The evolution of stars between the AGB and planetary nebula phases was investigated by sensitive radio continuum observations of a sample of 21 evolved stars with high mass loss rates and extended circumstellar envelopes, in a search for newly formed compa

8 After calibration of the u-v data, maps were made with natural weighting in the u-v plane and a cell size of 0:500 over a 51200 51200 eld encompassing the full-width half-maximum (FWHM) primary beamwidth (5:40 ). The resulting synthesized half-power beamwidth had a typical value of 300 . The r.m.s. thermal noise level of the total (right plus left) intensity near the eld center for each observed star is given in the last column of Table 1. Unresolved (< 300 ) emission was found close enough to the positions of several objects (CRL 618, CRL 915, VY CMa, R Leo, IRC+10216, CIT 6, IRAS 174231755 and IRAS 21282+5050) that we can be reasonably sure of its association with the star. The ux densities (corrected for primary beam attenuation) and positions of these sources are listed in Table 2, along with the o set between the position of the radio source and the stellar position in Table 1. The accuracy of the radio positions is 0:500 to 300, depending on the strength of the source. Other point sources which are unlikely to be associated with the observed star were found in several maps. The ux densities (corrected for attenuation by the primary beam of the array) and the positions are given in Table 3. The radio emission from CRL 618 (Table 2) is strong enough that the dynamic range of the map is improved by using the self-calibration procedure. The resulting maps were searched for extended emission; none was found. The data for the other stars were also searched for extended emission. This was done by applying a range of Gaussian weighting functions (or\tapers") before Fourier-transforming the u-v data to make maps with a range of spatial resolutions. Table 4 lists the results for three values of the full-width half maximum synthesized beam: 300, 1500, and 3000 . The errors for the detected sources are the r.m.s. noise values; upper limits