Vowels are the most important category of voiced sounds due to their crucial role in speech recognition (Fogerty & Humes, 2012). The spectral formant composition is the major phonetic aspect that determines vowel identity. Another important feature of vowels is their periodic structure and the associated pitch, which enables speaker identification (van Dommelen, 1990), auditory stream segregation (Oh, Zuwala, Salvagno, & Tilbrook, 2021), and vowel normalization (Andermann, Patterson, Vogt, Winterstetter, & Rupp, 2017). Abnormal processing of pitch and formant composition may contribute to receptive language problems associated with autism and specific language impairments (Arunachalam & Luyster, 2016; Demopoulos et al., 2015; Loucas et al., 2008; Rotschafer, 2021; Schelinski & von Kriegstein, 2020; Tager-Flusberg & Caronna, 2007; Yu et al., 2021). Therefore, studying how these properties of speech sounds are processed in the developing brain is important for understanding the mechanisms of speech abnormalities in developmental disorders.

Magnetoencephalography (MEG) has excellent temporal and good spatial resolution, which is crucial for studying the spatiotemporal dynamics of neural events associated with the analysis of speech in the auditory cortex. However, MEG studies of neural underpinnings of vowel processing are scarce and their results are inconsistent (Manca, Di Russo, Sigona, & Grimaldi, 2019; Manca & Grimaldi, 2016). Even less is known about MEG correlates of vowel processing during development because across-age comparison is complicated by the developmental changes in transient components of auditory responses (Albrecht, Suchodoletz, & Uwer, 2000; Donkers, Carlson, Schipul, Belger, & Baranek, 2020; Dwyer, De Meo-Monteil, Saron, & Rivera, 2021; Orekhova et al., 2013; Parviainen, Helenius, & Salmelin, 2019; Ponton, Eggermont, Kwong, & Don, 2000; Ruhnau, Herrmann, Maess, & Schröger, 2011).

Compared to the closely matched noise stimuli, the processing of sounds characterized by periodicity/pitch and/or formant structure is associated with a greater sustained ‘negative’ shift of cortical source current underlying MEG-recorded sustained magnetic field (‘sustained field’ – SF) (Gutschalk, Patterson, Scherg, Uppenkamp, & Rupp, 2004; Gutschalk & Uppenkamp, 2011; Hewson-Stoate, Schönwiesner, & Krumbholz, 2006). This differential sustained response persists throughout stimulus presentation and is evident already in the time range of the N100m component (∼100 msec) or even earlier (Gutschalk et al., 2004; Gutschalk & Uppenkamp, 2011; Hewson-Stoate et al., 2006). The differential SF response is not specific to vowel features processing but accompanies any regular acoustic pattern popping out of the ‘ground’ noise (Arunachalam & Luyster, 2016; Barascud, Pearce, Griffiths, Friston, & Chait, 2016; Southwell & Chait, 2018).

There are two features of the SF that make this neural response particularly suitable for studying temporal–spatial patterns of auditory processing at different developmental stages.

First, the sustained shift of potential (in EEG) or magnetic field (in MEG) associated with the processing of pitch and formant structure may reflect an integrated phonological representation of a time-varying complex acoustic signal in the auditory cortex (discussed in Stroganova et al., 2022). Animal studies show that this integrated representation is carried out by a certain type of neuronal population in the auditory cortex – the so-called non-synchronized neurons (Lu, Liang, & Wang, 2001; Wang, 2007, 2018; Wang, Lu, Bendor, & Bartlett, 2008; Wang, Lu, Snider, & Liang, 2005). Non-synchronized neurons increase tonic firing when driven by their respective preferred stimuli and are therefore highly dependent on sound characteristics (Wang et al., 2005). Similar to SF recorded by EEG or MEG, the activity of these neurons persists during stimulus presentation and is not time-bound to each periodic intensity pulse. It has been hypothesized that neurons of this type encode temporally integrated information by mean discharge rate (Wang, 2018; Wang et al., 2008). Thus, the SF response may reflect cortical processing of certain stimulus features by the non-synchronized populations of neurons, both through local cortical interactions and recurrent connections with higher-tier cortical areas.

Second, unlike the transient components of auditory ERP/ERF, SF has the same ‘negative’ polarity in children and adults (Stroganova et al., 2020) and is similarly enhanced in these age groups by sound periodicity and formant structure (Gutschalk & Uppenkamp, 2011; Stroganova et al., 2022). Therefore, shifts of sustained cortical source current induced in the auditory cortex by these stimuli can be directly compared between age groups.

Although neural networks involved in the processing of periodicity (pitch) and formant structure of the sound largely overlap (Bonte, Hausfeld, Scharke, Valente, & Formisano, 2014; Walker, Bizley, King, & Schnupp, 2011), there is some evidence that they are differently lateralized in the auditory cortex. Despite some inconsistent results (see (Hertrich, Mathiak, Lutzenberger, & Ackermann, 2004) for discussion), there is a general agreement that presence of periodicity (pitch) in any spectrally complex non-speech sound is associated with the right-lateralized activity in both EEG (Krishnan, Gandour, Ananthakrishnan, & Vijayaraghavan, 2015) and fMRI (Hyde, Peretz, & Zatorre, 2008; Patterson, Uppenkamp, Johnsrude, & Griffiths, 2002; Zatorre, Belin, & Penhune, 2002). However, natural vowel sounds that possess periodicity usually elicit bilateral and symmetrical brain responses in both MEG/EEG (see Manca for review) and fMRI (Bonte et al., 2014; Steinschneider, 2013; Uppenkamp, Johnsrude, Norris, Marslen-Wilson, & Patterson, 2006). It is likely that some vowel features are processed predominantly by the left hemisphere, but this left-hemispheric asymmetry is masked by the predominantly right-hemispheric processing of periodicity, which is typically present in natural vowels.

Although it is commonly accepted that speech is lateralized in the brain, there is an ongoing debate about the nature of this lateralization. The ‘domain general’ account suggests that the preferential processing of speech in the left hemisphere is due to an acoustic bias between the two hemispheres that favors the left-hemispheric processing of short speech sounds, such as consonants and/or rapid formant transitions in syllables. In contrast, the ‘domain specific’ account relates left hemispheric dominance to the linguistic features of speech (see (Bourke & Todd, 2021; Scott & McGettigan, 2013) for discussion). However, it is still unclear whether lateralization is related to semantic representation (Shtyrov, Pihko, & Pulvermüller, 2005) or occurs already during phonetic categorization (Bourke & Todd, 2021; Hornickel, Skoe, & Kraus, 2009). The issue is further complicated by changes in the relative contributions of the right and left hemispheres to speech perception during ontogeny (Nora et al., 2017; Olulade et al., 2020; Szaflarski, Holland, Schmithorst, & Byars, 2006).

There are two studies that have examined the lateralization of individual vowel features. Uppenkamp et al. (2006) independently manipulated vowel formant structure and periodicity/pitch in an fMRI study, but in both cases, activation was equally strong in the left and right hemispheres. However, if the functional asymmetry is limited to a short time interval, it can be blunted by the low temporal resolution of fMRI. Gutschalk et al. (Gutschalk & Uppenkamp, 2011) used the same stimuli in a MEG study, but also found no hemispheric asymmetry of SF during the processing of vowel periodicity or formant structure. There are at least two factors that might complicate the detection of hemispheric asymmetry in this MEG study. First, the authors did not take advantage of the high temporal resolution of the MEG and averaged the sustained neural activity over a wide time range (from 300 msec up to 800 sec of sound duration). Second, they used sequential fitting of a single dipole to localize the cortical sources of SF. Although dipole fitting approach is a powerful method for localization of locally generated cortical responses, the distributed spatiotemporal patterns of brain activity produced by spectrally complex periodic sounds (Allen, Mesik, Kay, & Oxenham, 2022; De Angelis et al., 2018; Norman-Haignere, Kanwisher, & McDermott, 2013) are likely to be more accurately described using distributed source modeling in combination with individual brain models.

Meanwhile, the perfect temporal resolution of MEG can help to detect lateralization and find out when it occurs in the auditory information-processing stream. The studies analyzing N100 (N100m) component of the auditory response in adults found early (at least at 100 msec after stimulus onset) rightward lateralization for pitch processing in the auditory cortex (Hertrich et al., 2004). Furthermore, vowel duration in natural languages rarely exceeds 200 msec (Jacewicz, Fox, & Salmons, 2007) suggesting that lateralization associated with processing of formant composition may occur soon after acoustic signal arrives to the cortex. Thus, at this point, it remains unclear how the cortical generators of sustained response to formant structure and vowels periodicity/pitch are related to each other, and whether the neural processing of these two features has the opposite hemispheric dominance.

In this study, we examined the sustained MEG response associated with vowel formant and periodicity/pitch processing in children and adults, using individual brain models and distributed source localization. We applied vowel-like stimuli previously used in Gutschalk and Uppenkamp (2011) and Uppenkamp et al. (2006). They included synthetic vowels and three types of sounds created by modifying synthetic vowels to selectively disrupt their periodicity, formant structure (‘vowelness’), or both while keeping other auditory signal characteristics (total energy, temporal profiles) similar to those of the synthetic vowels. The modified stimuli were of three types: sounds having only the formant structure of vowels (non-periodic vowels), only their periodic structure (periodic non-vowels), or none of these qualities (control stimuli). By contrasting structured and control stimuli and calculating cortical differential sustained responses corresponding to SF differences, we were able to identify neural activity specifically associated with the periodicity or formant structure. First, we sought to find out whether the effects of periodicity/pitch and stable formant structure on SF are similar in children and adults. Next, by contrasting an altered vowel devoid of periodicity with a normal periodic vowel, we tested whether adding periodicity to this speech sound results in the expected changes in hemispheric lateralization in these age groups.