Introduction:

Although developmental models propose that reading and spelling evolve together, evidence shows that deficits in these skills can occur independently. Furthermore, regression-based analyses suggest that reading and spelling may be influenced by partly distinct cognitive predictors. However, previous research has predominantly focused on reading development, with comparatively limited attention given to spelling. In particular, longitudinal studies investigating changes in spelling development over time are still rare. Therefore, the aim of this longitudinal study is to identify predictors of early spelling development and their changes over time. Based on developmental models, we expect that the shift from phoneme-grapheme based spelling to the direct retrieval of orthographic representations will be reflected in changes in predictive patterns over time. In addition, developmental changes over time will be examined by comparing the cognitive profiles of children at risk of spelling difficulties with those of typically developing spellers across time points.

Methods:

The analyses are based on four time points from the longitudinal Topp-Spell study, which investigates spelling development in a representative sample of 445 German-speaking children from kindergarten through Grade 3. Predictors were assessed prior to the onset of formal literacy instruction and again in Grades 1 and 2. These predictors included phoneme awareness, orthographic processing skills (i.e., graphotactic sensitivity and morphological awareness), rapid automatized naming, and nonverbal IQ. Spelling outcomes were measured in Grades 1 through 3. Regression-based analyses are used to examine the predictive relationships among these variables and to identify changes in predictive patterns with literacy experience and increasing orthographic knowledge. To assess group differences in cognitive profiles related to spelling performance, children who scored at or below the 25th percentile on a standardized spelling test in Grade 1 (at-risk group) were compared with children who scored above the 25th percentile (typically developing group).

Results and Discussion:

A first series of regression analyses revealed that phoneme awareness was the only significant predictor in kindergarten explaining individual differences in spelling in Grade 1. This finding is in line with developmental spelling models and highlights the role of phoneme awareness as foundational predictor for phoneme-grapheme based spelling at the beginning of literacy instruction. In line with our expectations, the importance of phoneme awareness as a predictor decreased over time. Instead, the relevance of predictors reflecting the processing of larger orthographic units increased over time. In particular, graphotactic sensitivity in Grade 2 emerged as the strongest predictor of spelling performance one year later in Grade 3. In addition, the importance of IQ as a predictor of spelling increased over time, which likely reflects the growing cognitive demands associated with more complex spelling phenomena during literacy acquisition. Similarly, rapid automatized naming (RAN) emerged as a predictor over time, although it accounted for a relatively small proportion of variance in spelling performance. While RAN is considered a core component of reading fluency, presumably indexing efficient serial processing of letters and larger orthographic units, its role in spelling remains less well specified. Efficient serial processing may also support spelling by facilitating the sequential retrieval and production of orthographic units; however, its contribution is likely smaller than in reading, given the slower nature of spelling.

Group comparisons revealed better performance in the typically developing group across measures and time points, with the only exception being graphotactic sensitivity in kindergarten, where no group differences were observed. This finding is consistent with models of literacy development that posit an increasing role of orthographic processing as literacy skills develop.

Educational significance of the research:

The findings advance cognitive models of spelling development by demonstrating changes in predictive patterns and their relation to shifts in spelling development, from phoneme-grapheme based spelling to the processing of larger orthographic units and the direct retrieval of orthographic representations. These changes in predictive patterns have important implications for diagnostics and targeted interventions. Consistent with research based on English-speaking samples, phoneme awareness provides the foundation for acquiring letter-sound knowledge and phoneme-grapheme based spelling and should therefore be assessed and targeted in kindergarten and at the onset of literacy instruction. From the end of Grade 1 onwards, however, spelling instruction should increasingly focus on larger spelling units, such as frequent letter cluster, syllables and morphemes, to support the build-up of orthographic representations.

Introduction. Arithmetic and written language processing are foundational academic competences. Despite they may seem distant skill domains, their performance is positively related. Research evidence indicates a shared cognitive basis explaining this association. Both skill domains rely on the ability to form memory traces that allow the switch from a procedure-oriented to a retrieval-based approach. On the neural level, a limited number of studies investigated associations and dissociations between arithmetic and written language. These studies focused on functional overlap between acquired skills like arithmetic and reading. However, examining functional overlap among acquired skills does not in itself reveal shared underlying neurocognitive mechanisms. The very same brain region may be active during different tasks for different reasons. A more straightforward way to investigate shared and unique neural mechanisms involved in arithmetic and written language processing, is to examine overlap of brain activity involved in learning dynamics.

The current study investigated the neurofunctional correlates of arithmetic and orthographic learning and the overlap of brain regions jointly engaged in verbal fact learning across these two skill-domains. The intraparietal sulcus, which is commonly active in tasks that involve number processing, was expected to be specifically related to arithmetic processing. Brain regions specific to orthographic processing were expected to be located in the left ventral occipito-temporal cortex including the visual word form area. As anticipated in our preregistration (https://osf.io/tpjhf/overview?view_only=cc531097b9bb4bfc98f5b10a18f29ffe), we hypothesized to observe functional overlap related to verbal fact learning in frontal, parietal and middle temporal areas, including the hippocampus, because these are crucial hubs for the manipulation and consolidation of memory traces.

Methods. The sample of the current fMRI study consisted in 63 adults (age range: 18-45 years) and is well in line with our apriori power analysis, in which we estimated a minimum of 52 participants for our within-subject design. During scanning time, participants performed an arithmetic learning task with multiplication problems in the format XX · X = XXX and an orthographic learning task with pseudowords (e.g., Hafschmief). Each task was split into three runs, each run included stimuli presented repeatedly and other stimuli that were presented only once.

Data analysis is currently ongoing. We plan to conduct within-subject analyses of repeated vs. unrepeated stimuli across the three runs of each task. In second-level analyses, we will contrast brain activity at the end vs. beginning of the tasks.

Results and Discussion. Our findings will highlight brain networks related to learning dynamics and disentangle unique and shared functional substrates of different skill domains. The results of the present study will advance our understanding of how verbal fact learning mechanisms interact with neural substrates for symbolic processing, which are relevant for acquiring foundational academic skills.

Educational Significance of the Research. By identifying overlapping brain networks involved in verbal fact learning across skill domains, this study can inform instructional approaches that address shared memory mechanisms and learning dynamics. This neuroscience-based evidence will contribute to a deeper understanding of the neural basis underlying associations among academic skills in typical adults. This knowledge will be the starting point to investigate neural substrates of learning dynamics possibly contributing to the comorbidity of learning disorders (e.g., dyslexia, dyscalculia) in developmental samples.

Introduction

Reading instruction and interventions (hereafter referred to as reading instruction) improve students’ short-term reading outcomes. A growing body of research suggests that instructional effects diminish in the long run, a phenomenon known as fadeout. Relatively little work has systematically examined the magnitude of long-term effects of reading instruction and the extent to which these effects persist or fade. The present research addresses this gap by integrating two complementary meta-analytic investigations. The goal of Study 1 was to estimate the overall weighted average long-term impact of reading instruction on reading outcomes from preschool through Grade 12. Study 2 builds on Study 1. The goal of the second study was to model the extent to which reading instruction effects evolve from post-test to follow-up, with a specific focus on persistence and fadeout across multiple time windows.

Methods

Study 1: We conducted a comprehensive meta-analysis of experimental and quasi-experimental studies evaluating the long-term effects of reading instruction on reading outcomes. A systematic literature search was conducted. Thirty-five studies met the inclusion criteria, yielding 46 independent samples and 143 effect sizes (10,208 students in grades K-12). Effect sizes were Hedges’ g. We estimated the overall average weighted long-term effect of reading instruction and the moderation effects by participant and instruction characteristics using random-effects robust variance estimation.

Study 2: Study 2 extended Study 1 by explicitly modeling the temporal dynamics of instruction effects. To do so, we restricted the analytic sample to studies that reported both post-test and follow-up outcomes for the same reading skill. This allowed us to conduct within-skill comparisons of effects over time. This resulted in 117 eligible effect sizes drawn from the same pool of 35 studies. Follow-up assessments were grouped into conceptually meaningful time windows based on the delay from post-test: 1 week-4 months, 5-8 months, 9-12 months, and 12-24 months. Each time window was treated as a distinct outcome. Effect sizes were Hedges’ g. To examine fadeout trajectories, we first estimated weighted average effects within each follow-up window. We then estimated meta-regression models linking post-test effect sizes to follow-up effect sizes for the same reading outcome. In these models, the slope parameter represented the conditional persistence rate (i.e., the proportion of the post-test effect that remained at follow-up), while the intercept captured any follow-up effects not explained by post-test magnitude. This analytical approach allowed us to distinguish between sustained impacts attributable to initial gains and additional effects that emerged beyond the post-test period.

Results and Discussion

Study 1: The meta-analysis indicated that reading instruction produced a statistically significant long-term effect on reading outcomes (g = 0.295). Moderator analyses showed limited evidence that long-term effects differed across participant or instructional characteristics.

Study 2: Results from Study 2 indicated that instruction effects declined over time, with evidence of partial persistence. Average effects decreased from post-test to follow-up. Meta-regression analyses showed that persistence at the 5-8 month interval was significantly predicted by post-test magnitude, indicating that approximately 52% of post-test gains were maintained. At shorter (1 week-4 months) and longer (beyond 8 months) follow-up time windows, estimates were less precise, with little evidence of effects beyond those explained by post-test.

Together, these findings provide a coherent story of the long-term effects of reading instruction on reading outcomes. Reading instruction produces meaningful gains that extend beyond post-test, but these gains are not fully stable. Instead, effects show partial persistence and gradual fadeout, with sustained impacts largely conditional on the magnitude of post-test effects. This pattern aligns with broader evidence in psychology and educational research indicating that early gains require continued instructional support to be maintained. The findings underscore the importance of interpreting fadeout not as instructional failure but as a characteristic of learning-to-read trajectories in the absence of reinforcement.

Educational Significance of the Research

This research is significant because, first, it demonstrates that evidence-based reading instruction provides benefits that persist beyond immediate outcomes, supporting its long-term value. Second, by explicitly modeling persistence and fadeout, this work provides a framework for anticipating the extent to which instruction effects might diminish. The findings suggest that sustaining reading gains requires intentional alignment across instructional periods, rather than assuming that early reading gains will persist on their own.