Working memory (WM), the ability to temporarily store information, underpins cognitive tasks such as learning, mathematical computation, and comprehension. Factors such as aging, prolonged stress, and neuropsychiatric disorders can impair WM, posing challenges to everyday activities. Identifying the neural mechanisms that support WM may help to develop interventions to enhance WM performance. Electroencephalogram (EEG) studies have indicated associations between parieto-occipital alpha oscillations and the maintenance sub-stage of WM, yet it remains unclear whether parieto-occipital alpha oscillations causally affect WM maintenance. Transcranial alternating current stimulation (tACS), which can non-invasively manipulate neural oscillations via entrainment, enables investigations into the causal impact of neural oscillations on cognitive function in humans. However, conventional tACS is unsuitable for examining the causal impact of alpha oscillations on WM maintenance because it cannot modulate neural oscillations within the short duration of the maintenance sub-stage. This limitation is possibly due to phase differences (PDs) between tACS and neural oscillations influencing tACS effects. PDs are not typically considered in conventional tACS techniques.

Therefore, this dissertation aims to develop a real-time phase-locked tACS system capable of applying tACS with a predefined PD relative to ongoing neural oscillations and to examine the causal impact of parieto-occipital alpha oscillations on WM maintenance using this innovative tACS system. Specifically, the dissertation aims to address five research questions: (1) How can PDs between tACS and neural oscillations be manipulated? (2) How do PDs between tACS and neural oscillations influence the behavioral and electrophysiological effects of tACS? (3) Do parieto-occipital alpha oscillations causally impact WM maintenance (i.e., can parieto-occipital alpha tACS applied during the maintenance sub-stage modulate WM performance)? (4) If parieto-occipital alpha tACS modulates WM performance, do modulation effects exhibit frequency-specificity and spatial-specificity, or do they result from factors (i.e., general effects independent of stimulation frequency and indirect effects by stimulating peripheral nerves on the scalp or retina) that may confound causal inference? (5) How does cognitive demand influence the effectiveness of parieto-occipital alpha tACS in modulating WM performance?

In order to apply tACS with a pre-defined PD relative to neural oscillations, Study 1 developed a real-time phase-locked tACS system. This system employed EEG recordings to monitor endogenous neural oscillations in real time, extracted phases of the neural oscillations of interest, and controlled tACS application based on the phase information to achieve phase alignment between tACS and ongoing neural oscillations. To evaluate the accuracy of phase alignment between tACS and ongoing neural oscillations, sinusoidal waveforms generated from a signal generator were used to simulate neural oscillations. PDs between tACS and sinusoidal waveforms were calculated in both the inphase condition (PD expected to be 0°) and antiphase condition (PD expected to be 180°) at multiple tACS frequencies. Results indicated that average PDs between tACS and simulated neural oscillations across stimulation frequencies were 3.94° (SD: 7.14°) in the inphase condition and 182.57° (SD: 7.14°) in the antiphase condition. These findings validated the accuracy of the phase alignment of the real-time phase-locked tACS system.

Using this system, Study 2 investigated the influence of PDs between tACS and neural oscillations on tACS effects and the causal impact of parieto-occipital alpha oscillations on WM maintenance. The study applied both inphase and antiphase stimulation at the individual alpha frequency over the parieto-occipital cortex during the maintenance sub-stage of a modified Sternberg WM task. EEG recordings were conducted at intervals without the application of tACS. Results showed that, compared to inphase stimulation, antiphase stimulation resulted in significant decreases in behavioral performance, parietal alpha power, and frontoparietal alpha coupling. Furthermore, a significant positive correlation between behavioral effects and electrophysiological effects was observed in the inphase condition. These results demonstrated that alpha oscillations exerted a causal impact on WM maintenance and that PDs between tACS and neural oscillations influenced the direction of tACS effects.

Study 3 controlled for the potential general effects independent of stimulation frequency that could confound the chain of causation in Study 2. Study 3 applied both inphase and antiphase stimulation at the individual theta frequency over the parieto-occipital cortex to target parieto-occipital theta oscillations irrelevant to WM retention. No differences in WM performance or parietal theta oscillations were found between inphase and antiphase stimulation. These findings suggested that the tACS effects observed in Study 2 originated from stimulation effects specific to alpha frequency rather than general tACS effects independent of stimulation frequency, demonstrating the frequency-specificity of tACS effects.

Study 4 controlled for the potential indirect effects of peripheral stimulation or retinal stimulation that could confound the chain of causation in Study 2. Study 4 applied both inphase and antiphase stimulation at the individual alpha frequency over the vertex region. As the vertex region has been reported to be unrelated to WM maintenance, Study 4 excluded direct cortical stimulation of parieto-occipital neurons and preserved the potential indirect effects. No differences in WM performance or parietal alpha oscillations were found between inphase and antiphase stimulation. These results suggested that the tACS effects reported in Study 2 originated from direct cortical effects rather than indirect effects, demonstrating the spatial specificity of tACS effects.

Study 5 explored how cognitive demand influenced the effectiveness of tACS in modulating WM performance. The trials of the Sternberg task were divided into 5-consonant trials with lower cognitive demand and 7-consonant trials with higher cognitive demand. Participants received both inphase and antiphase stimulation at the individual alpha frequency over parieto-occipital cortices. In the 5-consonant trials, no differences in WM performance were found between inphase and antiphase stimulation. In the 7-consonant trials, significant differences in WM performance were observed between inphase and antiphase stimulation. These results suggested that parieto-occipital alpha tACS was more effective in modulating WM at higher cognitive demands.

In summary, this dissertation develops a real-time phase-locked tACS system with the capacity to precisely control PDs between tACS and endogenous neural oscillations and demonstrates the causal impact of parieto-occipital alpha oscillations on WM maintenance. Considering its capacity to modulate neural oscillations within a short duration, the real-time phase-locked tACS system undoubtedly holds wide-ranging implications for both basic scientific research and the development of clinical treatments.