In a previous paper (Greenwood, 1972a), the measurement of masking by bands of noise of external origin led to determinations of signal/masker ratio that are applied in the present paper to the task of estimating combination band levels. Estimates of the levels of combination bands, given by the expression(n + 1)fl − nfh, were made by using the maximum masked threshold, read from a masking pattern produced by a given band, to calculate the denominator of the previously obtained signal‐to‐masker ratio. The shapes of the masking patterns and the estimates of level lead to the following conclusions: (a) Comparison of the masking effects of combination bands and bands of external origin further demonstrates that to assume that combination components, once generated, arenotequivalent to external stimuli in ultimate loci and subsequent effects in the cochlea would require complexad hocexplanations to account for the indistinguishable masking effects produced by the two kinds of bands. (b) Combination bands generated by mixtures of line and continuous spectra vary in level as functions of stimulus parameters in ways very similar to the variation in the level of combination tones produced by tonal primaries, as measured by cancellation techniques (Zwicker, 1955, 1968; Goldstein, 1967). (c) The present data extend the previously reported empirical functional relations, in that the level of the band2fl>−fhis (i) a less sensitive function of primary frequency separation at higher primary levels and (ii) a more nearly constant percentage of primary level at wider primary separations. (d) The generation of combination components in essentially the same way, whether they are produced by line, continuous, or mixed spectra, should advance the task of calculating empirically the amplitude characteristics of the combination spectra produced by complex primary spectra.