Classical deterministic (or ''mechanical'') developments of traffic st
ream models from car-following models are criticized on the following
basis. The evaluation of the constant of integration for initial condi
tions corresponding to speed zero and headway equal to the reciprocal
of jam density ignores large numbers of possibly relevant solutions of
the underlying car-following model. Some undesirable consequences, pa
rticularly the restriction to temporally and spatially constant flow c
onditions, of viewing traffic stream models as being deterministic con
sequences of car-following models are also noted. It is suggested that
an approach to traffic stream models via kinetic models possibly woul
d be more appropriate and useful.