Much of our understanding of modern astrophysics rest on the notion that the Initial Mass Function (IMF) is universal. Our observations of a sample of HI-selected galaxies in the light of H-alpha and the far-ultraviolet (FUV) challenge this result. The flux ratio H-alpha/FUV from these star formation tracers shows strong correlations with surface-brightness in H-alpha and the R band: Low Surface Brightness galaxies have lower H-alpha/FUV ratios compared to High Surface Brightness galaxies as well as compared to expectations from equilibrium models of constant star formation rate using commonly favored IMF parameters. I argue against recent claims in the literature that attribute these results to errors in the dust corrections, the micro-history of star formation, sample issues or escaping ionizing photons. Instead, the most plausible explanation for the correlations is the systematic variations of the upper mass limit and/or the slope of the IMF. I present a plausible physical scenario for producing the IMF variations, and suggest future research directions.