2006AJ....132..489L

OH/IR stars fall on a well-defined locus in two-color near-IR (NIR) plots. This locus is replicated by a sequence of self-consistent radiative-transfer models of shells with long-term constant mass loss provided that (1) the initial stellar spectral energy distribution has NIR colors matching those at the foot of the locus and (2) a cold-silicate dust opacity function is used. The models depend on the public code DUSTY. The NIR locus is also followed by models of detached shells: our models are based on shells generated by a constant dM/dt within which a central hole grows, so an expansion-time chronology can be attached to the color evolution of a detached shell. This also provides an upper limit on the time for 1612 MHz masers to disappear after dM/dt⟶0. The brevity of this timescale shows that the pumping of 1612 MHz masers is very dependant on the reprocessing of the stellar radiation field occurring within 5r_d of the star (where r_d_ is the radius of the hottest dust), which explains why they can disappear on timescales of less than two decades. Our second concern is to explain the distribution along the NIR locus of the OH/IR stars with the reddest IRAS colors, which appear to have ``detached'' shells, even though many are mid-IR (MIR) variables. Since the NIR locus is followed by shells generated by any form of dM/dt, whether constant, interrupted, or periodic, these MIR variables show that most thick, oxygen-rich, circumstellar shells exhibit a severe cyclical modulation in their mass-loss rates, as predicted by Simis in 2001.