3 Discussion

We have uncovered that the width of the peaks remains remarkably constant throughout a burst. The width should scales as $\Gamma(1-\beta\cos\theta)$. In the external shock model, such lack of temporal evolution implies that $\Gamma$ must be nearly constant. Peaks occur late in the burst because they are from regions off axis where the delay is caused by the curvature of the shell. The later peaks would be wider and delayed because off axis regions have larger $\theta$'s. A constant peak width indicates that all emitting entities must have similar $\theta$. Since the maximum angular size of the shell allowed by a differential spread of S is $S\Gamma^{-1}$, the entire size of the shell must be a few percent of $\Gamma^{-1}$. Thus, the only external shock model that is consistent with the observations is one where the overall size of the shell is much smaller than $\Gamma^{-1}$ and there is no deceleration during the classic GRB phase. This adds to our previous arguments ([Fenimore et al. 1996]; [Fenimore et al. 1998]) that a central engine (internal shocks) is the more likely explanation for the observed chaotic time history. In the context of the internal shock model, we cannot place a limit on angular extent but the lack of evolution of the peak width indicates that $\Gamma$ must be remarkably constant throught the internal shock phase.