The predictive validity and utility of assessment procedures can be increased by adding predictors to the prediction supplied by general ability tests. Of Jensen's early work comes the suggestion of focusing on the cognitive ability short-term memory (STM), especially for low-g Black children. Meta-analysis convincingly shows high predictive validities of STM tests for various criteria, but memory tests and g are not independent but substantially related. Therefore, the question is whether STM tests show incremental validity for school achievement measures over the validity supplied by g, whether incremental validity is higher on low- and medium-complexity measures than on high-complexity measures, and whether there is stronger incremental validity for low-g immigrant children than for high-g Dutch children. We compared immigrant primary school children (n=559) and Dutch children (n=604) who took the "Revisie Amsterdamse Kinder Intelligentie Test" (RAKIT), which is a cognitive ability test developed for primary school children, and regressed school achievement measures on g and STM. The most powerful single predictor of the criteria clearly is g. In general, STM scores add very little to the predictive validity supplied by g. In many cases, there is no incremental validity, the incremental validity decreases, or it increases when STM is weighed negatively. Incremental validity does not appear to be higher for low- and medium-complexity measures than for high-complexity measures. Incremental validity does not seem to differ for Dutch and immigrants. It is concluded that STM tests show very little promise in adding to the predictive validity supplied by g. © 2004 Elsevier Inc. All rights reserved.