptively engaged pro re nata, such that single or multiple systems may be active during any one timing task, depending on environmental conditions and behavioral requirements. One line of support for this hypothesis comes from a quantitative meta-analysis of 41 neuroimaging studies of time perception in which we found that different neural structures were engaged depending on stimulus duration and the “motor” or “perceptual” nature of the task (Wiener et al., 2010a). Of particular interest in this context, however, is the fact that the meta-analysis also demonstrated two areas engaged across all tasks: supplementary motor area (SMA) and right inferior frontal gyrus (rIFG). In subsequent analyses of this dataset, however, we found that even in regions active across several conditions there is evidence of multiple timing mechanisms at work. Consider the SMA for example. Recent observations suggest that the SMA is a heterogeneous structure that may be functionally divided into the SMA “proper” and pre-SMA (Nachev et al., 2008). A rostro-caudal gradient in the SMA has been proposed according to which SMA and pre-SMA subserve motor and cognitive processes, respectively. Consistent with this finding, we found evidence for a functional gradient in the SMA, wherein perceptual timing tasks are more likely to activate voxels within the pre-SMA while motor timing tasks are associated with SMA proper activation-likelihood (Figure ​(Figure11A). Figure 1 A subset of the results from our previous meta-analysis of neuroimaging timing studies. (A) Sagittal section of a rendered brain including SMA voxels from perceptual or motor timing tasks (regardless of duration length) and their overlap. Crosshairs are ... Fractionation of temporal processing may also be evident in the basal ganglia, a brain region often implicated in studies of time perception and with high connectivity to the SMA. Figure ​Figure1B1B depicts voxels from SMA and basal ganglia regions with significant activation-likelihood. Once again, different patterns of activation-likelihood were noted as a function of the duration of the stimulus and nature of the task. For example, there was a greater propensity for the basal ganglia to be activated during sub-second timing tasks. However, it is crucial to note that the basal ganglia interact with numerous other regions, and so these activation patterns must be considered in the larger context of interactive networks. Additional work beyond neuroimaging also argues for multiple timing systems. For example, we recently adopted a behavioral genetics paradigm to look at single-nucleotide polymorphisms in genes associated with different aspects of the dopamine system (Wiener et al., 2011). We found that a polymorphism affecting the expression of striatal D2 receptors was associated with poorer performance on a perceptual timing task, but only when the intervals tested were below 1 s. In contrast, subjects with a polymorphism affecting the expression of the enzyme catechol-O-methyltransferase (COMT), which is known to regulate prefrontal dopamine tone, were impaired during supra-second, but not sub-second timing. This work suggests that different dopaminergic systems may underlie distinct timing procedures. Another line of data supporting the claim that multiple mechanisms mediate timing comes from the fact that at least under some circumstances timing mechanisms appear to be both modality-specific and mediated by local neural structures. For example, adaptation to focal regions of the visual field produces duration distortions that are localized to that spatial region (Burr et al., 2007). Interestingly, modality-specific regions appear to be invoked for temporal expectations even in the absence of the stimuli themselves (Bueti and Macaluso, 2010), suggesting that the process may be mediated by simulation. The fact that subject strategies influence the neural circuits recruited for timing is also consistent with the hypothesis that multiple distinct procedures underlie timing. For example, a recent study demonstrated that subjects recruited different neural networks depending on whether they implicitly used a beat-based or duration-based strategy (Grahn and McAuley, 2009). Similarly, recordings from rodent striatum demonstrate that patterns of temporally varying neural activity may reflect an integration of the passage of time with its associated action (Portugal et al., 2011), further suggesting that the computations contributing to temporal control may critically depend on both environmental and behavioral context. The hypothesis that timing may be mediated by multiple distinct procedures also accounts for the puzzling lack of neurologic disorders characterized by a profound and selective impairment in temporal processing. Although syndromes characterized by selective deficits in vision, audition, language, attention, and multiple other faculties have been identified, we are unaware of a similar disorder involving temporal processing. Additionally, studies of patients and animals with brain lesions often demonstrate relatively mild deficits in temporal processing. The above discussion is not intended to be exhaustive. Differences in performance on tasks assessing timing for synchronized or syncopated beat timing (Jantzen et al., 2004), as well as explicit or implicit timing to temporal intervals (Coull and Nobre, 2008; Wiener et al., 2010b) have also been identified. A challenge for future research will be to identify these different timing networks and to clarify the functional relationship between them.

Parietal influence on temporal encoding indexed by simultaneous transcranial magnetic stimulation and electroencephalography.

Previous studies have suggested that contingent negative variation (CNV), as recorded by electroencaphalography (EEG), may serve as an index of temporal encoding. The interpretation of these studies is complicated by the fact that, in a majority of studies, the CNV signal was obtained at a time when subjects were not only registering stimulus duration but also making decisions and preparing to act. Previously, we demonstrated that repetitive transcranial magnetic stimulation (rTMS) of the right supramarginal gyrus (rSMG) in humans lengthened the perceived duration of a visual stimulus (Wiener et al., 2010a), suggesting the rSMG is involved in basic encoding processes. Here, we report a replication of this effect with simultaneous EEG recordings during the encoding of stimulus duration. Stimulation of the rSMG led to an increase in perceived duration and the amplitude of N1 and CNV components recorded from frontocentral sites. Furthermore, the size of the CNV amplitude, but not N1, positively correlated with the size of the rTMS effect but negatively correlated with bias (the baseline tendency to report a comparison stimulus as shorter), suggesting that the CNV indexes stimulus duration. These results suggest that a feedforward mechanism from parietal to prefrontal regions mediates temporal encoding and demonstrate a dissociation between early and late phases of encoding processes.

Fast forward: Supramarginal gyrus stimulation alters time measurement

The neural basis of temporal processing is unclear. We addressed this important issue by performing two experiments in which repetitive transcranial magnetic stimulation (rTMS) was administered in different sessions to the left or right supramarginal gyrus (SMG) or vertex; in both tasks, two visual stimuli were presented serially and subjects were asked to judge if the second stimulus was longer than the first (standard) stimulus. rTMS was presented on 50% of trials. Consistent with a previous literature demonstrating the effect of auditory clicks on temporal judgment, rTMS was associated with a tendency to perceive the paired visual stimulus as longer in all conditions. Crucially, rTMS to the right SMG was associated with a significantly greater subjective prolongation of the associated visual stimulus in both experiments. These findings demonstrate that the right SMG is an important element of the neural system underlying temporal processing and, as discussed, have implications for neural and cognitive models of temporal perception and attention.

Individual differences in the morphometry and activation of time perception networks are influenced by dopamine genotype

Individual participants vary greatly in their ability to estimate and discriminate intervals of time. This heterogeneity of performance may be caused by reliance on different time perception networks as well as individual differences in the activation of brain structures utilized for timing within those networks. To address these possibilities we utilized event-related functional magnetic resonance imaging (fMRI) while human participants (n=25) performed a temporal or color discrimination task. Additionally, based on our previous research, we genotyped participants for DRD2/ANKK1-Taq1a, a single-nucleotide polymorphism associated with a 30-40% reduction in striatal D2 density and associated with poorer timing performance. Similar to previous reports, a wide range of performance was found across our sample; crucially, better performance on the timing versus color task was associated with greater activation in prefrontal and sub-cortical regions previously associated with timing. Furthermore, better timing performance also correlated with increased volume of the right lateral cerebellum, as demonstrated by voxel-based morphometry. Our analysis also revealed that A1 carriers of the Taq1a polymorphism exhibited relatively worse performance on temporal, but not color discrimination, but greater activation in the striatum and right dorsolateral prefrontal cortex, as well as reduced volume in the cerebellar cluster. These results point to the neural bases for heterogeneous timing performance in humans, and suggest that differences in performance on a temporal discrimination task are, in part, attributable to the DRD2/ANKK1 genotype.

Temporal discrimination of sub- and suprasecond time intervals: a voxel-based lesion mapping analysis.

We used voxel-based lesion-symptom mapping (VLSM) to determine which brain areas are necessary for discriminating time intervals above and below 1 s. VLSM compares behavioral scores of patients that have damage to a given voxel to those that do not on a voxel-by-voxel basis to determine which voxels are critical for the given behavior. Forty-seven subjects with unilateral hemispheric lesions performed a temporal discrimination task in which a standard stimulus was compared on each trial to a test stimulus. In different blocks of trials, standard stimuli were either 600 or 2000 ms. Behavioral measures included the point of subjective equality, a measure of accuracy, and the coefficient of variation, a measure of variability. Lesions of the right middle and inferior frontal gyri were associated with decrements in performance on both durations. In addition, lesions of the left temporal lobe and right precentral gyrus were associated exclusively with impaired performance for subsecond stimuli. In line with results from other studies, these data suggest that different circuits are necessary for timing intervals in these ranges, and that right frontal areas are particularly important to timing.