2001A&A...370..351P

I show how an empirical variability - luminosity relationship for prompt gamma-ray bursts, first proposed by Fenimore and Ramirez-Ruiz, can be understood as a special-relativistic beaming effect in the ``cannonball model'' of Dar and De Rujula. In this scenario the variability is a measure of the direction of propagation and the Lorentz factor of the cannonball on which in turn the apparent luminosity of the prompt GRB depends sensitively. The observed absence of cosmological time dilation in the ``aligned peak test'' - when using redshifts derived with this relation - is also explained. The most direct evidence in favour of the cannonball model presented here is its correct description for the observed relation between narrow-spike width and amplitude within a given GRB. There seems to be an indication for cosmological time dilation in the total duration of GRBs, as expected in the cannonball model. Quantitative predictions for the luminosity function of GRBs and the ``spectral-lag luminosity relation'' are given.