Neurophysiological Evidence of Enhanced Cognitive Control Following Auditory Control Enhancement (ACE) Therapy: A Pilot RCT in Schizophrenia-Spectrum Disorders with Persistent Auditory Hallucinations

Abstract

Schizophrenia (Sz) involves positive symptoms such as auditory hallucinations, paranoia, and delusions, and are commonly linked to deficits in cognitive control. These deficits appear in tasks like the A-X version of the continuous performance test (AX-CPT), which reflects reduced proactive control, and in diminished auditory steady-state responses (ASSR) during selective attention. Notably, in Sz, ASSR to sounds presented during diverted attention increases with hallucination severity. We report results from a single-blind randomized controlled trial evaluating Auditory Control Enhancement (ACE) therapy—the combination of transcranial direct current stimulation (tDCS) paired with Auditory Cognitive Control Training (ACCT)—to improve cognitive control in Sz. Ten individuals with Sz participated: six received active ACE and four received sham tDCS with ACCT. All experienced auditory hallucinations at least three days per week. ACCT consisted of three 45-min computerized sensory reaction-time sessions over three consecutive days. All procedures, including baseline and follow-up assessments, were performed over a two-week period. Electroencephalography (EEG) was collected before and after training while participants completed AX-CPT and ASSR tasks. ASSR blocks alternated between attending to auditory stimuli and ignoring them while watching a silent video. Auditory hallucinations were assessed using the PSYRATS at baseline and post-intervention. Procedures were well tolerated, and acceptability was high. Participants receiving ACE showed increased attention modulation (larger ASSR attend–ignore difference), greater beta desynchronization during proactive inhibition, and reductions in physical aspects of auditory hallucinations. In contrast, the sham group showed no changes in EEG measures or hallucination severity. These preliminary results suggest ACE may enhance cognitive control and reduce physical aspects of auditory hallucinations in Sz. ACE could represent a promising intervention for cognitive dysfunction.

Auditory hallucinations (AH) associated with schizophrenia (Sz) impose an enormous burden on individuals and society. People with Sz commonly report AH as the most distressing and problematic systems, with the inability to concentrate and avoid distraction caused by AH being most commonly reported as a priority for treatment.^1,2 In about 25% of cases, AH only partially responds to antipsychotic medications. Cognitive symptoms, which most strongly predict global functioning,³ are unchanged or worsen due to metabolic side effects of antipsychotic medications.⁴ Thus, a novel biomedical treatment for improving cognition and reducing AH in Sz is a critical unmet clinical necessity.

People with Sz have problems planning, modifying, and executing goal-directed behavior,^5,6 devoting attention toward or away from sensory (particularly auditory) stimuli,^7,8 and focusing on items held in working memory^9,10 – a process collectively referred to as cognitive control. Although cognitive control is most frequently conceptualized in terms of behavior, it also plays a pivotal role in basic auditory perception. In infants learning language, cognitive control is critical for differentiating variation in native and non-native speech and enables the listener to ignore irrelevant variation.¹¹ Cognitive control is used when listening to speech in noise¹² or to non-native-accented speech.¹³ Auditory cognitive control is linked to the auditory control network, comprising auditory/verbal areas in superior temporal gyrus (STG) and temporoparietal junction (TPJ), and cognitive control systems in anterior cingulate cortex (ACC), midcingulate cortex (MCC), supplementary motor area (SMA), pre-SMA, and right ventrolateral prefrontal cortex (rVLPFC). rVLPFC is well-known for its role in behavioral inhibition, as underscored by the tale of Phineas Gage, whose personality changes after damage to the area were quoted as “giving the animal propensities absolute control”.^14,15 rVLPFC is active during attention to upcoming auditory and visual stimuli¹⁶ and has been linked to inhibition of auditory perceptual misrepresentations.¹⁷ For those with Sz, auditory control network impairment may contribute substantially to auditory hallucinations; behavioral measures of cognitive control such as response inhibition and sensory attention correlate strongly with auditory hallucination severity.¹⁸

Fluctuations in specific frequency bands of this signal have been linked to cognitive, behavioral, and physiological phenomena. An extensive and growing literature has documented cortical and subcortical abnormalities in beta and gamma band oscillations (15–80 Hz) in Sz, particularly in the context of tasks requiring cognitive control.^19–24 A common measure of cognitive control deficits in Sz comes from the AX-type continuous performance test (AX-CPT), where participants respond to X only when following an A, not when preceded by any other letter (BX error) or to an X not preceded by an A (AY error). Difference in prefrontal beta-band oscillatory power between preparing to inhibit versus preparing to execute a prepotent response in this task is widely accepted as a neurophysiological measure of cognitive control.^25,26 Another measure representing the link between high-frequency oscillations and cognitive control is reduced attentional control of gamma-band Auditory Steady-State Response (ASSR), which originates primarily in auditory cortex.^27,28 Auditory click series presented at a specific frequency (usually 40 Hz) cause the auditory cortex to oscillate at that frequency, termed the ASSR. Although the gamma-band ASSR does not in itself represent cognitive control, it can be modulated by devoting attention toward or away from the stimulus.^29,30 The ASSR itself is reduced in Sz.^31,32 Recent data shows that people with first-episode Sz are unable to modulate the ASSR with attention to the same degree as matched controls, and that this deficit strongly correlates with AH severity.³³ Based on this increasing body of evidence, frontal beta and gamma oscillations may reflect intrinsic defects of underlying neuronal circuits in Sz.

We have developed a potentially transformative outpatient treatment protocol that combines a time-tested psychotherapeutic computer-based training program with targeted transcranial Direct Current Stimulation (tDCS). To train auditory cognitive control, we have adapted Cognitive Enhancement Therapy, a behavioral treatment developed at Western Psychiatric Hospital that has repeatedly proven reliable for increasing cognitive performance,³⁴ with effects linked to changes in prefrontal network connectivity.^34–36 Cognitive control improvements occur primarily during the initial component of Cognitive Enhancement Therapy and most strongly predict positive outcomes.³⁴ Therefore, we updated and streamlined this initial portion to target auditory cognitive control. We call this computerized training Auditory Cognitive Control Training (ACCT). ACCT exercises target cognitive control processes that rely on rVLPFC, including reaction time training in a temporal mode using auditory cues (The Attention Reaction Conditioner),³⁷ intentional inhibition of motor responses training in a spatiotemporal mode with visual cues (the Zero Accuracy Conditioner),³⁸ and temporal vigilance and time estimation with auditory and visual cues (Time Estimates).³⁹

Animal studies show that anodal tDCS enhances and cathodal tDCS reduces neuronal excitability⁴⁰ and human studies confirm these results.⁴¹ tDCS of auditory/verbal areas in temporoparietal cortex has shown promise for alleviation of treatment-resistant AH in prior studies^42,43; however, findings are mixed. In most of these studies, tDCS is delivered at rest, meaning that the degree to which people are engaging in introspection, socialization, or cogitation is not controlled. This may be critically relevant to differences in neural target engagement during tDCS, as effects of tDCS on plasticity and learning are dependent upon adaptive utilization of the target brain circuitry.⁴⁴ By combining training with targeted tDCS of rVLPFC in healthy volunteers, we have demonstrated that the rate of learning to visually scan and identify objects from complex images can be enhanced by as much as 2x.^45,46 In a series of follow-up experiments, these highly-replicated effects of tDCS on training enhancement have been linked to cognitive control enhancement during learning.^47,48

Here, we present findings from a single-blind, sham-controlled, parallel-group pilot randomized controlled trial (RCT) with baseline and follow-up assessments evaluating Auditory Control Enhancement (ACE) therapy for improving cognitive control in individuals with a schizophrenia-spectrum disorder experiencing persistent auditory hallucinations. We hypothesized that ACE (the pairing of tDCS with ACCT, a behavioral training that engages the target system) would improve cognitive control, evinced by (1) increased beta-band ERD during proactive inhibition in the AX-CPT, and (2) increased attentional modulation of the ASSR in frontocentral EEG electrodes. We additionally examined changes in cognitive symptoms assessed via the MATRICS Consensus Cognitive Battery (MCCB) and positive symptoms assessed using the Psychotic Symptoms Rating Scales (PSyRatS).

Methods

Participants

This single-blind, sham-controlled, parallel-group pilot RCT with baseline and follow-up assessments included 10 individuals with current diagnosis of a schizophrenia-spectrum disorder, of which 6 were randomly assigned to receive active tDCS (3 male, mean age = 30.0 ± 3.1 yrs [SD]), and 4 to sham (3 male, mean age = 26.3 ± 8.9 yrs). We were unable to reach our target enrollment of 20 individuals in this study due to difficulties in recruitment of individuals meeting inclusion criteria during the funding period. This target was determined through power analysis based on prior pilot studies. Participants were recruited between June 2023 and December 2024. As inclusion criteria, all participants experienced persistent auditory hallucinations at least 4 days per week despite attempting three or more antipsychotic medications with greater than one month of medication compliance with duration of illness equal to or less than 10 years. Exclusion criteria included hearing deficit as assessed by audiometry (hearing threshold > 30 dB nHL), head injury with loss of consciousness > 10 min, medical illness affecting brain function or structure, significant neurologic disorder (eg seizure disorder), current or prior substance use disorder other than cannabis, current pregnancy, or current medications that affect alertness, other than antipsychotic medication (eg sedatives, tranquilizers, muscle relaxants, and sleeping aids). Participants provided voluntary informed consent and were compensated for participation. Detailed accounting of individuals screened, enrolled, and allocated to treatment conditions are included as a CONSORT flow chart (Figure 1)

Figure 1.

CONSORT flow diagram. Of the 134 subjects screened, 34 were assessed for eligibility and 10 were randomized into treatment groups. No participants were lost to follow-up during the 2-week period of data collection and assessment.

Broad symptoms were rated using the Positive and Negative Symptom Scale (PANSS).⁴⁹ Symptoms were also assessed with the AH scale of the PSYRATS, which provides scores for physical, emotional, and cognitive factors of AH using 11 Likert-scale (0–4) questions.⁵⁰ The delusions scale of the PSYRATS was also administered. Participants were asked to rate their responses according to symptom severity over the prior 2-week period. All interviews and tests were conducted by an expert clinical assessor. Of the 10 participants, six were diagnosed with Schizophrenia (paranoid, n = 4; unspecified, n = 2) and four with Schizoaffective Disorder (bipolar type, n = 2; depressive type, n = 2).

Procedures

Participants completed EEG testing, the MATRICS Cognitive Consensus Battery,⁵¹ and the PSYRATS at study entry (baseline) and again within two weeks (follow-up) of completing Auditory Cognitive Control Training (ACCT), accompanied by either active or sham tDCS. ACCT is an abbreviated and streamlined version of Cognitive Enhancement Therapy (CET), a behavioral treatment that has repeatedly proven reliable for increasing cognitive performance,³⁴ with effects linked to changes in prefrontal network connectivity.^34–36 ACCT exer cises target cognitive control processes that rely on rVLPFC, including reaction time training in a temporal mode using auditory cues (The Attention Reaction Conditioner),³⁷ intentional inhibition of motor responses training in a spatiotemporal mode with visual cues (the Zero Accuracy Conditioner),³⁸ and temporal vigilance and time estimation with auditory and visual cues (Time Estimates).³⁹ These three exercises were performed in three separate one-hour sessions separated by at least one day, with all three sessions completed within a two-week period.

Transcranial Direct Current Stimulation (tDCS)

tDCS was delivered during the initial portion of each of the three ACCT sessions using a multichannel Starstim tES/EEG system (Neuroelectrics) with 8 cm² circular sponge electrodes soaked in saline. Individual participants were randomly assigned to receive either active or sham tDCS during ACCT. Random allocation was computer generated, and allocation sequence was kept confidential by the principal investigator. Research staff was informed of a participant’s random assignment only after that participant was enrolled. Participants were not informed whether they received active or sham tDCS. In the active condition, tDCS was delivered at 2.0 mA for 30 min. In the sham condition, tDCS was ramped down after only 30 s. In both conditions, the anode was positioned over right inferior frontal gyrus at location F8, and the cathode was positioned over left temporoparietal junction at location CP5, according to the 10–10 EEG coordinate system. Electric field modeling for this stimulation montage was implemented using SimNIBS 4.1 (https://simnibs.github.io/simnibs/) using individualized finite-element models generated from T1 and T2 MRI as well as utilizing the default (MNI) head model available in SimNIBS. MNI-based electric field models and average finite element models across active tDCS participants are shown in Figure 2.

Figure 2.

Electric field models for applied tDCS current. Electric field models are shown based on generic MNI head models (upper panels) as well as individualized finite element models based on individual T1 and T2 MRI (lower panels). Individualized models were averaged across the six participants receiving active tDCS.

MRI

We acquired structural MRI from each participant using a 3 T Siemens Prisma MRI system. We obtained T1-weighted bias-field-corrected magnetization-prepared rapid gradient-echo (MPRAGE) images, with 208 slices acquired in the sagittal plane at 0.8 mm³ voxel resolution (TR/TE/TI (ms) = 2400/2.22/1000), as well as 3D T2-weighted images to aid in tissue segmentation and maximize the resolution of finite-element models of tDCS current (TR/TE(ms) = 3200/563)

EEG

EEG data were collected during AX-CPT and ASSR tasks using a 32-channel Starstim tES/EEG system (Neuroelectrics), with 30 scalp electrodes placed according to the 10–20 EEG coordinate system, one site below the right eye, and one site at the outer canthus of the right eye. The online bandpass was 0.01–125 Hz digitized at 500 Hz. Using the MATLAB-based EEGLAB Toolbox,⁵² channels/segments with excessive noise were removed via visual inspection and a high-pass filter (0.5 Hz; 12 dB/oct) was applied. ICA was then performed to detect and remove eye-blink and ECG components. Using the Brainstorm toolbox, scalp EEG data were then re-referenced to the average of all channels (average reference) and an 80 Hz low-pass filter was applied.

AX-CPT

The AX-CPT was constructed and implemented using Presentation (Neurobehavioral Systems, Inc.) and consisted of 425 trials. Each trial began with a cue (any letter other than X, K, or Y) presented in the center of the screen for 250 ms. After an inter-stimulus interval of 1000 ms, the probe stimulus (any letter except A, K, or Y) was presented in the center of the screen for 250 ms, followed by a 1000 ms inter-trial interval. Participants were instructed to press a button with the index finger of their right hand as quickly as possible whenever they observed an A cue followed by an X probe, and to withhold response whenever they observed any other letter pair. 40% of the trials in each task block consisted of an A followed by an X (AX trials), 20% of the trials in each block consisted of an A followed by a letter other than X (pseudo-randomly selected; AY trials), 20% of the trials in each block consisted of a letter other than A (pseudo-randomly selected) followed by an X (BX trials), and 20% of the trials in each block consisted of a letter other than A (pseudo-randomly selected) followed by a letter other than X (pseudo-randomly selected; BY trials). Trials were presented randomly. Behavioral performance was assessed by the signal detection measure d-prime, mean reaction time for responses to AX trials, and standard deviation of reaction times for responses to AX trials.

Using the Brainstorm toolbox, ‘A’ and ‘B’ trials were segmented from 500 ms before to 1800 ms after stimulus onset, average baseline voltage was subtracted, and trials were rejected when the voltage in any electrode exceeded 100 μV difference from baseline. For calculation of evoked power, Morlet wavelet deconvolution was applied to trial data using 7 cycles at 1 Hz increments from 13–50 Hz, then averaged across trials and baseline-corrected to the average of 200 ms prior to stimulus onset. Our primary outcome measures were average desynchronization (reduced power from baseline) in the beta-band (15–25 Hz) following ‘B’ trials across five frontocentral electrodes (Fz, FCz, FC1, FC2, and Cz) within two discrete time windows: 1) evaluation, between 200 and 500 ms after cue onset, and 2) preparation, between 500 and 1100 ms after cue onset.

ASSR

Participants listened to binaural auditory click stimuli generated with Audacity and presented using Presentation (Neurobehavioral Systems, Inc.) over Etymotic 3A insert earphones in two blocks (320 stimuli per block). Stimuli consisted of trains of 1 ms biphasic clicks presented at a stimulation rate of 40 Hz (500 ms duration, 750–1150 ms stimulus onset asynchrony). While most stimuli were presented at 85 dB, a subset (40 stimuli, 12.5%) were presented at 80 dB. Simultaneously, a silent nature video was presented. In one block, participants were instructed to ignore the video and press a button in response to “softer” stimuli presented at 80 dB. In the other, they were asked to ignore the stimuli and watch the video. Block order was counterbalanced.

Using the Brainstorm toolbox, trials presented at 85 dB were segmented from 200 ms before to 700 ms after stimulus onset, average baseline voltage was subtracted, and trials were rejected when the voltage in any electrode exceeded 100 μV difference from baseline. For calculation of evoked power, Morlet wavelet deconvolution was applied to trial-averaged ERPs using 5 cycles at 1 Hz increments from 25–55 Hz. ASSR evoked power were measured as the average across 35–45 Hz, 100–500 ms post-stimulus, and five frontocentral electrodes (Fz, FCz, FC1, FC2, and Cz).

Data Analysis

Group demographics were compared between groups using t-tests and chi-squared tests where appropriate. Due to the small number of subjects in this dataset, change from baseline was calculated for all neurophysiological dependent measures. Change in AX-CPT behavioral measures (d-prime, RT, and RT standard deviation) were compared between Active and Sham using t-test for independent samples. Change in beta desynchronization response to “B” cues in the AX-CPT was compared using repeated measures ANOVA with one between-subjects factor (group: Active vs Sham) and one within-subject factors (time window: evaluation vs preparation). Change in ASSR power was compared using repeated measures ANOVA with one between-subjects factor (group: Active vs Sham) and one within-subject factors (attention: attend or ignore). Significant effects were followed by pairwise t-tests between baseline and post-treatment assessments to determine whether significant changes from baseline were present in the active group. Changes in cognitive and positive symptom measures were assessed within each group using paired t-tests. Due to the exploratory nature of these comparisons, reported effects on cognitive and positive symptom measures were not corrected for multiple comparisons.

Results

Primary Behavioral and Neurophysiological Outcomes

AX-CPT behavioral measures (Table 1) generally showed greater change in active compared to sham groups, with effect sizes ranging from d = 0.9 to 1.5. Change in d-prime differed between groups, where the active tDCS group increased d-prime to a greater degree than sham (t₍₈₎ = 2.3, p = .05, d = 1.5); however, d-prime did not significantly change from baseline within either group. Group differences in reaction time (t₍₈₎ = −1.1, p > .1, d = −0.9) and reaction time variability (t₍₈₎ = −2.1, p = .08, d = −1.3) change from baseline achieved large effect sizes; however, these effects did not achieve statistical significance. No significant group differences were observed at baseline.

Table 1.

Clinical and Cognitive Changes in Active (N = 6) and Sham (N = 4) Groups.

	Active tDCS + ACCT (ACE)			Sham tDCS + ACCT
	Baseline	Follow-up	t / p	Baseline	Follow-up	t / p
AX-CPT
d′	4.0 ± 0.4	4.4 ± 0.8	1.4 / .231	4.2 ± 1.3	3.6 ± 0.9	1.9 / .150
RT (ms)	416 ± 42	400 ± 70	−0.9 / .390	365 ± 95	432 ± 59	−0.9 / .416
RT SD (ms)	143 ± 31	128 ± 36	−0.6 / .578	115 ± 18	174 ± 60	−2.2 / .110
Beta (μV)
Evaluation
‘A’ Trials	−0.07 ± 0.03	−0.07 ± 0.03	0.1 / .918	−0.06 ± 0.03	−0.06 ± 0.03	0.0 / .986
‘B’ Trials	−0.05 ± 0.04	−0.13 ± 0.11	−2.5 / .056	−0.05 ± 0.02	−0.04 ± 0.11	−0.8 / .460
Preparation
‘A’ Trials	−0.05 ± 0.05	−0.02 ± 0.03	1.4 / .223	−0.01 ± 0.02	−0.02 ± 0.03	0.5 / .678
‘B’ Trials	−0.02 ± 0.02	−0.04 ± 0.09	−0.4 / .683	0.01 ± 0.03	−0.01 ± 0.02	1.3 / .261
ASSR (μV)
Attend	0.09 ± 0.03	0.38 ± 0.34	2.1 / .094	0.14 ± 0.08	0.11 ± 0.07	1.4/ .261
Ignore	0.11 ± 0.03	0.06 ± 0.03	−2.1 / .092	0.07 ± 0.01	0.12 ± 0.10	−0.8/ .481
MATRICS Consensus Cognitive Battery
Processing speed	46.3 ± 10.4	53.3 ±4.1	2.2 / .081	39.3 ± 14.2	37.8 ± 12.9	−0.3 / .772
Attention	45.3 ± 12.1	43.0 ± 14.1	−0.7 / .518	36.8 ± 12.5	34.3 ± 14.5	−0.9 / .417
Working memory	47.3 ± 10.2	48.2 ± 8,4	0.2 / .836	38.3 ± 17.7	44.0 ± 13.4	2.3 / .102
Verbal learning	42.7 ± 8,7	41.2 ± 10.7	−0.4 / .688	36.8 ± 8.7	32.8 ± 3.2	−1.4/ .267
Visual learning	44.5 ± 7.9	48.7 ± 10.4	1.3 / .240	34.3 ± 8.8	38.0 ± 9.0	0.9 / .431
Reasoning	50.3 ± 4.5	52.7 ± 7.6	0.9 / .433	50.5 ± 10.1	52.5 ± 10.1	1.4/ .267
Social cognition	45.2 ± 15.1	49.8 ± 16.9	2.6 / .050	28.0 ± 17.9	27.0 ± 11.9	−0.1 / .899
Total	43.3 ± 6.3	46.8 ± 7.3	2.2 / .084	30.0 ± 18.0	30.8 ± 13.5	0.3 / .764
Symptoms
PSYRATS
AH	27.7 ± 4.8	26.2 ± 6.7	−0.8 / .439	28.0 ± 6.9	26.3 ± 6.7	−0.6 / .599
Physical	2.7 ± 0.4	2.3 ± 0.7	−3.1 / .028	2.1 ± 0.6	2.0 ± 0.5	−0.5 / .638
Emotional	2.6 ± 0.5	2.6 ± 0.8	0.0 / 1.000	2.6 ± 0.8	2.6 ± 1.0	0.0 / 1.000
Cognitive	2.4 ± 0.6	2.4 ± 0.8	0.0 / 1.000	2.9 ± 0.6	2.6 ± 0.6	−1.1 / .342
Delusions	6.3 ± 7.5	7.2 ± 6.1	0.6 / .590	14.0 ± 3.8	11.0 ± 4.8	−3.7 / .035

Descriptive statistics are reported for clinical and cognitive variables assessed at baseline and follow-up. Baseline and Follow-up values represent Mean ± SD. Significant and marginal inferential statistics are bolded.

Beta desynchronization responses to “B” cues were identified in both groups at baseline and following ACCT within evaluation and preparation time windows (Figure 3). Repeated measures ANOVA revealed a significant interaction between group and time window (F_(1,8) = 6.1, p = .04), where the active group showed greater change in beta desynchronization from baseline than sham within the evaluation time window (t₍₈₎ = 2.8, p = .04, d = 1.5), but not the preparation window (p > .1, d = 0.2). Within the active group, change in beta desynchronization from baseline within the evaluation time window trended toward significance (t₍₅₎ = 2.5, p = .06, d = 1.0), with greater beta desynchronization after ACE than baseline. No significant group differences were observed at baseline.

Figure 3.

Beta desynchronization during AX-CPT. Spectrograms are shown for the interval between cue and target stimuli in the AX-CPT for active (N = 6) and sham (N = 4) groups at baseline and post-treatment. Responses to ‘A’ and ‘B’ cues are shown separately. Black rectangles represent time and frequency windows of interest for evaluation and preparation responses.

ASSR responses were identified in both groups, timepoints, and attention conditions (Figure 4). Group effects in ASSR change from baseline differed significantly between attend and ignore conditions (F_(1,8) = 5.6, p = .04). This interaction was related to both a trend-level effect indicating greater ASSR increase from baseline in the attend condition (t₍₈₎ = 2.2, p = .07, d = 1.2) and non-significant greater ASSR decrease in the ignore condition (t₍₈₎ = −1.5, p > .1, d = −1.1) in the active group compared to the sham group. Within the active group, ASSR changes from baseline trended toward significance with large effect size in the attend condition (t₍₅₎ = 2.1, p = .09, d = 0.84) and ignore condition (t₍₅₎ = −2.1, p = .09, d = −0.85). Although no significant group differences were observed at baseline, there was a trend toward greater ASSR in the ignore condition in the active group compared to the sham group (t₍₈₎ = 2.2, p = .06, d = 1.57)

Figure 4.

Auditory steady-state responses (ASSR). Spectrograms are shown for active (N = 6) and sham (N = 4) groups at baseline and post-treatment. ASSR in attend and ignore conditions are shown separately. Black rectangles represent time and frequency windows of interest for evaluation and preparation responses.

Secondary Cognitive and Clinical Outcomes

Changes in cognitive symptom measures assessed with the MCCB were identified only for the active stimulation group, where overall MCCB scores were increased following ACE (t₍₅₎ = 2.2, p = .08, d = 0.88). Subtest comparisons revealed that this change was largely due to increases in processing speed (t₍₅₎ = 2.2, p = .08, d = 0.88) and social cognition (t₍₅₎ = 2.6, p = .05, d = 1.05). Physical aspects of auditory hallucinations assessed with the PSYRATS were reduced following ACE (t₍₅₎ = −3.1, p = .03, d = −1.25); however, there was no change in cognitive or emotional subscales (Table 1). Conversely, delusion severity was reduced in the sham tDCS group (t₍₃₎ = −3.7, p = .04, d = −1.84), but not in the active group; however, baseline delusion severity was much lower in participants assigned to ACE (6.3) compared to ACCT + sham tDCS (14.0), suggesting a potential floor effect in the active group.

Discussion

This single-blind, sham-controlled, parallel-group pilot RCT with baseline and follow-up assessments provides preliminary evidence that pairing targeted tDCS over rVLPFC with Auditory Control Enhancement (ACE: ACCT + tDCS) can enhance neurophysiological indices of cognitive control in individuals with schizophrenia or schizoaffective disorder with persistent auditory hallucinations. Two converging electrophysiological findings support target engagement of cognitive control systems: (1) greater post-treatment increase in beta-band desynchronization to “B” cues during the AX-CPT in the evaluation window for the active group relative to sham, and (2) a significant group × attention interaction for gamma-band ASSR change, reflecting stronger attentional up modulation when attending and down-modulation when ignoring in the active group. Behaviorally, AX-CPT d′ and response time variability showed large and significant between-group effects favoring active stimulation, although within-group changes did not reach statistical significance in this small sample.

The beta findings are notable because cue-locked beta suppression indexes the engagement of inhibitory control and the reconfiguration of sensorimotor sets that underlie proactive control. The selective enhancement in the evaluation window suggests that ACE may strengthen early context appraisal and rapid set shifting—processes plausibly supported by rVLPFC, ACC/MCC, and pre-SMA within the broader auditory control network described in the Introduction.^{16–18,53,54} That the preparation window did not change significantly may indicate that initial cue evaluation (rather than sustained preparatory maintenance) is the most plastic stage during brief ACE dosing, or simply that longer training is required to propagate effects into later, more sustained control epochs.

The ASSR results complement this interpretation. Here, the active group exhibited a larger post-treatment divergence between attend and ignore conditions—driven by a trend toward enhanced gamma power when attending and reduced power when ignoring. This pattern is consistent with improved top-down gating of auditory cortex, a core function attributed to rVLPFC-centered control circuitry projecting to STG/TPJ. Given the montage (anode F8; cathode CP5), ACE likely biases excitability in right inferior frontal regions while concurrently modulating temporoparietal cortex implicated in AH generation.^17,42,43 Critically, stimulation occurred during task-engaged training, aligning with evidence that tDCS effects on plasticity are state-dependent and strongest when targeted circuits are actively recruited.^44–48 Together, the beta and ASSR findings point to enhanced coupling between control and auditory systems—precisely the mechanistic target posited to mitigate hallucination vulnerability.

Behavioral changes trended in the expected direction (higher d′, reduced RT variability), with large between-group effect sizes (d≈0.9–1.5) but limited statistical reliability. This pattern – physiology moving more robustly than behavior in early-phase, underpowered samples – is common in cognitive remediation and neuromodulation studies, where proximal neural markers often show greater sensitivity than composite behavioral outcomes over short intervals.^{34–36,45–48} Given only three one-hour ACCT sessions and a single-blind design, the observed neurophysiological effects are encouraging signals of target engagement that may precede and enable durable behavioral and clinical benefits with longer or more intensive treatment protocols.

Several limitations temper interpretation. The sample was small (N = 10) and heterogeneous (schizophrenia and schizoaffective disorder), constraining power and generalizability. Single-blind procedures introduce potential expectancy bias. The brief intervention may have been insufficient to elicit reliable behavioral or symptom change, and multiple statistical tests raise family-wise error concerns. Montages using large sponge electrodes afford limited focality; future work could leverage individualized current flow modeling and high-definition tDCS to optimize engagement of auditory control systems. Although participants in this study were stable on medications with persistent symptoms, it is unknown whether ACE interacted with antipsychotic medication effects and whether this interaction may have impacted the results; future studies with larger samples might benefit from comparison between different medication types/classes to determine whether such interactions may exist. Hallucination-related changes were limited and domain-specific, meaning broader symptom improvement cannot be concluded from the present pilot data. Finally, this report focused on neural and cognitive endpoints; relating physiological changes directly to hallucination severity will be important for establishing clinical relevance.

These caveats notwithstanding, the present findings provide an initial mechanistic proof-of-concept for ACE: pairing rVLPFC-targeted, state-dependent tDCS with focused auditory cognitive control training can modulate beta-indexed inhibitory control and restore attentional gating of gamma responses in auditory cortex. Larger, double-blind trials with extended dosing, symptom endpoints (eg, PSYRATS-AH), and long-term follow-up are warranted. Incorporating neurophysiological bio-markers (beta ERD and ASSR attention effects) as proximal targets could guide adaptive dosing and personalize montage selection. If replicated, ACE could represent a scalable, outpatient approach to strengthen auditory cognitive control circuitry and, ultimately, reduce the burden of treatment-resistant auditory hallucinations in early psychosis.

Data Availability

Data presented in this manuscript is available by written request to Brian A Coffman, PhD.