We studied the radio properties of very young massive regions of star formation in HII galaxies , with the aim of detecting episodes of recent star formation in an early phase of evolution where the first supernovae start to appear .
Our sample consists of 31 HII galaxies , characterized by strong Hydrogen emission lines , for which low resolution VLA 3.5 cm and 6 cm observations were obtained .
We complemented these observations with archival data at 20 cm .
We found that the radio spectral energy distribution ( SED ) has a range of behaviours ; 1 ) there are galaxies where the SED is characterized by a synchrotron-type slope , 2 ) galaxies with a thermal slope , and , 3 ) galaxies with possible free-free absorption at long wavelengths .
The latter SEDs were found in a few galaxies ( e.g .
UM 533 , Tololo 1223-388 ) and represent a signature of heavily embedded massive star clusters closely related to the early stages of massive star formation .

Based on the comparison of the star formation rates ( SFR ) determined from the recombination lines and those determined from the radio emission we find that SFR ( H \alpha ) is on average five times higher than SFR ( 1.4 GHz ) .
We confirm this tendency by comparing the ratio between the observed flux at 20 cm and the expected one , calculated based on the H \alpha star formation rates , both for the galaxies in our sample and for normal ones .
This analysis shows that this ratio is a factor of 2 smaller in our galaxies than in normal ones , indicating that they fall below the FIR/radio correlation .
This result is further confirmed by the detection of high q-parameter values ( the ratio of infrared to radio fluxes ) in a few sources .
These results suggest that the emission of these galaxies is dominated by a recent and massive star formation event in which the first supernovae ( SN ) just started to explode .
This indicates that the radio emission is most likely dominated by free-free continuum , and that the emission at low frequencies may be optically thick , in agreement with the observed SEDs .

We combined the VLA data with age indicators based on optical observations ( e.g .
equivalent width of H \beta ) together with the ratio between the far infrared and the radio continuum fluxes and proposed an evolutionary scenario to explain the observed trends .
We conclude that the systematic lack of synchrotron emission in those systems with the largest equivalent width of H \beta can only be explained if those are young starbursts of less than 3.5Myr of age , i.e .
before the first type II SNe start to explode .