lable at ScienceDirect

Animal Behaviour 79 (2010) 787–796

Contents lists avai

Animal Behaviour

journal homepage: www.elsevier .com/locate/anbehav

A mechanism for antiphonal echolocation by free-tailed bats

Jenna Jarvis, Kirsten M. Bohn, Jedediah Tressler, Michael Smotherman*

Department of Biology, Texas A&M University

a r t i c l e i n f o

Article history:Received 8 June 2009Initial acceptance 3 September 2009Final acceptance 29 December 2009Available online 2 February 2010MS. number: A09-00381R

Keywords:acoustic maskinganimal communicationbatecholocationfree-tailed batnoiseTadarida brasiliensisvocalization

* Correspondence: M. S. Smotherman, DepartmUniversity, ms 3258 TAMU, College Station, TX 77843

E-mail address: smotherman@tamu.edu (M. Smot

0003-3472/$38.00 The Association for the Study of Adoi:10.1016/j.anbehav.2010.01.004

Bats are highly social and spend much of their lives echolocating in the presence of other bats. To reducethe effects of acoustic interferences from other bats’ echolocation calls, we hypothesized that bats mightshift the timing of their pulse emissions to minimize temporal overlap with another bat’s echolocationpulses. To test this hypothesis, we investigated whether free-tailed bats, Tadarida brasiliensis, echolo-cating in the laboratory would shift the timing of their own pulse emissions in response to regularlyrepeating artificial acoustic stimuli. Each bat tested showed a robust phase-locked temporal pattern inpulse emissions that included an initial suppressive phase lasting more than 60 ms after stimulus onset,during which the probability of emitting pulses was reduced by more than 50%, followed bya compensatory rebound phase, the timing and amplitude of which were dependent on the temporalpattern of the stimulus. The responses were nonadapting and were largely insensitive to broad changesin the acoustic properties of the stimulus. Randomly occurring noise bursts also suppressed calling for upto 60 ms, but the time course of the compensatory rebound phase was more rapid than when the batswere responding to regularly repeating patterns of noise bursts. These findings provide the first quan-titative description of how external stimuli may cause echolocating bats to alter the timing of subsequentpulse emissions.

The Association for the Study of Animal Behaviour. Published by Elsevier Ltd.

Bats are highly social and spend much of their lives echolocatingin the presence of other bats, but echolocation is susceptible to thedegrading influences of acoustic interference, especially fromconspecific vocalizations (Arlettaz et al. 2001; Jones 2008). Bats areknown to make subtle shifts in the acoustic structure of theirecholocation pulses (Obrist 1995; Gillam et al. 2007; Bates et al.2008) to help distinguish their own echoes from those of other bats,but it is unknown whether bats also possess mechanisms forcoordinating the timing of their pulse emissions with their neigh-bours to minimize acoustic overlap. Many animals regulate thetiming of their vocalizations to avoid acoustic interference,including frogs (Loftus-Hills 1974; Zelick & Narins 1985; Mooreet al. 1989), birds (Ficken & Ficken 1974; Knapton 1987; Brumm2006; Planque & Slabbekoorn 2008) and primates (Egnor et al.2007). Although echolocation serves a different function than otherforms of vocal communication, it is possible that when bats areecholocating in groups they too possess mechanisms for mini-mizing acoustic overlap in time.

Many species of bats live in large, dense colonies and may evenforage together at prime hunting grounds. Free-tailed bats, Tadaridabrasiliensis, partially mitigate the confounding effects of acoustic

ent of Biology, Texas A&M-3258, U.S.A.herman).

nimal Behaviour. Published by Els

interference from other bats by either calling louder (Simmonset al. 1978; Tressler & Smotherman 2009), or changing the spectralparameters of their echolocation pulses in a behaviour known asthe ‘jamming avoidance response’ (Ulanovsky et al. 2004; Gillamet al. 2007; Tressler & Smotherman 2009), but these adaptationsmay be energetically costly, are constrained by laryngealmechanics, and may require the bats to alter pulse characteristicsaway from the optimal acoustic values for foraging and navigation.Temporal strategies for echolocating in concert with other batswould have the benefit of allowing bats to continue to use opti-mized pulse acoustics without necessarily increasing energyexpenditure. The benefits of any such pulse-timing mechanismswould probably be minimal in the presence of large numbers ofecholocating bats, but that would also be true of the jammingavoidance response. However, the use of temporal shifts in pulsetiming performed in addition to the free-tailed bat’s jammingavoidance response might offer more comprehensive and flexiblesolutions to the problem of echolocating within the context of smallgroups of bats.

Here we present the results of a study in which we characterizedhow bats coordinated the timing of their echolocation pulses withdifferent types and temporal patterns of artificial acoustic stimuli.First, we tested the hypothesis that free-tailed bats would shift thetiming of their echolocation pulse emissions to minimize overlapwith a regularly repeating artificial stimulus, and furthermore thatthe behaviour would vary depending on the temporal dynamics of

evier Ltd.

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796788

the stimulus. Second, we tested the hypothesis that the bats mayrespond differently to regularly repeating stimuli versus randomlyoccurring stimuli, which would reflect the role of short-termmemory in the vocal control circuitry. Third, we examined the bats’vocal responses to more complex patterns of temporal stimuli byanalysing their vocal timing in response to either a simple paired-pulse temporal pattern, or an extended rapid pulse train thatmimicked the so-called feeding buzzes uttered by bats during preycapture, territorial displays and agonistic encounters. Finally, wetested whether key features of the acoustic structure of the stim-ulus, bandwidth and duration influenced the bats’ vocal responsesto the stimuli. The results provide the first evidence of a mechanismby which bats may shift the timing of their pulses in response toexternal acoustic stimuli, including the echolocation pulses of othernearby bats.

METHODS

Animals and Experimental Procedure

Fifteen male and five female adult Mexican free-tailed batspermanently housed in the Texas A&M Department of Biology batresearch vivarium were used for this study. All the bats had beencaptured locally and held in captivity for at least 6 months. Allhusbandry and experimental procedures were in accordance withNational Institutes of Health guidelines for experiments involvingvertebrate animals and were approved by the local InstitutionalAnimal Care and Use Committee (protocol no. 2007-254). Bats werekept on a reverse day/night light cycle (14:10 h light:dark cycle)with lights off at 1200 hours. Temperature and humidity within thevivarium were automatically regulated to simulate local naturalconditions. Fresh food and water were provided just before noondaily and the bats normally emerged from their artificial roostswithin 30 min of the lights going out, flew to feeding stations andfed themselves on a diet of mealworms (Tenebrio molitor) supple-mented with essential vitamins and minerals. The bats typicallybecame active within their roosts 1–2 h before noon. All experi-ments were conducted during the time of day when the bats werenormally most active within the vivarium (1000–1400 hours), andbats always received food after completion of each day’s experi-mental trials. All experiments were performed with the lights off.

The complete series of experiments described here took placeover a 2-year period during which groups of 5–10 bats at a timewere separated from the main colony and kept in small holdingcages (3.65 m2) for 1–2 weeks of behavioural testing before beingreturned to the main colony. This was done because free-tailed batsremain healthier in captivity when they are allowed to fly daily andto live in self-defined social groups. Bats used in experiments weremarked for identification and tracking before being returned to thecolony between experiments. Ten of the 20 subjects were chosen atrandom for each experiment. Thus, while all individuals contrib-uted to two or more experiments, the same 10 bats were not usedin all experiments.

During experimental trials the bats were placed in a 10 �10� 20 cm cage composed of 0.5-inch (1.27 mm) plastic-coated steelmesh that was then set on a small platform in the centre of a warm(30 �C, a typical local temperature outside at night) anechoicrecording chamber and left alone for the duration of the recordingsession (typically 1 h). We habituated bats to the handling proce-dures, cage and recording chamber by exposing them to a series ofa least three recording sessions on separate days that mimicked theconditions of the experimental trials. These habituating recordingsessions were used to obtain preliminary data about the acousticproperties of each bat’s echolocation pulses and to screen for indi-viduals that were unlikely to vocalize consistently enough to

complete the experimental trials. Prior to each experimental trial thebats were placed in the recording chamber and given roughly 20 minto become accustomed to the chamber’s climate before initiatingrecordings. During experimental trials we played a series of acousticstimuli to the bats and recorded all echolocation pulses produced bythe bats. All subsequent analyses investigated the temporal rela-tionship between the timing of each emitted pulse and the imme-diately preceding acoustic stimulus. The null hypothesis was that thetiming of pulse emissions would be unaffected by the presentation ofan acoustic stimulus. Under this hypothesis the probability of a pulsebeing emitted at any time between stimuli is random and thus shouldbe equiprobable with sufficient sampling. Instantaneous pulse rateswere highly variable during experiments, and so preliminary trialswere used to determine the sample size required to reliably detecta significant deviation of responses to repeated noise stimuli froma random distribution of stimulus pulse intervals under controlconditions (described below). Based on this preliminary data we useda minimum of 20 min and 1000 pulses per bat per stimulus condition.

Acoustic Apparatus

All acoustic stimuli were generated digitally with Tucker–DavisTechnology (TDT) system III hardware and the OpenEX softwarev5.4 (Tucker–Davis Technologies, Inc., Alachua, FL, U.S.A.). Stimuliwere played through a Sony amplifier (model no. STR-DE598)driving a four-speaker array composed of two Pioneer RibbonTweeters (ART-55D/301080) and two Pioneer Rifle Tweeters (ART-59F/301081) juxtaposed and oriented towards the cage holding thebat. The speakers provided a flat (�3 dB) output of 85 dB SPL overthe range of 15–60 kHz recorded at the position of the bat. The bats’echolocation pulses ranged in intensity from 90 to 115 dB SPL, asmeasured by a Bruel & Kjær free-field 0.25-inch (6.35 mm) micro-phone (Type 4939) placed just outside the cage and surrounded bya cone of acoustic foam that was oriented to minimize the recordedamplitude of the stimuli relative to the bat’s pulses. Incomingsignals were digitized with a National Instruments DAQmx card,NI PCI-6251 (200 kHz, 16-bit sample rate), viewed in real time withAvisoft Recorder v3.0 (Avisoft Bioacoustics, Berlin), and stored onthe computer hard drive for subsequent analysis. On a separatecomputer, the recorded bat pulses, the recorded stimuli and thedigital time stamps issued by the TDT hardware at the onset of eachstimulus were stored on separate channels for analysis with thehardware and software package Datapak 2K2 (Run Technologies,Mission Viejo, CA, U.S.A.). Transmission delays between the TDTsystem’s output and the actual recordings of sound stimuli at themicrophone (4.4 ms) were accounted for in post hoc analyses.Echolocation pulses were automatically separated from thesimultaneously recorded acoustic stimulus based on differences inamplitude and duration: bat pulses were on average 20 dB louderthan the stimulus at the microphone and were of shorter durationthan the stimuli (mean � SE bat pulse duration was 4.67 � 0.38 ms,N ¼ 10 bats, 300 pulses/bat). Experiments were conducted onindividually isolated bats under conditions that made use ofcommunication calls by subjects highly unusual. Communicationcalls are easily detected since they are significantly longer thanecholocation pulses (Bohn et al. 2008), and we found no commu-nication calls during a manual review of the acoustic recordingsobtained during the experiments. Thus, communication calls didnot contribute to the subsequent results or their interpretation.

Experimental Procedures

We first sought to establish whether free-tailed bats wouldchange the timing of their pulse emissions in response to regularlyrepeating noise stimuli and, if so, to characterize the general nature

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796 789

of that response. Pilot studies indicated that a noise burst couldcause a shift in the timing of a single echolocation pulse (Fig. 1a),but the effect was variable and it occurred to us that the nature andmagnitude of the effect could depend on precisely when the noiseburst appeared relative to the bat’s endogenous rhythms. Since wecould not predict when bats would emit pulses, we adopted thestrategy of using regularly repeating stimuli and analysing whenthe bats emitted pulses relative to the appearance of eachpreceding noise burst (illustrated in Fig. 1b). Normally the temporalpattern of a free-tailed bat’s pulse emissions reflects an endogenousperiodicity dictated by the bat’s respiratory rhythms; controlexperiments were therefore carried out to determine whethera combination of the bats naturally periodic pulse emissionsand a periodic stimulus could randomly result in the appearanceof a relationship where none actually existed. In the short term thisis likely to occur once in a while, but because the bats call discon-tinuously with frequent starts and stops, the probability ofobserving a significant association between breathing and thetiming of the computer-generated pulses would become increas-ingly unlikely with longer recording times. For the control trials weactivated the stimulus-generating software, recorded bat vocali-zations, and performed the experiments as usual except that thespeaker amplifier was turned off and no sounds were generatedfrom the speakers. Control data consistently showed that aftersufficient sampling (>1000 pulses), pulses were equally distributedover time and there was no correlation between the timing of thebats’ pulses and the temporal pattern of the stimulus with thespeaker turned off (see Results, Fig. 2a).

To investigate the effects of external acoustic stimuli on thetemporal patterns of pulse emissions, we conducted four experi-ments that incorporated 12 stimuli, including a control (Table 1).

Experiment 1 tested whether bats would modify the temporalpattern of their echolocation pulses differently in response toacoustic stimuli presented at different rates. The stimulus consistedof a continuous train of 10 ms ‘noise bursts’ repeated every 50, 100,200 or 1000 ms with intervening periods of silence. Noise bursts

100

80

60

40

20

Freq

uen

cy (

kHz)

100

80

60

40

20

0.1 0.2 0.3

0.1 0.2 0.3

Echolocation pulses

Tim

(a)

(b)

Figure 1. (a) Spectrogram of a series of Tadarida brasiliensis echolocation pulses interruptprominent harmonic components, but the stimulus was designed to overlap with just the lowused in this study: echolocation pulses from individual bats were recorded while the bat wathe bat’s pulses relative to each preceding stimulus was assessed with a poststimulus time

were produced by digitally band-pass filtering white noise toa range of 20–45 kHz, which completely overlaps with thefrequency range of the most prominent harmonic component ofthe free-tailed bat’s echolocation pulses. Onset and offset of allstimuli were gated with a 0.1 ms rise/fall time. Each bat wasexposed to only one of the four stimulus repetition rates duringa trial (one trial/day), and the order in which bats were exposed tothe different stimuli was varied pseudorandomly.

Experiment 2 tested whether the regularity of the intervalsbetween signals influences the bats’ temporal patterns of pulseemissions by measuring their responses to noise bursts presentedat randomly varying interstimulus intervals. As in experiment 1, thestimulus consisted of a 10 ms noise burst; however, instead of fixedrepetition intervals, the intervals were randomly varied between200 and 1000 ms. The range of intervals was constrained to allowfor a statistical comparison between the response to these semi-random intervals versus the responses to regular repeating inter-vals of 200 and 1000 ms, and because longer intervals requiredimpractically long recording times.

Experiment 3 examined how the bats would respond to stimuliwith more complex temporal patterns. The first complex stimulusconsisted of a ‘paired-pulse’ stimulus in which two 10 ms noisebursts separated by 50 ms were initiated every 200 ms. The secondstimulus was an artificially generated pulse train mimicking thefeeding ‘buzzes’ used by bats during the final stage of prey capture.The artificial buzzes consisted of trains of 15 downward-sweeping,frequency-modulated (FM) sounds (45–20 kHz, 8–4 ms duration)that shortened in duration and increased in repetition rate similarto natural patterns recorded in the field (Schwartz et al. 2007). Thebuzz stimulus was 200 ms long and was repeated every 500 ms.

Experiment 4 tested whether the duration or bandwidth of thestimulus influences the temporal patterns of the bats’ vocalresponse. All stimuli were repeated every 200 ms, since this repe-tition rate elicited a maximal response. Stimuli were noise burstswith durations of 5, 10, 25 and 50 ms, and a 25 kHz, 20 ms long puretone. The pure tone stimulus was used to test whether bandwidth

0.4 0.5 0.6

0.4 0.5 0.6

Noiseburst

e (s)

ed by a single 10 ms band-limited noise burst. The pulses shown here include threeest harmonic, which is typically the loudest. (b) Example of the experimental protocol

s exposed to noise bursts occurring at varied intervals, and subsequently the timing ofhistogram.

1.4

1.2

1

0.8

0.6

0.4

2

1.5

1

0.5

0

10 30 50 70 90 110 130 150 170 190

10 30 50 70 90 110 130 150 170 190Time (ms)

Nor

mal

ized

pro

babi

lity

of

pu

lse

emis

sion

Day 1Day 2

Day 3

200 ms intervalControl

(a)

(b)

Figure 2. (a) Normalized probability of pulses being emitted within successive timebins spanning the 200 ms time window between successive 10 ms noise burstsrepeated at 5 Hz for 20 min, and controls. Control data reflect experimental resultscollected with the stimulus speakers turned off. Data were normalized so thata value of one represents random chance (equal probability at every bin). Mean � SEbin values falling below one indicate that pulse emissions were suppressed belowwhat was expected due to random chance, and values above one indicate that pulseemissions occurred more frequently that expected. N ¼ 10 bats. (b) Average datafor three bats tested on three successive days with the same stimulus as in(a) (4 ms bins).

Table 1Summary of stimuli used in the four main experiments

Stimulus type Repetition interval (ms) Stimulus duration (ms) Experiment 1 (0–5

Control-none d d XNoise burst 50 10 XNoise burst 100 10 XNoise burst 200 10 XNoise burst 1000 10 XNoise burst Random* 10Paired pulse 50, 200y 10, 10Buzz 500 200Noise burst 200 5Noise burst 200 25Noise burst 200 5025kHz tone 200 20

* The interval between ‘random’ noise bursts varied continuously over a range of 200y The paired-pulse stimulus consisted of two 10 ms noise bursts separated by 50 ms a

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796790

affects the nature or magnitude of the response. We used 25 kHzand 20 ms duration because these values approximate the acousticstructure of one of the most common classes of communicationsyllables used by this species (Bohn et al. 2008) and because 25 kHzis centred within the most sensitive bandwidth of the bats auditorysystem (Pollak et al. 1978). We did not examine the effects ofchanging stimulus frequency or amplitude on the bats’ temporalresponses because we assumed that responses to variation in theseparameters would necessarily reflect the physiological propertiesof the bats’ auditory system (Pollak et al. 1978).

Finally, we also documented responses of 10 bats to an artificiallygenerated downward FM-sweep stimulus designed to mimic themost prominent harmonic of the free-tailed bats’ echolocation pul-ses. These FM-sweep stimuli swept downward from 45 kHz to 20 kHzover a period of 5 ms. The bats were tested separately with both theFM-sweep stimulus and a standard 10 ms noise burst stimulus, andboth stimuli were presented at a rate of 5 Hz. Because our analysissoftware was unable to reliably discriminate the bats’ pulse emissionsfrom the FM-sweep presentations, we excluded pulse counts fromthe first 8 ms after each stimulus onset (2� 4 ms time bins), whichprecluded direct comparison with the other data sets.

Statistical Analysis

We used poststimulus time histograms (PSTHs) to quantify andillustrate the proportion of pulses occurring in successive blocks oftime (time bins) relative to each preceding stimulus. We tested forstimuli effects using repeated measures ANOVA models on thepercentage of calls per bat in each time bin with stimulus type as aneffect and stimulus type by time bin as an interaction effect. Theinteraction effect of bin and stimulus was the primary target ofANOVA analyses, as this provided an indication of whether thedistribution of pulses across bins differed significantly betweenconditions. For experiment 1, we ran two tests because differentintervals resulted in different ranges in bins. First we examined the50 ms interval stimulus with the first 50 ms of the 100 ms, 200 ms,1000 ms and control stimuli. Then, we examined the 50–100 msrange of the 100 ms, 200 ms, 1000 ms and control stimuli. For thesecomparisons we used 4 ms bin widths, which resulted in 13 timebins. The remaining analyses (experiments 2, 3 and 4), were con-ducted on the 0–200 ms poststimulus time range, and so we used20 ms time bins for all of the statistical analyses (N ¼ 10 time bins).The 200 ms interval stimuli were used in comparisons in all fourexperiments, so we used a Bonferroni adjustment resulting in analpha (a) of 0.0125. We used least squares contrasts for specificcomparisons and additionally adjusted alpha accordingly. In thosecases alpha is provided, otherwise a ¼ 0.0125.

0 ms) Experiment 1 (50–100 ms) Experiment 2 Experiment 3 Experiment 4

X

XX X X XX X

XXX

XXXX

to 1000 ms.nd repeated together every 200 ms.

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796 791

For presentation purposes, PSTH bin counts were normalizedsuch that a value of 1.0 equalled the predicted mean number ofpulses per bin in the absence of stimuli (i.e. random chance). Forexample, for a PSTH composed of 50 bins, the random averagenumber of pulses per bin is expected to be 2.0% of the total numberof pulses recorded during the experimental trial; thus, experi-mental data were normalized by dividing actual bin percentages by2%. Normalization facilitated comparisons of data with differentstimulus intervals or different bin numbers and widths. All data arepresented as means � standard deviations unless noted otherwise.All statistical procedures were performed using SAS v9.2 andSAS-JMP 8.0 (SAS Institute, Inc., Cary, NC, U.S.A.).

RESULTS

Characterizing the Basic Nature of the Bats’ Temporal Responses toNoise Stimuli

Figure 2a presents the averaged response plot from 10 individualbats responding to 10 ms noise burst stimuli presented at 200 msintervals. This regularly repeating noise stimulus evoked a complexprobability waveform that included a rapid suppression of callingthat reached its lowest value between 40 and 60 ms after thestimulus onset, followed by a compensatory increase in the proba-bility of calling that peaked 100–120 ms after stimulus onset. Thecompensatory peak was followed by a sustained increase in prob-ability of calling above random chance that was sustainedthroughout the second 100 ms of the silent interval. We typicallyobserved a slight increase in the number of pulses during the first20 ms following stimulus onset, which we attributed to a phase lagin the compensatory response to the prior stimulus and not to animmediate response to the most recent stimulus, since a similarincrease in pulse emission probability was not commonly observedin the response to stimuli separated by longer intervals. Pulse ratesvaried among bats from 3.3 to 6.1 pulses per second for the durationof the 20 min recordings (4.7 � 1.2 Hz, N ¼ 10 bats), but we found noevidence that pulse rate influenced our measurements of the bats’behaviour. These results led us to conclude that the repetitive noisestimulus clearly appeared to modify the timing of the bats’ echolo-cation pulse emissions in a consistent and reliable way across allbats, which was confirmed statistically in experiment 1.

The time course and the relative degree of pulse suppression arekey metrics of the bats’ temporal response to noise stimuli, so weused histograms with 4 ms bins to provide a detailed analysis of thebats’ responses. The depression in pulse emission probabilityappeared to begin roughly 10–15 ms after stimulus onset, but theprobability of emitting pulses did not fall below one standarddeviation of the controls until the seventh 4 ms bin (24–28 ms), andremained lower than expected until the 20th 4 ms bin (76–80 ms).The relative probability of emitting a pulse reached its lowest meanvalue at bin 15 (56–60 ms), indicating that the likelihood of a batemitting a pulse within this time bin was reduced by an average ofz55% if a noise burst had occurred approximately 56–60 msearlier. The magnitude of this effect was consistently in the range of40–60% across all bats, although in one individual the suppressiveeffect was greater than 80% at 60 ms. The period of pulsesuppression was followed by a compensatory increase in thenumber of pulses emitted above random chance, which began withbin 26 (100–104 ms) and peaked at bin 28 (108–112 ms), at whichpoint the relative probability of emitting a pulse was increased byroughly 34% above random chance, and after which pulse emis-sions gradually returned to baseline levels.

Before beginning experiments 1–4, we examined the possibilitythat repeated exposure to the same stimuli might alter the bats’performance, which might in turn constrain our ability to use the

same bats for multiple experiments. Therefore, we compared theperformance of three naı̈ve bats tested with the 200 ms intervalstimulus on 3 consecutive days (Fig. 2b). The performance of thesethree bats was highly consistent across all 3 days and was quali-tatively identical to the mean performance of the 10 bats testedwith 10 ms noise burst stimuli presented at 200 ms intervals(Fig. 2a). Furthermore, we found no statistically significant changein the bats’ responses across all 3 days (repeated measures ANOVA,bin � day interaction: F8,58 ¼ 1.13, P ¼ 0.37), indicating thatrepeated exposure did not significantly alter the bats’ behaviour inthe presence of the stimulus.

Experiment 1: Effect of Stimulus Repetition Rate on Pulse EmissionPatterns

Experiment 1 tested whether the bats’ responses to regularlyrepeating noise bursts would vary with different stimulus intervals.Identical experiments were performed using 10 ms noise burstsseparated by silent intervals of 50, 100, 200 and 1000 ms intervals(20, 10, 5 and 1 Hz, respectively). Statistical comparisons of thebats’ pulse emission patterns just over the first 50 ms poststimulusacross all repetition rates and controls (no stimuli) revealed thatstimulus and repetition rate caused the bats to call differently duringthat time span (repeated measures ANOVA: F48,133¼ 2.6,P < 0.0001). Comparison of the distribution of pulses within thesecond 50 ms of the 100, 200 and 1000 ms interval stimuli andcontrols also confirmed that stimulus repetition rate significantlyaffected the distribution of pulses during this time span as well(repeated measures ANOVA: F33,77 ¼4.8, P < 0.0001). Post hoccomparisons of all stimuli versus controls were significant for thefirst 50 ms (F12,34 ¼ 3.5, P ¼ 0.002, a ¼ 0.008) and for the second50 ms (F11,26 ¼ 8.0, P < 0.0001, a ¼ 0.008). Collectively these resultsconfirm our observation that bats modify emission patterns in thepresence of noise bursts (Fig. 2) and demonstrate that free-tailedbats display different temporal patterns of pulse emissions whenpresented with the same noise stimulus at different repetition rates.

Figure 3a illustrates the mean responses of 10 bats to the 50 msinterval stimulus, and Fig. 3b illustrates and compares how the batsresponded differently to the 50, 100 and 200 ms interval stimuli.Bats’ responses to the three stimulus rates differed qualitativelyduring the first 50 ms in that the bats’ probability of callingincreased between each burst of noise in response to the 50 msinterval stimulus but showed an initial decrease in response to the100 and 200 ms interval stimuli. For the 50 ms interval stimulus,pulse emission probabilities fell below expected values immedi-ately preceding and during the stimuli, indicating that the batspulse emissions were more frequent between the noise bursts thanduring the noise bursts. Although evidence of phase locking wasapparent (see Fig. 3c), the average depth of modulation was rela-tively shallow. Nevertheless, a clear trend towards calling out ofphase with the stimulus was observed in all 10 bats, and the patternwas visibly different from the first 50 ms of the response to stimulipresented at 100 and 200 ms intervals. The response to the 100 msinterval stimuli (Fig. 3b, d) appeared as an intermediate betweenthe responses to the 50 ms and 200 ms stimuli. The 100 ms intervalstimuli evoked a suppressive period that initially followed a timecourse similar to that observed for the 200 ms interval data, butwith a shallower depth, and the following rebound phase wasinitiated earlier than in the 200 ms interval data.

Experiment 2: Effects of Random versus Predictable StimulusIntervals

Figure 4a compares the mean response plots for 10 batspresented with semi-randomly occurring noise bursts versus

1.15

1.10

1.05

1

0.95

0.90

0.85

1.6

1.4

1.2

1

0.8

0.6

0.4

0.250 ms interval

Nor

mal

ized

pro

babi

lity

of

pu

lse

emis

sion

Nor

mal

ized

pro

babi

lity

of

pu

lse

emis

sion

0 10 20 30 40 50 0 20 40 60 80 100

50 ms100 ms200 ms1000 ms

Time (ms)

1.4

1.2

1

0.8

0.6

1.4

1.2

1

0.8

0.6

0 50 100 150 200

0 50 100 150 200

Nor

mal

ized

pro

babi

lity

Nor

mal

ized

pro

babi

lity (d)

Noi

se b

urs

t

Noi

se b

urs

t

Noi

se b

urs

t

N=11583 calls

N=9191 calls

(c)

(a) (b)

Time (ms)

Time (ms)

Figure 3. (a) Mean � SE response plots for 10 bats calling in the presence of a 10 ms noise burst repeated at 50 ms intervals. (b) Comparison of the mean � SE responses of all 10bats to noise bursts presented at four stimulus presentation rates (only the first 100 ms of the longer intervals are shown). (c) Representative example of an individual responseusing a 200 ms poststimulus time histogram with 4 ms bins, to emphasize the apparent phase relationship between the 50 ms interval stimulus and the temporal pattern of pulseemissions. (d) Similar example of an individual bat calling in the presence of a noise burst repeated with silent intervals of 100 ms.

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796792

their responses to regularly repeating 200 and 1000 ms stimulusintervals. A repeated measures ANOVA indicated a significantdifference between responses to random, 200 ms and 1000 msinterval stimuli (F18,38 ¼ 3.6, P < 0.0005). In response to therandomly varying interval stimuli, the lowest probability ofcalling occurred at approximately 50 ms poststimulus, followedby a compensatory increase in the probability of calling thatpeaked at approximately 90 ms and ended with a rapid return torandom chance. Compared to the 200 ms and 1000 ms intervalstimuli, the random stimuli resulted in an earlier compensatorypeak and a more rapid return to random expectations followingthe peak (Fig. 4a). Thus, although the bats’ response to semi-random stimuli was similar to that for the 200 ms interval, itvisibly differed from the response to 200 ms intervals inmuch the same way that the 100 ms interval stimuli differedfrom the 200 ms interval stimuli, namely that the suppressionappeared more shallow and the compensatory response occurredearlier.

Experiment 3: Effects of More Complex Stimulus Patterns on PulseEmission Patterns

To explore the question of how bats would respond to morecomplex temporal patterns of stimulus presentations, wecompared the bats’ responses to three stimuli: (1) the regularlyrepeating 200 ms silent interval noise burst stimuli, (2) a ‘paired-pulse’ stimulus, in which a pair of pulses separated by 50 ms werepresented every 200 ms for 20 min and (3) an artificial ‘buzz’stimulus (see Methods). Comparing responses over the 200 mstime span immediately following the onset of each stimulus wefound significant differences between the average responses to thethree stimuli (repeated measures ANOVA: F18,38 ¼ 4.6, P < 0.0001).Bats’ responses to the 200 ms interval stimuli differed significantlyfrom their responses to the two more complex stimuli (post hoccomparison: F9,19 ¼ 3.1, P ¼ 0.0003, a ¼ 0.008); both the paired-pulse and buzz stimuli increased the duration of the suppressionphase (Fig. 4a, d). While the first 50 ms of the response to the

1.6

1.4

1.2

1

0.8

0.6

0.4

1.6

1.4

1.2

1

0.8

0.6

0.4

1.6

1.4

1.2

1

0.8

0.6

0.4

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

1.8

Nor

mal

ized

pro

babi

lity

Nor

mal

ized

pro

babi

lity

Nor

mal

ized

pro

babi

lity

Nor

mal

ized

pro

babi

lity

0 50 100 150 200 0 50 100 150 200

0 200 300 400 500100 0 100 150 20050

Time (ms) Time (ms)

Double-pulse stimulusBuzz stimulus

Double-pulse stimulus

Random interval200 ms interval1000 ms interval

Buzz stimulus

Noi

se b

urs

t

Noi

se b

urs

t

Buzz stimulus

(d)

(c)

(a) (b)

Figure 4. (a) Comparison of the mean � SE response to a semi-random interval stimulus versus the responses to regularly repeating noise bursts presented at either 200 or 1000 msintervals. (b) Mean response plots of 10 bats listening to a paired-pulse stimulus consisting of two identical noise bursts separated by 50 ms and repeated every 200 ms.(c) Mean � SE response of 10 bats calling in the presence of an artificial buzz stimulus consisting of 15 frequency-modulated pulses spread over a 200 ms period and repeated every500 ms. (d) Comparison of the paired-pulse stimulus data in (b) versus the first 200 ms of the buzz stimulus data in (c).

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796 793

paired-pulse stimulus was similar to that of the first 50 ms of theregular (single noise burst) 200 ms interval stimulus, the arrival ofthe second pulse at 50 ms postponed the timing of the compen-satory rebound in pulse emission probabilities by approximately anadditional 50 ms. Notably, the second stimulus did not suppresspulse emissions below that which had occurred in response to thefirst stimuli. It is also noteworthy that the bats apparently did nottreat the two intervals (50 and 150 ms) independently, since thepattern of pulse emissions within the 50 ms interval of the paired-pulse stimulus was more similar to that of the average response tothe initial 50 ms of the regularly repeating 200 ms interval stimulithan it was to that in response to the 50 ms intervals themselves(Fig. 3b).

Figure 4c presents the average response plots for 10 batscalling in the presence of the artificial buzz stimulus. As with theother stimuli, the buzz stimulus caused a rapid initial suppressionof pulse emissions that was greatest at 60 ms poststimulus onsetand lasted approximately 100 ms. Interestingly, the compensatoryrebound phase began approximately halfway through the buzz,indicating that the bats did not postpone emitting pulsesthroughout the duration of the buzz. The compensatory phaselasted up to 100 ms after the end of the buzz, although itappeared to contain a prominent notch (or brief reduction inpulse emissions) that coincided with the termination of the buzz.

Notably, the buzz stimulus was the most effective stimulus forprovoking a prominent prestimulus reduction in pulse emissions;during the final 100 ms leading up to the succeeding stimuluspresentation, pulse emission rates dropped progressively by up to20% (Fig. 4c).

To address the question of whether buzzes were more or lesseffective than the paired-pulse stimuli for suppressing pulseemissions, we compared the efficacy of the buzz stimulus duringthe first 200 ms poststimulus to that of the paired-pulse stimulus.The effect of the buzz stimulus did not differ significantly from thatof the paired-pulse stimulus (post hoc comparison: F9,19 ¼ 4.2,P ¼ 0.02, a ¼ 0.008), although responses to buzz stimuli tended tohave slightly shorter suppressive phases and less prominentcompensatory rebounds (Fig. 4d).

Experiment 4: Effects of Changing the Acoustic Parameters of theNoise Stimulus on Bat Temporal Patterns

We investigated whether changing the noise burst duration orbandwidth would significantly affect the bats’ responses. To testthe significance of stimulus durations, we repeated the experi-ments with regularly repeating 200 ms intervals but with 5, 10, 25and 50 ms duration noise bursts (Fig. 5a). To test whether band-width of the stimulus was important, we measured the bats’

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796794

responses to a regularly repeating narrowbandwidth tonalstimulus (25 kHz, 25 ms, 200 ms intervals). Neither changes instimulus duration nor changes in bandwidth significantly affectedthe bats’ average temporal responses to the stimuli (repeatedmeasures ANOVA: F45,128 ¼ 1.1, P ¼ 0.36). The average response tothe narrowband tonal stimuli appeared shallower relative to theeffects of the broadband stimuli presented at the same rate, butclearly the bats responded to the tonal stimuli (Fig. 5a), and theirresponses followed a time course similar to that of the broadbandstimuli. Finally, the bats’ average response plots (Fig. 5b)confirmed that the bats’ temporal responses to FM-sweep stimuliwere qualitatively identical to their responses to noise burststimuli. Collectively these results show that the onset of anyacoustic stimulus overlapping in frequency with the bandwidth ofthe echolocation pulses may be expected to elicit shifts in pulseemission timing.

1.6

1.4

1.2

1

0.8

0.6

0.4

(a)

1.6

1.4

1.2

1

0.8

0.6

0.4

1.8

0.2

2

0.2

Nor

mal

ized

pro

babi

lity

0 50 100 150 200

0 50 100 150 200Time (ms)

5 ms noise burst10 ms noise burst25 ms noise burst50 ms noise burst20 ms, 25 kHz tone

FM sweepNoise burst

(b)

Figure 5. (a) Comparison of the average temporal responses (mean � SE) to regularlyrepeating noise bursts of different durations repeated at 5 Hz and a regularly repeating25 kHz tonal stimulus lasting 20 ms also repeated at 5 Hz (N ¼ 10 bats). (b) Illustrationof the average temporal responses (mean � SE) of 10 bats echolocating in the presenceof a regularly repeating downward FM-sweep (45 to 20 kHz over 5 ms) repeated at5 Hz for 20 min, and separately while listening to a 10 ms noise burst also repeated at5 Hz. The first two 4 ms bins were excluded from the analysis because the FM-sweepstimulus was not reliably distinguishable from the bats’ own pulses using automatedmethods.

DISCUSSION

Our results demonstrate that free-tailed bats display a highlystereotyped shift in the timing of their echolocation pulse emissionpatterns in response to external acoustic stimuli. The temporalpatterns of the bats’ responses varied depending upon the temporalparameters of the stimuli but were apparently insensitive tochanges in the acoustic properties of the stimulus. In general,artificial noise bursts caused a pronounced but incompletesuppression of pulse emissions that began roughly 15–20 ms afterstimulus onset, peaked 60 ms poststimulus onset, and lasted up toapproximately 100 ms following each stimulus. Similar patterns ofvocal suppression were observed in every bat tested. The bats’behaviour did not require conditioning, did not show signs ofhabituation within trials and did not vary with repeated exposuresto the same stimulus. The timing of subsequent pulse emissionsduring a compensatory rebound phase was highly sensitive to thetemporal pattern and repetition rate of the acoustic stimulus. Wehypothesize that temporal patterning of responses to externalacoustic stimuli may provide a mechanism by which neighbouringbats coordinate the timing of their echolocation pulse emissionsantiphonally, thereby minimizing acoustic overlap and ambiguitiesin the returning echoes.

The soonest a noise burst led to a significant change in theprobability of a pulse being emitted was approximately 15–20 msafter stimulus onset, and the suppressive effect was greatest50–60 ms after stimulus onset, even with longer and more complexstimuli such as the paired-pulse and buzz stimuli. The range of15–60 ms is a critical time period in the context of active echolo-cation behaviour, given that the most salient echoes returning fromtargets or obstacles up to 10 m away would be returning over thistime frame. Thus, one of the net effects of the observed suppressioncould be reducing overlap between outgoing pulse emissions andreturning echoes generated by the external noise stimuli.

The time course of the suppressive phase may also offer someinsight into the underlying neural pathways. A 15 ms delay isindicative of a very short reflexive audio–vocal integrationpathway, such as the direct midbrain connections between theascending auditory system and descending vocal motor pathwaythat mediate the acoustic-laryngeal reflex (ALR) in little brownbats, Myotis lucifugus (Jen & Suga 1976). In the ALR, acoustic stimulievokes an excitatory response from laryngeal muscles with a mere6 ms latency in little brown bats (Jen & Ostwald 1977); however, thefunctional significance of this reflex remains vague. It is possiblethat the ALR may be contributing to the initial suppression phase byproviding a mechanism by which external acoustic stimuli couldinduce sufficient tension in the laryngeal muscles to constrainnormal vocalizing. If so, then all stimuli would be expected to haveessentially the same effect on pulse emissions. Although that wasnot found to be true in the present study, the initial time course ofthe induced suppression appeared similar for all stimuli except the50 ms interval stimuli, in response to which call rate increasedrather than decreased during the first 25 ms poststimulus. Sincethe responses to all stimuli began to differ from one another after25 ms poststimulus, it seems evident that the ALR alone cannotaccount for much of this behaviour, but it remains possible that theALR contributes to the initial suppression phase. If so, any contri-butions of an ALR during the responses to the 50 ms interval stimulimust have been masked by other contributing factors.

Alternatively, electrical stimulations of the vocal control regionof the anterior cingulate cortex (ACC) of the moustached bat,Pteronotus parnellii parnellii, evokes echolocation pulses withdelays greater than 50 ms (Gooler & O’Neill 1986), which allows forthe possibility that the arrival of inhibitory inputs from the auditorysystem to the ACC may explain why the suppressive phase peaked

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796 795

around 50–60 ms. It is also possible that the suppressive phase maybe composed of both rapid reflexive inputs to the larynx andmidbrain vocal pattern generators as well as slower forebrainpathways to the frontal cortex and ACC, but future studies will needto address this question directly.

The timing of the compensatory phase is also of critical impor-tance. The bats’ time course for rebounding from suppression wasclearly influenced by the temporal pattern of the stimuli, but itmust also be assumed that the bats’ pulse emissions were con-strained by respiratory patterns as well. The respiratory rate ofa stationary free-tailed bat is typically eight to nine breaths persecond (or 115–120 ms/breath), and the bats prefer to emit pulsesat the initiation of each expiration (Suthers 1988). If, for example,a bat had intended to call 50 ms after stimulus onset and insteadpostponed pulse emission because of an interfering noise, then thebat would have to (1) emit pulses at an irregular time point in therespiratory cycle, (2) shift the respiratory phase to accommodatea shift in pulse timing, or (3) wait until the next respiratory cycle toemit a pulse. Which of these a bat opts for would clearly be influ-enced by stimulus rate: for a stimulus repeated every 200 ms,a pulse postponed from 50 ms poststimulus could be shifted to thenext breath cycle (roughly 165–170 ms after stimulus onset)without altering respiratory phases and still supersede the nextstimulus presentation. That is not what our subjects did, however;when responding to the 200 ms stimulus interval, the bats showeda clear preference for delaying their pulses only until 100–120 mspoststimulus; therefore, we must assume that they either madechanges in their respiratory patterns or shifted the timing of theirpulses within the respiratory cycle. A stimulus repeated every100 ms would appear to pose more of a challenge for the bats sinceit would be difficult for the bats to vocalize continuously and alsomaintain a constant phase relationship with the stimulus withoutaltering respiratory rate. Respiratory constraints on pulse timingmay be one of the reasons the maximum depth of suppression inemission probability was typically only 55% in response to theslowest stimuli and significantly less with faster stimulus patterns.

In these experiments the bats were not performing motivatedbehaviours (i.e. there was no pressing need to maintain the highestpossible signal-to-noise ratios), and it is possible that they wouldshow far more rigid control over the timing of their pulses whenflying or foraging under more natural conditions. Alternatively, itmay be that, in flight, respiration imposes greater constraints onpulse timing in bats because breathing and calling are coupled withwing beat rhythms (Speakman & Racey 1991; Lancaster et al. 1993;Holderied & von Helversen 2003). It seems probable that flying batscould benefit from adjusting the timing of their pulses when theycan anticipate future interferences, but considering that suchinstances may be rare, it may be that this behaviour contributesonly modestly to echolocation in flight. On the other hand,temporal patterning of responses might be more relevant duringroosting, where temporal overlap with regularly vocalizing neigh-bours is likely. Although free-tailed bats often live in large, densecolonies, they also roost in smaller groups, and even within largecaves and buildings it is not unusual to find individuals segregatedinto smaller social groups. Thus, while it is difficult to imagine howtemporal patterning of responses would function in the context ofvery large groups, there are other social situations in which thisbehaviour could be beneficial. Because of the integral link betweenvocalization and respiration, all bats naturally show a periodicrhythmicity in their pulse emissions, and it is this natural rhyth-micity that would allow for one bat to anticipate the timing ofa neighbouring bat’s pulse emissions based on the timing andpattern of preceding pulse emissions. In the presence of muchlarger groups, bats may instead treat the many echolocation pulsesof other bats as continuous background noise, in which case they

may use other adaptive mechanisms for improving echolocation,such as simply calling louder.

The results of the paired-pulse and buzz stimulus experimentsprovided further insight into the free-tailed bats’ temporalpatterning of responses. The paired-pulse stimulus experimentsillustrated that the partial suppression of pulse emissions caused byan initial stimulus can be sustained with subsequent pulses, butimportantly, the overall depth of suppression was not increased byadditional stimuli. Similarly, the results of the buzz stimulusexperiments showed that while bats emitted significantly fewerpulses during the first half of the stimulus, they generally did notpostpone calling indefinitely. Notably, since none of the stimuliconsistently reduced the probability of pulse emissions by morethan approximately 55%, is seems likely that bats faced withcontinuous stimuli from many other bats would only be expected toreduce their average call rate by roughly half, and even then onlyfor limited periods of time, after which they might very well beginto emit pulses more rapidly than they would in response to silence(as in Fig. 4c). The temporal patterning of bats’ responses describedhere would not lead to the suppression of all vocalizations withinthe roost. Other contextual and motivational factors necessarilyinfluence whether and when bats emit pulses on a moment bymoment basis. Furthermore, the timing of pulse emissions may beaffected differently by persistent background noise versus periodicinterferences derived from a small number of vocalizing neigh-bours. To our knowledge there are no published reports describingthe effects of population density on pulse emissions rates in bats.

In conclusion, the results presented here describe how externalacoustic stimuli, possibly including those generated by other bats,may cause a postponement of imminent pulse emissions. Ourevidence showing that the temporal pattern of the stimulus influ-ences the time course of this delay shows that the bats recognizedand responded differently to changing stimulus temporal patternsand were not simply reacting to each preceding stimulus. It seemslikely that this mechanism could rapidly lead to asynchronouspulse emissions among small groups of actively echolocating bats,although we have yet to show that this actually happens. We hopethat these results propel future studies of this behaviour undermore natural conditions.

Acknowledgments

We thank Mr Clint Netherland and the Texas A&M UniversityAthletic Department for access to the bats of Kyle Field, the TexasParks and Wildlife Department for collection permits, and KristinDenton and Barbara Earnest for their help caring for the bats andtheir invaluable assistance and many helpful discussions. Thisstudy was funded by Texas A&M University and the NationalInstitutes of Health NIDCD grant no. DC007962 to M.S.

References

Arlettaz, R., Jones, G. & Racey, P. A. 2001. Effect of acoustic clutter on preydetection by bats. Nature, 414, 742–745.

Bates, M. E., Stamper, S. A. & Simmons, J. A. 2008. Jamming avoidance response ofbig brown bats in target detection. Journal of Experimental Biology, 211, 106–113.

Bohn, K., Schmidt-French, B., Ma, S. T. & Pollak, G. D. 2008. Syllable acoustics,temportal patterns, and call composition vary with behavioral context inMexican free-tailed bats. Journal of the Acoustical Society of America, 124,1838–1848.

Brumm, H. 2006. Signalling through acoustic windows: nightingales avoid inter-specific competition by short-term adjustment of song timing. Journal ofComparative Physiology A, 192, 1279–1285.

Egnor, S. E., Wickelgren, J. G. & Hauser, M. D. 2007. Tracking silence: adjustingvocal production to avoid acoustic interference. Journal of ComparativePhysiology A, 193, 477–483.

Ficken, R. W. & Ficken, M. S. 1974. Temporal pattern shifts to avoid acousticinterference in singing birds. Science, 183, 762–763.

J. Jarvis et al. / Animal Behaviour 79 (2010) 787–796796

Gillam, E. H., Ulanovsky, N. & McCracken, G. F. 2007. Rapid jamming avoidance inbiosonar. Proceedings of the Royal Society B, 274, 651–660.

Gooler, D. M. & O’Neill, W. E. 1986. The central control of biosonar signalproduction in bats demonstrated by microstimulation of anterior cingulatecortex in the echolocating bat, Pteronotus parnelli parnelli. In: The PhysiologicalControl of Mammalian Vocalization (Ed. by J. D. Newman), pp. 153–184. NewYork: Plenum.

Holderied, M. W. & von Helversen, O. 2003. Echolocation range and wingbeatperiod match in aerial-hawking bats. Proceedings of the Royal Society B, 270,2293–2299.

Jen, P. H. & Ostwald, J. 1977. Response of cricothyroid muscles to frequency-modulated sounds in FM bats, Myotis lucifugus. Nature, 265, 77–78.

Jen, P. H. & Suga, N. 1976. Coordinated activities of middle-ear and laryngealmuscles in echolocating bats. Science, 191, 950–952.

Jones, G. 2008. Sensory ecology: noise annoys foraging bats. Current Biology, 18,1098–1100.

Knapton, R. W. 1987. Intraspecific avoidance and interspecific overlap of songseries in the eastern meadowlark. Auk, 104, 775–779.

Lancaster, W. C., Henson, O. W. Jr. & Keating, A. W. 1993. Respiratory muscleactivity during vocalization in flying bats. American Zoologist, 33, 149A.

Loftus-Hills, J. J. 1974. Analysis of an acoustic pacemaker in Strecker’s chorus frog,Pseudacris streckeri (Anura: Hylidae). Journal of Comparative Physiology, 90, 75–87.

Moore, S. W., Lewis, E. R., Narins, P. M. & Lopez, P. T. 1989. The call-timingalgorithm of the white-lipped frog, Leptodactylus albilabris. Journal of Compar-ative Physiology A, 164, 309–319.

Obrist, M. K. 1995. Flexible bat echolocation-the influence of individual, habitat andconspecifics on sonar signal design. Behavioral Ecology and Sociobiology, 36, 207–219.

Planque, R. & Slabbekoorn, J. 2008. Spectral overlap in songs and temporalavoidance in a Peruvian bird assemblage. Ethology, 114, 262–271.

Pollak, G. K., Marsh, D. S., Bodenhamer, R. & Souther, A. 1978. A single-unitanalysis of inferior colliculus in unanesthetized bats: response patterns andspike-count functions generated by constant-frequency and frequency-modu-lated sounds. Journal of Neurophysiology, 41, 677–691.

Schwartz, C., Tressler, J., Keller, H., Vanzant, M., Ezell, S. & Smotherman, M. 2007.The tiny difference between foraging and communication buzzes uttered by theMexican free-tailed bat, Tadarida brasiliensis. Journal of Comparative PhysiologyA, 193, 853–863.

Simmons, J. A., Lavender, W. A., Lavender, B. A., Childs, J. E., Hulebak, K.,Rigden, M. R., Sherman, J., Woolman, B. & O’Farrell, M. J. 1978. Echolocationby free-tailed bats (Tadarida). Journal of Comparative Physiology A, 125, 291–299.

Speakman, J. R. & Racey, P. A. 1991. No cost of echolocation for bats in flight.Nature, 350, 421–423.

Suthers, R. A. 1988. The production of echolocation signals by birds and bats. In:Animal Sonar: Processes and Performance (Ed. by P. E. Nachtigall &P. W. B. Moore), pp. 23–45. New York: Plenum.

Tressler, J. & Smotherman, M. 2009. Context-dependent effects of noise onecholocation pulse characteristics in free-tailed bats. Journal of ComparativePhysiology A, 195, 923–934.

Ulanovsky, N., Fenton, M. B., Tsoar, A. & Korine, C. 2004. Dynamics ofjamming avoidance in echolocating bats. Proceedings of the Royal Society B,271, 1467–1475.

Zelick, R. & Narins, P. M. 1985. Characterization of the advertisement call oscilatorin the frog Eleutherodactylus coqui. Journal of Comparative Physiology A, 156,223–229.