The theories of recoil-induced resonances (RIR) [J. Guo, P. R. Berman, B. Dubetsky and G. Grynberg, Phys. Rev. A 46, 1426(1992] and the collective atomic recoil laser (CARL) [ R. Bonifacio and L. De Salvo, Nucl. Instrum. Methods A 341, 360 (1994)] are compared. Both theories can be used to derive expressions for the gain experienced by a probe field interacting with an ensemble of two-level atoms that are simultaneously driven by a pump field. It is shown that the RIR and CARL formalisms are equivalent. Differences between the RIR and CARL arise because the theories are typically applied for different ranges of the parameters appearing in the theory. The RIR limit considered in this paper is qP0/M ωq ≫ 1, while the CARL limit is qP0/M ωq < ∼ 1, where q is the magnitude of the difference of the wave vectors of the pump and probe fields, P0 is the width of the atomic momentum distribution and ωq is a recoil frequency. The probe gain for a probe-pump detuning equal to zero is analyzed in some detail, in order to understand how the gain arises in a system which, at first glance, might appear to have vanishing gain. Moreover, it is shown that the calculations, carried out in perturbation theory have a range of applicability beyond the recoil problem. Experimental possibilities for observing CARL are discussed.