Discussion
The aim of the present MEG study was to identify areas in the visual cortex of the human brain that contribute to an attentional enhancement of steady state amplitudes with attention. In a typical spatial attention task, after a baseline period, subjects were given an instructive cue to shift attention to the left or right visual hemifield and to perform a task at the to-be-attended side. Different from many studies, we controlled for the attentional deployment before the cue by instructing to either attend to central fixation or to both stimuli in the left and right visual hemifield, respectively. Compliance was assured by having a task at the respective location during the baseline as well as in the post-cue period. With this manipulation, we were able to test SSVEF sensory gain modulations for peripheral stimuli under the conditions that attentional resources were shifted to this peripheral stimulus (attend cross) or away from an already attended peripheral stimulus (attend rings). Behavioral data showed that subjects were compliant with the task, although the “attend cross” condition seemed to be more difficult compared to the “attend rings” baseline condition.
Comparing the baseline conditions, we found a differential activation in early visual cortex. In particular, when subjects were instructed to attend to both rings, activation was not just restricted to V1 and V2, as for the “attend cross” baseline, but also included occipito-parietal cortex, and hMT+. Given that hMT+ was only activated in the “attend rings” pre-cue baseline period is in line with previous reports on hMT+ activation by motion and also by flickering stimuli (Treue and Martinez-Trujillo 1999, Huk and Heeger 2002, Palomares et al. 2012). Surprising to us was that this additional activation in that baseline condition was more prominent in the right cortical hemisphere (but numerically in the left hemisphere as well, see Figure 3B), whereas V1/V2 activation was clearly located in both hemispheres.
The attentional activation pattern after the cue, relative to pre-cue baseline across all areas, was different for the two pre-cue conditions. When subjects first attended to both rings in the left and right visual hemifield, cortical activation slightly increased for the side for which attention was maintained, but significantly decreased for the side for which attention was withdrawn. If, however, participants first attended to central fixation, cortical activation was significantly increased for the to-be-attended stimulus, and data revealed a trend for this activation boost to be even bigger compared to when subjects attended to both rings first. For the to-be-ignored side, activation remained basically on the pre-cue baseline level.
However, a closer inspection on the individual activation pattern of the respective cortical areas resembled an interactive activation pattern that not only depended on the pre-cue condition but also on the visual area. When subjects attended to both rings first, we observed a reduction in activity in areas V1, V2, pre-cuneus, occipital parietal and inferior-temporal cortex for the to-be-ignored stimulus, whereas for the to-be-attended ring activity basically remained at the pre-cue level. Thus, all areas followed the pattern we observed in the test across all areas (see above). Critically, when subjects attended to the fixation cross first, the SSVEF modulation patterns differed significantly between the cortical areas within the visual cortex. In V1, V2, pre-cuneus and occipital parietal cortex SSVEF power followed a similar pattern as during the pre-cue attend to both rings baseline. The SSVEF amplitudes elicited by the to-be-ignored hemifield flickers were suppressed. However, the to-be-attended flickers generated a greater SSVEF power boost after participants attended the central fixation cross than after having split their attention between both hemifields during the ring pre-cue baseline task. On the contrary, hMT+ and inferior-temporal cortex did not show a further reduction in activity for the to-be-ignored ring and no enhanced boost of SSVEF amplitudes for the to-be-attended hemifield after the fixate central cross pre-cue task.
The present observations nicely match our very simple previous source reconstructions of EEG data in feature-based attention (Andersen et al. 2008, Andersen and Müller 2010). When subjects attended to red or blue spatially superimposed random dot kinematograms (RDKs) we also found the sources of the attention effect in early visual cortex only, including V1/V2. While it is quite clear that V1/V2 show prominent responses to flicker stimuli in the respective stimulation frequency plus its harmonics (Rager and Singer 1998) it is still unknown how far flicker responses propagate in the visual processing hierarchy. A number of physiological restrictions suggest that propagation is restricted to areas of early visual cortex, due to dendritic low-pass filtering, in particular for higher flicker frequencies (Fortune and Rose 1997, Vaidya and Johnston 2013), increase in receptive field size (Hubel and Wiesel 1968), or synaptic input from cells that respond to different flicker frequencies, resulting in so-called intermodulation frequencies (Zemon and Ratcliff 1984). The idea that propagation of the respective driving frequencies is limited to early processing stages was also demonstrated in fMRI (Di Russo et al. 2007, Palomares et al. 2012).
The activation pattern in early visual cortex when subjects first attended central fixation nicely resembles the pattern of SSVEP amplitude time courses in spatial shifting designs (Müller et al. 1998, Müller 2008). In these studies, SSVEP amplitude for the to-be-ignored side remained on pre-cue baseline level, whereas SSVEP amplitude for the to-be-attended side exhibited a significant increase. However, the pattern of amplitude modulations differed in study by Gundlach and colleagues (2020) that used the same design as in the present study. In this study, when subjects first attended to the fixation cross, an increase in SSVEP amplitudes for both, the to-be-attended and unattended side was found, of course with a significantly greater increase for the to-be-attended side. When subjects attended to both rings first, we found no change in amplitude for the side that needed to be ignored after the cue, but only a significant increase for the then selected side. The difference between the more recent and the previous studies may lie in the fact that in the most recent study we used a very broad range of posterior electrodes including midline electrodes, whereas in the two older studies we only used one posterior-temporal electrode left/right, respectively for SSVEP time course analysis. This broader cluster might have picked-up activity from many more cortical areas. In previous studies we have shown that the pattern of SSVEP amplitude modulation is not identical over a broader range of electrodes as used in the 2020 study (Andersen et al. 2012, Müller et al. 2018), and therefore, it might be of interest to reanalyze this data with a more temporal and much smaller electrode cluster, or even single electrodes. In addition, as the focus of the work by Gundlach and colleagues (2020) focused on the relationship between SSVEP and alpha-band modulations in a spatial cueing paradigm, the SSVEP amplitudes were derived from single-trial FFT spectra. With such an analysis, induced activity of non-phase locked ongoing neural oscillatory activity is not averaged out and may contribute to the SSVEP amplitude estimates in the same frequency range. In the current study, however, SSVEP amplitude values were derived from trial-averaged FFT spectra promoting evoked signals such as SSVEPs while minimizing induced signals such as endogenous neural activity not phase locked to the stimulation. While the relative amplitude patterns (attended > unattended) are comparable, differences in the absolute amplitude may stem from differences in foundations of the signals revealed by different analysis approaches.
To summarize: The present study replicated previous studies that located cortical sources of SSVEPs driven by flickering (frequency-tagged) stimuli in early visual cortex. Our results complement and extend these studies showing that early visual cortex also drives attentional modulations of SSVEP amplitudes. While the activation pattern in V1/V2 was consistent, the specific modulatory effects differed across different regions within early visual cortex with a distinct pattern further up in the processing hierarchy in areas such as hMT+ and inferior-temporal cortex in the present case. Nevertheless, results again demonstrate the power of frequency-tagged stimuli to investigate attentional competitive interactions in multi-element stimulus displays. To what extent a certain area that is specialized in processing of a certain feature, such as V4 for color, contributes significantly more to the observed SSVEP amplitude modulations in color processing, compared to the other areas that “just” follow the on/off is subject to a current project and we will report the results in the nearer future.