A test with many reading-heavy word problems predicts performance in a reading-intensive program, but the publisher markets it as a culture-free measure of general reasoning. What is the best conclusion?