Modelling Generic Problem-Solving Competence Underlying Domain-Specific and Domain-General Problem-Solving

Problem solving (PS) is currently regarded as an important skill for graduates entering the workforce (OECD, 2003a; Strijbos, Engels, & Struyven, 2015). In educational measurement, there are two distinguished perspectives in PS research, one pertaining to domain-specific PS competencies and the other to domain-general PS competencies. Although both PS perspectives are claimed to measure PS competence, there has been discussion as to whether performances in both measures (domain-specific PS and domain-general PS) reflect a common generic core aspect of PS competence (Fischer, Greiff, Wustenberg, Fleischer, Buchwald, & Funke, 2015; Raven, 2000). Using quantitative methods, the current study evaluates the extent to which Generic PS competence exists underlying both domain-specific and domain-general PS constructs. The Programme for International Student Assessment (PISA) Computer-Based Assessments (CBA) 2012 instruments, related reports, as well as secondary data were used in this study because they measured both domain-specific PS and domain-general PS abilities. In addition, the secondary data have the potential to reflect the diversity of participating countries. Therefore, the central question of the current research is “To what extent is the single Generic problem-solving construct, as manifested in the PISA 2012 CBA assessments, consistent across different fields, and how is it best modelled?” The answer to this research question was revealed in two main tasks devised in accordance with the model of the Standard for Educational, and Psychological Testing (AERA, APA, NCME, 1999, 2014) framework. The first task was developed to test content validity and, by means of examining PISA 2012 CBA instruments, sought to identify whether the conceptual meaning of Generic PS was present in three different PS constructs. The second task was based on the concept of structural validity in Messick’s validation framework. It was undertaken to examine three hypothesized models rep