, in press). Data Acquisition and Reduction The EEG signals were recorded using a 129-channel geodesic sensor net, filtered (0.1�C100 Hz), amplified, leave a message and digitized at a sampling rate of 250 Hz using an AC-coupled, high-input impedance (200 M��) amplifier (Geodesic EEG System 200; Electrical Geodesics Inc.). The EEG data were visually inspected offline to examine the overall data quality. On average, about 2% of the channels were contaminated by artifacts for more than 50% of the recording and were interpolated using spherical splines. Then, a spatial filtering method implemented in BESA (v 5.1.8.10, MEGIS Software GmbH) was used to remove eyeblink artifacts. Next, EEG data were transformed to average reference and segmented into 1850-ms epochs (800ms before through 1050ms after picture onset).

Within each epoch, channels contaminated by artifacts (e.g., absolute EEG difference larger than 100 ��V and difference between two contiguous data points larger than 25 ��V) were identified. If fewer than 90% of the channels in an epoch were artifact free, the epoch was excluded from further analyses. This criterion resulted in the loss of approximately 17% of all epochs. A continuous wavelet transformation was conducted using BrainVision Analyzer software (v 1.05.0005, Brain Products GmbH). A complex Morlet wavelet (Herrmann, Grigutsch, & Busch, 2005) was used to decompose the EEG signals for each channel in the range of 6�C30 Hz. The step between successive frequencies was set at 1 Hz, and the Morlet parameter c was set to 7.0.

For each picture-viewing trial, we extracted the time course of power information for signals falling in the range of alpha oscillations (8�C12 Hz). The data were baseline-corrected by converting power to a decibel (dB) scale (Delorme & Makeig, 2004), using the average power in the window lasting from 500 to 200ms before picture onset as the baseline. To examine whether baseline levels of alpha power changed over the course of the picture-viewing session, we tested for significant effects of block (first half or second half of the experiment) and valence category (neutral, pleasant, unpleasant, and cigarette related) on alpha power measured during the baseline window using a mixed model in which subject was modeled as a random effect (PROC MIXED; SAS v9.2, SAS Institute).

Data Analysis Following previously Brefeldin_A published procedures (De Cesarei & Codispoti, 2011), we computed picture-related alpha ERD by averaging across parieto-occipital sensors (P7, P5, P3, P1, P2, P4, P6, P8, PO7, PO5, PO3, PO2, PO4, PO6, O3, O1, O2, and O4). Then, we obtained a single estimate of alpha ERD per picture-category per-subject by averaging alpha ERD between 400 and 800ms after picture onset (Figure 1). These estimates were analyzed using mixed models (PROC MIXED) with subject modeled as a random effect. Figure 1. Picture-induced alpha ERD in the parieto-occipital area.