Linking assessment and curriculum in science education, particularly within the topic of matter and its changes, is often taken for granted. Some of the fundamental elements of the assessment, such as the choice of wording and visual representations, as well as its relation to the curricular sequence, remain understudied. In addition, very few assessment studies compare the reliability of assessments across different communities of students. In this study, we cross-examine the intersection between representation formats and educational contexts, in the analysis of students' performance on an assessment of the topic of matter and its changes. We developed two assessment forms with equivalent pictorial and verbal items to probe the reasoning of student groups in South Africa and Israel. Despite the vast differences between them, we found that in both groups, the pictorial versions of the items were generally easier than the verbal ones, except for items that included representations of motion using arrows. In addition, we found that the fitting of the pictorial items to the RASCH model was better than the verbal items in the South African group, indicating a lower reliability that is most likely related to language difficulties. We conclude that pictorial representations provide a better account of reasoning across cultural and curricular contexts, but when pictures include ambiguous symbols, especially when representing motion in a static picture -- verbal descriptions may be more accurate for revealing the actual reasoning of students.