Georg Rasch developed his probabilistic models and approach to measurement in the context of psychological and educational research. He began his work on psychological measurement in 1945 when he helped to standardize an intelligence test for the Danish government and went on in 1951 to investigate the impact of supplementary education for poor readers. The particular problem faced by Rasch was that the children, now 20 years old, had been tested using different instruments. He solved this problem by testing a sample of then current school children on the different tests. He was able to put both tests on the same scale, and was therefore able to establish whether, in fact, these children, who were now adults, had benefited from the tuition. The mathematical explanation was startling. Rasch published the results of this research in his seminal work “Probabilistic Models for Some Intelligence and Attainment Tests” in 1960, which since then has become a milestone in the development of test theory and methodology. Inspired by Rasch, Gerhard Fischer (1968) extended the applicability of the Rasch models to psychological measurement. Ben Wright from the University of Chicago, also inspired by Rasch’s methods instructed his students in the use of these models. Among others David Andrich, Geoffrey Masters, Graham Douglas and Mark Wilson, students of Wright, further developed Rasch’s models and helped to promote the application of this method in educational research (Wright, 1997). Proponents of the Item Response Theory (IRT) approach value the impact of Rasch’s work alongside further probabilistic approaches, that have been developed by Lord and Nowick (1968) and Birnbaum (1968). From the IRT perspective, Rasch models are regarded as special restrictive cases of IRT models, and they are often used as the “prototypes” when teaching psychometric modeling. From the Rasch perspective, the distinctive feature of the Rasch model, that the data are required to fit the model, distinguishes it from other IRT models, where the model is required to fit the data. As explained by Andrich; (to be continued)