1. Delay-tuned combination-sensitive neurons (FM-FM neurons) have been discovered in the dorsal and medial divisions of the medial geniculate body (MGB) of the mustached bat (Pteronotus parnellii). In this paper we present evidence for a thalamic origin for FM-FM neurons. Our examination of the response properties of FM-FM neurons indicates that the neural mechanism of delay-tuning depends on coincidence detection and involves an interaction between neural inhibition and excitation. 2. The biosonar pulse (P) and its echo (E) produced and heard by the mustached bat consist of four harmonics; each harmonic contains a constant frequency (CF) component and a frequency modulated (FM) component. Thus the pulse-echo pair contains eight CF components (PCF1-4, ECF1-4) and eight FM components (PFM1-4, EFM1-4). The stimuli used in this study consisted of CF, FM, and CF-FM sounds; paired CF-FM sounds were used to simulate any two harmonics of pulse-echo pairs. The responses of FM-FM neurons in the MGB were recorded extracellularly. We found that FM-FM neurons respond poorly or not at all to single sounds, respond strongly to paired sounds, and are tuned to the frequency and amplitude of each sound of the pair and to the time interval separating them (simulated echo delay). 3. All FM-FM neurons are facilitated by paired FM sounds and most are facilitated by paired CF sounds. Best facilitative frequencies measured with paired CF sounds fall outside the frequency ranges of the CF components of biosonar signals, whereas best facilitative frequencies measured with paired FM sounds fall within the frequency ranges of the FM components of biosonar signals. Thus FM-FM neurons are expected to respond selectively to combinations of FM components in biosonar signals. The FM components of pulse-echo pairs essential to facilitate FM-FM neurons are the FM component of the fundamental of the pulse (PFM1) in combination with the FM component of the second, third, or fourth harmonic of an echo (EFM2, EFM3, EFM4; collectively, EFM(n)). 4. The frequency combinations to which FM-FM neurons are tuned reflect small deviations from the harmonic relationship such as occurs in combinations of FM components from pulses and Doppler-shifted echoes. Compared with CF/CF neurons, however, FM-FM neurons are broadly tuned to stimulus frequency. Thus FM-FM neurons are Doppler-shift tolerant and relatively unspecialized for processing velocity information in the frequency domain. 5. In echolocation, echo delay is the primary cue for target distance. Best delays of FM-FM neurons range from 0 to 23 ms, corresponding to target distances from 0 to 3.9 m, which is the biologically important range of target distances in echolocation. 6. The MGB is the lowest nucleus in the ascending auditory pathway that contains FM-FM neurons; FM-FM neurons have been sought in the inferior colliculus, but none has been found. Response latencies of thalamic FM-FM neurons are 2.6 ms shorter, on the average, than those of their counterparts in the cortex. These data strongly suggest that the MGB is where FM-FM neurons are first created. 7. Previous studies have shown that FM1 is the weakest component in airborne biosonar signals but is present in the cochlear microphonic response to the bat's own pulse at suprathreshold levels for FM-FM neurons. The data presented here suggest that the bat's auditory system exploits the selective attenuation of FM1 to differentiate the bat's own pulse from echoes. Thresholds for facilitation of FM-FM neurons are, on the average, 17 dB higher for FM1 than for FM(n). These elevated thresholds for facilitation by FM1 are created within the central auditory system within or before the MGB. In principle, the elevated facilitative thresholds for FM1 can reject the FM1 component of echoes while passing the stronger FM1 component of pulses. As a consequence, different channels for pulses and echoes may be created within a tonotopically organized system such that pulses are represented by auditory pathways tuned to the fundamental, whereas echoes are represented by auditory pathways tuned to the second, third, and fourth harmonics. 8. Delay tuning in FM-FM neurons depends on coincidence detection. Best delays of FM-FM neurons that respond to FM1 and to FM(n), presented individually, closely match the difference in latency between FM1- and FM(n)-evoked excitation. Best delays correlate strongly and positively with latencies of FM1-evoked excitation, but do not correlate with latencies of FM(n)-evoked excitation. Thus the auditory system of the bat creates a range of delay-tuned neurons by delaying excitation in the pathway in which the fundamental of the pulse is represented. The range of FM1 excitatory latencies of thalamic FM-FM neurons is greater than the range reported for tectothalamic afferents, suggesting that additional FM1 transmission delays are created in the MGB. 9. For FM-FM neurons that have best delays of greater-than-or-equal-to 4 ms, stimulation by FM1 produces a period of inhibition that begins and ends before FM1-evoked excitation occurs. The duration of this inhibitory period is positively correlated with best delays of FM-FM neurons, suggesting that this inhibition is intrinsic to the mechanism by which FM1-evoked excitation is delayed. These results are consistent with a process in which FM1-evoked excitation is delayed in the MGB, at the site of convergence of the FM1 and FM(n) channels.