It was explored in two studies how primary school teachers employ the subtle structural properties of language to communicate distinct evaluations of students who attain higher versus lower marks, boys versus girls, and students with immigrant versus nonimmigrant origins. Written judgments of final evaluation records were coded for language abstraction. Outcomes displayed that students with higher marks are described with more abstract positive and more concrete negative terms than those receiving lower marks. By varying language abstraction teachers also communicate more favorable evaluations of girls than boys and of students with nonimmigrant than immigrant origins. These outcomes revealed a linguistic evaluation bias that implicitly increases girls’ qualities and performance, but impede improvements of boys and students of immigrant families by centering on their stable negative characteristics.