Beta and gamma binaural beats enhance auditory sentence comprehension

Binaural beats—an auditory illusion produced when two pure tones of slightly different frequencies are dichotically presented—have been shown to modulate various cognitive and psychological states. Here, we investigated the effects of binaural beat stimulation on auditory sentence processing that required interpretation of syntactic relations (Experiment 1) or an evaluation of syntactic well formedness (Experiment 2) with a large cohort of healthy young adults (N = 200). In both experiments, participants performed a language task after listening to one of four sounds (i.e., between-subject design): theta (7 Hz), beta (18 Hz), and gamma (40 Hz) binaural beats embedded in music, or the music only (baseline). In Experiment 1, 100 participants indicated the gender of a noun linked to a transitive action verb in spoken sentences containing either a subject or object-relative center-embedded clause. We found that both beta and gamma binaural beats yielded better performance, compared to the baseline, especially for syntactically more complex object-relative sentences. To determine if the binaural beat effect can be generalized to another type of syntactic analysis, we conducted Experiment 2 in which another 100 participants indicated whether or not there was a grammatical error in spoken sentences. However, none of the binaural beats yielded better performance for this task indicating that the benefit of beta and gamma binaural beats may be specific to the interpretation of syntactic relations. Together, we demonstrate, for the first time, the positive impact of binaural beats on auditory language comprehension. Both theoretical and practical implications are discussed.

A growing body of studies have demonstrated that binaural beat stimulation improves cognitive performance as well as psychological states (see Chaieb et al., 2015 andGarcia-Argibay et al., 2019a for a review and a meta-analysis respectively). For instance, Beauchene et al., (2016Beauchene et al., ( , 2017 showed that participants performed better on a visuospatial and an N-back working memory task while listening to a beta (15 Hz) binaural beat compared to silence. They also found that cortical connectivity along the frontoparietal network increased when listening to the 15 Hz binaural beat (Beauchene et al., 2016(Beauchene et al., , 2017. In addition, beta binaural beats have been shown to facilitate long-term memory (20 Hz: Garcia-Argibay et al., 2019b) and vigilance (16-24 Hz;Lane et al., 1998). Other studies have reported the positive effects of gamma binaural beats on attention (40 Hz: Colzato et al., 2017;Engelbregt et al., 2019Engelbregt et al., , 2021Reedijk et al., 2015;Ross & Lopez, 2020), theta binaural beats on verbal memory (5 Hz: Ortiz et al., 2008) and theta or alpha binaural beats on anxiety (9 Hz: Isik et al., 2017;4-7 Hz: Mallik & Russo, 2022;10 Hz: Wiwatwongwana et al., 2016). Nevertheless, there are some studies reporting lack of effects of binaural beats (e.g., a significant effect on accuracy but not on reaction time; Engelbregt et al., 2021). All in all, the binaural research field is still at its infancy necessitating more rigor for replication and corroboration.
In the present behavioral study, we investigated whether binaural beats facilitate auditory sentence processing, a cognitive domain that has never been explored with this noninvasive technique. We recruited 200 young adults across two language experiments (Fig. 1C), each of which entailed different syntactic operations while listening to a series of spoken sentences (see below for more detail). In both experiments, participants were presented with slow, non-rhythmic music mixed with or without binaural beats for 10 min prior to performing a language task (Fig. 1B, D). The exposure duration was chosen based on a recent study showing that 9 min of binaural beat stimulation induced connectivity changes across cortical regions (Ala et al., 2018).

Experiment 1
After the binaural beat stimulation, the participants underwent a language task involving syntactic interpretation of the noun-verb relation in a series of spoken sentences (Table 1). These sentences were adopted and modified from a previous study ) that contained either a centerembedded subject-(e.g., 'Kings that help queens are nice') or object-relative clause (e.g., 'Kings that queens help are nice') (see Methods for more details). Object-relative sentences are more difficult than subject-relative sentences to comprehend due to the syntactically non-canonical structure (MacDonald & Christiansen, 2002;Wells et al., 2009) and higher demand on working memory (Just & Carpenter, 1992;King & Just, 1991).
We used binaural beats of the theta (7 Hz), beta (18 Hz), and gamma (40 Hz) frequency bands because each of these bands has been associated with different aspects of sentence comprehension processing in previous research. For example, beta and gamma oscillations are thought to mediate core language functions-syntactic and semantic operations respectively-during sentence comprehension (Prystauka & Lewis, 2019). Bastiaansen et al., (2010) demonstrated a gradual increase in beta power over the course of syntactically well-formed sentences compared to randomly organized word sequences. Also, increased gamma power has been observed in response to semantically congruent sentences, but not to sentences with semantic anomalies (Hald et al., 2006). Critically, Bastiaansen and Hagoort (2015) demonstrated a functional segregation between beta and gamma frequencies in a single study wherein beta power was increased in response to syntactically correct sentences compared to incorrect sentences while gamma power was increased in response to semantically coherent sentences compared to meaningless sentences. Theta power has been shown to increase with a greater working memory load (Jensen & Tesche, 2002;Krause et al., 2000) and during on-line sentence processing (Bastiaansen et al., 2002(Bastiaansen et al., , 2010 likely due to its domain-general role in working memory during sentence comprehension (Bastiaansen & Hagoort, 2003).
Based on the implications of neural oscillations in the theta, beta, and gamma frequency bands during language comprehension, and especially with sentences containing object-vs. subject-relative clauses (Weiss et al., 2005), we hypothesized that binaural beats in any, if not all, of these frequency bands would enhance sentence comprehension performance. In particular, we expected to observe a more pronounced impact of these binaural beats in the more difficult object-relative, compared to subject-relative, sentences.

Participants
One hundred undergraduate students from the University of Texas at Dallas (63 females, 36 males, and 1 other/not specified, 18-37 years, mean age = 21.6 years, SD = 3.5 years) participated in Experiment 1 for course credits or monetary compensation ($15 gift card). All of the participants were native speakers of American English, had normal vision and hearing, and had no known history of cognitive, developmental, or neurological disorders. They consented to participating in the study, which was approved by the University of Texas at Dallas Institutional Review Board (IRB-21-109).

Stimuli and procedures
All sound stimuli were presented at a preset volume (70 dB SPL) via Sennheiser HD-280 headphones. For the binaural beat stimulation, a 250 Hz pure tone was presented to the right ear as a carrier frequency while another pure tone with 257 Hz, 268 Hz, or 290 Hz was presented to the left ear, eliciting theta (7 Hz), beta (18 Hz), and gamma (40 Hz) binaural beats, respectively (Fig. 1B). The two pure tones were mixed with an excerpt of slow-tempo, non-rhythmical music (Dangol, 2019) at a signal-to-music ratio of -2 dB (Online Resource 1) to help participants endure prolonged exposure to binaural beats. In the baseline condition, the same music was played without binaural beats. Auditory sentence stimuli were generated using the Google Text-to-Speech and the speaker voice was set to an American-English speaking male (Online Resource 2). Experiments were conducted using Matlab R2021 (Mathworks, MA) in a dimly lit sound-proof booth.
For the sentence comprehension task, the stimuli were comprised of 128 spoken sentences, each of which consisted of six words: a male noun (e.g., boys, uncles, kings), a female noun (e.g., girls, aunts, queens), a gender-neutral noun (e.g., children, students, doctors), a relative pronoun 'that', a transitive action verb (e.g., help, hug, bully), and one of four transitive preference verbs: love, adore, hate, and dislike. Each sentence contained either a subject or objectrelative center-embedded clause, which was solely determined by switching the temporal order of the same noun and verb in the relative clause within the sentence (Table 1). For each sentence trial, participants indicated the gender of the individuals performing an action, while disregarding those who love/adore/hate/dislike others, by pressing either the 'male' (left arrow) or 'female' (right arrow) key within 5 s. The sentence type (i.e., subject or object-relative) and the gender of the agent (i.e., female or male) were counterbalanced across trials.
Participants were randomly assigned to one of four conditions ( Fig. 1C): music + theta (7 Hz), music + beta (18 Hz), music + gamma (40 Hz), or music only (N = 25 for each The embedded clause is underlined, and the target action verb is in bold face Sentence type Example Correct answer Subject relative Gentlemen that assist ladies adore children Male Gentlemen that adore ladies assist children Male Object relative Gentlemen that ladies assist adore children Female Gentlemen that ladies adore assist children Male condition). Participants were first familiarized with the language task by undergoing 16 practice trials, during which they received feedback immediately after each response. Sentences used for the practice session were not presented during the main experiment. After the practice was completed, participants listened to each of these 4 sounds for 10 min while fixating their eyes onto a speaker icon on the screen. Participants were not explicitly informed about the presence of a binaural beat embedded in the music. Immediately after listening to the sound, participants underwent 64 trials of the sentence comprehension task (block1). A 15 s break was provided every 16 trials. This procedure was repeated with another 64 trials (block2). No feedback was provided during the two task blocks. Together, the duration of the experiment ranged from 50 to 60 min in total, including the tutorial/practice (10 min), binaural beat stimulation (2 × 10 = 20 min), and the main task (2 × 10-15 min per block = 20-30 min) (Fig. 1D).

Data analysis
We opted to use a logistic regression model to analyze the binomial accuracy data. The trial-by-trial accuracy data were entered into a mixed effects logistic regression model using the glmer function of the lme4 package (Bates et al., 2015) in R 4.2.1. The model included the factors of sentence type (i.e., subject-and object-relative), binaural beat condition (i.e., theta, beta, gamma, and music only), and an interaction between the two factors as fixed effects. In addition, the participant factor was included as a random intercept. The no-response trials or those with late (> 5 s)-responses were coded as incorrect (0.35% of the entire trials across participants). The statistical significance of the fixed effects was assessed using the Type III Wald chi-square tests of the car package (Fox et al., 2012). Upon a significant interaction effect (see below), we examined differences in accuracy within object-relative sentences between the music-only group and each of the three binaural beat groups using a one-sided general linear hypothesis test (glht) of the multcomp package (Hothorn et al., 2016). The multiple comparisons were corrected using the Dunnett's single-step method (Dunnett, 1955). Reaction time was not analyzed given large individual differences in accuracy.
To further examine the significant interaction effect, we compared performance between the baseline group and each of the binaural beat groups within object-relative sentences. The results showed higher accuracy in the beta (

Experiment 2
The findings in Experiment 1 naturally invite the following questions: could the binaural beat effect be generalized to different kinds of syntactic operations during auditory sentence comprehension? That is, would either beta or gamma binaural beats turn out to be significant when another syntactic task is used? To address this, we constructed another set of object-relative and subject-relative sentences, half of which contained a morpho-syntactic error on subject-verb agreement or verb tense (Table 2). For this type of syntactic analysis, both theta and beta frequency bands have been implicated in previous research, in which the presence of a syntactic violation elicited increased oscillatory power in the theta band and decreased power in the beta band (Kielar et al., 2015;Lewis et al., 2016;Pérez et al., 2012). While the increase of theta power is thought to reflect a greater working memory demand in processing sentences  Fig. 2 The accuracy of the sentence comprehension task on object and subject-relative sentences in the three binaural beat conditions and the music-only condition. Circles indicate individual data points. Participants in the beta and gamma binaural beat condition, but not theta, exhibit significantly better performance for object-relative sentences than those in the baseline music-only condition with syntactic violations (Prystauka & Lewis, 2019), the decrease of beta power has been interpreted as reflecting a disruption of building up syntactic structures. In addition, decreased gamma power has been associated with the presence of a semantic anomaly within syntactically well-formed sentences (e.g., 'The hose can bake water on the flowers') (Penolazzi et al., 2009;Rommers et al., 2013;Wang et al., 2012). Thus, we considered the following scenarios: (1) Based on the previous literature, theta and/or beta binaural beats would enhance grammaticality judgment performance.
(2) Because of the minimal influence of morpho-syntactic errors (e.g., go vs. goes) on the overall meaning of the sentence, we expected that a gamma binaural beat may yield little-to-no effects.

Participants
Another one hundred undergraduate students from the University of Texas at Dallas (67 females and 33 males, 18-27 years, mean age = 19.9 years, SD = 1.8 years) participated in Experiment 2 for course credits or monetary compensation ($15 gift card). None of the participants had participated in Experiment 1. All the participants were native speakers of American English, had normal vision and hearing, and had no known history of cognitive, developmental, or neurological disorders. They consented to participating in the study, which was approved by the University of Texas at Dallas Institutional Review Board (IRB-21-109).

Stimuli and procedures
The binaural beat stimuli and procedures were identical to those used in Experiment 1, except for the language materials and task. In Experiment 2, the language stimuli were comprised of 96 sentences, each of which contained either a subject-or object-relative clause and a time adverb phrase (e.g., every week, last year) ( Table 2). Half of the sentences were grammatical while the other half were ungrammatical, containing a morpho-syntactic error. There were three types of violations: a subject-verb agreement error within the relative phrase, a subject-verb agreement error in the main phrase, and a past tense error ( Table 2). The sentence and error types were counterbalanced across trials. For each trial, participants were instructed to indicate whether or not a given sentence was grammatically correct by pressing either the 'correct' (right arrow) or incorrect' (left arrow) key. There was no time constraint in this task to ensure obtainment of d' scores from every trial. Reaction time was not analyzed given large individual differences in accuracy. Participants were randomly assigned to one of three binaural beat conditions and a baseline condition: music + theta (7 Hz), music + beta (18 Hz), music + gamma (40 Hz), or music only (N = 25 for each condition) (Fig. 1B, C). Participants received 8 practice trials, during which they were given feedback immediately after each response. The sentence materials used for the practice were not included in the main experiment. After listening to the music, the participants underwent 48 trials of the grammaticality judgment task (block1). Participants took a 15 s break after every 12 trials. The task duration, including breaks, ranged from 10 to 15 min. This procedure was repeated with another round of the grammaticality judgment task consisting of 48 trials (block2). No feedback was provided during the actual task (Fig. 1D). The experiment took approximately 45 to 55 min before completion, including the practice (5 min), binaural beat stimulation (2 × 10 = 20 min), and the main task (2 × 10-15 min per block = 20-30 min) (Fig. 1D).

Data analysis
We computed a d-prime (d') score for each sentence type. A d′ score was calculated by subtracting the z-scores of false alarm rate from the z-scores of hit rate. To prevent an indefinite d' score, hits and false alarms of 0 and 1 were replaced with 1/24 and 23/24 respectively (Macmillan & Kaplan, 1985). The resulting data were submitted to a linear mixed effects model analysis using the lmer function of the lme4 package with fixed effects of sentence type and binaural beat condition, as well as a random intercept of participant. The embedded clause is underlined, and the morpho-syntactic violation is shown as italic in the parenthesis. SVA subject-verb agreement Sentence type Example Type of error Subject relative The customer that tips (tip) the waitress saddles horses every day SVA (relative) Every year, the criminals that avoid the police go (goes) to the jail SVA (main) Yesterday, the father that keeps the boys wanted (wants) toothpaste Tense Object relative The student that the nephew trusts (trust) fixed the mistake yesterday SVA (relative) Every week, the ladies that the baby loves watch (watches) the movie SVA (main) The animal that the children dislike hunted (hunts) a prey last month Tense

Results
Consistent with the Experiment 1, there was a significant main effect of sentence type on d' [F(1,99) = 31.36, p < 0.001], with lower accuracy in object-relative (M = 1.19) compared to subject-relative (M = 1.51) sentences (Fig. 3). Although the group mean accuracy was slightly higher in the three binaural beat groups than in the baseline group, there was neither a main effect of binaural beat condition [F(1,99) = 1.37, p = 0.255] nor a significant interaction [F(3,99) = 0.21, p = 0.892]. Thus, the binaural beat effects observed in experiment 1 did not translate to the grammaticality evaluation task used in experiment 2.

Discussion
In the present study with a large cohort of young adults (N = 200), we examined the impact of binaural beats on auditory sentence processing, a hitherto unexplored research domain. In Experiment 1, participants were presented with sentences containing either a subject or object-relative clause, which required a prompt interpretation of syntactic relations between noun and verb phrases in a given sentence. As expected, object-relative sentences yielded lower accuracy in comparison to subject-relative sentences due to their non-canonicity (Wells et al., 2009) and/or higher working memory load (King & Just, 1991). Importantly, we found that the beta and gamma binaural beats significantly enhanced the task performance for the object-relative sentences relative to the baseline. Experiment 2 was subsequently designed to see if the binaural beat effect could generalize to another linguistic operation (i.e., grammaticality judgment), which yielded non-significant results. This suggests that the effects of binaural beats on auditory sentence processing are rather specific. In what follows, we discuss the theoretical and practical implications of the current findings.

The functional role of beta binaural beat stimulation on auditory sentence processing
Beta binaural beat stimulation significantly enhanced auditory sentence comprehension involving syntactic interpretation of the relationship between an action verb and noun in a sentence, especially in the non-canonical object-relative sentences in Experiment 1. Indeed, beta band oscillations have been implicated in on-line sentence processing that involves various syntactic operations. For example, increased beta power has been observed while reading syntactically well-formed sentences (Bastiaansen & Hagoort, 2015). Also, object-relative sentences have been shown to elicit greater beta coherence compared to subject-relative sentences (Weiss et al., 2005), which may reflect a considerable processing burden towards analyzing the non-canonical syntactic structure of object-relative sentences. Our findings suggest that the beta binaural beat may have increased the beta power and/or coherence during sentence processing, which could help with the difficulty of analyzing objectrelative sentences-an important follow-up question raised by the current behavioral finding. It is also conceivable that the beta binaural beat might have enhanced verbal working memory (Beauchene et al., 2017). Note that there are two grammatical agents that must be maintained in working memory for objective-relative sentences, whereas only a single agent is present in subjectrelative sentences (King & Just, 1991; see Table 1). Relatedly, increased beta power has been elicited by listening to sentences with a long compared to a short subject-verb distance, indicating a consequence of working memory demand on beta frequency oscillations during sentence processing (Meyer et al., 2013). This explanation is consistent with a recent theoretical account that the increase in beta oscillations during sentence processing reflects an active maintenance of the ongoing sentence-level representations (Lewis et al., 2015;Weiss & Mueller, 2012). However, the working memory account may not be fully compatible with the findings in Experiment 1, given the lack of the beta binaural beat effect in Experiment 2, which also required listeners to maintain linguistic elements in a sentence (McDonald, 2000;Noonan et al., 2014). Although future studies are warranted, the beta binaural beat may have enhanced syntactic interpretation, rather than the domain-general working memory.

The functional role of gamma binaural beat stimulation on auditory sentence processing
We also found that the gamma binaural beat yielded performance enhancement during sentence comprehension  Fig. 3 The accuracy of the grammaticality judgment task in the three binaural beat conditions and the music-only baseline condition. Circles indicate individual data points. Performance was comparable across all the conditions in Experiment 1. Gamma band oscillations have been frequently associated with semantic components of sentencelevel language processing, evidenced by decreased gamma power in response to semantically anomalous words (Penolazzi et al., 2009;Rommers et al., 2013;Wang et al., 2012). Similarly, Bastiaansen and Hagoort (2015) reported that semantically congruent sentences elicited increased gamma power compared to semantically uninterpretable sentences or word lists, leading them to conclude that gammaband activities support semantic unification. In Experiment 1, although the agent linked to an action verb in a given sentence was determined solely by its syntactic structure (i.e., subject or object-relative), the task would inevitably require lexical-semantic processing of the nouns and verbs. Furthermore, each sentence might differ in the degree of semantic plausibility, which could have affected the interpretation of object-relative sentences (e.g., Gennari & MacDonald, 2008). As such, these semantic analyses may have benefited from gamma binaural beat stimulation. It is also plausible that the gamma binaural beat, along with beta, may have led to improved syntactic processing given both semantic and syntactic operations are often intertwined and influence each other (e.g., Gámez & Vasilyeva, 2015;Gunter et al., 2000). Unfortunately, the current sentence stimuli cannot allow us to dissociate the functional role of gamma frequency on semantic from syntactic processes, which necessitates a follow-up study. Another possibility is that the gamma binaural beat may have provided an attentional boost (Fell et al., 2003;Fries, 2015;Lakatos et al., 2008). Indeed, gamma binaural beats have been used to improve performance on various attention tasks, including a visual feature binding task (Colzato et al., 2017), an attentional blink task (Reedijk et al., 2015;Ross & Lopez, 2020), the Flanker task (Engelbregt et al., 2021), and a visual task in a dual-task paradigm (Hommel et al., 2016). For example, Engelbregt et al., (2021) exposed participants to a gamma binaural beat or pink noise while performing the Flanker task, which gauged their ability to selectively attend to a target stimulus in the presence of distractors. They found that the gamma binaural beat significantly reduced the number of incorrect responses compared to pink noise, indicating an enhancement of selective attention. Nevertheless, the lack of efficacy of the gamma binaural beat in Experiment 2 may weaken this interpretation.

The selective impact of beta and gamma binaural beats on auditory sentence processing
Our findings suggest that the beta and gamma binaural beats have functional specificity since they only enhanced syntactic interpretation of the noun-verb relation in Experiment 1, not morpho-syntactic error detection in Experi-ment2. Indeed, we have recently reported dissociable neuroanatomical systems devoted to different subsets of syntactic processing (Heard & Lee, 2020). Such dissociation within the language system was previously reported in clinical research as well. For example, aphasic patients preserved the ability to detect grammatical errors in sentences despite severe comprehension deficits (Linebarger et al., 1983;Wulfeck, 1988). Conversely, patients with cerebellar lesions were less impaired in sentence comprehension, compared to their grammaticality judgment ability (Justus, 2004). The present results are in line with the view that constructing a syntactic representation for grammaticality judgment is dissociable from the analysis of syntactic relations between words for sentence comprehension (Linebarger et al., 1983).

The functional role of theta binaural beat stimulation on auditory sentence processing
Although the theta binaural beat yielded numerically higher accuracy than the music-only baseline condition, the difference did not reach statistical significance in both experiments. Theta oscillations have been related to working memory functions, e.g., increased theta power is often associated with a greater working memory load (Hsieh & Ranganath, 2014;Jensen & Tesche, 2002;Scheeringa et al., 2009). Indeed, the theta binaural beat has been used to enhance participants' immediate recall of verbal items (Ortiz et al., 2008), leading us to hypothesize that the theta binaural beat would improve auditory sentence processing by facilitating the domain-general processing. However, theta was not as effective as beta or gamma in auditory sentence comprehension, negating the working memory hypothesis. Similar to our data, a past study showed that the theta binaural beat failed to improve performance on an N-back working memory task (Beauchene et al., 2017). Note, however, that this study used 5 min of theta binaural beat stimulation. As mentioned earlier, we chose 10 min of binaural beat stimulation based upon the evidence that a 9-min duration was sufficient to induce neural entrainment (Ala et al., 2018;Jirakittayakorn & Wongsawat, 2017). While our study failed to show a theta binaural beat effect, Ortiz et al. (2008) found a significant improvement in verbal memory using 15 min of theta binaural beat stimulation. A future study should consider employing various lengths of theta binaural beat stimulation to systemically determine the duration effect.

Other considerations
In the current study, we used binaural beats embedded in a slow, non-rhythmic music unbeknown to participants. Similarly, a recent study showed that theta binaural beat stimulation combined with personalized music reduced anxiety (Mallik & Russo, 2022). Importantly, however, they did not find a binaural beat effect without music, suggesting 1 3 an interaction between binaural beats and music. From one point of view, these findings suggest that binaural beats can be practically used along with music as a non-invasive sound therapy. However, the interaction effect must be explained by follow-up experiments that include a binaural beat only condition.
Another important investigation is to determine how long the after-effects of binaural beat stimulation would last. The current study was not designed to examine the sustainability of the binaural beat effects. From a therapeutic perspective, this is a critical and timely question. It would be informative in future studies to establish the duration of binaural beat effects, as is the case with the application of neurostimulation (Kasten et al., 2016;Neuling et al., 2013;Reinhart & Nguyen, 2019;Wischnewski et al., 2019).

Conclusion
In sum, we presented novel evidence that beta and gamma binaural beats positively impact performance on an auditory language comprehension task involving the syntactic analysis between a noun and a verb at the sentence level. Notably, these binaural beats helped overcome the difficulty of comprehending sentences with a non-canonical syntactic structure (i.e., object-relative). Nevertheless, such binaural beat effects did not generalize to another language task involving morpho-syntactic error detection. Future research is needed to elucidate the role of beta and gamma binaural beats on complex sentence comprehension that entails both core linguistic operations (e.g., syntax and semantics) and domaingeneral processing (e.g., working memory). Together, our behavioral study establishes the first step towards understanding the impact of binaural beats on auditory sentence processing, which can potentially serve as a novel therapeutic means for treating language disorders.