The Posner cuing paradigm \citep*{Posner_1980} has long been a standard tool used in the study of spatial attention. A wealth of research has validated that, amidst a task to detect or discriminate the identity of a visual target stimulus as quickly as possible, performance is affected by the appearance of non-target "cue" stimuli. While the precise nature of cue stimuli may vary, two features of the cue stimulus have been shown to discriminate two modes of attention. When cues appear at locations where subsequent targets may appear, an "exogenous" mode of attention is said to be engaged whereby the cue automatically attracts attention to that location. When cues have some feature (location, identity, etc) that provides the participant with information on the likely location of a subsequent target, an "endogenous" mode of attention is said to be engaged whereby the participant learns the predictive utility of the cue and attention shifts to the location predicted by the cue. While it is possible to construct cue stimuli that engage both modes, a pure measure of exogenous spatial attention is achieved by presentation of cues at potential target locations but randomly so, thereby providing no information with regards to subsequent targets, and a pure measure of endogenous spatial attention is achieved by presentation of cues at a central location (which is typically not a potential target location) with some feature (color, shape, etc) that predicts the likely location of each subsequent target \citep*{Klein}
By manipulating the time interval between presentation of the cue and subsequent target, it is possible to examine the timecourse of the influence of the cue on performance in responding to the target. In the context of purely endogenous cues, it is observed that at very short intervals of 100ms or less, performance at responding to the target is unaffected by whether the target appears at either the cued location (cued targets) or an uncued location (uncued targets). However, at longer intervals performance in response to cued targets becomes increasingly superior to performance in response to uncued targets, reaching an asymptote around 300ms after which cued targets maintain a constant performance benefits over uncued targets. This pattern suggests that after some minimum time required to decode the prediction of the cue stimulus, attention shifts towards the predicted target location, where it remains until the target is presented. This timecourse contrasts with that observed when purely exogenous cues are employed, where performance benefits at the cued location relative to uncued locations arise after shorter cue-target intervals, reaching a peak at intervals of around 100ms after which the benefits rapidly decrease and even become negative for intervals of 500ms and longer, reflecting performance costs at the cued location relative to uncued locations. This pattern suggests that attention shifts to the location of the cue relatively rapidly following cue presentation, then similarly rapidly shifts away from that location and thereafter experiences a residual impairment in returning to the cued location if the target appears there. The benefit to performance at the cued location after short cue-target intervals is often termed "facilitation", while the cost to performance at the cued location after longer cue-target intervals is often termed "Inhibition of Return" \cite{Posner1984,Klein2000}
One limitation to these behavioural demonstrations of the timecourse of spatial attention is that on any given trial only one cue-target interval can be employed, providing information on a small temporal sliver of the complete timecourse on each trial. Many trials, each probing one of a number of unique cue-target intervals, are therefore needed to build up a resolved picture of the timecourse. As many experimental contexts are limited in the total time available for testing a given participant, the desire to use many intervals and obtain a precisely resolved timecourse competes with the desire to reduce measurement noise at any given interval by presentation of many trials with that interval. In light of this limitation, \citet*{Müller1998} innovated an alternative approach to the characterization of the timecourse of spatial attention that involved the addition of task-irrelevant visual stimuli that are flickering at different rates and measuring the brain's response to these stimuli using electroencephalography (EEG). The rationale of the \citet{Müller1998}'s method builds on a long history of research on EEG amidst visual stimulation and manipulations of attention.
It has been well established that shortly after the onset of visual stimulation, a characteristic series of deflections are observed in EEG recordings over the occipital cortex (i.e. a visual evoked potential, or VEP, \citealt{Adrian_1934}), and that attention to the visual stimulus increases the magnitude of these VEPs  \cite{Eason_1969,Van_Voorhis_1977}. Furthermore, as the delay between multiple successive stimuli is reduced, the subsequent record of VEPs becomes sinusoidal (i.e. a steady-state visual evoked potential, or SSVEP) with a period matching that of stimulus presentation \citep{ADRIAN_1934,Regan_1966} and an amplitude that reflects relative attention to each stimulus. The location of attention may therefore be deduced by the presentation of multiple such SSVEP stimuli, each with it's own frequency, at different spatial locations; after frequency-domain decomposition of the subsequent EEG recordings, one may infer that those locations associated with frequencies with higher amplitude were allotted more attention. Furthermore, computing relative power of SSVEP stimulus frequencies at different windows of time during a trial permits characterization of the timecourse of shifts of attention. \citet{M_ller_1998} employed this strategy to study the timecourse of endogenous spatial attention, using an endogenous cuing paradigm with a single long cue-target interval on every trial and concurrent SSVEP stimuli at the locations of possible target appearance. Consistent with the behavioural timecourse detailed above, \citet{Müller1998} observed that power for the frequency of the SSVEP stimulus at the cued location was at first (immediately following the cue) equal to that of the uncued location, but then the power for the frequency of the SSVEP stimulus at the cued location grew with time until it reached an asymptote at which it was substantially larger than that of the uncued location at the time of target presentation. More recently, \citet{Li_2017} provided a similar exploration of the timecourse of exogenous spatial attention, yielding data that matched that obtained by the purely behavioural explorations demonstrating greater attention at the cued location at times shortly following the cue (facilitation) and greater attention at the uncued location at times more distant from the time of the cue (IOR). 
Limitations exist, however, to existing approaches to the use of SSVEPs in the study of spatial attention. First, existing approaches rely on the use of SSVEP stimuli of differing frequencies, which may provide an imbalance of stimulation that affects the operation of attention. For example, if one stimulus is presented at 8Hz while the other is presented at 20Hz (the rates used by \citealt{Li2017}), the slower-presented stimulus may act to grab attention, undermining the manipulation of attention induced by the cue. Second, in the reliance of a windowed frequency-domain analysis, existing approaches become limited in their temporal resolution. That is, insofar as frequency-domain computations necessitate windows that include at least 2-3 periods of the frequencies of interest, they cannot characterize the dynamics of shifts of attention that occur faster than that window size. For example, with an 20Hz stimulus, a window of at least 150ms must be used and dynamics of attention faster than this duration cannot be characterized. While it may be considered that a sliding window could serve to obviate this limitation, this approach simply serves to low-pass filter any timecourse information present in the EEG data, filtering out precisely those rapid dynamics that motivated use of such sliding windows. This paper therefore proposes an alternative approach to the use of SSVEPs for the study of the temporal dynamics of spatial attention that eliminates both these limitations. 
The new approach detailed here attempts to take advantage of the weak localization of both VEPs and SSVEPs. That is, both single-cell recordings \cite{Hubel_1959,Marg_1973} and VEPs measured via EEG \citep*{Spehlmann_1965} demonstrate that the location of the stimulation in the visual scene affects where in the visual cortex neurons fire and thereby where on the scalp VEPs are observed, with stimulation of the left hemifield tending to associate with larger VEPs in the right occipital cortex, and right hemifield stimulation tending to associate with larger VEPs in the left occipital cortex. However, this localization is weak, particularly when measured via EEG, where tissue between brain and sensors serves to spatially smear electrical potentials associated with neural activity. Thus, an electrode located centrally on the occipital cortex (ex. Oz) may evidence both VEPs and SSVEPs with lateralized visual stimulation, correlating more strongly with the neighboring electrode contralateral to the source of stimulation and less strongly with the neighboring electrode ipsilateral to stimulation. Amidst simultaneous stimulation of both left and right visual space and equal attention to both locations, the signal at Oz should reflect the simple sum of this both sources of stimulation and correlate equally well with both left and right neighboring electrodes. When one location is more strongly attended than the other, the signal at Oz should be more strongly influenced by the attended source of stimulation than the unattended source and consequently correlate more strongly with the electrode contralateral to to the attended source than with the electrode ipsilateral to the attended source. Critically, the correlation between the central and lateral electrodes can be computed for every time sample of EEG recordings, permitting a temporal resolution limited only by the rate of this recording and any low-pass filters applied during data collection and analysis (which are typically on the order of 50Hz, or 20ms). This permits the observation of any rapid shifts of attention that would otherwise go unobserved by existing approaches. Furthermore, so long as they are phase-offset from each other, the same frequency may be used for all SSVEP stimuli, providing a balanced stimulation.
As a demonstration of this new approach, the following experiment explores the timecourse of exogenous spatial attention. The central task around which the experiment was designed was a go/no-go task where some identity of the target would have two possible attributes (ex. color: white/black, shape: diamond/square, etc), one of which is mapped to a "go" response in which case participants must make a response, while the other is mapped to a "no-go" response in which case participants refrain from responding. This task is useful as it provides task demands that are greater than simple detection tasks, yet do not involve the extra complication of effector selection associated with 2-alternative forced choice tasks. The use of go/no-go task in a neuroimaging context also provides the benefit of having a subset of trials on which no motor response is made, which can often contaminate examination of brain activity immediately consequent to the appearance of a target (though such activity is not of central interest in this report). Modifications to the traditional stimulus design of the Posner cuing experiment were necessary to accommodate simultaneous presentation of steady-state stimuli. Lateralized visual stimuli were placed at a relatively large distance from central fixation to minimize any cross-hemispheric influences (i.e. stimulation from the left hemifield yielding left-hemisphere activity, stimulation from the right hemifield yielding right-hemisphere activity). To then diminish participants' need/desire to move their eyes to the target (eye movements are typically discouraged in such experiments to isolate effects of covert from overt visual attention), a luminance discrimination (white circles vs black circles amidst a grey background) was chosen as the focus of the go/no-go task. As a secondary means of incentivizing participants to avoid eye movements, an eye tracker was used to provide real-time feedback when participants either moved their eyes or blinked (which also impair inference on covert visual attention).
Typical experiments examining exogenous spatial attention provide place-markers at the locations where targets might appear and change the luminance of these markers to provide cue stimuli. In this experiment these placeholders will additionally be used to provide the steady-state stimuli such that the placeholders will consist of light/dark banded rings whose polarity (i.e which bands are light & which bands are dark) will alternate to present the steady-state stimuli. The cue will consist of a brief enhancement of the contrast of these rings, from light-grey-and-dark-grey to white-and-black. As is typical in such experiments, a "cueback" stimulus will be presented at fixation halfway between the time of cue presentation and target presentation. Such cueback stimuli are thought to enhance the behavioural observation of IOR by drawing any residual facilitatory attention away from the cued location. To implement a cueback, fixation will be a smaller version of the banded rings (though with no steady-state polarity changes), and the cueback will manifest as a brief contrast increase. Finally, a cue-target interval of 1s will be used on every trial, which should be sufficient to observe IOR in the subsequent response data, and both facilitation and IOR in the subsequent EEG data. 

Methods

Research Ethics

This research was approved by the Dalhousie University Department of Psychology & Neuroscience's Human Research Participants & Ethics Committee, file number R14/15.37.RK.

Subjects

A total of 10 subjects (age range: 19–23; 6 female; 1 left-handed) were recruited from a local research experience subject pool and compensated for their time with course credit. 

Equipment

Visual stimuli were presented on a Sony OLED display running in its "flicker-free" mode during which it provides image updates at 60Hz while producing a pulse-width modulation of 120Hz. This means that any non-black stimulus is presented at 120Hz with a 50% duty cycle (4ms-on, 4ms-off) and a minimum of 2 cycles. As expected for an OLED display, high speed photometry using a labjack and photosensor conducted prior to initiating this experiment validates that this display provides <1ms response time for transitions to and from any point in the color space, ideal for precise visual stimulus presentation as required for steady-state visual stimulation (c.f. LCDs which have response times of up to 50ms and response time variability dependent on the color/greyscale values being transitioned from and to).
Unreferenced EEG was recorded using a BrainVision ActiChamp amplifier and Acticap system with a ground electrode at location AFz, recording using the Pycorder software at 1KHz with a 260Hz low-pass filter applied pre-storage (post-storage filters described in the Analysis section below) and electrodes at the locations Oz, O1 & O2 as well as the right and left mastoids, which were used for the computation of offline average-mastoid reference (in fact 64 locations were recorded, but only data from Oz/O1/O2 were analyzed in this report). Impedances of all electrodes were brought to below 15kOhm before data collection. Event codes were sent from the stimulus presentation system to the ActiChamp via a labjack. For more precise time alignment of these event codes with actual presentation of visual stimuli, a photostimulus was also presented with critical stimuli for detection by a photometer plugged into the ActiChamp AUX port. Eye movements and blinks were monitored for real-time feedback using an Eyelink 1000 eye tracking system. Manual responses were recorded via the analog triggers on a wired xbox360 gamepad. All stimulus presentation and response registration was achieved via Python using the PySDL2 package and OpenGL rendering of visual stimuli.

Procedure

The task to be described in detail below involves a white/black color discrimination and consequent go/no-go response. To counterbalance the mapping of color to response requirement, participants were assigned to either "black-go" or "white-go", with this assignment determined by the order in which they were recruited to participate in the experiment, with even-numbered participants assigned to "black-go" and odd-numbered participants assigned to "white-go".
Following consent, participants were fitted with the EEG cap and impedance of all electrodes was brought to below 15 kOhm. The participant was then instructed with regards to task according to the following script:
[Instructions for "black-go" group] In this experiment, you will be playing a simple game. Since we are using this camera to watch your eyes, we have to check that the camera is working ok frequently.  A dark-grey-and-light-grey disc with a hole at the center will appear on the screen [at this point the computer shows the fixation stimulus] at which point you can initiate the camera check by looking at the disc's hole and pressing both triggers on this gamepad with your pointer fingers.   Sometimes you might see the disc blink momentarily, and that means you should try the check again. If the camera has lost track of your eyes, I'll pause the experiment and re-calibrate the camera. If the camera is ok, two large and flickering rings will appear to the left and right of center, both with dark-grey and light-gray bands [at this point the computer shows the peripheral rings]. When you see these appear,  your next job is to try to keep your eyes focused at the center disc's hole,  not moving your eyes or blinking. You'll only have to keep this focus for a couple seconds, but if you move your eyes or blink while the big rings are on the screen, the computer will beep at you and start over at the camera check again. 
Now, while you're staring at the center disc, other things will happen on the screen.  Eventually, either a white or black filled-circle will appear in one of the flickering rings. If the circle is white, you don't need to do anything,  but if the circle is black, you need to press both triggers as quickly as you can. When the filled-circle appears, don't look at it, just keep your eyes on the center disc,  but use your peripheral vision to notice the filled-circle's color.  To  help motivate you to respond quickly when the circle is black, after you respond a black number will appear at the center of the screen showing you how long it took you to respond. Smaller numbers are better and you want the numbers to be between 1-4. If you see numbers 5 or larger,  that means you're responding too slowly and you need to try to respond faster. If you fail to respond within one second, a black X will appear at the center of the screen, telling you that you took too long. If you press the triggers when you're not supposed to, after a white circle appears, a white X will appear at the center of the screen. If you correctly refrain from responding when a white circle appears, a white zero will appear at the center of the screen.
Before the filled-circle appears, two other things will happen on the screen. First, one of the flickering rings will change contrast briefly, going from being dark-grey and light-grey to being fully black-and-white. About a half second after this, the disc at the center will also change contrast briefly and the circle will appear another half-second later. The location of the ring that changes contrast is completely random and doesn't predict where the subsequent circle is going to appear.
So the full sequence of events will be (1) the camera check,  (2) flickering rings appear and stare at center disc,  (3) one of the rings will  change contrast briefly,  (4) the center disc will change contrast briefly, (5) the filled-circle appears and you press both triggers if it's black, do nothing if it's white, and finally (6) you receive feedback on how quickly you responded. After all this, everything will repeat, so you'll be back to the camera check. You can take a moment at each camera check to blink if you need to before continuing the camera check and initiating the next sequence. You'll be given the opportunity for a longer break every few minutes and the first few minutes will be just for practice. Do you have any questions before we start?
After reading these instructions, the eye tracker was then calibrated using a 9-point calibration procedure and participants were given a final opportunity to ask questions before the practice session began. The trial sequence is shown in Figure 1. Throughout the trial, SSVEP stimuli changed polarity every 100ms (10Hz), with a phase offset of 180° (i.e. the left-located SSVEP stimulus changes polarity 50ms after the right-located SSVEP stimulus, and vice versa). If at any point the eye-tracker detected either a blink or a >1° deviation of gaze from the central fixation, the trial was immediately terminated with an auditory alert and the word "Blink!" or "Moved!" (respective) presented at fixation for 500ms, after which a new trial would begin.
Manipulated variables included: cue location (left vs right), target location (left vs right), and target color (white vs black). Repeating the full combination of these variables four times yields 32 trials, which were presented in random order in each block, including one practice and 10 experimental blocks. The total time for the experiment including initial set up was 2 hours.