We report three studies of the effects of anodal transcranial direct current stimulation (tDCS) over auditory cortex on audition in humans. Experiment 1 examined whether tDCS enhances rapid frequency discrimination learning. Human subjects were trained on a frequency discrimination task for 2 days with anodal tDCS applied during the first day with the second day used to assess effects of stimulation on retention. This revealed that tDCS did not affect learning but did degrade frequency discrimination during both days. Follow-up testing 2–3 months after stimulation showed no long-term effects. Following the unexpected results, two additional experiments examined the effects of tDCS on the underlying mechanisms of frequency discrimination, place and temporal coding. Place coding underlies frequency selectivity and was measured using psychophysical tuning curves with broader curves indicating poorer frequency selectivity. Temporal coding is determined by measuring the ability to discriminate sounds with different fine temporal structure. We found that tDCS does not broaden frequency selectivity but instead degraded the ability to discriminate tones with different fine temporal structure. The overall results suggest anodal tDCS applied over auditory cortex degrades frequency discrimination by affecting temporal, but not place, coding mechanisms.
Perceptual learning is the improvement in perceptual performance with training not due to familiarity with the task. In audition, learning is specific to the trained stimulus and does not generalize widely to tasks with the same procedure but different stimuli (Wright et al., 1997; Wright & Fitzgerald, 2005). In particular, frequency discrimination learning is specific to the trained frequencies (Demany, 1985; Irvine et al., 2000; Demany & Semal, 2002), with evidence showing that the rapid improvements within the first hour of training are due to genuine perceptual learning rather than procedural familiarity (Hawkey et al., 2004; Ortiz & Wright, 2009). This is consistent with neurophysiological evidence showing frequency discrimination learning is associated with frequency-specific plastic changes in tonotopic maps (Recanzone et al., 1993; Menning et al., 2000; Jäncke et al., 2001; Polley et al., 2006). In humans, rapid auditory learning is associated with neuroplastic changes in auditory cortex (Alain et al., 2007, 2010). These changes underlie learning as greater long-term potentiation-mediated neuroplastic change is associated with increased learning (Rutkowski & Weinberger, 2005) and is independent of task familiarity (Reed et al., 2011).
Transcranial direct current stimulation (tDCS) is a relatively new non-invasive technique for manipulating cortical excitability in humans. Anodal tDCS increases local cortical excitability (Miranda et al., 2006) by decreasing membrane potential of stimulated neurons (Nitsche et al., 2003a; Stagg & Nitsche, 2011), as shown by changes in corticospinal excitability and cortical hemodynamic response (Nitsche & Paulus, 2000, 2001; Lang et al., 2004; Zaehle et al., 2011). This enhances the acquisition and retention of motor learning when applied over motor cortex (Nitsche et al., 2003b; Antal et al., 2004a; Reis et al., 2009), probably by increasing the associated long-term potentiation which underlies cortical plasticity (Stagg & Nitsche, 2011). These effects have been extended by research showing that anodal tDCS over Wernicke's or Broca's areas can enhance artificial language learning (Floel et al., 2008; Vries et al., 2010). Only one study has so far examined the effects of tDCS on perceptual learning, finding only a transient enhancement of visual learning by anodal stimulation before the effect dissipated with more training (Fertonani et al., 2011).
Our initial aim was to determine whether auditory perceptual learning, like motor learning, is enhanced by increasing excitability of its primary cortical representation, as it is underpinned by similar use-dependent plasticity, which anodal tDCS is thought to enhance. We instead found that anodal tDCS did not enhance frequency discrimination learning but unexpectedly degraded frequency discrimination. We conducted two further experiments examining the effects of anodal tDCS on the place and temporal coding processes that underlie auditory frequency discrimination to determine the cause of this degradation. Cumulative evidence suggests the auditory system uses duplex coding, with temporal processes dominant at lower frequencies (Moore, 1973; Moore & Glasberg, 1989) and place processes dominant at higher frequencies (Johnson, 1980). We took psychophysical measurements of place and temporal coding and showed that the degradation of frequency discrimination was probably due to interference with temporal coding.
Materials and methods
Twenty adult volunteers (14 females) aged between 18 and 27 years (median = 23 years), all of whom reported normal hearing, participated in the three experiments. This sample size is consistent with previous psychophysical studies (Demany & Semal, 2002; Mathys et al., 2010). Fifteen subjects completed the frequency discrimination learning task reported in Experiment 1. As learning was being examined, subjects with extensive psychoacoustic experience or with more than 1 year's musical experience were not recruited. Seven subjects (four of whom participated in Experiment 1) were recruited for Experiment 2A. Six subjects (two of whom competed both Experiments 1 and 2A and two who completed only Experiment 2A) participated in Experiment 2B. Author M.F.T. completed Experiments 2A and 2B but was blind to stimulation condition with the procedure being performed by another experimenter and was not informed of stimulation condition until after testing was completed. All other subjects were blind to the stimulation condition and Naïve to the experimental aims. No formal audiometric assessment was performed; instead stimulus levels were tailored to each subject's sensitivity. In all experiments, subjects were not recruited if they showed any neurological, psychiatric or medical condition that contraindicates the use of tDCS (Nitsche et al., 2008), as assessed using a checklist completed prior to testing. The protocol was in accordance with the Declaration of Helsinki and was approved by the Human Research Ethics Committee at The University of Western Australia. All subjects provided informed written consent prior to testing. No subjects reported adverse effects to the tDCS procedure, other than the reddening of skin under the electrode, and none withdrew from the study.
All testing took place in a sound-attenuated room. The acoustic stimuli were generated with a Creative SoundBlaster Live! Soundcard in Experiment 1 and with an ASUS Xonar Essence ST soundcard in Experiments 2A and 2B. Stimuli were presented monotically to the left ear by Sennheiser 280 Pro headphones.
The same procedure was used for all reported experiments, with anodal tDCS being delivered by a constant-current battery-driven stimulator (Dupel Iontophoresis System, MN) through two 6 × 4 cm electrodes in saline-soaked pouches placed on the scalp. The anode was placed 1 cm inferior to the midpoint of C4 and T4 in the International 10-20 system, corresponding to the right auditory cortex (Mathys et al., 2010) and the cathode was placed on the contralateral supraorbital region. This electrode montage has been shown to increase excitability in auditory cortex (Zaehle et al., 2011). Right auditory cortex was stimulated as frequency discrimination appears to be at least partially lateralized to this hemisphere (Lauter et al., 1985; Hyde et al., 2008). For anodal stimulation, the current was ramped up to 1 mA over 30 s, maintained at this level for 20 min, and then ramped off over 30 s. For sham stimulation, the current was ramped up to 1 mA over 30 s and immediately ramped off over 30 s. There is no ongoing sensation of stimulation after the initial ramp-up period so that sham stimulation produces the sensation of stimulation without inducing changes in cortical excitability (Ladeira et al., 2011; Kessler et al., 2012), making subjects blind to the stimulation condition. Subjects began the psychophysical procedures 30–60 s after stimulation had commenced.
We trained Naïve subjects for 2 days on a frequency discrimination task. To assess the effects of tDCS stimulation on rapid learning, we applied either anodal or sham tDCS stimulation during the first day of testing. The psychophysical procedure was repeated on the second day without tDCS to assess the effects of stimulation on retention of learning from the first day. The task followed that used by Hawkey et al. (2004) as they showed that the rapid decreases in frequency difference limens (DLFs) with training were genuine perceptual learning. A baseline measure could not be taken because this would prevent examination of rapid auditory learning that occurs during the early trials. It was hypothesized that DLFs for subjects receiving anodal tDCS would decrease more rapidly than those receiving sham stimulation, and that this decrease would be retained in the next day.
Frequency difference limens were determined on two successive days with a median time of 22 h between sessions (range = 18–24 h). On Day 1, one group (n = 7) was given anodal tDCS over right auditory cortex and another group (n = 8) was given sham stimulation over the same region. The subjects were randomly assigned to either tDCS or sham stimulation groups and were blind to the existence of a sham group until completion of testing. On Day 2, DLFs were determined in the same way as on Day 1 but without any tDCS for either group. DLFs were determined using an adaptive two-interval, two-alternative forced-choice (2I-2AFC) task with a two-down, one-up rule estimating the 70.7% point on the psychometric function (Levitt, 1971). One interval, selected at random, contained a 1000-Hz tone (the standard) and the other interval contained a tone with a frequency of 1000 + Δf Hz (the comparison). Tones were 100 ms long with 20-ms cosine rise/fall ramps, and were separated by a 500-ms interstimulus interval. The observation intervals were indicated by the numerals ‘1’ and ‘2’, which appeared successively on a computer screen coincident with the observation intervals. Subjects indicated the interval containing the comparison tone by clicking the left or right button on a mouse to indicate the first or second interval respectively. Response feedback (illumination of a green or red light on the screen) was given immediately after the response.
Following Hawkey et al. (2004), the initial frequency increment for the comparison stimulus (Δf) was 200 Hz. For the first six trials in each track, Δf was halved after two correct responses and doubled following an incorrect response; after the sixth trial, Δf was divided by √2 following two correct responses and multiplied by √2 after an incorrect response. Blocks of 180 trials were made up of three interleaved 60-trial tracks, with each track yielding an independent frequency discrimination threshold. Three 180-trial blocks were completed each day, with a self-paced break (typically < 1 min) between successive blocks. DLFs were calculated for each track as the geometric mean of Δf for the last eight reversals and for each block as the geometric mean of DLFs obtained from each of the three tracks (Hawkey et al., 2004). Response times were measured as the time (in ms) between the onset of the second tone and the response. Median response times were calculated for each track and response times for each block were taken as the geometric mean of the three tracks. Both DLFs and response times were positively skewed and were subject to natural logarithmic transformation for analysis; back-transformed values are reported. All stimuli were presented 20 dB above each subject's absolute threshold, which was determined immediately before testing each day with a 2I-2AFC procedure using a three-up, one-down rule to estimate a 79.4% threshold (Levitt, 1971).
Following the results of Experiment 1, we conducted two experiments to determine the effects of tDCS on the processes underlying frequency discrimination, place and temporal coding. We first examined the effects of tDCS on frequency selectivity, a psychophysical measure of place coding, at both 1000 and 2000 Hz. According to place coding theory (Zwicker, 1974), the bandwidth of frequency selectivity determines frequency discrimination, with smaller bandwidths producing smaller DLFs. Psychophysical tuning curves (PTCs) are commonly used to measure frequency selectivity, with wider PTCs indicating broader frequency selectivity (Moore et al., 1984). PTCs were determined at two frequencies to examine frequency-specific effects of tDCS on auditory perception. If tDCS degrades frequency discrimination by affecting place coding it will be evident in broader PTCs.
A within-subjects design was employed with the effects of tDCS on PTCs being assessed in separate sessions where either anodal or sham tDCS stimulation were applied over auditory cortex. A fast method was used to determine PTCs for 1000- and 2000-Hz test tones using the SWPTC program, which quickly and reliably measures frequency selectivity (Sęk et al., 2005; Sęk & Moore, 2011). A fast method was used rather than a lengthy constant-stimulus method as ethical guidelines recommend tDCS only be delivered for 20 min in a session (Bikson et al., 2009). The fast method allowed each frequency to be assessed in 10 min and was appropriate for the study. Tones were presented 10 dB above each subject's 70.7% absolute threshold, estimated immediately prior to each testing session. The sampled point of the psychometric function for absolute thresholds was changed from Experiment 1 for consistency with previous measures of frequency selectivity (Sęk et al., 2005; Sęk & Moore, 2011).
The PTC task required subjects to detect a test tone (with a frequency referred to as fc) in the presence of a narrow-band noise whose center frequency was gradually swept across a range of frequencies. As frequency selectivity represents the auditory system's ability to resolve frequencies, noise will interfere with detection only when it cannot be resolved from the test tone. The bandwidth of narrow-band noise was 200 Hz for the 1000-Hz test tone and 320 Hz for 2000-Hz test tone. Simultaneously with presentation of the test tone, which was pulsed on and off for 200 ms (with a 50% duty cycle), the center frequency of the narrow-band noise was swept at a constant rate from 0.5fc to 1.5fc over 5 min. At the start of the procedure the narrow-band noise was presented at 0.5fc at 25 dB above absolute threshold so it was clearly audible. Subjects were required to hold a key down while the tone was audible and release it when it was not. The amplitude of narrow-band noise increased by 2 dB/s if audible and decreased by 2 dB/s if inaudible. A change in amplitude of the narrow-band noise from increasing to decreasing (or vice versa) defined a reversal. The set of values of the amplitude of the narrow-band noise and its center frequency at each reversal defined the PTC.
Subjects were trained on the task for 2–4 h for both the 1000- and the 2000-Hz test tones to give consistent performance before the stimulation sessions. After training, PTCs were measured during two sessions in which either anodal or sham tDCS stimulation was applied for 20 min while subjects completed the task. In each experimental session, subjects first practised the task for 10 min, once for each 1000- and 2000-Hz test tone, before stimulation was applied. Two PTCs were determined for each test tone to give stable measurements, resulting in four PTC determinations per session. Anodal or sham stimulation was applied during four 5-min PTC determinations. All subjects had one anodal tDCS and one sham session with the order of stimulation counterbalanced. Sessions were separated by a week to avoid any carry-over stimulation effects. Each session lasted approximately 45 min with PTC measurements taking 20–25 min.
A rolling average of the amplitude of the narrow-band noise and its center frequency of two successive reversals was used to smooth the PTC and the frequency of the lowest point of the smoothed function (the LP) was found. The low-frequency slope was defined as 0.75× LP to LP and the high-frequency slope was defined as LP to 1.25× LP. Separate rounded exponential (roex(p)) functions were fitted to low- and high-frequency slopes using the equation (described in Patterson et al., 1982) for each slope:
where W is the shape of the PTC, g is the normalized deviation from the center frequency, p is the slope of the function and r is the shallower tail of the function. This produces low- and high-frequency slopes of the PTCs, with higher values indicating steeper slopes. The arithmetic mean for the low- and high-frequency slopes of the two determinations for each fc was taken. Equivalent rectangular bandwidths (ERBs) were determined using the products of the roex(p) fitting with the equation (Moore, 1995):
where fc is the frequency of the tone, pl is the slope of the low-frequency equation and pu is the slope of the high-frequency equation. Data were normally distributed and suitable for parametric analysis.
The second follow-up experiment measured the effects of anodal tDCS on temporal fine structure (TFS), which is dependent on the fidelity of temporal coding information (Rose et al., 1967). The experimental design was similar to Experiment 2A with TFS measured in separate tDCS and sham stimulation sessions for each subject. Sensitivity to TFS was measured using the method described in Hopkins & Moore (2007) and Moore & Sęk (2009). This method estimates a TFS threshold using an adaptive 2I-2AFC procedure with a two-up, one-down rule estimating the 70.7% point on the psychometric function (Levitt, 1971). Subjects were required to discriminate between two series of four harmonic tones. In one series, all four tones were drawn from the same harmonic whereas in the other they alternated between an inharmonic and harmonic complex. The harmonic tone complex had a fundamental frequency of 100 Hz, whereas the inharmonic complex had the same fundamental frequency but with each component shifted by the same amount (Δf). A harmonic complex with a 100-Hz fundamental frequency was used as it consists of components with frequencies around 1000 Hz, where frequency discrimination was measured in Experiment 1. Both complexes had the same harmonic envelope equal to a fundamental frequency of 100 Hz but with different TFS. In each trial, one interval, selected at random, contained the harmonic complex and the other contained the inharmonic complex. Intervals were indicated by numbered flashing boxes presented onscreen coincident with the presentation of the complexes. Subjects clicked with a computer mouse on the box corresponding to the interval containing the inharmonic tones. Following Moore & Sęk (2009), the duration of each complex was 200 ms and the two complexes were separated by a 300-ms interval. Feedback was given after each trial, with the selected observation period flashing either green for correct or red for incorrect. At the start of each block, Δf was set at 50 Hz and was adapted according to response. Again following Moore & Sęk (2009), blocks were terminated following eight reversals, and the threshold for the block was taken as the arithmetic mean of Δf for the last six reversals. To prevent subjects discriminating on place coding, all components were passed through a fixed band-pass filter set centered at 900 Hz with a width of 110 Hz rolling on at 30 dB per octave. Threshold equalizing noise, presented 15 dB below stimulus presentation level (SPL) and extending from 50 to 11 050 Hz, was used to mask components of the complexes falling outside the band-pass filter. Following Moore & Sęk (2009), SPL for the harmonic complex was set 20 dB SPL above each subject's 70.7% absolute threshold measured using an adaptive 2I-2AFC staircase method for 900 Hz immediately prior to each session. The sampled point of the psychometric function of absolute threshold was changed from Experiment 2A for consistency with previously established measures of TFS.
To give consistent performance, subjects had one initial training session prior to testing where they practised the TFS task for ~45 min. There were two counterbalanced TFS testing sessions after training where either anodal or sham tDCS was applied, separated by a week to avoid any carry-over effects of stimulation. Subjects completed seven threshold procedures during the 20 min of either tDCS or sham stimulation. Each staircase lasted ~2 min, varying with the subject's response times and number of trials needed for six reversals. The threshold for that session was taken as the arithmetic mean of the seven thresholds for the session, each of which lasted approximately 35 min. Data were normally distributed and suitable for parametric analysis.
Mean DLFs (± SEM) for both stimulation groups from each of the three blocks on both testing days are shown in Fig. 1. Because stimulation was only delivered on the first day, separate 3 (Block) × 2 (Stimulation) mixed-measures anovas were conducted on DLFs in each day. On the first day, mean DLFs rapidly decreased for both groups with training (F2,26 = 5.70, P = 0.009, = 0.31), showing rapid perceptual learning. DLFs decreased by 0.95 Hz for the tDCS group and by 0.86 Hz for the sham group. The interaction between Block and Stimulation did not approach significance, offering no evidence of a different rate of learning in the two groups (F2,26 = 1.04, P = 0.36, = 0.07). DLFs, however, were considerably higher in the tDCS than the sham group (F1,13 = 4.84, P = 0.046, = 0.27). The mean overall DLF for the tDCS group (1.46 Hz) was about double that of the sham stimulation group (0.65 Hz), although both groups improved to a similar extent with training. tDCS therefore degraded frequency discrimination without affecting perceptual learning. Most subjects in the tDCS group showed high DLFs during Block 1 that decreased by Block 2. Some subjects in this group, however, did not show smaller DLFs until Block 3. This variation in the effect of tDCS on auditory cortical functioning most likely caused the greater inter-individual variability of DLFs in the tDCS compared with sham stimulation group as evident in Fig. 1.
DLFs in the sham group became asymptotic by the third training block on Day 1 and remained stable on Day 2, whereas DLFs in the tDCS group returned to near initial levels on Day 2. There was no overall learning effect on Day 2 (F2,26 = 1.22, P = 0.31, = 0.09). The interaction between Block and Stimulation, however, was significant (F2,26 = 4.20, P = 0.03, = 0.24). This was due to the sham stimulation having asymptotic DLFs on all blocks whereas DLFs for the tDCS group decreased from Block 4 to 5. DLFs in the group given tDCS on Day 1 were still higher than those for the group given sham stimulation on Day 1 (F1,13 = 4.80, P = 0.047, = 0.27). The overall DLF for the tDCS group (1.19 Hz) was slightly lower than during stimulation on Day 1 but was still about double that of the sham stimulation group (0.59 Hz), showing a persistent effect of tDCS on frequency discrimination.
Fig. 2 shows that response times decreased monotonically over training blocks for both groups. Response times for both groups decreased over Blocks on Day 1 (F2,26 = 21.38, P < 0.001, = 0.62) and Day 2 (F2,26 = 4.88, P = 0.016, = 0.27). Stimulation did not differentally affect response times with training as the interaction of Stimulation and Block did not approach statistical significance on either Day 1 or Day 2 (both F <1). Although response times on both days were slightly shorter for the tDCS (M = 640 ms) than the sham stimulation group (M = 676 ms), there was also no overall effect of stimulation on response times on either day (both F <1). This suggests that while tDCS was interfering with frequency discrimination it did not interfere with the ability to perform the task.
Because DLFs were still significantly higher for the tDCS group than the sham group on Day 2, all subjects who received this treatment were contacted to complete a third day of testing without stimulation; all but one was re-tested between 48 and 109 days (median = 64 days) after the initial test day. To determine if the tDCS group's performance returned to normal levels, it was compared with the sham group's performance on Day 2. Fig. 3 shows DLFs (upper panel) and response times (lower panel) for the tDCS group's Day 3 results (n = 6) and those for the sham group's performance on Day 2 (n = 8). Re-tested DLFs for the tDCS group were similar to those for the sham group on Day 2 (F2,24 = 4.26, P = 0.06, = 0.49) and considerably smaller than the this group's DLFs 1 day after stimulation (0.85 and 1.19 Hz, respectively). Response times were also similar between the tDCS group's re-tested results and the sham group's performance on Day 2.
Contrary to expectations, anodal tDCS over auditory cortex did not accelerate rapid frequency discrimination learning, but did degrade frequency discrimination, with the mean DLF in the tDCS group about 0.8 Hz higher than that in the sham stimulation group. This degradation was still present on the testing session 1 day after stimulation with DLFs being ~0.6 Hz higher, showing that the effects of changing cortical excitability persisted for at least 24 h after stimulation, but was not present 2–3 months following stimulation, showing that the effect was not permanent. As response times for both groups were similar and decreased with training it is unlikely that the effect of stimulation was due to stimulation inhibiting task performance. The results overall suggest strongly that the increased DLFs for the tDCS group are a genuine perceptual degradation rather than a more general impairment in the ability to perform the task.
Frequency selectivity, quantified as ERB values, relies on place coding, which is thought to be one process that underlies frequency discrimination. We hypothesized that if tDCS degraded frequency discrimination by affecting place coding it would be evident in broader ERBs. Fig. 4 shows representative PTCs for the 1000- and 2000-Hz test tones during a tDCS and a sham stimulation session. As shown, the amplitude of the narrow-band noise was lower when it contained frequencies near that of the test tone. For this subject, PTCs for the 1000-Hz test tone were very similar during both tDCS and sham stimulation sessions. For the 2000-Hz test tone, the PTC was broader during tDCS than sham stimulation, showing that a wider range of noise frequencies interfered with detection of the test tone.
Mean ERB values for the tDCS and sham stimulation sessions for the 1000- and 2000-Hz test tones are shown in Fig. 5. Consistent with previous research (Moore & Glasberg, 1983; Glasberg & Moore, 1990, 2000), ERBs were broader for the 2000-Hz (M = 231.6 Hz, SE = 11.4) than 1000-Hz (M = 170. 2 Hz, SE = 13.4) test tone. At 1000 Hz, ERBs were similar in the tDCS and sham stimulation sessions (t6 = 1.15, P = 0.30, Cohen's d = 0.05). However, tDCS significantly broadened frequency selectivity at 2000 Hz (t6 = 2.80, P = 0.031, Cohen's d = 1.17).
We examined in this experiment the effects of anodal tDCS applied over primary auditory cortex on TFS thresholds, a psychophysical measure relying on temporal coding. Fig. 6 shows TFS thresholds were markedly larger during tDCS than sham stimulation sessions (t5 = 2.72, P = 0.04, Cohen's d = 0.62). TFS thresholds were consistently greater in the tDCS than the sham stimulation session with this effect shown in all but one subject.
Our hypothesis that increasing excitability of auditory cortex with anodal tDCS would enhance rapid frequency discrimination learning was not supported. Both tDCS and sham stimulation groups showed similar decreases in thresholds with training. We found unexpectedly that tDCS degraded frequency discrimination, with subjects receiving tDCS stimulation having mean DLFs more than double those receiving sham stimulation. This effect persisted for at least 24 h after stimulation but had dissipated on retesting 2–3 months later. Two follow-up experiments that investigated the source of the tDCS-induced degradation of frequency discrimination showed that although tDCS did increase the ERB of the PTCs measured at 2000 Hz, it had no effect at 1000 Hz (the frequency tested in Experiment 1), and that tDCS increased TFS thresholds by ~30%. Together, these results suggest that tDCS degrades frequency discrimination by affecting temporal, rather than place, coding mechanisms.
Anodal tDCS did not enhance auditory learning
It is unclear why anodal tDCS over auditory cortex did not enhance frequency discrimination learning during stimulation given the many reports that such stimulation over motor cortex enhances motor learning (Nitsche et al., 2003b; Antal et al., 2004a,b; Reis et al., 2009). It should be noted first the difference between the groups does not appear to be due to sampling error, biasing the allocation of differently hearing subjects. All subjects reported normal hearing and stimulus presentation levels were individually tailored to ensure consistency between subjects. There is additional evidence suggesting all subjects had normal frequency discrimination, as DLFs for all subjects during Block 1 were within normal levels (Moore, 2012) and subjects in both groups improved similarly with training. It is also unlikely the simultaneous degradation of frequency discrimination masked the enhancement of learning, as a previous study (Amitay et al., 2005) has demonstrated that subjects with initially poor frequency discrimination show the greatest improvements. The difference between groups is therefore likely to be a genuine experimental effect. We placed the cathodal electrode over the supraorbital region as we were replicating conventional practice in motor learning studies. The thick skull bone over frontal cortex partially attenuates the stimulation effect (Miranda et al., 2013) and placing the cathodal electrode over this region does not impair motor learning (Nitsche et al., 2003b; Reis et al., 2009). Importantly, the electrode montage used in the current study has been shown to increase the amplitude of an early component of the auditory event-related potential (Zaehle et al., 2011), strongly indicating that the technique used in the current study was suitable for increasing auditory cortical excitability. However, as in all studies with two cephalic electrodes, it is possible that the observed effects can be due to changes in cortical excitability under both electrodes.
The functional organization of auditory cortex, like that of motor cortex, is plastic and changes readily with experience (Weinberger & Diamond, 1987; Robertson & Irvine, 1989; Recanzone et al., 1993). More specifically, research in humans has shown that training with auditory stimuli increases the early components of auditory event-related potentials in parallel with rapid improvements in frequency discrimination, a finding that has been generally interpreted as showing rapid learning-induced changes in frequency representation in auditory cortex (Tremblay et al., 1998; Menning et al., 2000; Alain et al., 2007, 2010; Bosnyak & Gander, 2007). The failure of tDCS to enhance auditory learning does not therefore reflect an incapacity of the auditory cortex to change with experience. As we have shown here, anodal tDCS over auditory cortex degrades auditory frequency discrimination. In contrast, anodal tDCS over motor cortex immediately improves motor skill (Antal et al., 2004b; Vines et al., 2006) as well as enhancing motor learning and retention, and it is possible that an immediate stimulation-induced enhancement of performance is a necessary prerequisite for a stimulation-induced increase in learning and retention. Anodal tDCS over primary somatosensory cortex induces a rapid increase in spatial acuity measured on the tip of the index finger, an effect that persisted after stimulation (Ragert et al., 2008), suggesting stimulation-induced enhancement of perceptual discrimination and perceptual learning. Similarly, increasing the excitability of somatosensory cortex with high-frequency trains of transcranial magnetic stimuli induces an immediate increase in the spatial acuity of the index fingertip (Tegenthoff et al., 2005) and enhances tactile perceptual learning (Karim et al., 2006). In the current study, anodal tDCS may have failed to enhance perceptual learning because the sensory representation of the stimulus, unlike in previous studies, was also not simultaneously enhanced.
Although there was no evidence of an effect of anodal stimulation on frequency discrimination learning during Day 1, with DLFs similarly decreasing for both groups, there was a significant interaction between stimulation group and training on Day 2 when stimulation was not present. DLFs for the tDCS group returned to baseline levels between sessions while remaining at trained levels for the sham group, suggesting that stimulation degraded the consolidation of learning. This is unlike the effect of motor skill learning where anodal tDCS increases between-day consolidation (Reis et al., 2009). There is evidence from letter enumeration tasks, where subjects determine if the number of letters presented is odd or even, showing that learning is only retained if asymptotic performance is reached within each session (Hauptmann & Karni, 2002; Hauptmann et al., 2005). In the current study, the sham group had stable performance between Blocks 2 and 3, whereas DLFs for the tDCS group decreased in this period, suggesting asymptotic thresholds had not been reached in the session.
Temporal, but not place, coding is degraded by anodal tDCS
The lack of effect of tDCS on frequency selectivity around 1000 Hz, and the decreased sensitivity to TFS during tDCS, indicate that the degradation of frequency discrimination around 1000 Hz by anodal tDCS was probably due to interference with temporal coding. Imposing a transcortical DC current has been shown to immediately alter the spontaneous firing rate of cortical neurons in the rat (Bindman et al., 1964) and it is possible that tDCS interferes directly with temporal coding by disrupting the precision of the phase-locked firing pattern of active auditory neurons. Most evidence from both animals and humans points to the importance of temporal coding for frequency perception below 4000–5000 Hz (Rose et al., 1967; Johnson, 1980; Moore, 2012). Auditory neurons show millisecond-precise phase-locking firing rates to both complex and pure tone frequencies below 5000 Hz (Zatorre, 1988; Averbeck et al., 2006). Even small perturbations of temporal coding, in the scale of milliseconds, result in information loss for cortical neurons (Kayser et al., 2010). Changing the excitability of auditory cortex using tDCS could therefore sufficiently disrupt the fine structure information needed for precise temporal coding. There do not appear to be any studies in either animals or humans showing a dissociation of place and temporal coding processes following lesions to auditory cortex, although bilateral lesions impair perceptual discriminations relying on both temporal (Bowen et al., 2003) and place (Cooke et al., 2007) coding. These processes do appear to be at least partially lateralized, with the left hemisphere showing a preference for temporal information and the right showing a preference for place information (Zatorre & Belin, 2001; Schönwiesner et al., 2005).
Our findings that anodal tDCS over auditory cortex decreased frequency selectivity at 2000 Hz but not at 1000 Hz, and decreased sensitivity to temporal fine structure, show that altering auditory cortical excitability in this way has complex effects on auditory function. This diversity of effects is present in previous reports of the effects of tDCS over temporal regions on auditory function. Cathodal, but not anodal, tDCS over temporal cortex has been reported to interfere with frequency discrimination at 200 Hz, revealing an inhibitory effect of cathodal stimulation without a reciprocal excitatory effect of anodal stimulation (Mathys et al., 2010). The effects of tDCS on auditory event-related potentials similarly show complex effects, with anodal stimulation increasing the amplitude of the P50 component when delivered over temporal cortex and increasing the amplitude of the N1 component when delivered over temporo-parietal cortex (Zaehle et al., 2011). Anodal tDCS has been shown to enhance detection of temporal gaps in a 4000-Hz auditory carrier, without corresponding effects with carriers at lower frequencies (Ladeira et al., 2011). Although these authors report a frequency-specific effect of tDCS over auditory cortex, they did not measure the ability to discriminate different frequencies.
The diversity of effects of stimulation over temporal regions, in contrast to the consistent polarity-specific effects of stimulation over motor cortex, might reflect the structural and functional characteristics of auditory cortex. The primary auditory region is located on the transverse temporal gyri in the lateral sulcus. It is most responsive to narrow-band stimuli like pure tones (Bendor & Wang, 2006), and has at least two distinct tonotopic gradients with neurons with different characteristic frequencies probably having different orientations within the gyri (Talavage et al., 2004; Humphries et al., 2010; Da Costa et al., 2011; Langers & van Dijk, 2012). Neurons with characteristic frequencies of 1000 and 2000 Hz are located on different regions of the transverse temporal gyri, meaning each is differentially orientated relative to the scalp (Da Costa et al., 2011). The current flow generated in the brain by passing a direct current through scalp electrodes is complex, and depends on factors such as the morphology of the cortical surface and local variability in conductivity (Datta et al., 2009; Stagg & Nitsche, 2011). The deep location of the primary auditory region, and the variability in the orientation of frequency-specific cells in the multiple tonotopic representations to the direction of current flow, are likely to lead to diverse effects on tDCS on auditory perception.
It would be interesting to examine the effects of stimulating motor cortex on auditory functioning as a clear enhancement of motor functioning is evident with anodal tDCS over motor cortex. Recent evidence suggests an important role for interacting activity in sensory and motor cortical areas during perceptual discrimination. This work emphasizes the active role of the motor cortex in formulating a decision in even simple perceptual judgments, with activity in motor cortex linked directly to low-level sensory processing (Donner et al., 2009; Siegel et al., 2011; de Lange et al., 2013). Thus, recurrent activity between sensory and motor areas informs the motor response and modulates the interpretation of incoming sensory information. In addition, improvements on a sensory discrimination task reflect both perceptual learning (the increased ability to discriminate the specific trained stimuli) and procedural learning (understanding and dealing with the sequence of events in each trial, including formulating a decision) (Ortiz & Wright, 2009). Although fine motor skill was not required for the manual response (clicking the left or right mouse button) in the task used here, increasing the excitability of motor cortex could enhance learning of other aspects of the procedural component of the perceptual judgment task. Increasing excitability of motor cortex might therefore enhance both the recurrent processing of sensory input and the procedural component of perceptual learning.
A limitation of the current study is that only the effects of anodal tDCS on frequency discrimination were examined. It is possible that cathodal tDCS would enhance frequency discrimination and auditory learning, as cathodal and anodal stimulation have opposite effects on cortical excitability. Some previous studies of tDCS have, however, reported effects on visual function of one polarity of stimulation but not the other. Cathodal tDCS enhances global motion processing while anodal stimulation has no effects (Antal et al., 2004c), and visual attention is similarly enhanced by cathodal stimulation with no effect of anodal stimulation (Moos et al., 2012). In contrast, cathodal tDCS has no effect on contrast discrimination, which is enhanced by anodal stimulation (Olma et al., 2011).
Long-term effects of tDCS over auditory cortex
The persistent effect of tDCS on frequency discrimination thresholds reported here is a novel finding; previous studies of the effect of tDCS on perception have not looked for lasting effects (Antal et al., 2004b; Mathys et al., 2010). In a similar way to the current study, altering cortical excitability by low- and high-frequency alternation of visual stimulation has been reported to change visual discrimination thresholds for up to 10 days (Beste et al., 2011). Animal studies have shown that inducing neuronal depolarization with a direct current applied to the cortex can cause persistent synaptic changes with increased calcium ion and cyclic AMP levels, which are associated with cortical plasticity (Hattori et al., 1990; Moriwaki, 1991; Islam et al., 1995). This is consistent with extensive neurophysiological findings showing that frequency discrimination training causes long-term synaptic changes in neurons in primary auditory cortex (Weinberger & Diamond, 1987; Recanzone et al., 1993; Bosnyak & Gander, 2007). The effects of tDCS over motor cortex, measured by the amplitude of motor-evoked potentials elicited by transcranial magnetic stimulation, typically persist for up to about 90 min (Nitsche & Paulus, 2001). Transcranial magnetic stimulation may, however, be insufficiently sensitive to reveal persistent cortical excitability induced by tDCS with psychophysical tests providing a sufficiently sensitive behavioral measure of persistent synaptic change.
Our results show the first experimental dissociation between place and temporal coding processes in frequency discrimination in normal-hearing humans. The interference with temporal coding, but not with place coding around 1000 Hz, by tDCS could be a direct result of changed auditory cortical processing or an indirect result of auditory processing at lower levels of the neuraxis exerted through a corticofugal system. Generally, the dissociation of place and temporal coding processes by anodal tDCS offers a new means of exploring cortical processes in audition.
Funding was provided to M.F.T. by The University of Western Australia. We thank B. C. J. Moore and A. Sęk for providing programs we used to measure frequency selectivity and fine temporal structure. A. Sęk also provided technical assistance. The authors declare no competing financial interests.