Fluktuationen in den psychometrischen Kennwerten des hessischen „Kindersprachscreenings“ während und nach der Validierung

Abstract

Most German states assess German language skills of children of the late kindergarten age in large-scale language screening programs. In the state of Hesse, language test “Kindersprachscreening” (“Language screening for children”, KiSS) was developed for this purpose. Its target group are kindergarten children aged four- to four-and-a-half years. In the present study, changes in the psychometric characteristics of KiSS – level of difficulty, item-total correlations, internal consistency – and possible explanations for these changes were scrutinized. For this purpose, seven KiSS samples were re-analysed retrospectively: five from KiSS validation studies and two from the Hessian language screening program. Psychometric characteristics of KiSS improved for the monolingual German-speaking children in the language screening program in comparison with KiSS validation studies. Fluctuations in the item-total correlations and internal consistency were linked to changes in the difficulty level of KiSS tasks. In the data of bi-/multilingual children, no fluctuations of psychometric characteristics were detected. Some findings point at a selection bias – an unexpectedly high proportion of children needing language therapies – in the samples of KiSS validation studies and of the language screening program. This bias can result in incorrect KiSS cut-off values, if these are updated on the basis of data from the language screening program.