Beta oscillations define discrete perceptual cycles in the somatosensory domain

Thomas J. Baumgarten; Alfons Schnitzler; Joachim Lange

doi:10.1073/pnas.1501438112

Edited by Ranulfo Romo, Universidad Nacional Autonóma de México, Mexico City, D.F., Mexico, and approved July 31, 2015 (received for review January 22, 2015)

Significance

Our sensory system constantly receives multiple inputs, which are usually perceived as a seamless stream. Thus, perception is commonly regarded as a continuous process. Alternatively, a few phenomena and recent studies suggest that perception might work in a discrete and periodic sampling mode. In a human magnetoencephalography study, we challenged the common view of continuous perception. We demonstrate that neuronal oscillations in the alpha band and low beta band determine discrete perceptual sampling windows in primary somatosensory cortex. The current results elucidate how ongoing neuronal oscillations shape discrete perceptual cycles, which constitute the basis for a discontinuous and periodic nature of somatosensory perception.

Abstract

Whether seeing a movie, listening to a song, or feeling a breeze on the skin, we coherently experience these stimuli as continuous, seamless percepts. However, there are rare perceptual phenomena that argue against continuous perception but, instead, suggest discrete processing of sensory input. Empirical evidence supporting such a discrete mechanism, however, remains scarce and comes entirely from the visual domain. Here, we demonstrate compelling evidence for discrete perceptual sampling in the somatosensory domain. Using magnetoencephalography (MEG) and a tactile temporal discrimination task in humans, we find that oscillatory alpha- and low beta-band (8–20 Hz) cycles in primary somatosensory cortex represent neurophysiological correlates of discrete perceptual cycles. Our results agree with several theoretical concepts of discrete perceptual sampling and empirical evidence of perceptual cycles in the visual domain. Critically, these results show that discrete perceptual cycles are not domain-specific, and thus restricted to the visual domain, but extend to the somatosensory domain.

The sensory system continuously receives and processes numerous stimuli. Subjective experience implies that conscious perception, and thus cortical processing, of this stimulation is also continuous. This view of continuous cortical processing, however, has been challenged by several studies proposing that the brain operates discontinuously within a framework of discretely sampled “perceptual cycles” (1 ⇓⇓–4). This process of perceptual cycles is thought to create a temporally defined window, with discrete stimuli falling inside this window being consciously perceived as a single event (4). Discrete sampling of sensory information allows for the possibility of transforming perceptual input into temporal code (5, 6), metabolic efficiency (7), and the efficient organization of information, thereby preventing information overload (6). Over the past decades, however, there has been an ongoing discussion about the nature of perception. Several studies have argued against the theory of discontinuous perceptual cycles (8, 9). In recent years, the hypothesis of a discontinuous cyclic perception received new support by electroencephalography (EEG) and magnetoencephalography (MEG) studies investigating neuronal oscillations. This novel support is attributable to the theory that serial perceptual sampling is thought to depend on the temporal relationship between external stimuli and some ongoing internal neurophysiological process (4) providing a temporal reference frame (5), with neuronal oscillations representing a probable candidate measure for this underlying process.

There is growing evidence that oscillatory power and phase influence cortical processing (10, 11) and perception (3, 12 ⇓–14). Most of these studies investigated perception of single near-threshold stimuli. Although these studies demonstrate that neuronal oscillations play a critical role in defining neuronal states, which, in turn, influence perception and neuronal processing (5, 15, 16), these studies do not provide direct evidence for or against the theory of perceptual cycles. Recent studies, however, argued that parietooccipital alpha oscillations (∼8–12 Hz) might define cycles of perception (6, 15, 17 ⇓–19). However, they only provide evidence for discrete perceptual sampling in the visual domain. To claim that discrete perception is not domain-specific, it is critical to demonstrate discrete and cyclic perception also for other sensory modalities and whether different modalities work via the same mechanism (e.g., whether alpha cycles generally define critical perceptual cycles for all modalities). Because sensory modalities work on different time scales, there is some indication that the mechanisms might differ.

We investigated whether discrete perceptual cycles exist in the somatosensory domain. Contrary to most studies in the visual domain, we used discrete rather than continuous stimuli, which differed only in perceptual impact, yet not in physical stimulation parameters. By this method, we could study whether two successively presented stimuli are perceived as either one single or two separate sensory events, depending on their temporal relationship to discrete perceptual cycles defined by the ongoing neuronal oscillatory phase. This setup allowed us to test the theory of discrete perceptual sampling critically in the somatosensory domain, and thus whether cycles of perception represent a mechanism of conscious perception that exists beyond the visual domain.

Results

Behavioral Results.

Subjects received one or two electrical pulses separated by a specific stimulus onset asynchrony (SOA; nomenclature is provided in Materials and Methods) and had to perform a forced-choice temporal perceptual discrimination task (Fig. 1), wherein they had to report whether they perceived one or two stimuli. Subjects made negligible errors for the conditions 0 ms and 100 ms [SOA 0 ms: 97.7 ± 0.4% (mean ± SEM) reports of correctly perceiving one stimulus, SOA 100 ms: 94.6 ± 2.3% reports of correctly perceiving two stimuli]. Individually determined, intermediate SOAs yielded correct perception of two stimuli in ∼50% of the trials (58.0 ± 3.1% reports). For the condition intermediate − 10 ms, subjects perceived two stimuli in 25.6 ± 4.7% reports, and for intermediate + 10 ms, subjects perceived two stimuli in 79.1 ± 4.7% reports. A one-way repeated measures ANOVA comparing average hit rates between conditions demonstrated a highly significant difference [F(4,60) = 141.25, P < 0.001]. Post hoc t tests revealed significant differences between the condition 0 ms vs. intermediate − 10 ms [t(15) = 5.14, P < 0.01], 0 ms vs. intermediate [t(15) = 18.34, P < 0.001], 0 ms vs. intermediate + 10 ms [t(15) = 15.79, P < 0.001], 0 ms vs. 100 ms [t(15) = −37.15, P < 0.001], intermediate − 10 ms vs. intermediate [t(15) = 7.36, P < 0.001], intermediate − 10 ms vs. intermediate + 10 ms [t(15) = −7.15, P < 0.001], intermediate − 10 ms vs. 100 ms [t(15) = −13.1, P < 0.001], intermediate vs. intermediate + 10 ms [t(15) = −5.35, P < 0.001], intermediate vs. 100 ms [t(15) = −12.54, P < 0.001], and intermediate + 10 ms vs. 100 ms [t(15) = −3.79, P < 0.01].

Fig. 1.

Experimental paradigm. The sequence of events begins with presentation of a central fixation dot (500 ms). Luminance decrease signals start at the prestimulus epoch (900–1,100 ms), after which tactile stimulation is applied to the left index finger with varying SOAs (0 ms, intermediate − 10 ms, intermediate, intermediate + 10 ms, 100 ms). Stimulation is followed by a jittered poststimulus period (500–1,200 ms), after which written instructions signal subjects to report their respective perception of the stimulation by pressing a button.

Phase Angle Contrast.

To study the influence of oscillatory phase angles on perception, we sorted trials with intermediate SOA according to perceptual response (perceived one or two stimuli), resulting in two perceptual conditions (intermediate1 vs. intermediate2). We computed phase angles for each condition in source space by means of a virtual channel in the primary somatosensory cortex (S1) (Fig. 2A) and contrasted the phases of intermediate2 with intermediate1 (Fig. S1). The analysis revealed a significant positive cluster (P < 0.05; Fig. 2B) in the prestimulus epoch (−0.53 to −0.09 s) for frequencies in the alpha band and lower beta band (8–20 Hz). Notably, the effect was more prominent and temporally extended in the beta band (14–20 Hz, −0.53 to −0.09 s) compared with the alpha band (8–12 Hz, −0.39 to −0.24 s). That is, the phase difference between perceptual conditions differed significantly more in this time-frequency range compared with randomly distributed phases. For frequencies in the lower beta band, phase difference fluctuated around maximum (i.e., π) in the prestimulus period (Fig. 2C). To exclude any bias due to power differences, we analyzed power differences between perceptual conditions for those time-frequency elements exhibiting significant phase differences (analysis parameters are provided in ref. 14). The results did not reveal any significant power differences (P > 0.05, uncorrected). Regarding phase angle differences, we found an additional significant negative cluster (P < 0.05; Fig. 2B) between 2 and 28 Hz and between −0.1 and 0.24 s. Here, phase differences were significantly smaller compared with randomly distributed phases. This effect presumably resembles the phase resetting after stimulus presentation (18, 20).

Fig. 2.

Virtual sensor location and phase angle differences. (A) Virtual sensor location based on the voxel of maximum activity of the contrast M50 vs. prestimulus baseline. The voxel is highlighted on a slice plot of the Montreal Neurological Institute (MNI) template brain (MNI coordinates: 50 −10 50). (B) Time-frequency plot showing the results of the statistical analysis of phase angle differences between intermediate2 vs. intermediate1. Significant clusters (P < 0.05, corrected) are highlighted. Red colors indicate higher phase differences compared with randomly distributed phases. t = 0 indicates onset of the first stimulus. (C) Phase angle difference (black solid line) between intermediate2 (Intermed2) and intermediate1 (Intermed1) for an exemplary 14-Hz band. The upper dashed line indicates the maximum phase angle difference (π). (Insets) Phase angles for intermediate2 (blue lines) and intermediate1 (red lines) for exemplary time points (Left, t = −420 ms; Right, t = −50 ms) (D) Relationship between the momentary phase (Materials and Methods) and the normalized perceptual response rate. The probability of perceiving two stimuli significantly depends on the phase angle and differs maximally between opposite phase angles (ANOVA, P < 0.001).

Fig. S1.

Time-frequency plot showing the phase angle differences between intermediate2 vs. intermediate1. Red colors indicate a higher angular difference. Black outlines represent the extent of the significant group-level phase angle difference effect. t = 0 indicates the onset of the first stimulus.

Phase Angles and Perception.

To analyze the extent by which perception was influenced by phase, we computed for each subject the momentary phase for each single trial for both perceptual conditions at the time point showing the largest statistical phase difference effect (Materials and Methods). Trials were placed in one of six different phase bins and aligned for each subject so that the highest probability for perceiving two stimuli corresponded to a zero phase angle. For each subject, we calculated the normalized perceptual response rate per bin, and we then averaged normalized response rates across subjects (Fig. 2D). Although this analysis resembles a post hoc test (because it is based on the time-frequency points of maximal phase difference determined in the previous analysis), it quantifies the magnitude by which phase influences perception, as well as the grade by which performance varies over different phase bins. A one-way repeated measures ANOVA comparing normalized perceptual response rates between bins demonstrated a highly significant difference [F(4,60) = 6.53, P < 0.001]. Post hoc t tests revealed significant differences between bin 1 vs. bin 3 [t(5) = −4.17, P < 0.01], bin 1 vs. bin 5 [t(5) = −4.21, P < 0.01], bin 1 vs. bin 6 [t(5) = −4.13, P < 0.01], bin 2 vs. bin 3 [t(5) = −2.77, P < 0.05], and bin 2 vs. bin 5 [t(5) = −3.16, P < 0.01]. The results indicate a monotonic decrease of mean response rate from zero phase angle to π, with the response rates differing by 13% points between the lowest (−π, 38%) and highest (1/3 π, 51%) phase bins (with exclusion of the zero phase bin).

Beta-Band Cycles Determine Perceptual Cycles.

Fig. 3 illustrates a model derived from the analysis of phase angle contrasts and the theory of temporal framing (3, 19, 21). The model proposes that the temporal resolution of perception is defined by one cycle of a specific frequency. If presented within one cycle, the two stimuli are merged into one perceptual event and perceived as one single stimulus (Fig. 3A, white rectangles). If presented within two separate cycles, they will be perceived as two temporally separate perceptual events (Fig. 3A, black rectangles). Although the neural representation of the first stimulus can arrive at any point in the oscillatory cycle (21) for ongoing oscillations, the arrival of the second stimulus is determined by the SOA. For a cycle length twice as long as the respective SOA, a stimulus arriving in the first half of the cycle determines the arrival of the second stimulus in the same cycle (one perceived stimulus). Vice versa, a stimulus arriving in the second half of the cycle determines the arrival of the second stimulus in a subsequent cycle (two perceived stimuli). From the results of the phase angle contrast analysis, we derive that this critical frequency band lies in the alpha band and, particularly, the lower beta band between 8 and 20 Hz (Fig. 2B). Given these model preconditions, we can make two predictions. First, if the SOA between two stimuli equals half the length of the cycle of the critical frequency (e.g., 25 ms for a 20-Hz oscillation), mean phases for the perception of one stimulus (range: 0 to π for the example in Fig. 3A) and two stimuli (range: π to 2π) should differ maximally (∼π). More precisely, perception rates should critically depend on the phase at which the first stimulus arrives (Fig. 3 B–D). That is, if the stimulus arrives at a given phase φ, perception rates should differ significantly from φ + π. Second, if the critical frequency is known, we can predict behavioral response rates for different SOAs. The first prediction is confirmed by the analysis of phase angle contrast (Fig. 2 B and C). Based on these results, the post hoc phase binning analysis shows a monotonic decrease in perception over bins, and, thus, the dependence of perception rates on phase (Fig. 2D). The second prediction will be tested and presented below.

Fig. 3.

Model for perceptual cycles. (A) Red and blue lines illustrate two perceptual cycles. Two stimuli can occur within one (white rectangles, one stimulus perceived) or two (black rectangles, two separate stimuli perceived) perceptual cycles. (B) Same as in A, but for stimulus pairs with a longer SOA. The blue background illustrates the time frame in which the occurrence of the first stimulus results in one perceived stimulus (○), and the gray background illustrates the time frame in which the occurrence of the first stimulus results in two perceived stimuli (●). (C) Same as in B, but for stimuli with a shorter SOA. Note different lengths of blue and gray time frames. (D) Same as in B, but for examples of three different SOAs. Intermediate SOAs (rectangles) result in time frames for one (blue arrows) or two (gray arrows) perceived stimuli of approximately equal length. For longer SOAs (●), the time frame for two perceived stimuli (gray arrows) is bigger than for one perceived stimulus (blue arrows). For shorter SOAs (♦), the time frame for two perceived stimuli (gray arrows) is smaller than for one perceived stimulus (blue arrows).

Prediction of Perception.

Based on the model (Fig. 3), we predicted response rates for the different SOAs and computed linear regressions between predicted and behaviorally measured response rates. We computed predictions based on (i) group-level effect frequencies determined from MEG experimental data (8–20 Hz; Fig. 2B), (ii) based on single subject-level individual frequencies determined from MEG experimental data (Fig. S2 and Table S1), and (iii) based on frequencies determined from behavioral experimental data (i.e., the intermediate SOAs):

i) Based on group-level effect frequencies (8–20 Hz; Fig. 2B), the linear regression analysis for behavioral response rates and predicted response rates (Fig. 4) resulted in a highly significant correlation coefficient (r = 0.93, P < 0.01). The resulting slope estimate (0.83 ± 0.1) did not differ significantly from 1 [t(4) = −1.8, P > 0.05].
ii) Linear regression analysis of the individual behavioral and predicted individual response rates resulted in a significant correlation coefficient in all 16 subjects (r ranging from 0.69 to 0.96, P < 0.05). For 12 of 16 subjects, the resulting slope estimate did not differ significantly from 1 [t(4) ranging from −2.6 to 2.4, P > 0.05; Fig. S2 and Table S1]. We additionally predicted group-level response rates by averaging the individual response rates over subjects. The resulting predictions were virtually similar to the predictions based on the averaging over group-level effect frequencies (i) (Fig. 4). The resulting slope estimate (0.78 ± 0.1) did not differ significantly from 1 [t(4) = −2.23, P > 0.05].
iii) Predictions based on frequencies determined from behavioral experimental data yielded results highly similar to those results determined from MEG experimental data (details are provided in SI Results).

Fig. 4.

Model prediction of response rates. Proportion of “perceived two stimuli” reports for conditions with different SOAs for model predictions based on averaged individual response rates (gray bars), significant group-level frequencies (8–20 Hz, white bars), and behavioral data (black bars). Predicted responses based on frequencies were calculated per frequency bin and then averaged over all respective frequencies. Hit rates are presented as mean ± SEM.

Fig. S2.

Single-subject predicted and behavioral response rates. Proportions of perceived two stimuli reports for conditions with different SOAs for model predictions based on the individual critical frequency determined from MEG experimental data (white bars) and for individual behavioral data (black bars). Hit rates are presented as means. Individual critical frequencies are determined from MEG experimental data, the slopes specify for which the best fit between predicted and behavioral response rates can be achieved, and P values specify if the respective slope differs significantly from 1.

View this table:

Table S1.

Predicted individual response rates based on individual frequencies determined from MEG experimental data and individual behavioral response rates

SI Results

To compare the predictions of perception based on MEG experimental data (Results and Fig. 4) with a model based on the behavioral experimental data, we computed the critical frequency determined by the average intermediate SOA of 25.9 ms (22.4 Hz) and predicted response rates for this frequency. The linear regression analysis between predicted and behavioral response rates resulted in a highly significant correlation coefficient (r = 0.99, P < 0.01). The resulting slope (1.05 ± 0.04) did not differ significantly from 1 [t(4) = 1.29, P > 0.05]. Finally, we computed critical frequencies determined from behavioral experimental data on the single-subject level (i.e., based on individual intermediate SOAs and behavioral response rates) and averaged the resulting critical frequencies over subjects (23.9 ± 2.0 Hz). Linear regression analysis resulted in a highly significant correlation coefficient (r = 0.99, P < 0.01). The resulting slope estimate (1.08 ± 0.04) did not differ significantly from 1 [t(4) = 2.12, P > 0.05]. Thus, theoretically and experimentally determined frequencies yielded highly similar results: Both reveal significant correlation values and slopes not significantly different from 1.

Discussion

We investigated the neuronal mechanisms of varying conscious perception in the somatosensory domain. The results argue against a continuous perceptual mechanism and provide evidence that somatosensory perception operates in a discrete mode, with sensory input being sampled by discrete perceptual cycles in the alpha band and, in particular, the lower beta band (8–20 Hz).

Beta-Band Cycles Determine Discrete Perceptual Sampling.

We found that phase angles in S1 in the alpha band and lower beta band (8–20 Hz) before stimulus onset predicted whether subjects perceived two constant electrical stimuli with an SOA of ∼25 ms as one or two stimuli (Fig. 2B). Notably, this effect was most prominent in the lower beta band (14–18 Hz). We put forward a model proposing that somatosensory stimulation is discretely sampled and that the underlying perceptual cycles are determined by ongoing oscillatory alpha and beta cycles (Fig. 3). If multiple discrete stimuli fall within one perceptual cycle, the temporally fine-grained information is lost and the distinct stimuli are fused to a single percept, a phenomenon that has been labeled perceptual or temporal framing in the visual domain (3, 19, 21). The model was confirmed by two theoretical predictions. First, beta oscillations were found to be antiphasic (phase difference of π) for perception of one vs. two stimuli for intermediate SOAs (∼25 ms; Fig. 2 B–D). Based on these results, response rates were shown to depend on the specific phase at which the first stimulus arrives (Fig. 2D). Second, the model predicts behavioral performance on group (Fig. 4) and single-subject (Fig. S2) levels.

Based on behavioral response rates, the model predicted a theoretical critical sampling frequency of ∼23 Hz. The experimentally observed frequency range based on statistical analysis of phase angles revealed a significant effect between 8 and 20 Hz. Whereas the upper end of the experimental frequency range is close to the theoretically assumed frequency, the experimental frequency band also includes lower frequencies. A potential reason for this underestimation of the critical sampling frequency might be a decreased signal-to-noise ratio for higher frequencies. Noninvasive measurement (e.g., via EEG/MEG) of phase has been assumed to be especially susceptible to various interferences (e.g., delays in synaptic transmission) at higher frequencies (5). Likewise, phase differences in lower frequency bands could also resemble processes different than perceptual sampling (e.g., attentional processes) (22). This idea is in line with the different temporal distributions of phase angle differences for alpha- and beta-band frequencies. Finally, the presented model does not claim to cover all portions of the decision process determining the final response but, instead, focuses on early perceptual components. For example, the present data are derived from S1, thus not taking into account other cortical areas involved in the decisional process.

Discrete Perceptual Sampling Is Not a Domain-Specific Mechanism.

The theory of discrete perceptual cycles was introduced decades ago (1, 2). However, it has been controversially discussed (8, 9). Recently, the discussion on discrete perceptual cycles has gained new momentum by studies using EEG, which allows one to study potential neuronal mechanisms of discrete perceptual cycles noninvasively (5, 17, 22). Nonetheless, empirical evidence to support the theory of discrete perceptual cycles remains scarce and focuses mainly on the visual domain (3, 17), whereas evidence for discrete cycles in other domains is largely missing (19). The present study is thus, to our knowledge, the first to demonstrate the existence of perceptual cycles in the somatosensory domain, indicating that the cyclic characteristic of perception is not a domain-specific visual mechanism (19).

Modality-Specific Differences.

For the visual domain, EEG studies propose discrete cycles in perception and attentional updating defined by the alpha cycle (3, 17, 22). Our model agrees with these studies, albeit we propose perceptual cycles to be defined by alpha-band and, decisively, beta-band frequencies in the somatosensory domain. Although the significant group-level phase angle differences cover a rather broad band between 8 and 20 Hz, the major effect can be found in a narrower band between 14 and 18 Hz (Fig. 2B). Because subjects exhibit different individual intermediate SOAs, different individual frequencies for the discrete perceptual cycles are also to be expected (thereby blurring the group-level effect). In fact, the analyses based on individually determined frequencies confirmed that the individual narrow-band frequencies represent an appropriate predictor for individual response rates (Fig. S2). These domain-specific differences agree with a more prominent role of alpha oscillations in the visual domain for perception and neuronal processing (23, 24), whereas there is experimental evidence for a specific role of beta oscillations in the somatosensory domain (10, 13, 25 ⇓–27). The present findings are in line with studies investigating steady-state somatosensory evoked potentials (SSSEPs). These studies found that the largest SSSEP amplitudes can be achieved by a stimulation frequency of ∼18–26 Hz (i.e., in the beta band) (27 ⇓–29). Stimulation at this frequency would place every stimulus in a separate beta cycle, therefore enhancing SSSEPs and, consequently, facilitating perceptual detection (26). Finally, the proportion perceiving two stimuli differed by 13% between the lowest (−π) and the highest (1/3 π) phase bins (with exclusion of the zero phase bin). This difference agrees with ranges reported for visual stimuli (5, 15). Thus, both visual and somatosensory perception seems to be influenced by phase with a comparable magnitude.

What About Absolute Phase Angles?

Varela et al. (3) reported that the phase of occipital alpha oscillations determines whether subjects perceive two sequential visual stimuli as one or two stimuli. The respective phase for perceiving one vs. two stimuli was anticyclic (i.e., the phase difference was π). Although later studies failed to replicate this result (19, 30), our results support the finding by Varela et al. (3), because we find a phase difference of π between phases for perceiving one vs. two tactile stimuli. In contrast to Varela et al. (3), however, we do not claim that the specific phase (the peak or trough) is important for perception but, rather, whether two stimuli fall within a single cycle or separate cycles. The majority of studies investigating the influence of oscillatory phase on perception analyzed absolute phase angles within an oscillatory cycle at a specific moment, which are either favorable or unfavorable for subsequent perception (5, 11, 12, 15, 31). Thus, a potential concern might be that our results could be explained by favorable or unfavorable phases within one cycle. In such a framework, one stimulus might be presented at a favorable phase and the other stimulus might be presented at an unfavorable phase, thus leading to the erroneous perception of only one stimulus. The above-mentioned studies, however, used near-threshold stimuli. We presented stimuli with clearly suprathreshold intensities that are presumably perceived independent of the specific phase. Although a hypothesis proposing an influence of (un)favorable phases would predict that ∼50% of the stimuli with SOA 0 ms would be missed, subjects correctly perceived almost all stimuli. Similarly, such a framework would predict a higher percentage of trials with SOA 100 ms to be perceived as one stimulus than found in our behavioral data. Therefore, the present results cannot be explained by favorable or unfavorable phases within one oscillatory cycle.

Differentiating Effects of Phase and Power.

Recent studies demonstrated an influence of oscillatory power for perception of single (near-threshold) tactile stimuli, as well as for the temporal discrimination of two tactile stimuli (10, 25). The majority of these studies [including a previous study by our group on the dataset presented in this study (14)] found prestimulus power differences in the alpha band (8–12 Hz), whereas the present phase angles differed mostly in the lower beta band (14–18 Hz). Further, we found no significant power differences in those time-frequency elements showing significant phase angle differences between perceptual conditions. It is thus unlikely that the presented phase effect was biased by power differences. Indeed, there is experimental evidence for an influence of both oscillatory power (10) and phase (12, 31) for neuronal processing and perception, and recent studies could demonstrate that these measures act largely independently (5, 22). This differentiation is further supported by results showing that phase is able to transport more units of information per time than power changes (32) or spike counts (33), and represents a suitable candidate measure to encode fast-changing stimulus features (21).

Contradicting Subjective Experience.

There is accumulating evidence that our brain processes incoming stimulus information in a phasic mode (3, 17). However, personal experience does not intuitively match with a discrete sequencing approach but, rather, resembles a seamlessly updated percept. This divergence might explain why relatively few studies address this topic, although the concept of discrete perceptual sampling has been put forward at least since the middle of the 20th century (1, 2). It remains an open question how the brain transforms discretely sampled sensory information into a subjectively seamless impression. Although the mechanisms for such perceptual “smoothing” are unknown, there are, at least for the visual domain, several reports where the mechanisms fail to work (34). For example, in akinetopsia, subjects report perceiving a sequence of snapshots rather than a continuous motion (35, 36). Similarly, the ingestion of lysergic acid diethylamide often results in a perceptual disturbance wherein visual motion is perceived as a sequence of discrete stationary images (37, 38).

Conclusions

The present study demonstrates an influence of oscillatory phase on the temporal perception of two stimuli. We propose the existence of discrete perceptual cycles for the conscious perception of subsequently presented tactile stimuli. The perceptual cycles are determined particularly by frequencies in the beta band acting as the specific physiological correlate for perceptual cycles for the somatosensory modality. In combination with previous studies investigating similar paradigms in the visual domain (3, 30), the present results support the theory of temporal framing (1 ⇓–3, 19, 21) and indicate that perceptual cycles are no domain-specific visual phenomenon, albeit modality-specific frequencies that define perceptual cycles seem to be present.

Materials and Methods

Subjects.

The subjects, stimuli, paradigm, and MEG recording of the present study were previously reported in detail (14). Here, we present a comprehensive overview. Sixteen right-handed volunteers [seven males, age: 26.1 ± 4.7 y (mean ± SD)] participated in the study. Subjects provided written informed consent before the experiment in accordance with the Declaration of Helsinki and approved by the Ethical Committee of the Medical Faculty, Heinrich-Heine-University Düsseldorf.

Experimental Paradigm.

Details on the paradigm can be found in the study by Baumgarten et al. (14). A comprehensive overview is provided in Fig. 1 and SI Materials and Methods.

MEG Data Recording and Preprocessing.

Electromagnetic brain activity was continuously recorded using a 306-channel, whole-head MEG system (Neuromag Elekta Oy). Analysis was restricted to the gradiometers. Individual structural MRI scans were acquired using a 3-T MRI scanner (Siemens). Offline analysis of the data was carried out using custom-made MATLAB (MathWorks) scripts and the MATLAB-based open-source toolboxes FieldTrip (fieldtriptoolbox.org) (39), CircStat (40), and SPM8 (41). Continuously recorded data were segmented into trials. All trials were semiautomatically and visually inspected for artifacts, whereas artifacts caused by muscle activity, eye movements, or technical artifacts were removed semiautomatically using a z-score–based algorithm implemented in FieldTrip.

Virtual Channel Construction.

To focus on S1, we analyzed oscillatory activity in a predefined region of interest in source space (“virtual sensor”). Details regarding the construction of the virtual sensor are provided in SI Materials and Methods.