Current educational practices tasked with improving students’ math performance rely heavily on declarative learning techniques, which place a burden on students’ working memory. This working memory burden can be especially harmful for students with math anxiety. Implicit learning procedures could provide an alternative and effective mathematics intervention. Prior research has identified relative symbolic magnitude (i.e., one’s ability to accurately place numbers on a number-line) as a promising area for intervention, because it is both malleable and necessary for successful understanding of abstract mathematics. In the current study, we recruited second grade students to the experimental classroom in the university’s center for educational technology. Students were administered a pre-test which tested their existing number line estimation abilities. They were then randomly assigned to either the declarative learning group or the implicit learning group and given four training sessions, the last of which was followed by a post-test. Our results suggest that students in the implicit learning group did significantly better on the post-test than those in the declarative training group, even though the groups had equivalent performance on the pre-test. Although both groups were more accurate when estimating smaller numbers on the post-test, the implicit learning group was more accurate than the declarative learning group for all segments of the post-test number-line. The results provide evidence that the implicit number-line training was more effective at helping students develop an accurate representation of symbolic magnitude. Our results also provide exploratory evidence for the role of math anxiety as a moderator.