On the upper AGB, stellar evolution is determined by mass loss. It is now well
established from observations that mass loss is subject to large temporal
variations, or interruptions (Willems and de Jong 1988, A&A 196, 173; Zijlstra
et al. 1992, A&A 265, L5), probably related to the luminosity changes that
occur in thermally pulsating AGB stars. As a natural consequence of such mass
loss variability, detached shells are expected to develop repeatedly. The
existence of detached shells around C-stars is not only indicated by their
spectral energy distributions but has been confirmed by spatially resolved CO
maps (Olofsson et al. 1996, A&A 311, 587) and some high-resolution IRAS images
(Waters et al. 1994, A&A 281, L1). However, comparable observations of detached
shells around O-rich AGB stars have not been obtained up to now, suggesting
that this phenomenon is less frequent or less pronounced for these objects.
If confirmed, such a situation would imply significant differences in the
mass loss behavior between C-rich and O-rich AGB stars. The present proposal is
dedicated to address this question by searching for detached dust shells
around O-rich objects and analyzing their properties in comparison to those of
C-rich stars. Guided by our hydrodynamical models which show  that the
existence of a cold detached dust shell around O-rich sources is correlated with
weak or almost non-existing silicate features (see enclosed Fig. 1), we
propose a sample of five promising candidates (3 of them known to be extended)
for observation by ISO. We supplemented our sample by 3 new C-star candidates
(also known to be extended) showing significant 60 and 100 micron (um) excess
fluxes after correction for extension.
Data analysis will make use of elaborate time-dependent radiation
hydrodynamics models of models of AGB winds which shall serve as an important
new tool for a physical interpretation of the infrared spectra and spatially
resolved images. Our selection of targets gives us the possibility to
investigate the mass loss history in both types of AGB stars (O- and C-rich) in
a comparative study and should improve our knowledge about mass loss processes
on the upper AGB and may shed some light on the problem of C-star formation.