The question of the appropriate definition of temperature in the kinetic theory of dense gases is discussed. An apparent contradiction in the value of the bulk viscosity between two methods of making the transition from the kinetic to the hydrodynamic stage is resolved. It is shown that the definitions of temperature through the kinetic energy and through the total energy are equivalent. The general question of the appropriate choice of macroscopic variables in nonequilibrium statistical mechanics is discussed. It is pointed out that the form of the equations of hydrodynamics is covariant with the definition of the macroscopic variables, but the molecular-distribution functions are invariant to this choice. In spite of the equivalence of temperature definition between a quantity which is a local integral of motion and one which is not, the the principle that macroscopic observables are properly chosen to be "approximate single-valued integrals of motion" can still be maintained. The question of which temperature or bulk viscosity is measured is discussed in the context of specific experimental situations.