Observations tend to select mass loss rates near the critical rate, Ṁcrit = M/L. There are two reasons for this. In some situations, such as near the tip of the AGB, the mass loss rate is very sensitive to stellar parameters. In this case, stars with Ṁ ≪ Ṁcrit have dust-free, hard-to-measure mass loss rates while stars with Ṁ ≫ Ṁcrit do not survive very long and thus make up a small fraction of any sample. Selection effects dominate the fitting of empirical formulae; observations of mass loss rates tell us more about which stars are losing mass than about how a star loses mass. In other situations, such as for some of the stars along the RGB, a steady state situation occurs where the loss of mass leads to a decrease in mass loss rate while the evolutionary changes lead to an increase; the result is a steady state with Ṁ = Ṁcrit. To determine the envelope mass and composition at the end of a phase of intensive mass loss requires stellar evolution models capable of responding on a time scale ~ tKH and thus, a new generation of stellar modeling codes.