Investigating evidence in support of validity and reliability for data collected with the meaningful learning in the laboratory instrument (MLLI)

Abstract

The Meaningful Learning in the Laboratory Instrument (MLLI) was designed to measure students’ expectations before and after their laboratory courses and experiences. Although the MLLI has been used in various studies and laboratory environments to investigate students’ cognitive and affective laboratory expectations, the authors of the instrument reported a discrepancy between the intended factor structure of the MLLI and the factor structure suggested by the data collected in preliminary studies. Therefore, the aim of this study was to investigate the validity and reliability evidence related to data collected with the MLLI, especially that related to structural validity. Evidence to support structural validity would provide greater meaning for the reporting and interpretation of MLLI scores. In this study, two possible a priori models for the factor structure of data collected from multiple institutions with the MLLI were investigated using confirmatory factory analysis (CFA). This initial investigation found poor data-model fit for each of the two tested models. Cognitive interviews and free response items were then used to inform modifications to the two a priori structures, and a third alternative structure, which included a negative method factor, was also investigated. Once a best fitting model was identified, further model revisions were informed by a combination of modification indices and qualitative data. Evidence of adequate-to-good data model fit was found for the final revised version of the MLLI, deemed the MLLIv2. Additionally, evidence of both internal structure validity and single administration reliability were found for each of the MLLIv2 factors. The structure of the data from these items leads to scale scores that likely represent student expectations that contribute to meaningful learning and student expectations that detract from meaningful learning. As the results of this study provide the first psychometrically supported scales for MLLI data, they have implications on the future reporting and analyses of MLLI scores.