Abstract

Recent studies have demonstrated that prestimulus alpha-band activity substantially influences perception of near-threshold stimuli. Here, we studied the influence of prestimulus alpha power fluctuations on temporal perceptual discrimination of suprathreshold tactile stimuli and subjects' confidence regarding their perceptual decisions. We investigated how prestimulus alpha-band power influences poststimulus decision-making variables. We presented electrical stimuli with different stimulus onset asynchronies (SOAs) to human subjects, and determined the SOA for which temporal perceptual discrimination varied on a trial-by-trial basis between perceiving 1 or 2 stimuli, prior to recording brain activity with magnetoencephalography. We found that low prestimulus alpha power in contralateral somatosensory and occipital areas predicts the veridical temporal perceptual discrimination of 2 stimuli. Additionally, prestimulus alpha power was negatively correlated with confidence ratings in correctly perceived trials, but positively correlated for incorrectly perceived trials. Finally, poststimulus event-related fields (ERFs) were modulated by prestimulus alpha power and reflect the result of a decisional process rather than physical stimulus parameters around ∼150 ms. These findings provide new insights into the link between spontaneous prestimulus alpha power fluctuations, temporal perceptual discrimination, decision making, and decisional confidence. The results suggest that prestimulus alpha power modulates perception and decisions on a continuous scale, as reflected in confidence ratings.

Introduction

Decision making can be understood as a process in which sensory evidence is accumulated in a decision variable. If sensory evidence is sufficiently strong and available for a sufficiently long time, the decision variable accumulates until a decision bound for either decision is reached (see Gold and Shadlen 2007 for a review). In some situations, however, sensory evidence is ambiguous, providing equal sensory evidence for each decisional option. In other situations, sensory evidence is weak or presented insufficiently long for the decision variable to reach a decision bound. Consequently, decisions have to be made based on incomplete or equivocal sensory evidence, frequently causing incorrect decisions and low confidence in the decision. In addition, decision making is not only determined by sensory evidence, but also by trial-to-trial fluctuations of neuronal activity, usually interpreted as internal noise (Ratcliff and McKoon 2007; O'Connell et al. 2012).

Recent studies, however, demonstrated that fluctuations of neuronal activity can have a functional role for the perception of weak and ambiguous stimuli. Specifically neuronal oscillatory activity in the alpha band (∼8–12 Hz) has drawn much attention. Prestimulus alpha power is modulated by attention (e.g., Foxe et al. 1998; Worden et al. 2000) and prestimulus power and phase in early sensory areas are correlated with perception (Linkenkaer-Hansen et al. 2004; van Dijk et al. 2008; Mazaheri et al. 2009; Wyart and Tallon-Baudry 2009; Jensen and Mazaheri 2010; Romei et al. 2010; Keil et al. 2014). Furthermore, it has been shown that prestimulus oscillatory activity can influence poststimulus evoked responses (Başar et al. 1984; Brandt and Jansen 1991; Mazaheri and Jensen 2008; Jones et al. 2009, 2010; Anderson and Ding 2011; Lange et al. 2012). The influence of prestimulus oscillatory activity on decision variables remains largely unknown. In addition, the influence of prestimulus oscillatory activity on subjective confidence in perceptual decisions is unknown. Subjective confidence represents a measure of the degree to which a decision maker believes in the correctness of his decisions and thus provides an insight into decisional processes on a fine-grained scale (Kiani and Shadlen 2009). Moreover, it remains unexplained how the brain forms decisions when sensory evidence is insufficient to reach a decision bound, for example, due to sensory ambiguity.

To test how prestimulus alpha-band power biases perceptual decisions and the underlying neuronal decision variable in humans, we presented electrical stimuli with different stimulus onset asynchronies (SOAs) and compared 2 subjectively ambiguous experimental conditions in which physically identical tactile stimuli were perceived differently on a trial-by-trial basis. We used magnetoencephalography (MEG) and a forced-choice temporal perceptual discrimination task to investigate whether fluctuations of prestimulus neuronal oscillatory activity are related to the trial-to-trial variability of decisions and how prestimulus oscillatory activity influences the decision variable. We hypothesized that prestimulus alpha power correlates with temporal perceptual discrimination rate, with lower alpha power levels related to increased veridical temporal perceptual discrimination. Further, we expected that characteristics of the decision-making process would be evident in neural activity in the form of poststimulus event-related fields (ERFs). This should result in differences of neuronal activity for trials with different decisional outcomes, despite identical physical stimulation. Additionally, we hypothesized that prestimulus alpha power would influence this decision-related neuronal activity.

Materials and Methods

Subjects

Sixteen, right-handed subjects (7 males, age: 26.1 ± 4.7 years [mean ± SD]) participated in the study after providing written informed consent in accordance with the Declaration of Helsinki. All participants had normal or corrected-to-normal vision and reported no somatosensory deficits or known history of neurological disorders.

Experimental Design and Paradigm

The experimental task was designed to compare 2 conditions with identical physical stimuli, differing only in the participant's perception. Each trial started with the presentation of a start cue (500 ms; Fig. 1). Next, the cue decreased in luminance, indicating the prestimulus period (900–1100 ms), after which the subjects received either 1 or 2 short (0.3 ms) electrical pulses, applied by 2 electrodes placed between the 2 distal joints of the left index finger. The amplitude of the pulses was determined individually to a level clearly above subjective perception threshold, but below pain threshold (4.1 ± 1.2 mA [mean ± SD]). Note that all comparisons of conditions were performed at the within-subject level. Therefore, only conditions with identical stimulation parameters were compared (for details, see MEG Data Acquisition and Analysis). The electrical pulses were applied with varying SOAs: short (0 ms, i.e., only one stimulus was applied), long (100 ms), and 3 SOAs individually determined in a premeasurement. These 3 individual SOAs included a SOA for which subjects reported to perceive one electrical pulse in ∼50% of the trials, whereas in the other ∼50% of the trials 2 pulses were perceived (SOA: 25.9 ± 1.9 ms (mean ± standard error of the mean [SEM])). Subsequently, this condition will be labeled the intermediate SOA. The remaining 2 SOAs encompassed the intermediate SOA by ±10 ms and were included to minimize learning effects and response biases. A training phase of ∼5 min containing all possible SOAs preceded the experiment to familiarize subjects with the paradigm. The electrical stimulation was followed by a jittered poststimulus period of 500–1200 ms to minimize motor preparation effects, during which the fixation dot remained visible. Next, a written instruction indicated the start of the first response window. Subjects first reported whether they perceived the stimulation as 1 single or 2 temporally separate sensations. Responses were given by button-presses with the index or middle finger of the right hand, while button configurations were randomized from trial to trial to minimize motor preparation effects. Subjects were instructed to report within 3000 ms after presentation of response instructions. Due to the jittered poststimulus epoch (500–1200 ms) which determined the beginning of the subsequent response window, response speed was not taken into account. If no response was given after 3000 ms or the subject responded before the presentation of the instructions, a warning was presented visually. The respective trial was discarded from analysis and repeated at the end of the block. After reporting their subjective perception, written instructions indicated a second response window. Here, subjects rated their subjective confidence level regarding their first response. The confidence level was assessed via a 4-point rating scale, ranging from “very sure” to “very unsure.” Once both responses were given, the next trial started. With the exception of the aforementioned warning signal, no further feedback was given. All visual stimuli were projected on the backside of a translucent screen (60 Hz refresh rate) positioned 60 cm in front of the subjects.

Figure 1.

Experimental task. Sequence of events: A central fixation dot serves as start cue, after 500 ms a decrease in luminance signals the start of the prestimulus epoch, consisting of a jittered period of 900–1100 ms. Tactile stimulation is applied to the left index finger with varying SOAs (0 ms, intermediate – 10 ms, intermediate, intermediate + 10 ms, 100 ms). After a jittered poststimulus period (500–1200 ms), written instructions indicate the first response window and subjects report their perception of the stimulation by button-press. Subsequently, written instructions indicate the second response window and subjects report their decisional confidence by button-press.

Figure 1.

Experimental task. Sequence of events: A central fixation dot serves as start cue, after 500 ms a decrease in luminance signals the start of the prestimulus epoch, consisting of a jittered period of 900–1100 ms. Tactile stimulation is applied to the left index finger with varying SOAs (0 ms, intermediate – 10 ms, intermediate, intermediate + 10 ms, 100 ms). After a jittered poststimulus period (500–1200 ms), written instructions indicate the first response window and subjects report their perception of the stimulation by button-press. Subsequently, written instructions indicate the second response window and subjects report their decisional confidence by button-press.

Each SOA was presented in 50 trials. To increase statistical power, the intermediate SOA was presented 4 times as often as the other SOAs (i.e., 200 trials). 80 trials constituted one block with each block containing 10 repetitions (40 for the intermediate condition) of each SOA presented in pseudorandom order. Blocks were repeated 5 times, interrupted by self-paced breaks of ∼2 min, resulting in an overall 400 trials. The approximate total duration of the MEG measurement was ∼45–50 min (400 trials with a trial length of ∼6 s on average [4–8.6 s], interrupted by up to 4 self-paced breaks of ∼2 min).

Stimulus presentation was controlled using Presentation software (Neurobehavioral Systems, Albany, NY, USA). Before MEG recording, each subject received instructions of the task but remained naïve to the purpose of the experiment and the different SOAs used.

Behavioral Data Analysis

Behavioral data were analyzed with regard to correct responses and compared across conditions by means of a paired sample t-test. Prior to this, a Kolgomorov–Smirnov test was applied to ensure that the respective distributions did not differ from a Gaussian distribution. Further, we investigated learning/fatigue trends in the perceptual responses and confidence ratings by dividing experimental trials in 12 bins and computing the average temporal perceptual discrimination rate (i.e., perceived 2 stimuli or 1 stimulus) as well as the average confidence rating over subjects for each bin. Subsequently, we fitted a linear regression to the data in order to determine a linear trend.

MEG Data Acquisition and Analysis

Data Recording and Preprocessing

Ongoing neuromagnetic brain activity was continuously recorded at a sampling rate of 1000 Hz using a 306-channel whole head MEG system (Neuromag Elekta Oy, Helsinki, Finland), including 204 planar gradiometers (102 pairs of orthogonal gradiometers) and 102 magnetometers. Data analysis in the present study was restricted to the planar gradiometers. Additionally, electro-oculograms were recorded for offline artifact rejection by applying electrodes above and below the left eye as well as on the outer sides of each eye. Subjects' head position within the MEG helmet was registered by a head position indication system (HPI) built up of 4 coils placed at subjects' forehead and behind both ears. A 3-T MRI scanner (Siemens, Erlangen, Germany) was used to obtain individual full-brain high-resolution standard T1-weighted structural magnetic resonance images (MRIs). The MRIs were offline aligned with the MEG coordinate system using the HPI coils and anatomical landmarks (nasion, left and right preauricular points).

Data were offline analyzed using custom-made Matlab (The Mathworks, Natick, MA, USA) scripts, the Matlab-based open source toolbox FieldTrip (http://fieldtrip.fcdonders.nl; Oostenveld et al. 2011), and SPM8 (Litvak et al. 2011). Continuously recorded data were segmented into trials, starting with the appearance of the first fixation dot and ending with the second response of the subject. All trials were semiautomatically and visually inspected for artifacts, whereas artifacts caused by muscle activity, eye movements, or SQUID jumps were removed semiautomatically using a z-score-based algorithm implemented in FieldTrip. Excessively noisy channels were removed as well and reconstructed by an interpolation of neighboring channels. In addition, power line noise was removed from the segmented data by using a band-stop filter encompassing the 50, 100, and 150 Hz components. Further preprocessing steps were applied according to the respective analyses.

Time–Frequency Analysis

For exploratory reasons, we first performed a time–frequency analysis on all frequencies between 2 and 40 Hz for all time points (−900 to 500 ms, Fig. 2A). We focused our analysis on the effects of alpha power (8–12 Hz) in the prestimulus epoch (−900 to 0 ms) on perceptual decisions, that is, the responses to the temporal perceptual discrimination task. First, the linear trend and mean of every epoch were removed from each trial. Time–frequency representations for each trial were computed by applying a Fourier transformation on adaptive sliding time windows containing 7 full cycles of the respective frequency ft = 7/f), moved in steps of 50 ms and 2 Hz (van Dijk et al. 2008; Mazaheri et al. 2009; Lange et al. 2012). Data segments were tapered with a single Hanning taper, resulting in a spectral smoothing of 1/Δt. Spectral power was averaged over the alpha band separately for each trial. Alpha power was estimated independently for each of the 204 gradiometers. Subsequently, gradiometer pairs were combined by summing spectral power across the 2 orthogonal channels, resulting in 102 pairs of gradiometers. We sorted the trials with respect to the SOA for each subject separately. For all trials with intermediate SOA, we separated and compared trials with reports of 1 perceived stimulus to trials with 2 perceived stimuli. With this approach, we were able to compare 2 sets of decisional outcomes, which differed only in the subjects' temporal perceptual discrimination of the stimuli, though not regarding their physical properties. Due to the fact that, only for the intermediate condition, a sufficiently high number of trials for both decisional outcomes (perceived 1 stimulus or 2 stimuli) were available, only trials with intermediate SOA entered the analysis. In the following, trials in which stimulation was perceived as 2 temporally separate stimuli will be labeled correctly perceived trials, whereas trials in which stimulation was perceived as 1 single stimulus will be labeled incorrectly perceived trials. To test for statistically significant power differences between sets, we used a cluster-based nonparametric randomization approach (Maris and Oostenveld 2007). In a first step, we compared averaged alpha power between both sets of decisional outcomes (correct and incorrect, i.e., perceived 2 stimuli or 1 stimulus) for each subject independently in all channels and all time points in the prestimulus time window (−900 to 0 ms). Comparison between sets was performed by subtracting the power of both sets and dividing the difference by the variance (equivalent to an independent sample t-test). This step serves as a normalization of interindividual differences (Hoogenboom et al. 2010; Lange et al. 2011, 2013). The comparison was done independently for each time sample and channel, resulting in a time-channel map of pseudo-t-values for each subject. For group-level statistics, we analyzed the consistency of pseudo-t-values over subjects by means of a nonparametric randomization test identifying clusters in time-channel space showing the same effect. Neighboring channels were defined on the basis of spatial adjacency, with spatial clusters requiring a minimum amount of 2 neighboring channels. Spatially and temporally adjacent pseudo-t-values exceeding an a priori-defined threshold (P < 0.05) were combined to a cluster. t-values within a cluster were summed up and used as input for the second-level cluster statistic. Next, we computed a reference distribution by randomly permuting the data, assuming no differences between statistical conditions and exchangeability of the data. This process of random assignment was repeated 1000 times, resulting in a summed cluster t-value for each repetition. The proportion of elements in the reference distribution exceeding the observed maximum cluster-level test statistic was used to estimate a P-value for each cluster. This statistical approach effectively controls for the Type I error rate due to multiple comparisons across time points and channels (Maris and Oostenveld 2007).

Figure 2.

Results of the statistical comparison of correctly (perceived 2 stimuli) versus incorrectly (perceived 1 stimulus) perceived trials with intermediate SOA. (A) Time–frequency representation on sensor level averaged over all sensors. t = 0 indicates onset of sensory stimulation. (B) Time series of topographical representations on sensor level averaged over the alpha band (8–12 Hz). Significant sensors (P < 0.05) are marked by white circles. The lower right inset illustrates alpha power differences averaged across the whole time window (−900 to −250 ms; white dots represent channels of the anterior/somatosensory sensor-cluster; black crosses represent channels of the parieto-occipital sensor-cluster used for following analyses. See text for details on the separation of the clusters). (C) Source reconstruction projected on the MNI template brain for the significant effect in the alpha band (see B) viewed from the top (top row) and the right (bottom row). Source plots are masked to highlight significant clusters (P < 0.05). P-values in B and C are cluster corrected to account for multiple comparison corrections. The left color bar applies to A, the right color bar applies to B and C. For both color bars, blue colors indicate lower spectral power in correctly perceived trials compared with incorrectly perceived trials.

Figure 2.

Results of the statistical comparison of correctly (perceived 2 stimuli) versus incorrectly (perceived 1 stimulus) perceived trials with intermediate SOA. (A) Time–frequency representation on sensor level averaged over all sensors. t = 0 indicates onset of sensory stimulation. (B) Time series of topographical representations on sensor level averaged over the alpha band (8–12 Hz). Significant sensors (P < 0.05) are marked by white circles. The lower right inset illustrates alpha power differences averaged across the whole time window (−900 to −250 ms; white dots represent channels of the anterior/somatosensory sensor-cluster; black crosses represent channels of the parieto-occipital sensor-cluster used for following analyses. See text for details on the separation of the clusters). (C) Source reconstruction projected on the MNI template brain for the significant effect in the alpha band (see B) viewed from the top (top row) and the right (bottom row). Source plots are masked to highlight significant clusters (P < 0.05). P-values in B and C are cluster corrected to account for multiple comparison corrections. The left color bar applies to A, the right color bar applies to B and C. For both color bars, blue colors indicate lower spectral power in correctly perceived trials compared with incorrectly perceived trials.

Source Reconstruction

To identify the cortical sources of the statistically significant effects displayed on sensor level, we calculated source-level power estimates by means of an adaptive spatial filtering technique (DICS, Gross et al. 2001). To this end, a regular 3D grid with 1 cm resolution was applied to the Montreal Neurological Institute (MNI) template brain. Individual grids for each subject were computed by linearly warping the structural MRI of each subject onto the MNI template brain and applying the inverse of the warp to the MNI template grid. For one subject, no individual structural MRI was available; hence, we used the MNI template brain instead. A lead-field matrix was computed for each grid point employing a realistically shaped single-shell volume conduction model (Nolte 2003). Subsequently, the cross-spectral density (CSD) matrix between all MEG gradiometer sensor pairs was computed for the alpha band by applying a Fourier transformation on time windows of interest. Time windows of interest were based on the significant clusters of the group analysis on sensor level (Fig. 2B). Using the CSD and lead-field matrix, common spatial filters were constructed for each individual grid point. To this end, we pooled trials with intermediate SOA over both sets of decisional outcomes and computed a common spatial filter for each subject. CSD matrices of single trials were projected through those filters, resulting in single-trial estimates of source power (Hoogenboom et al. 2010; Lange et al. 2012), and further sorted according to decisional outcome. In line with the analysis on sensor level, power was contrasted between both sets of decisional outcomes. Similarly to the sensor-level analysis, the resulting individual source parameters were statistically compared across subjects by means of a nonparametric randomization test (Maris and Oostenveld 2007) which effectively controls for the Type I error rate. Group results were displayed on the MNI template brain in form of t-values. Finally, cortical sources were identified using the AFNI atlas (http://afni.nimh.nih.gov/afni), integrated into FieldTrip.

Since the time–frequency analysis and the source reconstruction demonstrated 2 spatiotemporally different activation clusters (see Results and Fig. 2B,C), we performed the subsequent analyses on 2 different sensor sets. First, we based the analyses on all channels showing a significant alpha power difference between correctly (perceived 2 stimuli) versus incorrectly (perceived 1 stimulus) perceived trials with intermediate SOA (as shown in Fig. 2B). Second, we based the analyses on 2 spatiotemporally separated sensor-clusters (see inset in Fig. 2B), 1 anterior/somatosensory sensor-cluster (MEG-sensors: MEG0712 + 13, MEG0722 + 23, MEG1042 + 43, MEG1112 + 13, MEG1122 + 23, MEG1132 + 33, MEG1142 + 43, MEG1312 + 13, MEG1342 + 43, MEG1832 + 33, MEG2012 + 13, MEG2022 + 23, MEG2212 + 13, MEG2222 + 23, MEG2232 + 33, MEG2242 + 43, MEG2412 + 13, MEG2422 + 23, MEG2612 + 13, MEG2642 + 43), and 1 parieto-occipital sensor-cluster (MEG-sensors: MEG2312 + 13, MEG2322 + 23, MEG2342 + 43, MEG2432 + 33, MEG2442 + 43, MEG2512 + 13, MEG2522 + 23).

Correlation of Prestimulus Power, Perceptual Decisions, and Confidence Ratings

To examine the relationship between prestimulus power and perceptual decisions, we averaged spectral power over time, frequency, and sensors and correlated averaged power values with perceptual decisions. To this end, we selected the sensors and time points showing a significant difference between decisional outcomes (see above, Fig. 2B). Note that this approach resembles a post hoc statistical analysis in the sense that sensor selection was based on those sensors showing a significant difference in the alpha band (see above, Fig. 2B). Averaging was done separately for each subject and trial, using a fixed time–frequency-sensor triplet resulting from the significant time-channel clusters derived from group-level statistics and the predetermined alpha frequency (8–12 Hz). Trials of each subject were sorted from low to high alpha power and divided into 5 bins (Linkenkaer-Hansen et al. 2004; Jones et al. 2010; Lange et al. 2012, 2013). For each bin and subject, we calculated the average temporal perceptual discrimination rate and normalized the resulting value for each bin to the individual average temporal perceptual discrimination rate across all bins by first subtracting and then dividing by the individual averaged temporal perceptual discrimination rate across all trials. This resulted in a percentage change relative to the normalized mean across all bins for each subject (Linkenkaer-Hansen et al. 2004; Lange et al. 2012, 2013). For each bin, average power and SEM were computed over all subjects. Linear and quadratic functions were fitted to the data to determine the best fit (Linkenkaer-Hansen et al. 2004; van Dijk et al. 2008; Jones et al. 2010; Lange et al. 2012, 2013). Average temporal perceptual discrimination rates in the respective bins were statistically compared by applying a one-way repeated-measures ANOVA and post hoc t-tests.

Additionally, we investigated the correlation of prestimulus power and confidence ratings. The analysis was conducted as stated above, with the following exceptions. To separately determine the relation between prestimulus power and confidence rating for correctly and incorrectly perceived trials, we divided the trials with intermediate SOA regarding their decisional outcome, that is, correctly and incorrectly perceived trials were analyzed separately. For each bin, we calculated the average confidence rating and normalized the result in each bin to the average confidence rating across all trials with the respective decisional outcome. Finally, the average confidence ratings were averaged over subjects. Likewise, linear and quadratic functions were fitted to the data.

Further, we separated the significant channels in 2 clusters (anterior/somatosensory vs. parieto-occipital; see above and inset of Fig. 2B) based on their spatiotemporal characteristics and performed the correlation analysis with power values averaged over the channels of these separated sensor-clusters.

Relation Between Decision Variable, Prestimulus Power, and Poststimulus ERFs

To examine the neural dynamics of perceptual decision making under conditions with suboptimal evidence accumulation and ambiguous stimulus perception, we studied the relation of poststimulus ERFs, prestimulus alpha power and decisional outcome. Perceptual decisions can be conceptualized as a process in which sensory evidence for a decision accumulates over time in a decision variable until a decision bound is reached, followed by a particular response selection (Gold and Shadlen 2007; Ratcliff and McKoon 2007; Kiani and Shadlen 2009). Recent works in human electrophysiology suggest that such decision variables are reflected in poststimulus event-related potentials (e.g., VanRullen and Thorpe 2001; Philiastides and Sajda 2006; Philiastides et al. 2006; O'Connell et al. 2012). Since event-related potentials/fields resemble a population-based measure of neuronal activity (Hari and Kaukoranta 1985), this is further supported by studies that identify signals from multiple neurons as basis of behavioral decisions (Britten et al. 1996). We hypothesized that, in trials with intermediate SOA, the total accumulation of sensory evidence would remain below any decisional bound due to insufficient sensory information in favor of any decision, therefore requiring forced-choice decisions. We aimed to assess these decision variables in poststimulus ERFs. Additionally, confidence levels should be a function of the distance of the decision variable to the decision bounds, with closer proximity of the decision variable to the respective decision bound resulting in higher confidence. Moreover, we hypothesized that prestimulus alpha power modulates the distance of the decision variable to the respective decision bounds.

To compute ERFs, preprocessed data were filtered between 2 and 40 Hz, the mean of each epoch was removed from each trial, and these data were averaged across trials. For each subject, ERFs were computed for all sensors that showed a significant difference between decisional outcomes (as shown in Fig. 2B). Additionally, we separated the significant channels in 2 spatial clusters (anterior/somatosensory vs. parieto-occipital, see inset of Fig. 2B) based on their spatiotemporal characteristics and calculated ERFs for all sensors of the respective sensor-cluster separately. To avoid cancelation effects when averaging across sensors and subjects, the signals of the 2 orthogonal sensors of each gradiometer pair were combined by taking the root mean square of the signals in the time domain (e.g., van Dijk et al. 2008; Lange et al. 2012), resulting in 102 gradiometer pairs. Poststimulus ERFs were baseline corrected by subtracting the mean of the prestimulus period (−900 to 0 ms). First, we determined potential poststimulus decision boundaries in the poststimulus ERFs. To this end, we computed ERFs for the 2 conditions with 0 and 100 ms SOA as they provided the most unambiguous perception of 1 and 2 stimuli. Only trials with correct responses (i.e., perceived 1 stimulus for trials with SOA 0 ms and perceived 2 stimuli for trials with SOA 100 ms) were included in this analysis, with conditions subsequently labeled as 0ms-1 and 100ms-2. We statistically compared the ERFs in the poststimulus period (0–300 ms) to identify time periods that maximally discriminated between these 2 reference conditions with 0 and 100 ms SOA. We used a nonparametric statistical test which effectively controls for the Type I error rate due to multiple comparisons across time points in line with the procedure described above (for details, see Time–Frequency Analysis). In brief, we calculated the difference between both ERFs for each subject, followed by a group-level statistic testing the consistency of the differences across subjects against a reference null distribution based on 1000 random sets of permutations regarding the 2 experimental conditions.

Next, we examined whether the ERFs reflect a decision variable that is independent of sensory input, but differing according to subject's decisional outcome. To this end, we sorted trials with intermediate SOA in trials with correct and incorrect responses. We hypothesized that due to their ambiguity and insufficient accumulation of sensory evidence, the decision bounds (i.e., ERFs of conditions 0ms-1 and 100ms-2) will not be reached in trials with intermediate SOA. Nonetheless, because of the implemented forced-choice task, subjects are forced to make the decision with a particular level of uncertainty. We hypothesized that confidence levels should be a function of the distance of the decision variable to the decision bounds, with closer proximity of the decision variable to the respective decision bound resulting in higher confidence. Moreover, we hypothesized that prestimulus alpha power has a distinguishable effect on the decision variable. Since prestimulus alpha power significantly influenced temporal perceptual discrimination and confidence ratings (Fig. 3A,B), an effect of prestimulus alpha power should be visible in the poststimulus decision variable. We hypothesized that prestimulus alpha power modulates the distance of the decision variable to the respective decision bounds (Fig. 4B). To this end, we averaged prestimulus alpha power across those time points and sensors that showed a significant difference between decisional outcomes (see above, Fig. 2B) and grouped the trials with intermediate SOA into correct and incorrect trials with either high and low prestimulus alpha power. This resulted in 4 different conditions: low prestimulus alpha power and perceived 2 stimuli (subsequently labeled low α-2), high prestimulus alpha power and perceived 2 stimuli (high α-2), low prestimulus alpha power and perceived 1 stimulus (low α-1), high prestimulus alpha power and perceived 1 stimulus (high α-1). We then computed poststimulus ERFs for each of these conditions.

Figure 3.

Results of the post hoc correlation analyses of averaged prestimulus alpha power (8–12 Hz) for significant sensors (as shown in Fig. 2B) and (A) normalized average temporal perceptual discrimination rate or (B) normalized confidence ratings, separated for correctly and incorrectly perceived trials. Insets show results of the linear regression analyses (black and gray lines). Higher number bins indicate higher spectral power. Error bars represent SEM. **P < 0.01, *P < 0.05.

Figure 3.

Results of the post hoc correlation analyses of averaged prestimulus alpha power (8–12 Hz) for significant sensors (as shown in Fig. 2B) and (A) normalized average temporal perceptual discrimination rate or (B) normalized confidence ratings, separated for correctly and incorrectly perceived trials. Insets show results of the linear regression analyses (black and gray lines). Higher number bins indicate higher spectral power. Error bars represent SEM. **P < 0.01, *P < 0.05.

Figure 4.

Results of the analysis of poststimulus ERFs. (A) Statistical comparison of poststimulus ERF amplitudes (averaged over all significant sensors, as shown in Fig. 2B) of correctly perceived trials with 0 ms (0ms-1) and 100 ms (100ms-2) SOA. Significant differences are indicated by shaded area (145–171 ms). The dashed line represents the point of maximum amplitude difference between the reference conditions 0ms-1 and 100ms-2 (150 ms). The blue arrow highlights the time point of stimulation for the 0ms-1 condition, while the red arrows highlight the time points of stimulation for the 100ms-2 condition. (B) Predicted poststimulus ERFs. Decision model illustrating the hypothesized order of poststimulus ERFs. Conditions 0ms-1 and 100ms-2 reflect the decision bounds for perceiving 1 and 2 stimuli, respectively. The other conditions are predicted to be between these bounds in the presented order. Distance to the bound is hypothesized to reflect confidence in the decision (indicated by gray-shaded background). The dashed line represents the point of maximum amplitude difference between the reference conditions 0ms-1 and 100ms-2. Beyond this point, the decision variables are thought to decline again to baseline. (C) MEG data of poststimulus ERFs. Close-up on the time window of significant difference (145–171 ms; shaded area) between poststimulus ERF amplitudes of 0ms-1 and 100ms-2 (averaged over all significant sensors, as shown in Fig. 2B). Shaded area and dashed line as in A. Color scheme as in B. (D) Same as A, but now for amplitude values averaged over the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 139–172 ms; time point of maximum amplitude difference: 151 ms). Blue and red arrows as in A. (E) Same as C, but now for amplitude values averaged over the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 139–172 ms; time point of maximum amplitude difference: 151 ms). Shaded area and dashed line as in D. (F) Same as A, but now for amplitude values averaged over the parieto-occipital sensor-cluster (as shown in Fig. 2B). Blue and red arrows as in A. No statistically significant difference was found. Significance values in AF are cluster corrected to account for multiple comparison corrections. t = 0 indicates onset of sensory stimulation, that is, the first stimulus of every stimulation.

Figure 4.

Results of the analysis of poststimulus ERFs. (A) Statistical comparison of poststimulus ERF amplitudes (averaged over all significant sensors, as shown in Fig. 2B) of correctly perceived trials with 0 ms (0ms-1) and 100 ms (100ms-2) SOA. Significant differences are indicated by shaded area (145–171 ms). The dashed line represents the point of maximum amplitude difference between the reference conditions 0ms-1 and 100ms-2 (150 ms). The blue arrow highlights the time point of stimulation for the 0ms-1 condition, while the red arrows highlight the time points of stimulation for the 100ms-2 condition. (B) Predicted poststimulus ERFs. Decision model illustrating the hypothesized order of poststimulus ERFs. Conditions 0ms-1 and 100ms-2 reflect the decision bounds for perceiving 1 and 2 stimuli, respectively. The other conditions are predicted to be between these bounds in the presented order. Distance to the bound is hypothesized to reflect confidence in the decision (indicated by gray-shaded background). The dashed line represents the point of maximum amplitude difference between the reference conditions 0ms-1 and 100ms-2. Beyond this point, the decision variables are thought to decline again to baseline. (C) MEG data of poststimulus ERFs. Close-up on the time window of significant difference (145–171 ms; shaded area) between poststimulus ERF amplitudes of 0ms-1 and 100ms-2 (averaged over all significant sensors, as shown in Fig. 2B). Shaded area and dashed line as in A. Color scheme as in B. (D) Same as A, but now for amplitude values averaged over the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 139–172 ms; time point of maximum amplitude difference: 151 ms). Blue and red arrows as in A. (E) Same as C, but now for amplitude values averaged over the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 139–172 ms; time point of maximum amplitude difference: 151 ms). Shaded area and dashed line as in D. (F) Same as A, but now for amplitude values averaged over the parieto-occipital sensor-cluster (as shown in Fig. 2B). Blue and red arrows as in A. No statistically significant difference was found. Significance values in AF are cluster corrected to account for multiple comparison corrections. t = 0 indicates onset of sensory stimulation, that is, the first stimulus of every stimulation.

To quantify the relation between these conditions, we chose 2 parallel approaches to determine a time window of interest. In the first approach, we averaged ERF amplitudes for each condition over those time points showing a significant difference between the conditions 0ms-1 and 100ms-2 (i.e., 145–171 ms; see above and Fig. 4A; see Supplementary Fig. 1 for the complete ERF time courses of all conditions). In the second approach, we determined the time point of maximum amplitude difference between the conditions 0ms-1 and 100ms-2 within those time points showing a significant difference between the conditions (150 ms). Please see the Discussion section for a further discussion on the selection criteria for the time window of interest. We averaged ERF amplitudes for each condition over the 10 ms that precede this time point of maximal difference (i.e., 140–150 ms). The rationale of this approach was that decision variables are thought to increase until a decision bound is reached and decline again afterwards to baseline (Kiani and Shadlen 2009; O'Connell et al. 2012). Thus, the time point of maximal difference between the reference conditions and the preceding time window should be the best predictor of the decision process (see model in Fig. 4B).

For the additional analyses based on separated sensor-clusters, significant differences between the conditions 0ms-1 and 100ms-2 could be demonstrated from 139 to 172 ms (see Fig. 4D) and the point of maximum amplitude difference was located at 151 ms for the anterior/somatosensory sensor-cluster. For the parieto-occipital cluster, no significant differences between the conditions 0ms-1 and 100ms-2 could be demonstrated (see Fig. 4F). To ensure that the absence of significant differences for the parieto-occipital cluster did not result from low statistical power due to a lower number of channels in this cluster (parieto-occipital cluster: 7 channel pairs, anterior/somatosensory sensor-cluster: 20 channel pairs), we further compared the conditions 0ms-1 and 100ms-2 for a random selection of 7 channel pairs from the anterior/somatosensory sensor-cluster. The results of this analysis reproduced the significant differences between the conditions 0ms-1 and 100ms-2 (139–169 ms; data not shown) as well as a significant negative linear correlation for the ordered averaged ERFs (i.e., 100ms-2, low α-2, high α-2, low α-1, high α-1, 0ms-1; r = −0.96, P < 0.01 for time window 139–169 ms; r = −0.87, P < 0.05 for time window 139–149 ms; data not shown). Based on these results, we conclude that the absent significant difference between the conditions 0ms-1 and 100ms-2 for the parieto-occipital cluster cannot be generally explained by the smaller number of channels in this cluster, but instead must be mainly attributed to the absence of decision-related ERF components in the parieto-occipital sensor-cluster.

For both sensor sets (all significant sensors and the anterior/somatosensory sensor-cluster), we subsequently ordered the conditions regarding the expected averaged ERF amplitudes (100ms-2, low α-2, high α-2, low α-1, high α-1, 0ms-1) and fitted a linear regression to the data to determine a linear trend (Fig. 5). Due to the a priori difference of the conditions 100ms-2 and 0ms-1, we performed an additional analysis in which we excluded these conditions from the regression analysis. Hence, the regression analysis was additionally calculated for the ordered intermediate conditions (low α-2, high α-2, low α-1, high α-1) only. Averaged ERF amplitudes were statistically compared by applying a one-way repeated-measures ANOVA. Because no time window showing a significant difference between the conditions 0ms-1 and 100ms-2 was found for the parieto-occipital sensor-cluster, we refrained from performing this analysis for the parieto-occipital sensor-cluster.

Figure 5.

Averaged amplitude values and confidence ratings of poststimulus ERFs. (A) Amplitude values (based on all significant sensors, as shown in Fig. 2B) averaged over the time window showing a significant difference between poststimulus ERF amplitudes of 0ms-1 and 100ms-2 (145–171 ms, see Fig. 4A,C). (B) Amplitude values (based on all significant sensors, as shown in Fig. 2B) averaged over the time window preceding the point of maximal amplitude difference (150 ms) between poststimulus ERF amplitudes of 0ms-1 and 100ms-2 (140–150 ms, see Fig. 4A,C). (C) Average confidence ratings per condition in relation to mean power difference to the respective decision bound (based on all significant sensors, as shown in Fig. 2B) for the time window 140 to 150 ms (see Fig. 4A,C). (D) Same as A, but now for amplitude values based on the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 139–172 ms, Fig. 4D,E). (E) Same as B, but now for amplitude values based on the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 141–151 ms, see Fig. 4D,E). (F) Same as C, but now for mean power difference based on the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window 141–151 ms, Fig. 4D,E). In A, B, D, and E conditions are ordered according to the hypothesized decision model (Fig. 4B). Insets in A, B, D, and E show results of the linear regression analyses (black lines) based on all 6 conditions (i.e., 100ms-2, low α-2, high α-2, low α-1, high α-1, 0ms-1). Note that the additional regression analyses excluding the 100ms-2 and 0ms-1 conditions similarly demonstrate a significant negative linear correlation (P < 0.05; regression lines not shown) for the ordered intermediate ERFs (i.e., low α-2, high α-2, low α-1, high α-1) for all 4 time windows (145–171, 140–150, 139–172, 141–151 ms). Insets in C and F show results of the linear regression analyses (black lines).

Figure 5.

Averaged amplitude values and confidence ratings of poststimulus ERFs. (A) Amplitude values (based on all significant sensors, as shown in Fig. 2B) averaged over the time window showing a significant difference between poststimulus ERF amplitudes of 0ms-1 and 100ms-2 (145–171 ms, see Fig. 4A,C). (B) Amplitude values (based on all significant sensors, as shown in Fig. 2B) averaged over the time window preceding the point of maximal amplitude difference (150 ms) between poststimulus ERF amplitudes of 0ms-1 and 100ms-2 (140–150 ms, see Fig. 4A,C). (C) Average confidence ratings per condition in relation to mean power difference to the respective decision bound (based on all significant sensors, as shown in Fig. 2B) for the time window 140 to 150 ms (see Fig. 4A,C). (D) Same as A, but now for amplitude values based on the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 139–172 ms, Fig. 4D,E). (E) Same as B, but now for amplitude values based on the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window: 141–151 ms, see Fig. 4D,E). (F) Same as C, but now for mean power difference based on the anterior/somatosensory sensor-cluster (as shown in Fig. 2B; time window 141–151 ms, Fig. 4D,E). In A, B, D, and E conditions are ordered according to the hypothesized decision model (Fig. 4B). Insets in A, B, D, and E show results of the linear regression analyses (black lines) based on all 6 conditions (i.e., 100ms-2, low α-2, high α-2, low α-1, high α-1, 0ms-1). Note that the additional regression analyses excluding the 100ms-2 and 0ms-1 conditions similarly demonstrate a significant negative linear correlation (P < 0.05; regression lines not shown) for the ordered intermediate ERFs (i.e., low α-2, high α-2, low α-1, high α-1) for all 4 time windows (145–171, 140–150, 139–172, 141–151 ms). Insets in C and F show results of the linear regression analyses (black lines).

Finally, we calculated the average confidence ratings per subject for each condition and averaged the mean confidence levels per condition over all subjects. Since confidence levels should be a function of the distance of the decision variable to the respective decision bounds (i.e., low α-2 and high α-2 to 100ms-2; low α-1 and high α-1 to 0ms-1, see Fig. 4B), we calculated the mean power difference of each intermediate condition (i.e., low α-2, high α-2, low α-1, high α-1) from the respective decision bounds averaged over the time window showing a significant difference between the conditions 0ms-1 and 100ms-2 (i.e., 145–171 ms) and the time window preceding the point of maximum amplitude difference between the conditions 0ms-1 and 100ms-2 (i.e., 140–150 ms, Fig. 5C). We plotted the distance of the decision variables to the respective decision bounds and related it to the mean confidence levels per condition over all subjects. Subsequently, we fitted a linear regression to the data to determine a linear trend. Additionally, we performed this analysis with amplitude values calculated for the time windows based on the anterior/somatosensory sensor-cluster (i.e., 139–172 ms; 141–151 ms, Fig. 5F). Due to the fact that, for the parieto-occipital sensor-cluster, no time window showing a significant difference between the conditions 0ms-1 and 100ms-2 was found, we refrained from performing this analysis for the parieto-occipital sensor-cluster.

Results

Behavioral Results

Subjects performed a forced-choice temporal perceptual discrimination task (Fig. 1) and had to report how many electrical stimulations applied to their left index finger they perceived. For SOAs of 0 and 100 ms, subjects made only negligible errors (SOA 0 ms: 92.3 ± 1.8% [mean ± SD] correct reports; SOA of 100 ms: 93.8 ± 2.7% correct reports). For intermediate SOAs, subjects correctly perceived stimulation in approximately half of the trials (56.7 ± 3.2% correct reports). The response distribution of each condition did not significantly differ from a Gaussian distribution (P > 0.05). Statistical testing revealed highly significant differences regarding temporal perceptual discrimination rates between the intermediate condition and the 0 ms (t(15) = 10.086, P < 0.0001) as well as the 100 ms condition (t(15) = 11.811, P < 0.0001). Overall, the absolute influence of learning/fatigue is negligible. No significant linear trends indicating learning or fatigue effects could be determined for average temporal perceptual discrimination rate (r = 0.49, P > 0.05, Supplementary Fig. 2A) or confidence ratings (r = 0.55, P > 0.05, Supplementary Fig. 2B).

Time–Frequency Analysis

We studied the role of prestimulus alpha-band oscillations (8–12 Hz) on temporal perceptual discrimination. We focused on trials with intermediate SOA and compared alpha power in the prestimulus period (−900 to 0 ms) between correctly and incorrectly perceived trials. The exploratory time–frequency analysis confirmed a prominent alpha effect in the prestimulus period (Fig. 2A). Prestimulus alpha power was found to be statistically significantly decreased if subjects correctly perceived the stimulation as 2 stimuli compared with incorrectly perceived trials (P < 0.05, Fig. 2B). Significant differences were most evident for anterior/somatosensory and parieto-occipital sensors contralateral to stimulation site between −900 and −250 ms. Particularly, the topographical location of the effect shifted over time, with significant decreases in both contralateral anterior/somatosensory and parieto-occipital sensors at the beginning of the prestimulus epoch (−900 to −500 ms), compared with a decrease of power in more posterior sensors in the later prestimulus epoch (−400 to −250 ms). Note that, although both sensor-clusters show a significant alpha power decrease in the prestimulus epoch, the decision-related effects of alpha power visible in the poststimulus ERFs could only be demonstrated for the anterior/somatosensory sensor-cluster (see Relation between Decision Variable, Prestimulus Power, and Poststimulus ERFs and Fig. 4).

Source Reconstruction

To identify the underlying cortical sources of the aforementioned significant effect, we applied a beamforming approach. We identified one source mainly located in contralateral postcentral gyrus (Brodmann area 3, Fig. 2C). A second cluster was found in the contralateral middle occipital region, encompassing Brodmann areas 19, 21, and 39.

Correlation of Prestimulus Power, Perceptual Decisions, and Confidence Ratings

To determine more precisely the relation of prestimulus alpha power and subjective perception, we performed a correlation analysis. We computed single-trial power averaged over alpha frequencies and significant sensor-time points (time window: −900 to −250 ms, see Fig. 2B). Trials were sorted from low to high power and divided into 5 bins. Response probabilities for each bin were calculated as the percentage change in temporal perceptual discrimination rate from the mean, normalized per subject to the individual mean temporal perceptual discrimination rate over all bins.

We found a significant negative linear relationship between prestimulus alpha power averaged over all sensors showing a significant alpha power difference between correctly (perceived 2 stimuli) versus incorrectly (perceived 1 stimulus) perceived trials with intermediate SOA and subjects' perceptual decisions (r = −0.94, P < 0.05, Fig. 3A). In other words, probability of correctly perceiving the stimulation as 2 temporally separate stimuli was greater during trials with lower prestimulus alpha power. A one-way repeated-measures ANOVA revealed a significant main effect (P < 0.05). Post hoc t-tests revealed significant differences between bin1 versus bin4 (t(15) = 3.049, P < 0.01), bin1 versus bin5 (t(15) = 3.096, P < 0.01), bin2 versus bin4 (t(15) = 2.545, P < 0.05), and bin3 versus bin4 (t(15) = 2.142, P < 0.05). No significant quadratic relationship between prestimulus alpha power and subject's perceptual decisions was found (r = 0.94, P = 0.11). In addition, we performed the same analysis with power values averaged over the sensors of the spatiotemporally separated sensor-clusters (anterior/somatosensory vs. parieto-occipital). For the anterior/somatosensory sensor-cluster, both linear (r = −0.97, P < 0.01) and quadratic (r = 0.99, P < 0.05) fits for the relationship between prestimulus alpha power and subjects' perceptual decisions were significant. A one-way repeated-measures ANOVA revealed an effect on trend level (P = 0.1). Post hoc t-tests revealed significant differences between bin1 versus bin4 (t(15) = 2.74, P < 0.05), bin1 versus bin5 (t(15) = 2.14, P < 0.05), and bin2 versus bin4 (t(15) = 2.32, P < 0.05). Similarly for the parieto-occipital sensor-cluster, both linear (r = −0.96, P < 0.05) and quadratic (r = 0.99, P < 0.05) fits for the relationship between prestimulus alpha power and subjects' perceptual decisions were significant. No significant effect was found by a one-way repeated-measures ANOVA (P = 0.15). Post hoc t-tests revealed significant differences between bin1 versus bin4 (t(15) = 2.47, P < 0.05).

In a similar analysis, we investigated the correlation between prestimulus alpha power and subjects' level of confidence regarding their perceptual decisions. We found a significant negative linear relationship between prestimulus alpha power averaged over all sensors showing a significant alpha power difference between correctly (perceived 1 stimuli) versus incorrectly (perceived 1 stimulus) perceived trials with intermediate SOA and confidence ratings for correctly perceived trials (r = −0.88, P < 0.05, Fig. 3B) and a strong trend toward a significant positive linear correlation for incorrectly perceived trials (r = 0.81, P = 0.095). No significant quadratic relationship between prestimulus alpha power and subjects' confidence ratings was found (correct trials: r = 0.95, P = 0.1; incorrect trials: r = 0.94, P = 0.11). For the anterior/somatosensory sensor-cluster, a significant negative linear relationship between prestimulus alpha power and confidence ratings for correctly perceived trials (r = −0.92, P < 0.05) could be demonstrated, while no significant effect was found for incorrect trials (r = 0.6, P = 0.28). For quadratic relationships between prestimulus alpha power and subjects' confidence ratings, no significant fit was found (correct trials: r = 0.92, P = 0.14; incorrect trials: r = 0.8, P = 0.35). Finally, no significant linear or quadratic relationship between prestimulus alpha power and confidence ratings could be demonstrated for the parieto-occipital cluster, neither for correct (linear: r = −0.49, P = 0.41; quadratic: r = 0.59, P = 0.65) or incorrect trials (linear: r = 0.73, P = 0.17; quadratic: r = 0.94, P = 0.11).

Relation Between Decision Variable, Prestimulus Power, and Poststimulus ERFs

We investigated if poststimulus ERFs show characteristics of a decision variable and the influence of prestimulus alpha power on these variables. We analyzed poststimulus ERFs by applying a boundary-crossing decision-making model (Philiastides et al. 2006; O'Connell et al. 2012). To this end, we estimated decision bounds for the unambiguous perception of 1 and 2 stimuli by calculating poststimulus ERFs from all correct trials of the 0 and 100 ms SOA conditions, subsequently labeled 0ms-1 and 100ms-2. Statistical comparison revealed a significant difference between both ERF amplitudes between 145 and 171 ms (P < 0.05), indicating that the 2 signals significantly diverge during this time window (Fig. 4A). The spatial distribution of the stimuli-evoked ERFs for those time points showing a significant difference between the conditions 0ms-1 and 100ms-2 (i.e., 145–171 ms) revealed highly similar patterns of activity over conditions (Supplementary Fig. 3).

According to our hypothesis, these ERFs should reflect the lower and upper boundaries for decisions toward 1 and 2 stimuli, respectively. ERFs of trials with intermediate SOA should be located in between these boundaries and the distance toward the respective boundary should reflect the perceptual decision as well as the confidence in the decision (Fig. 4B). The results demonstrate that, despite physically identical stimulation, the ERFs of trials with intermediate SOA differ with respect to subjects' perception and prestimulus alpha power (Fig. 4C). In line with our hypothesis, we found a significant negative linear correlation for the ordered averaged ERFs (i.e., 100ms-2, low α-2, high α-2, low α-1, high α-1, 0ms-1), indicating a monotonic decrease in amplitude from the 100ms-2 condition to the 0ms-1 condition (r = −0.93, P < 0.01 for time window 145–171 ms, Fig. 5A; r = −0.96, P < 0.01 for time window 140–150 ms, Fig. 5B). An additional regression analysis which excluded the 100ms-2 and 0ms-1 conditions also revealed a significant negative linear correlation for the ordered averaged intermediate ERFs (i.e., low α-2, high α-2, low α-1, high α-1), indicating a monotonic decrease in amplitude from the low α-2 condition to the high α-1 condition (r = −0.97, P < 0.05 for time window 145–171 ms, see also captions Fig. 5; r = −0.98, P < 0.05 for time window 140–150 ms, see also captions Fig. 5). A one-way repeated-measures ANOVA revealed a strong trend toward a significant main effect (P = 0.065) for the analysis of the time window 145–171 ms. No significant effect was found for the time window 140–150 ms.

For the additional regression analysis performed on the anterior/somatosensory sensor-cluster (Fig. 4D,E), a significant negative linear correlation for the ordered averaged ERFs (i.e., 100ms-2, low α-2, high α-2, low α-1, high α-1, 0ms-1) could be demonstrated (r = −0.99, P < 0.001 for time window 139–172 ms, Fig. 5D; r = −0.97, P < 0.01 for time window 141–151 ms, Fig. 5E). The negative linear correlations remained significant under exclusion of the 100ms-2 and 0ms-1 conditions (r = −0.99, P < 0.001 for time window 139–172 ms, see also captions Fig. 5; r = −0.97, P < 0.05 for time window 141–151 ms, see also captions Fig. 5). A one-way repeated-measures ANOVA revealed a significant main effect for both time windows (P < 0.05 for time window 139–172 ms; P < 0.05 for time window 141–151 ms). Because no time window showing a significant difference between the conditions 0ms-1 and 100ms-2 was found for the parieto-occipital sensor-cluster (see Fig. 4F), we refrained from performing the regression analysis for the parieto-occipital sensor-cluster.

We further related the average confidence ratings per condition to the distance of the decision variables to the respective decision bounds. According to our hypothesis, the average confidence ratings per condition should increase with closer proximity of the decision variables to the respective decision bounds (see Fig. 4B). While for the time window from 145 to 171 ms, no significant linear relation between confidence ratings and distance of the decision variables to the respective decision bounds could be demonstrated (r = 0.43, P = 0.57), a strong trend toward a significant negative linear relation (r = −0.95, P = 0.053) was evident for the time window from 140 to 150 ms (Fig. 5C). For the critical time windows based on the anterior/somatosensory sensor-cluster, a significant negative linear relation could only be demonstrated for the time window from 141 to 151 ms (r = −0.96, P < 0.05, Fig. 5F). For the time window from 139 to 172 ms, no significant linear fit was found (r = −0.24, P = 0.84). Regarding the time windows before the point of maximum amplitude difference (140–150 ms for all significant sensors, 141–151 ms for the anterior/somatosensory sensor-cluster), in agreement with our hypothesis a closer distance to the reference conditions resulted in higher confidence ratings. Because no time window showing a significant difference between the conditions 0ms-1 and 100ms-2 was found for the parieto-occipital sensor-cluster, we refrained from performing this analysis for the parieto-occipital sensor-cluster.

Discussion

We investigated the influence of prestimulus alpha activity on the temporal perceptual discrimination of suprathreshold tactile stimuli, the confidence in perceptual decisions and the underlying neuronal decision variable. Subjects received 1 or 2 tactile stimuli with different SOAs. In a forced-choice task, subjects reported their perceptual decision and their confidence in this decision.

Subjects frequently misperceived stimulation as 1 stimulus for trials with intermediate SOA, indicating perceptual ambiguity despite physically identical stimulation. For these trials with intermediate SOA, correct perception of 2 separate stimuli was correlated with a decrease of alpha power (8–12 Hz) relative to incorrectly perceived trials. This effect was evident before onset of stimulation (−900 to −250 ms) mainly in the contralateral postcentral gyrus (presumably primary somatosensory cortex) and the contralateral middle occipital region. Additionally, prestimulus alpha power correlated with subjects' confidence ratings. For correctly perceived trials, high confidence ratings correlated with low prestimulus alpha power. Contrarily, for incorrectly perceived trials, high confidence ratings correlated with high prestimulus alpha power. Finally, poststimulus ERFs at ∼150 ms revealed characteristics of a decision variable. In summary, we found: 1) Poststimulus ERFs at ∼150 ms reflect perceptual decisions and subjects' confidence in their decisions rather than pure sensory evidence. 2) ERFs for all conditions were in line with an accumulation-to-bound model in which sensory evidence is accumulated in a decision variable (Gold and Shadlen 2007). In trials with ambiguous, intermediate SOA, ERFs of correctly perceived trials were closer to the putative categorical decision bound for perceiving 2 stimuli while incorrectly perceived trials were closer to the categorical decision bound for perceiving 1 stimulus. 3) Due to their perceptual ambiguity, stimuli with intermediate SOA provided only incomplete sensory evidence, resulting in incomplete evidence accumulation and hence ERFs did not cross the decision bound. 4) Incomplete evidence accumulation resulted in lower confidence as reflected in the ERFs. 5) The variability of ERFs, decisions, and confidence ratings is biased by fluctuations of prestimulus alpha power. 6) Finally, the above-mentioned results could be replicated only for the anterior/somatosensory sensor-cluster after separating the sensors of interest. Therefore, it appears that mainly the somatosensory cortex areas account for the decision-related components visible at ∼150 ms.

We estimated the poststimulus categorical decision boundaries by calculating significant differences between ERFs of the reference conditions 0ms-1 and 100ms-2. One might argue that these conditions differ not only by subjects' decisions but also by sensory evidence (1 stimulus vs. 2 stimuli), and thus, our putative decision variable might reflect sensory input rather than decisional processes. However, we demonstrate that ERFs around ∼150 ms for trials with intermediate SOA, that is, with constant stimulation, correlate with perceptual decisions rather than sensory input.

Several studies have reported an inverted U-shaped relationship between prestimulus alpha power and perceptual performance, with intermediate alpha levels resulting in best performance levels (Linkenkaer-Hansen et al. 2004; Zhang and Ding 2009; Lange et al. 2012). On the contrary, other studies emphasize a linear relationship, with lower power levels being related to better performance (Thut et al. 2006; Hanslmayr et al. 2007; Schubert et al. 2008; van Dijk et al. 2008; Mathewson et al. 2009; Jones et al. 2010). In the present study, linear as well as quadratic fits were applied to the data. For most analyses, both linear and quadratic fits were significant for the correlation of prestimulus alpha power and perceptual decisions for the anterior/somatosensory and the parieto-occipital sensor-cluster. This demonstration of both linear and quadratic dependencies hinders a final conclusion on this matter. It remains to be seen if future studies can clarify the relevant factors in terms of neuroanatomical region or experimental conditions favoring one dependency over the other.

Notably, the majority of previous studies used near-threshold stimuli and relied on conditions where stimuli are either perceived or not perceived. Thus, subjects had to report whether or not stimulation is perceived, irrespective of its content. Here, we contrasted 2 different perceptual qualities with suprathreshold intensities, since subjects had to report whether they perceived 1 stimulus or 2 stimuli. Our paradigm therefore focuses on temporal discrimination and employs temporal ambiguity, with identical suprathreshold stimulation resulting in varying perceptual decisions. Hence, the present study provides critical extensions to the aforementioned studies.

Our results are in line with several studies reporting a correlation of prestimulus alpha power and detection or discrimination of near-threshold stimuli (e.g., Linkenkaer-Hansen et al. 2004; Zhang and Ding 2009; Jones et al. 2010). We critically extend these studies by demonstrating that alpha power influences also the temporal resolution of perception. Although formerly interpreted as correlate of cortical idling (Pfurtscheller et al. 1996), alpha activity has recently been suggested to gate neuronal processing by functional inhibition of task-irrelevant areas (Jensen and Mazaheri 2010; Jensen et al. 2012) and/or by modulating cortical excitability (Thut et al. 2006; Romei, Brodbeck et al. 2008; Romei, Rihs et al. 2008; Lange et al. 2013), resulting in more efficient neuronal stimulus processing in task-related neuronal groups. By using 2 clearly suprathreshold stimuli, we demonstrate that prestimulus alpha power extends the role of a simple binary switch between inhibition and processing. Rather, it modulates the quantity (1 stimulus or 2 stimuli, e.g., Lange et al. 2013; Keil et al. 2014) and the subjective quality (i.e., confidence) of perception continuously. This continuous modulation is reflected in confidence ratings, providing a more fine-grained scale of the decision process.

Prestimulus alpha power can be modulated by attention or expectation (Foxe et al. 1998; Worden et al. 2000; Jones et al. 2010; Anderson and Ding 2011; Haegens et al. 2012). In line with these results, recent studies demonstrated that prestimulus alpha power is predictive of perceptual performance in attention-based tasks (Kelly et al. 2009; O'Connell et al. 2009). While we did not explicitly modulate attention in our study, we suggest that spontaneous fluctuations of attention or arousal modulate prestimulus alpha power and thus influence perception and confidence. Further, it seems that such fluctuations are distinguishable from general training effects, since we did not find significant learning/fatigue trends for either perception or confidence.

We found alpha power to differ significantly in the prestimulus period in the contralateral postcentral gyrus and contralateral middle occipital region. Differential alpha-band activity in the postcentral gyrus (presumably primary somatosensory areas) has been found for other tactile tasks (e.g., Zhang and Ding 2009; Jones et al. 2010; Lange et al. 2012). Here, we extend the role of the postcentral gyrus to temporal perceptual discrimination of 2 subsequently presented stimuli. Since we applied only tactile stimuli and a tactile decision task, the significant alpha-band effect in visual areas might seem surprising. However, our results are in line with findings from a tactile spatial attention task, showing that in the absence of visual stimulation, attention to tactile stimulation resulted in suppression of alpha-band power in occipital areas (Bauer et al. 2006). Similarly, a recent study indicates that task-relevant spatial attention in one sensory domain affects oscillatory activity in other domains (Bauer et al. 2012). In accordance to these findings, a recent study demonstrated that parieto-occipital activation in the alpha band is linked to spatial attention across modalities (Banerjee et al. 2011).

In line with these results, the power differences in the contralateral middle occipital region can also be interpreted as correlate of global attention, thus not restricted to the somatosensory domain. This is supported by classical findings which localize the central generator of alpha rhythms in parieto-occipital areas (e.g., Salmelin and Hari 1994; Manshanden et al. 2002), independent of task requirements. The explanation is further strengthened by our findings that the decision-related ERF components could only be found for the anterior/somatosensory sensor-cluster, but not in the parieto-occipital cluster. This indicates that the parieto-occipital sensor-cluster, although showing significant power differences between perceptual conditions, is not central for decision-related processes. The influence of prestimulus alpha on decision variables is also in line with a recent EEG study (Lou et al. 2014). In this study, the influence of prestimulus activity is seen as top-down attention-based modulation, indicating that the sensory evidence is comprised of stimulus information and attentional state.

We demonstrate that prestimulus alpha power does not only correlate with perceptual decisions, but also with the subjective quality of such decisions. If alpha power was low, subjects were more confident with their decisions, but notably only for correctly perceived stimuli. Contrarily, if stimulation was perceived incorrectly, low alpha power correlated with low confidence. This seemingly contradictory result can be explained by a decision model. It has been proposed that sensory evidence is accumulated over time in a decision variable until a decision bound is reached (e.g., Shadlen and Newsome 2001). Here, we used such a decision-to-bound model to examine poststimulus decision variables. We hypothesized that due to the ambiguity of sensory evidence the decision variable does not cross a decision bound. Further, fluctuations of prestimulus alpha power should influence the decision variable and the confidence in perceptual decisions, if sensory evidence was insufficient to reach a decision bound. We identified this proposed pattern of a decision variable in poststimulus ERFs at ∼150 ms. Despite identical stimulation, poststimulus ERFs of trials with intermediate SOA differed according to the decisional outcome. While neither condition reached the categorical decision bound, the distance of the decision variable to the respective decision bounds determined the decisional outcome.

We identified perceptual decision-related components in the somatosensory domain, that is, differences in ERF amplitudes for conditions with physically similar stimulation parameters that discriminated between perceptual reports, as early as ∼150 ms. Other recent studies addressing perceptual decision making in the visual domain report decision-related neural activity at later time points (∼300 ms) and relate earlier components to low-level stimulus processing mechanisms (Philiastides et al. 2006; Lou et al. 2014). Such stimulus processing mechanisms can hardly fully explain our results, since our stimulation parameters remained constant for trials with intermediate SOA. An early decision-related component is further supported by studies where early components around ∼75–80 ms were shown to discriminate between high-level properties such as semantic category and components around ∼150 ms discriminate between target and nontarget conditions (and hence task-specific decision-related demands), independent of visual category (VanRullen and Thorpe 2001). In line with these results, the present components around ∼150 ms can be interpreted as a correlate of the subjects' perceptual recognition and subsequent decision, not merely as stimulus-related bottom-up processing. However, it is important to keep in mind that somatosensory processing presumably does not end after the aforementioned component, but it appears that at this time point sufficient information for a perceptual decision is accumulated.

Kiani and Shadlen (2009) recorded neuronal activity in monkey lateral intraparietal cortex during a decision-making task. If the monkey chose to opt out, that is, at low confidence levels, neural activity was at an intermediate level between decision bounds. We used a more detailed confidence rating and found that subjects' confidence correlated with the distance to a decision bound. This suggests that categorical decision making and confidence estimation can be a simple and fast inherent property of the same process (e.g., Kepecs et al. 2008; Kiani and Shadlen 2009), rather than a serial process requiring additional steps or higher (meta) cognitive functions (e.g., Grinband et al. 2006; Yeung and Summerfield 2012).

Additionally, we found poststimulus ERFs to interact with prestimulus alpha levels. Low prestimulus alpha levels shifted the decision variable towards the decision bound for 2 perceived stimuli, independent of decisional outcome. For correctly perceived intermediate SOA trials, low prestimulus alpha power increased confidence, because the distance between the decision variable and the decision bound for 2 perceived stimuli decreased. Contrarily, for incorrectly perceived intermediate SOA trials, low prestimulus alpha power decreased confidence, because the distance between the decision variable and the decision boundary for 1 perceived stimulus increased. The influence of prestimulus alpha power on poststimulus ERFs is in line with recent studies (Jones et al. 2009, 2010; Anderson and Ding 2011; Lange et al. 2012). The influence of prestimulus activity on decisions and the underlying decision variable is also in line with a recent study demonstrating that prestimulus firing rates bias decisions (Carnevale et al. 2012). While this study considers prestimulus activity as noise fluctuations, we argue that prestimulus alpha power is a functionally relevant marker of cortical excitability that can fluctuate over time or that can be endogenously or exogenously modulated by, for example, attention, arousal, or expectation (e.g., Foxe et al. 1998; Worden et al. 2000; Thut et al. 2006; Jones et al. 2010; Anderson and Ding 2011; de Lange et al. 2011).

In line with a recent study (de Lange et al. 2013), we suggest that prestimulus alpha power biases the starting point of the decision variable. Thus, the decision variable is the combination of the internal brain state (prestimulus activity) and the sensory evidence provided by the stimulus. If sensory evidence is weak or ambiguous, prestimulus activity can effectively bias decisions and confidence ratings by shifting the decision variable closer to either decision bound. The fact that prestimulus activity influences the decisional process implies that the decision-making process starts before stimulus presentation (Carnevale et al. 2012; de Lange et al. 2013). Such prestimulus fluctuations can also explain why decisions, confidence ratings, or response times can vary despite physically identical stimulation.

In conclusion, our results demonstrate that the brain state, characterized by alpha power, substantially modulates temporal perceptual discrimination of tactile stimuli despite identical physical stimulation, as well as confidence in perceptual decisions. Moreover, these fluctuations in prestimulus alpha power are visible in poststimulus ERFs mainly determined by somatosensory areas, reflecting the physiological correlate of evidence accumulation in a decision variable for perceptual decisions based on insufficient and suboptimal evidence. We conclude that alpha-band activity continuously modulates the quality of processing underlying perceptual decisions, resulting in differences in temporal perceptual discrimination.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

J.L. was supported by the German Research Foundation (LA 2400/4-1).

Notes

We thank Erika Rädisch for help with the MRI recordings. Conflict of Interest: None declared.

References

Anderson
KL
Ding
M
.
2011
.
Attentional modulation of the somatosensory mu rhythm
.
Neuroscience
 .
180
:
165
180
.
Banerjee
S
Snyder
AC
Molholm
S
Foxe
JJ
.
2011
.
Oscillatory alpha-band mechanisms and the deployment of spatial attention to anticipated auditory and visual target locations: supramodal or sensory-specific control mechanisms?
J Neurosci
 .
31
:
9923
9932
.
Başar
E
Başar-Eroglu
C
Rosen
B
Schütt
A
.
1984
.
A new approach to endogenous event-related potentials in man: relation between EEG and P300-Wave
.
Int J Neurosci
 .
24
:
1
21
.
Bauer
M
Kennett
S
Driver
J
.
2012
.
Attentional selection of location and modality in vision and touch modulates low-frequency activity in associated sensory cortices
.
J Neurophysiol
 .
107
:
2342
2351
.
Bauer
M
Oostenveld
R
Peeters
M
Fries
P
.
2006
.
Tactile spatial attention enhances gamma-band activity in somatosensory cortex and reduces low-frequency activity in parieto-occipital areas
.
J Neurosci
 .
26
:
490
501
.
Brandt
ME
Jansen
BH
.
1991
.
The relationship between prestimulus alpha amplitude and visual evoked potential amplitude
.
Int J Neurosci
 .
61
:
261
268
.
Britten
KH
Newsome
WT
Shadlen
MN
Celebrini
S
Movshon
JA
.
1996
.
A relationship between behavioral choice and the visual responses of neurons in macaque MT
.
Vis Neurosci
 .
13
:
87
100
.
Carnevale
F
de Lafuente
V
Romo
R
Parga
N
.
2012
.
Internal signal correlates neural populations and biases perceptual decision reports
.
Proc Nat Acad Sci USA
 .
109
:
18938
18943
.
de Lange
FP
Rahnev
DA
Donner
TH
Lau
H
.
2013
.
Prestimulus oscillatory activity over motor cortex reflects perceptual expectations
.
J Neurosci
 .
33
:
1400
1410
.
de Lange
FP
van Gaal
S
Lamme
VAF
Dehaene
S
.
2011
.
How awareness changes the relative weights of evidence during human decision-making
.
PLoS Biol
 .
9
:
e1001203
.
Foxe
JJ
Simpson
GV
Ahlfors
SP
.
1998
.
Parieto-occipital ∼10 Hz activity reflects anticipatory state of visual attention mechanisms
.
Neuroreport
 .
9
:
3929
3933
.
Gold
JI
Shadlen
MN
.
2007
.
The neural basis of decision making
.
Annu Rev Neurosci
 .
30
:
535
574
.
Grinband
J
Hirsch
J
Ferrera
VP
.
2006
.
A neural representation of categorization uncertainty in the human brain
.
Neuron
 .
49
:
757
763
.
Gross
J
Kujala
J
Hamalainen
M
Timmermann
L
Schnitzler
A
Salmelin
R
.
2001
.
Dynamic imaging of coherent sources: studying neural interactions in the human brain
.
Proc Nat Acad Sci USA
 .
98
:
694
699
.
Haegens
S
Luther
L
Jensen
O
.
2012
.
Somatosensory anticipatory alpha activity increases to suppress distracting input
.
J Cogn Neurosci
 .
24
:
677
685
.
Hanslmayr
S
Aslan
A
Staudigl
T
Klimesch
W
Herrmann
CS
Bäuml
K
.
2007
.
Prestimulus oscillations predict visual perception performance between and within subjects
.
Neuroimage
 .
37
:
1465
1473
.
Hari
R
Kaukoranta
E
.
1985
.
Neuromagnetic studies of somatosensory system: principles and examples
.
Prog Neurobiol
 .
24
:
233
256
.
Hoogenboom
N
Schoffelen
J
Oostenveld
R
Fries
P
.
2010
.
Visually induced gamma-band activity predicts speed of change detection in humans
.
Neuroimage
 .
51
:
1162
1167
.
Jensen
O
Bonnefond
M
VanRullen
R
.
2012
.
An oscillatory mechanism for prioritizing salient unattended stimuli
.
Trends Cogn Sci
 .
16
:
200
206
.
Jensen
O
Mazaheri
A
.
2010
.
Shaping functional architecture by oscillatory alpha activity: gating by inhibition
.
Front Hum Neurosci
 .
4
:
186
.
Jones
SR
Pritchett
DL
Sikora
MA
Stufflebeam
SM
Hämäläinen
M
Moore
CI
.
2009
.
Quantitative analysis and biophysically realistic neural modeling of the MEG mu rhythm: rhythmogenesis and modulation of sensory-evoked responses
.
J Neurophysiol
 .
102
:
3554
3572
.
Jones
SR
Kerr
CE
Wan
Q
Pritchett
DL
Hamalainen
M
Moore
CI
.
2010
.
Cued spatial attention drives functionally relevant modulation of the mu rhythm in primary somatosensory cortex
.
J Neurosci
 .
30
:
13760
13765
.
Keil
J
Müller
N
Hartmann
T
Weisz
N
.
2014
.
Prestimulus beta power and phase synchrony influence the sound-induced flash illusion
.
Cereb Cortex
 .
24
:
1278
1288
.
Kelly
SP
Gomez-Ramirez
M
Foxe
JJ
.
2009
.
The strength of anticipatory spatial biasing predicts target discrimination at attended locations: a high-density EEG study
.
Eur J Neurosci
 .
30
:
2224
2234
.
Kepecs
A
Uchida
N
Zariwala
HA
Mainen
ZF
.
2008
.
Neural correlates, computation and behavioural impact of decision confidence
.
Nature
 .
455
:
227
231
.
Kiani
R
Shadlen
MN
.
2009
.
Representation of confidence associated with a decision by neurons in the parietal cortex
.
Science
 .
324
:
759
764
.
Lange
J
Halacz
J
van Dijk
H
Kahlbrock
N
Schnitzler
A
.
2012
.
Fluctuations of prestimulus oscillatory power predict subjective perception of tactile simultaneity
.
Cereb Cortex
 .
22
:
2564
2574
.
Lange
J
Oostenveld
R
Fries
P
.
2011
.
Perception of the touch-induced visual double-flash illusion correlates with changes of rhythmic neuronal activity in human visual and somatosensory areas
.
Neuroimage
 .
54
:
1395
1405
.
Lange
J
Oostenveld
R
Fries
P
.
2013
.
Reduced occipital alpha power indexes enhanced excitability rather than improved visual perception
.
J Neurosci
 .
33
:
3212
3220
.
Linkenkaer-Hansen
K
Nikulin
VV
Palva
S
Ilmoniemi
RJ
Palva
JM
.
2004
.
Prestimulus oscillations enhance psychophysical performance in humans
.
J Neurosci
 .
24
:
10186
10190
.
Litvak
V
Mattout
J
Kiebel
S
Phillips
C
Henson
R
Kilner
J
Barnes
G
Oostenveld
R
Daunizeau
J
Flandin
G
et al
2011
.
EEG and MEG data analysis in SPM8
.
Comput Intell Neurosci
 .
2011
:
852961
.
Lou
B
Li
Y
Philiastides
MG
Sajda
P
.
2014
.
Prestimulus alpha power predicts fidelity of sensory encoding in perceptual decision making
.
Neuroimage
 .
87
:
242
251
.
Manshanden
I
de Munck
JC
Simon
NR
Lopes da Silva
FH
.
2002
.
Source localization of MEG sleep spindles and the relation to sources of alpha band rhythms
.
Clin Neurophysiol
 .
113
:
1937
1947
.
Maris
E
Oostenveld
R
.
2007
.
Nonparametric statistical testing of EEG- and MEG-data
.
J Neurosci Methods
 .
164
:
177
190
.
Mathewson
KE
Gratton
G
Fabiani
M
Beck
DM
Ro
T
.
2009
.
To see or not to see: prestimulus α phase predicts visual awareness
.
J Neurosci
 .
29
:
2725
2732
.
Mazaheri
A
Jensen
O
.
2008
.
Asymmetric amplitude modulations of brain oscillations generate slow evoked responses
.
J Neurosci
 .
28
:
7781
7787
.
Mazaheri
A
Nieuwenhuis
ILC
van Dijk
H
Jensen
O
.
2009
.
Prestimulus alpha and mu activity predicts failure to inhibit motor responses
.
Hum Brain Mapp
 .
30
:
1791
1800
.
Nolte
G
.
2003
.
The magnetic lead field theorem in the quasi-static approximation and its use for magnetoencephalography forward calculation in realistic volume conductors
.
Phys Med Biol
 .
48
:
3637
3652
.
O'Connell
RG
Dockree
PM
Kelly
SP
.
2012
.
A supramodal accumulation-to-bound signal that determines perceptual decisions in humans
.
Nat Neurosci
 .
15
:
1729
1735
.
O'Connell
RG
Dockree
PM
Robertson
IH
Bellgrove
MA
Foxe
JJ
Kelly
SP
.
2009
.
Uncovering the neural signature of lapsing attention: electrophysiological signals predict errors up to 20 s before they occur
.
J Neurosci
 .
29
:
8604
8611
.
Oostenveld
R
Fries
P
Maris
E
Schoffelen
J
.
2011
.
FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data
.
Comput Intell Neurosci
 .
2011
:
156869
.
Pfurtscheller
G
Stancák
A
Jr
Neuper
C
.
1996
.
Event-related synchronization (ERS) in the alpha band—an electrophysiological correlate of cortical idling: a review
.
Int J Psychophysiol
 .
24
:
39
46
.
Philiastides
MG
Ratcliff
R
Sajda
P
.
2006
.
Neural representation of task difficulty and decision making during perceptual categorization: a timing diagram
.
J Neurosci
 .
26
:
8965
8975
.
Philiastides
MG
Sajda
P
.
2006
.
Temporal characterization of the neural correlates of perceptual decision making in the human brain
.
Cereb Cortex
 .
16
:
509
518
.
Ratcliff
R
McKoon
G
.
2007
.
The diffusion decision model: theory and data for two-choice decision tasks
.
Neural Comput
 .
20
:
873
922
.
Romei
V
Brodbeck
V
Michel
C
Amedi
A
Pascual-Leone
A
Thut
G
.
2008
.
Spontaneous fluctuations in posterior α-band EEG activity reflect variability in excitability of human visual areas
.
Cereb Cortex
 .
18
:
2010
2018
.
Romei
V
Gross
J
Thut
G
.
2010
.
On the role of prestimulus alpha rhythms over occipito-parietal areas in visual input regulation: correlation or causation?
J Neurosci
 .
30
:
8692
8697
.
Romei
V
Rihs
T
Brodbeck
V
Thut
G
.
2008
.
Resting electroencephalogram alpha-power over posterior sites indexes baseline visual cortex excitability
.
Neuroreport
 .
19
:
203
208
.
Salmelin
R
Hari
R
.
1994
.
Characterization of spontaneous MEG rhythms in healthy adults
.
Electroencephalogr Clin Neurophysiol
 .
91
:
237
248
.
Schubert
R
Haufe
S
Blankenburg
F
Villringer
A
Curio
G
.
2008
.
Now you'll feel it, now you won't: EEG rhythms predict the effectiveness of perceptual masking
.
J Neurosci
 .
21
:
2407
2419
.
Shadlen
MN
Newsome
WT
.
2001
.
Neural basis of a perceptual decision in the parietal cortex (Area LIP) of the rhesus monkey
.
J Neurophysiol
 .
86
:
1916
1936
.
Thut
G
Nietzel
A
Brandt
SA
Pascual-Leone
A
.
2006
.
α-band electroencephalographic activity over occipital cortex indexes visuospatial attention bias and predicts visual target detection
.
J Cogn Neurosci
 .
26
:
9494
9502
.
van Dijk
H
Schoffelen
J
Oostenveld
R
Jensen
O
.
2008
.
Prestimulus oscillatory activity in the alpha band predicts visual discrimination ability
.
J Neurosci
 .
28
:
1816
1823
.
VanRullen
R
Thorpe
SJ
.
2001
.
The time course of visual processing: from early perception to decision-making
.
J Cogn Neurosci
 .
13
:
454
461
.
Worden
MS
Foxe
JJ
Wang
N
Simpson
GV
.
2000
.
Anticipatory biasing of visuospatial attention indexed by retinotopically specific α-band electroencephalography increases over occipital cortex
.
J Neurosci
 .
20
:
RC63
.
Wyart
V
Tallon-Baudry
C
.
2009
.
How ongoing fluctuations in human visual cortex predict perceptual awareness: baseline shift versus decision bias
.
J Neurosci
 .
29
:
8715
8725
.
Yeung
N
Summerfield
C
.
2012
.
Metacognition in human decision-making: confidence and error monitoring
.
Philos Trans R Soc Lond B Biol Sci
 .
367
:
1310
1321
.
Zhang
Y
Ding
M
.
2009
.
Detection of a weak somatosensory stimulus: role of the prestimulus mu rhythm and its top–down modulation
.
J Cogn Neurosci
 .
22
:
307
322
.