### Subjects

Seventeen right-handed participants (10 female) with mean age (29.9 +/- 6.4 years) and with normal or corrected-to-normal visual acuity gave informed consent to participate in the study. Subjects reported no history of photic epilepsy, had not experienced recent critical life events and had no history of psychotherapy or current psychopathology. The participants received 30 Euro for participation. The ethics committee of the University of Konstanz approved the procedures.

### Stimuli

Seventy-five colored pictures were chosen on the basis of their normative ratings from the International Affective Picture System [22]. Of these, 25 pictures presented unpleasant events (e.g. mutilations, assaults, etc.), 25 showed pleasant events (e.g. sports, erotic couples, children, etc.) and 25 showed neutral events (e.g. neutral faces, household objects, etc.). The three categories differed significantly from each other in IAPS normative valence ratings (pleasant: 7.4, neutral: 4.9, unpleasant: 2.4).

Normative arousal ratings did not differ for pleasant and unpleasant contents, but mean arousal levels for both emotional categories were significantly higher than for neutral contents (pleasant: 5.6, neutral: 2.9, unpleasant: 5.8.) Brightness, contrast and color spectra of the stimuli were matched across picture categories.

Pictures were presented with a video projector (JVC™, DLA-G11E) with a refresh rate of 100 Hz on a white plastic screen attached to the ceiling of the room. Pictures subtended a visual angle of 10° horizontally and 8° vertically to either side from the center of the screen. In each trial, one picture was presented in a flickering mode of 10 Hz for four seconds, resulting in 40 on/off cycles (same picture shown and not shown) of 50 milliseconds each. The inter-trial interval varied randomly between 6 to 8 seconds. In the inter-trial interval a grey screen with a fixation cross was presented to aid participants in maintaining gaze on the center of the screen.

### Procedure

Upon arriving at the laboratory, participants were familiarized with the MEG chamber and an informed consent form was signed. Handedness was determined using the Edinburgh Inventory [23]. For artifact control, four electrodes for the electro-oculogram (EOG) were attached; two near the left and right outer canthus and two above and below the right eye. Two electrodes attached at the left and right lower forearm recorded the electrocardiogram, which was monitored during the recording. As the aim of the current study was to introduce a rather new method of analyzing the spatial and temporal course of visual evoked brain activation, the presentation of ECG data would have gone beyond the scope of the paper. Results from the ECG recordings and correlations with several psychological and neural markers will be reported in an additional article. Subjects were then seated in a magnetically shielded chamber and their head shapes were digitized with a Polhemus 3Space Fasttrack (Polhemus, Colchester, VT, USA). Five index points (left and right periauricular points, nasion, pseudo-Cz and pseudo-inion point at the forehead) were determined to calculate the relative head position within the MEG helmet for source analysis. Finally, subjects were placed under the MEG sensors and instructed to avoid eye movement during picture presentation. A video camera monitored subjects' behavior and assured compliance throughout the experiment.

Then, the screen was positioned in front of the subjects and the presentation of 75 flickering (10 Hz) stimuli started. After MEG recordings, subjects rated each of the 75 affective pictures regarding emotional valence and arousal using the Self-Assessment Manikin self-report scale [22].

### MEG recording

Magnetic brain activity was recorded using a 148 channel whole-head system (Magnes™ 2500 WH, 4D Neuroimage, San Diego, USA). Vertical eye movements and blinks were recorded using Ag/AgCl-electrodes attached above and below the right eye (vertical electrooculogram). Lateral eye movements were recorded using two of the aforementioned electrodes at the outer canthi (horizontal electrooculogram). Electrocardiogram was recorded with two of the same electrodes on the left and right lower forearm. The ECG and EOG data were amplified using Synamps (Neuroscan™) Amplifiers. The MEG, ECG and EOG data was recorded with a sample rate of 678.17 Hz and filtered online with a band pass filter between 0.1 Hz and 200 Hz.

Procedures included in the MEG acquisition software package (Whole Head System software, version 1.2.5; 4D Neuroimaging) corrected global external noise and cardiac artifacts. Eye artifacts were corrected using the algorithm implemented in BESA™ software [24]. Trials containing large blink or EMG artifacts or maximum amplitudes above 3.5 pT were discarded from further analysis. The MEG data were digitally band pass filtered between 1 Hz and 25 Hz (slopes: 6 and 24 dB/octave, respectively) before averaging for picture category over 5000 ms (500 ms pre-stimulus, 4000 ms stimulus presentation and 500 ms post-stimulus).

### Data Analysis

The data analysis was carried out in two steps: First, the mean amplitude of the 10 Hz component was assessed using a moving window approach. Second, the time course of the modulation of the 10 Hz component over the four second interval of picture presentation was estimated using a complex demodulation technique.

### Moving Average

For each category average, the 10 Hz Fourier component was derived using a moving window averaging procedure [8]. To avoid contamination of results with the event related early activity, the initial 500 ms of the picture presentation interval were excluded. The resulting 500 – 4000 ms post stimulus part of each epoch was baseline-corrected using the 500 ms pre-stimulus interval. A 400-ms window containing four cycles of the 10 Hz flickering stimuli was shifted in steps of 100 ms (one cycle) across the epoch, and the magnetic field data within the shifting windows in the time domain were further averaged.

The resulting four cycles per category, subject and MEG channel were submitted to the fast Fourier-transformation (FFT) technique [25]. The real and the imaginary parts of the 10 Hz Fourier component were extracted for further analysis.

### Minimum Norm Estimation for the Moving Average Data

The real and imaginary parts of the 10 Hz Fourier component per condition resulting from the procedure mentioned above were submitted to minimum norm source estimation and subsequently recombined by taking the square root of the sum of the two squared dipole orientations. Cortical sources were estimated using the L2 minimum norm estimate (MNE), following the approach suggested by Hauk et al. [26] using EMEGS [27]. The L2-minimum-norm estimate enables enhanced resolution of brain activations generating the magnetic field without a priori assumptions regarding the location and number of current sources [21]. Calculation of the L2 minimum norm was based on a one-shell spherical head model with 2 (azimuth and polar direction) by 197 evenly distributed dipolar sources. This calculation was based on information on the center of a fitted sphere to the digitized head shape and the positions of the MEG sensors relative to the head. A spherical shell (1 shell, 6 cm, 197 dipoles) with evenly distributed dipole locations then served as source space. This shell was chosen as a compromise between depth sensitivity and spatial resolution [26]. The regularization parameter λ was .02 and thus identical across all subjects and conditions. After computing the minimum norm estimation for the real and imaginary parts of the 10 Hz Fourier component, both values were combined by using the square root of the sum of squares of the two Fourier parts as an estimate of absolute power [28].

### Minimum Norm Estimation for the assessment of time course

In order to assess the time course of the steady-state activation, a complex demodulation procedure was applied to minimum norm estimation data. Therefore, in a first step, the minimum norm estimation was computed for the four second interval of picture presentation. Here, we applied the same L2-minimum-norm technique as mentioned above, with the difference, that a minimum norm estimation was computed for every sample point in the raw data (3391 in total). All other parameters were kept equal. In a second step, the time course of the relevant 10 Hz component was extracted using a complex demodulation procedure. The detailed procedure is described below.

### Time course assessment

The time course of the amplitude of the 10 Hz steady-state component was computed separately for each dipole using the complex demodulation procedure. This procedure allows reliable extraction of the alterations of the amplitude of an ongoing waveform [

29,

30]. The complex demodulation mathematically extracts a modulating signal from a carrier signal by multiplying the raw data with a sine and cosine of the desired frequency and subsequent band pass filtration. The complex demodulation is computed as follows:

These two functions are applied to the averaged MEG raw data (MEG(t)). The frequency F in this case represents the driving frequency (in this case 10 Hz). Then, a 2 Hz Butterworth-filter is applied. The amplitude A(t) of the modulating signal is then described using the formula:

Finally, a baseline correction is applied in the same step using the 500 ms pre-stimulus baseline interval.

### Statistical Analysis

As a result of the aforementioned procedures, we obtained two different outcomes: First, we received the mean amplitude of the 10 Hz Fourier coefficient for every dipole as a measure for the averaged activation of the steady-state signal. Second, the complex demodulation procedure was used to derive the amplitude of the 10 Hz signal (component) for every sample point within the four second data interval as a measure of the time course of the activation for each of the 197 projected dipoles. The main goal of the statistical analysis of the MNE data was to show differences between the activation towards the different picture categories. For this purpose, we calculated pair-wise comparisons of the source activities for the three conditions. Condition-dependent activity was reflected by the contrast between activation towards affective (pleasant and unpleasant) and neutral pictures. To test for significant differences between the dipole activation of the three picture categories, we computed permutation tests. This procedure is qualified to cope with the high number of comparisons on dipole level without predetermined regions of interests [31].

Although no formal correction for multiple comparisons (Type 1 error) was made, only temporal and spatial regions comprising several sample points or dipoles respectively were interpreted, thus controlling for by-chance differences.

The advantage of the permutation test is that it does not require any a priori assumption about the distribution of the data, as it generates all possible permutations of the data to represent the data distribution. For each pair-wise condition comparison, we determined cut-off values for significant differences of the condition contrast at single dipole location based on 1000 (moving average) and 500 (time course) draws, respectively. For each draw, the individual condition contrast maps were randomly exchanged to generate data for a random condition composition. As we aimed at two-tailed tests, the maximum as well as the minimum of the differences at all dipole locations obtained from each draw entered the distributions of 1000 (500 respectively) maximum and minimum difference values. The upper and the lower critical values were determined as the 2.5% lowest and highest value in this distribution. Taken together, these two 2.5% tails represent critical limits of the 95% significance level (p < 0.05). In order to assess the time course, this was done successively for each sample point of the four second ssVEF interval. Difference values with permutation p < 0.05 were plotted onto a standardized brain. In order to accomplish this, the upper and lower critical difference values were subtracted from the original difference (unpleasant vs. neutral and pleasant vs. neutral) values. Thus, values greater than 0 for the upper critical value and less than 0 for the lower critical value represent the regions and epochs containing significant differences. This yielded maps of significant differences for each sample point between the two affective and the neutral conditions representing the main effect for condition under the null hypothesis that no difference between the conditions exists.

Müller et al.[17] noted that the greatest steady-state evoked potentials are found at occipital scalp sites. The aim of the present study was to evaluate the course of the activation over time. This includes the assumption, that the activation elicited by the steady-state stimulus is not spatially fixed.