How do we match up the desired sampling rate with the pre-filtering amounts?

I guess to answer the above question conceptually: the more sampling we are doing in texture space, the blurrier layer we want to choose form the mipmap?