Strong evidence of the ability of rTMS to induce frequency entrainment has been recently reported. Thut and colleagues [12] showed that rTMS over the posterior parietal cortex causes a local entrainment of the preferred frequency of the underlying generator (i.e., alpha), as revealed by a combination of TMS–EEG recordings. Furthermore, other rTMS studies (designed on a priori EEG knowledge of strict relationships between particular frequency bands and cognitive functions) have showed topographic- and frequency-specific effects of rTMS on the behavioural performances of participants [13], [14], [15], [16].

Even if the method remains controversial, the possibility of directly entraining oscillations in the brain has also been suggested for tACS [17], [18], [19], [20], [21], [22]. Recently rediscovered in the survey of cognitive neuroscience, tACS involves applying weak alternating electrical currents to the head via two electrodes, which are usually both located on the scalp. Kanai and colleagues [17], delivered tACS over the occipital cortex, which induced visual experiences (i.e., phosphenes) that were ascribed to the direct interaction between tACS and the on-going oscillations of the primary visual cortex. In particular, beta frequency stimulation was more effective in inducing phosphenes in an illuminated room, whereas alpha frequency stimulation was more effective in darkness. Researchers have argued, however, that tACS-induced phosphenes may actually be the result of activation of the retina [23] instead of the primary visual cortex. Indeed, electrical current could reach the orbital area and retina via volume conduction of the scalp irrespective of electrode configurations on the head [24]. Evidence consistent with this interpretation was already reported in the last century, when Brindley [25] observed that the intensity needed to elicit phosphenes increased with the distance from the eye and that they cannot be elicited after pressure blinding. What deserves attention, however, is not merely concern about the retinal versus cortical origin of tACS-induced phosphenes, but the possibility of tACS directly stimulating the cortex (but see [26]). The controversy is not new. Lippold and Redfearn [27] observed that, to obtain “psychological effects at imperceptible current strengths”, the current flow needed to enter the orbital fissure. In addition, Smitt and Wegener [28] and Hayes [29] recorded intracerebral voltage measurements in human cadavers and alive monkeys, respectively, and found evidence for a general diffusion of current through the brain. Widespread changes in neuronal activities of cortical and subcortical regions after electrical stimulation have recently been reported with modern neuroimaging techniques [30], [31]. Therefore, if tACS is effective for cortical stimulation, its effects appear to be generalised to all of the brain, instead of localised to a single cortical area.

Even if controversial, investigating the ability of tACS to induce direct cortical modulations of the natural brain oscillations remains an important issue because transcranial electric stimulation (tES) offers some advantages over TMS in terms of safety. Additionally, tES has a better translation in the clinical setting because the method is more feasible [32], [33]. To this end, the current study was designed to exploit the well-established relation between posterior alpha-oscillations (8–14 Hz) and visual attention/perception [15], [34], [35], [36], [37]. A role in regulating the incoming information at early stages of processing has been assigned to the posterior alpha rhythm through the functional inhibition of task-irrelevant regions [38], [39], [40]. Specifically, posterior alpha power before stimulus presentation is inversely related to the quality of perception [15], [41], [42], [43], [44] and retinotopically organised in accordance with the focus of attention [34], [45], [46].

The spreading of the effect across hemispheres could be ascribed to the poor spatial resolution of tES, and methodological accounts have to be considered. When using rectangular-pad electrode configurations, focality is considered to be limited, and the induced effects might also be able to modulate cortical areas adjacent to the target site [50], [58], [59], [60]. Even if we properly used an electrode with reduced size for the stimulation of the occipito-parietal areas and a larger reference electrode to improve spatial focality, as suggested by Nitsche and colleagues [50] (but see [61]), we could not eliminate this possibility. The use of a cephalic reference electrode is another factor implicated in spatial focality because the relative position of both electrodes is determinant for the induced current profile in the brain [62]. However, we placed the reference electrode over the vertex, as is conventional in most previous studies on visual perception [17], [51]. In addition, the stimulation intensity adopted in this study (1 mA) could have been too low to reach the cortical target and induce a clear focal effect. However, we had to consider the possibility of inducing phosphenes by increasing the stimulation intensity, as reported by previous studies [17], [20]. The occurrence of tACS effects over both visual fields could also be ascribed to the duration of the stimulation: in each block participants received tACS for five consecutive minutes. During this period of time, current could have first directly reached the target area and then spread to the contralateral hemisphere through cortico-thalamic and/or cortico-cortical connections. Evidence for a similar mechanism has already been reported in the motor domain, where electrical stimulation was able to directly alter the excitability of the stimulated region, and indirectly, the excitability of the homologous region of the opposite hemisphere [63]. However, in this case, we would have expected the opposite effect over the homologous areas, enhancement performance in the visual field ipsilateral to the stimulation site, which is consistent with the push–pull effect previously reported regarding the posterior alpha rhythm [15], [64].

Frequency-specificity was only marginally confirmed. In the present study, precise hypotheses were formulated according to knowledge relative to the active inhibitory role of the posterior alpha rhythm during on-going visual processing [38], [39], [40]. Theta-frequency stimulation was added to the experimental design as an additional control condition, and the stimulation was expected to be ineffective [15]. Instead, participants receiving tACS in the theta frequency showed performances comparable to those of the alpha group. Although we did not intend to investigate memory and learning, the sequential design of the study (with fixed order of the baseline and tACS sessions) might explain the involvement of the theta frequency. Theta band activity, indeed, has been closely associated with memory and learning [22], [65], [66], [67], [68], [69], as well as synaptic plasticity [70], [71]. There is also evidence linking theta oscillations to other cognitive functions, such as attention [72], [73], [74] and sensorimotor integration [75], [76], [77]. Because theta band oscillations reflect long-range communication between distant brain areas [5], these oscillations have been suggested to coordinate sensory and motor brain regions when the task requires updating a motor plan on the basis of incoming sensory information. To this regard, the task performed in the present study could involve changes in theta band activity, as participants were required to have two subsequent responses according to the features of the target stimulus. Functional involvement of the theta-band activity is therefore plausible considering both learning effects and the alternating responses. Most of the previous studies, however, found a positive correlation between theta activity and task performance: increase in theta band power was observed during the encoding and retrieval of successfully remembered items [65], [78] and when the memory load was systematically increased [66], [79]. Thus, after a direct entrainment of theta by tACS, an improvement of the visual performance should be expected, instead of a worsening of target detection as in our data. Based on all of these considerations, the lack of retinotopical specificity and frequency specificity suggest that the present results may not be ascribed to the direct entrainment of brain oscillations induced by tACS. However, these data cannot prove that tACS is unable to manipulate EEG oscillations, and this uncertainty represents a study limitation.

Another point that must be discussed is that all the tACS effects observed in the present study were related to the detection, but not to the discrimination response. The difficulty level was different between the two tasks: detecting the luminance change induced by the brief appearance of the target was easier than judging its orientation. Therefore, because performance was at chance level at the lowest contrast levels in the sham condition, the second response might be less sensitive to slightly worse performance induced by tACS. However, no tACS effect on the discrimination response emerged when we focused the analysis on the three highest contrast levels, in which performance was greater than the chance level. We also analysed the discrimination responses considering only those trials in which participants had correctly detected the Gabor patch for R1 Accuracy, but no tACS effect was observed. One could argue that tACS, as applied here according to the chosen montage, actually stimulated the dorsal visual stream, where the source of alpha activity was localised [12], [80]. The dorsal visual stream is the projection conveying the signal by the magnocellular system [81], which is particularly sensitive to stimuli at low-contrast and low-spatial frequency, like those used in the present study [82], [83], [84]. Electrical stimulation of the dorsal stream could have impaired target detection because the dorsal stream responds well to rapid changes in luminance contrast. On the contrary, the fine discrimination of visual features, such as the target orientation, is also under the control of the ventral visual stream, which was not affected by the stimulation leaving the discrimination task unchanged. Nevertheless, extensive literature shows that the attention bias associated with the parietal alpha modulation affects both detection [15], [41], [46] and discrimination [34], [45] tasks. Accordingly, an effective alpha entrainment over parietal regions should have induced the same effect on both responses.

A key point worth considering in the discussion of these results is the stimulation timing relative to the phase of the on-going oscillatory activity. Brain oscillations are not only characterised by power and frequency but also by their instantaneous phase. There is compelling evidence that phase dynamics reflect cyclic fluctuations of neural excitability and play a relevant functional role in cognitive processes [3], [85], [86], [87], [88]. Schyns and colleagues [89] have recently demonstrated that phase codes considerably more information than power during an emotion categorisation task. Moreover, an increasing number of studies show that processing of visual information is strongly dependent on the phase of the spontaneous EEG oscillations, such that a stimulus appearing at the optimal phase would be optimally registered and perceived, while at the opposite phase, the stimulus might be entirely missed [90], [91], [92], [93]. Thus, studies aiming to modulate participants' behaviours through the induction of an exogenous entrainment of brain rhythms should also take into consideration the temporal dynamics of phase of the underling brain oscillations and accordingly trigger tACS application. A similar approach has been recently followed by Neuling and colleagues [94], who applied oscillatory transcranial direct current stimulation at 10 Hz while subjects were performing an auditory detection task. Importantly, they presented the stimuli in specific phase bins relative to the electrical stimulation and found specific behavioural consequences dependent on the phase of the entrained oscillation. In the present study, tACS was simply applied for five consecutive minutes without a finer synchronisation with visual stimuli, and this application could have minimised the results. Another important aspect to take into account is the individual peaks of oscillations in a particular frequency-band. We have observed that alpha frequency shows a great inter-individual variability, and the frequency also changes across life-span [95]. In our study, we did not individually select the stimulation frequency; this point is a limit of the study.

On the whole, combining all aspects discussed above with a consistent interpretation of the ability of tACS to induce a direct entrainment of cortical oscillations is quite difficult because our data did not show conclusive proof or disclaimer of the point. Perhaps the most parsimonious explanation could be found by considering the possibility that the current flow spreads through the retina. Even if tACS did not induce a conscious perception of visual phosphenes in the present study, its action, which is subthreshold by definition due to the low stimulation intensity, could still affect the functioning of retinal cells. In this regard, we consider that contrast detection starts from the retina while orientation discrimination is a cortical process that occurs at the level of the primary visual cortex (i.e., V1). Indeed, research has established that different aspects of a visual scene are processed by separate parallel pathways, which run from the ganglion cells of the retina to the V1, passing through the lateral geniculate nucleus [96], [97]. The magnocellular and parvocellular cell systems differ significantly in their anatomical and physiological properties [84]. In particular, they differ with respect to contrast gain: the former systems are much more sensitive (8–10 times) to luminance contrast than the latter [82]. Thus, the magnocellular system is well suited to handle detection of rapid changes of low luminance stimuli, and this system is already in the retina. The recognition of stimuli orientation, instead, occurs in the V1 cortex, where the information is carried by both the magnocellular and parvocellular systems [96]. The results of the present experiment could be explained if tACS at low frequencies (6 Hz and 10 Hz) was able to selectively interfere with the magnocellular but not the parvocellular cells of the retina. In this case, detection would be impaired, while discrimination could be supported at the V1 level by the parvocellular system. Studies on the primate retinal ganglion cells showed that cells in the magnocellular system actually have temporal-frequency response characteristics distinct from cells in the parvocellular system [98], [99] and they peak at approximately 10 Hz [100]. Although intriguing, this theory is only a speculative explanation that needs to be investigated in further studies.

Another concern with tES in general (not only tACS) regards the way in which current flows through the brain. While the effects induced by electrical stimulation directly applied to the cortical tissue are well established [101], [102], [103], [104], [105], [106], [107], the same is not true when stimulation is applied transcranially over the scalp. During any tES modality, the current that reaches the cortex is strongly influenced by anatomical factors because of the different electrical conductivities of the intermediate tissues, such as the scalp, skull, cerebrospinal fluid and brain [59], [108]. Moreover, because the impedance of the skull is higher relative to that of the scalp, most of the current is shunted across the scalp [60]. Evidence provided by imaging and modelling studies [30], [31], [59], [109] as well as clinical studies [110], [111], [112] suggests a widespread modulation of multiple cortical and sub-cortical regions, independently of their anatomical connections. Considering these aspects, the range of possible interpretations for the data of the present study becomes wider and more elaborate.