The Language of Science and Readability. Correlations between Linguistic Features in TIMSS Science Items and the Performance of Different Groups of Swedish 8th Grade Students.

Abstract

This exploratory study examines how linguistic features of the Swedish TIMSS 2011 grade 8 science items correlate with results from different groups of students. Language use in different science subjects is analysed using four characteristic meaning dimensions of scientific language: Packing, Precision, Presentation of information, and Personification within the text, along with established measures of readability and information load. For each subject, one or more of the meaning dimensions show statistically significant correlations with students' performances with small to medium effect sizes. The results show that higher packing is positively correlated with students' results in earth science, negatively correlated in physics, and has no significant correlations in biology or chemistry. Placing items in everyday contexts reduces the likelihood of the items being answered correctly, and skilled readers are aided by higher precision in items, while less skilled readers seem unaffected. Many meaning dimensions that influence high performers' results do not influence those of low performers, and vice versa. The use of meaning dimensions is shown to be an enriching complementary method for analysing language use in science, as it connects the language use in items to student performance, while established measures do not.

Läsförmåga viktig för resultat på TIMSS

“…Tidigare studier tyder på att detta kan leda till sämre förståelse. I en studie publicerad 2016 har forskaren Tomas Persson vid Uppsala universitet granskat samband mellan resultat i TIMSS och språket i frågorna [1]. Resultatet visar att mycket information, i form av substantiv och långa ord, inte är ett hinder för eleverna.…”

ILSA: an automated language complexity analysis tool for French

“…The influence of language on item response has been demonstrated in several studies. For example, studies in Swedish (Persson, 2016); South African (Dempster & Reddy, 2007); and American (Martiniello, 2009) contexts have revealed the presence of language bias in standardised mathematics tests.…”