Steuerung zukünftiger Aufgabenentwicklung durch Vorhersage der Schwierigkeiten eines Tests für die erste Fremdsprache Englisch durch Dutch Grid Merkmale

The present study focuses on the construct validity of English as a foreign language reading and listening comprehension items. Corresponding items (N = 338 reading comprehension items, N = 289 listening comprehension items) were developed at the Institute for Educational Quality Improvement. One parameter logistic model item difficulties are predicted by item characteristics referring to cognitive demands, as well as content features of items, derived from the Dutch Grid catalogue (Alderson et al., 2006). A multilevel framework is used to identify the simultaneous impact of stimuli (i.e., the text heard or read) and item characteristics on empirical difficulties. Results show weak effects of Dutch Grid characteristics on reading comprehension item difficulties. More variance is explained for listening comprehension item difficulties, partially due to construct-irrelevant item features (i.e., answer formats). Implications for construct validity and systematic recommendations for future item development processes are discussed. [Editor]