This section discusses all available evidence to give a picture of galaxy evolution at high redshift consistent with luminosity-dependent luminosity evolution. This scenario neatly reconciles the various amounts of luminosity evolution seen in the surface brightness and internal kinematics studies.
Starting with internal kinematics, there are three studies which appear to be at odds with one another: Koo et al. koo95, Vogt et al. vogt96 and the CFHT sample presented in this thesis. At the low end of the galaxy mass spectrum, Koo et al. koo95 looked at compact, unresolved galaxies with narrow emission line widths. They found that the line widths of these galaxies were 2
3
lower than expected from the luminosity-linewidth relation of normal spiral galaxies. Those low-mass galaxies were identified with local HII galaxies on the basis that they fell on the luminosity-linewidth relation of local HII galaxies given by Telles and Terlevich telles93. At a fixed velocity width, the Koo et al. galaxies were 4 magnitudes brighter than expected from the local Tully-Fisher relation. At the high end of the galaxy mass spectrum, there is the Keck study of Vogt et al. vogt96. The galaxies in this study were intrinsically large (r
3.0 kpc) and bright (M
20.7). They were also intrinsically massive with typical rotation velocities of 200 km/s. Their rotation curves were similar to those of local normal galaxies, and the increase in luminosity with respect to the local Tully-Fisher was less than 0.6 magnitude in the B band.
The CFHT sample occupies a niche in size and mass right in between the above two Keck samples. The CFHT galaxies were typically a full magnitude fainter than the objects in the Vogt et al. study. They were also intrinsically smaller with typical disk scale lengths less than 2.0 kpc, and some of them were barely resolved. They were also less massive with rotation velocities 
100 km/s, but they were nonetheless more massive than the galaxies observed by Koo et al. The CFHT galaxies exhibited a diversity of internal kinematics not seen in the Keck study but consistent with local peculiar phenomena such as galactic supershells. The kinematically normal galaxies showed an increase in luminosity of 1.5 mag with respect with the local Tully-Fisher relation defined by the Mathewson et al. sample. If mass is taken as an indicator of the luminosity all the galaxies would have had in a quiescent phase, then all three internal kinematics studies can be understood with luminosity-dependent luminosity evolution. Low-mass galaxies are more susceptible to processes such as interactions/mergers which trigger star formation because they have a higher gas fraction. Supernova-driven winds and nuclear outflows will also be more damaging to the normal (i.e. rotational) kinematics of low-mass galaxies, so their internal kinematics should be more varied.
HII galaxies at high redshifts remain an unknown. There is no doubt that some galaxies in the Koo et al. sample are indeed HII galaxies with linewidths dominated by low velocity, ionized gas from a small number of star forming complexes; they lie too neatly on the luminosity-linewidth relation for HII galaxies. On the other hand, it is not inconceivable that some of the galaxies are low-mass, disk galaxies which may have been brightened up by 4.0 mag or less. These galaxies may have been lumped with HII galaxies due to the lack of spatial resolution. As far as the CFHT galaxies are concerned, their linewidths are clearly not dominated by a few star forming regions. First, [OII] follows the continuum light profile, and [OII] is symmetrically distributed about the galaxian centers. Smooth exponential [OII] distributions are unlikely to be produced by a small number of star forming complexes. Second, although some of our galaxies have a spatially-resolved spectrum with a rotation velocity consistent with zero, others clearly show a systematic rotation. Evidence for a giant HII region has been detected in one of the CFHT galaxies, but the rotational component could nonetheless be isolated and analyzed. An important question must be answered in order to accept that HII galaxies play as significant a role as Koo et al. claim over the redshift range 0.10.7 they observed. The timescale for a star burst is typically shorter than 1 Gyr. The redshift range 0.10.7 corresponds to a time interval of 3.5 Gyrs (H
=75 and q = 0.5). If HII galaxies are present in great numbers at all redshifts, then they must be forming at all redshifts, so that the newborn galaxies may replace the fading ones to keep them in sufficient numbers over many Gyrs.
The B-band surface brightness 
(B) of field disk galaxies undergoes a strong evolution over the redshift range 0.1 < z < 0.6 compared to a local z=0.06 relation and the Freeman law [\protect\astronciteSchade et al.1996a]. At redshifts of 0.43 and 0.55, 
(B) is equal to 1.22 and 0.97 respectively. This is consistent with or slightly less than the evolution seen in the CFHT sample, and it is certainly more than the amount of evolution seen in the Keck sample of Vogt et al.. Looking at Figure 1 of Schade et al. schade96a, there is a hint that surface brightness evolution depends on the disk scale length. Smaller galaxies evolve more drastically than large galaxies. The effect is particularly noticeable in the highest redshift bin where the log 
M relation at that redshift clearly curves ``down''. Taking M 
21 and <r> = 4.3 kpc (H = 75), one can see on the Schade diagram for 0.45 < z < 0.65 that a large number of galaxies at (log scale length = 0.8, 21) show little or no evolution as observed in the Keck sample.
Of course, the best way to reconcile the evolution seen in surface brightness and internal kinematics will be to study both in the same galaxies. For now, luminosity-dependent luminosity evolution is an elegant way to explain the available data.