The study determined the arithmetic verbal problem solving skills of primary school students and to examine them according to grade level and mathematics achievement variables. The study was designed according to the relational survey model, one of the quantitative research methods. The population of the research consists of students studying in public primary schools in the central districts of a large city in the south of Turkey. The sample consisted of 1,865 students determined according to the disproportionate stratified sampling method. The "arithmetic verbal problem test for students" prepared by the researchers was used as a data collection tool. As a result of the research, it was revealed that the students were most successful in the problems of combining and separating with "consequence unknown", then "with change unknown" and "initial unknown", but they were less successful in part-whole type problems. Thus, it was concluded that as the mathematics achievement and grade levels of the students increased, their success in solving verbal problems also increased.