I have often been frustrated with the following response to this admonition: 'Oh well, I see what you mean, and you're right in theory. There may be no necessary connection in logic, but isn't it more likely all the same that mean differences between groups would have the same causes as variation within groups.' The answer is still 'no.' Within- and between-group heredity are not tied by rising degrees of probability as heredity increases within groups and differences enlarge between them. The two phenomena are simply separate. Few arguments are more dangerous than the ones that 'feel' right but can't be justified.