Starbursts are certainly fascinating, but how important are they? An interesting way to frame the question is to ask: if we didn't know that major starbursts existed from direct observations, would we have difficulties explaining what we see in the universe? That is, are starbursts a necessary inference from other phenomena? The answer is an emphatic "yes," and here is a tentative and incomplete list of the essential fingerprints of starbursts based on issues raised in this conference:
Super star clusters: These very massive but compact systems, ranging from very young clusters still in dust cocoons to classic globular clusters, can evidently form only in a high pressure medium, with P / k ~ 108-9, over 104× higher than normal for disk galaxies. This requires abnormal, non-equilibrium conditions such as prevail in starbursts. Young SSC's are found to be mainly associated with interaction-induced starbursts.
Massive bulges and E galaxies: The high stellar
densities found
in the centers of nearby early type galaxies imply conversion of large
amounts of gas to stars at rates equivalent to the most extreme
starbursts known,
~ 1000
M
yr-1, if only a single event
was involved. There is good statistical evidence that many, if not
all, ETG's originate from gas-rich mergers, which are well known
to produce violent starbursts. A few large mergers rather than a
series of minor mergers are favored. (The best direct evidence for
bulge-producing starbursts at early times is probably the high
redshift sub-mm sample - e.g. from SCUBA - with
up to ~ 1000
M yr-1.)
Cosmic mass deposition in stars: For a decade, we
have been able to estimate the conversion rate of gas into long-lived
stellar mass at redshifts
z 
4. The
present-day mean mass density cannot
by itself place very good constraints on the range of
SFR
since over 13 Gyr have elapsed since the big bang (though, as noted
above, direct statistics on distant starburst progenitors can do so).
However, recent deep probes to z ~ 2, such as GDDS, reveal that a
considerable fraction of all stars then are in "dead and red"
systems comparable to local gE galaxies, with little star formation in
the preceding 1.5 Gyr. The data imply that massive galaxy assembly
begins early (zf
3-5) in dense
regions and, since there is so
little time for this to happen, massive starbursts with
300 M
yr-1 must be involved, possibly through gas-rich mergers.
The remaining items on the list involve superwinds generated by starbursts:
Chemical enrichment of the ICM and IGM: Although metal abundances at higher redshifts are generally lower than prevailing local values, much of the gas to the highest z's yet probed has been processed through stars. The primordial generations of stars responsible can be explored only theoretically now, but it is clear that the natural mechanism for dispersing new metals from the halos in which they form is a superwind.
The mass-metallicity relation in galaxies: The correlation between larger galaxy masses and higher metal abundances has been known for about 40 years. Optical, UV, and IR observations are currently providing much better information on metallicites and dust abundances in both local and high-redshift samples. Again, the natural explanation for the mass dependence is that starburst superwinds evacuate gas preferentially from lower mass systems.
Absence of super-supermassive black holes: Supermassive black holes (SMBH) are now thought to exist in all spheroidal systems, and their masses are linked to the surrounding stellar population. Their growth by gas accretion is self-limited to the Eddington rate. Nonetheless, in the absence of another inhibiting mechanism, SMBH masses would exponentiate with an e-folding time of about 100 Myr. Star formation in nuclei is often found associated with SMBH's, and it may be that starburst winds act to regulate SMBH growth in the same way they limit star formation itself.
Cosmic reionization: It is likely that massive
stars, rather than AGN's, are responsible for cosmic re-ionization at
redshifts
z 7. But
the optical depth in typical nearby galaxies
is such that only 3-10% of ionizing photons can escape. Superwinds
in protogalaxies may be necessary to clear out channels for ionizing
radiation.