3 Discussion

FC89 estimated the absolute position errors in the database at $\sim$0$.\!\!^{\prime\prime}$5 rms for both OH and H$_{\rm 2}$O maser reference features. The registration between the two species therefore has an uncertainty of $\sim$0$.\!\!^{\prime\prime}$7 rms. Comparison of the 23 GHz continuum peaks with 2 cm VLA images for 351.42+0.64 (Rodriguez et al. 1982) and 10.62-0.38 (Ho & Haschick 1981) indicated offsets in position of $\sim$1$^{\prime\prime}$. FC89 also noted that comparisons with previously published positions of seven OH masers and continuum sources from Garay et al. (1985) and Gaume & Mutel (1987) gave a maximum difference of $\sim$2$^{\prime\prime}$, with most <1$^{\prime\prime}$, consistent with an rms error of $\sim$0$.\!\!^{\prime\prime}$5. However, for more southerly declinations, especially south of -35$^\circ$, there may be somewhat larger errors. In particular, Caswell et al. (1995) remark that the OH FC89 position for 339.88-1.26 appears to be too far north by 4$^{\prime\prime}$. Similarly, for 347.63+0.15, Caswell (1997) suggests that the OH position as measured by FC89 may be too far north by 3$^{\prime\prime}$.

Hofner & Churchwell (1996) published positions and spectra for 21 H$_{\rm 2}$O masers associated with UCHii regions, eight of which overlap with FC89. Comparison of the maser positions for these eight sources gives a mean offset of 0$.\!\!^{\prime\prime}$85. By far the largest discrepancy (2$.\!\!^{\prime\prime}$48) was found for 34.26+0.15. Omitting this source gives a mean offset of 0$.\!\!^{\prime\prime}$64. This is the quadratic sum of the error in the two datasets, and is in agreement with an rms error $\sim$0$.\!\!^{\prime\prime}$5 for the FC89 data. We want to emphasize that in comparing the location of the masers in the FC89 database with other sources the uncertainty in absolute position must be taken into account.

Another important consideration is the size and orientation of the synthesized beam. The position of each H$_{\rm 2}$O maser spot listed in Table 3 and plotted in Fig.1 has been derived from a parabolic fit to the corresponding VLA channel map (the AIPS task MAXFIT). For the OH masers a Gaussian fit was used (the MAD equivalent of the AIPS task IMFIT). For a point source such a fit gives positions whose accuracy is determined by the signal-to-noise ratio and the size of the synthesized beam. For a signal-to-noise ratio of 10 the fits give positions accurate to $\sim$0$.\!\!^{\prime\prime}$2 rms for a beamsize of 5$^{\prime\prime}$. This means that the relative positions of maser spots of the same species brighter than 1 Jy are determined to $<0\hbox{$.\!\!^{\prime\prime}$}3$ rms for for a 5$^{\prime\prime}$ beam. For weak maser features the elongation of the synthesized beam may produce linear structure which is not real. Possible examples of this effect are the sources 341.22-0.21 and 347.63+0.15. One should therefore beware of apparent linear structure on scales <1$^{\prime\prime}$ which has the same orientation as the synthesized beam.

A related problem arises if several maser features with overlapping velocities but different positions are present. Owing to the simple fitting procedures used to derive the maser positions, beam blending can cause systematic spatial and kinematical effects which may not be real. A possible example is 351.42+0.64 (NGC 6334F). The H$_{\rm 2}$O masers in this source appear as many spots in a string $\sim$2$^{\prime\prime}$ long, offset $\sim$3$^{\prime\prime}$ north of a compact cluster of masers near the reference position (see Fig.1). The H$_{\rm 2}$O beam has a fwhm size of 2$.\!\!^{\prime\prime}$0 by 1$.\!\!^{\prime\prime}$3. Table 3 lists the spectrum and locations of the maser spots for this source, showing that several H$_{\rm 2}$O maser spots with the same velocity occur at different positions. Maser spots with the same velocity appear together in a single VLA channel map, and are therefore blended to some extent by convolution with the synthesized beam.

In 351.42+0.64 several spots in the compact group of H$_{\rm 2}$O masers near the reference position have small northern offsets. These spots have the same velocity as the strongest masers in the string, so it is possible that these small offsets are caused by beam blending. Likewise, the string itself may actually contain only 2 or 3 discrete groups with overlapping velocity ranges, which are blended in the VLA observations to appear as an almost continuous string. This ambiguity can only be resolved by new observations at higher resolution. The overall extent and general distribution of masers should not be seriously affected by beam blending. However, the effects of beam blending should be considered when interpreting apparent systematic spatial-velocity structure in these sources.