Conveying Temporal Information to the Auditory System via Transcranial Current Stimulation

This review paper investigates whether non-invasive application of electric current to the human scalp can be utilized to convey perceptually relevant temporal information to the auditory system. Recent studies have corroborated this notion by demonstrating that transcranial current stimulation (TCS) with temporally structured (sinusoidal and/or sound envelope-shaped) current biases neural processing and auditory perception toward the temporal pattern of the applied current. However, the perceptual benefits achieved with TCS so far are fairly modest. In sum, the temporally specific modulatory ability of TCS makes it a useful scientific tool for identifying temporal mechanisms for auditory perception. Practical or clinical applications (e.g., to enhance or restore auditory functions in normal or hearing-impaired populations) are currently still premature and require further optimization of stimulation parameters.


Introduction
TCS is an on-invasive and soundless brain stimulation technique that is becoming increasingly popular in auditory research, as reflected by ar apidly growing body of published studies overt he last years [1,2]. TCS involves the non-invasive and harmless application of lowintensity current (typically within ±2mA) via scalp electrodes. Current-flows imulations and intracranial recordings in humans [3,4] have shown that the induced current spreads mainly along the highly conductive skin and to as maller amount into the cranium, where it propagates widely across the brain. Extracellular electric fields applied to pyramidal neurons in vitro almost instantaneously elevate (orreduce, depending on the relative field orientation)s pontaneous neural firing (e.g., [5]). Thus, intracranial stimulation with temporally structured (alternating or complex-shaped)current can entrain neural excitability to the temporal pattern of that current [6]. Application of the current at the scalp (TCS)isthought to induce corresponding excitability changes at the neural population level [7,8]. Neural oscillations in neocortexa lign their highexcitability phases to expected relevant sensory events, thereby providing am echanism for temporal filtering of sensory input [9]. If TCS with temporally-structured cur-Received28February 2018, accepted 17 July 2018. rent wasalso able to control the timing of neural excitability,t hen this technique could be utilized (ina nalogy to neural oscillations)t oprovide the brain with temporal information that is critical for extracting specificevents from the sensory environment. The presumed ability of TCS to entrain neural oscillations [10] would be evidenced by demonstrating that neural excitability-and therewith the neural processing of sensory stimuli-covaries with the temporal pattern of TCS. Similarly,the presumed ability of TCS to convey perceptually relevant temporal information would require showing that perception of sensory stimuli covaries with the temporal pattern of TCS. Such evidence would have interesting consequences for both scientificresearch and clinical/practical application: It would confirm that TCS can effectively manipulate specifictemporal activity patterns (including neural oscillations), which would enable to experimentally identify causal roles of these patterns in auditory functions, such as hearing, auditory scene analysis, and speech comprehension. Moreover, it would suggest that TCS can be utilized to enhance or restore these functions in normal and hearing-impaired populations [11].
The goal of this reviewpaper is to assess whether TCS can be utilized to provide the human auditory system with perceptually relevant temporal information. The focus is on auditory studies that have applied TCS with temporally structured currents. Auditory studies using TCS variants that seem less suited for conveying temporal information (direct currents, random noise currents, or alternating cur-

TCS can entrain neural oscillations
Electric stimulation with alternating current can entrain spontaneous neural firing (measured intracranially)inanimal neocortex, even when the alternating current is applied transcranially (TACS) [12]. Studies attempting to translate these animal findings to humans found that TACS elevates power at the TACS frequencyi ns ubsequent scalp magneto-/electroencephalographic recordings (M/EEG) [ 13]. However, these aftereffects more likely reflect TACS-induced synaptic plasticity than oscillatory phase entrainment [14]. More convincing evidence of TACS-induced entrainment would showt hat oscillations followt he applied current waveform during the stimulation. Using sophisticated approaches to reduce TACSinduced artefacts in simultaneous M/EEG recordings (e.g., [15,16]), several studies have indeed shown an online increase in both power at the TACS frequency [17,18] and phase-locking between TACS waveform and M/EEG oscillation [16,17,19]. While the minimum current intensity required to reliably entrain human neocortical oscillations is still being debated [3,4,20,21], the aforementioned studies strongly suggest that TACS can entrain neural oscillations and thereby convey temporal information to sensory systems. In future studies, it needs to be shown whether TCS with more complex-shaped (aperiodically fluctuating)c urrents resembling naturally occurring sensory signals (e.g., speech)c an align endogenous neural activity as well [6].

TCS can modulate auditory perception in at emporally specific manner
Several studies have shown that TCS applied above temporal cortexc an modulate auditory perception in at emporally specificmanner.Neuling et al. [22] applied direct current modulated at 10 Hz. Theyreported that the threshold for detecting tone pips in asimultaneous noise masker depends on the relative phase of the simultaneously applied 10-Hz electric stimulation. The size of this phaseeffect on simultaneous-masked threshold was ∼0.3 dB. Using as imilar paradigm, Riecke et al. [23] found that detection of 4-Hz click trains in silence depends on the relative phase of the click train and simultaneously applied 4-Hz TACS (size of phase effect on click-detection accuracy: ∼2%). Consistent with these click-detection results, Riecke et al. [24] found that the perceptual buildup of 4Hzmodulated complext ones in informational maskers (i.e., the time required for listeners to perceptually segregate a rhythmic target stream from background noise)d epends on the relative phase of the target stream and 4-Hz TACS (size of phase effect on buildup time: ∼40 ms). Because the applied current carried temporal cues regarding the occurrence of the target stream constituents, this observation supports the notion that TCS can provide perceptually relevant temporal information to the auditory system.
Temporal information might be of particular importance for speech processing, as ar apid sequence of densely structured information has to be precisely deciphered. The amplitude envelope of the speech signal is critical for this, as its disruption strongly impairs speech comprehension (e.g. [25]). Indeed, several recent studies showt hat speech processing can be manipulated by applying external stimulation carrying speech-envelope information. E.g., Riecke et al. [26] investigated speech comprehension under 4-Hz TACS in ac ocktail party-liket wo-talker situation. Speech envelopes were artificially fixed at 4 Hz and enhanced, and the twos imultaneous speech signals were mixed so that their envelopes alternated. Consistent with results on click-and tone-detection, speechrecognition accuracyw as found to depend on the relative phase of the target-speech envelope and TACS (size of phase-effect on speech-recognition accuracy: ∼3%). Zoefel et al. [27] combined TACS at 3.125 Hz with functional magnetic resonance imaging (FMRI)w hile participants listened to rhythmic sentences (spoken at the TACS frequency) that were modified with vocoders to resemble either intelligible speech or only broadband noise fluctuating at the speech rhythm. Speech-evokedFMRI responses in superior temporal gyrus were found to depend on the relative phase of the auditory speech rhythm and TACS. Strikingly,t his effect waso nly observed for intelligible speech, butn ot fluctuating noise. Moreover, certain relative phases resulted in suppression of speech-evokedr esponses (compared with sham stimulation). Importantly, these results provide apotential neural mechanism underlying the aforementioned TACS-phase effects on speech recognition [26].
Twor ecent studies also applied TCS at different time lags relative to speech sounds; however, current waveforms matched the envelope of the natural speech stimuli. Wilsch et al. [28] observed that speech-in-noise recognition at listeners' individual "best" time lag (the lag revealing the smallest threshold)i sb etter than under sham stimulation, with at hreshold difference between individual best lag and sham of ∼0.4dB. In as econd experiment, Riecke et al. [26] removedthe low-frequencyenvelope from auditory speech stimuli and applied it via TCS to restore this "aurally missing" temporal information in the brain. Speech comprehension wasfound to depend on the lag between envelope-reduced speech and envelopeshaped TCS. The maximal benefitw as observed when TCS led the acoustic input by 375ms, with ad i ff erence in word-recognition accuracyo f∼ 4% (average best lag vs. average worst lag). In sum, these studies showconsistently that TCS applied above temporal cortexcan convey perceptually relevant temporal information to the auditory system.

TACS frequency modulates perceptual sampling frequency
It is conceivable that TACS transmits temporal information via not only its relative phase buta lso its frequency -i nf orm of as ampling rhythm that affects the temporal resolution of perception. Based on the observation that gap-detection thresholds correlate negatively with listeners' gamma resonance frequencyi na uditory cortex( defined as the modulation frequencyw ithin 20-70 Hz that evokes the strongest auditory steady-state EEG response), Baltus et al. used TACS near individual gamma-resonance frequencytomanipulate the temporal resolution of the auditory system [29,30]. Theyfound that TACS with afrequencyslightly above (vs. below) resonance frequencyreduced gap-detection threshold. This indicates that TACS can modulate the rate at which the human auditory system samples acoustic input.

Clinical utility of TCS and current limitations
There currently exists only little evidence to support an effective clinical/practical utilization of rhythmic auditory TCS. TCS effects observed so fara re rather modest (see section 3) and their actual size and direction (benefitv s disruption)are difficult to estimate. E.g., most studies correct for inter-individual differences in the best time lag (presumed to originate from anatomical variations [19]) to improve the power of group-levela nalyses. However, this approach requires sacrificing the listeners' maximumperformance data, which inevitably results in an underestimation of both average effect sizes and potential benefits. Indeed, beyond the effects of TCS timing, only few studies could validly detect significant differences between TCS and sham stimulation. Thus, it remains to be shown whether TCS can reliably provide sufficiently strong auditory perceptual benefits, as required for aclinical or practical application. Potential benefits may be strengthened by systematically improving TCS parameters to promote the specific spatiotemporal brain-activity patterns that underlie normal auditory functions [31]. While the TCS studies reviewed here have primarily considered basic temporal parameters, future research may focus on complementary (spatial and spectral)p arameters, e.g., by comparing the benefits of different electrode shapes, locations, and numbers. Clinical TCS utilization also requires more basic research on the mechanisms underlying the observed TCS effects, which are still being debated [20,21,32]. While it seems safe to conclude that TACS with sufficient intensity can entrain endogenous neural activity and modulate auditory perception (sections 2and 3),itisstill unclear howthis entrainment originates: directly in the cortex, indirectly via rhythmic peripheral responses elicited by the current, or both. An exclusive mediating role of peripheral responses seems unlikely because TACS applied above the mastoids does not entrain otoacoustic emissions [33] and TACSphase effects on auditory cortical speech processing have been observed exclusively for auditory stimuli that can be identified as speech [27]. Nevertheless, even though the TCS studies reviewed here attempted to reduce potential tactile or visual sensations [34], the applied currents might have induced subliminal tactile or visual temporal cues that potentially facilitated (ord isrupted)n eural activity. Future studies need to rule out this possibility by applying TCS also to non-auditory control locations (e.g., near the eyes or on the hand).

Conclusions
Our reviews hows that the timing of TCS modulates endogenous neural activity,auditory cortical speech processing, and various perceptual auditory functions (hearing in silence or noise, auditory streaming, and speech comprehension). This converging evidence indicates that TCS can convey temporal information to the auditory system (by entraining slowendogenous neural activity)and therewith alter processing and perception of acoustic input. This makes TCS ah ighly useful scientifict ool for identifying causal dynamic mechanisms underlying perceptual auditory functions. However, perceptual benefits achievedwith current stimulation settings are still modest, indicating that anyclinical or practical utilization of TCS (e.g., in assisted hearing)still requires substantial methodological improvement.