The Litkey Corpus: How a longitudinal corpus of primary school children’s texts can be used for studying spelling and writing development in German

We present research on the Litkey Corpus ( Laarmann-Quante et al., 2019c ), a longitudinal collection of German texts produced by primary school children between 2nd and 4th grade. With its multilevel annotations on the word level, including phonemes, syllables, morphemes, and error annotations, the corpus is particularly suitable for the study of orthography acquisition. This corpus-based approach complements experimental studies on spelling development, and offers a range of methodological approaches that come with their own strengths and weaknesses, as we will showcase in this contribution. Laarmann-Quante (2021) used machine learning methods to predict how likely words are to be misspelled and what (combination of) features make words particularly error-prone. In lemma-level analyses, the corpus has been used for studying vocabulary growth during the primary school years. In two pilot studies, we used a network science approach and explored various measures of lexical diversity to characterize how the lexical knowledge of children with different levels of orthographic proficiency develops ( Wieler, 2026 , Laâguidi et al., 2024 ). As a final example, we show how the Litkey Corpus can be used to study the development of syntactic complexity in the children’s texts, even in the absence of syntactic and morphosyntactic annotations ( Laâguidi et al., 2024 ). With each of the studies we review, we discuss important caveats that result from idiosyncrasies of working with corpus data. In closing, we discuss desirable extensions of the corpus that would allow it to be used to study morphosyntactic development.