A nationally representative sample of students (n=3507) were followed from the beginning of the 3rd to the end of 9th grade. About half of these also agreed to participate to a follow-up at the end of 12th grade. The development of mathematical performance and attitudes towards mathematics of these students were followed. In this presentation, I will focus on the development of a subsample of poor performing students in this sample. About 5 percentage of students showed a persistent poor performance based on a standardized test and a teacher’s evaluation, mathematical learning disabilities (MLD). After 5th grade this group of poor performers was divided into two: one continued following the mainstream curriculum (MLD/C), while the other were given an individualized educational plan (MLD/IEP). There was a significant difference between these groups in the learning gains from the beginning of the 6th to 9th grade. The MLD/IEP group did not differ from the typically performing students in the amount of learning in calculation skills (no gain in geometry or statistics), while the MLD/C did not show any improvement in learning during four years of mathematics education. A similar trend was found in attitudes. The MLD/IEP group had a stronger self-concept in mathematics and showed less math-anxiety than the MLD/C group, which had a clear declining trend to negative self-image in mathematics. There were also long-lasting effects to the end of the secondary education from the educational decisions done in the primary education. The results show that decisions done in the educational system have strong effects on numerical skills and attitudes of students with MLD. The educational history of students with MLD is often the largest unknown variable in cognitive and neuroscientific studies on MLD. Models how to control this unknown this is presented.