Selective Attention to Temporal Features on Nested Time Scales

Cerebral Cortex Advance Access published August 26, 2013 Cerebral Cortex doi:10.1093/cercor/bht240

Selective Attention to Temporal Features on Nested Time Scales Molly J. Henry, Björn Herrmann and Jonas Obleser Max Planck Research Group “Auditory Cognition”, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany Address correspondence to Dr Molly J. Henry, Max Planck Research Group “Auditory Cognition,” Max Planck Institute for Human Cognitive and Brain Sciences, Stephanstrasse 1a, Leipzig, Germany 04103. Email: [email protected]

Keywords: attention to time, auditory perception, fMRI, time perception Introduction Making sense of the acoustic environment is an inherently temporal process, as all acoustic events are necessarily extended in time. Although listeners rarely judge the temporal properties of auditory events explicitly (outside the laboratory, that is), such temporal cues play an implicit yet undisputed role in speech and music perception. A critical observation is that auditory events typically carry information simultaneously at multiple time scales. Thus, an important research question is how listeners selectively attend to auditory stimulus features unfolding at different rates. For example, natural speech contains temporal cues to manner of articulation or voicing which are embedded within more slowly unfolding durational cues to word-level stress (Rosen 1992). In the current study, we examined selective attention to the total duration of a stimulus (on average 500 ms, 2 Hz) versus amplitude modulations occurring at a somewhat faster rate (on average 8 Hz, 125 ms). In the visual domain, feature-selective attention has been dissected by requiring observers to attend to one feature of a stimulus, while simultaneously ignoring another (Chawla et al. 1999; Saenz et al. 2002; Schoenfeld et al. 2007). Correspondingly, one of the most common approaches to studying the neural mechanisms underlying time perception and attention to time involves comparing brain activation during duration judgments to activation during judgments about an orthogonally varying stimulus feature, such as frequency (Harrington © The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]

et al. 1998; Mangels et al. 1998; Rao et al. 2001; Nenadic et al. 2003) or spatial location (Coull and Nobre 1998; Bueti and Macaluso 2011). Contrasting timing with judgments about another stimulus feature is intended to control for more general processes such as attention, working memory, and response demands. That is, these “executive” processes should also be engaged during control tasks and should therefore be absent when the 2 tasks are directly compared. In general, temporal discrimination tasks (relative to discriminations of another stimulus feature) consistently activate temporal cortices (superior, middle, inferior, pole), supplementary motor area (SMA/pre-SMA) and premotor cortices, frontal cortex including lateral frontal cortex and operculum, inferior parietal regions, insula, basal ganglia, and cerebellum (Rao et al. 2001; Ferrandez et al. 2003; Lewis and Miall 2003; Nenadic et al. 2003; Coull et al. 2004; Morillon et al. 2009; Bueti and Macaluso 2011; for a recent meta-analysis, see Wiener et al. 2010). There are (at least) 2 complications involved in drawing conclusions about the roles of brain regions revealed using this strategy. The first is that, very often, the difficulty of the timing task is not carefully equated to the difficulty of the control task. For example, control tasks sometimes simply require participants to make a button-press response after the presentation of a stimulus without performing a task (Jueptner et al. 1995; Pouthas et al. 2005; Jahanshahi et al. 2006). Other studies have used a control task involving passively viewing stimuli without a task or response (Hinton and Meck 2004; Kudo et al. 2004). These difficulty differences have implications for the observed patterns of brain activation. For example, Livesay et al. (2007) compared a timing task to an easy and a difficult control (colordiscrimination) task. Brain areas that were activated more by the difficult than by the easy task (regardless whether the task required duration or color discrimination), included bilateral prefrontal cortices, inferior parietal cortex, pre-SMA, insula, putamen, and cerebellum. A similar network was observed when a difficult timing task was compared with an easy timing task: SMA, insula, dorsolateral prefrontal cortex, and putamen, all bilaterally (Tregellas et al. 2006). Importantly, in this context, activations associated with task difficulty differences largely overlap with those typically reported to underpin timing functions. Second, even when task difficulty is carefully matched, performance of a timing task is necessarily confounded with attention to time. That is, judgments about the temporal features of a stimulus require attending to time, but additionally rely on timing functions (e.g., duration estimation). Thus, neural networks supporting attention to temporal information have arguably thus far not been cleanly separated from those supporting timing functions per se. In the current human fMRI study, we overcame these limitations using a novel approach to studying timing performance

Downloaded from http://cercor.oxfordjournals.org/ at MPI Cognitive and Brain Science on August 27, 2013

Meaningful auditory stimuli such as speech and music often vary simultaneously along multiple time scales. Thus, listeners must selectively attend to, and selectively ignore, separate but intertwined temporal features. The current study aimed to identify and characterize the neural network specifically involved in this feature-selective attention to time. We used a novel paradigm where listeners judged either the duration or modulation rate of auditory stimuli, and in which the stimulation, working memory demands, response requirements, and task difficulty were held constant. A first analysis identified all brain regions where individual brain activation patterns were correlated with individual behavioral performance patterns, which thus supported temporal judgments generically. A second analysis then isolated those brain regions that specifically regulated selective attention to temporal features: Neural responses in a bilateral fronto-parietal network including insular cortex and basal ganglia decreased with degree of change of the attended temporal feature. Critically, response patterns in these regions were inverted when the task required selectively ignoring this feature. The results demonstrate how the neural analysis of complex acoustic stimuli with multiple temporal features depends on a frontoparietal network that simultaneously regulates the selective gain for attended and ignored temporal features.

Materials and Methods Participants Twenty right-handed, native German speakers (aged 21–31 years; n = 10 females) with self-reported normal hearing participated in the experiment in exchange for financial compensation. The procedure was approved of by the ethics committee of the medical faculty of the University of Leipzig and in accordance with the declaration of Helsinki. All participants provided written informed consent. Stimuli and Task Stimulus and task information is summarized in Figure 1. Stimuli were amplitude-modulated white noise bursts, which were always presented

in pairs. First, a “standard” stimulus was presented, which had modulation rate equal to 8 Hz and duration equal to 500 ms. One second later, a “comparison” stimulus was presented. Modulation rate and duration of the comparison stimulus were varied orthogonally, taking on 1 of 5 values for either stimulus feature (duration: 500 ms ± 0%, 20%, and 40%; modulation rate: 8 Hz ± 0%, 10%, 20%). These values of duration and AM rate were chosen on the basis of pilot testing conducted to match the difficulty of the tasks. The modulation depth of all sounds was 100%. The starting phase of the amplitude modulation was randomized from trial to trial, separately for the standard and comparison. In separate blocks, listeners had to judge either the duration or the modulation rate of the comparison stimulus relative to the standard. For duration judgments, comparison stimuli were judged as “shorter” or “longer” than the standard, while for modulation-rate judgments, comparison stimuli were judged as “faster” or “slower” than the standard. Note that judging one stimulus feature requires simultaneously ignoring the other. Although judgments of duration and modulation rate make use of different response categories, it is important that both judgments are inherently temporal; indeed, modulation-rate estimation could be conceptualized as an estimation of the duration between cycles.

Procedure Each participant completed a short familiarization session prior to scanning to ensure that they understood the task. Then, participants completed a total of 9 pseudorandomized blocks in the scanner (3 for duration judgments, 3 for modulation-rate judgments, and 3 for a control condition not reported here). Within each block, each of the 5 × 5 unique duration × modulation-rate combinations was presented once, resulting in 25 stimulus presentations per block. In addition, 5 silent trials were presented per block, where no acoustic stimulation occurred. On each trial, presentation of the standard stimulus was followed after a 1000-ms pause by the presentation of the comparison stimulus. A visual response prompt was displayed 1000 ms after the offset of the comparison, and the participant was given 2500 ms to indicate his response using the index finger of the right or left hand for the 2 response

Figure 1. Top: Schematic of the trial timeline. On each trial, a “standard” stimulus with fixed duration (500 ms) and fixed modulation rate (8 Hz) was presented, followed after 1000 ms by a “comparison” stimulus with a modulation rate of 8 Hz ± 0, 10, and 20% and a duration of 500 ms ± 0, 20, 40%. Listeners judged either the duration (by responding “shorter” or “longer”) or the modulation rate (by responding “faster” or “slower”) of the comparison relative to the standard. Bottom left: Behavioral results for the duration- and modulation-rate tasks. Proportion correct is shown as a function of duration (left) and modulation rate (right) separately for the duration-judgment (red) and modulation-rate-judgment (orange) tasks. The bar plot shows second-order coefficients estimated from quadratic fits to proportion correct. Bottom right: Brain regions activated more strongly by stimulation/task performance than during silence.

2 Selective attention to time

•

Henry et al.


that involved participants attending to and judging temporal stimulus features that unfolded on 2 different time scales. The goal was to isolate brain regions supporting feature-selective attention to time from those involved in task performance more generally. Participants discriminated either the duration or the amplitude modulation rate of auditory stimuli in separate blocks. Critically, auditory stimulation, working memory demands, and response requirements were identical across the 2 tasks. Moreover, we carefully controlled task difficulty across tasks. We identified brain regions supporting task performance generally using a correlational “searchlight”-based analysis (Kriegeskorte et al. 2006), which screened for neural activity matching the behavioral pattern associated with variations in both duration and modulation rate simultaneously in both tasks (i.e., duration or modulation-rate judgments). A second analysis based on attention-dependent variations in brain activity as a function of discrimination difficulty isolated those brain regions that were specifically involved in selective attention to temporal stimulus features.

categories (“shorter”/“longer” or “faster”/“slower”). The mapping of response category to hand was counterbalanced across participants. Auditory stimuli were delivered through MR-Confon headphones; Hearsafe HS-ER 20 earplugs provided an additional 16-dB attenuation. Presentation software was used to control stimulus delivery and collection of behavioral responses. Visual response prompts were projected on a screen that participants viewed via a mirror attached to the head coil.

Overall Effects of Stimulation and Task Performance First, we tested for brain regions that generally responded to auditory stimulation and/or task performance, independently of the temporal stimulus feature being judged. Single-subject design matrices were constructed that modeled trials during which stimulation was presented as one condition and null trials as a second condition; block was included as a regressor of no interest (block-specific mean normalization). A single contrast specified at the first level tested for effects of stimulation/ task against silence. The contrast result was tested at the second level using a 1-sample t-test against 0 and a threshold of PFWE < 0.05. Correlation Between BOLD and Single-Subject Behavioral Patterns (Searchlight Analysis) The goal of this analysis was to reveal brain regions that contributed to timing performance across both of the investigated tasks. This included regions supporting stimulus timing, selective attention, working memory,

Selective Attention to Temporal Stimulus Features Next, we were interested in isolating brain regions that were specifically involved in selective attending to and selective ignoring of individual temporal stimulus features. We reasoned that activation in such regions should be modulated by the difficulty of the stimulus comparison for either the task-relevant (attended) or task-irrelevant (ignored) feature. That is, we suspected that activation in these regions should be ramped up for difficult more than easy stimulus comparisons for the attended feature. On the other hand, the opposite might be true for the ignored feature, where easier stimulus comparisons are likely to be more distracting. Effectively, we tested for quadratic trends as a function of parametric stimulus level that differed depending on whether the tested stimulus feature was attended ( judged) or ignored. At the first level, we specified contrast coefficients corresponding to a quadratic trend [–2 1 2 1 –2] for the 5 parametric stimulus levels for either the duration or modulation-rate stimulus feature, separately for the duration- and modulation-rate tasks. This led to 4 contrast maps reflecting the degree to which the brain response as a function of stimulus level took on a quadratic trend in each of 4 experimental situations: attend duration, ignore duration, attend modulation rate, and ignore modulation rate. Then, at the second level, we conducted 2 separate paired-samples t-tests. First, we compared quadratic trends as a function of duration when duration was attended (i.e., during duration judgments) versus ignored (i.e., during modulation-rate judgments). Second, we compared quadratic trends as a function of modulation rate when modulation rate was attended (i.e., during modulation-rate judgments) versus ignored (i.e., during duration judgments). Effectively, these 2 second-level t-tests specifically targeted brain regions that show opposite quadratic trends as a function of stimulus level when a temporal stimulus feature was attended versus ignored.

Cerebral Cortex 3


Data Acquisition and Analysis Data were acquired on a 3-Tesla Siemens TIM Trio scanner with a 12-channel head coil. First, a Field map image was acquired for use during realignment. During functional scanning, T2*-weighted images were acquired using a gradient-echo echo-planar imaging (EPI) sequence with TE = 30 ms, flip angle = 90°, acquisition bandwidth = 116 kHz, matrix = 64 × 64, FOV = 19.2 cm, resulting in an in-plane resolution of 3 mm × 3 mm. Slice thickness was 3 mm, with a 1-mm interslice gap. A sparse (bunched) sampling protocol in combination with cardiac gating (von Kriegstein et al. 2006) was implemented, resulting in a repetition time (TR) ∼9 s; auditory stimuli were presented during the silent period between volume acquisitions. Existing highresolution T1-weighted magnetization-prepared rapid gradient-echo images were taken from the database of the Max Planck Institute for Human Cognitive and Brain Sciences. T1 images were acquired according to the following parameters: TR = 1.3 s, TA = 10 ms, TE = 3.93 ms, matrix = 256×240, FOV = 256 × 240, resulting in a resolution of 1 mm × 1 mm × 1.5 mm (interpolated to 1-mm isotropic during spatial normalization). Data were analyzed using SPM8 (Welcome Trust Centre for Neuroimaging, London, UK), Marsbar (Brett et al. 2002), and custom MATLAB scripts. The image preprocessing pipeline included rigid-body spatial realignment and unwarping using the Field map image, segmentation of the T1 image, coregistration to the T1 image according to spatial normalization parameters from segmentation, normalization to Montreal Neurological Institute (MNI) space and interpolation to 3 × 3 × 3-mm voxel size, and smoothing with an 8-mm isotropic Gaussian kernel. All first-level analyses were modeled based on a finite-impulse response (FIR) function. A high-pass filter with a cutoff of 1024 s was applied to eliminate low-frequency noise. No correction for serial autocorrelation was necessary because of the long TR in the sparse-sampling acquisition. Three main analyses were conducted. First, a confirmatory analysis contrasted brain activation associated with trials on which a stimulus was presented against null trials, without regard to which task was performed. Second, we used a searchlight-based analysis to identify brain regions where activation patterns across parametric stimulus levels and tasks were correlated with behavioral performance across the same stimulus levels and tasks. Finally, we isolated brain regions that contribute to selective attending/selective ignoring of individual temporal stimulus features, but were not specifically involved in time estimation, per se. We will describe these analyses in more detail in turn. Unless stated otherwise, second-level significance thresholds for all analyses were determined using a Monte Carlo simulation that estimated the cluster size necessary to achieve a whole-brain α = 0.05 (i.e., P < 0.005, k = 18; Slotnik et al. 2003; for application, see, e.g., Straube et al. 2008; Obleser et al. 2011; McGettigan et al. 2012).

and comparison and decision processes. In particular, we looked for brain regions where activation correlated with performance patterns across all combinations of duration and modulation-rate level and across both tasks. Thus, the analysis was sensitive to any potential interactions between factors. The current experimental design involved orthogonally crossing 5 durations with 5 modulation rates, resulting in a 5 × 5 matrix of unique stimulus conditions. For each listener, we calculated proportion correct in each of the 25 unique conditions, averaged across blocks, separately for the duration-judgment and the modulation-rate-judgment tasks. Then, in order to increase the resolution of our behavioral performance measure, we “smoothed” the 5 × 5 matrices into 4 × 4 grids (Schönwiesner and Zatorre 2009), where each cell of the grid corresponded to an average of behavioral responses to 12 stimuli (4 unique stimuli × 3 repetitions). After smoothing, each cell of the grid corresponded to an average of behavioral data for 2 duration levels and 2 modulation-rate levels. Examples of resulting grids for a single participant are shown in Figure 2. Then, for each subject, we constructed matching 4 × 4 grids containing estimated betas from a general linear model (GLM; Fig. 2). To do this, we constructed a single design matrix per subject. Individual volumes were assigned to the corresponding stimulus categories of the 4 × 4 grid, separately for each block. That is, for a single block, each of the 16 regressors comprised 4 stimuli, 2 adjacent duration levels, and 2 adjacent modulation-rate levels. An additional regressor per block was included for null trials, and individual blocks were separately normalized with respect to block-specific means. Betas were estimated for each of the 16 regressors, for each of the 6 experimental blocks. Then, betas were averaged across the 3 blocks corresponding to a single task (i.e., duration vs. modulation rate). Thus, parallel to the behavioral data, we formed a 4 × 4 grid of beta-weights, where each cell corresponded to a single beta (averaged over 3 blocks). We then used a whole-brain searchlight approach (searchlight radius = 3 mm) to look for regions where the pattern of betas over individual stimulus conditions and tasks significantly matched the behavioral performance over the same conditions (Kriegeskorte et al. 2006). Effectively, we calculated for each searchlight volume the Pearson’s correlation coefficient between the beta pattern and the behavioral pattern. Correlation coefficients per voxel were then Fisher’s z-transformed and tested against 0 at the second level.

ROI Analysis Following the quadratic trend analysis, we defined spherical region of interests (ROIs) (5-mm radius) centered on the peak voxel in each significant cluster. We then used Marsbar software to extract percent signal change for each of the 5 duration or modulation-rate conditions for the appropriate contrast and subsequently averaged the percent signal change across regions. We display quadratic trends for each of the effects we report, but we do not perform additional statistics on extracted percent signal change to avoid circularity in analysis.

Results Behavioral Results Figure 1 shows proportion correct as a function of duration and modulation rate, separately for duration judgments and modulation-rate judgments. It is clear that the task was more difficult (i.e., proportion correct was lower) when the degree of change between the standard and comparison stimulus was smaller, specifically for the attended ( judged) temporal stimulus feature. This is reflected in a quadratic performance profile as a function of stimulus level, with best performance corresponding to the most extreme stimulus manipulations (±40% for duration, ±20% for modulation rate). On the other hand, the level of the ignored stimulus feature did not have a strong influence on performance. In order to quantify these behavioral effects, we performed quadratic fits to single-participant data as a function of duration and modulation rate separately for duration and modulation-rate judgments. In particular, the second-order (quadratic) coefficients from these fits reflect the degree to which performance 4 Selective attention to time

•

Henry et al.

was coupled to stimulus level manipulations. We analyzed the second-order coefficients in a 2 (Stimulus Feature: duration, modulation rate) × 2 (Task: judge duration, judge modulation rate) repeated-measures analysis of variance. Critically, the interaction was significant (F1,19 = 122.03, P < 0.001). The interaction was driven by large second-order coefficients for attended stimulus features (i.e., duration in the duration-judgment task and modulation rate in the modulation-rate judgment task) and smaller second-order coefficients for the ignored stimulus features (i.e., modulation rate in the duration-judgment task and duration in the modulation-rate judgment task; see Fig. 1). Thus, variations in stimulus level for both stimulus features affected task performance (with best performance at the easiest stimulus levels), but specifically when the stimulus feature was attended. Neither the Stimulus Feature nor the Task main effect reached significance (P ’s ≥ 0.09). We also conducted 4 single-sample t-tests comparing quadratic coefficients against zero in order to test for significant quadratic trends in the behavioral data. Quadratic trends were significant for both stimulus features when they were attended (P ’s < 0.001) and for duration when it was ignored (P = 0.005), but not for modulation rate when it was ignored (P = 0.60). Notably, the quadratic coefficient as a function of ignored duration was opposite in sign to the coefficients for attended features, indicating that performance was better when the stimulus comparison of the ignored feature was less obvious (i.e., more difficult). Finally, we also conducted a paired-samples t-test on proportion correct in order to confirm that performance was very similar, and thus difficulty well equated, across tasks (duration


Figure 2. Top: A single-participant example of the 4 × 4 grid of behavioral data and estimated betas that were correlated in searchlights centered on single voxels; note that stimulus values reflect averages of 2 neighboring stimulus levels because of smoothing. The correlation for the example data is also shown for within a single significant cluster (left inferior temporal cortex). Bottom left: Brain regions where the behavior–brain correlation was significant across tasks. Bottom right: Fisher’s z-transformed correlation values for betas extracted from the peak voxel in each significant cluster, for illustration purposes only. IT, inferior temporal; IF, inferior frontal; SF, superior frontal; MT, middle temporal; MF, medial frontal; PC, posterior cingulate; MC, middle cingulate.

judgments: M = 0.80 ± SEM = 0.02; modulation-rate judgments: M = 0.77 ± 0.02), t (19) = 0.90, P = 0.38. Overall fMRI Effects of Auditory Stimulation and Task Performance Figure 1 shows brain areas that were activated more strongly during auditory stimulation and task performance than during null trials, irrespective of task (for peak voxel coordinates, see Table 1). Activated brain regions were consistent with performance of an auditory timing task, and included bilateral superior temporal gyri extending to insular cortex, basal ganglia, auditory thalamus (medial geniculate), cerebellum, pre-SMA/SMA, inferior frontal gyri (BA44), and auditory brainstem regions bilaterally including inferior colliculus and cochlear nucleus.

Table 1 MNI coordinates and statistics for peak voxels in significant clusters revealed by contrasting all trials during which stimulation was presented and a task was performed to null trials Brain region

L Superior temporal gyrus R Superior temoral gyrus R Caudate R Pallidum L Thalamus R Cerebellum (Lobule VI) L Cerebellum (Lobule VI) L Pre-SMA L Precentral R Inferior frontal cortex (p. opercularis) L Middle occipital cortex L Rolandic operculum R Calcarine cortex R Postcentral L Superior medial frontal cortex

MNI coordinates x

y

z

–39 45 21 21 –9 24 –24 –3 –45 54 –27 –48 12 48 –6

–22 –16 14 5 –16 –64 –58 2 5 11 –61 8 –67 –25 23

2 6 18 6 10 –26 –26 62 34 30 42 2 10 50 42

Z-value

No. of voxels

7.37

2149

6.72

995

5.92 5.38 5.04 5.03 4.98 4.94 4.93 4.85

170 23 2 3 3 5 4 2

Brain region

L Inferior temporal cortex L Cerebellum (VI) L Inferior frontal cortex (p. opercularis) L Superior frontal cortex L Middle temporal cortex R Medial frontal cortex R Inferior frontal cortex (p. opercularis) R Superior frontal cortex R Posterior cingulate R Middle cingulate L Insula L Inferior temporal cortex R Precentral R Medial frontal cortex L Medial frontal cortex

MNI coordinates x

y

z

−39 −33 −45 −24 −66 6 54 24 3 9 −30 −36 27 0 −6

5 −34 5 71 −4 62 8 2 −34 14 26 −1 −16 65 68

−50 −46 58 6 −22 46 34 74 26 46 2 −30 42 −10 22

Z-value

No. of voxels

5.09 5.09 5.09 4.27 4.05 3.86 3.8 3.57 3.49 3.25 3.23 3.16 2.94 2.92 2.8

38 6253 308 55 34 49 183 22 104 168 35 30 34 26 23

patterns in the same brain regions, regardless of the temporal stimulus feature to which attention was oriented. Thus, the correlational searchlight analysis revealed regions that underlie task performance generally, and which are common to the 2 temporal stimulus features studied here. However, this analysis cannot reveal whether individual regions correlated with performance because of their association with time estimation, attention to temporal information, retention of the standard in working memory, or stimulus comparison processes. Thus, we performed a second analysis in order to isolate regions underlying selective attention to temporal stimulus features. Selective Attention to Temporal Stimulus Features Figure 3 highlights brain regions that we interpret as supporting selective attention to temporal stimulus features. Critically, the analysis focused on differences in quadratic trends that depended on attention. That is, quadratic trends as a function of stimulus level (and thus related to discrimination difficulty; see Behavioral Results section) were inverted when one stimulus feature (i.e., duration, modulation rate) was ignored relative to when it was attended. In Figure 3, regions in red indicate responses to duration manipulations. Specifically, the pre-SMA bilaterally, left caudate, and right inferior parietal cortex responded more strongly when duration levels were less discriminable, specifically when duration was being judged. When duration was the ignored feature, however (i.e., when modulation-rate was being judged), these regions responded more strongly to more discriminable duration comparisons. A larger network of brain regions (blue) was significant when we evaluated responses to modulation-rate manipulations, although the 2 analyses were associated with a large degree of overlap ( purple). SMA (extending to anterior and middle cingulate and medial frontal cortex), bilateral inferior parietal cortices, bilateral inferior frontal gyri ( p. triangularis), bilateral basal ganglia ( putamen, pallidum), bilateral insular cortex, left orbitofrontal cortex, right supramarginal gyrus, and right cerebellum responded more strongly for less discriminable modulation rates, specifically when modulation rate was being judged. However, these same regions responded more strongly to more discriminable modulation-rate pairs when duration was being judged (i.e., when modulation rate was Cerebral Cortex 5


Correlation Between Brain Responses and Single-Subject Behavioral Patterns (Searchlight Analysis) Each participant had a unique pattern of behavior in the 16 stimulus conditions (4 duration × 4 modulation rate) in each of the 2 tasks, which we then correlated with betas estimated from design matrices comprising the same 4 × 4 stimulus conditions for the 2 tasks. Figure 2 shows the brain regions where behavior–brain correlations were statistically significant and Fisher’s z-correlation coefficients for each significant cluster (for illustration purposes only). Significant correlations with the behavioral patterns were observed for the bilateral cerebellum, bilateral auditory cortex, bilateral thalamus and basal ganglia (bilateral pallidum, right caudate), inferior colliculus, pre-SMA/middle cingulate cortex, bilateral calcarine, and middle occipital cortices, left inferior temporal cortex extending to temporal pole and fusiform gyrus, and right posterior cingulate cortex. Moreover, a number of clusters were observed in frontal cortex bilaterally, including inferior frontal gyri extending to insula, superior frontal, middle frontal, and orbitofrontal cortex. Peak voxel coordinates and statistics are provided in Table 2. To follow-up this analysis, we tested for brain regions where brain–behavior correlation strength depended on the task being performed: We compared correlation maps calculated only for the duration task with correlation maps calculated only for the modulation-rate task, and found no significant differences. This indicates that behavioral performance was related to activation

Table 2 MNI coordinates and statistics for peak voxels in significant clusters arising from correlating single-participant behavioral patterns with single-participant patterns of betas estimated from GLM

Table 3 MNI coordinates and statistics for peak voxels in significant clusters where quadratic trends as a function of parametric stimulus level differed depending on whether the analyzed stimulus dimension (duration or modulation rate) was attended or ignored Brain region

MNI coordinates

Duration R Inferior parietal cortex L Caudate R Pre-SMA Pre-SMA Modulation rate L Pre-SMA L Inferior parietal cortex L Inferior frontal cortex (p. triangularis) R Inferior frontal cortex (p. triangularis) L Orbitofrontal cortex R Cerebellum Crus I R Inferior parietal cortex R Supramarginal gyrus

Z-value

No. of voxels

x

y

z

54 −18 15 0

−37 2 8 26

50 −14 58 50

3.75 3.53 3.13 2.93

104 26 25 50

−3 −48 −54 48 −45 33 45 60

26 −46 20 29 44 −79 −52 −19

46 46 18 26 −10 −18 46 34

5.04 4.46 4.34 4.33 3.59 3.58 3.29 3.04

379 290 565 486 37 19 44 31

ignored). Peak voxel coordinates and statistics for both contrasts are provided in Table 3.

Discussion The current study aimed to identify and characterize the brain regions involved specifically in feature-selective attention to time. To this end, we used a novel paradigm that involved listeners judging either the duration or modulation rate of auditory stimuli where 2 temporal stimulus features were varied simultaneously. Critically, this paradigm overcame 2 limitations common to studies of time perception and attention to time. First, since both manipulated stimulus features were temporal, timing functions and attention to time were 6 Selective attention to time

•

Henry et al.

unconfounded. That is, both tasks involved attending to and judging temporal stimulus features, in contrast to previous studies where a timing task was compared with, for example, a frequency, location, or color judgment. Second, we carefully controlled task difficulty across the 2 temporal stimulus features. In addition, auditory stimulation, working memory demands, and response requirements were identical across tasks. Thus, the only difference between task conditions was the feature to which attention was allocated. A first analysis step identified all brain regions where activation for individual stimuli was correlated with performance. We interpret these regions as supporting general performance of the timing tasks. That is, these regions may have been involved in time estimation, selective attention to time, comparison processes, working memory, and so forth. A second analysis step then isolated those brain regions that were specifically involved in selective attention to temporal features. We discuss the results of these analyses in turn. Neural Correlates of Timing Functions Indexed by Behavior–Brain Correlation The first analysis involved a searchlight-based approach that correlated single-participant behavioral performance patterns across stimuli and tasks with single-participant brain responses. We expected the results of this analysis to include a conglomeration of timing-specific regions and higher-order “executive” mechanisms. That is, this analysis was intended to highlight brain regions that are associated with performance of a timing task, but were common to the 2 temporal features we investigated. We observed activation in a wide network that largely overlaps with brain regions typically involved in auditory timing tasks (see Fig. 2 and Table 2). In particular, the basal ganglia and cerebellum have both been suggested to play a critical role


Figure 3. Brain regions associated with selective attending to and selective ignoring of temporal stimulus features. All colored brain regions demonstrate significant differences in quadratic trends (as a function of stimulus level) between attending and ignoring temporal stimulus features. Responses during the duration task are shown in red: These regions responded most strongly to the most difficult duration comparison in the duration task, but responded most strongly to the easiest modulation-rate comparisons in the modulation-rate task. Responses during the modulation-rate task are shown in blue: These regions responded most strongly to the most difficult modulation-rate comparison during the modulation-rate task, but responded most strongly to the easiest duration comparison during the duration task. Purple color indicates overlap between the 2 contrasts. Note that plots on the right depict mean percent signal change averaged across regions that show significant effects in the whole-brain analyses displayed on the left.

Neural Correlates of Selective Attention to Temporal Stimulus Features The goal of our second analysis approach was to isolate brain regions supporting selective attention to time. We reasoned that brain activation in regions supporting selective attention might not only vary with discrimination difficulty, but would also be modulated by attention. That is, we expected that activation would be highest when temporal discrimination was the most difficult, specifically for the task-relevant feature. On the other hand, brain activation might be highest when discrimination was easiest for the task-irrelevant feature, that is, when irrelevant stimulus information was potentially the most distracting. Effectively, we tested for an inversion of a quadratic pattern depending on whether the temporal stimulus feature was attended or ignored. We observed activation in a fronto-parietal network that met these criteria; activation patterns in these regions did not depend quantitatively on which stimulus feature was task-relevant, but critically depended on whether attention was directed towards or away from a stimulus feature.

When participants attended to duration (i.e., when duration was judged), blood-oxygen-level dependent (BOLD) activation in right inferior parietal cortex, left caudate, and pre-SMA was strongest for the most difficult duration discriminations and for the easiest modulation-rate discriminations. When in turn participants attended to modulation rate, BOLD activation was strongest for the most difficult modulation-rate discriminations and the easiest duration discriminations in the same regions that responded when duration was judged and in a number of additional regions including left inferior parietal cortex, bilateral inferior frontal gyri, bilateral insular cortex, bilateral putamen and pallidum, and right cerebellum. Critically, the differences in the observed brain regions between features (duration vs. modulation rate) were a matter of strength, as the pattern for duration overlapped even more completely with the pattern for modulation rate when a slightly liberal threshold was used. Thus, the activated regions did not depend on which task the participant performed (i.e., duration vs. modulation rate), but rather patterns were only contingent on attentional allocation, such that the most difficult discriminations for the attended feature and the easiest (most distracting) discriminations for the ignored feature elicited the strongest activations (Fig. 3, Table 3). We interpret this fronto-parietal network activation as supporting selective attention to and selective ignoring of auditory temporal stimulus features. Indeed, the activated regions strongly overlap with those reported in a number of previous studies comparing timing to discrimination of orthogonally varying stimulus dimensions (e.g., frequency, intensity, color, spatial location; Rao et al. 2001; Ferrandez et al. 2003; Lewis and Miall 2003; Nenadic et al. 2003; Coull et al. 2004; Morillon et al. 2009; Bueti and Macaluso 2011). In particular, the observed network is almost identical to that observed by Livesay et al. (2007) when a difficult timing task was contrasted with an easier control (color-discrimination) task, or vice versa. In the current study, our analysis indeed picked out regions that were sensitive to discrimination difficulty. However, our results critically go beyond this by showing that task difficulty effects are sensitive to selective attention manipulations. We suggest that these regions are particularly associated with selective attention functions because the current task required selective attention to one of 2 “temporal” stimulus features, thereby ruling out timing-specific functions as driving brain activation. More generally, the same network (most frequently the frontal and parietal regions) is activated during selective attention to nontemporal, and in fact, nonauditory dimensions, in line with a domain-general “executive” role (e.g., Hopfinger et al. 2000; Culham et al. 2001; Pessoa et al. 2003; Power and Petersen 2013). Interestingly, in the current study, brain activation patterns completely reversed depending on whether an individual stimulus feature was attended or ignored. This suggests that the recruitment of top-down mechanisms depended as much on the necessity to ignore a stimulus feature as to attend to that feature. Generally, this finding is consistent with the idea that attention not only works to increase gain for task-relevant information, but also to attenuate task-irrelevant (ignored) information (Kanwisher and Wojciulik 2000; Pessoa et al. 2003). The Role of Insular Cortex in General Attention and Saliency Processing Although the frontal and parietal cortices are very commonly implicated in selective attention to a perceptual stimulus Cerebral Cortex 7


in timing functions (Ivry and Keele 1989; Harrington et al. 1998; Casini and Ivry 1999; Koch et al. 2007; for a review, see Coull et al. 2010 or for a meta-analysis, see Wiener et al. 2010). Moreover, Coull and colleagues (2010, 2004; see also Harrington et al. 2004) have advocated a timing-specific role for right inferior parietal cortex, while others have suggested that it is critical for timing (Coull et al. 2004; Wiener et al. 2010). Finally, SMA/pre-SMA has been proposed as underpinning a neural component of a clock mechanism (Macar et al. 1999 2002, 2006; for a recent meta-analysis, see Schwartze et al. 2012). Consistent with these proposals, we observed in all of these brain regions activation that was correlated with behavioral performance across the 2 timing tasks. This analysis also revealed correlations with performance in auditory cortex, auditory thalamus, and inferior colliculus along the auditory pathway, and visual areas including bilateral middle occipital and calcarine cortices. First, although the current experiment made use of auditory stimulation, auditory pathway activation nevertheless changed with performance, suggesting a more active role in performance of the timing tasks. In this regard, it is notable that a modality-independent role for auditory sensory cortices has been suggested for time perception, as superior temporal cortices are often implicated in timing tasks involving visual, rather than auditory, stimulation (Coull et al. 2004; Morillon et al. 2009). Moreover, transcranial magnetic stimulation (TMS) to auditory cortex disrupts both auditory and visual duration discrimination (Kanai et al. 2011). Second, a number of studies report visual activation during auditory timing tasks (Bengtsson et al. 2009; Grahn et al. 2011). However, the implications of visual activation are unclear. Bueti and Macaluso (2010) showed visual activation in response to an auditory timing task, but auditory stimuli in this study easily evoked visual imagery (rhythmic pounding of a hammer or hands clapping). More similar to the current study, Bramen (2008) observed visual activation during auditory timing that correlated with performance, indicating a specific role for timing. Visual activation in this study was observed already at the level of the brainstem, suggesting very early crosstalk between modalities. We suggest that visual cortical areas, like auditory cortical areas, may play a modality-independent role in time perception. Future research will need to address more clearly the role of the visual pathway in timing functions.

Basal Ganglia’s Role in Executive and Timing Function The activation patterns we observed in the left caudate, bilateral pallidum, and bilateral putamen are consistent with basal ganglia involvement in executive processes, despite their implication as a critical component of human temporal processing (Wiener et al. 2010). Outside of timing literature, holding the basal ganglia accountable for executive processes is much more common (Packard and Knowlton 2002; Gruber et al. 2006; McNab and Klingberg 2007). One line of evidence comes from general cognitive and executive impairments observed for patients with Parkinson disease (Grahn et al. 2008, 2009). In particular, Parkinson patients tend to have trouble switching between the features of a task, failing to disengage a now irrelevant feature or goal, while simultaneously failing to engage a newly relevant feature or goal. The current task required online adjustment of feature weighting, as the relative difficulties of discriminations along the 2 tasks determined the amount of attentional allocation each feature should receive. Thus, one possibility is that basal ganglia involvement is necessary to flexibly adjust the amount of attention that should be allocated to an attended versus an ignored stimulus feature. In this regard, caudate activation has previously been observed to increase with increasing task difficulty, presumably reflecting increasing attentional allocation for more difficult task conditions (Dagher et al. 1999). Linking the basal ganglia with the 8 Selective attention to time

•

Henry et al.

remainder of our observed attention network, Grahn et al. (2008, 2009) have reviewed data indicating both structural and functional connections between basal ganglia and frontal and cingulate cortices, and found that the caudate nucleus shares connections with the lateral prefrontal cortex and pre-SMA. Selective Attention to Temporal Information on Different Time Scales The current study involved selective attention to as well as selective ignoring of auditory temporal stimulus features, where the manipulated features (i.e., duration and modulation rate) were distinguishable on the basis of their relevant time scales. That is, stimuli were varied around 500 ms in duration, while amplitude modulations occurred on a somewhat faster time scale (i.e., 125 ms or 8 Hz). Thus, the time scale corresponding to amplitude modulation was nested within the time scale on which duration was manipulated. In this regard, the stimuli we used share an important feature with natural speech, which contains important temporal information at a number of nested time scales (Rosen 1992). The necessity of processing temporal information on different time scales has been suggested to determine the neural mechanisms underpinning timing functions, with different time scales being assigned to different brain regions (Lewis and Miall 2003; Schwartze et al. 2012), hemispheres (Kagerer et al. 2002; Poeppel 2003), neural oscillatory frequency bands (Pöppel 1994; 1997), or neural implementations more generally (Ivry and Schlerf 2008). Thus, it could be suggested that the duration and modulation-rate tasks may have relied on partly dissociable neural mechanisms because of their dissociable time scales. However, we did not observe any neural differences between tasks when we performed a direct comparison. There is little consensus on boundaries between time scales that recruit different neural mechanisms (Pöppel 1997; Lewis and Miall 2003; Poeppel 2003; Buhusi and Meck 2005; Ulbrich et al. 2007; Schwartze et al. 2012). However, we suggest that the temporal manipulations examined here share a functionally similar time scale (to be contrasted with, in an extreme case, millisecond versus circadian timing). That is, the durations and modulation rates tested here were chosen to be within the range of time scales which are relevant for perceiving the syllable envelope and durational stress cues in speech (Rosen 1992; Ghitza and Greenberg 2009). Thus, the duration and modulation-rate tasks were unlikely to recruit different neural mechanisms on the basis of time scale differences, but were still nested in a way that arguably allows generalization of the current results to more natural stimuli (see next section). Generalizing the Current Results to Perception of Natural Sounds A natural question that emerges from the current work is the extent to which the results can be generalized to more ecologically valid stimuli. That is, to what extent can we assume that the brain regions we have identified as underpinning selective attention to temporal stimulus features also support selective attention to, for example, voicing versus duration-stress cues in natural speech? One hint comes from a recent set of studies by Erb et al. (2012, 2013). The authors showed that psychophysical thresholds for a modulation-rate discrimination task correlated with degraded speech comprehension (2012). That is, modulation-rate discrimination abilities predicted success in


feature, the roles of the anterior insula and basal ganglia are perhaps less clear in this context. This is especially true, as the insular cortex is very often activated during timing tasks (Kosillo and Smith 2010), and the basal ganglia have been assigned a crucial role underpinning human timing functions (Wiener et al. 2010). However, activation in both regions scaled with task difficulty in an attention-dependent manner, mimicking the responses of the frontal and parietal cortices. Although the anterior insula has commonly been associated with, for example, emotion perception and interoception, an emerging view is that the anterior insula plays a role in attention and salience perception (Eckert et al. 2009; Menon and Uddin 2010; Sterzer and Kleinschmidt 2010). For example, in timing tasks in particular, anterior insula activation is modulated by task difficulty (Harrington et al. 2004; Tregellas et al. 2006). More generally, anterior insula activation was a marker of single-trial discrimination difficulty in an auditory discrimination task, where activation scaled with signal-to-noise ratio and reaction time (Binder et al. 2004). Recent work from our own laboratory found the anterior insula to be critically involved both in coping with degraded speech as well as in difficult auditory nonspeech discrimination (Erb et al., 2013), supporting earlier work by Eckert et al. (2009), who demonstrated a relation between anterior insula activation and speech intelligibility. Critically, this latter study conducted functional connectivity analyses that revealed correlations between time courses in anterior insula (seed region) and inferior frontal regions, pre-SMA/ACC, and inferior parietal cortices bilaterally, consistent with the concurrent activations observed in the current study. Further connectivity analyses using diffusion tensor imaging and resting-state fMRI reinforce the strong connections between the anterior insula and the other regions that, in the current study, we have attributed to a selective attention network (Menon and Uddin 2010; Sterzer and Kleinschmidt 2010).

Conclusions The current human fMRI study applied a novel behavioral paradigm involving judging auditory stimuli with 2 simultaneously varying and nested temporal stimulus features. In a correlational, searchlight-based analysis approach, we identified a network of brain regions where activation was significantly correlated with behavioral performance across individual stimuli and tasks. The activated regions were consistent with many previous timing studies, but this analysis was critically unable to inform associations of brain regions with specific functions. Therefore, a second analysis focusing on modulation of activation patterns by selective attention isolated a fronto-parietal neural network including basal ganglia and anterior insula that supports selective attention to individual temporal stimulus features. Critically, activation patterns were inverted when the task involved ignoring the same temporal stimulus feature. The results demonstrate how the neural analysis of complex acoustic stimuli with multiple temporal features depends on a fronto-parietal network that simultaneously regulates the selective attentional gain as well as the selective ignoring of temporal features. Notes This work was supported by the Max Planck Society, Germany, through a Max Planck Research Group grant to J.O. We are grateful to Mandy Jochemko, Anke Kummer, and Simone Wipper for help with data acquisition, to Matthias Powelleit for help with experimental setup and data analysis, and to Merav Ahissar for fruitful initial discussions. We further thank 2 anonymous reviewers for their helpful comments on this manuscript. Conflict of Interest: None declared.

References Bengtsson SL, Ullen F, Ehrsson H, Hashimoto T, Kito T, Naito R, Forssberg H, Sadato N. 2009. Listening to rhythms activates motor and premotor cortices. Cortex. 45:62–71. Binder JR, Liebenthal E, Possing ET, Medler DA, Ward BD. 2004. Neural correlates of sensory and decision processes in auditory object identification. Nat Neurosci. 7:295–301. Bramen JE. 2008. Funtional mechanisms of auditory interval timing. Los Angeles: [ProQuest LLC] University of California.

Brett M, Anton J-L, Valabregue R, Poline JB. 2002. Region of interest analysis using the MarsBar toolbox for SPM 99. NeuroImage. 16: S497. Bueti D, Macaluso E. 2010. Auditory temporal expectations modulate activity in visual cortex. NeuroImage. 51:1168–1183. Bueti D, Macaluso E. 2011. Physiological correlates of subjective time: evidence for the temporal accumulator hypothesis. NeuroImage. 57:1251–1263. Buhusi C, Meck WH. 2005. What makes us tick? Functional and neural mechanisms of interval timing. Nat Rev Neurosci. 6:755–765. Casini L, Ivry RB. 1999. Effects of divided attention on temporal processing in patients with lesions of the cerebellum or frontal lobe. Neuropsychologia. 13:10–21. Chawla D, Rees G, Friston KJ. 1999. The physiological basis of attentional modulation in extrastriate visual areas. Nat Neurosci. 2:671–676. Coull JT, Cheng R-K, Meck WH. 2010. Neuroanatomical and neurochemical substrates of timing. Neuropsychopharmacology. 36:3–25. Coull JT, Nobre AC. 1998. Where and when to pay attention: the neural systems for directing attention to spatial locations and to time intervals as revealed by both PET and fMRI. J Neurosci. 18:7426–7435. Coull JT, Vidal F, Nazarian B, Macar F. 2004. Functional anatomy of the attentional modulation of time estimation. Science. 303:1506–1508. Culham JC, Cavanaugh P, Kanwisher NG. 2001. Attention response functions: characterizing brain areas using fMRI activation during parametric variations of attentional load. Neuron. 32:737–745. Dagher A, Owen AM, Boecker H, Brooks DJ. 1999. Mapping the network for planning: a correlational PET activation study with the Tower of London task. Brain. 122:1973–1987. Eckert MA, Menon V, Walczak A, Ahlstrom JB, Denslow S, Horwitz A, Dubno JR. 2009. At the heart of the ventral attention system: the right anterior insula. Hum Brain Mapp. 30:2530–2541. Erb J, Henry MJ, Eisner F, Obleser J. 2012. Auditory skills and brain morphology predict individual differences in adaptation to degraded speech. Neuropsychologia. 50:2154–2164. Erb J, Henry MJ, Eisner F, Obleser J. 2013. The brain dynamics of rapid perceptual adaptation to adverse listening conditions. J Neurosci. 33:10688–10697. Ferrandez AM, Hugueville L, Lehericy S, Poline JB, Marsault C, Pouthas V. 2003. Basal ganglia and supplementary motor area subtend duration perception: an fMRI study. NeuroImage. 19:1532–1544. Ghitza O, Greenberg S. 2009. On the possible role of brain rhythms in speech perception: intelligibility of time-compressed speech with periodic and aperiodic insertions. Phonetica. 66:113–126. Grahn JA, Henry MJ, McAuley JD. 2011. FMRI investigation of crossmodel interactions in beat perception: audition primes vision, but not vice versa. NeuroImage. 54:1231–1243. Grahn JA, Parkinson JA, Owen AM. 2008. The cognitive functions of the caudate nucleus. Prog Neurobiol. 86:141–155. Grahn JA, Parkinson JA, Owen AM. 2009. The role of the basal ganglia in learning and memory: neuropsychological studies. Behav Brain Res. 199:53–60. Gruber AJ, Dayan P, Gutkin BS, Solla SA. 2006. Dopamine modulation in the basal ganglia locks the gate to working memory. J Comp Neurosci. 20:153–166. Harrington DL, Boyd LA, Mayer AR, Sheltraw DM, Lee RR, Huang M, Rao SM. 2004. Neural representation of interval encoding and decision making. Brain Res Cogn Brain Res. 21:193–205. Harrington DL, Haaland KY, Knight RT. 1998. Cortical networks underlying mechanisms of time perception. J Neurosci. 1:1085–1095. Hinton SC, Meck WH. 2004. Frontal-striatal circuitry activated by human peak-interval timing in the supra-seconds range. Cogn Brain Res. 21:171–182. Hopfinger JB, Buonocore MH, Mangun GR. 2000. The neural mechanisms of top-down attentional control. Nat Neurosci. 3:284–291. Ivry RB, Keele SW. 1989. Timing functions of the cerebellum. J Cogn Neurosci. 1:136–152. Ivry RB, Schlerf JE. 2008. Dedicated and intrinsic models of time perception. Trends Cogn Sci. 12:273–280. Jahanshahi M, Jones CRG, Dirnberger G, Frith CD. 2006. The substantia nigra pars compacta and temporal processing. J Neurosci. 26:12266–12273.

Cerebral Cortex 9


processing more natural stimuli (i.e., speech), in particular under circumstances of degradation. Moreover, as mentioned above, Erb et al. (2013) found that modulation-rate discrimination and degraded speech processing rely on shared functional neural networks. In particular, brain activation scaled with task difficulty for modulation-rate discriminations and showed increased activation for degraded relative to clear speech in the SMA/ACC and insula. In the current study, we observed that activation in these same regions (in addition to parietal and frontal cortices) scaled with task difficulty for both duration and modulation-rate discrimination. From a more clinical perspective, a clear link exists between temporal processing deficits measured in psychophysical tasks and negative changes in speech comprehension abilities resulting from aging and hearing loss (for a review, see PichoraFuller 2003). Taking all of the reviewed findings together, we consider it likely that the same regions we observed in the current study support selective attention to temporal aspects of natural stimuli, such as speech, in particular when temporal cues are critical to comprehension (i.e., under degradation).

10 Selective attention to time

•

Henry et al.

Pessoa L, Kastner S, Ungerleider LG. 2003. Neuroimaging studies of attention: from modulation of sensory processing to top-down control. J Neurosci. 23:3990–3998. Pichora-Fuller MK. 2003. Processing speed and timing in aging adults: psychoacoustics, speech perception, and comprehension. Int J Audiol. 42:S59–S67. Poeppel D. 2003. The analysis of speech in different temporal integration windows: cerebral lateralization as asymmetric sampling in time. Speech Commun. 41:245–255. Pöppel E. 1997. A hierarchical model of temporal perception. Trends Cogn Sci. 1:56–61. Pöppel E. 1994. Temporal mechanisms in perception. Int Rev Neurobiol. 37:185–202. Pouthas V, George N, Poline JB, Pfeuty M, VandeMoorteele P-F, Hugueville L, Ferrandez AM, Lehericy S, LeBihan D, Renault B. 2005. Neural network involved in time perception: an fMRI study comparing long and short interval estimation. Hum Brain Mapp. 25:433–441. Power JD, Petersen SE. 2013. Control-related systems in the human brain. Curr Opin Neurobiol. 23:223–228. Rao SM, Mayer AR, Harrington DL. 2001. The evolution of brain activation during temporal processing. Nat Neurosci. 4:317–323. Rosen S. 1992. Temporal information in speech: acoustic, auditory, and linguistic aspects. Philos T Roy Soc B. 336:367–373. Saenz M, Buracas GT, Boynton GM. 2002. Global effects of featurebased attention in human visual cortex. Nat Neurosci. 5:631–632. Schoenfeld MA, Hopf J-M, Martinez A, Mai HM, Sattler C, Gasde A, Heinze H-J, Hillyard SA. 2007. Spatio-temporal analysis of featurebased attention. Cereb Cortex. 17:2468–2477. Schönwiesner M, Zatorre RJ. 2009. Spectro-temporal modulation transfer function of single voxels in the human auditory cortex measured with high-resolution fMRI. Proc Natl Acad Sci USA. 106:14611–14616. Schwartze M, Rothermich K, Kotz SA. 2012. Functional dissociation of pre-SMA and SMA-proper in temporal processing. NeuroImage. 60:290–298. Slotnik SD, Moo LR, Segal JB, Hart J Jr. 2003. Distinct prefrontal cortex activity associated with item memory and source memory for visual shapes. Brain Res Cogn Brain Res. 17:75–82. Sterzer P, Kleinschmidt A. 2010. Anterior insula activations in perceptual paradigms: often observed but barely understood. Brain Struct Funct. 214:611–622. Straube B, Green A, Weis S, Chatterjee A, Kircher T. 2008. Memory effects of speech and gesture binding: cortical and hippocampal activation in relation to subsequent memory performance. J Cogn Neurosci. 21:821–836. Tregellas JR, Davalos DB, Rojas DC. 2006. Effect of task difficulty on the functional anatomy of temporal processing. NeuroImage. 32:307–315. Ulbrich P, Churan J, Fink M, Wittmann M. 2007. Temporal reproduction: further evidence for two processes. Acta Psychol. 125: 51–65. von Kriegstein K, Warren JD, Ives DT, Patterson RD, Griffiths TD. 2006. Processing the acoustic effect of size in speech sounds. NeuroImage. 32:368–375. Wiener M, Turkeltaub P, Coslett HB. 2010. The image of time: a voxelwise meta-analysis. NeuroImage. 49:1728–1740.


Jueptner M, Rijntjes M, Weiller C, Fais JH, Timmann D, Mueller SP, Diener HC. 1995. Localization of a cerebellar timing process using PET. Neurology. 45:1540–1545. Kagerer FA, Wittmann M, Szelag E, Steinbüchel NV. 2002. Cortical involvement in temporal reproduction: evidence for differential roles of the hemispheres. Neuropsychologia. 40:357–366. Kanai R, Llyod H, Bueti D, Walsh V. 2011. Modality-independent role of the primary auditory cortex in time estimation. Exp Brain Res. 209:465–471. Kanwisher NG, Wojciulik E. 2000. Visual attention: insights from brain imaging. Nat Rev Neurosci. 1:91–100. Koch G, Oliveri M, Torriero S, Salerno S, Lo Gerfo E, Caltagirone C. 2007. Repetitive TMS of the cerebellum interferes with millisecond time processing. Exp Brain Res. 179:291–299. Kosillo P, Smith AT. 2010. The role of the human anterior insular cortex in time processing. Brain Struct Funct. 214:623–628. Kriegeskorte N, Goebel R, Bandetti P. 2006. Information-based functional brain mapping. Proc Natl Acad Sci USA. 103:3863–3868. Kudo K, Miyazaki M, Kimura T, Yamanaka K, Kadota H, Hirashima M, Nakajima Y, Nakazawa K, Ohtsuki T. 2004. Selective activation and deactivation of the human brain structures between speeded and precisely timed tapping responses to identical visual stimuli. NeuroImage. 22:1291–1301. Lewis PA, Miall RC. 2003. Brain activation patterns during measurement of sub- and supra-second intervals. Neuropsychologia. 41:1583–1592. Livesay AC, Wall MB, Smith AT. 2007. Time perception: manipulation of task difficulty dissociates clock functions from other cognitive demands. Neuropsychologia. 45:321–331. Macar F, Coull J, Vidal F. 2006. The supplementary motor area in motor and perceptual time processing: fMRI studies. Cogn Process. 7 (2):89–94. Macar F, Lejeune H, Bonnet M, Ferrara A, Pouthas V, Vidal F, Maquet P. 2002. Activation of the supplementary motor area and of attentional networks during temporal processing. Exp Brain Res. 142(4):475–485. Macar F, Vidal F, Casini L. 1999. The supplementary motor area in motor and sensory timing: evidence from slow brain potential changes. Exp Brain Res. 125:271–280. Mangels JA, Ivry RB, Shimizu N. 1998. Dissociable contributions of the prefrontal and neocerebellar cortex to time perception. Cogn Brain Res. 7:15–39. McGettigan C, Faulkner A, Altarelli I, Obleser J, Baverstock H, Scott SK. 2012. Speech comprehension aided by multiple modalities: behavioral and neural interactions. Neuropsychologia. 50:762–776. McNab F, Klingberg T. 2007. Prefrontal cortex and basal ganglia control access to working memory. Nat Neurosci. 11:103–107. Menon V, Uddin LQ. 2010. Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct. 214:655–667. Morillon B, Kell CA, Giraud A-L. 2009. Three stages and four neural systems in time estimation. J Neurosci. 29:14803–14811. Nenadic I, Gaser C, Volz HP, Rammsayer T, Hager F, Sauer H. 2003. Processing of temporal infomration and the basal ganglia: new evidence from fMRI. Exp Brain Res. 148:238–246. Obleser J, Meyer L, Friederici AD. 2011. Dynamic assignment of neural resources in auditory comprehension of complex sentences. NeuroImage. 4:2310–2320. Packard MG, Knowlton BJ. 2002. Learning and memory functions of the basal ganglia. Annu Rev Neurosci. 25:563–593.