The integrated bolometric effective surface brightness S-e distributions of starbursts are investigated for samples observed in (1) the rest frame ultraviolet (UV), (2) the far-infrared and H alpha, and (3) 21 cm radio continuum emission. For the UV sample we exploit a tight empirical relationship between UV reddening and extinction to recover the bolometric flux. Parameterizing the S-e upper limit by the 90th percentile of the distribution, we find a mean S-e9.0=2.0x10(11)L.kpc(-2) for the three samples, with a factor of 3 difference between the samples. This is consistent with what is expected from the calibration uncertainties alone. We find little variation in S-e.90 with effective radii for R-e similar to 0.1 - 10 kpc, and little evolution out to redshifts approximate to 3. The lack of a strong dependence of S-e.90 on wavelength, and its consistency with the pressure measured in strong galactic winds, argue that it corresponds to a global star formation intensity limit ((Sigma) over dot(e.90)similar to 45 - M. kpc(-2)yr(-1)) rather than being an opacity effect. There are several important implications of these results: (1) There is a robust physical mechanism limiting starburst intensity. We note that starbursts have S-e consistent with the expectations of gravitational instability models applied to the solid body rotation portion of galaxies. (2) Elliptical galaxies and spiral bulges can plausibly be built with maximum intensity bursts, while normal spiral disks cannot, (3) The UV extinction of high-z galaxies is significant, implying that star formation in the early universe is moderately obscured. After correcting for extinction, the observed metal production rate at z similar to 3 agrees well with independent estimates made for the epoch of elliptical galaxy formation. (C) 1997 American Astronomical Society.