The present study is a first attempt to find an effect of transfer between multiplications of diverging difficulty. 58 students at the age of 18 to 35 were tested in a pretest training posttest paradigm. The pretest consisted of 90 multiplications, of which 30 were easy (3x4), 30 were of moderate difficulty (13x4) and the last 30 were difficult (23x4). The testing was performed on two consecutive days. On the first day, participants were presented with a sociodemographic questionnaire, three subtests taken from the IST- 2000R, the pretest, a questionnaire about applied strategies and the first training session. During the pretest participants solved multiplications in random order in two sets. The pretest consisted of 180 multiplications from all difficulties. During each training session, participants were confronted with 15 multiplications (easy, moderate or difficult) that they had to train in 40 sets. The following day, training was repeated and the posttest was presented. The posttest consisted of the same items as the pretest. Analyses of Variance showed an effect of training in the moderate and the difficult condition. An effect of transfer had not been detected. Additionally strategies of solving and intelligence had been measured. Participants showed clear preferences in strategies depending on difficulty of multiplications.With respect to intelligence results show, that level of numerical intelligence, measured by IST-2000R has an impact on performance. Persons with a higher level of numerical intelligence showed a better effect of training than persons with a lower level of numerical intelligence.An impact of numerical intelligence on the effect of transfer had not been detected.The results of the present study are discussed with reference to the current literature.