Sensorimotor synchronization refers to the rhythmic co-ordination of perception and action (Repp, 2005; Repp and Su, 2013), and is most often studied using finger tapping to a rhythmic sequence of auditory stimuli. Such rhythmic tapping is hereafter referred to as beat synchronization. The temporal co-ordination of rhythmic movement with an external rhythm tends to improve with age (e.g., Drake et al., 2000; McAuley et al., 2006), and is reliably enhanced with musical training (e.g., Thompson et al., 2015). The precision of beat synchronization is reduced in children with developmental dyslexia (Overy et al., 2003; Thomson and Goswami, 2008; Flaugnacco et al., 2014). In studies of child populations, the precision of beat synchronization is related to individual differences in both phonological awareness (children's ability to reflect on the sound structure of words, a cross-language predictor of the efficiency of reading acquisition) and to individual differences in progress in reading (e.g., Dellatolas et al., 2009 [French]; Corriveau and Goswami, 2009 [English]; Flaugnacco et al., 2014 [Italian]). Individual differences in beat synchronization are also related to measures of reading readiness (Woodruff Carr et al., 2014, English).

The mechanisms underlying the relationships between individual differences in beat synchronization, reading, and phonological awareness in children are currently unclear. To explore these relationships in more detail, here we adapt an EEG paradigm developed by Nozaradan and her colleagues for measuring selective neuronal entrainment to beat and meter based on steady-state evoked-potentials (SS-EPs, see Nozaradan et al., 2012, 2015). Nozaradan and colleagues were able to measure the emergence of an internal (neural) representation of beat at beat and meter frequencies that were not physically present in the stimulus. For example, Nozaradan et al. (2015) recorded EEG while adult participants listened to a 2.4 Hz metronome beat and tapped on every second beat (1.2 Hz). Neurally, both a 1.2 Hz SS-EP related to motor entrainment to the beat and a 2.4 Hz SS-EP related to auditory beat-related entrainment were observed, as would be expected. Crucially, an interaction between the sensory (2.4 Hz) and motor (1.2 Hz) SS-EPs was evidenced by an additional SS-EP at 3.6 Hz, compatible with sensorimotor integration and the internal representation of a beat. This approach offers a way of separating the sensory (auditory) and motor (manual) components of beat synchronization in child populations. Our question in the current study was whether children with developmental dyslexia, who are known to show deficits in both perceiving auditory rhythm (e.g., Huss et al., 2011) and in tapping tasks (e.g., Thomson and Goswami, 2008) would show neural differences compared to typically-developing children in either the motor (1.2 Hz SS-EP) or the auditory (2.4 Hz SS-EP) domain, or in both domains. A group difference for the 3.6 Hz SS-EP would suggest that sensorimotor integration and the internal representation of the beat is impaired in developmental dyslexia.

When neural measures are used to index developmental mechanisms that may underlie impaired performance by children with developmental disorders, it is important that the groups being compared are matched on the behavior under study (Shaywitz et al., 2007; Olulade et al., 2013). For example, phonological processing is known to be impaired in children with dyslexia, so to measure atypical neural processing during a phonological task, the children with dyslexia must be matched in performance to the typically-developing children in the task being studied (usually achieved via a reading-level match group design, see Olulade et al., 2013). In the current study, matching the children with dyslexia for tapping behavior to the typically-developing controls permits the inference that group differences in beat-related neural activity reflect developmental differences rather than differences in sensorimotor expertise. Accordingly, in the current study we took advantage of an ongoing training study in our laboratory in which children with developmental dyslexia received practice in tapping to a 2 Hz beat on a weekly basis (as part of a larger battery of phonological remediation activities). The children who volunteered for EEG were matched in their behavioral tapping performance to the age-matched control children, and both groups were then compared in Nozaradan et al.'s (2015) task. On the basis of our prior data (Power et al., 2013), we expected that neural differences in preferred phase in beat-related auditory entrainment would drive any group differences that we might observe. As our prior EEG studies revealed auditory neural differences between children with dyslexia and control children in the time domain rather than in the frequency domain, no a priori prediction was made regarding whether group differences may be found in the SS-EPs.

For children with developmental dyslexia, it is unknown whether their impairments in beat synchronization arise from a primary sensory deficit in auditory rhythm perception which in turn affects the temporal precision of action, or whether beat synchronization is impaired via an independent deficit related to the developing motor system, or both. As impairments in both rhythm perception and the precision of beat synchronization are found in a range of developmental disorders of language, including stuttering (Falk et al., 2015; Wieland et al., 2015) and speech and language impairment (Corriveau and Goswami, 2009; Cumming et al., 2015) as well as developmental dyslexia, it is of theoretical interest to ascertain the sensory and/or motor sources of impaired sensorimotor synchronization in children with different disorders. The identification of underlying neural mechanisms would help to optimize remediation. Synchronizing to a beat is a complex process, requiring accurate internal time-keeping, the production of steady movements, and the use of auditory feedback to predict and correct the timing of action (Sowinski and Dalla Bella, 2013; Cason et al., 2015; Tierney and Kraus, 2016). To take the example of dyslexia, if the primary sensory impairment is auditory rather than motor, then rhythmic interventions involving a motor component (e.g., drumming) may improve rhythmic synchronization in affected children via sensorimotor coupling, with consequent effects on auditory and language processing driven by the motor practice (Goswami, 2012). Linguistic improvements are indeed found for rhythmic interventions with children that involve drumming, or other forms of rhythm production (tapping with a pencil), or playing a musical instrument (e.g., Degé and Schwarzer, 2011; Bhide et al., 2013; Slater et al., 2014; Flaugnacco et al., 2015; Serrallach et al., 2016). For example, in Bhide et al. (2013), the rhythmic intervention (based on bongo drumming) improved children's discrimination of amplitude envelope “rise time,” phonological awareness and reading. Bhide et al. (2013) interpreted this cross-domain (motor-to-language) enhancement in terms of Temporal Sampling theory (Goswami, 2011).

Temporal Sampling (TS) theory has been developed as a sensory/neural explanatory framework for the linguistic (phonological) deficits that characterize children with developmental dyslexia across languages. TS theory proposes that phonological difficulties in developmental disorders of language may be related to atypical neural entrainment by auditory oscillatory networks to the rhythms of speech (the syllable “beats”) that are carried by slower amplitude modulations (AMs) <10 Hz. This atypical neural entrainment is thought to be related to impaired sensory discrimination of amplitude envelope (AE) rise times. Impaired discrimination of AE rise times is found in children with developmental dyslexia in a range of languages (English, French, Dutch, Spanish, Chinese, Hungarian, and Finnish, see Goswami, 2015, for a summary). Amplitude envelope rise time is an important perceptual correlate of rhythm. The onsets of successive modulations in the amplitude envelope (AE) of a sound, and the rates of change of these modulation onsets (AE rise times), are a key temporal component of rhythmic perceptual structure for both speech and music (e.g., Greenberg et al., 2003; Patel, 2010, 2012; Goswami, 2011; Giraud and Poeppel, 2012). As well as showing consistent relations with phonological awareness, the impairments in AE rise time discrimination in dyslexic populations are also usually linked behaviorally to impaired non-speech rhythmic processing, for example as measured by musical tasks requiring sensitivity to metric structure (see Huss et al., 2011; Goswami et al., 2013; Flaugnacco et al., 2015).

By TS theory, these well-documented behavioral relationships between AE rise time discrimination, rhythm discrimination, and phonological development are linked to atypical oscillatory neuronal entrainment to the AM patterns in speech and music that carry rhythm. Neural studies in adult populations have shown that speech is encoded in part via the entrainment of neuronal oscillations in auditory cortex by AMs in the speech signal at multiple temporal rates simultaneously [delta (~1–3 Hz), theta (~4–8 Hz), beta (~15–30 Hz), and gamma (>30 Hz); see (Giraud and Poeppel, 2012; Poeppel, 2014)]. These oscillating neural networks phase-reset their activity to phase-align or entrain with AMs in the speech signal, using AE rise times as acoustic landmarks for accurate alignment (see Gross et al., 2013; Doelling et al., 2014). As children with developmental dyslexia show impaired discrimination of AE rise times, their phase resetting process is likely to be atypical. Indeed, Power et al. (2013) demonstrated a different preferred phase in delta-band entrainment (~1–3 Hz) in children with dyslexia aged on average 13 years compared to control children in a rhythmic speech listening task (to syllable repetition at 2 Hz). Power et al.'s data suggested neuronal phase alignment to less informative temporal points in the incoming speech signal by children with dyslexia, which would affect the extraction of phonological information.

Interestingly, studies of beat synchronization in children reveal that 2 Hz (500 ms) is the preferred spontaneous tempo (the rate at which children choose to tap in the absence of an external timekeeper, see McAuley et al., 2006, see also Moelants, 2002). Accordingly, by TS theory, the impaired synchronization to the beat found in children with developmental dyslexia may arise from atypical auditory phase entrainment to acoustic rhythm in the delta band, which in turn affects sensorimotor synchronization. Related studies reveal that children with developmental dyslexia also show atypical neural entrainment to continuous speech in the delta band (centered on ~2 Hz, Power et al., 2016, sentence repetition, English; Molinaro et al., 2016, story comprehension, Spanish).

A complementary theoretical framework that is relevant to the development of accurate beat synchronization, but based on the mechanisms governing attention rather than linguistic processing, is Dynamic Attending Theory (DAT; Large and Jones, 1999; Jones, 2008). DAT proposes that better sensory processing should be found when the brain can utilize a regular rhythm for temporal prediction. By DAT, if an auditory rhythm is present, such as a 2.4 Hz beat train, then this can be thought of as an external oscillator at 2.4 Hz, with which hypothetical internal oscillators (neural rhythms with similar phase and period relations) can synchronize, thereby coupling their neural activity with the external oscillator. By DAT, successful rhythmic entrainment would dynamically modify attention in time, with more attentional resources being allocated to informative positions, such as the temporal point at which the beat occurred. As an example, a network of neurons whose preferred oscillatory rate was 2.4 Hz (the oscillations reflect the concentration of neuronal electrical discharges to particular phases of a temporal cycle) would entrain to an external auditory rhythm at 2.4 Hz, so that maximal neural activity (peak phase) coincided with the beat. Perception of other sensory information (e.g., a flash of light) that occurred during this enhanced attentional phase would then be optimized. Neural networks that oscillate quasi-rhythmically at different temporal rates are found in many areas of the brain (e.g., Buzsáki, 2004; Lakatos et al., 2008; Schroeder and Lakatos, 2009). It has further been demonstrated that these oscillators indeed affect attention. For example, adults are unaware of visual stimuli that occur during the trough (least excitable phase) of a parietal alpha oscillation (~10 Hz) and are most likely to detect targets at the oscillatory peak (Mathewson et al., 2009).

Although DAT focuses on attention and TS theory focuses on linguistic processing, both theoretical frameworks predict that children who can entrain successfully to an acoustic rhythm will be more likely to tap in time in beat synchronization tasks. These two theoretical frameworks used in the child beat synchronization literature are largely complementary, and enable some neural predictions to be made for the current study. Firstly, on TS theory, group differences in the phase of entrainment in the delta band (here, 2.4 Hz) would be expected in the time domain. Following Power et al. (2013), the children with dyslexia would be expected to show a difference in preferred phase angle compared to the typically-developing control children. Secondly, both DAT and TS theory would predict group differences in the internal representation of the beat, documented by the 3.6 Hz SS-EPs. Such a group difference would be indicative of disrupted sensorimotor integration in developmental dyslexia. Identification of the source/s of disrupted beat synchronization have important implications for optimizing educational remediation.

Here, we set out to investigate the neural mechanisms that may underpin the developmental relationships between the precision of beat synchronization, children's phonological awareness, and their progress in reading (Thomson and Goswami, 2008; Corriveau and Goswami, 2009; Dellatolas et al., 2009; Flaugnacco et al., 2014; Woodruff Carr et al., 2014). Precise specification of neural mechanisms should enable the optimization of remedial programmes based on improving temporal synchronization in children, for example programmes based on drumming and other forms of rhythm production, which have been shown to enhance both children's phonological awareness and their reading development (Overy et al., 2003; Degé and Schwarzer, 2011; Bhide et al., 2013; Slater et al., 2014; Flaugnacco et al., 2015; Serrallach et al., 2016). Following prior work by Nozaradan et al. (2015), we aimed to use beat-related SS-EPs to disentangle sensory- and motor-related neural beat entrainment in children with developmental dyslexia and age-matched control children. Consistent with the prior dyslexia literature on neural entrainment, which has revealed atypical entrainment to both speech and non-speech rhythmic stimuli in the delta band (~1–3 Hz, see Hämäläinen et al., 2012; Power et al., 2013, 2016; Soltész et al., 2013; Molinaro et al., 2016), we also analyzed neural phase alignment (preferred phase angle) and phase consistency for the 2.4 Hz rhythmic stimulus.

The data showed atypical neural beat-driven entrainment in dyslexia in terms of preferred phase angle, and greater neural power when passively tracking the beat (2.4 Hz SS-EP in the Auditory only condition, p < 0.05, BF₊₀ = 3.102). The data were suggestive of an impaired internal representation of the beat during sensorimotor integration during right handed tapping (SS-EP at 3.6 Hz, p = 0.051, BF₊₀ = 2.862). There were no group differences for motor-related neural activity alone (1.2 Hz SS-EPs). As all children were right handed and still relatively young, it is likely that the variability of responding with the left hand precluded any group differences for the Left hand tapping condition (in their study with neurotypical adults, Nozaradan et al. (2015) also found significant effects for Left hand tapping). Both the neural time domain (preferred phase) and frequency domain (SS-EP at 2.4 Hz, Auditory only condition) measures showed significant correlations with the development of single word reading and phonological awareness.

Overall, our neural data support the view that there is an auditory rhythm perception deficit in developmental dyslexia, and that this sensory deficit affects the temporal precision of action. Given the results obtained in temporal remedial programmes based on practicing beat synchronization (see Degé and Schwarzer, 2011; Bhide et al., 2013; Slater et al., 2014; Flaugnacco et al., 2015; Serrallach et al., 2016), these auditory sensory deficits do not seem to prevent children from benefitting from rhythmic practice in the motor modality. However, the current data leave open the theoretical question of whether such behavioral programmes are beneficial because they improve a primary sensory deficit in auditory rhythm perception in dyslexia via cross-domain sensorimotor coupling (TS theory; Goswami, 2011), or because they improve sensorimotor temporal prediction (DAT theory; Jones, 2008), or both. The likely conclusion is that both theoretical frameworks are related to children's synchronization behavior, because both attention and language processing are supported by oscillatory neural processes.

Regarding TS theory, it has been argued that rhythmic remediation for dyslexia may improve phonological and reading skills by “entraining the oscillators,” as discussed by Goswami and Szucs (2011). This general proposal is also consistent with DAT theory. Goswami and Szucs (2011) proposed that both rhythmic linguistic interventions requiring motor production of speech, such as programmes based on singing, oral learning of nursery rhymes and reciting metrical poetry, and rhythmic interventions involving rhythmic motor practice such as beat synchronization, might support the accuracy of entrainment of lower frequency oscillatory networks (theta- and delta-rate networks) in primary auditory cortex, those important for accurate rhythm perception (see Giraud and Poeppel, 2012; Patel, 2012). We have also argued from behavioral data in a musical task based on a 2 Hz pulse (Huss et al., 2011) that the dyslexic brain may not be able to set up a reliable internal acoustic representation of a 2 Hz beat. A 2 Hz beat may represent an approximate periodic structure in language (based on stressed syllables, see Goswami et al., 2013) to which the infant brain may entrain across languages, facilitating language acquisition. Sensitivity to speech rhythm is considered a cross-language precursor of language acquisition (e.g., Mehler et al., 1988). If sensitivity to linguistic beat-based temporal structure is impaired in developmental dyslexia, this would lead to phonological impairments. By TS theory therefore, beneficial effects of rhythmic tapping on phonological development and reading would occur via improved temporal precision of auditory oscillatory networks via coupling with motor oscillators (see also Putkinen et al., 2013; Cason et al., 2015).

Meanwhile, DAT expects that any linguistic improvements that follow rhythmic practice are mediated by improved temporal prediction in the attention system. For language development, improving temporal prediction would improve expectations concerning the location of the most informative portions of the speech signal, such as stressed syllables. As discussed by Kotz and Schwartze (2010), speech perception is inherently linked to rhythmic timing via the perception of rhythmically prominent (stressed) syllables. The patterning of the stressed syllables provides a temporal structure that can be used by the brain in encoding and understanding speech, with these quasi-rhythmic events guiding auditory attention to important information. Indeed, it is known that beat-based and musical primes can improve children's syntactic processing as well as their phonological processing (Przybylski et al., 2013; Gordon et al., 2015). The most conservative conclusion from our data is that both auditory oscillators and any other oscillators involved in temporal prediction benefit from practice in beat synchronization via cross-modal integration.

It should be noted that a related account of why beat synchronization may enhance children's phonological awareness, but based on more rapid timescales, has been proposed by Tierney and Kraus (2014, 2016), the Precise Auditory Timing Hypothesis (PATH). Tierney and Kraus noted that the development of phonological awareness depends on precise timing perception, particularly at rapid timescales such as the differences in voice onset time or in formant transitions that distinguish consonants like /b/ and /d/. They noted that children with musical training are better able to distinguish such consonants (Strait et al., 2012), and that musical training may support fine-grained temporal discrimination by enhancing auditory-motor timing. The PATH is also consistent with the current data. The main difference in focus between TS theory and the PATH is in the temporal rates proposed to mediate the audio-motor effects on children's phonological development, namely slow (TS theory, which foregrounds syllabic information <10 Hz) vs. faster (the PATH, which foregrounds phonetic information, usually considered >30 Hz). While PATH focuses on rapid timescales, TS theory accords a special explanatory role to delta-band information. The 2.4 Hz beat utilized here falls within the delta band.

Oscillations in the delta band (~1–3 Hz, see Poeppel, 2014) sit at the top of the neural oscillatory hierarchy for speech encoding in auditory cortex (Gross et al., 2013). Amplitude modulations in the delta band play a key role in rhythmic child-directed speech (English nursery rhymes, Leong and Goswami, 2015) and in infant-directed speech (Leong et al., 2017). Adult studies reveal that neural oscillatory networks for speech act as a hierarchy via both phase-phase and phase-power relations, with delta rate oscillators governing the activity of oscillators at faster rates (Gross et al., 2013). This oscillatory hierarchy matches the acoustic hierarchy of AMs in English nursery rhymes (Goswami and Leong, 2013). Both auditory and neural studies suggest that oscillations in the delta band are related to the extraction of syllable stress patterns (e.g., Ghitza and Greenberg, 2009; Ghitza et al., 2013; Poeppel, 2014; Leong and Goswami, 2015). Accordingly, slower delta-rate temporal information may be an important target for remedial rhythmic programmes involving young children. One way to explore the temporal rate issue in more depth may be to contrast the effects on phonological awareness of general musical remediation with remediation focused specifically on delta-band rhythms.

In conclusion, the data reported here are supportive of atypical neural beat entrainment by the dyslexic brain in the delta band (~1–3 Hz), with the children with dyslexia showing a significantly different preferred phase both during passive listening and during beat synchronization. Brain-behavior correlations were found between measures of both reading and phonological development and individual differences in both preferred phase and in the power of SS-EPs at the auditory stimulation rate (2.4 Hz). Overall, the data are supportive of an interpretation of developmental difficulties in beat synchronization driven by impaired auditory perception of the beat. However, whether this conclusion would also apply to other developmental disorders of language learning which present with beat synchronization difficulties (stuttering, SLI) and whether it would apply in languages that are syllable-timed rather than stress-timed, requires further investigation.

LC and UG designed the experiments, performed the analyses, and drafted the manuscript. LC and HN collected the data.