Nearby dwarf irregular galaxies were searched for compact star clusters using data from the HST archives .
Eight of the 22 galaxies in our sample were found to host compact clusters of some type .
Three of these have populous clusters , with M _ { V } < -9.5 at a fiducial age of 10 Myr , and the same three also have super-star clusters , with M _ { V } < -10.5 at 10 Myr .
Four other dwarf galaxies , two of which contain populous and super-star clusters , are also considered using data in the literature .
The results suggest that galaxies fainter than M _ { B } = -16 or with star formation rates less than 0.003 M _ { \sun } yr ^ { -1 } kpc ^ { -2 } do not form populous or super-star clusters , and that even the brighter and more active dwarfs rarely form them .
Yet when they do form , the associated star formation activity is very high , with numerous compact clusters of similar age in the same complex and evidence for a galaxy-wide perturbation as the trigger .
This tendency to concentrate star formation in localized regions of high column density is consistent with previous suggestions that self-gravity must be strong and the pressure must be high to allow a cool phase of gas to exist in equilibrium .

Statistical considerations emphasize the peculiarity of super star clusters in dwarf galaxies , which are too small to sample the cluster mass function to that extreme .
We suggest that triggered large-scale flows and ambient gravitational instabilities in the absence of shear make the clouds that form super-star clusters in small galaxies .
This is unlike the case in spiral galaxies where density wave flows and scale-free compression from turbulence seem to dominate .
Further comparisons with spiral galaxies give insight into Larsen & Richtler ’ s relation between the star formation rate per unit area and the fraction of young stars in massive dense clusters .
We suggest that this relation is the result of a physical connection between maximum cluster mass , interstellar pressure , interstellar column density , and star formation rate , combined with a size-of-sample effect .