Background: Recent studies indicate that school-age children's patterns of performance on measures of verbal and visuospatial short-term memory (STM) and working memory (WM) differ across types of neurodevelopmental disorders. Because these disorders are often characterized by early language delay, administering STM and WM tests to toddlers could improve prediction of neurodevelopmental outcomes. Toddler-appropriate verbal, but not visuospatial, STM and WM tasks are available. A toddler-appropriate visuospatial STM test is introduced. Method: Tests of verbal STM, visuospatial STM, expressive vocabulary, and receptive vocabulary were administered to 92 English-speaking children aged 2-5 years. Results: Mean test scores did not differ for boys and girls. Visuospatial and verbal STM scores were not significantly correlated when age was partialed out. Age, visuospatial STM scores, and verbal STM scores accounted for unique variance in expressive (51%, 3%, and 4%, respectively) and receptive vocabulary scores (53%, 5%, and 2%, respectively) in multiple regression analyses. Conclusion: Replication studies, a fuller test battery comprising visuospatial and verbal STM and WM tests, and a general intelligence test are required before exploring the usefulness of these STM tests for predicting longitudinal outcomes. The lack of an association between the STM tests suggests that the instruments have face validity and test independent STM skills.