Listeners discriminated between pairs of complex sounds, each consisting of two groups of components. Two harmonic complexes were played out through separate channels, and each filtered to obtain a ‘‘lower’’ and a ‘‘higher’’ group. The ‘‘carrier fundamental frequencies (F0s)’’ of both groups were usually 125 Hz; only those components in the lower group were resolvable by the peripheral auditory system. For the standard stimulus, the F0s of the two groups were frequency modulated coherently with each other, so that they were always equal. For the signal, the F0s of the two groups were modulated incoherently (π modulator delay), so that they differed by an amount that varied sinusoidally between values proportional to the depth of FM (the dependent variable). Stimuli were usually presented in continuous pink noise. The results showed that (i) when the components were added in sine or cosine phase, the mean threshold across listeners corresponded to a zero-peak modulation depth of 6%–7% (rms mistuning=8.5%–10%); (ii) performance dropped to chance when the upper components were added in alternating sine–cosine phase, but was only moderately affected by the phase of the lower components; (iii) threshold for sine-phase stimuli improved by a factor of 1.6 when noise in the frequency region of the two component groups was removed; (iv) threshold increased moderately with increases in the frequency separation between the two component groups; (v) threshold dropped markedly when the F0s of both groups of components were increased so as to be resolvable by the peripheral auditory system; and (vi) performance dropped to chance when the nominal carrier F0s of the two groups of components differed from each other. It is concluded that listeners can perform simultaneous comparisons of F0s derived from resolved and unresolved harmonics, and that their performance on this task is fairly robust. Implications for the perceptual segregation of concurrent complex sounds, and for models of pitch perception, are discussed.