Merging is hard to prove at redshifts z 1.5; cosmological dimming renders tidal tails nearly invisible, while bandshifting effects complicate interpretation of the observations . But circumstantial evidence implicates merging in various high-z objects.
2.1. Starburst Galaxies
The most extensive and unbiased sample of high-redshift galaxies are the ``Lyman-break'' objects at z ~ 3, which have rest-frame UV luminosities consistent with star formation rates of ~ 101 M yr-1 . The actual rates could be several times higher, since much of the UV emitted by young stars may be absorbed by dust (eg. ). Spectra show gas outflows with velocities of ~ 500 km sec-1 , atypical of quiescent galaxies but fairly normal for starburst systems. Heavily obscured high-z starbursts have been detected at sub-mm wavelengths [15, 16]. These have IR spectral energy distributions similar to ultra-luminous starburst galaxies like Arp 220 and appear to be forming stars at rates of ~ 102 M yr-1.
At low redshifts, luminous starbursts are often triggered by mergers of gas-rich galaxies . The gas in such systems is highly concentrated; H2 surface densities of 103 to 105 M pc-2 are typical of nearby starbursts , and similar surface densities are indicated in high-z starbursts . In the potential of an axisymmetric galaxy, gas becomes ``hung up'' in a disk several kpc in radius (Frenk, these proceedings) instead of flowing inward. Violently changing potentials in merging galaxies enable gas to shed its angular momentum and collapse to as little as ~ 1% of its initial radius .
But models based on mergers of low-z disk galaxies may not apply to high-redshift starbursts . First, bar instabilities in isolated galaxies can drive rapid gas inflows without external triggers . Second, disks forming at higher redshifts are more compact  and thus may already have the surface densities associated with starbursts. Third, the starbursts in Lyman-break galaxies occur on scales of several kpc (Weedman, these proceedings), whereas inflows concentrate gas into much smaller regions. Nonetheless, these objects also have irregular morphologies suggestive of mergers, and deep HDF images reveal faint asymmetric features which may be due to tidal interactions [9, 23]. Mergers seem to be the ``best bet'' for high-z starbursts, but something more than naive extrapolation from low-z is needed to test this conjecture.