•The present study was designed to investigate the best modeling procedure to teach the cardinality principle (CP).•Three intervention conditions involved: (a) label and then count (label-first), (b) count, label, and emphasize the last word (count-first), and (c) counting only.•How many—a direct measure of the CP was used as the main task to investigate CP understanding, and give-me-n—an indirect measure of the CP was used as transfer task to investigate meaningful application of the CP.•The count-first strategy provides a more efficacious way to model the CP, because it entails a closer temporal connection between the last number word of a count and the stated cardinal value of a collection.•Modeling the CP should capitalize on children’s subitizing ability to reinforce the purpose of counting for determining the number of objects in a collection and to promote understanding of the CP. The cardinality principle (CP), which specifies that the last number word used in the counting process indicates the total number of items in a collection, is a critically important aspect of numeracy. Only one published study has focused on how best to teach the CP, and its results are uncertain (Mix, Sandhofer, Moore, & Russell, 2012). The present study was designed to investigate several modeling procedure to teach the CP. Forty-nine 2–5-year olds were randomly assigned to one of the three interventions: (a) label and then count (label-first), (b) count with an emphasis on the last word and label (count-first), and (c) counting only. At a delayed posttest, the count-first intervention was substantively more efficacious than the other interventions at promoting success on the CP task and a transfer task (as measured by effect size). The results underscore the need for early childhood educators and parents to reinforce the purpose of counting by building on children’s subitizing ability and explicitly labeling the total number of items after a collection is counted.