Behavioral Ecology Vol. 11 No. 1: 71-83
© 2000 International Society for Behavioral Ecology
The transmission of advertisement calls in Central American frogs
a Department of Zoology, University of Texas at Austin, Austin, TX 78712, USA b Smithsonian Tropical Research Institute, Balboa, Panama
Address correspondence to N. M. Kime. E-mail: nmkime{at}mail.utexas.edu .
Received 1 December 1998; accepted 2 July 1999.
 |
ABSTRACT |
|---|
 TOPFOOTNOTES ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Acoustic communication signals change over distance due to loss of amplitude and fidelity, and it is assumed that signal degradation influences the receiver's ability to detect and decode signals. The degree of degradation depends on the signal's structure and the environment through which it transmits. We broadcast the advertisement calls of 22 species of Central American frogs at two heights within forested and open environments in Panama. We recorded these calls at five distances from the source and estimated signal degradation with a cross-correlation analysis, a measure that combines the effects of decrement in signal amplitude and fidelity. Calls degraded less when broadcast higher above the ground compared to on the ground, and less in open habitat compared to forested habitat; there was an additional interaction between height and environment. Furthermore, calls with low dominant frequencies experienced less degradation than calls with high dominant frequencies. There was no evidence, however, that the calls of these frogs have evolved to maximize habitat-specific transmission.
Key words: animal communication, anuran, attenuation, degradation, environmental acoustics, frog.
|
INTRODUCTION |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Animals often use acoustic signals for communication over long distances (e.g., Brown, 1989
;
Langbauer et al., 1991;
Stebbins and Cohen, 1995). As
these signals travel from sender to receiver, they are altered in a manner
that can reduce the signal's ability to elicit a response from a receiver.
Signal amplitude usually decreases at a minimum of 6 dB per doubling of
distance due to spherical spreading, and additional or "excess"
attenuation results from absorption and scattering of sound waves by the air,
ground, and surrounding vegetation (Wiley
and Richards, 1978). Additionally, the signal's temporal and
spectral fidelity is disrupted by frequency-dependent attenuation,
reverberations, and irregular amplitude fluctuations
(Michelsen and Larsen, 1983;
Richards and Wiley, 1980;
Wiley and Richards, 1978). Changes in both signal amplitude and fidelity over distance can degrade signal efficacy. Sufficient loss of amplitude can decrease signal-to-noise ratio to a level at which the receiver cannot detect the signal. Sufficient change in signal fidelity can make the signal unrecognizable to the receiver. There can also be important interaction effects; a certain decrease in fidelity might result in an effective signal at a high but not a low signal-to-noise ratio. Many studies partition the effects of attenuation and fidelity (the latter often referred to as degradation). We address degradation of signal efficacy over distance, which is dependent on changes in both amplitude and fidelity. Thus we analyze their combined effects in our measure of signal degradation. Throughout, we use the term "degradation" to refer to this combined effect of signal attenuation and loss of fidelity.
Signal degradation can be influenced by spectral and temporal
characteristics of the signal, broadcast height, characteristics of the
habitat through which the sound propagates, background noise, and
meteorological conditions (Bradbury and
Vehrencamp, 1998; Wiley and
Richards, 1978). The effect of each of these factors on sound
attenuation and degradation has been examined using transmission experiments.
Some generalizations can be drawn from these studies. Attenuation and
degradation are affected by both the height from which the signal is broadcast
or received (Brenowitz et al.,
1984; Dabelsteen et al.,
1993; Henwood and Fabrick,
1979; Marten and Marler,
1977; Mathevon et al.,
1996; Waser and Waser,
1977) and the environment through which the signal travels
(Brown and Gomez, 1992;
Marten and Marler, 1977;
Morton, 1975;
Ryan et al., 1990;
Waser and Brown, 1986).
Not all effects on signal degradation are external to the signal; various
sounds are more or less susceptible to degradation. For example, signal
degradation can vary with dominant frequency
(Brown and Gomez, 1992;
Marten and Marler, 1977;
Marten et al., 1977;
Morton, 1975;
Ryan, 1986;
Waser and Brown, 1986;
Waser and Waser, 1977) or
temporal characteristics (e.g., Mathevon
et al., 1996; Ryan and
Sullivan, 1989). Furthermore, the effects of call structure on
degradation can differ among environments. Thus, for long distance
communication, selection should favor signal characteristics that degrade less
in their local transmission habitat (e.g.,
Endler, 1992;
Morton, 1975;
Ryan and Brenowitz, 1985;
Sorjonen, 1986;
Wiley, 1991).
Differences in call structure among species that call in different
environments were first identified in comparative studies of bird song
morphology (Bowman, 1979;
Hunter and Krebs, 1979;
Gish and Morton, 1981;
Morton, 1975;
Sorjonen, 1986;
Wiley, 1991). In these
studies, the observed differences in song structure supported predictions
based on the acoustics of the local signaling habitat. A few studies in birds,
primates, and frogs have also demonstrated differences within a species
between calls that are used for long- and short-distance communication
(Brown and Gomez, 1992;
Ryan, 1986;
Sorjonen, 1983;
Waser and Waser, 1977). In
contrast, some recent studies on anurans have failed to find the predicted
differences in call structure among species that call in different
environments (e.g., Penna and Solis,
1998; but see Ryan et al.,
1990). In addition, some authors suggest that factors such as body
size, phylogenetic constraints, and background noise can also contribute to
habitat-specific differences in signal structure
(Ryan and Brenowitz, 1985;
Zimmerman, 1983).
In this study, we measured advertisement call degradation for 22 species of
frogs. The advertisement call is usually a loud, conspicuous signal used by
males to advertise their presence to other males and to attract females over
long distances. Only a few studies have investigated the role of habitat
acoustics in the evolution of frog call structure
(Penna and Solis, 1998;
Ryan, 1986;
Ryan et al., 1990;
Zimmerman, 1983). We first
asked to what extent transmission height and habitat type influence the amount
of call degradation across species. We then investigated how the dominant
frequency of advertisement calls influences the amount of degradation a call
experiences. Finally, we tested the hypothesis that advertisement calls have
evolved in response to selection generated by the acoustic properties of a
species' calling habitat.
|
METHODS |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Transmission experiments
Exemplars of the advertisement calls of 22 species of Panamanian frogs, including representatives from 11 genera, were selected from the call libraries of A. S. Rand and M. J. Ryan (Figure 1, Figure 1). Several calls were selected from each species and copied onto a single metal audio tape. All calls were recorded at the maximum amplitude without clipping. Thus, the relative peak amplitudes of the calls were similar; because the natural call amplitudes of most of these species are unknown, no attempt was made to scale them to their relative amplitudes in nature. We created two synthetic sounds using an Eico analog pulse generator (model 377) and transferred them to the same tape; these were a continuous tone with a fundamental frequency of 300 Hz and 10 harmonics (dominant frequency 900 Hz) and a 250-ms pulse with a dominant frequency of 1000 Hz (Figure 1, Figure 1). This master tape, which we used for all subsequent transmission studies, thus consisted of a sequence of different frog calls and sounds approximately 30 min in length.
|
|
M.J.R. conducted transmission experiments 3 June-27 July 1983 in Soberania National Park near Gamboa, Panama. All experiments were performed at night in one of two environments within the park, a field of tall grass or forest. The master tape was played from a Sony TCD 5M tape recorder through a 5-inch, full-range dynamic speaker packed in a plastic box and backed with a sound-absorbing medium. The sound pressure level (SPL; re. 20 µPa) of the pure tone stimulus measured at 50 cm in front of the speaker was adjusted to 90 dB SPL at the beginning of each experiment. Each broadcast was subsequently recorded at a distance from the source on one channel of metal tape using a Sennheiser ME80 microphone with a K3U power module and a second Sony TCD 5M recorder. Calls were broadcast and recorded at two heights within each of the two environments, at ground level and at 1.5 m above the ground, with both the speaker and the microphone at the same height. There were thus a total of four transmission locations. Recordings of the transmitted calls were made at distances of 1, 2, 3, 10, and 20 m from the broadcast source.
Cross-correlation analysis
We digitized recordings of transmitted calls at a sampling rate of 25 kHz
with Signal (Engineering Design, 1997). We then saved up to four calls from
each of the 22 frog species (n = 82 calls) for later analysis, using
both the sonogram and amplitude envelope plotted on the same time axis. The
same calls were systematically saved from each transmission tape so that
subsequent analyses could be conducted on identical calls across all distances
and locations. Only single calls were analyzed; for species that arrange their
calls in groups, we took four individual calls from a variety of groups. In
addition, we chose two 500-ms segments of the tone and two synthetic pulses
haphazardly from the digitized transmission recordings and saved them in
separate files. Some calls could not be discerned from the background noise at
10 and 20 m. We classified these as missing data for analysis. In addition, if
another species of frog or insect was calling loudly during the target call,
we also regarded it as missing data because the random presence of these
individuals interfered with subsequent data analysis.
We digitally band-pass filtered the individual calls to decrease the amount of noise outside the frequency range of the signal. For calls with energy in frequencies <625 Hz, we applied only a low-pass filter due to a mathematical constraint of digital filtering functions. Filter ranges varied from species to species; they were determined from visual evaluation of the sonogram and power spectrum of undegraded calls transmitted over a short distance (1 m) at a height of 1.5 m in an open environment.
We made sonograms of identical calls from each of the five transmission distances (10.2 ms time resolution, 97.7 Hz frequency resolution). If necessary, we additionally trimmed the sound files so that all five sonograms were approximately the same length and contained the same call segment from the master tape. We then calculated the amplitude-normalized covariances between the sonograms of two signals as a function of the time offset between them. Sonogram cross-correlations were performed over a frequency range of 200 Hz-10 kHz. Four cross-correlations were performed for each call exemplar within each of the transmission locations; these were between a call recorded at the 1 m transmission distance (a relatively undegraded control) and the same call recorded at a transmission distance of 2, 3, 10, or 20 m in the same location. The maximum cross-correlation coefficient from each comparison was saved for statistical analyses.
In a sonogram cross-correlation analysis, two stimuli are slid past one another along the time axis. The cross-correlation coefficient is recorded for each point in time. The maximum cross-correlation coefficient for two sonograms should thus occur at the point at which the two signals are time-aligned. Cross-correlation coefficients are interpreted in the same way as other correlation coefficients; they vary between -1 and +1, where -1 signifies antisimilarity, 0 signifies no similarity, and +1 indicates that the two signals are indistinguishable. Cross-correlation coefficients computed from comparisons of sonograms reflect differences in the spectral and temporal characteristics of the call, as well as differences in signal-to-noise ratio. They are therefore indicators of both attenuation (change in call amplitude) and loss of fidelity (change in call structure). These two types of signal change cannot be distinguished from one another using this technique, but both contribute to the degradation of signal efficacy over distance, which is the goal of our analysis. Thus we analyzed their combined effects in our measure of signal degradation.
There are two methodological limitations of the sonogram cross-correlation procedure. Cross-correlation coefficients are calculated from the entire range of the digitized sound. Consequently, the cross-correlation coefficient does not reflect changes in the call alone; any random disturbance present in one transmission but not another can affect the correlation. We partially addressed this problem by filtering sound outside the frequency range of each call, as well as by using several call exemplars from each species. Second, as noted above, cross-correlation coefficients reflect differences between the control and degraded signal with respect to both attenuation and fidelity. Cross-correlation coefficients decrease as signal-to-noise ratio decreases and as the fidelity of the signal's temporal and spectral structure decreases. The effects of these two signal changes on the cross-correlation coefficients are not equal, but their combined effect will influence the degradation of signal efficacy. Thus, this procedural limitation can be viewed as an advantage for the question we address: How does signal efficacy degrade with distance?
Statistical analyses
We used the mean values of the maximum cross-correlation coefficients for
each species to assess differences in call degradation between transmission
heights, between habitats, among species grouped by call dominant frequency,
and among species grouped by natural calling location. Each of the 22 species
was thus represented only once in each comparison. Separate analyses were
conducted for each of the four transmission distances. We used nonparametric
methods for all statistical tests, as unequal variances between groups and
non-normal data distributions violated the assumptions of analysis of
variance. In accordance with these nonparametric analyses, median values of
the cross-correlation coefficients from each location or group are reported as
indicators of the general tendencies within groups.
We also used a second method of analysis to investigate differences in call degradation among locations and species. For each species, we calculated the regression of cross-correlation coefficients plotted with respect to transmission distance within a single location. We used a linear regression as the simplest model, although degradation with distance was not linear for some of the species included in this study. In most (80 out of 95) cases, there was a significantly negative correlation between cross-correlation coefficients and transmission distance (Pearson product moment correlation, p <.05 for 12 cases, p <.005 for the remaining 68 cases). We used the slopes of those regressions for which the correlation was significant to estimate the rate of degradation with distance; highly negative slopes indicated high rates of degradation. Each species was again represented only once in the comparisons among locations and species, although the slopes were computed from all of the analyzed call exemplars from that species. We compared slopes among locations or species using nonparametric statistical tests.
|
RESULTS |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Call degradation with distance
We first examined changes in cross-correlation coefficients over distance in order to test the validity of the cross-correlation technique in measuring call degradation. Cross-correlation coefficients decreased with distance in all four transmission locations (Table 1, Figure 2). At a recording distance of 2 m, cross-correlation coefficients for all species were high. At the 10 and 20 m recording distances, however, cross-correlation coefficients diverged among species; some calls remained similar to the 1 m control, and others became highly masked or degraded. Sample variance generally increased over distance (Table 1, Figure 3).
|
|
|
We compared cross-correlation coefficients between all possible pairs of transmission distances; there were thus six comparisons within each of the four transmission locations. In all 24 comparisons, the median cross-correlation coefficient was higher for the shorter transmission distance (Table 1, Figure 2). The difference between distances was significant in 20 out of these 24 comparisons (Mann-Whitney, p <.05 for 6 comparisons, p <.005 for the remaining 14 comparisons). Although increasing degradation with distance is certainly not surprising, our results are critically important because they indicate that the cross-correlation coefficients used in the following analyses are a useful measure of call degradation.
Differences in call degradation among locations
We investigated the effects of transmission height and habitat type on call
degradation. Statistical analyses were conducted across all species and
included the synthetic tone and pulse. Removing the tone and pulse from these
analyses did not change the interpretation of any of the results. The
following comparisons thus provide indications of how transmission height and
environment affect signal degradation, without consideration of the structure
of the sound.
We compared cross-correlation coefficients between the two heights from which calls were broadcast and recorded. Analyses were conducted separately for each of the two transmission environments as well as for each of the four transmission distances; eight height comparisons were made. Calls generally experienced less degradation when transmitted at a height of 1.5 m than when transmitted at ground level, but the effect of height was more notable in the tall grass environment than in the forest environment. In tall grass, cross-correlation coefficients from the 1.5 m transmission height were significantly higher, or nearly so, than those from ground level transmission at the 3, 10, and 20 m transmission distances (Mann-Whitney, p <.001, p <.06, p <.005 for the 3, 10, and 20 m transmission distances, respectively; Table 1, Figure 2A,B). In the forest environment, median cross-correlation coefficients from the 1.5 m transmission height were again higher than those from ground level at the three longer transmission distances, but the differences between the heights were not significant in this environment (Mann-Whitney, p =.40, p =.07, and p =.11 for the 3, 10, and 20 m transmission distances, respectively; Table 1, Figure 2C,D). At the 2 m distance within the forest environment, cross-correlation coefficients from calls transmitted at ground level were significantly higher than those of calls transmitted at a height of 1.5 m (Mann-Whitney, p <.05).
Comparisons of the regressions of cross-correlation coefficients over distance complemented the above direct comparisons of cross-correlation coefficients. Calls transmitted at ground level experienced a greater rate of degradation than calls transmitted at a height of 1.5 m (Mann-Whitney, p <.001 in the tall grass environment, p <.05 in the forest environment; Table 2, Figure 3). As above, the effect of height was much more pronounced in tall grass; median slopes differed between the two heights in tall grass more than in forest (Table 2, Figure 3).
|
We next compared cross-correlation coefficients between the two environments through which calls were transmitted. Separate analyses were conducted for each of the two transmission heights, as well as for each of the four transmission distances. At ground level, calls transmitted through forest and tall grass fared equally well at three of the four transmission distances. There were significant differences between the two environments at only the 3 m transmission distance, where the cross-correlation coefficients of calls transmitted through the forest environment were higher than those of calls transmitted through tall grass (Mann-Whitney, p <.05, Table 1, Figure 2B,D). The influence of environment on call degradation was slightly more pronounced at a transmission height of 1.5 m. Above the ground, cross-correlation coefficients from the tall grass environment were significantly higher than cross-correlation coefficients from forest for both the 2 m and 20 m transmission distances (Mann-Whitney, p <.01 for the 2 m transmission distance, p <.005 for the 20 m transmission distance; Table 1, Figure 2A,C).
Comparisons of the regressions of cross-correlation coefficients over distance again complemented the direct comparisons of cross-correlation coefficients outlined above. The rate of call degradation did not significantly differ between the two environments when calls were transmitted at ground level (Mann-Whitney, p =.96; Table 2, Figure 3). At a transmission height of 1.5 m, however, calls transmitted through the tall grass environment exhibited a trend of lower rate of degradation than calls transmitted through the forest environment (Mann-Whitney, p =.06; Table 2, Figure 3).
Differences in call degradation among species
In a second set of analyses, we tested the hypothesis that advertisement
calls evolved in response to selection generated by habitat acoustics. We
first asked whether there were differences in degradation among the different
species. We then asked whether there were differences in degradation among
calls with different dominant frequencies. Finally, we asked whether these
differences were reflected in differences among species that normally call
from different heights and environments. The synthetic tone and pulse were not
included in any of these analyses. Separate analyses were conducted for each
of the four transmission locations; no comparisons were made between
transmission heights or environments.
We first evaluated differences among the 22 frog species. Cross-correlation
coefficients from all call exemplars from each species were included in this
analysis, rather than the species means used in all other comparisons. We
found highly significant differences among species at all four transmission
distances in all four locations (Kruskal-Wallis, p 
.001 in all 16
cases). The cross-correlation coefficients of some species ranked consistently
higher than the cross-correlation coefficients of other species, especially at
the 10 and 20 m transmission distances. Pleurodema brachyops (see
Figure 1,
Figure 1 for representative
sonograms) had consistently higher cross-correlation coefficients than most
other species in all four transmission locations. The cross-correlation
coefficients for both Physalaemus pustulosus and Hyla
foliomorta were higher than those of most other species when calls were
transmitted at ground level within both environments. Leptodactylus
pentadactylus fared better than most other species at both heights within
the tall grass environment. Leptodactylus fragilis had consistently
higher cross-correlation coefficients than many other species for transmission
at both heights within the forest environment, as well as within the tall
grass environment at a transmission height of 1.5 m.
The cross-correlation coefficients of some species were consistently lower than those of other species at the 10 and 20 m transmission distances. Colosthesus flotata and Hyla microcephala had lower cross-correlation coefficients than most other species in all four transmission locations. Both Scinax boulengeri and Scinax rubra had lower cross-correlation coefficients than most other species in the forest environment at a height of 1.5 m and in the tall grass environment at ground level. Colosthesus inguinalis had consistently lower cross-correlation coefficients than many other species in the tall grass environment at a transmission height of 1.5 m.
Species with calls of low dominant frequency (<2000 Hz, n = 11) were compared to species with calls of high dominant frequency (>2000 Hz, n = 10). The classifications for each species are summarized in Table 3. For ground level transmission, there were significant differences between the two groups of species at all four transmission distances within both environments (Mann-Whitney, p <.05 for three comparisons, p <.005 for the remaining five comparisons). In all cases, calls with low dominant frequencies experienced less degradation than calls with high dominant frequencies (Figure 4B,D). There were, in addition, significant differences between the two groups in the regression slopes of cross-correlation coefficients plotted with respect to distance (Mann-Whitney, p <.05 for forest, p <.005 for tall grass). In both environments, calls with dominant frequencies >2000 Hz had less negative slopes, and thus exhibited lower rates of degradation, than calls with dominant frequencies >2000 Hz.
|
|
For transmission above the ground, there were significant differences in degradation between calls with low and high dominant frequencies for only the longer transmission distances. In forest, calls with low dominant frequencies had higher cross-correlation coefficients than calls with high dominant frequencies only when transmitted over a distance of 20 m (Mann-Whitney, p <.05; Figure 4C). In the tall grass environment, calls with low dominant frequencies had significantly higher cross-correlation coefficients than calls with high dominant frequencies at the 10 and 20 m transmission distances (Mann-Whitney, p <.05 for the 10 m transmission distance, p <.001 for the 20 m transmission distance; Figure 4A). The regression slopes of the cross-correlation coefficients plotted with respect to distance differed between the two groups in both environments (Mann-Whitney, p <.05 for forest, p <.001 for tall grass). In both cases, calls of lower dominant frequency exhibited lower rates of call degradation than calls of higher dominant frequency.
Cross-correlation coefficients tended to decrease with increasing dominant frequency (Figure 5). For transmission at ground level in both environments, cross-correlation coefficients were significantly correlated with dominant frequency at all four transmission distances (Pearson product moment correlation, r >.6, p <.01 for all four transmission distances in both environments). The slopes of these regressions generally increased with transmission distance (Figure 5B,C). For transmission at a height of 1.5 m in tall grass, cross-correlation coefficients were significantly correlated with dominant frequency for the 3, 10, and 20 m distances (Pearson product moment correlation, r =.45, p <.05 for the 3 m transmission distance, r =.83, p <.001 for the 10 m transmission distance, r =.74, p <.001 for the 20 m transmission distance). Again, the slopes of the regressions generally increased with transmission distance (Figure 5A). For transmission at a height of 1.5 m above the ground through a forested environment, cross-correlation coefficients were in no case correlated with dominant frequency (Pearson product moment correlation, r <.45, p >.1 for all four distances).
|
Finally, the 22 frog species were categorized into three natural calling habitats (forest, n = 9; open, n = 6; stream, n = 5) and two calling heights (ground, n = 12; above ground, n = 9). We compared degradation among groups to test the hypothesis that frog calls have evolved in response to selection for decreased degradation within their home environment. Classifications for each species are summarized in Table 3. As in the above comparisons among species grouped by call dominant frequency, we compared the groups within a single transmission location.
There were no significant differences in degradation among species that normally call from the three different environments; this was true for all four transmission distances within all four transmission locations (Kruskal-Wallis, p >.2 for all 16 comparisons; Figure 6). In addition, species that normally call from different environments did not significantly differ in the regression slopes of cross-correlation coefficients plotted with respect to transmission distance (Kruskal-Wallis, p >.3 for all four locations). There were also no significant differences between species that normally call from the ground and those that normally call from above the ground at any transmission distance within any transmission location (Mann-Whitney, p >.1 for all 16 comparisons; Figure 7). There were no significant differences between these two groups in the slopes of the regression lines drawn through the cross-correlation coefficients over distance (Mann-Whitney, p >.3 for all four transmission locations).
|
|
|
DISCUSSION |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Our estimates of signal degradation using cross-correlation coefficients revealed that in 22 species of Central American frogs the amount of degradation differed between habitats and between transmission heights. Degradation also differed among species and among calls with different spectral characteristics. These differences were not specific to the species' local calling habitat; thus there is no evidence to support the hypothesis that the calls of these species have evolved in response to selection generated by habitat acoustics. We thus suggest that morphological constraints or selection in other contexts have been more important in determining frog call structure than the need for transmission efficiency.
The influence of height and environment on call degradation
We found that both transmission height and environment influenced call
degradation and that there was an interaction between the effects of height
and environment. Calls transmitted at ground level suffered greater
degradation than those transmitted above the ground. This difference was more
pronounced in tall grass than in forest. Calls fared equally well in tall
grass and forest when transmitted at ground level, but, above the ground,
calls experienced slightly lower rates of degradation in the tall grass
relative to the forest.
A number of factors, including scattering from dense vegetation, ground
attenuation, and temperature and wind gradients, can cause increased
attenuation and degradation at ground level
(Wiley and Richards, 1978).
The effect of caller height on signal transmission has been accordingly
demonstrated in a number of transmission studies (see Introduction). The
results of the current study are consistent with the results of these previous
studies.
Such simple generalizations cannot be made about the differences between
open and forested environments. The results of previous experiments have not
been consistent with respect to the overall influence of environment on call
degradation. In some transmission studies, signals broadcast through forested
environments experienced less attenuation or degradation than the same signals
transmitted through open environments
(Brown and Gomez, 1992;
Morton, 1975;
Waser and Brown, 1986). Others
found the opposite to be true: signals transmitted through open environments
experienced less attenuation or degradation than the same signals transmitted
through forest (Marten and Marler,
1977; Ryan et al.,
1990; Waser and Brown,
1986). In addition, Marten and Marler
(1977) showed that the
influence of environment on call attenuation was less than the influence of
height. Similarly, Wiley and Richards
(1978) suggested that the
effects of stationary heterogeneities from objects such as trees in a forest
could balance the effects of nonstationary heterogeneities caused by
micrometerological instability in open areas, and that this trade-off could
obscure differences in signal transmission between open and forested
environments. In light of these previous findings, it is not surprising that
we did not find larger differences in degradation between the two
environments.
We observed two related interactions between the effects of height and environment on call degradation. Caller height was more influential in the tall grass environment than in the forest environment. In addition, an effect of environment was evident above the ground but not at ground level. In general, calls transmitted through tall grass at a height of 1.5 m experienced less degradation than calls transmitted through any of the other broadcast locations. In our study area, the density of intervening vegetation decreased appreciably with increased height in the tall grass environment but not in the forest environment. This difference between the two environments in the distribution of vegetation is one likely cause of the observed interactions between height and environment. Calls transmitted above the ground in the tall grass environment were freed of much of the influence of intervening vegetation, likely resulting in lower levels of degradation in this location relative to the three locations in which vegetation was more dense.
The influence of dominant frequency on call degradation
Species with high and low frequency calls differed in the amount of
degradation they experienced. As in a number of previous transmission studies
(see Introduction), we found that call degradation generally increased with
increasing call dominant frequency. We also observed an interaction between
the influence of call dominant frequency and the effects of transmission
height and environment. At ground level, sound degradation increased with
dominant frequency at all transmission distances in both environments. Above
the ground, on the other hand, sound degradation was higher for calls with
high dominant frequencies at only the longer transmission distances. Molecular
absorption and scattering disproportionately affect higher frequencies at all
heights (Wiley and Richards,
1978), but the effect of scattering should be greatest near the
ground where vegetation is most dense. Our results are consistent with this
prediction. We also found that, for above-ground transmission, the difference
in the amount of degradation experienced by high and low frequency calls was
greater in the tall grass environment than in the forest environment. Above
the ground, irregular amplitude fluctuations caused by atmospheric turbulence
disproportionately affect higher frequencies
(Richards and Wiley, 1980). It
may be true that the influence of this environmental factor was stronger in
the tall grass environment than in forest.
In our sample, other call characters correlated with dominant frequency may
have influenced the results of the cross-correlation analysis. For example,
all four of the species with dominant frequencies lower than 1000 Hz also had
relatively long calls. However, because previous experimental results point to
a strong effect of frequency on signal transmission
(Brown and Gomez, 1992;
Marten and Marler, 1977;
Marten et al., 1977;
Morton, 1975;
Waser and Brown, 1986;
Waser and Waser, 1977), we
believe that our results correctly indicate an effect of dominant frequency on
call attenuation and degradation.
The effects of environmental selection on call degradation
We found no evidence to support the hypothesis that the advertisement calls
of the frogs under study evolved in response to selection to maximize
transmission distance within their local calling habitat. This was surprising
because it has been repeatedly suggested that the signals used for
long-distance communication are selected upon for increased transmission
efficiency within the habitat from which a species normally calls (e.g.,
Endler, 1992;
Morton, 1975;
Ryan and Brenowitz, 1985;
Sorjonen, 1986;
Wiley, 1991).
A number of previous studies, conducted mostly on bird song, showed that
structural differences in the calls of species from forested and open
environments match predictions based on transmission experiments conducted in
these environments (Bowman,
1979; Hunter and Krebs,
1979; Gish and Morton,
1981; Morton,
1975; Sorjonen,
1986; Wiley,
1991). It has been concluded that the observed differences in call
morphology were the result of evolution in response to environmental
selection. Studies of anurans, however, have tended not to support this
conclusion. One recent study failed to find predicted differences in call
morphology among species that broadcast calls through different environments
(Penna and Solis, 1998). In
addition, Zimmerman (1983)
showed that observed differences in call structure among Amazonian frogs could
be more readily explained by differences among species in body size or by
phylogenetic relationship than by differences in natural calling
environment.
There is evidence that some of the fine-scale call differences between
subspecies of cricket frogs, Acris crepitans crepitans and A. c.
blanchardi, might derive from an evolutionary response to selection
generated by habitat acoustics (Ryan et
al., 1990; Wilczynski and
Ryan, 1999). A. c. blanchardi resides in open habitat
that has little effect on call degradation. The forest environment of A.
c. crepitans has a more severe effect on calls. Both subspecies exhibit
little call degradation in the open habitat, with the call of A. c.
crepitans performing better than that of its sister subspecies; both
showed substantial degradation in the forest habitat. There was also an
interaction effect; the calls of A. c. crepitans experienced a less
severe increase in degradation between the open and forest habitats than the
calls of A. c. blanchardi. Thus the calls of the forest subspecies,
A. c. crepitans, appears to have evolved in response to selection for
transmission efficiency in forest and thus transmit with less degradation in
both environments.
We expected similar findings. The results of our transmission studies indicated that the strength of environmental selection is especially high for ground-level callers and that changes in call morphology as a result of this selection could result in increases in transmission efficiency for these species (see previous sections). We thus predicted that the advertisement calls of species that normally call from the ground have evolved in response to selection for more efficient transmission within this location. Furthermore, we expected that the resulting changes in call structure would be reflected in less call degradation, within the ground-level transmission locations, among ground-level callers than among species that normally call from above the ground.
Our failure to find differences in degradation among species partitioned by
calling locations can be attributed to a number of different factors,
including statistical sampling error, the relatively small active space of the
anuran mating signal, morphological constraints on call structure, or
selection on signal morphology in other contexts. One possibility is that our
failure to find significant differences among the groups of species may have
been due to a type II statistical error. Similarly, differences among species
might have been obscured by the coarseness of the cross-correlation technique,
or by the groups into which we classified species being too broad to detect
differences among species that call from more narrowly defined microhabitats
(as cautioned by Morton, 1975;
Zimmerman, 1983).
A second possibility is that frog advertisement calls simply may not need
to travel long distances. The distance over which anuran advertisement calls
can be heard varies greatly among species
(Loftus-Hills and Littlejohn,
1971; Stebbins and Cohen,
1995; Zimmerman,
1982), and it is possible that the calls of some of the species
included in this study have not been selected upon for transmission over the
distances we investigated. However, we observed significant differences in
degradation among individual species over distances as short as 2 m. Because
the calls of most of the anuran species included in this study probably must
travel at least this far to reach female receivers (personal observation), we
suggest that this alternative is not likely with respect to the current study.
Furthermore, in most species of anurans tested there is preference for calls
of greater amplitude (reviewed in Ryan and
Keddy-Hector, 1992). Decreased degradation might not only increase
the active space of the signal but also increase the amplitude and thus the
attractiveness of the signal at closer distances.
It is also possible that the morphology of the vocal apparatus constrains
call structure (Cocroft and Ryan,
1985) and that phylogenetic constraints on this physical trait can
limit call evolution in response to environmental selection
(Ryan and Brenowitz, 1985;
Zimmerman, 1983). Our results
and others (Brown and Gomez,
1992; Marten and Marler,
1977; Marten et al.,
1977; Morton,
1975; Waser and Brown,
1986; Waser and Waser,
1977) demonstrate that dominant frequency is one call
characteristic that may be selected upon in the context of habitat acoustics.
However, the dominant frequency of a frog's call is partially constrained by
its body size. Larger frogs generally call with lower dominant frequencies
than smaller frogs (Littlejohn,
1977; Martin,
1967). Frogs can lower dominant frequency independent of body size
by direct weighting of the vocal folds. Without a concomitant increase in body
size, however, there will be even greater impedance in coupling acoustic
energy to the environment. Many frogs use call wavelengths that are several
times their body length but exhibit very low energetic efficiencies for
calling (Ryan, 1988).
Advertisement call dominant frequency may thus be more constrained by the
morphology of the caller than by habitat acoustics.
One further possibility is that selection on call structure in the context
of species recognition or sexual selection has had a greater influence on call
morphology than selection for decreased call degradation. Evolution of call
structure in response to environmental selection would necessarily result in
some convergence of call structure among the different species occurring in a
given habitat. However, some studies have suggested acoustic niche
partitioning among species of frogs occurring in a single community
(Drewry and Rand, 1983;
Duellman and Pyles, 1983;
Hödl,
1977). Selection for divergence of signal structure, in order to
avoid acoustic interference among species, may have been more influential than
selection for decreased degradation in determining the structure of frog
calls.
Sexual selection can also influence advertisement call structure. Female
frogs show mating preferences for many characteristics of the frog's
advertisement call, and female mate choice with respect to these preferences
can exert selection upon male call morphology (e.g.,
Andersson, 1994;
Ryan, 1985). Sexual selection
and environmental selection may operate in the same direction (such as toward
lower dominant frequencies). If sexual selection opposes environmental
selection, however, it may be strong enough to balance or counteract the
effects of the latter on call structure.
We suggest that phylogenetic constraints on call morphology, or selection on call structure in other contexts, have been more influential in determining the structure of anuran advertisement calls than selection for increased transmission efficiency. Our experiments included a diverse mix of frog taxa. Further studies investigating the role of environmental selection on call morphology should focus on differences among the calls of closely related species or populations of a single species that have experienced abrupt habitat shifts (e.g., A. crepitans, as discussed above). Such studies would allow a more fine-scale analysis of differences in signal structure and call transmission than the current study. In addition, closely related groups should share some morphological and behavioral constraints. The differences in call morphology among these groups could be more effectively used to test hypotheses about the evolution of frog call structure in response to environmental selection.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
We thank Andre de la Reza for his extensive work on early call analysis and Kate Levering for discussions on the cross-correlation methodology. We also thank an anonymous reviewer for helpful comments on an earlier draft of the manuscript. The transmission experiments were supported by a National Geographic Award to M.J.R. Call analysis was supported in part by a National Research Science Award training grant awarded to N.M.K.
|
FOOTNOTES |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
W. R. Turner is now at the Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, USA.
|
REFERENCES |
|---|
|
TOPFOOTNOTESABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Andersson M, 1994. Sexual selection. Princeton, New Jersey: Princeton University Press.
Bowman RI, 1979. Adaptive morphology of song dialects in Darwin's finches. J Ornithol 120:353-389.
Bradbury JW, Vehrencamp SL, 1998. Principles of animal communication. Sunderland, Massachusetts: Sinauer Associates.
Brenowitz EA, Wilczynski W, Zakon HH, 1984. Acoustic communication in spring peepers. Environmental and behavioral aspects.J Comp Physiol A 155:585-592.
Brown CH, 1989. The active space of blue monkey and grey-cheeked mangabey vocalizations. Anim Behav 37:1023-1034.
Brown CH, Gomez R, 1992. Functional design features in primate vocal signals: the acoustic habitat and sound distortion. In:Topics in primatology, vol. 1, Human origins (Nishida T, McGrew WC, Marler P, Pickford M, deWaal F, eds). Tokyo: Tokyo University Press;177 -198.
Cocroft RB, Ryan MJ, 1995. Patterns of advertisement call evolution in toads and chorus frogs. Anim Behav 49:283-303.
Dabelsteen T, Larsen ON, Pederson SB, 1993. Habitat-induced degradation of sound signals: quantifying the effects of communication sounds and bird location on blur ratio, excess attenuation, and signal-to-noise ratio in blackbird song. J Acoust Soc Am 93:2206-2220.
Drewry GE, Rand AS, 1983. Characteristics of an acoustic community: Puerto Rican frogs of the genus Eleutherodactylus. Copeia 1983:941-953.
Duellman WE, Pyles RA, 1983. Acoustic resource partitioning in anuran communities. Copeia 1983:639-649.
Endler JA, 1992. Signals, signal conditions, and the direction of evolution. Am Nat 139:S125-S153.[Web of Science]
Gish SL, Morton ES, 1981. Structural adaptation to local habitat acoustics in Carolina wren song. Z Tierpsychol 56:74-84.
Henwood K, Fabrick A, 1979. A quantitative analysis of the dawn chorus: temporal selection for communicatory optimization. Am Nat 114:260-274.
Hödl W, 1977. Call differences and calling site segregation in anuran species from central Amazonian floating meadows. Oecologia 28:351-363.
Hunter ML, Krebs JR, 1979. Geographical variation in the song of the great tit (Parus major) in relation to ecological factors. J Anim Ecol 48:759-785.
Langbauer WRJ, Payne KB, Charif R, Rapaport L, Osborn F,1991
. African elephants respond to distant playbacks of
low-frequency conspecific calls. J Exp Biol
157:35-46.
Littlejohn MJ, 1977. Long range acoustic communication in anurans. In: The reproductive biology of amphibians (Taylor DH, Guttman SI, eds). New York: Plenum Press;263 -294.
Loftus-Hills JJ, Littlejohn MJ, 1971. Mating-call sound intensities of anuran amphibians. J Acoust Soc Am 49:1327-1329.
Marten K, Marler P, 1977. Sound transmission and its significance for animal vocalization I. Temperate habitats. Behav Ecol Sociobiol 2:271-20.
Marten K, Quine D, Marler P, 1977. Sound transmission and its significance for animal vocalization II. Tropical forest habitats.Behav Ecol Sociobiol 2:291-301.
Martin WF, 1967. The mechanism and evolution of sound production in the toad genus Bufo (MA dissertation). Austin: University of Texas.
Mathevon N, Aubin T, Dabelsteen T, 1996. Song degradation during propagation: importance of song post for the wren Troglodytes troglodytes. Ethology 102:397-412.
Michelsen A, Larsen ON, 1983. Strategies for acoustic communication in complex environments. In: Neuroethology and behavioral physiology (Huber F, Markl H, eds). Berlin: Springer-Verlag;321 -331.
Morton ES, 1975. Ecological sources of selection on avian sounds. Am Nat 109:17-34.
Penna M, Solis R, 1998. Frog call intensities and sound propagation in the South American temperate forest region. Behav Ecol Sociobiol 42:371-381.
Richards DG, Wiley RH, 1980. Reverberations and amplitude fluctuations in the propagation of sound in a forest: implications for animal communication. Am Nat 115:381-399.
Ryan MJ, 1985. The túngara frog: a study in sexual selection and communication. Chicago: University of Chicago Press.
Ryan MJ, 1986. Environmental bioacoustics: evaluation of a commonly used experimental technique. Anim Behav 34:931-933.
Ryan MJ, 1988. Energy, calling, and selection.Am Zool 28:885-898.
Ryan MJ, Brenowitz EA, 1985. The role of body size, phylogeny, and ambient noise in the evolution of bird song. Am Nat 126:87-100.
Ryan MJ, Cocroft RB, Wilczynski W, 1990. The role of environmental selection in intraspecific divergence of mate recognition signals in the cricket frog, Acris crepitans.Evolution 44:1869-1872.
Ryan MJ, Keddy-Hector A, 1992. Directional patterns of female mate choice and the role of sensory biases. Am Nat 139:S4-S35.
Ryan MJ, Sullivan BK, 1989. Transmission effects on temporal structure in the advertisement calls of two toads, Bufo woodhousii and Bufo valliceps. Ethology 80:182-189.
Sorjonen J, 1983. Transmission of the two most characteristic phrases of the song of the thrush nightingale Luscina luscina in different environmental conditions. Ornis Scand 14:278-288.
Sorjonen J, 1986. Factors affecting the structure of song and the singing behaviour of some northern European passerine birds.Behaviour 98:286-302.
Stebbins RC, Cohen NW, 1995. A natural history of amphibians. Princeton, New Jersey: Princeton University Press.
Waser PM, Brown CH, 1986. Habitat acoustics and primate communication. Am J Primatol 10:135-154.
Waser PM, Waser MS, 1977. Experimental studies of primate vocalization: specializations for long-distance propagation. Z Tierpsychol 43:239-263.
Wilczynski W, Ryan MJ, 1999. Geographic variation in animal communication systems. In: Geographic variation in behavior (Foster SA, Endler JA, eds). Oxford: Oxford University Press;234 -261.
Wiley RH, Richards DG, 1978. Physical constraints on acoustic communication in the atmosphere: implications for the evolution of animal vocalizations. Behav Ecol Sociobiol 3:69-94.
Wiley RH, 1991. Associations of song properties with habitats for territorial oscine birds of eastern North America. Am Nat 138:973-993.
Zimmerman BL, 1982. A survey of Central Amazonian forest frog species and an analysis of structural and behavioural features of their vocalizations (PhD dissertation). Guelph, Ontario: University of Guelph.
Zimmerman BL, 1983. A comparison of structural features of calls of open and forest habitat frog species in the central Amazon. Herpetologica 39:235-245.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  V. C. K. Couldridge and M. J. van Staaden Habitat-dependent transmission of male advertisement calls in bladder grasshoppers (Orthoptera; Pneumoridae) J. Exp. Biol., July 15, 2004; 207(16): 2777 - 2786. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||