A scatter plot shows study hours (x-axis) versus final exam scores (y-axis) for 30 students, color-coded by major. Biology students (12 points) show correlation $r ≈ 0.85$, and Chemistry students (18 points) show $r ≈ 0.88$. However, when a single regression line is fit to all 30 points, the overall correlation is only $r ≈ 0.40$ (weak). Which best explains this paradox?