In order to put constraints on the possible mechanism(s) for the mass loss it is important to observationally determine the mass loss characteristics of stars with different pulsational behaviour. A major survey of circumstellar CO radio line emission of irregular and semiregular variables have therefore been performed. The result is a significant increase in the number of detected objects, and essentially all objects are observed in more than one transition. Work on detailed modelling of the data has started. Some objects exhibit very peculiar line profiles, and the Owens Valley mm-wave interferometer has been used to study in detail the CO line emission from one object, RV Boo. As an alternative to probing the circumstellar properties by CO radio line emission, methods are developed to use resonance line scattered light from NaI and KI, and light scattered in the vib-rotation lines of CO. The first positive results have already been obtained.
Based on a survey of circumstellar CO radio line emission towards a sample of optically bright carbon stars it has become evident that some of these stars must have experienced a recent period of highly variable mass loss. The IRAM PdB mm-wave interferometer has been used to study in great detail the CO emission towards two of these stars, TT Cyg and U Cam. The TT Cyg observations reveal a remarkably thin (width/radius<0.05), very close to spherically symmetric, shell consisting of CO radio line emitting clumps that is centred on the star. The cause of the shell is not clear, but a mass ejection about 7000 year ago due to a He-shell flash is an attractive explanation. Also here alternative methods for probing the circumstellar medium are developed, and the first results in terms of images of these detached shells in dust scattered light have been obtained.
Molecules, some of them quite complex by astronomical standards, are present in the circumstellar envelopes as a result of chemical reactions in the stellar atmosphere as well as in the envelope itself. These molecules may be used also as probes of the stars, e.g., their elemental composition. This is not an easy task, but it has been shown that even simple observational quantities like line intensity ratios may be used to discriminate between stars of different chemical compositions, e.g., whether C/O<1 or >1. Of particular interest in this connection are the S-stars whose C/O-ratio balances very close to unity. Some initial results on these objects have been obtained. The results of a major survey of circumstellar molecular line emission from a sample of optically bright carbon stars will shed further light in this area. More than 300 lines from five different molecular species have been detected and the survey is observationally completed.
A model for interpreting in more detail the circumstellar molecular line
emission has been developed. It is based on the versatile Monte Carlo
method, and it includes the energy balance equation for the gas. The
effects of over-lapping lines are also included. Using this a successful
modelling of CO far-infrared (from ISO LWS) and radio lines from the
extreme carbon star IRAS15194-5115 have been done. The star is peculiar
because of its very low 
C/ 
C-ratio ( 
5) despite it being
highly evolved on the AGB. The model is now being applied to HCN and CN
interferometer data on five carbon stars, and CO and CO data
on a number of carbon stars (for which there exists photospheric
C/ C-ratio estimates that have been intensively debated
recently in the literature).
Since the launch of the ISO, there have been many discoveries in the field of late-type stars. Spectroscopy of these stars reveals many absorption bands which are inaccesible from the ground. Fitting the molecular absorption bands leads one to estimate the excitation temperatures where they arise and hence gives us clues as to where in the circumstellar envelope the absorption occurs.
One of the main discoveries is the pump line (in absorption at 34.6 
m)
for the OH maser seen in the radio at 1612MHz. This, along with OH lines
in the far-IR seen by other groups, confirms the radiative pumping theory.
The detection was done towards a supergiant with a very high mass loss
rate. For supergiants with low mass loss rates emission lines due to
atomic fine structure lines of [Fe II] and [Si II], which are thought to
radiatively cool the gas in the circumstellar envelope, have been detected.
Dust grains condensed in oxygen-rich (C/O<1) outflows show signatures at
10 and 18 m which are due to silicates. Generally, these features are
broad and are thought to be associated with an amorphous (disordered)
structure of the dust. However, emission bands in the far-IR due to
crystalline silicates have been detected in a few high mass loss rate
stars. The formation process of crystalline grains is unclear.
A subgroup of low mass loss rate AGB-stars are known to exhibit a weak dust
emission feature at 13 m, as well as weak silicate emission at
10 m. These stars are found to exhibit emission bands between
13-16 m which are due to gaseous CO 
. The association between
dust and and gas emission is thought to be due to shocks in the stellar
atmosphere. As the mass loss rate gets higher, the signatures of both are
either swamped by stronger 10 m silicate dust emission, or the carrier
of the 13 m dust feature and the CO are destroyed.
Some work on early post-AGB objects has also been done. Strong maser emission in rotational lines in vibrationally excited states of SiO have proven to be an important probe of the conditions in the dynamical atmospheres of AGB-stars. In a survey of such emission from objects believed to be at the tip of the AGB and beyond (i.e., they are highly obscured), it was clearly shown that SiO masers are only associated with pulsating stars. CO observations of the peculiar object HD101584 revealed a circumstellar molecular envelope with properties very similar to those of wellknown early post-AGB objects. A substantial fraction of the molecular gas has been accelerated to very high velocities (>100km/s), by a mechanism that must be different from radiation pressure.