Behavioral assessment of acoustic parameters relevant
to signal recognition and preference in a vocal fish
Jessica R. McKibbena)
Section of Neurobiology and Behavior, Cornell University, Ithaca, New York 14853
Andrew H. Bass
Section of Neurobiology and Behavior, Cornell University, Ithaca, New York 14853
and University of California Bodega Marine Laboratory, Bodega Bay, California 94923
?Received 19 April 1998; revised 11 June 1998; accepted 26 August 1998?
Acoustic signal recognition depends on the receiver’s processing of the physical attributes of a
sound. This study takes advantage of the simple communication sounds produced by plainfin
midshipman fish to examine effects of signal variation on call recognition and preference. Nesting
male midshipman generate both long duration ??1 min? sinusoidal-like ‘‘hums’’ and short duration
‘‘grunts.’’ The hums of neighboring males often overlap, creating beat waveforms. Presentation of
humlike, single tone stimuli, but not grunts or noise, elicited robust attraction ?phonotaxis? by gravid
females. In two-choice tests, females differentiated and chose between acoustic signals that differed
in duration, frequency, amplitude, and fine temporal content. Frequency preferences were
temperature dependent, in accord with the known temperature dependence of hum fundamental
frequency. Concurrent hums were simulated with two-tone beat stimuli, either presented from a
single speaker or produced more naturally by interference between adjacent sources. Whereas
certain single-source beats reduced stimulus attractiveness, beats which resolved into unmodulated
tones at their sources did not affect preference. These results demonstrate that phonotactic
assessment of stimulus relevance can be applied in a teleost fish, and that multiple signal parameters
can affect receiver response in a vertebrate with relatively simple communication signals. © 1998
Acoustical Society of America. ?S0001-4966?98?02412-6?
PACS numbers: 43.66.Gf, 43.80.Lb ?DWG?
The function and underlying mechanisms of communi-
cation in the acoustic modality have remained largely unex-
plored in the largest of extant vertebrate taxa, teleost fish.
Studies of teleost hearing have demonstrated fundamental
similarities with auditory processing in higher vertebrates, in
spite of differences in their auditory endorgans ?review: Pop-
per and Fay, 1993?; thus one might expect similar acoustic
dimensions to have been exploited by both fish and terrestrial
vertebrates for encoding behaviorally relevant information.
We used playback of synthetic signals to test which acoustic
features affect call recognition and attractiveness in a sound-
producing teleost, the plainfin midshipman ?Porichthys nota-
Acoustic signals, generated by vibration of intrinsic
swimbladder muscles, are a prominent feature of social in-
teractions in midshipman fish. Midshipman breed along the
west coast of North America ?Walker and Rosenblatt, 1988?
where the parental, or ‘‘type I,’’ males establish nests under
rocks in the intertidal zone ?see Bass, 1996?. From these
nests, the type I males emit long-duration, multi-harmonic
signals known as ‘‘hums’’ ?Ibara et al., 1983; Brantley and
Bass, 1994; Fig. 1A–C?. When a female enters his nest, a
male stops humming, and spawning may begin. Spawning
can take several hours as the eggs are affixed to the under-
surface of the rock, after which the female departs and the
male alone guards and maintains the developing embryos.
Whereas, in any breeding season, a female probably spawns
only once, depositing all her eggs in one nest ?DeMartini,
1988; Brantley and Bass, 1994?, a male may continue to hum
and attract mates and may have several clutches at different
stages of development in his care. Midshipman also have a
smaller, ‘‘type II,’’ male morph, which pursues sneak
spawning strategies, does not hum, and lacks the type I
male’s secondary sexual characteristics, including its special-
ized vocal system ?Bass, 1992, 1996?.
Midshipman hums can be continuous for minutes to an
hour or more, often with minimal variation in frequency or
amplitude ?Ibara et al., 1983; Brantley and Bass, 1994; Bass
et al., in press?. Hum fundamental frequency is linearly re-
lated to water temperature, increasing about 5 Hz/°C, and is
approximately 100 Hz at 16 °C ?Brantley and Bass, 1994;
also see Bass and Baker, 1991?. Ibara et al. ?1983? found
with simple playback experiments that recorded hums, or
pure tones in the same frequency range, were sufficient to
attract gravid females ?phonotaxis?. This finding, along with
aquarium observations of midshipman nesting behavior
?Brantley and Bass, 1994?, suggests the hum functions as a
mate call. Midshipman also produce brief ?50–200 ms?
‘‘grunt’’ sounds ?Fig. 1D–F? that probably serve an agonistic
function. Both females and type II males have been recorded
making infrequent, single grunts; but only type I males pro-
duce trains of grunts ?Fig. 1D–F?, emitted, during aquarium
observations, in response to intruder males ?Brantley and
a?Electronic mail: firstname.lastname@example.org
3520 3520J. Acoust. Soc. Am. 104 (6), December 1998 0001-4966/98/104(6)/3520/14/$15.00© 1998 Acoustical Society of America
This study first tests the hypothesis that the hum func-
tions as a mate call by examining the phonotactic responses
of gravid and spent females and of both male morphs. Then,
one- and two-choice tests with gravid females are used to
assess the importance of certain acoustic parameters to hum
recognition and attractiveness. Finally, we examine the ef-
fects of signal overlap. Midshipman nests are often clustered,
and males hum simultaneously; thus receivers must com-
monly process concurrent acoustic signals ?Bodnar and Bass,
1997; Bass et al., in press?. We chose to present playback
choices as they would be encountered in a natural situation,
that is, with alternative calls originating from closely spaced
sources and overlapping in time. Therefore, inherent in every
choice task presented here is the problem of identifying, lo-
cating and evaluating concurrent acoustic signals. The effect
of overlap on stimulus effectiveness is further investigated
by presenting two-tone beat stimuli.
I. MATERIALS AND METHODS
A. Experimental animals
Midshipman, including females and type I and type II
males, were collected from nests exposed at low tide along
Tomales Bay and San Quentin Point in Marin County, Cali-
fornia. Females and males ?type I and II? are distinguished
by the size and shape of the urogenital papilla, while the two
male morphs are easily further separated on the basis of size
and coloration. Gravid females typically have a very dis-
tended abdomen, which becomes flaccid and often darker in
coloration in spent females ?e.g., see Brantley and Bass,
1994; Bass, 1996?. Females taken from nests were in various
stages in the spawning process, with many having begun egg
deposition, but the majority were still conspicuously gravid.
Each fish was measured, weighed, and labeled either
with subcutaneous injections of poster paint, creating a
unique color/location pattern, or with a numbered tag sutured
just rostral to the dorsal fin. Most experiments were con-
ducted using gravid females (n?265), which ranged in size
from 8.9 to 20.4 cm ?mean 13.5?2.0 cm? standard length
and 9.3 to 121.8 g ?mean 33.8?18.1 g?. Fish were held for at
least 24 h prior to testing in outdoor, running sea water tanks
at the University of California Bodega Marine Laboratory,
where all experiments were conducted. The temperature the
fish were housed at varied with the temperature of the in-
coming seawater ?usually between 10 and 15 °C?. Live grass
shrimp, brine shrimp and goldfish, and chopped anchovies
were occasionally offered as food. Under these conditions,
females usually would retain their eggs and remain respon-
sive for up to several weeks.
Females and both type I and type II males were tested
for responses to humlike, continuous tones; but all tests of
comparative responses were done with gravid females. Due
to limited numbers of subjects, individual females were used
for multiple tests ?median number of tests per fish: 6, median
number of responses: 3?, and may have encountered the
same stimulus as one of the choices in different tests. Except
for serial one-choice tests, fish were used for no more than
one response per night; and, because some fish were used for
the same test on different nights or for tests that were later
grouped, only the first response of a fish to a stimulus pair is
included in the analysis. Variables such as length of time in
captivity and hormonal state could also have affected indi-
vidual responsiveness. However, since all tests involved
choices or response comparisons, loss of responsiveness was
controlled for. It is not possible to rule out experiential ef-
fects on preference strengths, although it seems likely that
any habituation to the artificial sound sources or stress due to
time constraints on egg viability would tend to decrease
rather than increase selectivity.
B. Experimental layout
Tests were done outdoors in a cylindrical concrete tank
?4-m diameter, 0.75-m water depth? supplied with running
seawater. Since midshipman normally call next to the sub-
strate in water a few meters or less in depth, the boundary
conditions in the experimental tank were not entirely unnatu-
ral. Underwater loudspeakers ?UW-30; University Sound,
Buchanan, MI? were suspended just above the bottom from
water-filled PVC frames, so that no direct contact substrate
conduction was possible. The speakers were placed near the
center of the tank facing outward ?Fig. 2?, well removed
from the wall to reduce both the influence of reflected sound
and incidental approach by the fish, which tended to hug the
tank perimeter in the absence of sound stimuli. For experi-
ments conducted in 1995, the speakers were 137 cm apart
center to center and angled toward the release site ?configu-
ration C in Fig. 2?. These experiments include one of the
FIG. 1. Midshipman calls recorded in the field. ?A? Segment ?1 s? of a hum
waveform showing the nearly flat envelope. ?B? An expansion of the hum
waveform. ?C? Frequency spectrum of same hum showing peaks at the
107-Hz fundamental frequency and harmonics. Amplitude values are rela-
tive only. ?D? Waveform for segment ?6 s? of a grunt train. ?E? Expanded
waveform of single grunt. ?F? Spectrum of grunt shown in ?E?.
3521 3521 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998 J. R. McKibben and A. H. Bass: Signal recognition in fish
response specificity experiments ?Table II, experiment 1?,
one of the beat test stimuli ?Table VI, test C?, and the 90- vs
110-Hz frequency preference tests ?Fig. 7A?. However, for
most experiments reported here ?conducted in 1996–97?, the
speakers were side by side and 50 cm apart center to center
?configuration A in Fig. 2?. Fish are highly mobile subjects
and tend to swim along barriers, such as that formed by the
release cage, rather than simply orienting to a stimulus.
Thus, their initial swimming direction may not reflect local-
ization of a sound but it can quickly bias any response. The
goal in placing the speakers side by side was to minimize
such random responses by increasing the likelihood of a fish
being able to perceive both stimuli upon release. Natural
nests are often as close or closer together ?see Fig. 1 in Bass,
1996?. Around each speaker, 50 by 50 cm squares were
marked with tape to allow quantification of time spent in the
proximity. Two-choice tests of multi-source beat versus tone
preferences utilized two pairs of speakers. One speaker in
each pair played the test tone with the adjacent ‘‘neighbor’’
speaker either playing a second tone or silent ?configuration
B in Fig. 2?.
C. Acoustic stimuli
This study involved both one-choice playbacks, to test
for recognition and response specificity, and two-choice
playbacks, to test for preferences. One-choice tests were used
to compare elicitation of phonotaxis by continuous tones,
grunts and noise, by tones at different frequencies, and by
pulsed, short duration tones. Two-choice tests were used to
assess preferences for tones with different frequencies or in-
tensities as well as to compare the attractiveness of tones
with harmonic, frequency modulated, and beat stimuli.
Stimuli were synthesized ?except for the natural grunt
train? and played ?SoundEdit 16 software? using a portable
computer ?Apple Macintosh PowerBook 540c?. All stimuli
were looped for continuous, transient-free playback; and, for
two-choice tests, individual stimuli were stored as separate
audio tracks in the same file. The stereo output was con-
nected to Nagra battery powered amplifiers, which drove the
UW-30 loudspeakers. Speaker output was monitored and re-
corded for analysis using a hydrophone ?Cornell Bioacous-
tics Research Program, Laboratory of Ornithology; CBRP;
response flat ?1 dB, 60–500? Hz?, suspended with the
sensing element approximately 7.5 cm above the bottom and
15 cm in front of a speaker, connected to a Sony Professional
Walkman. Recordings were digitized at 11 kHz and spectra
computed using Canary, a sound analysis program ?CBRP?.
Table I summarizes the types and ranges of synthetic
stimuli used in this study. The basic humlike stimulus con-
sisted of a continuous pure tone with a frequency between 80
and 140 Hz. Tone levels were set at 130–140 dB re: 1 ?Pa,
15 cm in front of each speaker and matched between speak-
ers for any given test. These stimulus levels are comparable
to what has been recorded at midshipman nests in the field
?CBRP hydrophones; A. Bass and M. Marchaterre, unpub-
lished observations? and to the reported levels for advertise-
ment calls in the related oyster toadfish ?Tavolga, 1971; Ba-
rimo and Fine, 1998?. Amplitudes were equalized across
frequencies by adjusting synthesis amplitudes. Harmonic dis-
tortion at the second harmonic was less than ?40 dB re: the
fundamental 15 cm in front of the speaker. For tests with
amplitude as the variable parameter, both synthesis and am-
plification adjustments were made to achieve 3 dB ??0.5
dB? pressure differences between 90- or 100-Hz tones played
from the two speakers. Levels were verified by measuring
the spectral peaks of stimuli recorded in the tank.
Pulsed stimuli were synthesized by reiteration of a unit
composed of a specified silent interval added to a 0.5-, 0.75-,
1-, or 2-s tone. The tone segment had approximately 45-ms
rise and fall times. Both harmonic stimuli and single-source
beat stimuli consisted of two digitally mixed tones. For the
harmonic stimulus, these were the fundamental frequency,
F1 ?90 or 100 Hz?, and its second harmonic, 2•F1. The
second harmonic is generally very prominent in midshipman
calls, often containing at least as much energy as the funda-
mental ?e.g., Fig. 1C?, thus harmonic stimuli were synthe-
FIG. 2. Diagram of playback tank showing speaker positions for two-
speaker tests ?A, black bars; and C, open bars? and four-speaker tests ?B,
hatched bars?. Squares around speakers in position A represent 50?50 cm
areas marked with tape. DR ? drain pipe; FL ? floodlight; N ? neighbor
speaker; RS ? release site.
TABLE I. Summary of synthetic stimuli and variables presented.
components Variables Range
0.5–2 s ?0.1 s?a
0.25–0.5 s ?0.4 s?a
F1 and 2•F1 Harmonic phase0° or 90°
2 or 5 Hz
?12 to 0 dB re: F1
0 to ?3 dB re: tone
F1 and F2
FM 100 Hz?modulation
?5 or 10 Hz
Noise Noise band
aFor one-choice, specificity tests.
bFor single-source beat.
35223522J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
sized with equal energy at F1 and 2•F1. Beat stimuli were
composed of F1 ?90 or 100 Hz? plus a second tone, F2,
equal to F1?2 Hz or F1?5 Hz ?Fig. 3?.
Frequency-modulated ?FM? stimuli consisted of a
100-Hz carrier modulated by ?5 or ?10 Hz at 10 modula-
tions per second ?i.e., frequency was continually either in-
creasing or decreasing over a 10- or 20-Hz range?. In order to
compensate for the frequency response of the speakers, the
computer-synthesized FM waveform was low-pass filtered,
and then re-digitized and amplified for playback. The noise
stimulus used for response specificity tests was generated
within SoundEdit, then low-pass filtered to produce a 5 s
looped stimulus with peak power, upon playback, between
95 and 130 Hz. Finally, the natural grunt train stimulus was
recorded ?CBRP hydrophone into Sony Professional Walk-
man? near a midshipman nest in Tomales Bay, CA, digitized
?11 kHz sampling rate?, and a 10 s segment looped for con-
Sound pressure in the tank was mapped with measure-
ments taken at 29 points, forming a grid with 6 in. ?15.2 cm?
steps moving away from and perpendicular to the speaker
faces and 10.5 in. ?26.7 cm? steps from side to side in par-
allel with the speaker faces. The hydrophone was positioned
3.5 cm ?transducing element approximately 7.5 cm? above
the tank bottom and stimuli were recorded on a Sony Profes-
sional Walkman and then digitized ?11 kHz sampling rate,
16 bit? for analysis. Sound pressure values at the stimulus
frequency were computed using Canary software based on
stimulus spectra ?FFT size 8192 points?. Over the relevant
frequency range, sound pressure dropped off logarithmically
?0.25 dB/cm? from the speaker to the release site with no
apparent discontinuities. Thus reflections or standing waves
were unlikely to have affected responses. It was not possible
to eliminate all background noise, which sometimes had
peaks within the frequency range of interest ?i.e., 60 and 120
Hz?. However, noise peaks in the 20–200-Hz band were at
least 30–50 dB below the stimulus frequency peak level 15
cm in front of a speaker.
D. Experimental protocol
Testing was conducted during the midshipman breeding
season, from June through September, between 18:00 and
03:00 h, when midshipman are normally most active. Three
red floodlights around the tank perimeter ?FL, Fig. 2? al-
lowed observation and videotaping of responses. Although
this light was apparently visible to the fish ?fish at the perim-
eter sometimes swam to the surface beneath the floodlights?,
it did not seem to affect phonotaxis. The water temperature
in the test tank varied with the temperature of the incoming
sea water and solar heating and could be ‘‘regulated’’ from
night to night by adjusting the rate of water flow ?off during
all tests?. Prior to testing, females were held, several to a
bucket, in water from the test tank at the test temperature and
allowed to acclimate for at least 15 min when the test tem-
perature differed from that in their holding tanks. Tests be-
gan with one fish placed in a 30 cm diameter, plastic mesh
cylinder approximately 60 cm in front of the speakers ?RS,
Fig. 2?. Fish were released by manually raising the cylinder.
The release protocol was modified over the course of
experimentation in order to reduce the occurrence of fish not
leaving the release site or retreating to the tank perimeter.
Thus, one of three release procedures was followed:
?1? The fish was placed in the mesh cage, allowed to accli-
mate for at least 2 min, the stimuli were then turned on
for 30 s, and the fish released.
?2? The fish was placed in the mesh release cage while the
stimuli were playing and then released after 30 s.
?3? For most experiments reported here, the fish was placed
in the mesh release cage while the stimuli were playing
and released as soon as it swam to the front center of the
cage ?used with speakers in configuration A?.
Releasing the fish without an acclimation period seemed to
increase the likelihood of a response, especially in oft tested
fish. Likewise, releasing the fish when they were at the front
of the release cage appeared to increase response rates. This
was presumably because of the steep drop-off in stimulus
level with distance from the speakers. In all cases the stimuli
played continuously after the fish was released. Trials ended
either when the fish swam away from the speaker area or
when the sound from one or both speakers was shut off, the
latter being after at least 30 s in nearly all cases. Most tests
were videotaped ?Sony Handycam? to allow verification of
observations and detailed response analysis.
Preliminary experiments showed that fish sometimes re-
sponded after initially bypassing the speakers and swimming
around the tank. Although such free exploration would likely
be the case in nature, the position of the fish when it began
attending to the playback was not controlled. Thus, only fish
that swam directly to a speaker without leaving the area be-
tween the release site and the speakers were counted as re-
sponding. A positive response was scored if a fish ap-
proached a speaker, and then touched or circled in front of or
under the speaker.
For one-choice tests, a single stimulus was presented out
of one speaker, with a second speaker serving as a silent
control. In order to compare responses to different, individu-
ally presented signals while controlling for various levels of
responsiveness across fish, multiple stimuli were presented
serially to the same fish. Such tests with individual fish were
separated by at least 30 min to avoid problems with handling
FIG. 3. Synthesis of beat stimuli. Two tones of slightly different frequency,
F1 and F2, are added together, resulting in amplitude and phase modula-
tions, known as beats, at their difference frequency ?dF?.
3523 3523J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
stress or short-term experience, and the active speaker was
switched randomly between trials.
For two-choice tests, the alternative stimuli were pre-
sented simultaneously from the two speakers and stimulus
locations were switched either pseudo-randomly to achieve
equal responses in each configuration or alternated from re-
sponse to response. A preference was considered significant
if the distribution of responses was significantly (p?0.05)
different from the two-tailed binomial distribution expected
given the null hypothesis that the proportions of fish ap-
proaching each speaker were equal. Fish that did not ap-
proach either of the stimulus options for a given test were
considered ‘‘nonresponders.’’ Since nonresponses could not
be attributed to the acoustic stimuli, they were excluded from
analysis. Further details of stimulus presentation for each
experiment are presented in the results section.
A. Morph-specific phonotactic responses
Gravid females showedphonotacticresponsesto
humlike continuous tone stimuli, and most of the tested fe-
males responded at some point during their captivity ?81% of
74 in 1995, 64% of 101 in 1996, and 69% of 90 in 1997?.
When spent females were tested with the same stimuli, none
approached the speaker (n?13). For a representative
sample of consecutive trials, the videotaped behavior of
gravid females was analyzed in detail. Phonotactic responses
by gravid females included several typical behaviors ?sum-
marized in Fig. 4A?, and usually began with a straight ap-
proach to one speaker. Most fish approached the speaker
with a distinctive pause-and-glide swimming pattern. Once
at the speaker, the specific movements of a responding fish
reflected the structure of the speaker and stand; but circling
?CI, AF?, and a tendency to go under objects ?UN, TU? are
probably general female midshipman response characteris-
tics. Most responses were unambiguous, incorporating both
physical contact ?TU, TF? and prolonged active interest in
the chosen speaker ?Fig. 4B?. However, in a small percentage
??10%? of trials, fish swam toward a speaker but then con-
tinued immediately toward the back of the tank or stopped
near the speaker and sat on the bottom. Including these ap-
proaches, which lacked speaker contact or circling, did not
change data trends; however, since the fish were not clearly
responding to the sound, they were not counted as responses.
Figure 4B shows the time spent within a 50 by 50 cm
marked square around the speaker for 79 approaches meeting
the response criteria ?either touching the speaker or circling
or both; ‘‘Responses’’? and 13 that did not ?‘‘Nonre-
sponses’’?. Most responding fish spent at least 30 s around
Although not tested as systematically as females, several
type I males approached and showed prolonged interest in a
speaker playing a 90- or 100-Hz continuous tone ?n?9 of 24
fish; 13.1–28 cm standard length?. The responses elicited
from type I males were quite different from those of gravid
females. Whereas females always approached the speaker
head-on, type I males were sometimes observed to back up
to and under the speaker and frequently performed ‘‘dig-
ging’’ motions with their tail and pectoral fins. These move-
ments resemble the nest-building behavior typical of type I
males ?Brantley and Bass, 1994?. Type II males rarely ap-
proached a speaker playing humlike, 90- or 100-Hz tones,
but the few observed phonotactic responses (n?5 responses
from 3 fish out of 22 tested; 7–11 cm? included ‘‘sneaker-
like’’ behaviors such as sidling up to the speaker frame or
working their tails underneath the frame ?see Brantley and
Bass, 1994?. Unlike females, neither type I nor type II males
touched the speaker face or circled rapidly back and forth in
front of the speaker.
FIG. 4. ?A? Histogram showing frequencies of occurrence of different be-
haviors in a sample of 87 videotaped phonotactic responses by gravid fe-
males. Only trials which included either circling ?CI? or contact with the
speaker ?TU, TF? are included in this analysis. CI?circling: swimming in
circles or back and forth in front of or under the speaker. TU?touching
underside of speaker: approaching speaker head-on or from side and making
contact with underside of projecting speaker face. UN?under: going under
the speaker from front to back, back to front, or from the side. BE?behind:
approaching or lingering behind the speaker within the boundaries of the
speaker stand. TF?touching speaker face: contacting, often appearing to
nuzzle or swim against speaker face. AF?around frame: swimming near the
bottom around the square frame supporting the speaker. ?B? Time spent
within 50?50 cm area around speaker during first 30 s after entering area.
Data are shown for 92 trials including 79 responses ?the 87 in A minus 8
responses without the full 30 s of time data? and 13 instances where fish
crossed into the speaker area but did not circle or touch the speaker ?non-
responses?. Five-second bins; values are upper limits.
35243524 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
B. Response specificity
Table II shows results for two one-choice experiments
where gravid females responded selectively to humlike
stimuli. For each experiment, stimuli were presented one at a
time, in random order, so that each fish was tested once with
each stimulus. The speakers were in the widely spaced con-
figuration ?configuration C in Fig. 2? for the first experiment
and side by side for the second ?configuration A in Fig. 2?. In
the first experiment, 8 of 11 fish approached a continuous
tone ?matched to the 110-Hz fundamental frequency of the
natural grunt train? but none approached the grunts. Like-
wise, in the second experiment, 10 of 14 fish approached a
speaker playing a continuous 100-Hz tone, but no fish ap-
proached a 100-ms pulsed tone, approximating a grunt train,
or noise centered around 100 Hz. No fish approached the
silent control speaker.
C. Stimulus duration
The importance of signal duration to hum recognition
was further investigated with one-choice tests of phonotaxis
using pulsed tone stimuli with various pulse durations and
intervals ?Fig. 5A?. Experimentally naive gravid females
were presented with one of various sets of two to four dif-
ferent pulsed 90-Hz stimuli. Stimuli were presented indi-
vidually; and, within each set, the order of presentation was
varied across subjects to reduce any experiential bias. Fol-
lowing a set of pulsed stimuli, each fish was tested with a
continuous 90-Hz tone. This served as a control for respon-
siveness and was presented last to prevent expectations from
influencing the attractiveness of the pulsed signals. For
analysis, only data from fish that responded to the continuous
control ?80 of 101 fish tested? were included; and results for
specific stimuli were combined across the different stimulus
sets. Thus the same individual may be represented in the data
for various combinations of stimuli ?no fish were tested with
all five stimuli?, but never more than once for the same
The results graphed in Fig. 5B show that pulsed stimuli
of 2 s or less elicited phonotaxis by at least some fish. How-
ever, the percentage of fish approaching the sound tended to
decrease as pulse duration was shortened from 2 to 0.5 s
?stimulus a to e?. Significant differences between pairs of
stimuli ?Fisher’s exact test, p?0.05? are indicated by
matching symbols above the corresponding bars. The pulsed
stimuli also varied in interpulse interval length and, thus,
total energy. A higher percentage of fish approached a 1 s
pulse when the interpulse interval was 0.25 s ?stimulus b?
than when it was increased to 0.5 s ?stimulus c? ?Chi-square
?3.107, p?0.078?. Over the range of stimuli tested, the
percentage of fish responding correlated well with the pro-
portion of the stimulus the sound was on ?duty cycle? ?Fig.
The attractiveness of a pulsed tone relative to a continu-
ous tone was directly compared using a two-choice protocol.
Given a choice between a 90-Hz stimulus consisting of 1 s
pulses with 0.5 s intervals ?same as stimulus c in Fig. 5? and
a 90-Hz continuous tone, 10 of 13 fish approached the con-
tinuous tone (p?0.05). The three fish that initially ap-
proached the pulsed signal all subsequently went to the
speaker playing the continuous tone.
FIG. 5. ?A? Pulsed stimuli ?a–e? consisted of 90-Hz tones with the pulse and
interval durations indicated. Envelopes are shown at right for three seconds
of each stimulus. ?B? Percent of fish that subsequently responded to a con-
tinuous tone ?responsive fish? that approached each of the five different
pulsed stimuli. Significant differences ?Fisher’s exact test, p?0.05? be-
tween responses to pairs of stimuli are indicated with matching symbols
over the corresponding bars. Total number of responsive fish tested with
each stimulus: a, 10; b, 22; c, 19; d, 12; e, 22. ?C? Same response data
plotted in terms of duty cycle ?proportion of stimulus sound was on?. Test-
ing conducted at 14.4–15.6 °C.
TABLE II. Results of one-choice playback tests showing specificity of pho-
notactic response by gravid female midshipman. Experiment 1: Two stimuli
individually presented ?in random order? to each fish ?16.3–16.8 °C?. Ex-
periment 2: Three stimuli individually presented to each fish ?15.7–
16.2 °C?. The number of fish responding to each stimulus is indicated for
each position of that stimulus in the presentation order.
Presentation order: Total responses:
123 ExperimentStimulus:Total fish tested
1 Continuous tone
Field recorded grunts
?110 Hz fundamental?
2 Continuous tone
Pulsed tone 100 Hz
?100 ms on, 400 ms off?
Filtered noise000 0:14
3525 3525 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
Although natural hums contain considerable energy at
the first two or three harmonics ?Fig. 1C?, pure tones ?second
harmonic ??40 dB re: fundamental frequency? were suffi-
cient to elicit phonotaxis. Two-choice experiments tested
whether harmonics might affect the attractiveness of a hum.
Fish were presented with a choice between a pure tone and
that same tone plus its second harmonic, at the same inten-
sity and at the same starting phase as the fundamental. De-
spite the greater intensity of the harmonic signal, fish showed
no preference in tests with either a 90- or 100-Hz fundamen-
tal frequency ?Table III, tests A and B?. Because harmonic
phase, which affects fine waveform structure, including peak
amplitude, might have affected signal attractiveness ?e.g., as
shown for anurans; Bodnar, 1996?, fish were also given a
choice between harmonic stimuli that differed in the phase of
the harmonic. With a 90-Hz fundamental, no preference for
relative phase of the second harmonic, 0° or 90° starting
phase relative to the fundamental at synthesis, was detected
?Table III, test C?.
E. Fundamental frequency
In order to test the frequency selectivity of hum recog-
nition, gravid females were presented individual tones span-
ning a 60-Hz range. Figure 6 shows the results of one-choice
tests with 80-, 100-, 120-, and 140-Hz continuous tones at
two temperatures. Each fish was presented one of the four
FIG. 6. Approaches to individually presented tones at two temperatures.
Each fish was tested with 80, 100, 120, and 140 Hz and the results are
divided ?horizontal dashed lines? according to whether the fish were tested at
just cool ?fish a–c?, just warm ?fish h–l?, or both temperatures ?fish d–g?.
Responses to a given frequency at the cool temperature ?14.2–14.4 °C? are
indicated with filled circles, and, at the warm temperature ?19.4–21.5 °C?,
with open squares.
FIG. 7. Two-choice, continuous tone frequency preferences at different tem-
peratures. ?A? Choice of 90 or 110 Hz over three temperature ranges: 13.6–
15.5, 15.6–17.5, and 17.6–19.5 °C. The number of fish choosing that stimu-
lus is shown next to each bar. Dashed lines in ?A? and ?B? indicate expected
hum frequency at each temperature ?based on Brantley and Bass, 1994; and
unpublished observations?. ?B? Frequency preferences between pairs of
tones differing by 10 Hz at indicated temperatures. For each test, paired
frequency choices are connected with vertical lines, and the preferred ?p
?0.05; large filled circles? and nonpreferred ?small filled circles? frequen-
cies are indicated.
TABLE III. Results of two-choice tests showing approaches to stimuli with
or without second harmonic (2•F1) ?A and B?, and to stimuli with phase of
second harmonic at 0° or 90° relative to fundamental (F1) ?C?.
35263526 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
frequencies per trial, in random order, for a total of four trials
per fish; only data from fish that responded to at least one
frequency at the specified temperature are included. Subjects
are grouped ?separated by dashed lines? according to whether
they were tested at either the cool ?14.2–14.4 °C; fish a–c?
or warm ?19.4–21.5 °C; fish h–l? temperature, or at both
temperatures ?fish d–g?. Approaches to each frequency tone
are plotted with filled circles ?cool temperature? or open
squares ?warm temperature?. At a single temperature ?i.e.,
either cool or warm?, 7 of the 12 responding fish ?fish a, c–g,
k? approached two or more of the frequencies tested, a
20–40 Hz range. An effect of temperature on the attractive-
ness of different frequencies was also suggested. Fish tended
to respond more to higher frequencies ?120 and 140 Hz? at
the higher temperature, although, due to small numbers of
responding fish, the differences were not significant.
To further test the effect of temperature on frequency
preference, fish were given choices between two continuous
tones differing by 10 or 20 Hz at various temperatures. Prior
to testing, subjects were acclimated to the test temperature
for at least 15 min and up to approximately 1 h. The first
experiment ?Fig. 7A? tested preferences for 90- or 110-Hz
tones as temperature varied. Results are grouped into 2 °C
bins and plotted with bars extending over the applicable tem-
perature range. Preferred frequencies are indicated with thick
filled bars and nonpreferred frequencies with thin filled bars.
At the lowest temperatures ?13.6–15.5 °C?, 9 of 10 respond-
ing fish chose the 90-Hz tone (p?0.011); and at the highest
temperatures, ?17.6–19.5 °C?, 9 of 11 fish chose the 110-Hz
tone (p?0.033). There was no clear preference at interme-
diate temperatures ?open bars; 15.6–17.5 °C?.
Figure 7B shows the results of two-choice tests with
stimuli differing by 10 Hz at specific temperatures. Stimulus
choices consisted of continuous tones at 80 and 90 Hz, 90
and 100 Hz, 100 and 110 Hz, and 105 and 115 Hz. Paired
frequency choices are linked by vertical lines, and preferred
frequencies (p?0.05) are marked with large filled circles;
nonpreferred frequencies, with small filled circles. The num-
ber of fish approaching each stimulus is indicated to the left
of each marker. Frequency preference, significant for all
tests, was neither absolute nor simply directional but de-
pended upon temperature. For example, whereas fish chose
80 Hz over 90 Hz near 12 °C, that preference was reversed at
14.5 °C. The 80-, 90-, and 100-Hz stimuli were each, in dif-
ferent tests, both the preferred and nonpreferred stimulus.
Also plotted in Fig. 7A and B is the expected hum frequency
?dashed line; Brantley and Bass, 1994; and unpublished ob-
servations?, which increases linearly over this temperature
range. Frequency test pairs ?Fig. 7B? were chosen to bracket
expected hum frequency, and, for all six tests, fish preferred
the frequency closest to the expected hum frequency at the
test temperature. In four cases this was the lower frequency,
and, in two cases, the higher.
F. Stimulus amplitude
Fish were given a choice between continuous tonal
stimuli that differed by 3 dB ??0.5 dB?. Table IV shows
results for three tests demonstrating preferential approach to
the more intense of the two signals. Although speaker output
and the steep drop in sound intensity with distance limited
the range of levels that could be presented, this preference
was apparent over a 6 dB range at 90 Hz as well as at 100 Hz
and thus was clearly based on amplitude.
G. Frequency modulation
In order to test the importance of frequency constancy
?which is also an indicator of fine temporal waveform peri-
odicity? to signal effectiveness, we compared attraction to
constant frequency and frequency-modulated tones ?Table
V?. FM signals consisted of 100 Hz modulated ?5 or ?10
Hz with a modulation rate of 10 Hz. In two-choice tests, fish
preferred a 100-Hz tone to the 100?10-Hz FM stimulus
?Table V, test A; p?0.038?, although, in separate tests, four
fish presented with just the ?10-Hz FM stimulus approached
the speaker. When the FM stimulus was modulated only ?5
Hz, no preference for the tone or FM was observed ?Table V,
H. Beat stimuli
When two concurrent tones, or hums, of slightly differ-
ent frequencies summate, constructive and destructive inter-
ference results in beats at their difference frequency ?dF?.
Because female midshipman approaching a group of hum-
ming males will often encounter beats in the summated
acoustic waveform ?see Bodnar and Bass, 1997?, we used
TABLE IV. Results of two-choice tests with tonal stimuli that differed by 3 dB. Response values are numbers
of individual fish approaching each stimulus.
?dB re: 1 ?Pa?
133 and 136
131 and 134
128 and 131
TABLE V. Results of two-choice tests between continuously frequency-
modulated ?FM? and constant frequency tones. Response values are num-
bers of individual fish approaching each stimulus. Testing conducted at
14.5–15.1 °C ?test A? and 15.1–15.3 °C ?test B?.
FM 100 Hz
A 100 Hz?10 Hz
at 10 Hz 412 16 0.038
B 100 Hz?5 Hz
at 10 Hz
148 22 0.143
35273527 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
two-choice tests to assess what impact beats might have on
hum attractiveness. First we compared the attractiveness of a
beat waveform played from one speaker ?single-source beat?
with a single tone from a second speaker. Then we compared
the attractiveness of a single tone with or without beat inter-
ference from an adjacent speaker ?multi-source beat?.
Single-source beat stimuli were generated by adding two
tones, F1 and F2 ?equal to F1?2 or ?5 Hz?, together ?Fig.
3?. Tests paired the beat stimulus played out of one speaker
with a single tone stimulus, F1, played out of the second
speaker. Depending on the test, beat stimuli were either
equal in intensity ?averaged over a complete beat cycle? to
the tone alternative or equivalent to the tone intensity plus
the F2 intensity. In the latter case, this meant beat stimuli
were up to 3 dB more intense than the tone stimulus. The
depth of the beat modulation was varied by changing the
relative contribution of F2 to the beat stimulus.
Table VI summarizes the results of the two-choice beat
tests. When fish were given a choice between F1 and an
equal intensity 5-Hz beat signal composed of F2 equal in
amplitude to F1 ?maximum beat depth?, 11 of 13 fish ?p
?0.011? went to the nonbeat signal ?Table VI, test A?. When
the depth of beating was reduced, by reducing the relative
amplitude of the F2 component to one third that of F1,
while keeping total signal energy the same ?test B?, 10 of 11
(p?0.006) fish approached the nonbeat stimulus. In tests
C–F, the beat F1 component was not attenuated, resulting in
beat stimuli with higher average intensity than the single
tone stimulus. With F1 equal in amplitude to F2?Test C?, 10
of 10 fish preferred the nonbeat, lower intensity stimulus.
?This one test was conducted with speaker arrangement C in
Fig. 2?. When F2 was reduced to half the amplitude of F1,
producing shallower envelope modulation ?test D?, fish still
tended to approach the nonbeat stimulus ?8 of 10, p?0.055?.
However, there was no preference when F2 was only one
quarter the amplitude of F1 ?test E?. Likewise, when F2 was
equal to the amplitude of F1 but the dF was only 2 Hz ?test
F?, there was no detectable preference.
Unlike the single-source beat stimuli presented above,
naturally occurring beat waveforms result from interference
between separate sources, which are themselves nonbeating.
In order to separate out the attractiveness of beat stimuli at a
distance from acceptability at the source, fish were tested
with multi-source beat stimuli. This experiment used two
pairs of speakers ?Fig. 2, configuration B? with one speaker
in each pair playing the test tone and the neighboring speaker
either playing a second tone, creating a multi-source beat, or
silent. When the test tone, 90, 95, or 100 Hz ?chosen to be
attractive at test temperature? was paired with a single tone
‘‘neighbor’’ of equal amplitude and ?5 Hz from the test
tone, there was no detectable preference among 12 fish for
the tone or beat signal ?Table VII, tests A–C?. Likewise,
with a multi-source beat consisting of a 90-Hz test tone plus
a 92-Hz neighbor ?test D?, fish went equally to the tone and
beat speakers. When fish approached the beat speakers dur-
ing these tests, they responded to either or both of the tones.
TABLE VI. Results of two-choice playback tests comparing approaches to single tone and two-tone beat stimuli. Beat stimulus tones ?F1 and F2? were
combined and presented out of a single speaker.
3528 3528 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
A. Playback and acoustic communication
This study has examined attraction by females and type
I and type II males to synthetic type I male calls in the
plainfin midshipman fish. By isolating acoustic signals from
the potentially complex behavioral milieu, playback allows
more direct comparison between the physical properties of a
sound and the receiver’s response. Whereas playback has
been broadly exploited in terrestrial vertebrates ?e.g., reviews
by Gerhardt, 1988; Searcy, 1992; Hauser, 1996?, the study of
midshipman reported here is the first we are aware of in fish
that uses two-choice presentations to test the relative attrac-
tiveness of signals that vary in precisely specified param-
This study has used one-choice tests of call recognition
and two-choice tests of preferences between simultaneously
presented variants to elucidate several acoustic dimensions
important to female midshipman attraction to male hums.
Recognition and preference are likely to be measures of the
same underlying sensory and neural processing ?Ryan and
Rand, 1993?. A call may evoke a response ?recognition? over
a range of values for an acoustic parameter ?e.g., frequency?;
and, within that range, certain values may be chosen over
B. Fish phonotaxis
Historically, much more has been known about the pro-
duction of sounds by fish than about their communicatory
significance ?reviews by Fine et al., 1977; Myrberg, 1981;
Ladich, 1997?. A small number of playback experiments
with vocal fish have been used to test call function, species
specificity of responses, and behavioral relevance of call pa-
rameters; however, only a very few experiments have tested
differential phonotaxis. Experiments with sunfish ?Gerald,
1971?, and cyprinids ?Delco, 1960? showed greater attraction
to conspecific than to heterospecific calls. Likewise, in bi-
color damselfish, a conspecific call was more effective than
heterospecific calls in facilitating courtship by males ?Myr-
berg and Spires, 1972? and more effective than heterospecific
calls or noise in attracting females ?Myrberg et al., 1986?.
Tavolga ?1958? found that female and male gobies, Bathygo-
bius soporator, increased activity or approached the trans-
ducer, respectively, in response to playbacks of male court-
ship sounds. Tavolga then presented, one at a time, synthetic
call variants and concluded that the fish were not very selec-
tive as long as stimulus fundamental frequency, pulse dura-
tion and pulse repetition rate fell within certain ranges. By
playing back the calls of a large and a small male bicolor
damselfish near the natural territories of several females,
Myrberg et al. ?1986? demonstrated preferential attraction to
the larger male’s call. The larger male’s call was lower in
frequency ?see also Myrberg et al., 1993?, although it is pos-
sible that temporal differences between these two natural,
amplitude-modulated signals could have influenced female
An Atlantic relative of the midshipman, the oyster toad-
fish, Opsanus tau, also has nesting males that produce adver-
tisement calls. These ‘‘boatwhistle’’ calls are generally less
than 500 ms in duration and have fundamental frequencies
similar to midshipman signals, though warmer ambient tem-
peratures tend to make boatwhistle frequencies higher ?Fine,
1978; Bass and Baker, 1991; Barimo and Fine, 1998?. Most
playback experiments with toadfish have been directed at
calling males, which will increase or decrease their calling
rates in response to certain signals ?Winn, 1967, 1972; Fish,
1972?. Winn ?1972? showed that gravid female, and very
occasionally male, toadfish approach speakers playing back
pre-recorded boatwhistles. Only one signal was used, so no
conclusions about call preference could be made.
Ibara et al. ?1983? tested plainfin midshipman responses
to recorded hums and pure tones. Gravid, but not spent, fe-
males were found to become more active in response to in-
dividually presented tones between 80 and 115 Hz, and five
fish for which data were collected were most responsive to
frequencies around 95 Hz ?playback temperatures were not
given?. Midshipman were apparently unable to localize
sound sources in the small test tanks, but in additional tests
in a larger tank, fish approached the active speaker. Again,
fish were not given a choice between signals; and only
stimulus frequency was varied.
C. Phonotaxis and function
Aquarium observations of midshipman acoustic and
spawning behavior ?Brantley and Bass, 1994? and both the
present and previous ?Ibara et al., 1983? playback studies
strongly support the hypothesis that the midshipman hum
functions as a mate call. Gravid females show the most ro-
bust and consistent phonotactic responses and are clearly
able to localize sound sources even in a non-ideal acoustic
environment, such as the shallow tank used in this study.
TABLE VII. Results for two-choice tests with single tone versus multi-source beat stimuli. For beat stimuli,
equal intensity tones with dF’s of 5 or 2 Hz were projected from adjacent speakers. Response numbers indicate
individual fish approaching each stimulus.
90 & 95
90 & 95
100 & 105
Totals for 5 Hz dF
1090 & 92
aTest temperatures ?°C?: A, 12.7–14; B, 14; C, 14.2; D, 14.4–14.8.
35293529J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
The observation that both type I and type II males will
sometimes approach a loudspeaker playing humlike signals
suggests that the hum has a general meaning of ‘‘nesting
male present,’’ and probably plays a role in male–male in-
teractions as well. As noted above, in the related oyster toad-
fish, playback of male calls influenced the calling behavior
of nearby males ?Winn, 1967; Fish, 1972? as well as attract-
ing gravid females ?Winn, 1972?. Although all three mid-
shipman morphs showed some degree of phonotaxis to
hums, they differed in the motor patterns elicited. This im-
plies that a neural mechanism couples auditory input to the
evocation of sex- and morph-specific reproductive behaviors
?as described by Brantley and Bass, 1994?.
D. Spectral and temporal waveform selectivity
Female midshipman were selective with regard to the
type of sound they found attractive. Pure tones or tones with
harmonics evoked phonotaxis, but noise with comparable en-
ergy at relevant frequencies was never attractive. Thus, au-
dible sound is not sufficient for mate recognition. Noise, by
definition, lacks periodicity in its fine waveform structure,
and the periodicity of the hum waveform is one of its most
salient characters ?Fig. 1B?. Hums can be remarkably con-
stant in frequency ?Bass et al., in press?, and recordings in
the midshipman peripheral auditory system have shown that
inner ear afferents can precisely synchronize ?i.e., phase
lock? to the periodicity of tones over the range of frequencies
that attract gravid females ?McKibben, 1998?.
To investigate the importance of fine temporal wave-
form periodicity to female preference, fish were given a
choice between constant frequency and frequency-modulated
signals. Continuous frequency modulation produced an ap-
proximately sinusoidal signal with varying intervals between
successive peaks in the waveform. While a pure tone was
preferred over a ?10-Hz modulated signal, the FM signal
was attractive to at least some fish, and four fish presented
with the FM stimulus alone approached the speaker. Fish
went equally to the tone and FM signal when the modulation
was only ?5 Hz. Thus, hum recognition does not require
strict frequency constancy, although increasing degrees of
modulation are likely to make a signal less attractive. A con-
founding aspect of frequency modulation in this system is
the relationship between temperature and frequency prefer-
ence. Although frequency variation over 20 Hz does not ex-
ceed the demonstrated range of recognition at a given tem-
perature ?Fig. 6?, a preference for lower or higher
frequencies could affect the relative attractiveness of the FM
signal depending on temperature and apart from modulation
Hums generally last on the order of minutes ?Ibara et al.,
1983; Brantley and Bass, 1994?, but we observed females
responding to trains of stimuli as short as 500–750 ms ?Fig.
5?. Although much shorter than most hums, 500 ms is con-
siderably longer than a grunt ?Brantley and Bass, 1994?. In-
deed, duration may be the critical difference between calls
perceived as hums and as grunts. ?Fine et al. ?1977? review
both graded and categorical differences in duration in the call
repertoires of several fish species.?
Over the range of pulsed stimuli tested with midshipman
there was a positive correlation between the duty cycle ?per-
cent of time the sound was on? and the attractiveness of the
stimulus ?Fig. 5C?. However, in the two cases where stimuli
with different pulse durations had the same duty cycle ?Fig.
5, stimuli a and b, c and e?, the stimulus with longer pulse
durations attracted a higher percentage of fish ?Fig. 5B?. This
suggests that the duration component of hum recognition
cannot be solely a matter of temporal integration ?over hun-
dreds of milliseconds? but depends on the duration of indi-
vidual sound pulses. The threshold duration for hum recog-
nition appears to be around 0.5–1 s. In two-choice tests,
females preferred a continuous tone over a 1 s pulsed stimu-
lus. This result, along with the observation that female search
behavior usually ceases when a hum stops, suggests that fe-
males would be more likely to locate and approach a con-
tinuously humming male.
The hum fundamental frequency alone is sufficient to
attract female midshipman, and adding the second harmonic
had no effect on stimulus preference. The midshipman’s pe-
ripheral auditory system can encode frequencies up to at
least the third harmonic ?around 300 Hz; McKibben, 1998?,
so harmonics could have some behavioral relevance. For ex-
ample, harmonics could increase the detectability of a near-
threshold hum, especially given constraints on sound trans-
mission presented by the physical environment. Midshipman
often nest in the intertidal zone where the water depth is
variable but often considerably less than the 1/4 to 1/2 wave-
length ?15 m for a 100 Hz signal? necessary for efficient
propagation of the fundamental ?Forrest et al., 1993?. In such
a constricted space the harmonics’ shorter wavelengths may
be able to propagate further than the fundamental ?Fine and
G. Frequency preference and temperature coupling
In this study, gravid female midshipman approached
tones with fundamental frequencies of 80 to 140 Hz. This is
greater than the 80–115-Hz range found by Ibara et al.
?1983?, and the full range of attractive frequencies, especially
as temperature is varied, is presumably greater. The funda-
mental frequency of a male’s hum is a function of muscle
contraction rate, determined by the output of a pacemaker-
motoneuron circuit in the hindbrain ?Cohen and Winn, 1967;
Bass and Baker, 1990?. The frequency of this rhythmic out-
put and, hence, of the hum, increases linearly with ambient
temperature ?Bass and Baker, 1991; Brantley and Bass,
The two-choice experiments reported here show that
midshipman are capable of discriminating and choosing be-
tween two signals differing by 10 or 20 Hz. Although these
tests used tones well within the frequency range of midship-
man calls, analysis of field recordings has shown that the
differences in hum fundamental frequency that a female is
3530 3530 J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
likely to encounter at any one temperature are less than 10
Hz ?Bodnar and Bass, 1997?. Thus, the question of whether
midshipman can discriminate frequency differences that
could influence choice between humming males remains
open. In psychophysical tests with tones around 100 Hz,
goldfish are able to detect frequency differences down to
approximately 5 Hz ?Fay, 1970?.
Female frequency preference was temperature depen-
dent and in accord with the temperature dependence of hum
frequency. In two-choice tests with tones differing by 10 or
20 Hz, fish preferred the frequency closest to the expected
hum frequency at the test temperature. However, when fish
were given a choice between 90 and 110 Hz at intermediate
temperatures ?15.6–17.5 °C? where the expected hum fre-
quency would be around 100 Hz, they showed no preference.
These results suggest that at a given temperature there will
be a most attractive frequency, with attractiveness falling off
as the frequency is either increased or decreased. Similar
temperature coupling of male calling song and female re-
sponse preferences has been described in crickets ?Pires and
Hoy, 1992? and frogs ?e.g., Gerhardt and Doherty, 1988? and
has been linked to underlying sensory processes ?Brenowitz
et al., 1985?. Temperature dependence of call characteristics
is common in poikilothermic species, and temperature-
dependent reception presumably can enhance the sensitivity
or accuracy of recognition of conspecific calls.
higher intensity of two tones that differed by just 3 dB. Us-
ing our experimental layout, the responding fish were pre-
sumably able to hear both signals and choose the higher
amplitude. Since both sources were equi-distant from the re-
lease site, this response could reflect assessment either of
source intensity or of relative intensity at the fish’s location
?see Forrest and Raspet, 1994?. Three dB sensitivity is better
than the 3.7–5.2 dB intensity discrimination thresholds ob-
tained for 110-Hz pulsed tones in conditioning experiments
with cod and haddock ?Chapman and Johnstone, 1974?. Like
the midshipman, these species lack specialized connections
between the swimbladder and inner ear. Goldfish, known as
‘‘hearing specialists’’ for just such an enhanced transmission
pathway, are sensitive to differences in pulsed tone ampli-
tude of only 2.1 dB at 100 Hz ?Fay, 1989?. We did not test
the limits of midshipman discrimination, but the demonstra-
tion of a preference based on a 3 dB difference suggests that
midshipman sensitivity is likely to be comparable to that of
goldfish and to other vertebrates ?Fay, 1989?. It remains to be
seen whether there are true interspecies difference in fish
intensity discrimination abilities or just differences in experi-
I. Concurrent hums and beat stimuli
Perhaps the most striking aspect of the hum signal is its
essential lack of gross temporal structure. Whereas acoustic
communication signals typically have some form of distinct
envelope modulation that contributes to the signal’s behav-
ioral effectiveness ?e.g., Winn, 1972; Myrberg et al., 1978;
Van Tasell et al., 1987; Gerhardt, 1988; Crawford et al.,
1997; Penna, 1997?, the midshipman hum is effectively con-
tinuous and flat in amplitude. However, females commonly
encounter acoustic waveforms that are more complex than
those of individual hums. Neighboring males often hum si-
multaneously, and small differences in their fundamental fre-
quencies ??8 Hz? produce beats in the summated acoustic
waveform ?Bodnar and Bass, 1997?.
In both the two-choice, frequency preference tests ?Fig.
7? and the beat stimulus experiments ?Tables VI and VII?
reported here, fish were able to discriminate and localize
sounds in the presence of beat waveforms. For the frequency
tests, the concurrent tones originated from separate speakers
and differed by 10 or 20 Hz. Thus, their interference pro-
duced beats at 10 or 20 Hz. Just as these overlapping signals
did not effectively jam each other and eliminate phonotaxis,
fish were also able to localize and respond to individual
sources in tests with single- or multi-source beat stimuli with
dF’s of 2 or 5 Hz. Thus, female preference for the hums of
individual males must be based on auditory processing that
can extract overlapping hums ?Bodnar and Bass, 1997?.
Fish preferred pure tones over certain single-source beat
stimuli. Both beat frequency ?dF? and depth of beating ap-
pear to be limiting factors for stimulus attractiveness. In two-
choice tests, fish rejected 5-Hz dF, 100% modulated ?F1 and
F2 equal amplitude? beat stimuli in favor of F1 presented
alone ?Table VI, tests A, C?. When F2 was attenuated, re-
ducing the depth of modulation, this preference gradually
disappeared ?Table VI, tests D, E?. Fish did not prefer tone
over beat stimuli when the dF was only 2 Hz ?Table VI, test
Beat stimuli differ from single tones in their envelope
modulation and in their spectrum and fine waveform struc-
ture. Any of these differences could have reduced the attrac-
tiveness of certain beat stimuli relative to a tone. As the
frequency components of rejected beat stimuli differed by
only 5 Hz, and FM stimuli that varied by ?5 Hz ?a 10-Hz
range? were not distinguished from tones, envelope modula-
tion may be the critical factor.
When we presented females with a choice between a
beat waveform created by the summation of single tones
from adjacent speakers and a tone stimulus, no preference
was apparent ?Table VII?. Thus, beats did not affect stimulus
attractiveness as long as at their sources they resolved into
individual tones. Multi-source beats more closely mimic
natural beats resulting from the overlap of hums from adja-
Comparing responses to single- and multi-source beats
suggests strategies females may use to accomplish the real
world listening task, extracting individual calls from sum-
mated waveforms. Single- and multi-source beats are likely
to be perceived similarly at some distance from the speakers.
If this is the case, the differences in response could reflect
differential perception based on spatial sampling. The appar-
ent depth of modulation for multi-source beats is dependent
upon the relative amplitudes and distances to the two signals.
Thus, not only can the effective magnitude of the beating
change, and become smaller, with position for a multi-source
beat, but, by attending to beats, receivers could gauge their
35313531J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
relative proximity to an individual source. Given the re-
sponse criteria applied in this study, it is also possible that
rejection of single-source beat stimuli occurs only after the
fish has approached the source. Indeed, fish were sometimes
observed to orient toward and begin to approach a speaker
playing a strongly beating waveform and then to swim away,
implying recognition of the summated waveform that oper-
ates at a distance but not at the source.
Alternatively, although the two speakers used here to
create the multi-source beat were side by side, it is possible
that, due to directional selectivity of sound transducing hair
cells in the fish ear, the angular separation of the sources was
sufficient to diminish the effective interference ?review of
directional masking data in Fay, 1988; Fay and Edds-
Walton, 1997?. In effect, each tone would be processed by a
separate channel, and beating would not be a problem.
IV. SUMMARY AND CONCLUSIONS
First, this study supports the conclusion of Ibara et al.
?1983? that the plainfin midshipman hum functions as a mate
call, and extends that finding by describing the phonotactic
response of both male morphs as well as of gravid females.
Second, results from one and two-choice tests show that fe-
male midshipman fish respond selectively to audible sounds
and are capable of differentiating and choosing between
acoustic signals that differ in duration, frequency, amplitude,
and spectral/temporal content. Although it is most straight-
forward to examine preference functions along one dimen-
sion, in reality there are probably interactions between vari-
ous signal parameters such that they create a multi-
dimensional preference space ?Doherty, 1985; Ryan and
Rand, 1993; Forrest and Raspet, 1994?. This study has not
directly addressed the influences of one parameter on an-
other, but it is apparent, for example, that, although fish can
discriminate small amplitude differences, they do not always
prefer the more intense signal.
In the absence of any data correlating hum characteris-
tics with male or nest quality, we do not know whether fe-
males would benefit from acoustic choosiness beyond that
necessary to locate a conspecific male. However, females do
make a large investment in one clutch of eggs that is appar-
ently deposited in a single nest ?DeMartini, 1988; Brantley
and Bass, 1994?, and search behavior could be costly. Thus,
selection would likely have favored acoustic discrimination
that enabled females to extract any available information
about mate or nest quality from male hums.
The midshipman’s simple acoustic signals and unam-
biguous phonotactic responses make it a promising system
for understanding the receiver mechanisms of communica-
tion, both in terms of behavioral decisions and the underly-
ing neural coding.
We would like to thank Margaret Marchaterre and
Deana Bodnar for logistical support and advice during this
study, and Deana Bodnar, Christopher Clark, Ronald Hoy,
and the reviewers for their comments on the manuscript.
This research was supported by a training grant from NIMH
?5T32GM07469?. a Clare Booth Luce Fellowship, and NIH
Grant No. DC-00092.
Barimo, J. F., and Fine, M. L. ?1998?. ‘‘Relationship of swim-bladder shape
to the directionality pattern of underwater sound in the oyster toadfish,’’
Can. J. Zool. 76, 134–143.
Bass, A. H. ?1992?. ‘‘Dimorphic male brains and alternative reproductive
tactics in a vocalizing fish,’’ Trends Neurosci. 15, 139–145.
Bass, A. H. ?1996?. ‘‘Shaping brain sexuality,’’ Am. Sci. 84, 352–363.
Bass, A. H., and Baker, R. ?1990?. ‘‘Sexual dimorphisms in the vocal con-
trol system of a teleost fish: Morphology of physiologically identified
neurons,’’ J. Neurobiol. 21, 1155–1168.
Bass, A. H., and Baker, R. ?1991?. ‘‘Evolution of homologous vocal control
traits,’’ Brain Behav. Evol. 38, 240–254.
Bass, A. H., Bodnar, D. A., and Marchaterre, M. A. ?in press?. ‘‘Comple-
mentary explanations for existing phenotypes in an acoustic communica-
tion system,’’ in Neural Mechanisms of Communication, edited by M.
Hauser and M. Konishi ?MIT, Cambridge?.
Bodnar, D. A. ?1996?. ‘‘The separate and combined effects of harmonic
structure, phase, and FM on female preferences in the barking treefrog
?Hyla gratiosa),’’ J. Comp. Physiol. A 178, 173–182.
Bodnar, D. A., and Bass, A. H. ?1997?. ‘‘Temporal coding of concurrent
acoustic signals in auditory midbrain,’’ J. Neurosci. 17, 7553–7564.
Brantley, R. K., and Bass, A. H. ?1994?. ‘‘Alternative male spawning tactics
and acoustic signals in the plainfin midshipman fish, Porichthys notatus
?Teleostei, Batrachoididae?,’’ Ethology 96, 213-232.
Brenowitz, E. A., Rose, G., and Capranica, R. R. ?1985?. ‘‘Neural correlates
of temperature coupling in the vocal communication system of the gray
treefrog ?Hyla versicolor?,’’ Brain Res. 359, 364–367.
Chapman, C. J., and Johnstone, A. D. F. ?1974?. ‘‘Some auditory discrimi-
nation experiments on marine fish,’’ J. Exp. Biol. 61, 521–528.
Cohen, M. J., and Winn, H. E. ?1967?. ‘‘Electrophysiological observations
on hearing and sound production in the fish, Poichthys notatus,’’ J. Exp.
Zool. 165, 355–370.
Crawford, J. D., Cook, A. P., and Heberlain, A. S. ?1997?. ‘‘Bioacoustic
behavior of African fishes ?Mormyridae?: Potential cures for species and
individual recognition in Polliymrus,’’ J. Acoust. Soc. Am. 102, 1200–
Delco, E. A., Jr. ?1960?. ‘‘Sound discrimination by males of two cyprinid
fishes,’’ Tex. J. Sci. 12, 48–54.
DeMartini, E. E. ?1988?. ‘‘Spawning success of the male plainfin midship-
man. I. Influences of male body size and area of spawning site,’’ J. Exp.
Mar. Biol. Ecol. 121, 177–192.
Doherty, J. A. ?1985?. ‘‘Temperature coupling and ‘trade-off’ phenomena in
the acoustic communication system of the cricket, Gryllus bimaculatus De
Geer ?Gryllidae?,’’ J. Exp. Biol. 114, 17–35.
Fay, R. R. ?1970?. ‘‘Auditory frequency discrimination in the goldfish ?Car-
assius auratus?,’’ J. Comp. Physiol. Psychol. 73, 175–180.
Fay, R. R. ?1988?. Hearing in Vertebrates: A Psychophysics Databook ?Hill-
Fay Associates, Winnetka, IL?.
Fay, R. R. ?1989?. ‘‘Intensity discrimination of pulsed tones by the goldfish
?Carassius auratus?,’’ J. Acoust. Soc. Am. 85, 500–502.
Fay, R. R., and Edds-Walton, P. L. ?1997?. ‘‘Directional response properties
of saccular afferents of the toadfish, Opsanus tau,’’ Hearing Res. 111,
Fine, M. L. ?1978?. ‘‘Seasonal and geographical variation of the mating call
of the Oyster Toadfish, Opsanus tau L.,’’ Oecologia 36, 45–57.
Fine, M. L., and Lenhardt, M. L. ?1983?. ‘‘Shallow-water propagation of the
toadfish mating call,’’ Comp. Biochem. Physiol. A 76, 225–231.
Fine, M. L., Winn, H. E., and Olla, B. L. ?1977?. ‘‘Communication in
fishes,’’ in How Animals Communicate, edited by T. A. Sebeok ?Indiana
U.P., Bloomington?, pp. 472–518.
Fish, J. F. ?1972?. ‘‘The effect of sound playback on the toadfish,’’ in Be-
havior of Marine Animals, Volume 2: Vertebrates, edited by H. E. Winn
and B. L. Olla ?Plenum, New York?, pp. 386–434.
Forrest, T. G., Miller, G. L., and Zagar, J. R. ?1993?. ‘‘Sound propagation in
shallow water: Implications for acoustic communication by aquatic ani-
mals,’’ Bioacoustics 4, 259–270.
Forrest, T. G., and Raspet, R. ?1994?. ‘‘Models of female choice in acoustic
communication,’’ Behav. Ecol. 5, 293–303.
Gerald, J. W. ?1971?. ‘‘Sound production during courtship in six species of
sunfish ?Centrarchidae?,’’ Evolution ?Lawrence, Kans.? 25, 75–87.
35323532J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998J. R. McKibben and A. H. Bass: Signal recognition in fish
Gerhardt, H. C. ?1988?. ‘‘Acoustic properties used in call recognition by
frogs and toads,’’ in The Evolution of the Amphibian Auditory System,
edited by B. Fritzsch, M. J. Ryan, W. Wilczynski, T. E. Hetherington, and
W. Walkowiak ?Wiley, New York?, pp. 455–483.
Gerhardt, H. C., and Doherty, J. A. ?1988?. ‘‘Acoustic communication in the
gray treefrog, Hyla versicolor: Evolutionary and neurobiological implica-
tions,’’ J. Comp. Physiol. A 162, 261–278.
Hauser, M. D. ?1996?. The Evolution of Communication ?MIT, Cambridge?.
Ibara, R. M., Penny, L. T., Ebeling, A. W., Van Dykhuizen, G., and Cailliet,
G. ?1983?. ‘‘The mating call of the plainfin midshipman fish, Porichthys
notatus,’’ in Predators and Prey in Fishes, edited by D. L. G. Noakes, D.
G. Lindquist, G. S. Helfman, and J. A. Ward ?Dr. W. Junk Publishers, The
Hague, The Netherlands?, pp. 205–212.
Ladich, F. ?1997?. ‘‘Agonistic behavior and significance of sounds in vocal-
izing fish,’’ Mar. Fresh. Behav. Physiol. 29, 87–108.
McKibben, J. R. ?1998?. ‘‘A neuroethological analysis of acoustic commu-
nication in the plainfin midshipman fish, Porichthys notatus,’’ Ph.D. the-
sis, Cornell University, Ithaca, NY.
Myrberg, A. A., Jr. ?1981?. ‘‘Sound communication and interception in
fishes,’’ in Hearing and Sound Communication in Fishes, edited by W. N.
Tavolga, A. N. Popper, and R. R. Fay ?Springer-Verlag, New York?, pp.
Myrberg, A. A., Jr., Ha, S. J., and Shamblott, M. J. ?1993?. ‘‘The sounds of
bicolor damselfish ?Pomacentrus partitus?: Predictors of body size and a
spectral basis for individual recognition and assessment,’’ J. Acoust. Soc.
Am. 94, 3067–3070.
Myrberg, A. A., Jr., Mohler, M., and Catala, J. C. ?1986?. ‘‘Sound produc-
tion by males of a coral reef fish ?Pomocentrus partitus?: Its significance
to females,’’ Anim. Behav. 34, 923–933.
Myrberg, A. A., Jr., and Spires, J. Y. ?1972?. ‘‘Sound discrimination by the
bicolor damselfish, Eupomacentrus partitus,’’ J. Exp. Biol. 57, 727–735.
Myrberg, A., Spanier, E., and Ha, S. ?1978?. ‘‘Temporal patterning in acous-
tic communication,’’ in Contrasts in Behavior, edited by E. Reese and F.
Lighter ?Wiley, New York?. pp. 137–179.
Penna, M. ?1997?. ‘‘Selectivity of evoked vocal responses in the time do-
main by frogs of the genus Batrachyla,’’ J. Herpetology 31, 202–217.
Pires, A., and Hoy, R. R. ?1992?. ‘‘Temperature coupling in cricket acoustic
communication: I. Field and laboratory studies of temperature effects on
calling song production and recognition in Gryllus firmus,’’ J. Comp.
Physiol. A 171, 69–78.
Popper, A. N., and Fay, R. R. ?1993?. ‘‘Sound detection and processing by
fish: Critical review and major research questions,’’ Brain Behav. Evol.
Ryan, M. J., and Rand, A. S. ?1993?. ‘‘Species recognition and sexual se-
lection as a unitary problem in animal communication,’’ Evolution
?Lawrence, Kans.? 47, 647–657.
Searcy, W. A. ?1992?. ‘‘Song repertoire and mate choice in birds,’’ Am.
Zool. 32, 71–80.
Tavolga, W. N. ?1958?. ‘‘The significance of underwater sounds produced
by males of the gobiid fish, Bathygobius soporator,’’ Physiol. Zool. 31,
Tavolga, W. N. ?1971?. ‘‘Sound production and detection,’’ in Fish Physi-
ology, edited by W. S. Hoar and D. J. Randall ?Academic, New York?,
Vol. 5, pp. 135–205.
Van Tasell, D. J., Soli, S. D., Kirby, V. M., and Widin, G. P. ?1987?.
‘‘Speech waveform envelope cues for consonant recognition,’’ J. Acoust.
Soc. Am. 82, 1152–1161.
Walker, H. J., and Rosenblatt, R. H. ?1988?. ‘‘Pacific toadfishes of the genus
Porichthys ?Batrachoididae? with descriptions of three new species,’’
Copeia 1988, 887–904.
Winn, H. E. ?1967?. ‘‘Vocal facilitation and the biological significance of
fish sounds,’’ in Marine Bioacoustics, edited by W. N. Tavolga ?Perga-
mon, Oxford?, Vol. 2, pp. 213–230.
Winn, H. E. ?1972?. ‘‘Acoustic discrimination by the toadfish with com-
ments on signal systems,’’ in Behavior of Marine Animals, Volume 2:
Vertebrates, edited by H. E. Winn and B. L. Olla ?Plenum, New York?,
3533 3533J. Acoust. Soc. Am., Vol. 104, No. 6, December 1998 J. R. McKibben and A. H. Bass: Signal recognition in fish