Globally, 34 million children (∼0.2% of all children aged 0 to 18 years) have disabling hearing loss, i.e., hearing loss greater than 30 decibels (dB) in the better ear (report of Organization, 2021). The prevalence rises to 5% when mild and unilateral hearing losses are also considered (Wang et al., 2019). Unaddressed hearing loss has been proven to affect children’s speech and language development, educational attainments and social skills. Through early detection and interventions many of these impacts can be mitigated, highlighting the importance of accurate hearing diagnostics (O’Donoghue, 2013). Evaluation of speech intelligibility is a fundamental component of hearing loss assessment and rehabilitation. It can determine speech intelligibility and discrimination of speech features, and provides insight into the perceptual abilities of an individual (Eggermont, 2017). The current gold standard in measuring speech intelligibility relies heavily on behavioral tests. While these tests are reliable and fast in healthy adults, it can be difficult to assess speech intelligibility in children. A child’s abilities and limitations in the language domain as well as their cognitive abilities strongly affect the results of a behavioral test. Moreover, the type of response task, the speech material, and children’s motivation and involvement should also be considered (Mendel, 2008). In addition, measures of speech intelligibility are often restricted to testing in quiet, due to a lack of appropriate and reliable speech in noise tasks (van Wieringen and Wouters, 2022). However, the high prevalence of noise in children’s natural listening environments, especially in settings where children learn and play (e.g., clamorous classrooms and rowdy playgrounds), poses serious challenges for a still immature auditory system (e.g., Ambrose, Van Dam, Moeller, 2014, Neuman, Wroblewski, Hajicek, Rubinstein, 2010). Therefore it is essential that we can measure the ability to understand speech in these noisy, everyday environments.

An objective approach, using EEG to measure neural tracking in response to natural running speech, could overcome the current challenges in pediatric hearing assessment and provide a more reliable measure of a child’s speech intelligibility (for a review see Gillis et al., 2022b). Neural tracking refers to the process by which neural responses in a listener’s brain time-lock to dynamic patterns of the presented speech, such as the speech envelope. The speech envelope contains acoustical information (Rosen, 1992) and reflects phoneme, syllable and word boundaries (Peelle et al., 2013), which are critical for speech intelligibility (Shannon et al., 1995).

Indeed, a rich literature has found that during speech perception, auditory neural activity tracks the temporal fluctuations of the speech envelope in frequency bands matching the important occurrences of speech information, i.e., phrases/sentences (below 2 Hz) and syllables (2-8 Hz) (Ahissar, Nagarajan, Ahissar, Protopapas, Mahncke, Merzenich, 2001, Bourguignon, De Tiége, de Beeck, Ligot, Paquier, Van Bogaert, Goldman, Hari, Jousmäki, 2013, Ding, Melloni, Zhang, Tian, Poeppel, 2016, Gross, Hoogenboom, Thut, Schyns, Panzeri, Belin, Garrod, 2013, Meyer, Henry, Gaston, Schmuck, Friederici, 2017, Molinaro, Lizarazu, 2017, Vander Ghinst, Bourguignon, Niesen, Wens, Hassid, Choufani, Jousmäki, Hari, Goldman, De Tiége, 2019). More importantly regarding diagnostic purposes, many researchers have shown that neural envelope tracking is affected by the intelligibility of the presented speech (Di Liberto, Lalor, Millman, 2018, Gross, Hoogenboom, Thut, Schyns, Panzeri, Belin, Garrod, 2013, Peelle, Gross, Davis, 2013) and is significantly correlated with behavioral measures of speech intelligibility (Ding, Chatterjee, Simon, 2014, Lesenfants, Vanthornhout, Verschueren, Decruy, Francart, 2019, Vanthornhout, Decruy, Wouters, Simon, Francart, 2018, Verschueren, Vanthornhout, Francart, 2021).

Most of the above-mentioned studies have been conducted in adults. Research on neural envelope tracking in children is scarce, despite the functional relevance that neural tracking might have for objectively measuring children’s speech intelligibility. Moreover, cross-sectional evidence shows that auditory neural activity changes drastically during childhood (e.g., Cragg, Kovacevic, McIntosh, Poulsen, Martinu, Leonard, Paus, 2011, Panda, Emami, Valiante, Pang, 2020, Schneider, Maguire, 2019, Vander Ghinst, Bourguignon, Niesen, Wens, Hassid, Choufani, Jousmäki, Hari, Goldman, De Tiége, 2019), which makes it difficult to extrapolate outcomes from neural tracking research in adults to children. Furthermore, previous research in children mostly involved older children (> 10 years) with dyslexia, focusing on the relation between neural tracking and reading development/experience. These studies show consistent coherence in the delta frequency range (0.5-4 Hz) (Molinaro et al., 2016) and theta (4-8 Hz) (Abrams, Nicol, Zecker, Kraus, 2009, Molinaro, Lizarazu, Lallier, Bourguignon, Carreiras, 2016) using natural speech. In addition, they found that dyslexic children have impaired tracking at low frequencies (< 2 Hz) compared to typically developing children (Destoky, Bertels, Niesen, Wens, Vander Ghinst, Rovai, Trotta, Lallier, De Tiége, Bourguignon, 2022, Di Liberto, Lalor, Millman, 2018, Molinaro, Lizarazu, Lallier, Bourguignon, Carreiras, 2016, Power, Mead, Barnes, Goswami, 2013, Power, Colling, Mead, Barnes, Goswami, 2016).

Although evidence of the neural tracking mechanisms in typically developing children is limited, more recent studies have been able to show successful neural tracking of continuous natural speech in both infants (<1 year) (Attaheri, Choisdealbha, Di Liberto, Rocha, Brusini, Mead, Olawole-Scott, Boutris, Gibbon, Williams, Grey, Flanagan, Goswami, 2022, Kalashnikova, Peter, Di Liberto, Lalor, Burnham, 2018, Tan, Kalashnikova, Di Liberto, Crosse, Burnham, 2022) and young children between 4 and 9 years old (Ríos-López, Molinaro, Bourguignon, Lallier, 2020, Tan, Kalashnikova, Di Liberto, Crosse, Burnham, 2022, Vander Ghinst, Bourguignon, Niesen, Wens, Hassid, Choufani, Jousmäki, Hari, Goldman, De Tiége, 2019). For example, Ríos-López et al. (2020) showed that speech-brain coupling already occurs at 4 years of age in the delta-band frequency range, but not at theta frequencies. Similarly, Vander Ghinst et al. (2019) found significant speech tracking in children aged 6-9 at < 1 Hz frequencies, while neural envelope tracking was reduced or even absent in the theta band.

Furthermore, neural envelope tracking in children has typically been evaluated under optimal, noiseless listening conditions. Findings from multiple, behavioral studies provided evidence that mature performance on a wide range of speech-in-noise measures is established by about 9-10 years of age, while younger children require a higher signal-to-noise ratio (SNR) to achieve adult-like performance (for a review see Leibold 2017). Thus, children seem more susceptible to the detrimental effects of noise on speech intelligibility. Only one study to our knowledge has studied neural tracking of speech-in-noise in typically developing children using magnetoencephalography (MEG). In accordance to behavioral measures, Vander Ghinst et al. (2019) found that neural tracking differs between typically developing children and adults and, that noise differentially corrupts neural tracking in children. In adults, their results showed a clear effect of noise at delta frequencies (<1 Hz), that is, a decrease in coherence as SNR decreased and speech is less intelligible. However, in children increasing noise decreased coherence more strongly than in adults. Additionally, children’s coherence was drastically reduced or even absent in comparison with adults in the theta band regardless of SNR. Generally, these results are in line with previous behavioral studies showing children’s poorer speech intelligibility in adverse listening conditions (Johnson, 2000, Leibold, 2017, Neuman, Wroblewski, Hajicek, Rubinstein, 2010, Talarico, Abdilla, Aliferis, Balazic, Giaprakis, Stefanakis, Foenander, Grayden, Paolini, 2007). However, we cannot conclude from this study that delta and/or theta coherence is directly related to speech-in-noise intelligibility, since only indirect behavioral measures (e.g., intelligibility rating) were included. In addition, MEG-based recordings were used and various practical aspects of MEG in its current form (e.g., cost) pose a limitation to its large-scale usability in clinical practice, compared to electroencephalography (EEG) (Destoky et al., 2019).

Our recent research provides a framework for investigating neural tracking of different speech features, including the speech envelope (Gillis, Van Canneyt, Francart, Vanthornhout, 2022, Lesenfants, Vanthornhout, Verschueren, Francart, 2019, Verschueren, Vanthornhout, Francart, 2021), using EEG. The approach combines both linear decoding (backward-modelling) and encoding (forward-modelling) models, providing complementary information about neural tracking. The backward model is a model to reconstruct the speech envelope from the associated EEG recording, whereas the forward model predicts the EEG responses to speech and can be used to study the spatio-temporal dynamics of the response similarly to event related potentials (ERPs). This is the first study to use this method in preschoolers and we aim to investigate the validity of a measure that has previously been used only with adults (e.g., Vanthornhout et al., 2018). More specifically, in this study, we investigate (i) the effect of SNR on neural envelope tracking in preschoolers, and (ii) whether neural envelope tracking reflects speech intelligibility by evaluating the correspondence between neural envelope tracking and behavioral measures of speech intelligibility in noise. We have two specific hypotheses. First, we predicted that as SNR increases, neural envelope tracking increases, given that previous research has shown that stronger neural responses are associated with better speech intelligibility (Ahissar, Nagarajan, Ahissar, Protopapas, Mahncke, Merzenich, 2001, Ding, Chatterjee, Simon, 2014, Peelle, Gross, Davis, 2013, Vanthornhout, Decruy, Wouters, Simon, Francart, 2018). Secondly, we hypothesize that, similar to studies in adults (Lesenfants, Vanthornhout, Verschueren, Decruy, Francart, 2019, Vanthornhout, Decruy, Wouters, Simon, Francart, 2018), behaviorally measured speech intelligibility in noise is significantly correlated with our neural, objective measure.