Introduction:

Spelling proficiency is commonly defined as the ability to write many words correctly and largely automatically across different writing situations (Hinney et al., 2018). German orthography is based on an external, codified norm that learners largely internalize during literacy acquisition; consequently, deviations from this norm are particularly salient in educational contexts, where attention often focuses on orthographic errors. At the same time, spelling performance has been shown to differ between integrated writing tasks (text composition) and non-integrated tasks such as dictations or gap-fill exercises (Lemke, 2021). Consequently, for the teaching and improvement of spelling, it is essential to know which orthographic areas are the most difficult. Ideally, for students with weak spelling skills, a standardized assessment is complemented by a qualitative error analysis in which errors are assigned to linguistic categories, e.g., lowercase instead of uppercase (<*haus> for <Haus> house), missing double vowels (<*dof> for <doof> stupid) in order to identify problematic orthographic domains and derive targeted instructional measures (Thomé & Thomé, 2020).

Traditional qualitative error analyses, however, face two methodological limitations: i) correctly spelled items within the same category are typically not considered, and ii) orthographic categories occur with unequal frequency. Attempts to compensate for this bias, for example by determining base rates (Reichardt, 2015), have so far relied on small samples, as manual analysis is time-consuming.

Methods:

This presentation reports empirical, corpus-based findings on deviations between actual spellings and target norms based on an aggregated corpus of 1.8 million tokens from approximately 13,000 freely written texts produced by students aged eight and above. The aim is to identify difficulty-determining features of German orthography independently of individual learner characteristics. The analysis is enabled by the data infrastructure of the writing platform iderblog.eu, funded by the German Federal Ministry of Education and Research (BMBF) within the DigitalPakt Schule (2019–2024), which applies learning-analytics methods to collect and analyze learner data in order to better understand and optimize learning processes (Solar, 2022). Due to the structure of the database, orthographic errors are assigned to higher-level categories and subordinate phenomena; in addition, for each misspelled word form, the corresponding frequency of the correctly spelled word form is also available. In addition to traditional error analysis, error rates are calculated by relating each misspelled form to the corresponding correctly spelled forms in the corpus.

Results and Discussion: The results show that these different methodological approaches lead to markedly different rankings of orthographic problem areas. Selected orthographic categories, such as umlaut spelling, are examined in greater detail to determine which specific phenomena (e.g., inflectional versus derivational contexts) are particularly prone to errors. Although there are also limitations regarding the data basis, since only already categorized misspellings can be included in the analysis, this evaluation represents a completely new approach and contributes to a deeper understanding of orthographic challenges in the German language.

Educational significance:

The findings provide an empirical basis for the development of a computer-based, gamified approach to spelling instruction, the basic design of which will be outlined. For this purpose, depending on the results of the various evaluations of the qualitative error analysis, additional explanations and exercises are provided, with particular emphasis on the most problematic categories. Teachers can also benefit from the results, as they enable them to focus more strongly on problematic orthographic areas in their teaching on an empirical basis.

Introduction: Written language contains multiple levels of statistical regularities to which skilled readers become highly sensitive. These regularities operate at both lexical and sublexical levels and jointly shape the efficiency of visual word recognition. At the lexical level, word frequency reflects how often a word occurs in a language and is known to facilitate access to stored lexical representations, resulting in faster and more efficient processing. At the sublexical level, readers are sensitive to orthographic legality, that is, whether a letter string conforms to the permissible letter combinations of a given writing system, and to bigram frequencies, which captures how often adjacent letter pairs co-occur.

Previous behavioral and electrophysiological research has shown that these statistical properties influence distinct stages of word processing. Early event-related potential (ERP) components, such as the N1, are associated with visual-orthographic analysis, while later components, most notably the N400, are linked to lexical-semantic access. However, the extent to which lexical and sublexical statistical regularities interact across these temporal stages remains an open question.

Thus, the present study aimed to disentangle the contributions of lexical frequency, orthographic legality, and bigram frequency to the neural dynamics of visual word recognition. Using ERPs, we examine how these factors modulate early and late processing stages to address the following questions: (1) How do lexical frequency effects manifest in early and late ERP components? (2) How do orthographic legality and bigram frequency influence early visual-orthographic processing? (3) How do lexical frequency and sublexical regularities interact across early and late stages of word recognition?

Methods: Thirty-two native Hungarian-speaking adults (18–36 years, 10 male) with normal or corrected-to-normal vision participated in the study. EEG was recorded while participants performed an implicit reading task in which they indicated on a subset of trials presented on a yellow background whether the stimulus was identical to the immediately preceding one.

Stimuli consisted of five categories: high-frequency words, low-frequency words, legal pseudowords with high bigram frequency, legal pseudowords with low bigram frequency, and orthographically illegal pseudowords with either high- or low-frequency bigrams. All stimuli were matched for length and basic visual properties. ERP analyses focused on the N1 time window, indexing early visual-orthographic processing, and the N400 time window, indexing lexical-semantic processing.

Results and Discussion: In the N1 time window, high-frequency words are predicted to elicit smaller amplitudes than low-frequency words, consistent with more efficient early orthographic processing of familiar lexical items. For pseudowords, orthographic legality is expected to exert a robust effect, with illegal strings producing larger N1 amplitudes than legal pseudowords, reflecting increased processing demands for illegal letter combinations. Within legal pseudowords, lower bigram frequency is supposedly associated with larger N1 amplitudes than higher bigram frequency, indicating sensitivity to sublexical statistical structure during early visual analysis. Bigram frequency effects are predicted to be most pronounced for real words and legal pseudowords, suggesting that sublexical statistics are most effectively exploited when letter strings conform to the orthographic system.

In the N400 window, both legal and illegal pseudowords are expected to elicit larger amplitudes than real words, reflecting the absence of lexical-semantic representations. Lexical frequency effects are predicted to be reduced at this later stage, while differences between words and pseudowords dominated the response pattern. Together, these findings indicate temporally distinct contributions of lexical and sublexical regularities to visual word recognition.

Educational Significance: Understanding how readers extract and utilize statistical regularities from print has direct implications for literacy education. The present findings highlight that sensitivity to orthographic structure and sublexical statistics emerge at early stages of processing and support efficient word recognition. Educational approaches that emphasize exposure to well-formed orthographic patterns and frequent letter combinations may facilitate the development of fluent reading. Moreover, distinguishing between early visual-orthographic analysis and later lexical-semantic processing may help refine instructional strategies and interventions, particularly for readers who struggle with reading acquisition despite adequate exposure to print.

Introduction

General orthographic knowledge – the implicit knowledge of rules and regularities within a writing system – is a predictor of reading skills (Conrad et al., 2013). In our increasingly multilingual society, bilingual children often learn to read in both their instructional and their heritage language, being exposed to two sources of orthographic input. It remains unclear, how general orthographic knowledge from two languages is organized. Research on lexical processing in bilingual individuals has proposed two major frameworks that we base our predictions on: The selective access hypothesis proposes that bilinguals develop separate, language-specific systems to store lexical representations (Kroll & Stewart, 1994). Either language is then selectively activated or inhibited, depending on task demands, limiting interactions between them. The non-selective hypothesis assumes that representations of all languages are stored in a shared, integrated system, with all languages activated simultaneously during word recognition (Dijkstra & Van Heuven, 2002). This allows for cross-language interaction, as well as possible interference between languages. The aim of the current study was to investigate whether general orthographic knowledge is stored in separate, language-specific systems, or one integrated system. To this end, we assessed general orthographic knowledge in bilingual primary school children and examined whether there are cross-language contributions of general orthographic knowledge to literacy skills in both languages.

Methods

A total of 80 bilingual children who read and speak in German and Croatian participated in this study. All children were in 3rd or 4th grade of primary school in Austria. Reading, spelling, vocabulary and naming speed were assessed in both languages.

General orthographic knowledge was assessed via a forced-choice letter detection task. Participants were presented with a series of bigrams and asked to indicate, whether each bigram contains a target letter or not by pressing corresponding response keys. Bigram frequency in both languages was orthogonally varied, resulting in four conditions: High frequency in both orthographies (HH), high in one and low in the other (HL/LH), and low frequency in both (LL). Accuracy and reaction times were recorded.

Reading fluency was assessed with a one-minute reading task. Participants read out as many words as they could from a list of words within one minute. The score is the number of correctly read words per minute.

Spelling was assessed by verbally presenting participants with a target word and a simple example sentence. The score is the number of correctly spelt words.

Vocabulary was assessed with the Peabody Picture Vocabulary Test. Children were asked to indicate which of four pictures corresponds to a verbally presented word. The score is the number of correctly recognized words.

Naming speed or rapid automatized naming (RAN) was assessed by presenting the children with 30 single-digit numbers and asking them to name them as quickly as possible. The score is the time needed to successfully name all numbers in seconds.

Phonological awareness was assessed using a nonword phoneme elision task. Participants were verbally presented with 25 nonwords and asked to repeat it. They were then asked to repeat it again while eliciting a specified phoneme. The score is the number of correct phoneme elisions divided by the number of correct repetitions.

Results and Discussion

Hierarchical regression analyses revealed that both German and Croatian word reading were predicted by children’s knowledge about Croatian orthographic patterns, beyond naming speed and phonological awareness. This was not the case for German orthographic knowledge. This pattern may indicate that orthographic knowledge derived from a highly transparent writing system (e.g., Croatian) can serve as a reliable source of sublexical information that supports word reading across languages. German spelling was predicted by German and Croatian orthographic knowledge. Croatian spelling was only predicted by Croatian naming speed. These findings provide evidence for the cross-language activation of bilingual orthographic knowledge in literacy skills.

Educational Significance

As bilingual children can demonstrate cross-language activation of general orthographic knowledge, supporting the development of orthographic knowledge in the heritage language may benefit the development of literacy skills in the instructional language. On the one hand, the Croatian-German bilingual participants in this study constitute one of the largest ethnic minorities in Austria. More broadly, these implications are of significance for children with different native languages as the instructional language, as well as their educators.