The purpose of this study was to investigate the modeling and generalizing processes of 3rd through 5th grade students while they are solving a non-routine additive word problem. The study utilized an experimental design with no control group. A total of 278 students from 7 different schools participated in the study. Students were asked a non routine word problem as a pretest. Then, they were given a semi structured worksheet based on modeling similar problems with small numbers. Finally, they were asked a non-routine problem similar to the one asked as the pretest. Results showed that the students had very low level of success on this type of problem. After the intervention, only the fifth graders made some improvement. Implications for mathematics teaching and learning are discussed.