Introduction
Steady state visual evoked potentials (SSVEPs) and their magnetic counterpart (SSVEFs) experienced a remarkable evolution in the cognitive neurosciences and brain computer interface literature. The number of publications using or investigating SSVEPs increased significantly during the last 30 years. For example, a snapshot of the last three years, i.e., from 2020 to 2022, Web of Science lists 298 publications with the words “SSVEP” or “steady state visual evoked” in the title. Ten years ago, i.e., between 2010 and 2012 the list contains 81 publications, and this went down to 14 between 2000 and 2003 and to 16 publications between 1990 and 1992, respectively.
This evolution is not too surprising, given the advantage that SSVEPs provide to investigate competitive attentional selection mechanisms in the visual (see for summaries Müller 2014, Norcia et al. 2015), auditory (cf. Linden et al. 1987, Woldorff and Hillyard 1991, Makeig et al. 1996, Picton et al. 2003) and somatosensory domain (Snyder 1992, Tobimatsu et al. 1999, Giabbiconi et al. 2007, Adler et al. 2009, Goltz et al. 2015, Pang and Müller 2015) and as a robust signal for brain computer interfaces (for a review see Zhu et al. 2010). The biggest advantage of SSVEPs is their high signal-to-noise ratio, in particular for frequencies in the range of 8-15 Hz, that allows to extract them even in single trials. Their ongoing oscillatory nature permits to track exact time courses of competitive interactions or attentional resource allocation in multi element stimulus designs, in particular in the visual domain (cf. Müller et al. 1998, Antonov et al. 2020, Gundlach et al. 2020, Gundlach et al. 2022). The power of frequency-tagged stimuli as a valuable tool in scientific and applied settings was nicely reviewed in a 2015 paper by Norcia and colleagues (Norcia et al. 2015), and we would like to refer interested readers to that paper, rather than listing all the areas here.
In visual attention research, it was demonstrated that the amplitude of SSVEPs increased significantly when a flickering stimulus was attended, compared to when it was ignored. This modulation pattern has been observed for spatial (cf. Morgan et al. 1996, Müller et al. 1998, Müller and Hübner 2002, Müller et al. 2003, Walter et al. 2014), feature-based (cf. Müller et al. 2006, Andersen et al. 2008, Andersen et al. 2013, Störmer and Alvarez 2014, Andersen et al. 2015) and object-based attention (cf. Brummerloh et al. 2019, Brummerloh and Müller 2019, Adamian et al. 2020). Despite the wealth of studies using SSVEPs as a tool, surprisingly few studies were interested in the underlying cortical generators. Among the earliest studies was a study by Müller and colleagues (Müller et al. 1997), using a 37-channels magnetometer and a single equivalent current dipole model for steady state responses at 6, 12 and 15 Hz. Results revealed sources in early visual cortex, and suggested differential activation patterns as a function of stimulation frequency. One year later, Hillyard and colleagues (Hillyard et al. 1997), in a combined electroencephalographic (EEG) and functional magnetic resonance imaging (fMRI) study, reported sources in ventral and lateral extrastriate visual areas in a visual spatial attention task, that were basically replicated in another spatial attention task study implementing a pure EEG-based source analysis of SSVEP signals (Müller et al. 1998). While these early studies found sources of SSVEPs predominately in extrastriate cortex, studies that were conducted after the turn of the millennium consistently found generators of the SSVEP in V1 as well. Pastor and colleagues (Pastor et al. 2003) presented their subjects a broad range of flickering stimuli from 2-90 Hz while they recorded EEG and used some of these frequencies during positron emission tomography (PET) recordings. They reported significant activation in primary visual cortex (see also Pastor et al. 2007). These results are in line with an MEG study from Fawcett and colleagues (Fawcett et al. 2004) who also presented their subjects a broad range of frequencies up to 21 Hz, and by Kaufmann and colleagues (Kaufmann et al. 2001) for the same frequency range in an fMRI study. The most extensive study, however, was presented by DiRusso and colleagues (Di Russo et al. 2007). In their study a 6 Hz Gabor stimulus was presented in a passive viewing task. They recorded EEG and fMRI, and used retinotopic mapping to estimate SSVEP sources in visual cortex. They found SSVEP generators in V1, V3A, V4/V8 and hMT+/V5.
In summary, these studies provided clear evidence, that the generators of SSVEPs are located in early visual cortex that are linked to low-level visual processing. This holds for a broad range of flicker frequencies and stimulus types (for flickering facial stimuli see Wieser and Keil 2011). However, when it comes to the question of what are the sources that generate attention effects, e. g. the attentionalSSVEP amplitude enhancement, these studies have only limited value. First, almost all studies used a passive viewing task, and therefore were agnostic to attentional modulations. Second, in the fMRI study by Hillyard et al. (1997) conditions in which attention was deployed to one side of the screen were contrasted with passive viewing. This did not allow to control for attentional deployment during passive viewing (e. g. did subjects attend to the fixation cross, or to both flickering stimuli, or attended to both flickers in an arbitrary alternating manner during 36 seconds of recording in blocks of passive viewing?). In order to get a better estimate of cortical sources that contribute to attention effects, it is mandatory to control for subjects’ attentional state/deployment during a reference or baseline stimulation period.
In the current MEG study, we used a spatial attention design with a flickering stimulus presented in the left and right visual hemifield, respectively. In a trial-by-trial fashion, in an interval between 1500 and 2000 ms after flicker onset, subjects were cued to shift attention to the left or right stimulus and to perform a task at the attended location/stimulus. In order to control for attentional deployment during the pre-cue period (i.e. baseline), subjects either attended to the fixation cross or to both stimuli in the left and right visual hemifields and performed a detection task associated with the respective spatial location(s). With this manipulation we were able to (a) estimate the cortical generators while attention was focused away from the peripheral stimuli (i.e. attend fixation cross), (b) when they attended to both stimuli (left and right), and (c) the contribution of cortical areas of attentional amplitude enhancement in the post-cue period, relative to the two pre-cue baseline conditions,.