In the present study, we recorded behavioral and EEG responses from human subjects performing a covert visuospatial attention task, where they either attended to a 50%-contrast flickering checkerboard stimulus presented at 1 of 22 possible locations along an elongated hexagonal grid spanning ~ 2.58° x 9.03° visual angle across the entire computer screen or attended to a 100%-contrast central fixation (see detail in Figs. 1A-B and the Materials and Methods section). In the attend-stimulus blocks, they were instructed to fixate at the central fixation, covertly shift their attention to the stimulus location without producing eye movements and sustain their attention at the stimulus location in order to detect a contrast dimming that could occur in 20% of the entire trials. To monitor SSVEPs, the 50%-contrast checkerboard stimulus was flickered at 18.75 Hz for 1,000ms. In order to ensure that subjects sustained their attention throughout the entire stimulus duration, the contrast dimming event could appear any time from 150–650 ms after the stimulus onset for 350ms. Intertrial intervals (ITI) were also jittered from 500-1,000 ms so that subjects could not predict stimulus onset. The stimulus presentation in the attend-fixation blocks was identical to that in the attend-stimulus blocks except that the contrast dimming appeared at the central fixation and subjects were instructed to attend to the fixation throughout the entire block and ignore the presence of the 50%-contrast checkerboard stimulus that could appeared at 1 of the 22 possible locations. Here, the subjects’ task was to detect contrast dimming at the central fixation that could in 20% of the entire trials.
In order to control the level of task difficulty across attention conditions and eccentric locations, we adjusted the degrees of contrast decrement (i.e., contrast thresholds) on a block-by-block basis to maintain hit rate at around 0.7. To do this, we grouped the 22 stimulus locations into 7 sets of eccentricities and adjusted the contrast thresholds separately for the visual stimuli that fell into these individual groups: +8.45°, -3.72°, -1.94°, 0°, + 1.94°, + 3.72°, and + 8.45° visual angle, where 0° is the fovea and - and + values indicate the averaged eccentricity relative to the left and right of the fovea (Fig. 1B). Here we termed stimuli that fell into 0°, ± 1.94°, ± 3.72°, and ± 8.45° eccentricities foveal, parafoveal, near peripheral, and far peripheral stimulations, respectively.
Behavioral results
As illustrated in Fig. 1C, in the attend-fixation condition, we found that the level of contrast thresholds measured at the fixation did not change significantly as the eccentricity of the competing visual stimulus increased (F(6, 162) = 0.13, p = 0.992). However, in the attend-stimulus condition, the contrast threshold significantly increased as a function of eccentricity (F(6,162) = 26.11, p < 0.001). This result is consistent with the increased receptive field (RF) sizes and decreased cortical magnification factors of neurons in the early visual areas as a function of eccentricity (Cowey and Rolls, 1974; Gattass et al., 1981, 1988; Sereno et al., 1995; Engel et al., 1997; Duncan and Boynton, 2003; Wandell et al., 2007; Dumoulin and Wandell, 2008; Freeman and Simoncelli, 2011; Harvey and Dumoulin, 2011).
The difficulty levels of contrast detection were well maintained at ~ 0.7 hit rate across the different attention conditions and eccentricities (mean hit rate = ± SD = 0.70 ± 0.016) (Fig. 1D). There was a slight but significant main effect of eccentricity on hit rate (range = 0.69–0.73 across all locations; F(6, 162) = 2.80, p = 0.013). Post-hoc paired t-tests show that hit rate in the foveal locations were significantly higher than all parafoveal and peripheral locations (t(27)’s = 2.19–3.22, p’s = 0.003–0.037, two-tailed due to the known direction of the attention effect, Holm-Bonferroni corrected with the threshold of 0.05). However, there was neither a significant main effect of attention F(1, 27) = 3.21, p = 0.085) nor a significant interaction between attention and eccentricity on hit rate F(6, 162) = 0.35, p = 0.912).
To examine the influence of eye movement, we sorted trials based on the level of EOG signals (see Fig. 2 and Materials and Methods). We found that there was a significant interaction between the level of eye movements and attention F(1, 27) = 4.32, p = 0.047). This is driven by a significant main effect of attention in trials with low levels of eye movements (F(1, 27) = 4.77, p = 0.038) but no significant main effect of attention in trials with high levels of eye movements F(1, 27) = 0.02, p = 0.899). Post-hoc t-tests revealed that in trials with lower levels of eye movements, the attention effects on hit rate were significant at the parafoveal locations (t(27) = 2.74, p = 0.005, one-tailed due to the known direction of the attention effect, Holm-Bonferroni corrected with the threshold of 0.0125) and nearly significant at the peripheral locations (t(27)’s = 1.67–1.71, p’s = 0.049–0.053, one-tailed) but not significant at the foveal locations (t(27) = 0.79, p = 0.218, one-tailed). On the other hand, there was no significant attention effect on hit rate at any location for trials with high levels of eye movements (t(27)’s = 0.38–0.87, p’s = 0.197–0.354, one-tailed), suggesting that small saccadic responses were necessary for subjects to maintain similar levels of hit rate across eccentricities in the attended-stimulus condition due reduced contrast sensitivity to visual stimuli in the parafoveal and peripheral locations (see the contrast thresholds results where higher degrees of contrast changes were required for subjects to perform the contrast detection task in Fig. 1C).
As illustrated in Fig. 1D, the false alarm rate in general was relatively low (mean false alarm rate ± SD = 0.036 ± 0.008). There was a slight but significant eccentricity effect of false alarm rate (range = 0.024–0.048 across all locations; F(6, 162) = 5.37, p < 0.001). That said, there was no significant main effect of attention F(1, 27) = 0.51, p = 0.481) and no significant interactions between attention, eccentricity, and the level of eye movements (F’s = 1.78–1.97, p’s = 0.073–0.194).
Steady-state visually evoked potential (SSVEP) results
To monitor the effect of sustained visual attention on the stimulus-evoked neural activity in the early visual cortex, we measure SSVEPs, which were the phase-locked visually evoked EEG responses that oscillated at the same frequency as of the frequency of the flickering visual stimulus (i.e., 18.75 Hz). We used the SSVEP technique here because it is thought to capture the population-level visually evoked responses generated from the early visual areas (reviewed in Refs [58–59]; also see Refs (Regan, 1989; Müller et al., 1997; Di Russo et al., 2007; Andersen et al., 2008, 2012; Fuchs et al., 2008; Andersen and Müller, 2010; Störmer et al., 2013, 2014; Itthipuripat et al., 2013, 2014b, 2018, 2019)). Moreover, many past studies have consistently shown that SSVEP is a sensitive measure for neural gain in many variants of sustained visual attention tasks (Müller et al., 1998, 1997; Di Russo et al., 2007; Andersen et al., 2008, 2012; Fuchs et al., 2008; Andersen and Müller, 2010; Störmer et al., 2013, 2014; Itthipuripat et al., 2013, 2014a, 2018, 2019).
Consistent with previous reports (Müller et al., 1998, 1997; Di Russo et al., 2007; Andersen et al., 2008, 2012; Fuchs et al., 2008; Andersen and Müller, 2010; Störmer et al., 2013, 2014; Itthipuripat et al., 2013, 2014a, 2018, 2019), we observed robust SSVEP signals peaking at the flicker frequency of 18.75Hz and the spectral power at the SSVEP frequency peaked at the posterior occipital electrodes (Figs. 3–4). For the visual stimuli appearing within the fovea, the SSVEP signals were distributed bilaterally at the posterior occipital sites. However, as the stimuli were presented at the parafoveal and peripheral locations, the SSVEP signals shifted towards the contralateral compared to the ipsilateral electrodes, resulting in a significant interaction between eccentricity and channel location (i.e., left and right posterior occipital electrodes) (F(6,162) = 23.97, p < 0.001) (see Fig. 5). Overall, the mean SNR of the bilateral SSVEP signals elicited by the foveal stimuli were comparable to those of the contralateral SSVEP signals elicited by the parafoveal stimuli (t(27) = -0.39, p = 0.698, two-tailed, not passing the Holm-Bonferroni-corrected threshold of 0.05). However, the SNR of the contralateral SSVEP signals reduced significantly as the visual stimuli appeared in the eccentric locations further in the periphery, resulting in a significant main effect of eccentricity on the SSVEP SNR (F(6,162) = 34.49, p < 0.001). Post-hoc paired t-tests showed that the SSVEP SNR for the foveal and parafoveal locations were significantly higher than those for the near and far peripheral locations (t(27)’s = 4.28–7.06, p’s < 0.001, two-tailed). In addition, the SSVEP SNR was significantly higher for the near compared to the far peripheral locations (t(27) = 4.30, p < 0.001, two-tailed, all tests passed the Holm-Bonferroni-corrected threshold of 0.025). Taken together, these results were consistent with the fact that foveal vision has higher cortical magnification than peripheral vision (Cowey and Rolls, 1974; Sereno et al., 1995; Engel et al., 1997; Duncan and Boynton, 2003; Wandell et al., 2007; Harvey and Dumoulin, 2011).
As expected, attention significantly increased the SNR of the SSVEP signals (F(1,27) = 18.17, p < 0.001). Importantly, there were significant interactions between attention and eccentricity (F(6,162) = 7.23, p < 0.001) as well as between these factors and channel locations (F(6,162) = 8.62, p < 0.001), showing that the pattern of attentional modulations of SSVEP data was heterogeneous across eccentricities (Figs. 5A-D). As illustrated in Figs. 5D and 5H, these interactions could be described by two characteristics of attentional gain patterns of the SSVEP results across eccentricities: (i) the reduction of attentional modulations of the SSVEP signals at the foveal compared to the parafoveal locations (t(27) = -2.85, p = 0.008, two-tailed) and (ii) the reduction of the attentional modulations at the more peripheral direction compared to the parafoveal locations that occurred in a graded fashion (t(27) = 3.59, p = 0.001 for far peripheral vs. parafoveal location; t(27) = 2.90, p = 0.007 for far vs. near peripheral locations, all tests were two-tailed and passed the Holm-Bonferroni-corrected threshold of 0.025).
The SNR of the SSVEP signals also depends on the level of eye movements as the SSVEP SNR was significantly higher in trials with low compared to high levels of eye movements (F(1,27) = 21.6, p < 0.001) (Compare Figs. 5A and 5C to 5E and 5G). Importantly, the heterogeneous pattern of attentional modulations across eccentricities reported above was relatively more robust in trials with low levels of eye movements compared to those with high levels of eye movements, resulting in a significant three-way interaction between attention, eccentricity, and the level of eye movements (F(6,162) = 2.44, p = 0.028) (Compare Figs. 5B and 5D to 5F and 5H). As illustrated in Fig. 5D, in trials with low levels of eye movements, there was a significant reduction in the SSVEP SNR at the foveal compared to the parafoveal locations (t(27) = -2.80, p = 0.009, two-tailed). Also, the SSVEP SNR significantly reduced as visual stimuli appeared in the locations peripheral to the parafoveal locations (near peripheral vs. parafoveal location: t(27) = 2.60, p = 0.015; far peripheral vs. parafoveal location: t(27) = 3.97, p < 0.001; far vs. near peripheral locations: t(27) = 3.47, p = 0.002, all tests were two-tailed and passed the Holm-Bonferroni-corrected threshold of 0.05). For trials with high levels of eye movements, there was only a marginal difference between the SNR level of the SSVEP signals across the foveal and parafoveal locations (t(27)’s = -2.04, p = 0.052, two-tailed) (Fig. 5H). The reduction of the SSVEP SNR towards the periphery was significant but relatively less robust as there were significant SNR differences between parafoveal and far peripheral locations (t(27) = 3.55, p = 0.001, two-tailed) and between the near and far peripheral locations (t(27) = 2.38, p = 0.025, two-tailed) but no difference between the parafoveal and near peripheral locations (t(27) = 1.77, p = 0.088, all tests were two-tailed with the Holm-Bonferroni-corrected threshold of 0.025).
Overall, the SSVEP results suggested that the attention effects of early sensory responses as measured by SSVEPs were heterogeneous across eccentricities. Moreover, eye movements affected the SNR level of the SSVEP signals and the robustness of the heterogeneous pattern of attentional modulations of the SSVEP data measured as a function eccentricity.
Sustained negative deflection (SND) results
Next, we examined the slow negative-going wave in the ERP data, termed here as the sustained negative deflection (SND). The negative deflection in the ERP like the SND component has been found to track the locus of spatial attention in behavioral tasks where the stimulus appeared at the peripheral locations (Grent-’t-Jong and Woldorff, 2007; Kelly et al., 2009; Itthipuripat et al., 2018, 2019; Hakim et al., 2019). Specifically, these studies found that attending to the peripheral locations increased the amplitude of the negative deflection (i.e., it becomes more negative) in the contralateral posterior occipital electrodes (Itthipuripat et al., 2018, 2019; Hakim et al., 2019). Importantly, a seminal study combining EEG and fMRI demonstrated that the SND component reflects the top-down attentional biasing signals from the fronto-parietal cortex onto the upstream visual areas (Grent-’t-Jong and Woldorff, 2007).
In the present study, we found the SND component arising from ~ 300ms to 1000ms after the stimulus onset (Figs. 6–7). For the parafoveal stimuli, the SND were distributed centrally at the posterior occipital electrodes and moved more contralaterally as the locations of the stimuli were further away from the fovea. As illustrated in Fig. 8, we observed the higher bilateral SND amplitude (i.e., more negative) elicited by the parafoveal stimulation compared to the contralateral SND amplitude elicited by the more peripheral locations, resulting in a significant interaction between eccentricity and channel location (left and right posterior occipital electrodes) (F(6,162) = 18.72, p < 0.001). Post-hoc paired t-tests showed that the SND amplitudes for the foveal locations were significantly more negative than those in the more peripheral locations (t(27)’s = 2.24–3.36, p’s = 0.002–0.034, two-tailed, passing the Holm-Bonferroni-corrected threshold of 0.05). Additionally, the SND amplitudes in the far peripheral locations were significantly less negative than those in the near peripheral location (t(27) = -2.81, p = 0.009, two-tailed, passing the Holm-Bonferroni-corrected threshold of 0.0167) but the SND amplitudes in the parafoveal location did not differ from those in the far peripheral locations (t(27)’s = 0.49–1.61, p’s = 0.120–0.631, two-tailed). Together, these data suggested that the amplitude reduction of the SND component towards the peripheral direction occurred in a graded fashion.
Similar to the SSVEP results, we observed significant interactions between attention and eccentricity (F(6,162) = 2.94, p = 0.009) as well as between these two factors and channel location on the SND amplitudes (F(6,162) = 27.88, p < 0.001). Importantly, we found a significant three-way interaction between the level of eye movement, attention, and eccentricity, suggesting that the patterns of attentional modulations of the SND amplitudes across eccentricity were different between trials with the low and high levels of eye movements. Specifically, in trials with low levels of eye movements, we found the heterogeneous pattern of attentional modulations of the SND amplitudes similar to the SSVEP results (Figs. 8B and 8D). That is there was a significant reduction of attentional modulations of the SND amplitudes at the foveal compared to the parafoveal locations (t(27) = 2.45, p = 0.021, two-tailed) and the degree of attentional modulations of the SND amplitudes also decreased significantly in the peripheral compared to the parafoveal locations (near peripheral vs. parafoveal location: t(27) = -2.25, p = 0.033 ; far peripheral vs. parafoveal location: t(27) = -3.76, p < 0.001, two-tailed). In contrast, we found that the attentional modulations of the SND amplitudes were comparable across all eccentric locations in trials with high levels of eye movements (t(27)’s = 0.01–0.11, p’s = 0.911–0.990, two-tailed) (Figs. 8F and 8H).
Taken together, the SDN data suggested that the degrees of attentional modulations of the top-down attention signals from the frontoparietal regions also differed across eccentricities when eye movements were carefully controlled and the observed heterogeneity of the pattern of attentional modulations of the SDN signals depend on the level of eye movement.