Large HI shells, with diameter of hundreds of pc and expanding with velocities of order 10-20 kms^-1 have been detected in their hundreds in the Milky Way and are well observed features of local gas rich galaxies. These shells could well be predicted as a result of the impact of OB associations on the ISM, but doubt has been cast on this scenario by the apparent absence of OB stars close to the centres of a large fraction of these shells in recent observations of the SMC. Here we present a combination of observational and theoretical work which strongly supports the scenario in which OB associations do produce the giant HI shells. Using Fabry-Pérot scanned Hα emission line mapping of nearby galaxy discs we have detected, in all the HII regions where the observations yield sufficient angular resolution and S:N ratio, dominant shells with radii a few tens of pc, expanding at velocities of 50-100 kms^-1 , and with gas masses of 10^4-10^5 M (sun). In previous studies (Relaño et al. 2005), we found that stellar winds alone can account for the energetics of most of the shells, which form initially before the stars explode as SNe. Using the Hα luminosities of the regions to estimate their ionizing stellar populations, and standard assumptions about the surrounding ISM, we have applied, using Starburst99, dynamical modelling to project the properties of the observed shells at a few x10^7 yrs after the formation of the OB stars. The results are in good agreement with the ranges of masses (up to 10^6 M(sun)), velocities (up to ~20 kms^-1), and diameters (up to 500 pc) of representative HI shells observed in nearby galaxies. The combined effects of stellar winds, acting during the first few 10^6 yrs, and SN explosions, ¨switching on¨ subsequently, are required to yield the observed parameters. By the time the HI shells have expanded to the diameter range considered, most of the stars causing the expansion will have exploded as SNe, so that the stellar association which gave rise to the shell may not be readily detected.