Participants 
22 participants were tested in the current study. 4 participant datasets were excluded due to insufficient trial numbers to form a meaningful condition average and one dataset was excluded due to strong contamination throughout the scalp originating from anterior/facial areas during the trials which introduced a distinct 10 Hz distortion and separate broadband distortion rising at 15 Hz and extending through to approximately 50 Hz, matching artefacts observed in previous studies of frontalis muscle activity [26, 27]. Datasets were analysed from
the remaining 17 participants (13 females, 17 right-handed), aged between 19 and 44 years (mean age = 26.5 years, SD = 7). All participants had normal or corrected-to-normal vision, and gave their informed consent before participating in the study. Participants were offered 50 CHF for their time or course participation credits.

Stimuli and Procedure 
Participants were instructed to follow a moving target as it moved across a computer screen and to press a keyboard button when the target entered a spherical “goal” portion of the screen. The targets consistently travelled along a diagonal path from the top-left part of the screen to bottom-right goal section (see Fig 1) at a speed of 3.75 deg/s. The target stimuli consisted of a black and white rectangle (1.05˚ x 2.10˚ visual angle) checkerboard pattern alternating (reversing between black and white) at a consistent rate of 30 Hz against a white background. The 30 Hz flicker created the frequency tag used for the subsequent EEG analysis of
visual attention. Two experimental conditions were created by manipulating what participants expected to see in the trials. In one condition block, participants were instructed that only one target would travel across the screen in each trial, and that they should press the keyboard button when it reached the goal area. In the other condition block, participants were instructed that a second target might appear while the first target was still travelling across the screen (occurring in 2/3 of the trials in this condition). The second target (also flickering at the 30 Hz) appeared in the same location as the first target at the onset of the trial so that participants had a specific and predictable target for a covert shift of attention. The second target appeared at either 2136
or 2270 ms into the trial, providing a consistent time range for participants to predict when to direct covert attention with a small range of variability. Participants were instructed that if a second target did appear they were to then track that second target with their eyes and press the button when the second target reached the goal area. This was to provide a task-related division of attention, while balancing all low-level visual properties between the conditions up until the appearance of a second target in the periphery. Participants were informed at the beginning of each condition block whether to expect either only one or more than one target,
creating two experimental conditions; an undivided attention condition and a divided attention condition. The time-window leading up to the possible presentation of a second target thus formed the period of interest for our analysis, where shifts of attention relating to participants condition-related expectations were predicted to occur. Analyses were performed on the entire trial time-range, with particular focus on the period of interest where the participants' eye-gaze was directly over the moving targets. Analyses on the time period preceding this period are provided with the caveat that the conditions during this time period were uncontrolled for low-level visual properties relating to where the stimuli appeared in the participants' visual field.

There were 204 trials in total, with 102 in each of the divided and undivided attention conditions. The trials were divided into 4 alternating homogenous blocks, and the presentation order of these blocks was counter-balanced to avoid fatigue or order-effects by creating two block-orders presented to two participant groups (8 and 9 participants in the two counterbalanced groups). The experimental trials began with a fixation cross in the top left corner of the screen, corresponding to the region where the target would initially appear. When participants fixated on this cross area (1.3˚ x 1.3˚), the cross would disappear and the target would begin to emerge from the top-left corner 266 ms later, becoming fully visible at 667 ms. The trials ended when the participant made their decisions relating to the targets completely entering the goal area (3170 ms into the trial in the undivided condition, 5003 or 5136 ms into the trial in the divided condition) by making a keypress. The experimental stimuli were presented on a 24 inch VIEWPixx/3D monitor (1920 x 1080 pixels, 120 Hz refresh rate) at a distance of 75 cm, and presented through Experiment Builder (v1.10.1630) software.

Eye movement recording
Eye-positions were recorded through a desktop-mounted Eyelink 1000 monocular (left) eye-tracker sampling at 1000 Hz. Calibrations of the eye-tracker (13-points, average position error < 0.5˚) were performed at the beginning of the experimental blocks and after breaks in the trials. The onset of a trial was triggered by a fixation in a specified region in the top left part of the screen; if this was not fixated upon within 4 seconds after presentation then a re-calibration sequence was entered, ensuring effective calibration throughout each of the trials. The trials began with the flickering targets emerging near the upper-left portion of the screen 266 ms
after trial onset. The early part of the trials was characterised by the target stimuli approaching and passing the participants’ fixated gaze, and the subsequent orienting of their gaze to these moving targets through catch-up saccades. This orienting phase generally took approximately 500 ms before participants were able to align their smooth-pursuit eye movements with the movement of the targets. To allow for this, a time-window of analysis for eye-gaze and EEG was created, beginning 1000 ms after the onset of the trial (734 ms after the onset the first target) and ending at 2000 ms (shortly before a second target might appear at either 2136 or 2270 ms).

EEG recording and processing
Electrophysiological responses were recorded through a Biosemi Active-Two amplifier system, using 128 Ag/AgCl electrodes sampling at 1024 Hz. Additional electrodes were placed at the outer canthi and above of each eye, to register ocular movements and blinks. EEG data was processed offline through EEGLAB (14.1.0b) running in the MATLAB 2016B environment. After an initial bandpass filtering process (0.1 - 75 Hz, zero phase shift, linear finite impulse, Hamming window), epochs of 5000 ms duration were created, beginning at a -1000ms baseline period at the onset of the trial. To isolate and remove blink and eye-movement distortions, the 5000 ms epochs were subjected to Independent Component Analysis (ICA, using the ‘runica’ algorithm through EEGLAB) [36]. Independent components corresponding to frontal blink and saccade topographic distortions were isolated and removed from the data, as well as slow drift in EEG corresponding to smooth pursuit activity (see S2 Fig). However, in a number of datasets this slow drift was not able to be isolated through ICA, even though a clear drift could be observed in the raw data. This was not problematic in the current experimental design, however, as the slow drift was not related to frequencies overlapping the 30 Hz frequency tag utilised in the study (see S2A and S2B Fig for examples of the frequency responses of independent components associated with blink and smooth pursuit distortions).