Why is empiricism so hard to do consistently well?