Furukawa and Moore [S. Furukawa and B. C. J. Moore, J. Acoust. Soc. Am.100, 2299–2312 (1996)] found that the detection of frequency modulation (FM) imposed on two inharmonically related carriers was better when the FM was coherent across carriers than when it was incoherent. Here, “coherence” refers to whether the pattern of frequency change over time was identical or different across carriers. The present paper was designed to explore three possible mechanisms underlying this effect. Thresholds were measured for the detection of a single cycle of sinusoidal FM imposed on two sinusoidal carriers. The FM of each carrier was equally detectable, as determined in preliminary experiments. A continuous pink-noise background was used to mask the outputs of auditory filters tuned between the two carrier frequencies. The modulation rate was either 2.5, 5, or 10 Hz. Three combinations of carrier frequencies were used, varying in the extent to which the carriers were harmonically related (1050 and 2069 Hz; 1100 and 2000 Hz; and 1100 and 1925 Hz). The carrier frequencies were either fixed at these values, or were randomly varied (roved) from one trial to another (±10%) keeping the frequency ratio constant. Performance for coherent FM was generally better than for incoherent FM. The effect of FM coherence was greater at the lowest modulation rate and was slightly greater when the carrier frequencies were fixed throughout a block of trials than when they were roved. For the two lowest modulation rates, the effect of FM coherence was greater for carriers that were (nearly) harmonically related. It is proposed that sensitivity to FM coherence depends partly on comparing patterns of phase locking to the carriers; this is done most effectively at low modulation rates. However, two other factors may play a small role. These are: sensitivity to the coherence of amplitude modulation induced in the auditory system by the FM (which is somewhat disrupted by roving the carrier frequencies); and sensitivity to fluctuations in the residue pitch evoked by the two carriers (the residue pitch being less salient for inharmonically related carriers).