Several studies report that the durations of faint bursts and the widths of their peaks are factor of two longer than for bright bursts (Norris et al. 1994; Davis et al. 1994; Fenimore et al. 1995b). However, Mitrofanov et al. (1993, 1994) find no difference in the widths of the peaks of faint and bright bursts.
The reported time stretching of a factor of two has been interpreted as due to cosmological time dilation. However, the difference is not uniquely a signature of cosmological time dilation, and therefore of a cosmological origin for the bursts. The difference can be intrinsic to the bursts, be due to correlations among other burst properties, or arise from the inhomogeneous spatial distribution of sources.
Moreover, when the spectral energy dependence of the peak width is
carefully taken into account, the reported time stretching of a factor
of two requires that the burst sources lie at a redshift 
,
if one attributes the difference to cosmological time dilation (see
Figure 20) (Fenimore and Bloom 1995). The bursts would then originate
before the first generation of stars or galaxies; in other words, they
would be "orphans." Furthermore, consistency with the observed
brightness distribution of bursts and an origin at
requires strong source evolution; then the -3/2 slope of the brightness
distribution of bright bursts is coincidental. It is then no longer
evidence that the spatial distribution of nearby sources is uniform,
and that nearby space is Euclidean.
Alternatively, the sources could lie at redshifts 
, but
then at least 70% of the reported difference in the widths of the
peaks of faint and bright bursts must be intrinsic to the burst. This
raises the question: if 70%, then why not all? This is a
particularly relevant question, given we know that more intense spikes
are narrower than less intense spikes within individual bursts.
Consequently, the reported time stretching is neither evidence for nor against the cosmological or Galactic hypotheses.