Is Alpha Asymmetry a Byproduct or Cause of Spatial Attention? New Evidence Alpha Neurofeedback Controls Measures of Spatial Attention

Abstract

Cued spatial attention differentially modulates alpha power in attended relative to non-attended brain representations, termed the alpha asymmetry. Yet a causal role for alpha in attention is debated. In this issue of Neuron, Bagherzadeh et al., (2019) utilize neurofeedback to train alpha asymmetry and causally impact measures of spatial attention.

Hans Berger, the father of modern human electrophysiology, began studying the macroscale electrical dynamics of the brain because he thought it was key to understanding information transfer via telepathic communication. In doing so, he discovered “Berger’s Wave,” or the alpha rhythm (8–12 Hz), which remains one of the most robust signatures of brain activity measured today. Decades of follow-up studies debunked Berger’s initial ideas, rather proposing that alpha simply reflects a passive idling state of the brain because it is most prominent during rest with eyes closed. Since then, it has been widely demonstrated that alpha rhythms are actively and focally modulated with attention, indicating greater functional relevance. However, a causal role for alpha still remains debated a century later, not for telepathic transfer of information between people, but for optimal processing of information within one’s own body.

Much of the research on alpha and attention has focused on the visual system, where decreased alpha power correlates with enhanced processing of attended stimuli in the contralateral hemifield, and increased power is correspondingly associated with diminished processing of unattended ipsilateral stimuli (for review, see Jensen and Mazaheri, 2010); this highly replicated finding is called the alpha band laterality, or alpha asymmetry. Similar phenomena have also been observed in the somatosensory (e.g., Jones et al., 2010) and auditory (e.g., Müller and Weisz, 2012) domains, indicating a generalizable effect on sensory processing.

Although it is well established that attention correlates with shifts in alpha power, it is difficult to unequivocally demonstrate if and how alpha causally impacts information processing (Figure 1A). In humans, non-invasive brain stimulation aimed at artificially entraining alpha provides support that driven oscillations can suppress behavior, though the efficacy of such methods is debated (Lakatos et al., 2019). Alternatively, neurofeedback training provides a means to endogenously modulate activity in a more natural manner. Prior studies have shown that neurofeedback training based on rhythmic asymmetry can yield changes in motor action (Kajal et al., 2017) and sensory processing (Okazaki et al., 2015). Okazaki et al. (2015) suggested that attention-induced changes in alpha were causal to observed improvements in visual detection. However, it was not clear if alpha asymmetry was the cause of improvement or a byproduct of attention. In this issue of Neuron, Bagherzadeh et al. (2019) test for a causal role of alpha asymmetry directly on attention by designing a similar neurofeedback task without spatial cues and assessing sustained effects on neurophysiological and behavioral correlates of spatial attention.

Figure 1. Neurofeedback Training to Assess Whether Alpha Asymmetry Causally Controls Spatial Attention.

(A–C) (A) It is well-established that cued spatial attention drives alpha asymmetry in cortex, yet if and how alpha is causal to shifts in attention is debated. To address this debate, Bagherzadeh et al., 2019 tested whether alpha asymmetry neurofeedback training in the absence of cued spatial attention can induce changes in neurophysiological and behavior measures of visual spatial attention. Alpha asymmetry training enhanced (B) lateralized visual processing in the online training period, and (C) sustained changes in behavioral measures of spatial attention, including reaction times and fixation bias in the post-training period. These findings support a causal role of alpha asymmetry on spatial attention.

During neurofeedback training, participants fixated on a grating stimulus in the middle of a screen and were instructed to use “mental effort” to enhance its visibility. The contrast of the grating was controlled by an alpha asymmetry index (AAI), calculated as the normalized difference in parietal alpha power between hemispheres. Two separate groups of participants were trained to increase alpha power in the left relative to right parietal cortex or vice versa, in order to make the grating stimulus visible for a subsequent match to sample task (Figure 1B).

Over the course of training, all participants successfully learned to modulate alpha asymmetry in the trained direction. Although participants’ strategies differed, with some increasing ipsilateral alpha while others reduced contralateral alpha, changes in left parietal cortex were consistently the primary driver of effects. Notably, gamma (30–60 Hz) lateralization was also observed in the direction opposite alpha asymmetry (Figure 1B).

An important distinction between this and prior studies is that all aspects of spatial attention were removed from the task during neurofeedback training. Alpha-driven changes in spatially specific visual processing were assessed by measuring evoked responses to an un-cued probe stimulus (task-irrelevant gray dot) that flashed in the left or right hemifield in a random subset of trials (Figure 1B). Neurofeedback training led to weaker probe-evoked responses in the hemisphere trained to have higher alpha power, suggesting that alpha asymmetry directly impacted visual processing.

To test if alpha asymmetry neurofeedback yields sustained effects, one of two additional tasks was used to compare neurophysiological and behavioral measures of spatial attention during the pre- versus post-training periods (Figure 1C). The majority of participants (14/20) took part in a classic Posner attention paradigm (Figure 1C, left), in which they were instructed to fixate on the center of a screen, and then cued to covertly attend with a left, right, or neutral stimulus. A visual grating stimulus was then presented on either side of the screen, and subjects indicated that it was in one of two orientations in a 2-alternative forced choice task. Attentional modulation indexes (AMIs), calculated as the normalized difference in alpha power during the left- minus right-cue periods, were positive in left and negative in right parietal regions for all participants, reflecting alpha asymmetry as expected. However, pre- versus post-training changes in AMIs were found only in the trained hemisphere, suggesting the training selectively enhanced lateralized alpha modulation, which was used as a neurophysiological measure of spatial attention.

Changes in alpha power were also observed during the post-cue delay period in “neutral” cue trials that lacked explicit instruction for covert attentional allocation. Neutral cue-evoked alpha power decreased more in the post- than pre-training periods, specifically in regions contralateral to the trained hemisphere. Correspondingly, reaction times to visual targets in the hemifield ipsilateral to the trained hemisphere were faster after neurofeedback training, but again only on neutral cue trials. Both measures are proposed as further evidence of sustained changes in spatial attention without spatial cues.

In the remaining 6/20 subjects (Figure 1C, right), eye tracking during a free-viewing task was used to test for a fixation bias training effect. For all but one participant, a bias in horizontal eye fixation was observed after training in the hemifield ipsilateral to the trained hemisphere, providing further evidence of sustained changes in spatial attention without a spatial cue.

These findings clearly demonstrate that alpha asymmetry is a trainable brain dynamic that causes plastic changes in visual information processing and motor behaviors associated with spatial attention, but are they sufficient to show that alpha causally controls spatial attention? Several open questions suggest further research is needed to definitively establish a causal link. For example, alpha asymmetry induced by neurofeedback training was accompanied by a lateralization of gamma (~30–60 Hz) power. Gamma’s potential influence on measures of attention was not assessed, and it is possible that alpha asymmetry could be a non-functional consequence of gamma modulation. Additionally, in the Posner task, enhanced alpha lateralization was used as a measure of spatial attention, in order to show that alpha controls spatial attention, creating a circular argument in which true causality is impossible to disentangle. It also remains unclear whether participants were indeed relying on alpha asymmetry feedback signals without an explicit attention-based strategy. Although there was a significant difference in fixation bias between the Posner task, which employed cued covert attention, and neurofeedback training without cued attention, this single measure of eye movement cannot rule out the possibility that participants utilized explicit attentional strategies.

Finally, the observation that participants increased alpha asymmetry in different ways during training raises the question of if and how inconsistent neuromodulation strategies can lead to a solitary mechanism of causal control. Although the suppressive effects of alpha have been linked to GABA-mediated inhibition (Jensen and Mazaheri, 2010), consensus regarding a mechanism for the generation of the alpha rhythm is not yet agreed upon. Before we can tackle this challenge, we first need to appreciate the various signal features that could underlie changes in alpha power. Brain rhythms, including alpha, can emerge as transient high power “bursts” of activity, as opposed to ongoing oscillations (e.g., Jones et al., 2009; Ossadtchi et al., 2017). As such, increased signal features other than amplitude, e.g., burst probability, duration, or frequency span, can also manifest as high averaged power (Shin et al., 2017). Alpha-based neurofeedback training has been shown to modify the rate, but not the duration or amplitude, of alpha frequency transients (Ossadtchi et al., 2017). The neural mechanisms underlying changes in rate are likely different than that of duration or amplitude, and how these signal features relate to measures of attention like reaction time and fixation bias is unknown.

Ultimately, Bagherzadeh et al. (2019) get us one step closer to defining the causal role of alpha asymmetry with regard to spatial attention. However, revealing the specific neural mechanisms underlying this relationship is the key to establishing true causality. Further work is needed for a more refined understanding of the signal properties underlying power modulation. From there, a link to cellular and network-level mechanisms can be made using animal models and/or computational neural modeling. In sum, Berger’s legacy carries on as we continue to investigate the functional role of his infamous alpha wave.