Re-examining the content validation of a grammar test: The (im)possibility of distinguishing vocabulary and structural knowledge