Constraints on the prevalence of luminous, dusty starbursts at very high redshifts

At a conference devoted to ever deeper surveys hunting for ever more distant galaxies, I posed a question for which a concensus view has been difficult to reach: `Is there evidence for, or can we rule out, a significant population of dust-obscured starbursts at z > 3?' If, as seems likely, submm-selected galaxies are proto-ellipticals, one of the biggest unanswered questions is whether a significant fraction form at very high redshift - perhaps by the collapse of single gas clouds - or whether the entire population forms over a range of redshifts, primarily via merging. The latter scenario is favoured strongly by existing data for the majority of bright submm galaxies - mergers are common; typical spectroscopic redshifts are in the range 1 to 3. However, our reliance on radio imaging to pinpoint submm galaxies leaves open the possibility of a significant population of very distant, massive starbursts. To rule out such a scenario requires a completely unbiased redshift distribution for submm-selected galaxies and this is unlikely to be forthcoming using conventional optical/infrared spectroscopic techniques. Here, I summarise recent attempts to close that door, or pass through to an early Universe inhabited by a significant population of collosal, dust-obscured starbursts. I conclude that the door is only barely ajar; however, the idea of galaxies forming, near instantaneously, on a collosal scale is not dead: I show an SMG imaged by MERLIN+VLA for >1Ms - the deepest high-resolution radio image so far obtained. The data reveal a galaxy-wide starburst covering >=10kpc. Thus interpreting SMGs in terms of compact ULIRG-like events may not always be appropriate, as one might expect when considering the Eddington limit for >~1000M(sun)/yr starbursts.