Evaluation of cognitive ability in heterogeneous student populations: Development of the Test of Cognitive Potential

Evaluation of cognitive potential of individuals with diverse backgrounds often presents a challenge. Individuals’ imperfect mastery of the language of instruction in particular presents a major threat to validity of results and prohibits comparison to performance with better language skills. Eliminating language from test instruction may be a solution to this problem. The present thesis presents the development and validation of the innovative Test of Cognitive Potential (TCP), a language-free, tablet-based group-assessment tool designed to measure fluid intelligence within children in grade 4 (regular age between 9 and 10). Using dynamic visual instruction videos, task requirements are presented visually, without adding any verbal clarification. In order to sample a wide array of cognitive processes on different contents (domain-general, quantitative and semantic-syntactic reasoning as well as visual processing), the TCP includes a variety of subtests. Using confirmatory factor analysis, the overall structure of the test is evaluated within student samples from Luxembourg, Germany and Brazil. In a 2x2 design, measurement invariance of a latent TCP g-factor is further shown to largely hold across combinations of verbal vs. visual dynamic instruction and tablet-based vs. paper & pencil assessment. For the target TCP version (tablet and dynamic visual instruction), only one subtest is found to present increased residual variance and thus needs revision. Overall, a strong association of the TCP with criterion measures (educational achievement and a well-established test of cognitive ability) is found across samples. Measurement of a latent TCP factor could further be shown to be invariant across students with differential opportunity to learn. No strong evidence of biased measurement is found for students with different language backgrounds and socio-economic status. The Test of Cognitive Potential proofs to be a valid and reliable tool for the measurement of general cognitive ability. It will be particularly useful for the evaluation of children with limited understanding of the local language. Avoiding language bias, the TCP may contribute to the reduction of inequalities in educational opportunities that are produced by structural characteristics of many school systems.