Journal of Comparative Physiology A

, Volume 193, Issue 1, pp 13–20

FM signals produce robust paradoxical latency shifts in the bat’s inferior colliculus


    • Department of Molecular and Integrative Physiology and Beckman InstituteUniversity of Illinois
  • Alexander V. Galazyuk
    • Department of Molecular and Integrative Physiology and Beckman InstituteUniversity of Illinois
    • Northeastern Ohio University College of Medicine
  • Albert S. Feng
    • Department of Molecular and Integrative Physiology and Beckman InstituteUniversity of Illinois
Original Paper

DOI: 10.1007/s00359-006-0167-9

Cite this article as:
Wang, X., Galazyuk, A.V. & Feng, A.S. J Comp Physiol A (2007) 193: 13. doi:10.1007/s00359-006-0167-9


Previous studies in echolocating bats, Myotis lucifugus, showed that paradoxical latency shift (PLS) is essential for neural computation of target range and that a number of neurons in the inferior colliculus (IC) exhibit unit-specific PLS (characterized by longer first-spike latency at higher sound levels) in response to tone pulses at the unit’s best frequency. The present study investigated whether or not frequency-modulated (FM) pulses that mimic the bat’s echolocation sonar signals were equally effective in eliciting PLS. For two-thirds of PLS neurons in the IC, both FM and tone pulses could elicit PLS, but only FM pulses consistently produced unit-specific PLS. For the remainder of PLS neurons, only FM pulses effectively elicited PLS; these cells showed either no PLS or no response, to tone pulses. PLS neurons generally showed more pronounced PLS in response to narrow-band FM (each sweeping 20 kHz in 2 ms) pulse that contained the unit’s best frequency. In addition, almost all PLS neurons showed duration-independent PLS to FM pulses, but the same units exhibited duration-dependent PLS to tone pulses. Taken together, when compared to tone pulses, FM stimuli can provide more reliable estimates of target range.


EcholocationTarget rangingFrequency modulationOscillationFirst-spike latency


Microchiropteran bats use a sonar system to orient and navigate in 3D space and to detect, discriminate, and locate prey through analysis of echoes of their emitted sonar signals. Target ranging is particularly important for localizing and tracking a moving prey and has been extensively studied behaviorally and physiologically. Behavioral studies by Simmons (1973, 1979) have demonstrated that the most important cue for ranging is the echo delay, the time between bat’s sonar emission and its echo. Physiological studies have shown that a substantial population of neurons in the central auditory system of echolocating bats is tuned to echo delays, exhibiting response facilitation to pulse–echo pairs at specific echo delays (Feng et al. 1978; O’Neill and Suga 1979; Sullivan 1982a, b; Mittmann and Wenstrup 1995; O’Neill 1995). Bats’ echolocation signals are highly adaptive but two major classes are distinguishable depending on the species and foraging behavior (Neuweiler 1984): FM bats primarily emit brief frequency-modulated (FM) sonar signals, whereas CF/FM bats emit longer constant frequency (CF) sonar pulses ending with a downward FM.

Suga and colleagues (Suga 1983) showed that, in the mustached bat (a CF/FM bat), delay-tuned cortical neurons are sensitive to the delay of the “FM” (and not the CF) component of bat’s echolocation signal. This suggests that FM plays a more important role in target ranging and processing of echo delays than does the CF component. For little brown bats (a FM species), Sullivan (1982a, b) proposed that delay-tuned responses in the auditory cortex are attributed to a coincidence detection mechanism, and neurons exhibiting paradoxical latency shift (PLS, i.e., longer response latencies for more intense sounds than for weaker sounds) play a critical role in the construction of delay-tuned responses. For the majority of cortical neurons showing PLS, the magnitude of the latency shift for each neuron corresponds to the echo delay to which the neuron is best tuned (Sullivan 1982b; Berkowitz and Suga 1989). The phenomenon of PLS has since been observed in other parts of the bat’s central auditory system, primarily at the level of the inferior colliculus (IC) and above, as well as in the IC of non-echolocating vertebrates (Olsen and Suga 1991a, b; Covey 1993; Klug et al. 2000; Galazyuk and Feng 2001; Galazyuk et al. 2005). Some of the above studies used FM and others utilized CF as sound stimuli. The relative efficacy of FM versus CF for eliciting PLS is unknown.

Galazyuk et al. (2005) recently showed that, in the IC of little brown bats, high-threshold leading inhibition of a unit’s oscillatory discharge is a building block for PLS; for PLS neurons, application of bicuculline abolishes PLS and unmasks the unit’s periodical firing pattern. Furthermore, the periodical firing pattern of IC neurons is unit-specific, as is the units’ PLS. However, it is unclear whether or not the cell-specific PLS (and oscillatory discharge) is stimulus dependent, and in particular whether FM stimuli can also elicit it. In the present study, we compared the effectiveness of FM stimuli and tone bursts in the generation of PLS.


Surgical and recording procedures

Experimental subjects comprised 20 little brown bats, Myotis lucifugus. Details of experimental methods are given in Galazyuk and Feng (2001). Briefly, for surgery, the animal was anesthetized via halothane inhalation (4% halothane administered by a precision vaporizer). After incision of the skin and clearing of the tissues above the skull, a small metal rod was glued to the skull using glass ionomer cement. Following the surgery, animals were allowed to recover for 2–4 days in individual holding cages.

Recordings were made from awake bats in a sound-attenuating chamber. The metal rod on the bat’s head was secured to a small holder for restraining the animal’s head atraumatically, leaving the ears unobstructed for free-field acoustic stimulation. A small hole (∼50 μm) was then made in the skull overlying the IC through which a recording electrode was inserted to reach the IC. Throughout the recording session, the animal was offered drinking water periodically and monitored for signs of discomfort. After a recording session of 6–8 h, the exposed skull was covered with sterile bone wax, and the animal was returned to its holding cage. Such experiments proceeded every 2–3 days for a maximum of 3 weeks.

Extracellular single-unit recordings were made with glass micropipettes (10–20 MΩ, 2–3 μm tip) filled with horseradish peroxidase (5% in 0.2 M potassium acetate in Tris buffer). Such electrodes produce stable single-unit recordings over long periods and thus are preferred over other types of recording electrodes. The electrode was positioned above the IC. The relative position of each electrode was monitored from the readouts of digital micrometers using a common reference on the skull. Vertical advancement of the electrode was made by a precision piezoelectric microdrive from outside the sound-attenuating chamber. Spikes were amplified (Dagan 2400), band-pass filtered (0.3–10 kHz, Krohn-Hite 3700), stored on the computer hard drive, and processed off-line using BrainWare (Tucker-Davis Technologies).

Our experimental protocols for bats are in compliance with the “Guide for the Care and Use of Laboratory Animals” (publication no. 86–23 of the National Institutes of Health) and with the Animal Welfare Act of 1966 and its amendments of 1970 and 1976. These were reviewed and approved by the University of Illinois Lab Animal Use and Care Committee.

Acoustic stimulation

Two standard acoustic stimuli were used in the present study: tone pulses at the unit’s best frequency (BF) and FM pulses with linear downward frequency sweep from 80 to 20 kHz, i.e., FM80–20 that approximated bat’s echolocation pulse during the search phase—unless specified otherwise these stimuli had a duration of 2 ms and a rise–fall time of 0.4 ms. In addition to the standard FM pulses above, three narrow-band FM stimuli with a frequency sweep of 20 kHz were also used in some experiments: FM80–60, FM60–40, and FM40–20. Depending on the unit’s absolute BF, one of these narrow FM stimuli encompassed the BF and the remainder outside the BF. Typically we presented 20 tone and FM pulses at each sound level.

Acoustic stimuli were delivered to the bat from a free-field ultrasonic loudspeaker located 60 cm in front of the bat. Sound pulses of a wide range of sound levels, from 10 to 90 dB SPL, were presented at a rate of 1 s−1, in increments of 5 dB. The range of sound level was extended to lower values if the unit’s response threshold was below 10 dB SPL. The absolute sound pressure level (dB re: 20 μPa) was measured with a 1/4 in. condenser microphone (Brüel & Kjær 4135) and a RMS-measuring amplifier (Brüel & Kjær 2610).

Tone pulses were used to determine a unit’s frequency-tuning curve, the units’ BF, and its rate-level function to BF (from which we determined the unit’s temporal discharge patterns and response latencies at different sound levels). Then, the standard FM pulses were presented to generate a separate rate-level function; the unit’s temporal discharge patterns and response latencies to the tone and FM pulses were later compared. Time permitting, the unit’s rate-level functions to narrow FM pulses were made to extend the comparison further.

Data analysis

Dot raster histograms were used to depict a unit’s temporal discharge pattern at each level. Each dot in a raster histogram indicated a spike at the relative time instant with respect to the stimulus onset. Dot raster histograms from different sound levels were combined to create a unit’s composite dot histogram for visualization of how the response latency and the firing pattern changed globally with sound level. To quantify the response latency, we determined the mean first-spike latencies over 20 presentations within the “response window” (defined as the window wherein the spike count is >25% above the background spontaneous firing). The spontaneous firing rate was the average response over a 50 ms window before each stimulus epoch; IC neurons generally showed < 1 spike over this window. To assess the unit’s temporal discharge patterns to different stimuli, for each stimulus we constructed either a composite peri-stimulus-time (PST) histogram across all sound levels or a separate PST histogram and/or inter-spike interval (ISI) histogram at each level.


PLS neurons

A total of 186 IC neurons were studied. Of these, 39 neurons (21%) exhibited PLS to tone pulses at the unit’s BF and/or standard FM pulses—this was characterized by an increase in the first-spike latency with an increase in sound level (Fig. 1). As reported previously (Galazyuk et al. 2005), the level-dependent latency shift was generally quantal (Fig. 1a, b), but sometimes showed a smooth progressive increase (Fig. 1d).
Fig. 1

Examples of IC units for which both FM and tone pulses at the unit’s BF elicited PLS. Shown are the temporal firing patterns of two representative IC neurons in response to FM and tone pulses at various stimulus levels. Each dot in these composite dot histograms represents a spike at a relative time instant (re: stimulus onset time). For each sound level, responses to 20 trials are shown in the dot histograms. Unit 20R31 in a and b shows quantal level-dependent shifts in the first-spike latency in response to both FM and tone pulses. Unit 19L51 in c and d shows differential PLS (quantum PLS to FM pulses and gradual shift in first-spike latency to tone pulses)

The majority of PLS neurons (25/39) exhibited latency shifts to both FM and tone pulses. For example, unit 20R31 in Fig. 1a, b showed an average latency of 8.5 ms in response to FM pulses at 10–55 dB SPL, and the response latency was shifted to 12.7 ms in a quantum step when the sound level was increased to 60 dB SPL and to 17.1 ms at ≥ 65 dB SPL. In response to tone pulses at the unit’s BF, its first-spike latency showed a shift from 9.5 to 14.3 ms at 50 dB SPL (Fig. 1b). In both cases, the shift in the first-spike latency was quantal (∼4.2–4.5 ms). Unit 19L51 in Fig. 1c, d exemplified neurons showing differential PLS for FM and tone pulses. For FM stimuli, its first-spike latency shifted quantally from 11 to 16 and 21 ms when the sound level was elevated to 80 and 90 dB SPL, respectively. In contrast, in response to tone pulses, its first-spike latency showed a progressive increase from 10 to 17 ms when the sound level was increased above 60 dB SPL. In nature, the consequence of the PLS is that the intense bat’s sonar signal (the “pulse”) can be expected to elicit relatively long response latencies, whereas the returning echo (the “echo”) is relatively weak and thus would generate relatively short latencies. The differential effects of intense versus weaker signal allow the auditory system to evaluate the elapse time between the pulse and the echo.

For each PLS neuron, we compared the difference in the average first-spike latencies at 90 and at ∼50 dB SPL for each stimulus. We found that the average differences in response latencies for tone and standard FM pulses were 3.5 and 4.8 ms, respectively; these were statistically significant (paired Student’s t test, P < 0.005).

For the remaining 14 PLS neurons, FM was the more effective stimulus for eliciting the PLS: these cells exhibited PLS in response to FM, but showed either no PLS or no response to tone pulses. For example, unit 8R21 in Fig. 2a responded to FM pulses at 10–70 dB SPL with an average first-spike latency of 14 ms; the response latency jumped to ∼20 ms in a quantum step when the sound level was increased to 75 dB SPL and beyond. The same unit responded to tone pulses at its BF but only at high sound levels above 50 dB SPL (Fig. 2b); an increase in the sound level produced an increase in spike count, accompanied by a slight decrease in the first-spike latency, with no evidence of PLS. Five of the 14 neurons exhibited PLS in response to FM stimuli, but showed poor or no responses to tone stimuli. For example, unit 7L21 in Fig. 2c showed a tonic discharge to FM pulses at 40–65 dB SPL with a first-spike latency of ∼17 ms; the latency shifted to 21.7 ms at a sound level of 70 dB SPL and beyond (Fig. 2c). This neuron did not respond to tone pulses (Fig. 2d) and thus, its BF could not be determined.
Fig. 2

Examples of IC neurons for which FM pulses elicited distinct PLS but tone pulses did not. Unit 8R21 in a and b exhibits a quantum shift in first-spike latency in response to FM pulses; it responds to tone pulses at high sound levels with no evidence of PLS. Unit 7L21 in c and d shows PLS in response to FM pulses, but no response to tone pulses at any frequency from 20 to 80 kHz and at all sound levels tested. Shown in d is the response to a randomly selected frequency of 32 kHz

For eight PLS neurons, in addition to the standard FM sweep of 80–20 kHz, we also tested the unit’s responses to narrow-band FM stimuli, i.e., FM80–60, FM60–40, and FM40–20. We found that the FM pulses containing the unit’s BF consistently elicited more distinct PLS (and oscillatory responses) when compared to the standard FM and other narrow-band FM pulses, as exemplified by unit 10L41 (BF = 28 kHz) in Fig. 3. This unit exhibited periodic discharges in response to standard FM pulses, with no evidence of PLS (Fig. 3a, e). When stimulated with narrow-band FM pulses that did not contain the unit’s BF, there was no evidence of PLS or firing periodicity (Figs. 3c, g, 4d, h); its response threshold to FM80–60 was 75 dB SPL, higher than for all other FM stimuli (Fig. 3b). In contrast, the unit’s responses to FM40–20 (that contained the unit’s BF) were distinctly periodic (Fig. 3d) and showed a shift in the first-spike latency at ≥ 80 dB SPL (Fig. 3h).
Fig. 3

Composite dot histograms (left column) of an IC unit (BF = 28 kHz) in response to the standard FM pulse (FM80–20) and three narrow-band FM pulses showing that the FM segment containing the unit’s BF elicited a most pronounced PLS and oscillatory response. Responses to the standard FM pulses are shown in a. Responses to FM80–60, FM60–40, and FM40–20 are shown in b, c, and d, respectively. In the right column, shown are post-stimulus time (PST) histograms at 90 and 75 dB SPL using a bin width of 1 ms; data from the standard FM pulse, FM80–60, FM60–40, and FM40–20 are shown in e, f, g, and h, respectively

During echolocation, bats actively control and optimize their sonar signals (Surlykke and Moss 2000; Ghose and Moss 2003). In particular, when a bat approaches a target, its sonar emission rate is progressively increased and at the same time the signal duration is reduced to allow rapid updating of target information at short ranges. To determine whether or not signal duration has an influence over a unit’s PLS, for 19 IC neurons we compared each unit’s responses to BF and standard FM pulses having different stimulus durations, ranging from 1 to 8 ms (the range of durations these bats produce in nature). Of these, 17 units showed robust PLS to FM stimuli, largely independent of the stimulus duration, despite the fact that changing the duration of FM pulses alters the rate of frequency sweep. This is exemplified by the results of unit AGC14 in Fig. 4. This unit showed PLS of 5.1 ms to tone pulses (at the unit’s BF of 38 kHz) and standard FM pulses when the stimulus duration was 1 and 2 ms (top two panels of Fig. 4a, b). With stimulus durations of 4 and 8 ms, whereas it ceased to respond to tone pulses at high sound levels (bottom two panels of Fig. 4a), its response to FM pulses continued to exhibit PLS (bottom two panels of Fig. 4b); the magnitude of PLS was unchanged in spite of an overall increase in the response latency (the latter was due to the delay in the time of occurrence of the 38 kHz component when the duration of FM sweep was increased). The remaining two IC neurons had very weak and irregular responses to both tone and FM pulses. Therefore it was difficult to define their duration response functions.
Fig. 4

Effects of stimulus duration on PLS for a representative IC neuron (unit AGC14). Shown are composite dot histograms (left column) and the corresponding composite PST histograms (right column), in response to tone pulses at the unit’s BF of 38 kHz (a) and standard FM pulses (b) having different stimulus durations. A change in stimulus duration compromised the unit’s PLS to tone pulses, but not to FM pulses. The stimulus duration, ranging from 1 ms (top panels) to 8 ms (bottom panels), is shown as a thick horizontal bar below each dot histogram. The unit’s responses throughout the entire dynamic range (from 10 to 90 dB SPL) were used to construct the composite PST histograms (bin width = 1 ms) for each stimulus


The FM is a prominent feature in the sonar signals of echolocating bats. As described earlier, it is an important sound component for target ranging, for both FM bats and CF/FM bats. Interestingly, in spite of its importance in coding of target range, the underlying reason has not been determined physiologically. In theory, any sound pulse with a clear-cut onset and offset (i.e., having distinct amplitude modulations) can be used to estimate the elapse time between the pulse and its echo, whether the sound is a constant frequency tone or a broadband FM stimulus. However, behavioral evidence indicates that FM signal is better suited for temporal processing than pure tone because it is broadband and thus is processed by an array of neurons tuned to different narrow frequency bands (Simmons 1979; Roverud 1993)—these studies showed that bats do not monitor echo onset and instead they make use of information contained in the internal structure of the signal.

Results of the present study support the behavioral studies showing that FM stimuli are well suited for encoding echo delay. We found that FM stimuli produce oscillatory discharges and PLS more consistently and reliably than CF stimuli (Figs. 2, 3). Specifically, over one-third of the 39 PLS neurons in the IC exhibited oscillatory discharges and PLS to FM stimuli only; these cells either did not respond to tone pulses or showed no PLS in response to tone pulses. For little brown bats, previous studies showed that neurons exhibiting oscillatory discharges comprise a building block for PLS (Galazyuk and Feng 2001; Galazyuk et al. 2005), and that in the cortex PLS is involved in coincident detection of echo delay and the creation of delay-tuned responses (Sullivan 1982b). Taken together, FM signal not only permits multiple estimations of target range by a large network of neurons, but also more reliable estimates of target range (see below).

Our study also showed that the PLS resulting from FM stimuli was largely independent of stimulus duration (Fig. 4). By comparison, the PLS to tone pulses was duration dependent. During a target pursuit, an echolocating bat typically increases its sonar emission rate and concomitantly it progressively decreases the signal duration (Surlykke and Moss 2000; Wadsworth and Moss 2000; Ghose and Moss 2003). The consistency of PLS for FM stimuli across a wide range of stimulus durations can offer an invariant estimate of target range over the entire target pursuit sequence. Our results support the suggestion of Sanderson and Simmons (2005) that the response selectivity of neurons in the bat IC to tone duration (Pinheiro et al. 1991; Casseday et al. 1994; Ehrlich et al. 1997) probably plays little role in guiding the bat’s target pursuit.

At this time, the mechanisms that underlie differential dynamic responses to broadband FM and narrow-band tone pulses are unclear. Various studies have investigated the response selectivity of central auditory neurons to FM sounds (see review by Langner 1992). Along the ascending auditory pathways, neurons become increasingly more selective to FM sounds. In particular, many IC neurons are tuned to narrow ranges of FM (Suga 1965, 1968, 1969; Schuller 1979; Bodenhamer and Pollak 1981; Rees and Moller 1983; Suga et al. 1983; Langner and Schreiner 1988; Fitzpatrick et al. 1991; Casseday and Covey 1992; Poon et al. 1992; Fuzessery 1994; Casseday et al. 1994, 1997; Gordon and O’Neill 2000; O’Neill and Brimijoin 2002), and this response selectivity is abolished when GABAergic inhibition is blocked (Fuzessery and Hall 1996). Although many of these studies investigated the relative efficacy of FM versus tone stimuli for eliciting neural responses at several stimulus levels, the investigations focused on the spike count metric, thereby overlooking the units’ temporal discharge patterns. As such, further studies, and particularly intracellular studies, are needed to elucidate the mechanisms underlying the differential PLS responses to FM and tone bursts.

Wong and colleagues (Shannon-Hartman et al. 1992; Maekawa et al. 1992) previously showed that, in the primary auditory cortex of the little brown bat, most neurons are FM sensitive and display delay-tuned responses to paired-FM sweep containing the units’ BF (but not to paired tone pulses). Results of the present study in the IC of little brown bats are consistent with their findings. Specifically, we found that a number of IC neurons gave robust oscillatory discharges and PLS in response to FM stimuli that contained the unit’s BF (Fig. 3). As oscillatory discharges serve as a building block for PLS, and the latter is the foundation for delay-tuned responses (Sullivan 1982b; Galazyuk and Feng 2001; Galazyuk et al. 2005), these neurons likely contribute to the delay-tuned responses to paired-FM sweep in the cortex.


This research is supported by a grant from the National Institute on Deafness and Communication Disorders of the NIH (R01DC04998). We thank Wenyu Lin and Karla Melendez for their comments on earlier versions of this manuscript.

Copyright information

© Springer-Verlag 2006