The concurrent validity of standardized achievement tests (the Stanford 9 and the Iowa Tests of Basic Skills) was examined using data from different school districts nationwide and a latent variable modeling approach. Items in the standardized achievement tests in several content areas were divided into parcels. Parcel scores were used to create latent variables. Students' grade point average, teachers' ratings, and other achievement scores were also used to create external-criterion latent variables. The standardized achievement latent variable was correlated with the external-criterion latent variables. The results suggest that: (1) there is a strong correlation between the standardized achievement and external-criterion latent variables; (2) this relationship is much stronger when latent variables rather than measured variables are used; and (3) the correlation between standardized achievement and external criterion latent variables is significantly larger for the population of students not of limited English proficiency (LEP) than for the LEP population. It is speculated that the low correlation between the two latent variables in the case of the LEP group is due to the impact of language factors. That is, language factors act as construct irrelevant sources. (SLD)