Abstract

Introduction:

Integration of written and spoken information is crucial for reading acquisition. Correspondingly, individuals with reading difficulties exhibit deficiencies in audiovisual (AV) congruency processing. The timeline of AV congruency processing in children and the influence of reading skills on this process, however, remain largely unclear. Therefore, we examined when and how reading skills modulate AV congruency processing for orthographic (words, pseudowords) and non-orthographic conditions (objects).

Methods:

Eighty-two native German-speaking 2nd and 3rd graders completed an explicit task involving the matching of AV congruent and incongruent orthographic and non-orthographic stimuli, while EEG was recorded.

Results:

Behaviorally, poorer reading skills were associated with lower performance and slower responses for orthographic conditions. Neurally, topographic EEG analyses revealed congruency effects emerging after 300 ms for orthographic conditions and around 200 ms for objects. ERP analyses showed that reading skills modulated the N400 incongruency effect more strongly for orthographic than non-orthographic stimuli.

Discussion:

In summary, poorer reading skills were associated with slower AV matching and a weaker N400 incongruency effect for orthographic conditions. These findings suggest that while reading skills might not broadly affect AV congruency processing, they critically impact the AV congruency processing of orthographic information, potentially hindering struggling readers’ ability to effectively use preceding auditory information to process print.

1 Introduction1.1 Audiovisual processing as a foundation for reading acquisition

Humans rely on multiple senses to understand and interact with their environment. In language comprehension, audiovisual (AV) processing is of particular relevance. Visual speech cues, such as lip movements, help disambiguate ambiguous speech sounds in spoken language, and in written language, phonemes must be mapped onto graphemes to decode print (Ehri, 2005; Frost, 2012; McGurk and Macdonald, 1976; van Wassenhove et al., 2005). In alphabetic scripts, beginning readers first learn letter–speech sound correspondences and then use this knowledge to decode words in a slow, effortful, letter-by-letter fashion. This letter–speech sound learning process specifically reflects AV integration, as it requires binding auditory phonemes to their corresponding visual graphemes during reading acquisition (Blomert, 2011; Froyen et al., 2008). AV information integration is therefore considered fundamental for reading acquisition and operational literacy in alphabetic languages (Blomert and Froyen, 2010; Romanovska et al., 2022). With practice and repeated decoding (i.e., reading letter-by-letter), children gradually build a mental lexicon of stored word form representations for rapid access to familiar words without serial decoding, which is referred to as “direct word recognition” or “sight-word reading” (Coltheart, 2005; Ehri, 2005). Nevertheless, even skilled readers continue to rely on decoding for unfamiliar words, irregular words, or non-words (Coltheart, 2005). Once skilled in reading, visual orthographic stimuli (i.e., printed words or letter strings) can facilitate auditory comprehension, and conversely, spoken language can facilitate visual word processing.

1.2 Behavioral evidence for AV facilitation

AV integration is typically studied in paradigms that manipulate AV congruency in simultaneous presentations or crossmodal priming, contrasting audiovisually congruent (AVcon) and incongruent (AVinc) stimulus pairings. In these paradigms, auditory and visual stimuli can be presented simultaneously (synchronously) or sequentially, with one modality preceding the other (asynchronous presentation of AV priming) (van Atteveldt et al., 2007a). Across tasks, congruent AV input enhances behavioral performance: for instance, adults showed faster speech-in-noise identification, quicker letter target detection, and better working memory with AV versus unimodal stimuli (Blau et al., 2008; Frost et al., 1988; Raij et al., 2000; Xie et al., 2017). Although stronger in adults and still maturing through childhood, children also benefit: synchronous AV letters, words, or abstract stimuli yield higher accuracy and faster reactions than unimodal presentations (Barutchu et al., 2010; Jost et al., 2014; Kronschnabel et al., 2014). Similar facilitation by AV congruency has been observed for both synchronous and asynchronous presentations. Children and adults show better performance for AVcon than AVinc word pairs in crossmodal priming tasks (Clahsen and Fleischhauer, 2014; Holcomb et al., 2005; Quémart et al., 2018), as well as in tasks requiring synchronous AV processing (adults: Hocking and Price, 2009; children: Jost et al., 2014; adults: Yin et al., 2008).

1.3 Event-related potential (ERP) markers of AV congruency processing across reading development

On the neural level, electrophysiological event-related potentials (ERPs) provide temporally precise markers for probing AV processing during reading. Early components such as the visual N1 reflect perceptual encoding of orthographic features and become increasingly specialized as children learn to read (Brem et al., 2010; Fraga-González et al., 2021; Maurer et al., 2005a, 2005b, 2010; Maurer and McCandliss, 2007). Later components such as the N400 are associated with lexico-semantic and congruency processing. Together, these ERPs offer useful indices of how AVcon and AVinc information is processed across reading development (Blackford et al., 2012; Diaconescu et al., 2011; Holcomb et al., 2005).

The visual N1 (often also referred to as N170) is characterized by bilateral negative deflection over occipitotemporal electrodes in response to a visual stimulus peaking at around 170 ms to 220 ms in children and shows an experience-dependent tuning to orthographic stimuli with reading acquisition (Brem et al., 2010; Maurer et al., 2006; Zhao et al., 2014). Only a few studies have examined the N1 directly in AV orthographic paradigms, but several have used the mismatch negativity (MMN) to study AV congruency (Froyen et al., 2008, 2009b, 2010; Näätänen, 2001; Žarić et al., 2014). The MMN is a distinct ERP component indexing automatic change detection in oddball paradigms, typically occurring in a similar early time window and thus potentially overlapping with N1-related processing. Using oddball paradigms in adults, early neural differences between AVcon and AVinc letter–speech sound pairs have been observed around 150 ms after stimulus onset in implicit letter–speech sound paradigms (Andres et al., 2011). Early visual N1 congruency differences (around 160 ms) have also reported in Swiss-German speaking first-graders with varying reading skills in a target detection task involving AVinc and AVcon words (Jost et al., 2014), whereas MMN congruency effects in children are less consistent and depend on age and reading skills (Froyen et al., 2009a, 2011; Žarić et al., 2014, 2015). In addition, MMN paradigms have revealed late AV congruency effects around 650 ms (late negativity) in both typically reading and poor-reading children across grades (Froyen et al., 2009a, 2011; Žarić et al., 2015). Overall, these findings suggest AV congruency effects at both early sensory (N1, MMN) and later integrative stages, with their timing and strength varying by reading development and skill level.

The N400 is an extended centro-parietal negative component peaking roughly between 200 and 600 ms and is elicited by semantic conflict or unexpected events (Duncan et al., 2009). Its amplitude reflects the effort required to integrate multiple units of lexicosemantic information, largely independent of the modality (Kutas et al., 1987). Consequently, the N400 has been extensively used to study semantic violations in sentences and is also sensitive to crossmodal repetition and priming (Kiyonaga et al., 2007; Kutas and Federmeier, 2011; Kutas and Hillyard, 1984). The unimodal N400 in response to words and objects is generally similar in adults, supporting the view that the N400 indexes conceptual memory access (Nigam et al., 1992). Pseudowords typically elicit N400 amplitudes comparable to or larger than real words, reflecting extended lexical–semantic search (Coch and Holcomb, 2003; Tzeng et al., 2017; Varga et al., 2021). AV N400 effects have been reported in children and even when prime–target relations are newly learned shortly before the experiment, although the timing may differ in younger children (Jost et al., 2014; Kaganovich and Ancel, 2019; Liu et al., 2012).

Despite some inconsistency in the direction of AV congruency effects on ERP amplitudes (Herdman et al., 2006; Jost et al., 2014; van Atteveldt et al., 2007b), processing of AVcon stimuli typically leads to attenuated ERP amplitudes compared to unimodal or AVinc stimuli (Caffarra et al., 2021; Hocking and Price, 2009; Raij et al., 2000). Depending on whether AVcon or AVinc elicit the more pronounced response, the effect is labeled an AV congruency or AV incongruency difference, respectively. The factors underlying these direction differences remain subject of ongoing debate (Caffarra et al., 2021; Plewko et al., 2018). One plausible account is that reduced amplitudes for AVcon stimuli reflect optimized neural processing of overlearned, familiar AV associations, whereas AVinc pairs fail to benefit from such tuning (Raij et al., 2000).

1.4 Developmental differences and reading skill effects on AV congruency processing

A deficit in letter-speech sound (LSS) integration despite adequate reading instruction has been discussed as a possible reason underlying poor reading skills (Aravena et al., 2013; Blau et al., 2009, 2010; Blomert, 2011). Supporting this view, reduced AV suppression and congruency effects in classical integration areas have been observed in children, adolescents, and adults with poor reading skills (Blau et al., 2009, 2010; Froyen et al., 2011; Kronschnabel et al., 2014). EEG findings similarly indicate deviant AV congruency processing in individuals with poor reading skills across development (Froyen et al., 2011; Kronschnabel et al., 2014; Mittag et al., 2013; Žarić et al., 2014). Longitudinal studies starting in pre-reading children have shown reduced developmental changes in the emergence, strength, and timing of the AV integration effects (Caffarra et al., 2021; Froyen et al., 2009a; Karipidis et al., 2021) in children who later develop poor reading skills compared with those who become typical readers (Karipidis et al., 2018, 2021). However, the nature of the AV integration deficit remains debated (Gori et al., 2020; Hahn et al., 2014). It is still unclear whether the deficit is predominantly restricted to orthographic or linguistic contexts (Desroches et al., 2010; Shaywitz and Shaywitz, 2005; van Laarhoven et al., 2018; Ye et al., 2017) or whether it reflects a more general AV integration impairment that also affects non-linguistic stimuli (Goswami et al., 2013; Hairston et al., 2005; Harrar et al., 2014; Widmann et al., 2012).

1.5 AV processing of orthographic- vs. non-orthographic stimuli

Despite a growing interest in AV processing and integration, relatively few studies have directly compared AV congruency processing across orthographic and non-orthographic categories (e.g., letters, words, pseudowords, objects) in children and examined how such effects relate to reading skills. Most work has focused on letters (Andres et al., 2011; Blau et al., 2009, 2010; Froyen et al., 2008, 2009a, 2011; Herdman et al., 2006; Kronschnabel et al., 2014; Raij et al., 2000; van Atteveldt et al., 2004, 2007a), with fewer studies using syllables, consonant-vowel-consonant strings, words, or pronounceable non-words (Caffarra et al., 2021; Eberhard-Moscicka et al., 2021; Jost et al., 2014; Kronschnabel et al., 2014; Okano et al., 2016; Varga et al., 2021; Wang et al., 2020). AV processing of lexical (words) versus non-lexical (non-words) stimuli (Varga et al., 2021) as well as of orthographic versus non-orthographic stimuli has been investigated separately, including using functional MRI, but direct comparisons across these categories remain scarce (see Hocking and Price, 2009). In the present work, we therefore aim to better tease apart different visual categories.

1.6 Auditory–visual congruency, speech–print mapping, and reading skills

Research to date has mainly focused on the “decoding” of print to speech at the letter-by-letter level, and on direct whole-word recognition and semantic access (e.g., Coltheart, 2005; Ehri, 1997, 2005; Gonzalez-Frey and Ehri, 2021; Reitsma, 1983). Most EEG studies on AV orthography or object congruency present auditory stimuli concurrently with or following visual stimuli (Froyen et al., 2010; McNorgan et al., 2013; Žarić et al., 2015). Priming studies with auditory primes for visual targets typically used very short or masked primes and focus on adults (Grainger et al., 2003; Holcomb and Anderson, 1993; Holcomb et al., 2005; Justus et al., 2009; Kiyonaga et al., 2007). Conceptually, however, learning to read and write also heavily relies on the reverse direction: transforming spoken language into its written form and aligning spoken and written representations. Instruction in alphabetic languages usually introduces written forms via its familiar spoken forms at multiple levels of complexity—ranging from simple phonemes to syllables, morphemes, words, sentences, and text. Classroom activities such as dictation or shared story reading, where spoken information precedes or accompanies written text, leverage auditory cues to enhance speech–print mappings. Beyond initial acquisition, auditory-to-visual processing remains crucial for spelling, using (asynchronous) video captions, following spoken instructions referring to text, scanning text for specific information, vocabulary learning, and even disambiguating similarly sounding words (e.g., “ice cream” vs. “I scream”) (Iordanescu et al., 2011). These demands highlight the tight links between spoken and written language, suggesting that phonological and orthographic systems influence each other in both directions across development. Behavioral and neuroimaging studies support this view by demonstrating bidirectional influences between phonological and orthographic processing, including orthographic neighborhood effects on spoken word recognition and recruitment of reading-related brain areas such as the fusiform gyrus during purely auditory tasks (Desroches et al., 2010; Ziegler and Muneaux, 2007). Reading and decoding skills, in turn, modulate AV processing and automatic orthographic access during auditory processing (Desroches et al., 2010; McNorgan et al., 2013). Auditory primes may thus enhance the processing of written words more effectively than the processing of objects, given the tighter link between written words and their phonological forms (Glaser and Glaser, 1989; Price et al., 2006). At the same time, phonological processing and rapid automatized naming of objects are closely associated with reading skills and can predict future reading outcomes in pre-readers (McWeeny et al., 2022; Moll et al., 2009), suggesting some overlap in how reading skills influence the AV processing orthographic and non-orthographic items.

1.7 The present study and hypotheses

In this EEG study, 2nd- and 3rd-grade children with varying reading skills performed an AV priming task in which congruent (AVcon) and incongruent (AVinc) auditory primes preceded visual word, pseudoword, or object stimuli. The study’s purpose was to investigate the influence of reading skills on the processing of AV congruency of lexical orthographic, non-lexical orthographic, and non-orthographic stimuli.

In the present work, we refer to visually presented letter strings (words and pseudowords; W and PW) as “orthographic stimuli,” whereas object images (Obj) are referred to as “non-orthographic stimuli” (see Supplementary section 6 for abbreviations). Furthermore, we refer to stimuli as “lexical” if they have established lexical–semantic representations (words and familiar objects) and as “non-lexical” if they lack such representations (pseudowords). AV perceptual binding is thought to require close temporal proximity, typically within a temporal binding window up to approximately 300 ms (Stevenson and Wallace, 2013). Because the auditory stimulus precedes the visual stimulus by more than 500 ms in our task, the present paradigm does not index low-level perceptual AV binding. Instead, we interpret observed effects as reflecting higher-level “AV congruency processing” and this term is used throughout to describe the experimental paradigm and results. Importantly, such AV congruency effects are assumed to depend on intact AV integration, defined here as the establishment of learned associations between orthographic and phonological representations.

First, we examined behavioral congruency effects on task performance across stimulus types and in relation to reading skills. We hypothesized AVcon stimuli to elicit higher accuracy and faster responses than AVinc stimuli, reflecting behavioral AV congruency facilitation (Clahsen and Fleischhauer, 2014; Jost et al., 2014; Quémart et al., 2018), and that this facilitation would be stronger in children with higher reading skills for orthographic items (words, pseudowords) but not for object images.

Our second hypothesis pertained to the temporal emergence of AV congruency effects in the ERP time course for orthographic and non-orthographic stimuli. Specifically, we compared mean ERP amplitudes for AVcon and AVinc trials (congruency difference) and tested how reading skills modulate effects for lexical versus non-lexical (W vs. PW) and orthographic versus non-orthographic (W/PW vs. Obj) conditions. Our second hypothesis was that better readers would show larger N400 incongruency effects, i.e., greater negativity for AVinc than AVcon stimuli from around 300 ms, particularly for orthographic, but not object, conditions, indexing more efficient matching of visual information with the preceding auditory prime (Caffarra et al., 2021; Jost et al., 2014; Kaganovich and Ancel, 2019; Karipidis et al., 2017; Raij et al., 2000).

Finally, independent of congruency, our third hypothesis was that auditory primes would modulate early occipitotemporal N1 responses in a reading-skill-dependent manner, with differences between lexical and non-lexical orthographic stimuli, as well as potential effects of reading skills on N1 lateralization (Araújo et al., 2012).

As a complementary, data-driven, and exploratory analysis, we performed TANOVA analyses to examine AV congruency and reading-skill-related effects outside of the predefined time windows and independently of a priori electrode selections, focusing on global differences in scalp topography over time. TANOVA is well suited for detecting neural effects that may not be captured by mean amplitude comparisons alone. In contrast to the ERP analyses, which modeled reading skills as a continuous variable, the TANOVA analyses were conducted at the group level, complementing continuous analyses of individual variability. This group level perspective may allow the detection of effects that do not scale linearly with reading skills and provides a directly interpretable comparison between children with lower and higher reading skills.

2 Methods and materials2.1 Participants

A total of 95 native German-speaking children in 2nd to 3rd grade participated in the first session of a longitudinal neuroimaging study which included a grapheme-phoneme intervention and multiple time points including behavioral assessments and neuroimaging recordings (EEG, fMRI). The present article focuses on the behavioral and EEG results of the first time point (Table 1). The data of 13 children were excluded from the final analyses due to not meeting our stringent EEG quality criteria (see below), resulting in a final sample of 82 participants (M = 8.86 y, SD = 0.65 y; see Table 1 for more detailed demographic information). For the linear mixed model (LMM) analyses, one participant was excluded because the control covariate was missing for all observations, resulting in an analytic sample of 81. Inclusion criteria were nonverbal IQ index (NVIQ) scores > 80 (non-verbal subpart of the Reynolds Intellectual Assessment Scales test battery; RIAS) (Reynolds and Kamphaus, 2003), normal or corrected-to-normal vision, normal hearing, and the absence of neurological, neurodevelopmental, or psychiatric impairments, except for dyscalculia (5 self-reported cases, 1 of which diagnosed by a specialist) and attention deficit (hyperactivity) disorder (AD(H)D, 9 cases with diagnosis, of which 4 under medication). These criteria were assessed in a telephone screening with the children’s parents, except for IQ, which was tested in the behavioral session. Individuals with AD(H)D were either unmedicated or were asked to discontinue medication at least 24 h before behavioral and EEG sessions. Parents filled in the Child Behavioural Checklist (CBCL) questionnaire and the subscore on attention-deficit/hyperactivity symptoms was used to estimate children’s ADHD symptoms (CBCL/4–18 subscore) (Achenbach and Edelbrock, 1983; Steinhausen and Winkler Metzke, 2011). Further, parents also provided information on their reading history in the Adult Reading History Questionnaire (ARHQ) (Lefly and Pennington, 2000) which was used as an estimate of the familial dyslexia risk of the children: The highest parental value determined the degree of familial risk with values greater than 0.3 indicating an increased familial risk for developmental dyslexia. Familial risk level for developmental dyslexia was low in 24 children (29.3%; ARHQ < 0.3), moderate in 26 children (31.7%; ARHQ range 0.3–0.4), and high in 32 children (39.0%; ARHQ > 0.4) (see Table 1; Fraga-González et al., 2022; Lefly and Pennington, 2000). Prior to the study, parents gave written informed consent, and children were asked for their oral consent. Children were reimbursed with vouchers and presents. The study was approved by the local ethics committee of the Canton of Zurich and neighboring cantons in Switzerland (BASEC No. 2018-01261). All experiments were performed in accordance with relevant guidelines and regulations of the approving local ethics committee.

MeasureM (SD) [min, max]Spearman’s rankUncorrectedCorrectedN82School class (2nd:3rd)31:51Sex ratio (female:male)41:41Handedness (right:left:both)73:8:1Age8.86 (0.66) [7.48, 10.25]rs = −0.18puncorr = 0.106pcorr = 1Months since school enrolment30.00 (7.15) [16.53, 39.98]rs = 0.01puncorr = 0.965pcorr = 1ReadingReading composite measure: Average perc. of word reading fluency, pseudoword decoding fluency, and reading comprehension37.90 (28.40) [0.87, 99.13]Word reading fluency (perc.)35.22 (30.81) [1, 99]Pseudoword decoding fluency (perc.) [N = 81]39.27 (30.43) [1, 99]Reading comprehension (perc.)39.21 (31.45) [1, 99]Silent sentence reading fluency (RQ)89.09 (18.63) [62, 138]rs = 0.92puncorr < 0.001pcorr < 0.001**RAN (Rapid automatized naming) Short animal names0.85 (0.20) [0.42, 1.39]rs = 0.51puncor r < 0.001pcorr < 0.001**RAN Long animal names0.63 (0.18) [0.20, 1.04]rs = 0.48puncorr < 0.001pcorr < 0.001**Letter knowledge, lower-case (letter names)24.08 (3.42) [5, 26]rs = 0.29puncorr = 0.009pcorr = 0.141Letter knowledge, lower-case (letter sounds)24.45 (3.27) [6, 26]rs = 0.17puncorr = 0.121pcorr = 1Spelling (perc.)34.60 (29.67) [0, 100]rs = 0.76puncorr < 0.001pcorr < 0.001**Child Behavioural Checklist (CBCL) Attention-deficit/hyperactivity subscore (T-scores) [N = 81] (normal range: T-scores <65, T-scores 65–69, clinical range: T-scores >69)54.99 (7.01) [50, 84]rs = −0.36puncorr = 0.001pcorr = 0.014*Adult Reading History Questionnaire (ARHQ) [N = 80]0.38 (0.13) [0.09, 0.68]rs = −0.19puncorr = 0.091pcorr = 1IQ Nonverbal104.38 (7.37) [88, 120]rs = 0.26puncorr = 0.019pcorr = 0.291IQ Verbal98.98 (12.23) [53, 121]rs = 0.37puncorr = 0.001pcorr = 0.010*Receptive Vocabulary (perc.) [N = 77] (performed in a separate session 12 or 24 weeks later)53.61 (30.25) [8.10, 98.90]rs = 0.12puncorr = 0.304pcorr = 1Digit Span Forward4.63 (0.84) [3, 7]rs = 0.26puncorr = 0.018pcorr = 0.277Digit Span Backward3.33 (0.89) [2, 6]rs = 0.27puncorr = 0.013pcorr = 0.199

Sample characteristics, behavioral test scores, and Spearman’s rank correlations with the reading composite measure.

Raw scores represent the number of correct items (Letter-knowledge, maximum 26 letters) or the number of correct items named per second (RAN). Age-standardized scores were used when available: Perc., percentile scores; IQ, intelligence quotient; RQ, reading quotient; digit span: longest number of digits correctly recalled by the participant until the end criterion of test was reached (range = 2–9). Stars indicate correlations that remain significant after Bonferroni correction for multiple comparisons: *pcorr ≤ 0.05; **pcorr ≤ 0.001.

2.2 Cognitive assessments

The behavioral sessions were held in person (N = 73) or (during the COVID-19 pandemic) via online video sessions (N = 9) with a duration of approximately 3 h. Percentile scores of reading comprehension (Ein Leseverständnistest für Erst-bis Siebtklässler-Version II; ELFE-II) (Lenhard et al., 2018), word reading fluency (Salzburger Lese-und Rechtschreibtest 2; SLRT-II W) (Moll and Landerl, 2014), and pseudoword decoding fluency (SLRT-II PW) (Moll and Landerl, 2014) were averaged to form a reading skill composite measure. Percentile scores index children’s relative standing among age-matched peers. Silent sentence reading fluency (Salzburger Lese-Screening; SLS 2–9) (Wimmer and Mayringer, 2014) was also assessed. Additionally, we evaluated lower-case letter knowledge, rapid automatized naming of short and long animal names (RAN), spelling (Schreib.on® Online; May, 2008, 2010; Valtin and Hofmann, 2009), nonverbal and verbal IQ (Reynolds and Kamphaus, 2003), digit span (subtest of Wechsler Intelligence Scale for Children; WISC-V) (Wechsler and Petermann, 2017), and vocabulary (Peabody picture vocabulary test; PPVT-4) (Dunn and Dunn, 2015). Table 1 provides an overview of the mean performance on behavioral assessments and correlations between behavioral assessments and the reading composite measure. Correlations were corrected for multiple comparisons using Bonferroni (15 comparisons).

2.3 Experimental design and task

EEG data with a high-density 128-channel coverage was acquired in an air-conditioned, electrically shielded, and sound-attenuated room, with participants seated at a distance of 92 cm from the display.

While EEG was recorded, participants performed an explicit audiovisual (AV) forced choice paradigm adapted from Jost et al. (2014), Figure 1, see Supplementary section 5 and Supplementary Videos for example videos of the W and PW conditions. The task tested the ability to evaluate AV congruency by presenting an auditory stimulus, followed by an AVcon or an AVinc visual item. AVcon spoken and visually presented items were identical across modalities (e.g., Ball–Ball), whereas AVinc items differed in their visual and auditory forms but always shared at least the initial speech sound (e.g., Ball–Bär; in English: ball–bear). The first speech sound was held constant to prevent children from solving the task by detecting an immediate mismatch at word onset, thereby requiring them to process the full orthographic form when judging AV congruency for orthographic conditions.

Experimental paradigm. Examples of audiovisually congruent (AVcon) and incongruent (AVinc) stimulus pairs are shown for each condition.

2.3.1 Paradigm procedure

For each trial, children had to indicate via button press (with their dominant hand) whether the visually presented stimuli matched (AVcon) or did not match (AVinc) the auditory reference presented in advance (see