A question of time?—Relations between temporal auditory processing abilities and literacy skills over the course of primary school and how they are mediated by phonological processing

The role of temporal auditory processing for literacy development is still under debate. Current temporal auditory processing theories and models assume that relations between temporal auditory processing and literacy are mediated by phonological processing. However, no previous study with unselected children has simultaneously considered temporal auditory processing abilities from different timescales when testing this mediation. Additionally, it remains unclear whether mediation via phonological processing becomes weaker with increasing literacy experience. In the cross-sectional study presented here, German school children from grades 1 to 4 (N = 277) completed tablet-based tasks measuring rapid auditory processing, rhythmic auditory processing and phonological processing. Alphabetic and orthographic literacy were assessed with reading and spelling tests. Data were analyzed separately for children with early (grade 1 and 2) vs. more advanced (grade 3 and 4) literacy experience using multi-group structural equation modeling (SEM). In early readers and spellers, phonological processing was found to mediate the relations between temporal auditory processing (both rapid and rhythmic) and alphabetic literacy. In more experienced readers and spellers, phonological processing was found to mediate the relationship between rapid auditory processing and orthographic literacy. Direct relations between rhythmic auditory processing and literacy were significant only in early readers and spellers. Direct relations between rapid auditory processing and literacy were found only in more experienced readers and spellers. Thus, the findings indicate that in early literacy development, both rapid and rhythmic auditory processing are linked to (alphabetic) literacy via their effect on phonological processing. These results support key assumptions of temporal auditory processing theories and developmental models. In early literacy development, rhythmic auditory processing—presumably affecting suprasegmental (prosodic) speech processing—appears particularly relevant for literacy. In later literacy development, rapid auditory processing—likely influencing segmental (phonemic) speech processing—shows a stronger connection with literacy, partly mediated by phonological processing.

Reading is a complex skill and its acquisition, which relies on a network of motor, auditory and cognitive processes, can be viewed as one of the most complex human abilities (Tierney et al., 2021). To uncover the exact processes that are involved in reading development is crucial—not only for basic research on the foundations of reading and developmental dyslexia, but also for applied research in the fields of diagnosis and treatment. Only if the exact perceptual and cognitive mechanisms influencing reading development are known can diagnostic tools and treatment approaches be tailored to the specific needs of children at risk for or suffering from reading difficulties (Tierney et al., 2021). In the last decades, empirical research has firmly established that phonological processing is an important predictor of reading skills (Caravolas et al., 2012; Melby-Lervåg et al., 2012) and is impaired in developmental dyslexia (Ramus et al., 2003; White et al., 2006). However, whether temporal auditory processing abilities influence reading development and do so via their impact on phonological processing skills is still a matter of debate (see Protopapas, 2014; Zhang and McBride-Chang, 2010).

Are temporal auditory processing and literacy related?

It has been suggested that temporal processing impairments such as deficits in rapid temporal processing (Tallal, 1980) or deficits in rhythm and stress perception (Goswami, 2011; Goswami et al., 2002) might underlie developmental dyslexia (see also Protopapas, 2014): The rapid temporal processing theory of dyslexia by Tallal (1980) views stimulus duration and rapidity of stimulus change as crucial temporal parameters. The ability to process short or rapidly changing auditory information might be a prerequisite for intact phonological development and phoneme perception, as acoustic cues differentiating between phonemes may lie in time windows as short as 20–40 ms (Bishop, 1997; Tallal, 2000). Thus, the theory assumes that the basic temporal processing deficit disrupts the normal development of the phonological system. The deficits of phonological processing result in subsequent failure to read normally (Tallal, 1980; Tallal et al., 1998). The Temporal Sampling Framework by (Goswami 2011, 2022), on the contrary, focuses on the role of acoustic rhythm and views slower temporal modulations approximating syllable rate as central temporal parameters. Specifically, the Temporal Sampling Framework proposes that the automatic alignment (or sampling) of endogenous brain rhythms with rhythm patterns in speech is atypical regarding low-frequency envelope information in children with dyslexia (Goswami, 2022). With respect to the processing of non-linguistic/musical rhythms, this leads to problems in rhythmic timing and musical metrical sensitivity. With respect to the processing of speech rhythm, this results in poorer encoding of stressed syllables, syllables in general and onset-rime units in words, ultimately leading to impaired phonological learning. A study by Goswami et al. (2011) has shown that slower temporal modulations on the syllable level do also contain information that is important for phonetic discrimination. Thus, deficits concerning the processing of slower temporal modulations might also impair the development of phonological representations on the phoneme level (Goswami et al., 2011). The two temporal theories need not be mutually exclusive but might act on different timescales and might be relevant for different aspects of speech processing (Protopapas, 2014; Zhang and McBride-Chang, 2010).

Besides their relevance for developmental dyslexia, temporal auditory processing abilities might also be an important foundation for literacy development in general. (Zhang and McBride-Chang 2010) suggest in their developmental model that both rapid and rhythmic auditory processing abilities contribute to reading development. Indeed, various studies with predominantly unselected samples from different age groups show that rapid auditory processing abilities (compare rapid temporal processing theory above) are related to reading development (Hood and Conlon, 2004; Malenfant et al., 2012; Share et al., 1984; Walker et al., 2006) or to the development of both reading and spelling (Steinbrink et al., 2014a). In recent years, there is an increased research interest in the way that rhythmic auditory processing abilities (processing of auditory non-linguistic/musical rhythms, as exemplified for the Temporal Sampling Framework above) are related to reading and/or spelling development (see Ozernov-Palchik and Patel, 2018, for review). A multitude of studies shows that rhythmic auditory processing is related to reading and/or spelling abilities (e.g., David et al., 2007; Douglas and Willatts, 1994; Holliman et al., 2010; Steinbrink et al., 2019; Tierney et al., 2021) in unselected children.

But how exactly do rapid and rhythmic auditory processing influence reading development? (Zhang and McBride-Chang 2010) suggest that rapid and rhythmic auditory processing affect speech perception, which influences phonological processing that, ultimately, affects literacy development. According to their model, rhythmic auditory processing influences suprasegmental (prosodic) speech perception, while rapid auditory processing affects segmental (phonemic) speech perception (Note: This is in accordance with assumptions made in the context of temporal auditory processing theories of developmental dyslexia, Tallal, 1980; Goswami, 2011). Due to differences in the speech perception processes affecting phonological processing, the exact pattern of relations between temporal auditory processing and literacy might depend on the temporal auditory processing skill under consideration. However, to the best of our knowledge, existing studies with unselected children measured either rapid or rhythmic auditory processing abilities. Temporal processing on both timescales was only measured in studies including children with developmental dyslexia: In a study evaluating the discrimination of phonetic contrasts based on rapidly vs. more slowly changing temporal cues, children with dyslexia were impaired in processing slower, but not rapid temporal aspects of the phonetic contrast under consideration, supporting Goswami's temporal sampling framework (Goswami et al., 2011). Results of another study by Goswami et al. (2002) presenting correlational data of a large sample of children incorporating typically reading children as well as children with dyslexia suggest, however, that both rapid and rhythmic auditory processing are related to reading and spelling. Thus, it remains to be seen whether the same is true for unselected children.

Are relations between temporal auditory processing and literacy mediated by phonological processing? Does this mediation depend on literacy experience?

As outlined above, both the two temporal processing theories of developmental dyslexia (Goswami, 2011; Tallal, 1980) and the developmental model proposed by (Zhang and McBride-Chang 2010) assume that relations between temporal auditory processing and literacy are mediated by phonological processing. Models of literacy development (e.g., Frith, 1985, 1986; Ehri, 2005, 2014) suggest that children progress through various phases of literacy development. Alphabetic reading and spelling strategies (i.e., phonological strategies that rely on the use of phonemes and graphemes and their correspondences) are established before orthographic reading and spelling strategies (i.e., non-phonological strategies that rely on larger units such as morphemes; see Frith, 1985, 1986).

Thus, one might ask whether the mediating effect of phonological processing on relations between temporal auditory processing and literacy becomes less relevant with increasing literacy experience, as reading and spelling become less phonological. Few studies with unselected children have, however, addressed the question of how relations between temporal auditory processing and reading and spelling change with literacy experience. Most studies with school children concentrated on single school grades (e.g., Douglas and Willatts, 1994; Lê et al., 2020; Steinbrink et al., 2019) or combined children from different grades in the same analysis without focusing on potential differences resulting from differences in reading and/or spelling experience (Holliman et al., 2010; Malenfant et al., 2012; Tierney et al., 2021). In a longitudinal study by David et al. (2007), rhythmic auditory processing in grade 1 was related to reading ability in grades 1–5. No other studies have systematically investigated the role of temporal auditory processing for reading and spelling abilities in children with varying levels of literacy experience.

Ozernov-Palchik and Patel (2018) suggest that longitudinal and cross-sectional studies comparing readers with varying levels of reading experience would be a valuable step toward identifying the mechanisms underlying the associations between rhythmic auditory processing and reading: If rhythmic auditory processing influences phonological development, which ultimately influences reading acquisition, the mediating role of phonology should decrease with growing reading experience, as the involvement of the phonological system in reading decreases with increasing reading level. If more abstract mechanisms such as the ability to use rich contextual regularities to form predictions are responsible for the rhythm-to-reading link, then direct relations between rhythm and reading should strengthen with increasing reading experience (Ozernov-Palchik and Patel, 2018). Interestingly, Sousa et al. (2022) showed that phonological processing mediates the relation between rhythmic auditory processing and reading in first graders, whereas Lê et al. (2020) concluded that the relationship between rhythmic auditory processing and literacy is not mediated by phonological processing in third graders. With respect to rapid auditory processing, Malenfant et al. (2012) found that relations between rapid auditory processing and reading were mediated by phonological processing in 8-year-old children from second and third grade. As these three studies did not aim to compare children with different levels of literacy experience, they cannot inform us about potential changes concerning the mediation via phonological processing. They indicate, however, that (a) the way that relations between temporal auditory processing and literacy are mediated by phonological processing might differ for rapid vs. rhythmic auditory processing and that (b) at least for rhythmic auditory processing, a decrease of mediation with increasing literacy experience within the first school grades is a conceivable scenario (compare Ozernov-Palchik and Patel, 2018). Thus, future studies concerning relations between temporal auditory processing abilities and literacy should incorporate both rapid and rhythmic auditory processing in a systematic comparison of children differing in reading and spelling experience.

Are temporal auditory processing skills more strongly related to alphabetic than to orthographic literacy skills?

Even competent reading and spelling makes use of phonological processing (see, e.g., Ehri, 2014; Ziegler et al., 2001). Different literacy tasks do, however, differentially depend on these processes. For pseudoword reading, for example, phonological processing can be viewed as more central than for real word reading. In the following, we will refer to literacy skills that crucially depend on phonological processing (e.g., pseudoword reading) as alphabetic reading and spelling skills, while we refer to literacy skills that are less dependent on phonological processing (e.g., real word reading) as orthographic reading and spelling skills.

The strength of connections between temporal auditory processing and literacy might depend on the literacy skill under consideration. Temporal auditory processing theories of dyslexia (Tallal, 1980; Goswami, 2011) as well as the developmental model by (Zhang and McBride-Chang 2010) suggest that temporal auditory processing abilities influence phonological processing, which ultimately influences literacy development. Rapid auditory processing is assumed to influence the processing of smaller phonological units, i.e., phonemes (Tallal, 1980; Zhang and McBride-Chang, 2010). Rhythmic auditory processing is assumed to influence the processing of both smaller and larger phonological units (Goswami, 2011; Goswami et al., 2011; Zhang and McBride-Chang, 2010). Thus, both rapid and rhythmic auditory processing might be more strongly linked to alphabetic reading and spelling than to orthographic reading and spelling. Most of the pertinent literature is, however, not in favor of this notion: In studies with unselected samples measuring both alphabetic/non-word and orthographic/word reading, alphabetic as well as orthographic reading were related to temporal auditory processing (David et al., 2007; Tierney et al., 2021; Walker et al., 2006), without systematic differences in relationship strength. For spelling, Steinbrink et al. (2019) found stronger relations between rhythmic auditory processing and alphabetic spelling than between rhythmic auditory processing and orthographic spelling. The interpretation of these divergent results is unclear.

The current study

To the best of our knowledge the current study is the first to investigate rapid as well as rhythmic auditory processing and their relation to reading and spelling abilities in a sample of unselected school children. German primary school children from four grade levels (grades 1–4) were included into this cross-sectional study, allowing to analyze whether potential links between temporal auditory processing and literacy skills are dependent on reading and spelling experience. As both alphabetic and orthographic reading and spelling were assessed, we were also able to test whether alphabetic and orthographic literacy are differentially linked to temporal auditory processing abilities. Additionally, the inclusion of a measure of phonological processing allowed to test whether relations between temporal auditory processing and literacy are mediated by phonological processing. The following assumptions were made:

The pertinent research literature is in favor of the notion that temporal auditory processing abilities from different timescales are related to reading and spelling in unselected school children. Thus, we expect that both rapid and rhythmic auditory processing are associated with reading and spelling abilities in German primary school children.

Theories and models concerning temporal auditory processing and literacy (e.g., Goswami, 2011; Goswami et al., 2011; Tallal, 1980; Zhang and McBride-Chang, 2010) suggest that temporal auditory processing abilities influence reading and spelling via their influence on phonological processing abilities. As the involvement of the phonological system in reading and spelling decreases with increasing reading and spelling experience (see e.g., Frith, 1985, 1986), relations between temporal auditory processing and literacy skills should become less mediated via phonological processing over the course of primary school (see also Ozernov-Palchik and Patel, 2018).

Due to the suggested effect of temporal auditory processing on phonemic-level phonological abilities (Goswami et al., 2011; Tallal, 1980), relations between temporal auditory processing and literacy should be stronger for reading and spelling skills that especially draw on phonemic representations, i.e., alphabetic literacy skills, than for orthographic literacy skills (see the results of Steinbrink et al., 2019, for spelling in German).

Participants

Primary school children from six local schools in Thuringia/Germany were recruited for the study. In total, 395 first to fourth graders from 29 different classes participated. Out of these, 118 children had to be excluded from analysis (main reasons for exclusion: native language other than German, incomplete data sets; see section Statistical analyses for details). Thus, the final sample used for statistical analyses comprised 277 children (M age = 107.6 months, SD age = 14.3 months, 62% girls) of which n = 60 were first graders, n = 76 second graders, n = 59 third graders and n = 82 fourth graders. For further analyses, children were grouped with respect to reading and spelling experience (grade 1 and 2 vs. grade 3 and 4). Please note that formal reading and spelling instruction starts in Germany only after school entry, i.e., at the age of 6. Most children cannot read when they start school (see e.g., Schmitterer and Schroeder, 2019). In the current study, reading and spelling experience reflects children's experience with formal reading instruction. In German primary schools, the acquisition of written language in the first two years of school is focused on basic skills and takes place in the sense of “learning to read,” while from grade 3 onwards, reading and spelling are primarily used as tools for learning in the sense of “reading to learn” (see Thüringer Ministerium für Bildung, Wissenschaft und Kultur, 2010). Table 1 presents age means and standard deviations for early (grade 1 and 2; n = 136) vs. more experienced (grade 3 and 4; n = 141) readers and spellers.

The children's parents gave written informed consent for children's participation in the study, which was carried out in accordance with the Declaration of Helsinki. Before data collection, the purpose of the research was explained in detail to the children in an age-appropriate way. They were further informed about the option to terminate participation at any time. Families received no compensation for participation. Children's class teachers received with parental consent a report on students' results in the reading and spelling tasks. The study protocol was approved by the ethics committee of the University of Erfurt as well as the local school authorities (responsible for central Thuringia).

General procedure

Children were visited at their schools around the end of the school year. Data collection took place during morning hours of regular school days on two separate days. On the first day, participating children's phonological processing and temporal auditory processing abilities were assessed in a quiet classroom separate from their non-participating classmates. If more than eleven children from one class participated, participants were divided into two small groups. In that case, group sessions were administered simultaneously in two separate classrooms. Children's literacy skills were measured on the second day. While reading was assessed in individual sessions, spelling abilities were tested in group sessions including all participating children of a class. On both days, sessions lasted about 45 min. Assessments were carried out by the first author and a team of five trained research assistants. All task procedures were standardized based on fixed instructions. Research assistants underwent a two-week training phase right before the study started, where all tasks and instructions were practiced.

Procedure of measuring phonological processing and temporal auditory processing

To assess children's phonological processing and temporal auditory processing ability, a vowel length discrimination task (phonological processing and rapid auditory processing) followed by a rhythm reproduction task (rhythmic auditory processing) were conducted (see below for description). For both tasks a tablet application was used. Hence, each child received headphones and an iPad that was connected via Wi-Fi to an offline server in the classroom. Using headphones and iPads during classroom testing ensured that the items of all auditory tasks were presented with high audio quality and perfect time resolution. The connection to the offline server allowed the research assistants to control task initiations and to monitor task procedures. Technical as well as individual issues during testing could therefore be detected immediately. Both tasks used relatively short stimulus sequences and were low in task complexity, yet attractive to primary school children due to the tablet procedure. By this, we aimed at reducing potential confounds such as lack of motivation, misinterpretation of the task or short-term- or working memory limitations.

Both tasks started with a verbally given introduction explaining the subsequent task and, in case of the vowel length discrimination task, the different response options. Practice items were presented before each task. For the vowel length discrimination task, practice items involved visually presented feedback. In case of an incorrect answer, children had the opportunity to listen to the practice item again, ensuring that each participant had an exact idea of task requirements and tasks procedures. During the testing phase, no feedback was given, and children worked individually and at their own pace.

Measures

Prior to commencing the study, general socio-demographic information including children's age, gender and family languages was collected using a short parental questionnaire (adapted from Haase and Steinbrink, 2022). Only children speaking German as their native language were included into data analyses.

Vowel length discrimination task

To assess children's phonological processing and rapid auditory processing ability, a vowel length discrimination task was used (see Groth et al., 2011, and Steinbrink et al., 2014b, for a more detailed description of general characteristics of the task). This task makes use of the fact that vowel length is phonemic in German (e.g., /k a m/ Kamm, English comb vs. /k a: m/ kam, English came). Thus, vowel length discrimination is a phonological task in German. The task incorporates manipulated vowels besides natural vowels and can be used to measure phonological vs. rapid temporal vs. spectral processing. It was developed to reduce potential methodological confounds resulting from differences in task and stimulus complexity between phonological and temporal auditory processing tasks and was successfully applied in behavioral studies with dyslexic adults (e.g., Groth et al., 2011) and children (Steinbrink et al., 2014b), as well as an fMRI study with dyslexic adults (Steinbrink et al., 2012).

The task uses German monosyllabic pseudowords (fVp, nVp) that are presented in pairs. If the two pseudowords differ, they differ only with respect to vowel length. In German, vowel length is not only characterized by temporal, but also by spectral information. Besides syllables incorporating natural vowels, syllables in which a long vowel was shortened to the length of its short counterpart and syllables in which a short vowel was lengthened to the length of its long counterpart are used in the task. This allows the creation of three vowel length discrimination conditions for items of the category different: (1) a phonological condition in which vowels differ with respect to temporal and spectral information, (2) a temporal condition in which vowels differ only with respect to temporal information (note: the temporal characteristics differentiating long from short German vowels are in an appropriate range for the investigation of rapid auditory processing, compare Tallal and Piercy, 1975) and (3) a spectral condition in which vowels differ only with respect to spectral information. In former studies, differential response patterns for the phonological, temporal and spectral condition of the task reflecting individual item difficulties per vowel type were obtained, indicating that the three conditions of the task yield independent and valid results (see Groth et al., 2011, and Steinbrink et al., 2014b, for details). The task has already been successfully used with older primary school children (Steinbrink et al., 2014b). A pilot study amongst preschoolers and first graders that used a tablet-based adaption of the task as incorporated by Steinbrink et al. (2014b) revealed that the task is also appropriate for testing younger age groups.

In the iPad version of the task, German vowel pairs of different vowel heights (/i:-I/, /o:-O/ and /a:-a/) that are naturally characterized by a linear decline in temporal difference from the low vowel pair /a-a:/ (67 ms) over the mid vowel pair /o: - O/ (53 ms) to the high vowel pair /i: - I/ (40 ms) have to be discriminated. The task consists of two separate assessment blocks and a short rest period in-between. Within each block, 36 pseudoword pairs are presented acoustically. Children have to determine whether two consecutive pseudowords are either the same or different from each other by choosing a specific response button on the iPad screen. Out of the 72 items, 36 refer to the category same. These items are mainly used as control items to avoid response biases. The other 36 items account for the category different, of which 12 items belong to each condition (phonological vs. temporal vs. spectral condition), yielding a maximum score of 12 per condition. Each condition consists of four items from each vowel type (/i:-I/, /o:-O/ and /a:-a/). Given the rationale of the current study, only data from the phonological and the temporal condition were used in subsequent analyses. Internal consistency for the current sample was acceptable, with Cronbach's α = 0.657 for phonological vowel length discrimination and α = 0.676 for temporal vowel length discrimination.

Rhythm reproduction task

Tapping and rhythm production tasks have been used before in studies in the context of the Temporal Sampling Framework (Goswami, 2011; Kertész and Honbolygó, 2023). In our study, rhythmic auditory processing was assessed using a rhythm reproduction task from the Music Screening for Children I (Jungbluth and Hafen, unpublished) as used before in an iPad version by Steinbrink et al. (2019). In the previous study, the iPad version of the task was found to be a reliable (α = 0.902) instrument for preschool children and third graders (Steinbrink et al., 2019). Short-term- or working memory capacity does not influence the results obtained in this task (see Steinbrink et al., 2019).

The task includes 10 different rhythmic patterns that are presented in form of clave sounds. Each pattern consists of a minimum of 4 and a maximum of 8 pulses, which are composed of eighth notes, sixteenth notes, quarter notes, dotted eight and quarter notes as well as triplets in a 4/4 beat (see Appendix A1). After listening to a target rhythm, children have to repeat the rhythmic pattern as precise as possible by tapping on a blue button presented on the tablet screen. On completion of the reproduction task, children are able to hear their responses as their tapping also produces a clave sound. Children's responses were transferred immediately to the offline server via Wi-Fi and were saved to a database file. At a later point in time, the quality of the rhythm reproduction was rated by two trained assistants. Training involved professional guidance by the second author (a professor of music education) as well as more than 100 ratings of independent test items. The training period continued as long as necessary until a high level of consistency between the trainer and trainee was achieved. Intra-class correlations (ICCs) were used to calculate inter-rater reliability, which yielded satisfying ICCs (0.972 > ICC >0.766). Comparing children's responses with the original sound file, the raters assigned holistic ratings on a scale from 0 (rhythm is not recognizable) to 4 points (perfect solution). Ratings took into account the accuracy of the reproduced rhythmic pattern, stability of tempo, and the number of reproduced pulses (see Steinbrink et al., 2019, for details of the scoring system). Sum scores (with a maximum score of 40) were only calculated with missings ≤ 10% (max. one item). Internal consistency was again excellent (Cronbach's α = 0.902) in the current sample.

Spelling

Spelling was assessed using the Hamburger Schreibprobe (HSP; May et al., 2018). The HSP is a standardized test that was designed to measure fundamental spelling skills and the usage of spelling strategies (alphabetic strategy, orthographic strategy, morphological strategy). For each grade level, a separate version was applied (HSP 1+, HSP 2+, HSP 3+ and HSP 4–5). In the test, single words and short sentences are dictated. The analysis of spelling strategies is performed by focusing on the spelling of specific parts of words that is indicative of the use of the respective strategy. The manual of the HSP reports for the end of grades 1–4 an internal consistency of 0.82–0.88 for the alphabetic spelling strategy, of 0.91–0.92 for the orthographic spelling strategy and of 0.83–0.87 for the morphological spelling strategy. In the current study, orthographic and morphological strategy as measured in the HSP were analyzed together and subsumed under the label orthographic spelling as they both refer to spelling processes that are based on larger units for which phonological processing is less central. Raw scores for alphabetic and orthographic spelling strategies were calculated using an online tool provided by the HSP. Please note that the versions of the HSP differ per grade level in the number of word parts that are analyzed to determine the use of spelling strategies. This results in different maximum raw scores for the alphabetic and orthographic spelling strategy for different grades (see Table 1 for details).

Reading

Reading fluency was measured using the Salzburger Lese- und Rechtschreibtest (SLRT II, Moll and Landerl, 2014). The SLRT II distinguishes between two aspects of reading: direct word recognition (real word reading fluency, also termed orthographic reading in the following) and synthetic reading based on grapheme-phoneme correspondence rules (pseudoword reading fluency, also termed alphabetic reading in the following). The manual reports parallel form reliabilities of 0.94–0.98 for word reading and 0.90–0.96 for pseudoword reading for second to fourth grade. During testing, children were asked to accurately read as many words/pseudowords (presented on a paper card) as possible within one minute. Both reading subtests (word and pseudoword) were audio recorded to allow for later analysis. Raw scores for the number of correctly read (pseudo-)words were calculated. For both real word and pseudoword reading, the maximum score was 156.

Statistical analyses

In a first step, data from 102 children were excluded as they did not speak German as their native language and/or did not participate in both study sessions and all tasks. To allow for comparisons across grades, raw scores were then transformed into grade-specific z-scores for all further analyses. In a second step, we identified outliers (≥3 SD below/above mean scores; n = 16 children) which were then removed from all subsequent analyses. There were no missing data, except for 0.3% of missing values across single items of the rhythm reproduction task. We formed two groups to allow for comparisons between children varying in literacy experience, i.e., early readers and spellers (grade 1 and 2) vs. more experienced readers and spellers (grade 3 and 4). To determine basic associations between study measures, correlations were calculated through Pearson's product–moment correlations applying Šidák-correction to account for multiple comparisons (Type I error). Correlation analyses were executed separately for each of the two groups. We conducted structural equation modeling (SEM) using maximum likelihood (ML) estimation to test our main hypotheses (see Figure 1), i.e., whether measures of temporal auditory processing have a direct or indirect (mediated by phonological processing) effect on alphabetic and/or orthographic literacy skills and whether effects differ across groups of varying literacy experience. Thus, we included paths between temporal auditory processing variables (rapid vs. rhythmic auditory processing), phonological processing and two latent variables: alphabetic literacy (measures of alphabetic reading and spelling) and orthographic literacy (measures of orthographic reading and spelling) into our model. We further added covariances between the two measures of temporal auditory processing and between the two latent variables and we allowed correlated errors terms between spelling variables (HSP) and reading variables (SLRT II) to account for methodological similarities and shared method variances. We conducted multi-group SEM to test whether our model operates similarly or differently across groups.

Prior to calculations, we evaluated the data for general assumptions of SEM. While no violations were found for the linearity assumption, multivariate normality was not met [ χ (14) 2 = 279.06, p < 0.001; Doornik and Hansen, 2008]. We used bootstrapping (500 resamples) to address non-normality and limited sample size. Bootstrapping provides a powerful method for testing mediation effects even in smaller sample sizes (Cheung and Lau, 2008; Preacher and Hayes, 2008). Approximate fit indexes (goodness-of-fit measures) were assessed using root mean square error of approximation (RMSEA), comparative fit index (CFI) and standardized root mean square residual (SRMR). CFI ≥ 0.97 and RMSEA, SRMR ≤ 0.05 indicate good model fit, RMSEA, SRMR 0.05 ≤ 0.08 are indicators for acceptable model fit (e.g., Hu and Bentler, 1999). The analyses sequentially tested configural, metric and scalar invariance using established criteria (e.g., Putnick and Bornstein, 2016). All analyses were performed using STATA 17.0 (StataCorp LLC, College Station, TX: US) applying a significance level of α = 0.05.

Group-specific descriptive results of study variables are summarized in Table 1. In all auditory measures (phonological processing, rapid auditory processing, rhythmic auditory processing) mean scores of task performances increased over groups indicating improvements in phonological and temporal auditory processing abilities with increasing reading and spelling experience. Mean scores of the rapid auditory processing task remained at moderate level even in more experienced readers and spellers (mean = 4.9) indicating that the task was quite difficult for participants in general. The task is not characterized by floor effects, though, as even in early readers and spellers the achieved scores range between 0 and 9 points. A check of response patterns per vowel type within the rapid temporal vs. phonological condition of the vowel length discrimination task revealed that the results of this task were also in the current study valid at the level of individual items and independent for the two conditions (see further explanations in the Materials and Methods section). For most measures, ceiling effects were not observed. Please note, however, that for the alphabetic spelling strategy in grade 3, scores were close to the maximum possible value, suggesting reduced variability.

General associations between all variables of interest were computed separately for each group. Bivariate correlation coefficients (Šidák-corrected) are reported in Table 2. Correlation analyses showed significant relations between all literacy measures across groups. While phonological processing was significantly associated with alphabetic reading and spelling in less experienced readers and spellers, it was related to both alphabetic and orthographic reading, but not to spelling in more experienced readers and spellers. Rapid auditory processing was not associated with literacy skills in early readers and spellers but was significantly associated with both reading and spelling in more advanced readers and spellers. Rhythmic auditory processing was in both groups only related to spelling, not to reading, with stronger relations for less experienced readers and spellers. In this group, phonological processing was associated with both types of temporal auditory processing abilities (rapid as well as rhythmic auditory processing). In more experienced readers and spellers, phonological processing was associated only with rapid auditory processing. Rapid and rhythmic auditory processing were related only in children with greater reading and spelling experience.

SEM results with standardized parameter estimates per group are shown in Figure 1. Measurement invariances across groups are reported in Table 3. The model fit was good [ χ (10) 2 = 11.66, p = 0.308; CFI = 0.998, RMSEA = 0.035, SRMR = 0.028]. Given that the configural model demonstrated good fit, the same factor structure can be applied across groups. Subsequent models imposing equality constraints on factor loadings (metric invariance) and item intercepts (scalar invariance) all yielded (still) acceptable changes in fit indices, supporting full measurement invariance (Cheung and Rensvold, 2002; Hu and Bentler, 1999). Thus, group comparisons found in the subsequent structural model can be interpreted meaningfully. As the data fits well to the hypothesized model, no post-hoc modifications were conducted.

In early readers and spellers, rapid auditory processing (β = 0.27, SE = 0.07, p < 0.001, CI 95% = [0.14; 0.41]) and rhythmic auditory processing (β = 0.33, SE = 0.07, p < 0.001, CI 95% = [0.18; 0.47]) had a significant direct effect on phonological processing, which itself significantly influenced alphabetic literacy (β = 0.35, SE = 0.12, p = 0.004, CI 95% = [0.11; 0.58]). Furthermore, results showed a significant direct effect of rhythmic auditory processing on alphabetic literacy skills (β = 0.30, SE = 0.11, p = 0.006, CI 95% = [0.08; 0.50]) as well as orthographic literacy skills (β = 0.31, SE = 0.10, p = 0.003, CI 95% = [0.11; 0.51]). The direct paths from rapid auditory processing to the literacy skills factors did not reach significance in this model. Thus, rapid auditory processing only indirectly affected alphabetic literacy skills via phonological processing.

In more experienced readers and spellers, rapid auditory processing showed again a significant effect on phonological processing (β = 0.51, SE = 0.07, p < 0.001, CI 95% = [0.38; 0.64]), which itself affected orthographic literacy skills (β = 0.22, SE = 0.10, p = 0.032, CI 95% = [0.02; 0.43]), but not alphabetic literacy skills. However, no such indirect effect was found for rhythmic auditory processing, nor did we find any direct effects of rhythmic auditory processing on alphabetic and orthographic literacy skills (note, however, that the path from rhythmic auditory processing to alphabetic literacy was marginally significant, p = 0.051). Instead, rapid auditory processing directly influenced both alphabetic (β = 0.37, SE = 0.13, p = 0.003, CI 95% = [0.12; 0.61]) and orthographic (β = 0.27, SE = 0.11, p = 0.012, CI 95% = [0.06; 0.49]) literacy skills in this group. In addition, literacy experience (group variable) can be seen as a moderator of the covariance between both temporal auditory processing measures: A significant association between rapid and rhythmic auditory processing was only found in more experienced readers and spellers (β = 0.27, SE = 0.07, p < 0.001, CI 95% = [0.13; 0.40]).

We assessed rapid auditory processing, rhythmic auditory processing and phonological processing as well as alphabetic vs. orthographic reading and spelling in German children from the first four school grades and used SEM to gain new insights into the question whether rapid vs. rhythmic auditory processing are related to alphabetic vs. orthographic literacy and whether these potential relations are mediated by phonological processing. To address the role of literacy experience, data were analyzed separately for two groups, i.e., beginning readers and spellers (grades 1 and 2) vs. more experienced readers and spellers (grades 3 and 4).

Relations between rapid vs. rhythmic auditory processing and literacy are differential and depend on literacy experience

Correlational and SEM analyses showed that, as expected, both rapid and rhythmic auditory processing abilities were related to literacy skills. Unexpectedly, however, our analyses revealed that the relevance of the two aspects of temporal processing depended on literacy experience: Rhythmic auditory processing predominantly showed relations to literacy abilities in less experienced readers and spellers (exception: alphabetic spelling, see Table 2). Rapid auditory processing was associated with literacy variables only in more experienced readers and spellers. This result cannot be directly evaluated with respect to former studies as our study is, to the best of our knowledge, the first assessing both rapid and rhythmic auditory processing in unselected children and also the first specifically focusing on differences between more experienced and less experienced readers and spellers. The notion that rhythmic auditory processing might be especially important for early literacy development while rapid auditory processing gains in importance in later literacy development, can, however, be evaluated in light of the developmental model by (Zhang and McBride-Chang 2010; see also Wood et al., 2009): As outlined in the introduction, this model suggests that rhythmic auditory processing affects suprasegmental (prosodic) speech perception and that rapid auditory processing affects segmental (phonemic) speech perception. Sensitivity to speech prosody promotes spoken word recognition and spoken vocabulary growth, which ultimately fosters phonemic awareness (see lexical restructuring hypothesis, Metsala and Walley, 1998). Thus, it is conceivable that rhythmic auditory processing is of special importance earlier in development, as it promotes suprasegmental (prosodic) speech perception and indirectly the establishment of phonemic awareness. In later development, the ability to process fine temporal differences in the speech signal might be of special importance, as it fosters the fine-tuning of segmental (phonemic) speech perception.

Results from SEM revealed that rhythmic auditory processing directly affected alphabetic and orthographic literacy in less experienced readers and spellers. The direct rhythm-to-reading links might reflect children's ability to make predictions based on rich contextual structure in rapidly unfolding sequences, which is necessary both in rhythmic and in linguistic processing (Ozernov-Palchik and Patel, 2018; see also Lê et al., 2020). They might also be explained by prosodic awareness acting as a mediating variable between rhythmic auditory processing and literacy (Lê et al., 2020). Rapid auditory processing was directly associated with alphabetic and orthographic literacy in more experienced readers and spellers. This result is in accordance with those from a path model by Malenfant et al. (2012) finding direct effects of rapid auditory processing on reading comprehension in children from grade 2 and 3. These direct effects support the notion that rapid auditory processing does not solely influence literacy development via its effect on phonological processing. Instead, rapid auditory processing and phonological processing seem to form separate, but functionally related cognitive prerequisites for literacy acquisition (Malenfant et al., 2012).

Nevertheless, the result that rapid auditory processing was not related to literacy in early readers and spellers does not conform to findings from former studies (Hood and Conlon, 2004; Share et al., 1984; Steinbrink et al., 2014a). One might suspect that the lack of relations between rapid auditory processing and literacy might stem from an insensitivity of the rapid auditory processing task to measure temporal auditory processing skills in early readers and spellers. Indeed, this task was quite difficult for children from grade 1 and 2. Still, it yielded reliable results (α grade 1 = 0.626; α grade 2 = 0.645). Additionally, there was sufficient variability between children (range grade 1: 0–8 points, grade 2: 0–9 points; SD grade 1: 2.1, grade 2: 2.4). The task made use of temporal information in speech within a same-different paradigm (temporal vowel length discrimination), while former studies applied tonal stimuli in a repetition test or temporal order judgment tasks. In early grades, the ability to distinguish fine temporal differences in the speech signal might only influence literacy development via its impact on phonological processing and might only later directly assist the further promotion of literacy skills.

Mediation by phonological processing depends on literacy experience and temporal auditory processing skill addressed

To the best of our knowledge, our study is the first testing whether phonological processing mediates relations between rapid as well as rhythmic auditory processing and literacy in unselected beginning vs. more experienced readers and spellers. In beginning readers and spellers, relations between rapid/rhythmic auditory processing abilities and alphabetic literacy were mediated via phonological processing. This suggests that temporal auditory processing abilities support alphabetic literacy abilities via their effect on phonological processing at the beginning of literacy development. This suggestion is in accordance with assumptions made in the context of temporal processing accounts of developmental dyslexia (Goswami, 2011; Goswami et al., 2011; Tallal, 1980) and the developmental model by (Zhang and McBride-Chang 2010). The fact that mediation via phonological processing does not occur for rhythmic auditory processing in more experienced readers and spellers might be a result of reading and spelling being increasingly less dependent on phonological variables (Ozernov-Palchik and Patel, 2018). In this context, our results are in accordance with those of former studies showing that relations between rhythmic auditory processing and literacy skills are mediated by phonological processing in children from grade 1 (Sousa et al., 2022) but not in children from grade 3 (Lê et al., 2020). Rhythmic skills might help children to segment language into rhymes and syllables earlier in development (see also Goswami, 2011) but might stop playing a decisive role once phonological awareness is initiated (Lê et al., 2020).

In more experienced readers and spellers, relations between rapid auditory processing and orthographic literacy were mediated by phonological processing. This result is in principle compatible with that of Malenfant et al. (2012), although these authors focused on reading comprehension. Mediation via phonological processing specifically affected orthographic but not alphabetic literacy skills, indicating that in more experienced readers and spellers rapid auditory processing abilities assist in using fine phonological differences in the context of orthographic reading and spelling processes. In German, orthographic markings of vowel length can be exploited to produce vowels of correct vowel length in reading (e.g., a double consonant following a vowel marks a short vowel phoneme, a “silent h” following a vowel marks a long vowel phoneme: La mm (engl. lamb) vs. la h m (engl. lame) = > /l a m/ - /l a: m/). In the same vein, the detection of subtle spectral and temporal differences between vowel phonemes is a prerequisite for learning and using such orthographic markings of vowel length in spelling.

Relations between temporal auditory processing and literacy are not stronger for alphabetic literacy skills

Our hypothesis that relations between temporal auditory processing and literacy are stronger for alphabetic than for orthographic literacy due to effects of temporal auditory processing on phonemic-level phonological abilities (Goswami et al., 2011; Tallal, 1980), was predominantly not confirmed. Firstly, the correlation analysis revealed no differential pattern of relations between temporal auditory processing and alphabetic vs. orthographic reading and spelling skills. Secondly, in SEM analyses, rhythmic auditory processing directly affected both alphabetic and orthographic literacy in early readers and spellers. Similarly, rapid auditory processing was directly associated with alphabetic as well as orthographic literacy in more experienced readers and spellers. Finally, mediation via phonological processing was observed not only for alphabetic but also for orthographic literacy. These findings indicate that temporal auditory processing abilities are related both to literacy skills that strongly rely on phonological processing and to those that depend on it to a lesser extent. This is exactly in line with results from former studies with English-speaking participants (David et al., 2007; Tierney et al., 2021; Walker et al., 2006). It is interesting, though, that the only finding supporting our hypothesis, i.e., a restriction of relations between rhythmic auditory processing and literacy to alphabetic spelling in more experienced readers and spellers, is coherent with the results by Steinbrink et al. (2019).

How can the result that temporal auditory processing abilities are not only connected to alphabetic, but also to orthographic literacy skills be explained? On the one hand, alphabetic skills can be viewed as a foundation for the later development of orthographic skills in literacy, as an efficient storage of orthographic patterns requires multiple associations between phonological segments of a spoken word and the corresponding graphemes of its written form (Moll et al., 2014; Caravolas et al., 2001). On the other hand, temporal auditory processing seems to influence literacy skills not only via its impact on phonological processing, but also in a direct fashion. If a more abstract mechanism—such as the ability to detect and exploit regularities across rhythmic and linguistic input—underlies the association between rhythmic auditory processing and literacy (Ozernov-Palchik and Patel, 2018), this ability should also aid at learning and using orthographic regularities. Similarly, if rapid auditory processing assists in detecting subtle temporal differences in the speech signal, this may contribute to the development and application of orthographic competencies.

Limitations and suggestions for future research

Our results must be evaluated with respect to limitations of the study. Based on theoretical considerations and research questions, we decided to distinguish between alphabetic vs. orthographic literacy in the SEM in the current study. Given that reading and spelling are sometimes referred to as separate constructs (e.g., Steinbrink et al., 2019) or are analyzed together as one literacy variable (e.g., Lê et al., 2020), we also tested alternative models. However, these models could not be estimated due to convergence issues (reading vs. spelling) or showed unsatisfactory fit indices [one literacy variable: χ (22) 2 = 127.35, p = < 0.001; CFI = 0.846, RMSEA = 0.186, SRMR = 0.080]. This provides indirect support for the proposed model.

SEM generally requires large sample sizes to provide reliable and stable estimates. Therefore, we used bootstrapping technique to improve the stability of parameter estimates and to correct for bias in our sample. However, future studies should consider larger sample sizes.

Our study shares the general problem of all studies in the field that the tasks employed to measure rapid auditory processing and rhythmic auditory processing are rather heterogeneous (see, e.g., Farmer and Klein, 1995; Hämäläinen et al., 2013 and Ozernov-Palchik and Patel, 2018, for review). A greater consensus about the tasks employed would be beneficial, as it would heighten the interpretability of the results (see also Lê et al., 2020). In this context, the proposed multidimensionality of rhythmic skills has to be taken into account (e.g., Fiveash et al., 2022; Tierney and Kraus, 2015). In future studies investigating rhythmic auditory processing, motor demands of the respective tasks might be reduced or eliminated by using rhythm production tasks requiring synchronous tapping (see Lê et al., 2020) or by focusing on rhythm perception tasks (see Sousa et al., 2022).

Furthermore, the current study did not control for socioeconomic background. Both musical experience and early literacy support tend to be more prevalent in families with higher socioeconomic status (e.g., Bradley and Corwyn, 2002; Corrigall and Schellenberg, 2015; Niklas and Schneider, 2013). Thus, relations between rhythmic auditory processing and literacy might be partly confounded by socioeconomic factors. Future studies should include measures of socioeconomic status and musical expertise to disentangle the contributions of temporal auditory processing abilities from those of environmental factors.

For the current study, we cannot exclude the possibility that differences in stimuli and in task demands between the rapid vs. rhythmic auditory processing task influenced the results. We would, however, like to stress that the focus of the current study was on differences with respect to literacy experience and that the results of the study altogether seem plausible as they were widely consistent with former empirical evidence and interpretable in the light of current theoretical accounts.

Finally, in the current study, we used a vowel length discrimination task to measure phonological processing. This task was developed to keep task demands easy and comparable between phonological and temporal processing conditions (see Groth et al., 2011, and Steinbrink et al., 2014b, for details). Future studies should test whether our results concerning mediation via phonological processing are confirmed when more conventional phonological processing tasks involving a manipulation component (e.g., phonological awareness tasks) are employed. Additionally, they should systematically explore whether temporal processing acting at different timescales is specifically linked to specific aspects of phonological processing (see also Protopapas, 2014). It is, for example, theoretically plausible to assume that rhythmic auditory processing might be more closely related with phonological awareness for larger phonological units, while rapid auditory processing might be more closely associated with phonological awareness for smaller phonological units.

The current cross-sectional study gave new insights into relations between temporal auditory processing and literacy and the possible mediating role of phonological processing by (a) addressing temporal auditory processing abilities from different timescales (rapid vs. rhythmic auditory processing) in the same sample of primary school children, (b) investigating the role of literacy experience by comparing beginning and more experienced readers and spellers and (c) testing differential links between processing abilities and alphabetic vs. orthographic literacy. In accordance with a developmental model by (Zhang and McBride-Chang 2010) and temporal processing theories of developmental dyslexia (Goswami, 2011; Tallal, 1980), phonological processing mediated the relations between both aspects of temporal auditory processing and (alphabetic) literacy in beginning readers and spellers. In more experienced readers and spellers, however, mediation of relations between temporal auditory processing and literacy via phonological processing was restricted to rapid auditory processing, which corroborates results from former studies. The pattern of direct relations between temporal auditory processing and literacy was also in favor of the notion that rhythmic auditory processing is especially important for early literacy skills, while rapid auditory processing gains in importance in more experienced readers and spellers. This might be explained by differential links between rapid vs. rhythmic auditory processing and the processing of speech: Rhythmic auditory processing might influence suprasegmental (prosodic) speech processing, thereby indirectly fostering the development of phonemic awareness in early reading and spelling. Rapid auditory processing, on the contrary, might specifically help to further shape segmental (phonemic) speech processing in children with more literacy experience.