We have seen that our very reliable observing technique implies rates of dMe radio bursts per hour which are slightly different from what can be inferred from the "reliable" former papers. However, these rates may be under- or over-estimated for various reasons. We have compared the stars with each other, but their rate of activity presumably depends on their instrinsic properties.
In reality, the peak flux density might not be the right indicator of rate of activity. In the visible
domain, the activity rate is defined by Doyle & Mathioudakis (1990) as the total energy released during
a flare, divided by the total observing time. In order to calculate an equivalent quantity at radio
wavelengths - although it might not be the best indicator - we need to know the total bandwidth of the
emission, in addition to the distance of the source and duration of the burst. It is therefore not possible to
estimate the total released energy from the papers. The energy of the bursts detected in 1989-1993 is of
the order of 
, but it could be higher if the burst's bandwidth is much larger than the observing
bandwidth.
Since dMe's flares are often compared to the Solar White Light Flares (WLF), it would also be interesting to compare the activity of dMe's in radio with the activity in the visible and X-rays. Indeed, the Solar WLF of type I are found to be correlated with X-rays and microwaves, whereas it is not the case for type II WLF. Although the last paragraph of Sect. 4 suggests that more optical flares than radio bursts are observed from flare stars, we point out that only numerous simultaneous observations can bring reliable statistical correlations between various wavelength domains: these are necessary to provide information on the triggering of flares and on the acceleration process(es). The correlation with other wavelengths and the rate of occurrence of each type of stellar radio burst is also necessary to understand the whole phenomenon of flares in dMe's.
Establishing reliable rates of activity for stars requires a large amount of observing hours. A long term study of dMe's radio bursts is necessary to establish the various types of radio bursts and the occurrence of each type. Collecting such data will be possible when the Arecibo and Nan‡ay single-dish radiotelescopes are renovated. We have also extended these studies to decametric wavelengths, at which an array is easily accessible to us in Nan‡ay (France): these observations are performed simultaneously with a complementary decametric array in Kharkov (Ukraine) and with an optical telescope in Nauchny (Crimea).
Acknowledgements
We are very grateful to N. Meyer-Vernet for encouraging this study. We thank H.J. Wendker for providing useful references, J. Aboudarham, T.S. Bastian, J.A. Bookbinder, S.M. White, and J. Lim for fruitful discussions, and the referee for suggesting improvements for the paper.