Abstract

Perceptual decision making requires the comparison and integration of sensory evidence to generate a behavioral response. We used magnetoencephalography to investigate the temporal dynamics of decision making during an auditory task that required forced-choice decisions about whether a pair of syllables S1 and S2 differed either in their acoustic patterns or in the perceived position of their sound sources. Conditions with easy and difficult decisions were created by varying the similarity of S1 and S2. Statistical probability mapping showed enhanced gamma-band activity (GBA) over posterior parietal cortex for spatial and over left inferior frontal cortex for pattern changes (at ∼120 to 220 ms after S2 onset). Activations were stronger for easy than difficult decisions. GBA over dorsolateral prefrontal cortex was more pronounced at ∼280 to 430 ms for easy than difficult decisions regardless of type of change, possibly reflecting decision-relevant networks that integrate information from higher sensory areas representing the perceptual alternatives. Sensorimotor beta desynchronization as a measure of motor preparation peaked at ∼460 ms for easy and at ∼520 ms for difficult decisions, thus reflecting the reaction time difference between both conditions. In summary, the analysis of oscillatory activity in magnetoencephalogram served to elucidate the temporal dynamics of perceptual decision making in humans.

Introduction

Perceptual decisions rely on the comparison and integration of sensory inputs to select a behavioral response. Single-cell studies in monkeys have shown that responses of neurons in visual cortex serve as predictors of psychophysical judgments (Newsome et al. 1989). During visual direction-of-motion tasks, decisions could be predicted by subtracting the integrated firing rates from pools of neurons preferring one of 2 perceptual choices (e.g., rightward or leftward movement) (Shadlen et al. 1996). Similarly, the frequencies of tactile vibratory stimuli can be discriminated by comparing the activity of sensory neurons selectively tuned to either slower or faster frequencies (Romo and Salinas 2003). Neurons that appear to integrate signals from these feature-selective units have been found in the lateral intraparietal area (Shadlen and Newsome 1996, 2001), the frontal lobe (Kim and Shadlen 1999), and the superior colliculus (Horwitz and Newsome 1999).

In humans, functional magnetic resonance imaging (fMRI) studies have focused predominantly on the processing of uncertainty or reward or on the emotional correlates of decision making in choice situations with a higher complexity than simple perceptual decision tasks (Paulus et al. 2001, 2002; Krawczyk 2002; Zysset et al. 2006). In contrast, a recent study by Heekeren et al. (2004) was designed by analogy to the animal experiments mentioned above. During a visual face- versus house-detection task with varying levels of difficulty, hemodynamic activation in feature-selective inferior temporal areas was modulated by the detectability of stimuli of either category. Left dorsolateral prefrontal cortex responded more strongly to easier than more difficult decisions, and its activation correlated with the difference between the activations of face- and house-selective ventral temporal regions. These results were interpreted in terms of a decision-making mechanism based on the integration of the responses of areas selectively tuned to the different perceptual alternatives (Heekeren et al. 2004; Rorie and Newsome 2005).

Tracing the time course of activations would help to further elucidate the role of a possible prefrontal decision area mediating between sensory input and motor output regions. The activation of this structure should follow the activations of sensory networks in time but precede activity in motor preparation areas. Although the time resolution of fMRI is insufficient to reveal the subsecond dynamics of these activations, electroencephalography (EEG) or magnetoencephalography (MEG), respectively, may serve to assess the fine temporal structure of decision processes in humans. The analysis of event-related potentials (ERPs) during visual categorization tasks has shown that whereas category-specific ERP differences may occur as early as 75 ms after stimulus onset, decision-related activity does not appear until about 150 ms (VanRullen and Thorpe 2001). A recent single-trial EEG study of face versus car discrimination (Philiastides and Sajda 2006) supported the relevance of early activity at about 170 ms poststimulus onset for decision performance. However, a second component starting at 300 ms showed a better match to the psychometric function and, importantly, shifted forward in time with more difficult decisions. It may thus have reflected a temporal accumulation mechanism whose processing time depended on the strength of sensory evidence.

In the present study, we investigated cortical oscillatory activity during a simple auditory decision task. Oscillatory signals may be recorded in MEG with a good spatial and high temporal resolution (Kaiser and Lutzenberger 2003; Bauer et al. 2006; Hoogenboom et al. 2006). In particular, gamma-band activity (GBA, >30 Hz) has been associated with the oscillatory synchronization of cortical networks involved in mental representations. For example, GBA enhancements have been reported during gestalt perception (Kaiser et al. 2004; Tallon-Baudry et al. 1996), multisensory integration (Kaiser et al. 2005, 2006; Senkowski et al. 2005), selective attention (Gruber et al. 1999; Müller and Keil 2004; Tallon-Baudry et al. 2005), and memory (Tallon-Baudry et al. 1998; Lutzenberger et al. 2002; Kaiser, Ripper, et al. 2003; Gruber et al. 2004; Herrmann et al. 2004; Gruber and Müller 2005). The wide variety of processes associated with increases of cortical oscillatory activity suggests that GBA may represent a general signature of activated cortical networks. In line with this assumption, recent studies have supported a link between GBA and hemodynamic cortical activity (Fiebach et al. 2005; Niessing et al. 2005).

We employed an auditory forced-choice task with 2 levels of difficulty where subjects had to decide whether 2 sounds differed either in their sound pattern or in the perceived spatial direction from which the sound was presented. A series of previous work both on passive auditory change detection and auditory working memory has consistently yielded GBA increases between 50 and 90 Hz in MEG sensors over anterior temporal/inferior frontal areas during auditory pattern processing (Kaiser et al. 2002; Kaiser, Ripper, et al. 2003) and over posterior temporoparietal cortex during auditory spatial processing (Kaiser et al. 2000; Lutzenberger et al. 2002; Leiberg et al. 2006). The topography of these activations is consistent with the putative auditory ventral and dorsal pathways thought to be involved in the processing of sound identity and sound source position, respectively (Rauschecker 1998; Romanski et al. 1999; Tian et al. 2001; Arnott et al. 2004, 2005). On the basis of our previous work and Heekeren et al. (2004) findings, we formulated the following hypotheses. Higher correct response rates and faster responses were expected for easy than for difficult decisions. Higher GBA amplitude would be found over posterior parietal cortex (putative auditory dorsal stream) for spatial changes and over anterior temporal/inferior frontal cortex (putative auditory ventral stream) for pattern changes. Greater GBA increases were expected when changes between the 2 stimuli were more pronounced, that is for the easier compared with the more difficult decisions. For a putative prefrontal decision-making area, we hypothesized higher GBA amplitude for easy than for difficult trials. The maximum of this activity should follow the peak in the higher sensory areas in time but precede the motor activity. Finally, we expected reductions in beta activity (∼13 to 30 Hz) over central areas to reflect motor preparation processes (Pfurtscheller and Lopes da Silva 1999; Kaiser, Ulrich, et al. 2003). The latency of motor-related beta desynchronization should be shorter for easy than difficult decisions.

Methods and Materials

Participants

Fourteen healthy adults (8 females, 6 males; mean age 25.9, standard deviation [SD] = 5.2 years) gave their written informed consent to participate in the study. One subject was excluded who had achieved an insufficient number of correct trials in the difficult pattern change condition (only 7 correct responses out of 100). Participants were paid for their participation. Thirteen subjects were right handed, one left handed. The study was conducted in accordance with the Declaration of Helsinki.

Procedure and Stimuli

Subjects were seated upright in a magnetically shielded room (VAC, Hanau, Germany). They were instructed to sit still and fixate a cross in the center of their visual field ∼2 m in front of them. Auditory stimuli were presented binaurally via air-conducting tubes with ear inserts.

In each trial, 2 acoustic stimuli S1 and S2 (duration 200 ms each) were presented with an interstimulus interval of 200 ms (Fig. 1). Participants were instructed to decide whether these 2 stimuli differed either in their spectral composition, that is their sound pattern, or in their spatial position (direction). Responses in this 2-alternative forced-choice task were given by pressing one of 2 buttons. Button 1 represented the answer “pattern change,” button 2 represented “direction change.” Buttons were placed at the left and right index fingers, and the assignment of buttons to the type of change was balanced across participants. Subjects were asked to respond immediately after the offset of S2. S1 was a sound halfway between the syllables /da/ and /ta/ (labeled /da′/) that was presented simultaneously to both ears, giving the percept of a sound source in the midsagittal plane. Whereas S1 was identical in all trials, S2 varied along the dimensions “type of change” (pattern vs. direction) and “task difficulty” (easy vs. difficult decisions). For easy pattern changes, S2 could be the syllable /te/ or /de/, representing easily recognizable deviations. Difficult pattern changes were created by presenting the syllables /da/ or /ta/ that were more similar to S1 (/da′/) and therefore more difficult to identify. Direction changes were implemented by presenting the same sound as S1 with an interaural time delay, creating the impression of a sound source lateralized either to the left or to the right from the midsagittal plane. For easy direction changes, a larger interaural time difference of 1 ms was used, whereas the interaural time difference amounted to only 0.2 ms for the difficult direction changes. Pretests were conducted to ensure that the chosen stimulus combinations differed in difficulty. In addition, a fifth condition was employed where S1 and S2 were identical. However, this last condition was not included in the analyses because the vast majority of subjects tended to give only one type of response (indicating either a perceived spatial or a pattern change), so that a comparison between both response types was not possible. A total of 500 trials were presented in 2 separate recording blocks of 250 trials each. Trials for each condition occurred at equal probabilities and in randomized order in each of the 2 recording blocks. The total number of trials per condition thus amounted to 100.

Figure 1.

Trial structure of the task. In each trial, 2 acoustic stimuli S1 and S2 of 200 ms duration each were presented with an interstimulus interval of 200 ms, see section on “Procedure and Stimuli” for details. Mean response (resp.) latencies varied between 630 and 700 ms after the onset of S2. The 3 horizontal bars above and below the time axis show the latency windows for spectral analyses (a1 300–900 ms, a2 300–1200 ms, and a3: 600–1200 ms).

Figure 1.

Trial structure of the task. In each trial, 2 acoustic stimuli S1 and S2 of 200 ms duration each were presented with an interstimulus interval of 200 ms, see section on “Procedure and Stimuli” for details. Mean response (resp.) latencies varied between 630 and 700 ms after the onset of S2. The 3 horizontal bars above and below the time axis show the latency windows for spectral analyses (a1 300–900 ms, a2 300–1200 ms, and a3: 600–1200 ms).

Data Recording

MEG was recorded with a 151 first-order magnetic gradiometer whole-head system (CTF Inc., Vancouver, Canada) with an average distance between sensors of ∼2.5 cm. The signals were sampled at a rate of 312.5 Hz with an antialiasing filter at 100 Hz. Recording epochs lasted from 400 ms before to 1400 ms after trial onset, allowing 1 s after the onset of S2 for the response. The subject's head position was determined with localization coils fixed at the nasion and the preauricular points at the beginning and end of each recording block to ensure that head movements did not exceed 0.5 cm.

Data Analysis

Three sets of spectral analyses were performed. The first served to identify the frequency ranges where activity differed between pattern and spatial changes. The second set of analyses focused on the comparison of easy and difficult decisions. The third analysis assessed motor preparation processes. The analyses followed a procedure that has been applied in a series of previous studies on MEG oscillatory responses (Lutzenberger et al. 2002; Kaiser, Ripper, et al. 2003; Kaiser et al. 2005). First, spectral analysis was performed to identify the frequency ranges with the most robust differences between conditions. Significance of the observed spectral power values for each frequency bin and MEG sensor was tested with a statistical probability mapping including corrections for multiple comparisons. Second, topography (sensors) and time courses of activations were assessed after filtering in the frequency ranges with the most pronounced differences between conditions.

Spectral analysis was conducted on single-trial basis for frequencies up to 90 Hz for the following time windows. For the comparison of pattern and direction changes, an analysis window a1 was chosen from 300 to 900 ms after trial onset, thus excluding the first 100 ms immediately after the offset of S1 to avoid contamination with the “off response” to S1. For the comparison of easy and difficult decisions, a longer time window was used between 300 and 1200 ms (a2) after trial onset. The time period for the analysis of activity related to motor processes was chosen from 600 to 1200 ms (a3) after trial onset. To reduce the frequency leakage for the different frequency bins, the records were multiplied by Welch windows. For example, a time window of 600 ms resulted in records of 150 points, which were zero-padded to obtain 256 points. Then Fast Fourier Transforms were carried out, and square roots of the power values were computed to obtain more normally distributed spectral amplitude values. These values were averaged across epochs to obtain measures of the total spectral activity for each condition. Spectral activity contrasts were evaluated with a statistical probability mapping procedure that has been used repeatedly in previous studies (e.g., Kaiser et al. 2005). It included corrections both for multiple comparisons and for possible correlations between data either from neighboring frequency bins (for spectral analysis) or from time points (for time course analysis). Significance criteria (corrected t values tcorr) were determined on the basis of permutation tests (Blair and Karniski 1993). Permutation tests allow to identify the probability to observe a difference of a certain size between 2 experimental conditions on the basis of the distribution obtained by randomly assigning the recorded data to the conditions. In general, the significance criteria obtained from the present procedure correspond to approximately P = 0.003 for 2 neighboring frequency bins.

Starting point was the comparison of group average spectral amplitude values for 2 conditions at each sensor and each frequency bin. This yielded the observed distributions of the t values for all frequency bins i × sensors j. To avoid spurious findings in individual frequency bins, we introduced the requirement that 2 neighboring frequency bins differ significantly between conditions. To ensure that tests for 2 consecutive frequency bins were significant, a new distribution of the minimal t values tm was computed for all pairs of neighboring frequency bins (time points) i and i + 1 at all sensors j: 

graphic

The next analysis step was designed to take into account possible correlations between neighboring frequency bins. The t value tm and its corresponding P value P0.05 were determined for which 5% of the observed tmi,j were larger. In the case of highly correlated data, P0.05 would be close to or smaller than 0.05, whereas for highly independent data, P0.05 would be greater than 0.05. The next step was to assess the random distribution of maximal t values in the present data set by exchanging the values for each trial type (or the signs of the differences between 2 conditions) at a time for all sensors j and frequency bins (time points) i on a subject-by-subject basis. This was done for 212 permutations of the 12 subjects. Each of these permutations now yielded a new maximum t value. The distribution of these maximal t values tmax for each of the nrand = 212 permutations was computed as follows: 

graphic

The corrected t value tcorr was now defined as the value where P0.05 × nrand of the obtained tmax were greater. This corrected t value tcorr was then applied as significance criterion to the observed data.

To explore the time course and the topographical localization of the observed spectral amplitude differences between conditions, the signals across the recording interval were multiplied with cosine windows at their beginnings and ends and filtered in the frequency ranges in which the statistical probability mapping had yielded significant effects. Noncausal, Gaussian curve–shaped Gabor filters (width ± 2.5 Hz, length in the time domain 100 ms) in the frequency domain were applied to the signals on a single-epoch basis for both conditions. The filtered data were amplitude-demodulated by means of a Hilbert transformation (Clochon et al. 1996) and then averaged across epochs for each condition. Differences in amplitude between conditions in the filtered frequency band were assessed with the statistical probability mapping procedure described above. In addition, relative effect sizes (ES) are given as 

graphic
where A and B are the GBA spectral amplitudes in 2 conditions and N is the technical noise of the MEG system at a given sensor and frequency band.

To depict the topographical localization of the observed differential spectral amplitude enhancements, we assigned the sensor positions with significant spectral amplitude effects of each subject to common spatial coordinates (common coil system). Sensor positions with respect to the underlying cortical areas were determined using a volumetric magnetic resonance image of one subject. The error that is introduced by not using individual sensor locations was estimated in previous studies by using a single dipole for somatosensory evoked fields and 2 dipoles for the localization of the first auditory evoked component (N1m) (Kaiser et al. 2000). The comparison of individual sensor locations and the “common coil system” revealed differences ranging below the spatial resolution determined by the sensor spacing of 2.5 cm. This justified the application of a common coil system for the purpose of the present study where no exact source localization was attempted. In our opinion, techniques like beamformer (Hillebrand and Barnes 2005) may be applied successfully to the source localization of temporally sustained activations with a sufficient signal-to-noise ratio (Hoogenboom et al. 2006), whereas they may not yield reliable results for more transient effects like the ones described in the following. Moreover, surface GBA patterns observed with the present method have not suggested simple dipolar source structures. Although single dipole sources would produce 2 patches with strong magnetic fields, the single patches typically found in our previous work could possibly be attributed to a more complex structure of combined local sources that would generate a relatively weak field that is maximal over the area between the dipoles (see Kaiser et al. 2000 for a detailed discussion of the possible source structure). According to this model, the cortical generators would thus have to be localized in the vicinity of the sensor showing the strongest activation.

For the analysis of oscillatory activity related to motor processes, sensor positions were mirrored to the contralateral hemisphere for those subjects who had been instructed to respond to pattern changes by pressing a button with their left index finger and to direction changes with their right index finger. Assuming that there are no systematic differences between the motor activation patterns of both hemispheres, we thus created a new data set where right-hemisphere responses were always related to (left-hand) direction responses and left-hemisphere responses were always associated with (right-hand) pattern responses. To selectively assess motor activity specifically related to the lateralized response (instead of a combination of specific and unspecific motor activation), we followed the rationale of the lateralized readiness potential (Coles 1989; Leuthold et al. 1996) by subtracting first activity to direction from activity to pattern responses and then subtracting activity at right-hemisphere sensors from activity at homologous left-hemisphere sensors. The significance of the amplitude of the resulting signals was then tested against zero.

Results

Behavioral Data

Analysis of variance with the within-subject factors task difficulty (easy vs. difficult decisions) and type of change (sound pattern vs. spatial direction) was employed to evaluate correct response rates and reaction times for correct responses. A main effect of task difficulty was found for correct response rate (F1,12 = 17.9, P < 0.001) with a higher rate of correct responses for easy than difficult trials (easy 92.6% [SD = 1.3], difficult 80.4% [SD = 3.2]). In contrast, there was neither a main effect of task type nor an interaction task type × task difficulty (F1,12 = 1.1, P = 0.31), that is decisions on pattern or direction changes were of comparable difficulty. A similar main effect of task difficulty was observed for reaction times (F1,12 = 12.0, P < 0.005) with shorter mean response latencies to easy (628 ms, SD = 28 ms) than difficult decisions (685 ms, SD = 34 ms). Again there was neither a difference between reaction times to pattern and spatial changes nor an interaction task type × task difficulty. These findings demonstrate that the manipulation of difficulty was successfully implemented.

MEG Analyses

Comparison of Pattern and Direction Changes

The results of frequency analysis for the comparison of pattern and direction changes are depicted in Figure 2. Direction changes were associated with a relative enhancement of GBA at ∼75 Hz at a right parietal sensor. In contrast, higher spectral amplitude for pattern changes was observed at ∼64 Hz at a left frontal sensor. Both effects were found during the time window of 300–900 ms posttrial onset. They met the criterion of tcorr = 3.28 in the frequency range of 55–90 Hz. To explore the time course and the topography of these spectral amplitude differences, the data records were Gabor-filtered (filter width 5 Hz) in frequency ranges with center frequencies of 64 and 75 Hz, respectively. The topography and time courses of the GBA differences between pattern and direction changes in these frequency ranges are depicted in Figure 3. For spatial changes, the spectral amplitude enhancement at 75 ± 2.5 Hz at a right parietal sensor (Fig. 3, left map) was maximal between 100 and 130 ms after the onset of S2. The difference amplitude for this sensor during this time window amounted to 0.35 fT (SD = 0.07 fT), t12 = 5.05, P < 0.001, ES = 0.53. At the left frontal sensor with the most pronounced difference for pattern changes (Fig. 3, right map), the relative GBA enhancement at 64 ± 2.5 Hz peaked between 200 and 230 ms after the onset of S2, with an average difference between both conditions of 0.75 fT (SD = 0.15 fT), t12 = 4.96, P < 0.001, ES = 0.36.

Figure 2.

Comparison of oscillatory responses to pattern (patt.) versus direction (dir.) changes. The top curves show the spectral amplitudes (fT) for both types of changes for frequencies between 55 and 90 Hz at the frontal (fr.) and the parietal (pa.) sensor with the statistically most pronounced differences between conditions (see Fig. 3 for the localization of these sensors). The graph at the bottom shows the results (P values) of t-tests comparing spectral amplitudes between both types of change at those 2 sensors in the same frequency range. The solid line gives P values for the comparison of pattern minus direction responses (p – d) at the frontal sensor, and the dotted lines gives P values for the comparison of direction minus pattern responses (d – p) at the parietal sensor. Values below the horizontal dotted line at P = 1 correspond to inverse differences.

Figure 2.

Comparison of oscillatory responses to pattern (patt.) versus direction (dir.) changes. The top curves show the spectral amplitudes (fT) for both types of changes for frequencies between 55 and 90 Hz at the frontal (fr.) and the parietal (pa.) sensor with the statistically most pronounced differences between conditions (see Fig. 3 for the localization of these sensors). The graph at the bottom shows the results (P values) of t-tests comparing spectral amplitudes between both types of change at those 2 sensors in the same frequency range. The solid line gives P values for the comparison of pattern minus direction responses (p – d) at the frontal sensor, and the dotted lines gives P values for the comparison of direction minus pattern responses (d – p) at the parietal sensor. Values below the horizontal dotted line at P = 1 correspond to inverse differences.

Figure 3.

Topography of GBA differences between pattern and direction changes in the frequency ranges where statistical probability mapping had revealed significant effects. Each circle represents one of the 151 MEG sensors projected onto a 2-dimensional cortical surface map with some major anatomical landmarks (dorsal view, nose up). The size of the circle reflects the statistical strength of the GBA difference between both conditions. Filled circles symbolize relative spectral amplitude increases in response to direction changes, whereas open circles stand for relative spectral amplitude enhancements for pattern changes. The largest circles represent the sensors with the most robust GBA differences between conditions. The left and the right maps depict the topography of spectral amplitude differences in the 75-Hz and the 64-Hz ranges, respectively. The graph at the bottom depicts the results of t-test comparisons between conditions, that is the time courses of the P values of the spectral amplitude differences in the filtered frequency bands. The solid curve represents the time course of relatively higher spectral amplitude to direction than pattern changes at ∼75 Hz at the parietal sensor (largest filled circle in left map). The dotted curve represents the time course of relatively higher spectral amplitude to pattern than direction changes at ∼64 Hz at the frontal sensor (largest open circle in right map).

Figure 3.

Topography of GBA differences between pattern and direction changes in the frequency ranges where statistical probability mapping had revealed significant effects. Each circle represents one of the 151 MEG sensors projected onto a 2-dimensional cortical surface map with some major anatomical landmarks (dorsal view, nose up). The size of the circle reflects the statistical strength of the GBA difference between both conditions. Filled circles symbolize relative spectral amplitude increases in response to direction changes, whereas open circles stand for relative spectral amplitude enhancements for pattern changes. The largest circles represent the sensors with the most robust GBA differences between conditions. The left and the right maps depict the topography of spectral amplitude differences in the 75-Hz and the 64-Hz ranges, respectively. The graph at the bottom depicts the results of t-test comparisons between conditions, that is the time courses of the P values of the spectral amplitude differences in the filtered frequency bands. The solid curve represents the time course of relatively higher spectral amplitude to direction than pattern changes at ∼75 Hz at the parietal sensor (largest filled circle in left map). The dotted curve represents the time course of relatively higher spectral amplitude to pattern than direction changes at ∼64 Hz at the frontal sensor (largest open circle in right map).

Separate P value time courses for easy and difficult conditions at the sensors with the most pronounced responses to spatial and pattern changes are depicted in Figure 4A,B. At the parietal site (Fig. 3, left map), an early relative GBA enhancement was observed to easy spatial change decisions, whereas difficult decisions elicited a second peak ∼300 ms thereafter (Fig. 4A). The GBA increase to easily detectable direction changes peaked almost immediately after the onset of S2, that is at 50–80 ms. The mean 75 ± 2.5 Hz spectral amplitude difference in this latency range between easy direction and pattern decisions amounted to 0.69 fT (SD = 0.15 fT), t12 = 4.67, P < 0.001, ES = 1.1. In contrast, at 50–80 ms there was no difference between difficult direction and pattern decisions, and the direct comparison of easy and difficult decisions in this latency range showed that the amplitude difference between spatial and pattern changes was greater for easy than difficult decisions (F1,12 = 6.0, P = 0.031). In contrast, for difficult decisions, the amplitude difference between direction and pattern responses peaked at 350–380 ms after S2 onset. It amounted to 0.56 fT (SD = 0.19 fT), t12 = 3.0, P = 0.011, ES = 1.0.

Figure 4.

Time courses of spectral amplitude differences between easy and difficult (diff.) decisions. P values are shown for comparisons of activity in the frequency ranges and at the sensors with the most pronounced differences for each comparison at latencies between 100 ms before the onset of S2 until 700 ms thereafter. Values below P = 1 correspond to inverse differences. (A) Comparisons of easy and difficult direction versus pattern changes (solid and dotted curves, respectively) at ∼75 Hz at the parietal sensor with the most pronounced difference between direction and pattern stimulus changes (stim. par.). (B) Comparisons of easy and difficult pattern versus direction changes (solid and dotted curves, respectively) at ∼64 Hz at the left frontal sensor with the most pronounced difference between pattern and direction changes (stim. fr.). (C) Comparison of easy and difficult decisions at the 2 frontal sensors (f1: more anterior sensor, f2: more posterior frontal sensor) with the most pronounced enhancements to easy compared with difficult decisions (easy > diff.) at 80 Hz (f1) and 54 Hz (f2). (D) Comparison of easy and difficult decisions at the frontal (fr) and the parietal (pa) sensor with the most pronounced enhancements to difficult compared with easy decisions (easy < diff.) at 40 Hz (fr) and 58 Hz (pa). (E) Time course of response-related beta desynchronization (resp.) at ∼30 Hz over sensorimotor areas to easy decisions (solid curve) and difficult decisions (dotted curve).

Figure 4.

Time courses of spectral amplitude differences between easy and difficult (diff.) decisions. P values are shown for comparisons of activity in the frequency ranges and at the sensors with the most pronounced differences for each comparison at latencies between 100 ms before the onset of S2 until 700 ms thereafter. Values below P = 1 correspond to inverse differences. (A) Comparisons of easy and difficult direction versus pattern changes (solid and dotted curves, respectively) at ∼75 Hz at the parietal sensor with the most pronounced difference between direction and pattern stimulus changes (stim. par.). (B) Comparisons of easy and difficult pattern versus direction changes (solid and dotted curves, respectively) at ∼64 Hz at the left frontal sensor with the most pronounced difference between pattern and direction changes (stim. fr.). (C) Comparison of easy and difficult decisions at the 2 frontal sensors (f1: more anterior sensor, f2: more posterior frontal sensor) with the most pronounced enhancements to easy compared with difficult decisions (easy > diff.) at 80 Hz (f1) and 54 Hz (f2). (D) Comparison of easy and difficult decisions at the frontal (fr) and the parietal (pa) sensor with the most pronounced enhancements to difficult compared with easy decisions (easy < diff.) at 40 Hz (fr) and 58 Hz (pa). (E) Time course of response-related beta desynchronization (resp.) at ∼30 Hz over sensorimotor areas to easy decisions (solid curve) and difficult decisions (dotted curve).

At the left frontal sensor (Fig. 3, right map), easy pattern decisions gave rise to a more pronounced GBA increase than difficult decisions (Fig. 4B). The 64 ± 2.5 Hz amplitude difference between easy pattern and direction decisions peaked at 250–280 ms after the onset of S2. The mean spectral amplitude difference in this latency range amounted to 1.06 fT (SD = 0.18 fT), t12 = 5.77, P < 0.001, ES = 0.37. In contrast, the amplitude difference between difficult pattern and direction responses did not reach significance (mean 0.20 [SD = 0.18], t12 = 1.10, P = 0.29). In the latency window of 250–280 ms, the spectral amplitude difference between pattern and spatial changes was greater for easy than difficult decisions (F1,12 = 17.3, P = 0.001).

Comparison of Easy and Difficult Decisions

Trials with direction and pattern changes were pooled to investigate oscillatory components differentiating between easy and difficult decisions. The frequency ranges with the most pronounced oscillatory activity increases to easy decisions were 54 and 80 Hz (Fig. 5). These effects met the significance criterion of tcorr = 4.02 in the latency window of 300–1200 ms after trial onset and the frequency range of 53–80 Hz. The greatest GBA increases for difficult decisions were identified at frequencies of 40 and 58 Hz that met tcorr = 3.81 for the time window of 400–1300 ms after trial onset and in the frequency range of 40–60 Hz (Fig. 6). The topography of GBA increases for either type of decision is depicted in Figure 7. Easy decisions gave rise to GBA enhancements in 2 frontal sensors, whereas relative GBA increases to difficult decisions were found in a frontal and a parietal sensor.

Figure 5.

Comparison of oscillatory responses to easy versus difficult (diff.) decisions for the 2 frontal sensors showing increased responses to easy compared with difficult decisions. The top curves show the spectral amplitudes (fT) for both types of decisions for frequencies between 53 and 85 Hz at the more anterior and the more posterior frontal sensor (f1 and f2, respectively) (see top maps in Fig. 7 for the localization of these sensors). The graph at the bottom shows the results (P values) of t-tests comparing spectral amplitudes between both types of decisions at those 2 sensors in the same frequency range. The thin line gives P values for the comparison of easy minus difficult decisions (e – d) at the more anterior frontal sensor f1, and the bold line gives P values for the same comparison at the more posterior frontal sensor f2. Values below the horizontal dotted line at P = 1 correspond to inverse differences.

Figure 5.

Comparison of oscillatory responses to easy versus difficult (diff.) decisions for the 2 frontal sensors showing increased responses to easy compared with difficult decisions. The top curves show the spectral amplitudes (fT) for both types of decisions for frequencies between 53 and 85 Hz at the more anterior and the more posterior frontal sensor (f1 and f2, respectively) (see top maps in Fig. 7 for the localization of these sensors). The graph at the bottom shows the results (P values) of t-tests comparing spectral amplitudes between both types of decisions at those 2 sensors in the same frequency range. The thin line gives P values for the comparison of easy minus difficult decisions (e – d) at the more anterior frontal sensor f1, and the bold line gives P values for the same comparison at the more posterior frontal sensor f2. Values below the horizontal dotted line at P = 1 correspond to inverse differences.

Figure 6.

Comparison of oscillatory responses to easy versus difficult (diff.) decisions for the frontal (fr) and the parietal (pa) sensor showing increased responses to difficult compared with easy decisions. The top curves show the spectral amplitudes (fT) for both types of decisions for frequencies between 35 and 70 Hz at both sensors (see bottom maps in Fig. 7 for the localization of these sensors). The graph at the bottom shows the results (P values) of t-tests comparing spectral amplitudes between both types of decisions at those 2 sensors in the same frequency range. The thin line gives P values for the comparison of difficult minus easy decisions (d – e) at the frontal sensor fr, and the bold line gives P values for the same comparison at the parietal sensor pa. Values below the horizontal dotted line at P = 1 correspond to inverse differences.

Figure 6.

Comparison of oscillatory responses to easy versus difficult (diff.) decisions for the frontal (fr) and the parietal (pa) sensor showing increased responses to difficult compared with easy decisions. The top curves show the spectral amplitudes (fT) for both types of decisions for frequencies between 35 and 70 Hz at both sensors (see bottom maps in Fig. 7 for the localization of these sensors). The graph at the bottom shows the results (P values) of t-tests comparing spectral amplitudes between both types of decisions at those 2 sensors in the same frequency range. The thin line gives P values for the comparison of difficult minus easy decisions (d – e) at the frontal sensor fr, and the bold line gives P values for the same comparison at the parietal sensor pa. Values below the horizontal dotted line at P = 1 correspond to inverse differences.

Figure 7.

Topography of GBA differences between easy and difficult decisions in the frequency ranges where statistical probability mapping had revealed significant effects. Each circle represents one of the 151 MEG sensors projected onto a 2-dimensional cortical surface map with some major anatomical landmarks (dorsal view, nose up). The size of the circle reflects the statistical strength of the GBA difference between both conditions. Filled circles symbolize relative spectral amplitude increases in response to easy decisions, whereas open circles stand for relative spectral amplitude enhancements for difficult decisions. The largest circles represent the sensors with the most robust GBA differences between conditions. The top maps reflect the topography of spectral amplitude differences in the frequency ranges where the most significant increases to easy decisions were observed (top left 80 Hz, top right 54 Hz), and the bottom maps reflect the topography of spectral amplitude differences in the frequency ranges where the most significant increases to difficult decisions were observed (bottom left 40 Hz, bottom right 58 Hz).

Figure 7.

Topography of GBA differences between easy and difficult decisions in the frequency ranges where statistical probability mapping had revealed significant effects. Each circle represents one of the 151 MEG sensors projected onto a 2-dimensional cortical surface map with some major anatomical landmarks (dorsal view, nose up). The size of the circle reflects the statistical strength of the GBA difference between both conditions. Filled circles symbolize relative spectral amplitude increases in response to easy decisions, whereas open circles stand for relative spectral amplitude enhancements for difficult decisions. The largest circles represent the sensors with the most robust GBA differences between conditions. The top maps reflect the topography of spectral amplitude differences in the frequency ranges where the most significant increases to easy decisions were observed (top left 80 Hz, top right 54 Hz), and the bottom maps reflect the topography of spectral amplitude differences in the frequency ranges where the most significant increases to difficult decisions were observed (bottom left 40 Hz, bottom right 58 Hz).

The time courses of filtered activity in these frequency ranges are shown in Figure 4C,D. For easy decisions, activity at 80 ± 2.5 Hz at the more anterior frontal sensor peaked between 280 and 310 ms after the onset of S2 (mean 0.48 fT [SD = 0.11 fT], t12 = 4.33, P = 0.001, ES = 0.33), whereas the spectral amplitude increase at 54 ± 2.5 Hz at the more posterior frontal sensor reached its maximum later at 400–430 ms after the onset of S2 (mean 0.35 fT [SD = 0.08 fT], t12 = 4.49, P = 0.001, ES = 0.16). For difficult decisions, 58 ± 2.5 Hz oscillatory activity at the parietal sensor was most pronounced between 370 and 400 ms after S2 onset (mean 0.64 fT [SD = 0.16], t12 = 4.10, P = 0.001, ES = 0.15), whereas the frontal 40 ± 2.5 Hz GBA increase peaked at latencies of 400–430 Hz (mean 0.52 fT [SD = 0.14], t12 = 3.85, P = 0.002, ES = 0.19).

Activity Related to Motor Preparation

After computing the double difference between pattern and direction responses and between the left- and right-hemisphere sensors, the resulting signals were tested against zero with the statistical probability mapping procedure described above in latency windows between 200 and 800 ms after the onset of S2. This analysis revealed spectral amplitude reductions in the frequency range around 30 Hz in the high beta range (beta event-related desynchronization [ERD]) that met the criterion of tcorr = 3.81 for frequencies between 20 and 80 Hz. Figure 8 depicts the topography and time course of P values for motor-related activity in the 30 ± 2.5 Hz range. Both axial gradiometers (left map) and virtual planar gradiometers (right map) are shown. The latter were computed on the basis of spatial derivatives of MEG signals, leading to more focal ERD maps for spatially widespread signals (Bastiaansen and Knösche 2000). In both cases, activations with comparable time courses were observed in sensors close to the central sulcus. Figure 4E shows separate time courses for motor activity during trials with easy and difficult decisions. For easy decisions, beta ERD peaked at latencies between 460 and 470 ms after S2 onset (mean decrease 2.61 fT [SD = 0.43 fT], t12 = 6.05, P < 0.001, ES = 0.23), whereas difficult decisions were associated with beta ERD at longer latencies of 520 and 530 ms after the onset of S2 (mean decrease 3.61 fT [SD = 0.67 fT], t12 = 5.35, P < 0.001, ES = 0.32).

Figure 8.

Topography and time courses of activity related to motor response preparation. The top maps show the topography of the lateralized beta ERD (∼30 Hz) collapsed across hemispheres for all trials. Each circle represents one MEG sensors projected onto a 2-dimensional cortical surface map of one hemisphere with some major anatomical landmarks (dorsal view, nose up). The size of the circle reflects the statistical strength of beta ERD. The left map shows effects on the basis of the axial gradiometer sensors (s), whereas the right map shows the same effects for virtual planar gradiometers (pg) computed as spatial derivatives of MEG signals based on spline approximation (Bastiaansen and Knösche 2000; Kaiser et al. 2001) with a more focal distribution over somatosensory cortex. The graph at the bottom indicated the relatively similar time courses of beta ERD for sensors (s) and virtual planar gradiometers (pg). Peaks about 200 ms prior to the motor response (resp.) support the notion that beta ERD reflects motor preparation processes.

Figure 8.

Topography and time courses of activity related to motor response preparation. The top maps show the topography of the lateralized beta ERD (∼30 Hz) collapsed across hemispheres for all trials. Each circle represents one MEG sensors projected onto a 2-dimensional cortical surface map of one hemisphere with some major anatomical landmarks (dorsal view, nose up). The size of the circle reflects the statistical strength of beta ERD. The left map shows effects on the basis of the axial gradiometer sensors (s), whereas the right map shows the same effects for virtual planar gradiometers (pg) computed as spatial derivatives of MEG signals based on spline approximation (Bastiaansen and Knösche 2000; Kaiser et al. 2001) with a more focal distribution over somatosensory cortex. The graph at the bottom indicated the relatively similar time courses of beta ERD for sensors (s) and virtual planar gradiometers (pg). Peaks about 200 ms prior to the motor response (resp.) support the notion that beta ERD reflects motor preparation processes.

Discussion

The temporal dynamics of sensory decision making were investigated during a 2-alternative forced-choice task where subjects had to decide whether 2 syllables presented with a 200-ms interstimulus interval differed in their sound pattern or the perceived lateralization of their sound source. Two difficulty levels were created for each task by varying the similarity of S1 and S2. The successful implementation of the 2 difficulty levels is reflected by the significant differences in correct response rates and reaction times between easy and difficult tasks, whereas pattern and spatial task did not differ in these parameters. The present correct response rates of about 93% and 78% for easy and difficult trials, respectively, are consistent with previous studies on perceptual decision making (Heekeren et al. 2004).

Statistical probability mapping of MEG signals yielded several components of task-specific activation in the gamma and beta frequency ranges. These effects were obtained for the analysis of spectral activity on the basis of single trials, that is for the total spectral activity, whereas an analysis of evoked (phase-locked) activity did not yield task-specific differences. The topography and time courses of the identified activations are summarized separately for easy and difficult decisions in Figure 9. In summary, GBA components between ∼50 and 280 ms after the onset of S2 reflected the processing of sound pattern and direction changes in ventral prefrontal and parietal putative higher auditory processing areas, respectively. This was followed by oscillatory activations over prefrontal and parietal cortex in the latency range of ∼280 to 430 ms after S2 onset that differentiated between easy or difficult decisions. Finally, beta ERD over sensorimotor regions contralateral to the response side peaked between 460 and 530 ms after the onset of S2, thus preceding the motor responses that were given at ∼630 to 700 ms.

Figure 9.

Summary of findings. The top panel depicts time course and topography of the activations to easy decisions, whereas the bottom panel shows the same for difficult decisions. The sensors and corresponding statistical activation time courses (P values) are coded as follows: d, parietal sensor activated in response to direction changes; p, left inferior frontal sensor activated by pattern changes; f1 and f2, frontal sensors showing enhanced GBA to easy compared with difficult decisions; fr and pa, frontal and parietal sensors showing enhanced GBA to difficult compared with easy decisions; r, sensors (left and right hemisphere, dependent on response side and instruction) showing beta ERD during motor preparation. There is a clear temporal sequence of 1) change processing in areas belonging to the putative auditory dorsal and ventral streams (note that d and p show more pronounced responses to the more pronounced [easy] compared with less pronounced [difficult] changes), 2) responses in decision-relevant frontal and possibly attentional frontoparietal local networks, and 3) response preparation in sensorimotor regions. Values below P = 1 correspond to inverse differences.

Figure 9.

Summary of findings. The top panel depicts time course and topography of the activations to easy decisions, whereas the bottom panel shows the same for difficult decisions. The sensors and corresponding statistical activation time courses (P values) are coded as follows: d, parietal sensor activated in response to direction changes; p, left inferior frontal sensor activated by pattern changes; f1 and f2, frontal sensors showing enhanced GBA to easy compared with difficult decisions; fr and pa, frontal and parietal sensors showing enhanced GBA to difficult compared with easy decisions; r, sensors (left and right hemisphere, dependent on response side and instruction) showing beta ERD during motor preparation. There is a clear temporal sequence of 1) change processing in areas belonging to the putative auditory dorsal and ventral streams (note that d and p show more pronounced responses to the more pronounced [easy] compared with less pronounced [difficult] changes), 2) responses in decision-relevant frontal and possibly attentional frontoparietal local networks, and 3) response preparation in sensorimotor regions. Values below P = 1 correspond to inverse differences.

Auditory decisions were accompanied by GBA increases at ∼75 Hz over right parietal cortex for spatial changes and at ∼64 Hz over left inferior frontal cortex for acoustic pattern changes. These areas have been proposed to form part of the putative auditory dorsal and ventral, “where” and “what” processing streams suggested by animal electrophysiology and anatomy studies (Rauschecker 1998; Romanski et al. 1999; Tian et al. 2001) and corroborated by human fMRI (Alain et al. 2001; Maeder et al. 2001; Arnott et al. 2004) and MEG work (Kaiser et al. 2000, 2002; Lutzenberger et al. 2002). Activations over putative higher auditory spatial and pattern processing areas may be interpreted as the counterparts to the activations of areas along the visual ventral stream during visual categorization task reported both in EEG (VanRullen and Thorpe 2001) and fMRI (Heekeren et al. 2004). Relative GBA enhancements over these areas were more pronounced for easy than difficult decisions, that is, when S1 and S2 differed more clearly from each other. This response pattern supports the selectivity of these regions to the perceptual alternatives that were relevant in the present task.

GBA to direction changes peaked on average about 100 ms faster than to pattern changes. This difference is consistent with faster spatial than pattern mismatch responses during passive acoustic change processing (Schröger 1995; Schröger and Wolff 1997) that may be attributable to the fact that interaural time delays are present at sound onset, whereas differences in spectral composition take time to unfold. Moreover, auditory spatial cues are already processed in the superior colliculi (King and Hutchings 1987), whereas more complex sound features may be extracted only at a cortical level (Griffiths and Warren 2002). Moreover, the present sound patterns were highly similar, particularly following sound onset, possibly resulting in a slower change detection than when sounds are contrasted that differ more strongly from each other (Murray et al. 2006). For spatial changes, there was a pronounced latency difference between easy and difficult decisions. Easily detectable changes elicited a fast GBA peak at only ∼50 to 80 ms after the onset of S2, whereas more difficult decisions were accompanied by a weaker GBA enhancement about 300 ms thereafter. For pattern changes, task difficulty had an effect on the amplitude but not on the latency of oscillatory activity. These findings suggest that participants may have attended primarily to an obvious spatial difference. If such a change did not occur, they may then have searched for a pattern change. However, subjects did not indicate spontaneously any clear strategies.

Next in the sequence of activations were components differentiating between easy and difficult decisions across the 2 types of acoustic change. Easy decisions gave rise to relative GBA enhancements in 2 sensors over left dorsal prefrontal cortex at latencies of ∼280 to 310 ms for the more anterior sensor at ∼80 Hz and at ∼400 to 430 ms after S2 onset for the more posterior of the 2 sensors at ∼54 Hz. The time course of these activations is consistent with ERP findings of decision-related components at latencies >300 ms (Philiastides and Sajda 2006). By showing greater activity when evidence for a perceptual choice was most pronounced, these components thus met one criterion for integrator networks involved in decision making on the basis of sensory information (Shadlen and Newsome 1996). Another criterion is the association between activity of the putative prefrontal integrator area and the output of the sensory areas (Rorie and Newsome 2005). To explore this relationship, we computed correlations between GBA amplitude at the dorsal prefrontal sensors (sensors f1 and f2 in Fig. 9) on the one hand with the difference of standardized GBA amplitudes between the posterior parietal and the left inferior frontal sensors (sensors d and p, respectively, in Fig. 9) for the 4 conditions (easy and difficult pattern and spatial changes). The resulting moderate positive correlations between r = 0.35 and 0.37 add further support to the interpretation of the dorsal prefrontal components as integrator networks (Fig. 10). Moreover, the temporal dynamics of the present oscillatory components with peaks in between the activation maxima of putative higher auditory regions and motor networks suggest that prefrontal regions may guide activity from lower level sensory areas to those that plan and execute responses (Kim and Shadlen 1999; Miller and Cohen 2001; Koechlin et al. 2003).

Figure 10.

Correlations between GBA recorded over dorsal prefrontal cortex (sensors f1 and f2 in Fig. 9) and over posterior parietal and left inferior frontal (sensors d and p, respectively, in Fig. 9) putative higher sensory areas. GBA amplitude (after subtraction of sensor noise) at the dorsal frontal sensors is given on the ordinate. For the anterior sensor f1, 80 ± 2.5 Hz spectral amplitude at 250–350 ms after S2 onset was selected, whereas for the posterior sensor f2, 54 ± 2.5 Hz spectral amplitude at 350–450 ms after S2 onset was entered into the analysis. The abscissa represents GBA amplitude differences between the stimulus representations in parietal and inferior frontal sensors that were computed as follows: baseline-corrected 75 ± 2.5 Hz spectral amplitude at 0–100 ms after S2 onset at the parietal sensor constituted the activity related to direction changes, whereas baseline-corrected 64 ± 2.5 Hz spectral amplitude at 200–300 ms after S2 onset at the inferior frontal sensor constituted the activity related to pattern changes. For easy and difficult direction changes, the differences between parietal, inferior frontal GBA were computed, whereas the inverse differences were calculated for pattern changes. Positive values on the abscissa thus correspond to larger amplitudes in response to the preferred type of change. Data points are included for the 4 conditions, resulting in n = 52.

Figure 10.

Correlations between GBA recorded over dorsal prefrontal cortex (sensors f1 and f2 in Fig. 9) and over posterior parietal and left inferior frontal (sensors d and p, respectively, in Fig. 9) putative higher sensory areas. GBA amplitude (after subtraction of sensor noise) at the dorsal frontal sensors is given on the ordinate. For the anterior sensor f1, 80 ± 2.5 Hz spectral amplitude at 250–350 ms after S2 onset was selected, whereas for the posterior sensor f2, 54 ± 2.5 Hz spectral amplitude at 350–450 ms after S2 onset was entered into the analysis. The abscissa represents GBA amplitude differences between the stimulus representations in parietal and inferior frontal sensors that were computed as follows: baseline-corrected 75 ± 2.5 Hz spectral amplitude at 0–100 ms after S2 onset at the parietal sensor constituted the activity related to direction changes, whereas baseline-corrected 64 ± 2.5 Hz spectral amplitude at 200–300 ms after S2 onset at the inferior frontal sensor constituted the activity related to pattern changes. For easy and difficult direction changes, the differences between parietal, inferior frontal GBA were computed, whereas the inverse differences were calculated for pattern changes. Positive values on the abscissa thus correspond to larger amplitudes in response to the preferred type of change. Data points are included for the 4 conditions, resulting in n = 52.

In the same latency window, difficult compared with easy decisions gave rise to GBA enhancements at ∼58 and 40 Hz in a parietal and a prefrontal sensor, respectively. Increased activation in frontal and parietal cortex during the more demanding sensory decisions was also observed in the fMRI study by Heekeren et al. (2004). In accordance with these authors, we would interpret these components as reflecting structures involved in the recruitment of attentional resources necessary to compensate for the poorer sensory evidence in the more difficult trials.

A measure of lateralized motor activity was computed by pooling the data across response sides and types of decisions. Beta ERD over sensorimotor regions demonstrated a latency difference of ∼60 ms between easy and difficult decisions. Spectral amplitude reductions in the alpha and beta frequency ranges in EEG (Pfurtscheller et al. 1998; Pfurtscheller and Lopes da Silva 1999; Stancak et al. 2000), MEG (Taniguchi et al. 2000; Kaiser, Ulrich, et al. 2003), and intracranial recordings (Szurhaj et al. 2003) have been shown to be closely related to the preparation of motor responses. Moreover, the peak of beta ERD appears to reflect the timing of response selection (Kaiser et al. 2001). This interpretation is supported by the timing of beta ERD in the present study: its peak latencies of ∼460 to 530 ms followed the activations of the putative decision-relevant or attentional prefrontoparietal components in time, and the beta ERD latency difference of ∼60 ms between easy and difficult decisions matched the mean reaction time difference of 68 ms between both conditions.

Although the present approach served to elucidate the fine temporal structure of decision-related oscillatory activity, as in our previous studies we did not attempt to localize the underlying cortical generators. Surface activation patterns of GBA with single sensors showing significant effects do not suggest simple dipole sources but could be explained better by more complex configurations of local sources that would produce relatively weak magnetic fields with a maximum over the area between the sources (Kaiser et al. 2000). In contrast, activations in lower frequency bands like the beta ERD in the present study that typically show more widespread surface topographies can be localized more reliably by computing spatial derivatives of MEG signals based on spline approximation that lead to more focal ERD maps (Bastiaansen and Knösche 2000; Kaiser et al. 2001) as shown in Figure 8. On the basis of these considerations, we assume that local networks showing task-specific increases of GBA are localized in the vicinity of sensors where these effects are observed. We thus interpret the dorsal prefrontal topography of the present decision-relevant activations as consistent with the previously proposed dorsolateral prefrontal decision-making area (Heekeren et al. 2004).

In summary, the investigation of oscillatory activity in MEG has revealed the temporal dynamics of the cortical mechanisms underlying auditory perceptual decision making. Decision-relevant sensory information about the 2 perceptual choices was first processed in higher areas of the putative auditory dorsal and ventral streams, where activations were stronger for the less similar and thus easier-to-discriminate S1–S2 pairs. Subsequently, separate local networks in prefrontal cortex showed enhanced spectral amplitude during easy decisions, suggesting that these components reflect decision-relevant structures that integrate information from higher sensory areas. During the same time window, difficult decisions were accompanied by enhanced activation of frontoparietal areas putatively involved in attentional control. Prefrontal areas may have mediated between sensory regions and motor networks where response preparation formed the last step in the cascade of cortical activations.

This work was supported by Deutsche Forschungsgemeinschaft (SFB 550/C1). Conflict of Interest: None declared.

References

Alain
C
Arnott
SR
Hevenor
S
Graham
S
Grady
CL
“What” and “where” in the human auditory system
Proc Natl Acad Sci USA
 , 
2001
, vol. 
98
 (pg. 
12301
-
12306
)
Arnott
SR
Binns
MA
Grady
CL
Alain
C
Assessing the auditory dual-pathway model in humans
Neuroimage
 , 
2004
, vol. 
22
 (pg. 
401
-
408
)
Arnott
SR
Grady
CL
Hevenor
SJ
Graham
S
Alain
C
The functional organization of auditory working memory as revealed by fMRI
J Cogn Neurosci
 , 
2005
, vol. 
17
 (pg. 
819
-
831
)
Bastiaansen
MC
Knösche
TR
Tangential derivative mapping of axial MEG applied to event-related desynchronization research
Clin Neurophysiol
 , 
2000
, vol. 
111
 (pg. 
1300
-
1305
)
Bauer
M
Oostenveld
R
Peeters
M
Fries
P
Tactile spatial attention enhances gamma-band activity in somatosensory cortex and reduces low-frequency activity in parieto-occipital areas
J Neurosci
 , 
2006
, vol. 
26
 (pg. 
490
-
501
)
Blair
RC
Karniski
W
An alternative method for significance testing of waveform difference potentials
Psychophysiology
 , 
1993
, vol. 
30
 (pg. 
518
-
524
)
Clochon
P
Fontbonne
J
Lebrun
N
Etevenon
P
A new method for quantifying EEG event-related desynchronization:amplitude envelope analysis
Electroencephalogr Clin Neurophysiol
 , 
1996
, vol. 
98
 (pg. 
126
-
129
)
Coles
MG
Modern mind-brain reading: psychophysiology, physiology, and cognition
Psychophysiology
 , 
1989
, vol. 
26
 (pg. 
251
-
269
)
Fiebach
CJ
Gruber
T
Supp
GG
Neuronal mechanisms of repetition priming in occipitotemporal cortex: spatiotemporal evidence from functional magnetic resonance imaging and electroencephalography
J Neurosci
 , 
2005
, vol. 
25
 (pg. 
3414
-
3422
)
Griffiths
TD
Warren
JD
The planum temporale as a computational hub
Trends Neurosci
 , 
2002
, vol. 
25
 (pg. 
348
-
353
)
Gruber
T
Malinowski
P
Müller
MM
Modulation of oscillatory brain activity and evoked potentials in a repetition priming task in the human EEG
Eur J Neurosci
 , 
2004
, vol. 
19
 (pg. 
1073
-
1082
)
Gruber
T
Müller
MM
Oscillatory brain activity dissociates between associative stimulus content in a repetition priming task in the human EEG
Cereb Cortex
 , 
2005
, vol. 
15
 (pg. 
109
-
116
)
Gruber
T
Müller
MM
Keil
A
Elbert
T
Selective visual-spatial attention alters induced gamma band responses in the human EEG
Clin Neurophysiol
 , 
1999
, vol. 
110
 (pg. 
2074
-
2085
)
Heekeren
HR
Marrett
S
Bandettini
PA
Ungerleider
LG
A general mechanism for perceptual decision-making in the human brain
Nature
 , 
2004
, vol. 
431
 (pg. 
859
-
862
)
Herrmann
CS
Lenz
D
Junge
S
Busch
NA
Maess
B
Memory-matches evoke human gamma-responses
BMC Neurosci
 , 
2004
, vol. 
5
 pg. 
13
 
Hillebrand
A
Barnes
GR
Beamformer analysis of MEG data
Int Rev Neurobiol
 , 
2005
, vol. 
68
 (pg. 
149
-
171
)
Hoogenboom
N
Schoffelen
JM
Oostenveld
R
Parkes
LM
Fries
P
Localizing human visual gamma-band activity in frequency, time and space
Neuroimage
 , 
2006
, vol. 
29
 (pg. 
764
-
773
)
Horwitz
GD
Newsome
WT
Separate signals for target selection and movement specification in the superior colliculus
Science
 , 
1999
, vol. 
284
 (pg. 
1158
-
1161
)
Kaiser
J
Birbaumer
N
Lutzenberger
W
Event-related beta desynchronization indicates timing of response selection in a delayed-response paradigm in humans
Neurosci Lett
 , 
2001
, vol. 
312
 (pg. 
149
-
152
)
Kaiser
J
Bühler
M
Lutzenberger
W
Magnetoencephalographic gamma-band responses to illusory triangles in humans
Neuroimage
 , 
2004
, vol. 
23
 (pg. 
551
-
560
)
Kaiser
J
Hertrich
I
Ackermann
H
Lutzenberger
W
Gamma-band activity over early sensory areas predicts detection of changes in audiovisual speech stimuli
Neuroimage
 , 
2006
, vol. 
30
 (pg. 
1376
-
1382
)
Kaiser
J
Hertrich
I
Ackermann
H
Mathiak
K
Lutzenberger
W
Hearing lips: gamma-band activity during audiovisual speech perception
Cereb Cortex
 , 
2005
, vol. 
15
 (pg. 
646
-
653
)
Kaiser
J
Lutzenberger
W
Induced gamma-band activity and human brain function
Neuroscientist
 , 
2003
, vol. 
9
 (pg. 
475
-
484
)
Kaiser
J
Lutzenberger
W
Ackermann
H
Birbaumer
N
Dynamics of gamma-band activity induced by auditory pattern changes in humans
Cereb Cortex
 , 
2002
, vol. 
12
 (pg. 
212
-
221
)
Kaiser
J
Lutzenberger
W
Preissl
H
Ackermann
H
Birbaumer
N
Right-hemisphere dominance for the processing of sound-source lateralization
J Neurosci
 , 
2000
, vol. 
20
 (pg. 
6631
-
6639
)
Kaiser
J
Ripper
B
Birbaumer
N
Lutzenberger
W
Dynamics of gamma-band activity in human magnetoencephalogram during auditory pattern working memory
Neuroimage
 , 
2003
, vol. 
20
 (pg. 
816
-
827
)
Kaiser
J
Ulrich
R
Lutzenberger
W
Dynamics of sensorimotor cortex activation to spatial sounds precueing ipsi- versus contralateral manual responses
Cogn Brain Res
 , 
2003
, vol. 
17
 (pg. 
573
-
583
)
Kim
JN
Shadlen
MN
Neural correlates of a decision in the dorsolateral prefrontal cortex of the macaque
Nat Neurosci
 , 
1999
, vol. 
2
 (pg. 
176
-
185
)
King
AJ
Hutchings
ME
Spatial response properties of acoustically responsive neurons in the superior colliculus of the ferret: a map of auditory space
J Neurophysiol
 , 
1987
, vol. 
57
 (pg. 
596
-
624
)
Koechlin
E
Ody
C
Kouneiher
F
The architecture of cognitive control in the human prefrontal cortex
Science
 , 
2003
, vol. 
302
 (pg. 
1181
-
1185
)
Krawczyk
DC
Contributions of the prefrontal cortex to the neural basis of human decision making
Neurosci Biobehav Rev
 , 
2002
, vol. 
26
 (pg. 
631
-
664
)
Leiberg
S
Kaiser
J
Lutzenberger
W
Gamma-band activity dissociates between matching and nonmatching stimulus pairs in an auditory delayed matching-to-sample task
Neuroimage
 , 
2006
, vol. 
30
 (pg. 
1357
-
1364
)
Leuthold
H
Sommer
W
Ulrich
R
Partial advance information and response preparation: inferences from the lateralized readiness potential
J Exp Psychol Gen
 , 
1996
, vol. 
125
 (pg. 
307
-
323
)
Lutzenberger
W
Ripper
B
Busse
L
Birbaumer
N
Kaiser
J
Dynamics of gamma-band activity during an audiospatial working memory task in humans
J Neurosci
 , 
2002
, vol. 
22
 (pg. 
5630
-
5638
)
Maeder
PP
Meuli
RA
Adriani
M
Bellmann
A
Fornari
E
Thiran
JP
Pittet
A
Clarke
S
Distinct pathways involved in sound recognition and localization: a human fMRI study
Neuroimage
 , 
2001
, vol. 
14
 (pg. 
802
-
816
)
Miller
EK
Cohen
JD
An integrative theory of prefrontal cortex function
Annu Rev Neurosci
 , 
2001
, vol. 
24
 (pg. 
167
-
202
)
Müller
MM
Keil
A
Neuronal synchronization and selective color processing in the human brain
J Cogn Neurosci
 , 
2004
, vol. 
16
 (pg. 
503
-
522
)
Murray
MM
Camen
C
Gonzalez Andino
SL
Bovet
P
Clarke
S
Rapid brain discrimination of sounds of objects
J Neurosci
 , 
2006
, vol. 
26
 (pg. 
1293
-
1302
)
Newsome
WT
Britten
KH
Movshon
JA
Neuronal correlates of a perceptual decision
Nature
 , 
1989
, vol. 
341
 (pg. 
52
-
54
)
Niessing
J
Ebisch
B
Schmidt
KE
Niessing
M
Singer
W
Galuske
RA
Hemodynamic signals correlate tightly with synchronized gamma oscillations
Science
 , 
2005
, vol. 
309
 (pg. 
948
-
951
)
Paulus
MP
Hozack
N
Frank
L
Brown
GG
Error rate and outcome predictability affect neural activation in prefrontal cortex and anterior cingulate during decision-making
Neuroimage
 , 
2002
, vol. 
15
 (pg. 
836
-
846
)
Paulus
MP
Hozack
N
Zauscher
B
McDowell
JE
Frank
L
Brown
GG
Braff
DL
Prefrontal, parietal, and temporal cortex networks underlie decision-making in the presence of uncertainty
Neuroimage
 , 
2001
, vol. 
13
 (pg. 
91
-
100
)
Pfurtscheller
G
Lopes da Silva
FH
Event-related EEG/MEG synchronization and desynchronization: basic principles
Clin Neurophysiol
 , 
1999
, vol. 
110
 (pg. 
1842
-
1857
)
Pfurtscheller
G
Zalaudek
K
Neuper
C
Event-related beta synchronization after wrist, finger and thumb movement
Electroencephalogr Clin Neurophysiol
 , 
1998
, vol. 
109
 (pg. 
154
-
160
)
Philiastides
MG
Sajda
P
Temporal characterization of the neural correlates of perceptual decision making in the human brain
Cereb Cortex
 , 
2006
, vol. 
16
 (pg. 
509
-
518
)
Rauschecker
JP
Cortical processing of complex sounds
Curr Opin Neurobiol
 , 
1998
, vol. 
8
 (pg. 
516
-
521
)
Romanski
LM
Tian
B
Fritz
J
Mishkin
M
Goldman-Rakic
PS
Rauschecker
JP
Dual streams of auditory afferents target multiple domains in the primate prefrontal cortex
Nat Neurosci
 , 
1999
, vol. 
2
 (pg. 
1131
-
1136
)
Romo
R
Salinas
E
Flutter discrimination: neural codes, perception, memory and decision making
Nat Rev Neurosci
 , 
2003
, vol. 
4
 (pg. 
203
-
218
)
Rorie
AE
Newsome
WT
A general mechanism for decision-making in the human brain?
Trends Cogn Sci
 , 
2005
, vol. 
9
 (pg. 
41
-
43
)
Schröger
E
Processing of auditory deviants with changes in one versus two stimulus dimensions
Psychophysiology
 , 
1995
, vol. 
32
 (pg. 
55
-
65
)
Schröger
E
Wolff
C
Fast preattentive processing of location: a functional basis for selective listening in humans
Neurosci Lett
 , 
1997
, vol. 
232
 (pg. 
5
-
8
)
Senkowski
D
Talsma
D
Herrmann
CS
Woldorff
MG
Multisensory processing and oscillatory gamma responses: effects of spatial selective attention
Exp Brain Res
 , 
2005
, vol. 
166
 (pg. 
411
-
426
)
Shadlen
MN
Britten
KH
Newsome
WT
Movshon
JA
A computational analysis of the relationship between neuronal and behavioral responses to visual motion
J Neurosci
 , 
1996
, vol. 
16
 (pg. 
1486
-
1510
)
Shadlen
MN
Newsome
WT
Motion perception: seeing and deciding
Proc Natl Acad Sci USA
 , 
1996
, vol. 
93
 (pg. 
628
-
633
)
Shadlen
MN
Newsome
WT
Neural basis of a perceptual decision in the parietal cortex (area LIP) of the rhesus monkey
J Neurophysiol
 , 
2001
, vol. 
86
 (pg. 
1916
-
1936
)
Stancak
A
Jr
Feige
B
Lucking
CH
Kristeva-Feige
R
Oscillatory cortical activity and movement-related potentials in proximal and distal movements
Clin Neurophysiol
 , 
2000
, vol. 
111
 (pg. 
636
-
650
)
Szurhaj
W
Derambure
P
Labyt
E
Cassim
F
Bourriez
JL
Isnard
J
Guieu
JD
Mauguiere
F
Basic mechanisms of central rhythms reactivity to preparation and execution of a voluntary movement: a stereoelectroencephalographic study
Clin Neurophysiol
 , 
2003
, vol. 
114
 (pg. 
107
-
119
)
Tallon-Baudry
C
Bertrand
O
Delpuech
C
Pernier
J
Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual responses in human
J Neurosci
 , 
1996
, vol. 
16
 (pg. 
4240
-
4249
)
Tallon-Baudry
C
Bertrand
O
Henaff
MA
Isnard
J
Fischer
C
Attention modulates gamma-band oscillations differently in the human lateral occipital cortex and fusiform gyrus
Cereb Cortex
 , 
2005
, vol. 
15
 (pg. 
654
-
662
)
Tallon-Baudry
C
Bertrand
O
Peronnet
F
Pernier
J
Induced gamma-band activity during the delay of a visual short-term memory task in humans
J Neurosci
 , 
1998
, vol. 
18
 (pg. 
4244
-
4254
)
Taniguchi
M
Kato
A
Fujita
N
Hirata
M
Tanaka
H
Kihara
T
Ninomiya
H
Hirabuki
N
Nakamura
H
Robinson
SE
, et al.  . 
Movement-related desynchronization of the cerebral cortex studied with spatially filtered magnetoencephalography
Neuroimage
 , 
2000
, vol. 
12
 (pg. 
298
-
306
)
Tian
B
Reser
D
Durham
A
Kustov
A
Rauschecker
JP
Functional specialization in rhesus monkey auditory cortex
Science
 , 
2001
, vol. 
292
 (pg. 
290
-
293
)
VanRullen
R
Thorpe
SJ
The time course of visual processing: from early perception to decision-making
J Cogn Neurosci
 , 
2001
, vol. 
13
 (pg. 
454
-
461
)
Zysset
S
Wendt
CS
Volz
KG
Neumann
J
Huber
O
von Cramon
DY
The neural implementation of multi-attribute decision making: a parametric fMRI study with human subjects
Neuroimage
 , 
2006
, vol. 
31
 (pg. 
1380
-
1388
)