Competence-based education has become a favourite model for innovations in VET and stimulated research on assessment for learning. Efforts have also been made to construct tools for large scale assessment - along three main lines. One is to identify compe tencies on the basis of job requirements and specific work tasks. The second strategy is represented by a German initiative to develop a so called PISA-VET or VET-LSA that combined the previous one with an attempt to measure cross-occupational competenci es like social skills. A third alternative is to be found in the COMET-model which is a platform for competence diagnostics that underlines the need for a comprehensive (holistic) measurement instrument in the field of VET. The COMET and the VET-LSA-proje cts differ in terms of principles for test construction. Whereas the later selects samples of pre-defined occupational tasks from a database, the procedure in the COMET-project is to construct a limited set of multidimensional tasks for each occupations in workshops. The output is open task assignments and a set-up for rating based on judgmental consensus and criterion-related guidance - a logic also to be found in recent OECD prototypes like AHELO (higher education generic skills) and PIAAC (adult comp etencies). In the MECVET we will address the methodological challenges associated with prototyping a measurement tool for competence development that builds on the COMET-model and introduce methods for validating both the problem solving strategies of tra inees and the judgments of the raters. Our study will have a cross over longitudinal design over the 2., 3. and 4. year in VET and the learning trajectories of the students/trainees will be compared with questionnaire data on background factors, vocation al commitment, school environment, and the work place as learning environment.