Earplugs

Before the trial starts, a white plus sign (+, for fixation) appears in the center of the screen against a gray background along with black and white checkerboards on either side of the plus sign. During the trial, one of the checkerboards is briefly replaced (240 ms duration) in a pseudorandom manner, with one of four types of salient black and white visual stimuli: a Mooney face (Mooney and Ferguson, 1951 (link)) or a distorted version of it (randomly picked from a set of 30 each); a Kanizsa triangle (Kanizsa, 1979 ) or a distorted image with the same components as the Kanizsa triangle (randomly picked from a set of two each) as shown in Figure 1. In each block, a different version of the oddball paradigm is used by presenting the salient image types as follows: One of the four salient image types is presented on one side of the fixation sign for 80% of the time (Frequent image category) while two other image types are presented on the other side for a total 20% of the time (10% for each type, forming the Rare category). Figure 2. shows two examples of blocks with different stimulus types.
Additionally, each trial is presented with multiple audio-visual distractors as detailed in Figure 3. Auditory tones (75 dB) are delivered using earphones with disposable ear plugs. Standard auditory tones (80% trials, 1000 Hz, 15 ms long, 500 ms apart) and Deviant auditory tones (10% trials, 1500 Hz, 15 ms long) are presented in each block. The auditory tones are presented in one of three sets timed with reference to the visual stimulus onset: 240 ms before, 40 ms before, or 160 ms after, the start of the visual oddball trial. In addition, small checkerboards appear bilaterally on top or bottom (randomly allocated) for different trials as visual distracters as can be seen in Figure 1.
Each trial is 1550 ms long on average (varied between 1200 and 1900 ms). Participants have up to 700 ms to respond (using the index fingers of both hands) to the stimuli after which it is considered a missed response.
In each block, there are 25 active trials followed by three baseline trials [where only the fixation sign (+ sign) is presented on a gray background, without any checkerboard masks].
The average luminous intensity was tested to be similar across blocks using the SHINE Toolbox (Willenbockel et al., 2010 (link)) and MATLAB Image processing toolbox. The Mooney face images and their distorted versions were kindly provided by Peter Uhlhaas (Uhlhaas et al., 2006 (link)). We prepared the Kanizsa triangles and their distorted versions using the open source graphics editor Inkscape (http://www.inkscape.org/).

In dyadic conditions (see below), each subject used a robotic manipulandum, a twin visuomotor and haptic interface system (Tvins)^{37 (link)}, to control a simulated virtual beam (Fig. 1a). Each robot’s arm consists of parallel links which are driven by electric motors placed under the display board on which the task was visually rendered by a projector. The handles of the manipulanda are aero-magnetically floated on the support table to minimize friction, such that they move freely on the flat surface of the support table. In the present study, the handle was programmed to move only in forward–backward dimension (y; Fig. 1b). The forces exerted by a participant on the robot handle were measured by a six-axis force/torque sensor. Data were collected at a sampling rate of 2 kHz.
Participants sat on a reclining adjustable chair and wore a seat belt. We set the chair’s height so as to align the shoulder-arm-hand with the handle’s height on the horizontal plane. We positioned each participant by having his/her shoulder 45 cm away from the origin of the hand position. Each participant's forearm rested on a cuff that was mechanically supported on the same horizontal plane. Therefore, participants were not required to hold their arm against gravity. For safety reasons, movement of the Tvins could be stopped if one of the following criteria was met: when the emergency hand switch was pressed by the participant, when the force/torque sensor measured force greater than 20 N, or when two participants’ hands were 20 mm away from each other on the y- axis. The latter criterion was used to ensure that both subjects would initiate the movement at similar onset latencies from the time at which they saw the start area color change (movement start cue), thus avoiding one subject only starting the virtual beam transport task. Depending on the stiffness condition, the virtual beam and hand position may be apart. In such a case, as a safety measure, the 20-mm criterion was used for two reasons: to avoid causing a large driving force to the beam (see Eq. 3) and prevent the robot from exerting large feedback force on the subjects (see Eq. 4). Each participant grasped the robot handle under the display board with his/her dominant (right) hand. Additionally, we placed a partition to prevent participants from seeing each other. Participants wore earplugs and soundproof earmuffs to mute the robot sound.

After insertion of earplugs, the individual resting motor threshold (rMT) was determined by locating a “hot spot” on the motor cortex for stimulation of the right abductor pollicis brevis using single pulses at 0.2 Hz. For treatment the H-coil was advanced 6 cm anteriorly over the prefrontal cortex, with the H1 (labeled as “A”) remaining over the LPFC and the H7 coil (labeled as “B”) adjusted to a centralized position before treatment. Patients and operators were not aware of the differences between coil A and B. The rMT was rechecked at least once a week. Treatment intensity was 120% of rMT.
During 4 consecutive weeks (5 sessions/wk; Figure 1) patients were treated daily while during the 2 additional weeks patients were treated biweekly (with at least 48 hours between treatments). Patients in both treatment groups received a similar Deep TMS protocol (120% of rMT, 18 Hz, 2 seconds on and 20 seconds off over a 20-minute period; total of 1,980 stimuli per session), applied over the lateral (H1 coil) or medial (H7 coil) PFC. Maps of the electric field distribution generated in both treatment modes are provided in Supplemental Figure 5.