A hierarchy of line-by-line, narrow-band and broadband infrared radiation models are discussed for ozone, a radiatively important atmospheric trace gas. It is shown that the narrow-band (Malkmus) model is in near-precise agreement with the line-by-line model, thus providing a means of testing narrow-band Curtis-Godson scaling, and it is found that this scaling procedure leads to errors in atmospheric fluxes of up to 10%. Moreover, this is a direct consequence of the altitude dependence of the ozone mixing ratio. Somewhat greater flux errors arise with use of the broadband model, due to both a lesser accuracy of the broadband scaling procedure and to inherent errors within the broadband model, despite the fact that this model has been tuned to the line-by-line model. ¿ American Geophysical Union 1988