In Fig. 1 we have also shown the positions of a sample of oxygen-rich and carbon-rich Miras. At the low temperatures characteristic of the Miras, a part of the emission at 12 μm comes from the photosphere. For a blackbody at 2000 K, the ratio of fluxes at wavelengths of 12 and 2 μm (f12/f2)∼0.18. The Miras shown in Fig. 1 have (f12/f2) ratios larger than twice the above value. It is clear that the three groups of objects populate three different regions of the diagram. Hacking et al. (1985) have already noticed that there are distinct differences between the IRAS colours of oxygen-rich and carbon-rich objects. On the basis of an analysis, using a bigger sample of bright giant stars in the IRAS catalogue, this has been interpreted by Zuckerman & Dyck (1986) as being due to a systematic difference in the dust grain emissivity index. U Mon shows the 10-μm silicate emission convincingly and, in most of the other objects for which low-resolution spectra in the near-infrared have been reported (Gehrz 1972; Olnon & Raimond 1986), the 10-μm emission may be partly attributed to silicates. Hence it is reasonable to expect that, in the envelopes around at least some of the RV Tauri stars, the dust grains are predominantly of silicates, as in the case of oxygen Miras (Rowan-Robinson & Harris 1983a). The fact that none of the RV Tauri stars is found in the region of the two-colour diagram occupied by the oxygen Miras indicates that the emissivity indices of the silicate grains in the two cases are different. Because of the higher temperatures and luminosities, the environment of grain formation will be different in RV Tauri stars.