Development of CDI-III for children over the age of 3 years

Thematic Section: Using parental reports to examine early language development in bilingual children: CDIs and beyond

early language development, vocabulary acquisition, first words, language input, parental reports

Grzegorz Krajewski, Faculty of Psychology, University of Warsaw
Karolina Mieszkowska, University of Warsaw
Magdalena Łuniewska, University of Warsaw
Agnieszka Kacprzak, University of Warsaw
Ewa Haman, University of Warsaw

In this talk we will present the work on the Polish adaptation of CDI-III: the proposed form and structure of the tool, the adopted procedure for its development, and the current state of the work with the focus on the results of the pilot study.

Pl-CDI-III will be based on the design of Swedish adaptation, keeping its division into a vocabulary scale with 100 items, a shorter grammar scale, and one general question on pronunciation (Eriksson, 2017). We will discuss the decision to drop a metalinguistic awareness scale and consider alternative wordings of the pronunciation question.

The general procedure for constructing the vocabulary and grammar scales is to generate initial pools of items approximately twice as large as the final scales, estimate their psychometric characteristics through piloting, and select the final sets based on those estimates. In generating potential items we have drawn from both SCDI-III and the large dataset from the norming study of the Polish CDI W&S (IRMIK SiZ; Smoczyńska et al., 2015), in particular the corpus of sentences gathered in the “sample utterances” and “longest sentences” sections (IRMIK sentences corpus).

Lexical items come from four sources: (i) translations of words used in SCDI-III and/or culturally appropriate alternatives from the same thematic categories, (ii) words from IRMIK SiZ that remain particularly difficult for the oldest children in the norming sample, (iii) content words from the IRMIK sentences corpus based on their frequency, and (iv) frequency lists from Haman et al. (2011). Grammatical structures and constructions are selected based on the analysis of the IRMIK sentences corpus.