Recent radio surveys have identified numerous objects which are morphologically similar to FR-II radio galaxies but are appreciably smaller. Those sources that are less than 500pc in extent are termed Compact Symmetric Objects (CSOs; Wilkinson et al. 1994). This class also contains many Gigahertz Peaked Sources (GPSs), sources whose radio spectrum is seen to peak at GHz frequencies (O'Dea, Baum & Stanghellini 1991). The spectral form of GPSs is thought to be due to either free-free absorption by an inhomogeneous foreground screen, or synchrotron self-absorption in the source itself. Slightly larger sources, those in the range 0.5-15kpc, have been termed Medium Symmetric Objects (MSOs) by Fanti et al. (1995). These various classes of small sources are found to constitute 10-30 per cent of all sources in a flux limited sample.
It is tempting to consider an evolutionary picture in which CSOs evolve into full size FR-II radio galaxies, passing through the MSO stage. Since we would expect the sources to remain small for a relatively short period of time, there must be strong luminosity evolution for us to see so many small sources (Begelman 1996; Readhead et al. 1996). Begelman (1996; hereafter B96) showed how this luminosity evolution could be understood in terms of a declining source pressure as the source evolves.
O'Dea & Baum (1997; hereafter OB97) have recently studied a combined
sample of objects including CSOs, MSOs and classical FR-II radio galaxies.
In particular, they examine the distribution of (projected) linear sizes.
Using the B96 evolution model, and assuming physically realistic
interstellar medium (ISM) density profiles, they find a clear overabundance
of CSOs and MSOs as compared with the classical FR-II radio galaxies. This
is seen as a `plateau' in the size distribution of their sample of sources
between 
and 
. They suggest that either (1) a
large fraction of the small sources are transient or frustrated and never
evolve into large sources, or (2) the luminosity evolution is much stronger
than that predicted by B96, possibly due to a decline in the efficiency of
conversion of jet kinetic energy into radio power.
In this letter we suggest that radio sources are intermittent, and that the source statistics examined by OB97 can be understood in the context of a simple evolutionary picture if this intermittency is taken into account. In Section 2, we develop a simple model of radio source evolution including intermittency. Theoretical source statistics are calculated in Section 3. Section 4 discusses predictions of this scenario. Our conclusions are summarized in Section 5.