when the effect of one independent variable on subjects' scores seems to depend upon the levels of the other independent variable, we are likely to have something called a(n)