Behavioral Ecology Advance Access published February 2, 2010 Behavioral Ecology doi:10.1093/beheco/arp203
Assessing ‘‘false’’ alarm calls by a drongo (Dicrurus paradiseus) in mixed-species bird flocks S. Harsha K. Satischandra,a Prasanna Kodituwakku,a Sarath W. Kotagama,b and Eben Goodaleb Department of Natural Resource Management, Faculty of Applied Science, University of Sabaragamuwa, Belihuloya, Sri Lanka and bField Ornithology Group of Sri Lanka, Department of Zoology, University of Colombo, Colombo, Sri Lanka
a
The suggestion that some members of mixed-species bird flocks use alarm calls when predators are not present in order to startle other species and thereby gain access to additional prey, first postulated by Munn (Munn CA. 1986. Birds that ‘cry wolf’. Nature. 391:143–145.), has generated considerable interest due to its implication that the calling birds are intentionally deceiving listeners. Despite this interest, ‘‘false alarms’’ have been studied rarely and without detailed acoustical analysis. We explored whether false alarms are used by Greater Racket-tailed Drongos, which produce a distinctive set of notes when alarmed. We found that drongos did indeed make false alarms, defined as an alarm vocalization made either while the bird was on the perch or in the air, and followed within 10 s by a foraging attempt. However, the acoustic features of the call notes used, particularly of those calls made in the air, were more similar to aggressive calls, made when drongos chased each other, than to actual alarms produced when predators were present. Drongo foraging success was greater after false alarm calls than after silence or nonalarm vocalizations, and playback of false alarms in the air induced escape behavior in other species, though at a lower level than actual alarms. Thus, although drongos can use false alarms to startle other birds and gain foraging opportunities, such calls cannot be called ‘‘false’’ with certainty because they may also signal aggressive intent. Indeed, aggression and alarm may be intertwined in this family of birds as drongos actively mob or chase many predators, including bird-eating hawks. Key words: aggressive interactions, alarm calls, deception, Dicrurus paradiseus, false alarms, interspecific communication, mixed-species flocks. [Behav Ecol]
larm calling has long been of theoretical interest because it seems apparently ‘‘altruistic’’ in that the caller may incur a cost for producing the call without receiving a direct benefit to itself (Sherman 1977). Indeed, alarm calling played a key role in the development of kin selection theory, as researchers argued that most animals that alarm call are in groups of related individuals (Maynard Smith 1965; Sherman 1977). However, it also became apparent that an alarm caller could benefit either by directly communicating to the predator that it had been seen (Caro 1995) or by inducing other individuals to flee, thereby confusing the predator or distracting attention from itself (Charnov and Krebs 1975). Related to the latter hypothesis, Munn (1986) published a widely cited article suggesting that birds in mixed-species flocks make alarm calls when there are no predators present, startling other birds into dropping food that the caller then consumes. Munn (1986) came to this conclusion by defining ‘‘false alarms’’ as foraging attempts made when no predator was observed and directly after an alarm call; he then used a playback experiment to demonstrate that false alarms elicited escape responses from other species much as did actual alarms to predators. Subsequently, false alarms have also been reported in other mixed-species flock systems (Ridley and Raihani 2007) and in birds at feeding tables (Møller 1988; but see Evans et al. 2004). Note that these kinds of false alarms are different from false alarms in which the caller simply provides inaccurate information because it has perceived the environment inaccurately
A
Address correspondence to E. Goodale, who is now at c/o Nieh Lab, Biological Sciences, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0116, USA. E-mail: eben.goodale @gmail.com. Received 5 August 2009; revised 15 December 2009; accepted 21 December 2009. The Author 2010. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail:
[email protected]
(Beauchamp and Ruxton 2007). Rather, such behavior implies that the caller has accurate knowledge of the actual environmental conditions but transmits false information regardless. Hence, this behavior has been largely cited in literature about deception in animals (Cheney and Seyfarth 1991; Semple and McComb 1996; Rowell et al. 2006). Despite the interest in false alarms and the influence of the article by Munn (1986), Munn’s work has not been replicated in many other mixed-species bird flock systems, even though alarm calling is known to occur in, and to be important to, such flocks (Morse 1970; Gaddis 1980; Sullivan 1984; Goodale and Kotagama 2005). Although some studies have compared spectrograms of false and true alarms (Munn 1986; Ridley et al. 2007, see supplementary information), none have systematically investigated the duration and frequency of the 2 types of calls. Two studies have demonstrated that false alarms are effective by showing that they induce response when played back (Munn 1986; Møller 1988). However, alternate hypotheses for the birds’ response, such as that the calls indicate aggression, and not necessarily alarm, have not been explored. Here, we investigated the potential occurrence of false alarm calling in the Greater Racket-tailed Drongo (Dicrurus paradiseus), a frequent participant in mixed-species flocks of lowland Sri Lanka (Kotagama and Goodale 2004). Throughout the Old World tropics, drongos often associate with other species and make alarm calls (Chen and Hsieh 2002; Trivedi 2006), and they have been suspected of making false alarm calls (Herremans and Herremans-Tonnoeyr 1997; King and Rappole 2001; Ridley and Raihani 2007). In the Sri Lankan system, drongos occasionally kleptoparasitize other species in this system and frequently catch insects disturbed by other birds (Satischandra et al. 2007). The advantages of using the Sri Lankan drongos to investigate false alarms include that alarm calls in this mixed-species flock system have been described in detail (Goodale and Kotagama 2005) and that drongos
2
produce distinctive, well-characterized notes in alarm contexts (Goodale and Kotagama 2006). We defined a false alarm as a vocalization consisting of a high percentage of these notes, produced in the apparent absence of a predator and quickly followed by a foraging attempt. We present observational data on the occurrence of such behavior by drongos in mixed-species flocks, compare the acoustic characteristics of these notes with alarms in response to naturally occurring predators, and investigate whether this behavior influenced the foraging success of drongos. We also conducted a playback experiment to determine how other species participating in flocks responded to such false alarms. MATERIALS AND METHODS
Behavioral Ecology
Figure 1 The 7 types of species-specific drongo alarm notes, in the order they usually appear in. HP, high pitched; DS, downslur; US, upslur; MU, multiple note upslur; CK, crack; CR, crescent; GR, growl.
Study site and species The study was conducted in the Sinharaja World Heritage Reserve (lat 626#N long 8021#E, 450–600 m), the largest remaining relatively undisturbed lowland rainforest in Sri Lanka, categorized as a tropical wet evergreen forest (de Zoysa and Raheem 1987; Gunatilleke et al. 2004). The mixed-species flocks of the reserve are large (averaging 12 species and 40 individuals) and are characterized by 2 species that are found in .90% of flocks: Orange-billed Babblers (Turdoides rufescens, average of 16 individuals per flock) and Greater Racket-tailed Drongos (average of 3 individuals, Kotagama and Goodale 2004). The subspecies of Greater Racket-tailed Drongo (Dicrurus paradiseus lophorhinus) present in the forest is considered by some to be a separate species, endemic to Sri Lanka (Rasmussen and Anderton 2005). Observational study The study was conducted on unmarked drongos in 7 different areas of the reserve, with each site at least 1.5 km from all others to ensure independence as radiotracked drongos did not travel more than this distance (Goodale and Kotagama 2006). The observations were made by 2 observers, one watching the bird with binoculars and the other recording the birds’ vocalizations with a Marantz PMD 222 or 430 cassette recorder and a Seinheiser ME 62 omnidirectional microphone mounted on a Telinga parabola. Drongos were observed when they were in mixed-species flocks. A focal bird was chosen that was currently silent and perched, and recording was commenced and timed, with the observer with binoculars noting when the bird vocalized. Any movements of the bird were then noted as either 1) a foraging attempt, a twisting flight obviously directed toward a flying insect (referred to as ‘‘sallies’’ in Remsen and Robinson 1990) or 2) a straight flight to another perch. The observation was continued until the drongo returned to the perch after a foraging attempt. After the observation was over, the following information was recorded: 1) any individually distinctive physical features of the focal bird (especially its crest and tail), 2) whether the drongo was successful in the foraging attempt and how that information was judged (e.g., the insect was seen or the drongo was seen manipulating food with its bill), and 3) whether any predator was seen to be present or whether there was any alarm reaction (sudden scattering) by any bird in the flock. When following a mixed-species flock, we attempted to gather data from different individual drongos in the flock (judging individual identity by the birds’ crests and tails); within an area, we also attempted to sample different flocks, which tend to follow regular routes. Using this sampling scheme, we believe the chance of making repeated observations on the same individuals was minimized. We digitized the recordings (at 22 050 Hz sampling rate) and rendered spectrographs using the software package Raven
1.2 (copyright Cornell Laboratory of Ornithology, Ithaca, NY, USA, 2005). We then scored each drongo note (defined as a continuous line on a spectrogram) as either alarm or nonalarm. Drongos use 7 types of species-specific alarm notes, which collectively comprise ;75% of the notes included in alarm vocalizations and which are rarely found outside of alarms (Goodale and Kotagama 2006, Figure 1). Alarm vocalizations also sometimes include imitations of raptors, nest predators, and other species’ alarm calls. We then scored each complete sample that contained vocalizations as either alarm or not, depending on whether at least 30% of the notes were drongo species-specific alarm notes or not. The 30% criterion was adopted because in previous work most recordings clustered into 2 groups: 1) vocalizations with occasional (,30%) species-specific alarm notes in them but with very few examples of alarm call mimicry and 2) vocalizations with .30% speciesspecific alarm notes, and with many examples of alarm call mimicry (see Figure 3 in Goodale and Kotagama 2006); in practice, almost all samples classified as alarms had proportions of alarm notes much higher than 30%. Observations were then categorized into the following categories: 1) observations without any vocalizations, 2) observations with nonalarm vocalizations, 3) observations with alarm vocalizations in which a predator was present or an alarm response of a nonfocal bird was noted prior to the focal bird’s alarm vocalization (‘‘true alarm’’), 4) observations in which the alarm vocalization preceded foraging attempts by more than 10 s, but in which no predator or prior alarm reaction was seen (‘‘alarms of unknown context’’; the thick vegetation of the rainforest sometimes precludes identification of a threat by human observers), 5) observations in which the alarm vocalization preceded foraging attempts by less than 10 s, but in which no predator or prior alarm reaction was seen (‘‘false alarm on the perch’’), 6) observations in which the alarm vocalization occurred in flight, before a foraging attempt, in which no predator or prior alarm reaction was seen (‘‘false alarm in air’’), and 7) observations in which drongos aggressively chased each other through the air (‘‘aggressive interactions’’). The 10 s criterion was chosen because birds respond to drongo alarms within 5 s (Goodale and Kotagama 2008) and then remain in place for at least 5 s longer (in practice most responses are within 1 s; however, there is a dynamic, as described in Goodale and Kotagama 2005, that early calls are unreliable, whereas later calls emphasize true threats, and the 5 s rule was chosen in earlier studies to allow for the possibility that birds might wait to assess reliability). No timing criterion was used for false alarms in air or aggressive interactions because the foraging attempt or aggressive behavior always followed the calling by less than 10 s. We began the analysis by first comparing the number of observations in each site of 4 categories of vocalizations—true alarms, false alarms on the perch, false alarms in the air, and
Satischandra et al.
•
Assessing ‘‘false alarms’’
aggressive interactions. We then compared the 4 categories in the proportion of all notes that were of alarm type and also in the proportion of alarm notes that were of each type (7 drongo note types and 2 different kinds of mimicry). Because there were very few observations of true alarms and aggressive behaviors, we pooled among areas when conducting this analysis; however, we first ensured that the acoustic characteristics of vocalizations for which sample sizes were high (false alarms on the perch or in air) did not differ significantly by area. Comparisons between the 4 vocal categories were made using a generalized linear model with a binomial distribution and a logit link (Crawley 2002), using the GLM function in R 2.7.1 (R Foundation for Statistical Computing 2008). For true alarms, we supplemented the data with 8 recordings of predator encounters made more than 10 years previous to this study. In making multiple comparisons, we reduced the alpha level using the Dunn–Sˇida´k method (Sokal and Rohlf 1995). We also tested for differences among the 4 vocal categories in their acoustic variables: the duration of the alarm, the number of notes within the first second of the alarm (the ‘‘intensity’’ of the alarm), and the frequency of the first second of the alarm. We focused on the first second of the alarm because most response occurred after this time period. In many cases, the alarm itself did not last a second, leading to collinearity between the intensity (number of notes in first second) and the duration of the call. Therefore, we added another variable, which was the number of notes in the first second, if the alarm lasted that long (this variable could thus only be measured on a subsample of the calls). For the timing attributes (duration and number of notes), the full number of recorded exemplars from each category was used, although we did not include the older recordings of hawk attacks in this analysis because they were not made in accordance with the current study’s timing protocol. The duration of the alarm was defined as the time from the start of the first alarm note to the end of the last one, provided that the bird did not forage during this time. The distribution of duration values was highly nonnormal, so we analyzed them with the nonparametric Kruskal–Wallis test, followed by Mann–Whitney U tests for multiple comparisons (Sokal and Rohlf 1995). Frequency measurements, which were more time intensive, were made on 15 randomly chosen exemplars per category. The measured exemplars for true alarms and false alarms in air were also used for making playback tapes (see below). To measure frequency, we noted the frequency of peak amplitude (‘‘peak frequency’’) and the lowest frequency and highest frequency that were 10 db less in amplitude, modifying the technique of Podos (2001). Frequency and intensity measurements were analyzed using analysis of variance (ANOVA) as they met the assumptions of normality and heteroschedasticity, and multiple comparisons were made using the Tukey Honestly Significant Difference (HSD) method. Given that birds were unmarked, we acknowledge that there is the possibility that one individual may have contributed more than one exemplar within the analyzed sample (a risk higher in the larger sample for timing variables than in the smaller sample for frequency variables). However, as discussed above, we tried to minimize this possibility by rotating between individual birds within a flock and following different flocks within an area. The sample size of observations of foraging success was large enough to do the analysis separately for the 7 areas. For each area, we determined the percentage of successful foraging observations after a false alarm (on the perch or in the air) and compared that with the percentage of successful foraging observations after silence or nonalarm vocalizations, comparing the categories using Fisher Exact tests.
3
Playback experiment Three treatments were compared in a playback experiment: vocalizations from true alarms, false alarms in air, and control sounds. We used vocalizations from false alarms in air rather than from false alarms on perch because the in-air notes were more acoustically distinct from true alarms, and hence, if they could be shown to effectively elicit an escape response, it is probable that false alarms on the perch would do so as well (see Results and Discussion). Each of the 45 playback exemplars was made from a different recording. Fifteen true alarm playback exemplars were made from recordings of flock responses to instances of actual predators attacking flocks—13 Accipiter hawks and 2 Crested Serpent Eagles (Spilornis cheela), representing all examples of such attacks in our recording database. We edited the exemplars so that only drongo alarms were included. Fifteen false alarms in air exemplars were made from randomly selected recordings of good sound quality. Fifteen control exemplars were made from recordings of Yellow-fronted Barbets (Megalaima flavifrons), which are loud birds not known to make alarm calls. Each exemplar began with 15 s of flock noise without any alarm calls, followed by the stimulus and continuing for 30 s. True alarms were all 30 s long, but false alarms in air were between 2 and 5 s long (see Results). To isolate any effect of duration on response, the barbet exemplars were made so that 7 of them were 30 s long and the rest were between 2 and 5 s long. Playback exemplars were standardized for amplitude and played at a peak amplitude of 90–95 db at 1 m from the speaker, a amplitude that is similar to the natural one. In the field, a 15 km route of footpaths was walked until a flock was encountered. Then this flock was followed off road for at least 15 min until it habituated to the presence of the observers. The observers attempted to position themselves in front of the flock so that the birds would need to reverse direction in order to move away from the speaker. Also, the observers attempted to stay uphill of the flock when possible; this makes the playback broadcast from the observer’s shoulder height sound more natural. Orange-billed Babblers were chosen as subjects for the playback experiment because they are the most numerous species in the flocks and because they responded clearly to drongo alarms in a previous experiment (Goodale and Kotagama 2008). For each trial, a focal Orangebilled Babbler was chosen that was clearly visible on an exposed branch and foraging within 10 m of the observers. As the observer’s watched the focal bird, recording was started, and 15 s of baseline vocalizations were recorded before starting the 45 s playback stimulus. After a trial, we recorded the following information: 1) Did the bird fly or jump within 5 s of the start of the alarm? 2) If so, did it fly or jump away from the speaker, or toward the speaker, or neither? 3) Did the bird make a head turn greater than 90 within 5 s of the start of the alarm? After making these notes, the observers began to follow the flock again. After an additional 15 min of habituations, another trial was conducted. For one flock, only one trial of each treatment was performed (hence, 3 playback trials per flock). The order of the trials was randomly selected. In order for the trials to be independent, different flocks were chosen that were at least 250 m from each other. In the statistical analysis of the playback experiment, we categorized the response of a bird as 0—no response, 1—head turn within 5 s of the beginning of the alarm, but no other movement, and 2—a jump or a flight away from the speaker within 5 s. Because a head turn response was considered intermediate between no response and a movement, ordinal logistic regression was used for analysis using the logistic regression model procedure in R, version 2.7.1 (R Foundation for Statistical Computing 2008).
Behavioral Ecology
4
Table 1 The number of observations of the different categories in the 7 sampling areas, in order of the total number of recordings per area
Location
Silent
Nonalarm
True alarms
Research center Wathurawa Deniya to Heendola Pitakalle Murakelle Sinhagala Deniyaya Total
126 124 100 109 86 81 37 663
26 23 17 9 12 5 2 94
0 0 1 2 4 2 0 9
Alarms of unknown occasion
False alarm on perch
False alarm in air
Aggressive
Total
5 6 5 10 4 3 2 35
26 19 31 8 12 11 6 113
20 5 13 24 19 8 7 96
6 12 0 4 6 4 0 33
209 189 167 166 143 114 55 1043
For details on how the observations were classified, see text.
RESULTS Observational study False alarms were commonly produced by Greater Rackettailed Drongos that participated in mixed-species flocks. In a total of 64.5 field h, we made 1043 observations of drongos, including data from all 7 sampling areas. In the majority (64%) of observations, drongos were completely silent (Table 1). True alarms were very rare, although when observations were pooled over the 7 sampling areas, a total of 1% of all observations were true alarms, including 6 attacks on mixed-species flocks by Accipiter hawks, one attack by a Crested Serpent Eagle, and 2 incidents in which nonfocal birds clearly scattered but the stimulus was not observed. Observations of aggressive interactions were also rare, representing 3% of all observations. False alarms were more common and represented a relatively constant percentage of observations (ranging from 12% to 27%) in each area (see Table 1), including observations of both false alarms made on the perch and in air. Different locations, however, differed in the types of false alarms that occurred: for example, drongos at the Waturava location had many more false alarms on the perch (19) than in the air (5), whereas drongos at the Pitakelle location had fewer false alarms on the perch (8) than in the air (24; Fisher’s Exact test, P , 0.001). False alarms and aggressive calls contained a greater percentage of alarm notes than the true alarms themselves. True alarms consisted of 75% alarm notes, whereas the other 3 categories had even higher percentages, with a significant difference among categories (Table 2, generalized linear model, log-likelihood v23 ¼ 8.09, Dunn–Sˇida´k corrected P ¼ 0.045). False alarms in the air had a significantly higher percentage of alarm notes than true alarms (Wald’s statistic
z ¼ 2.67, Dunn–Sˇida´k corrected P ¼ 0.045), likely due to false alarms being shorter (see further details on duration, below) and more likely to include just one note, in which case all notes in the alarm are of alarm type, by definition. Thirtyeight percent of exemplars of false alarms in air included only one note, whereas 0% of true alarms were of that length. The relative proportion of the different types of alarm notes also differed between true alarms and false alarms, with false alarms, and particularly false alarms in air, being more similar in composition to aggressive calls than to true alarms. True alarms were composed of all the alarm call types, in relatively even proportions, and some alarm mimicry (Figure 2). In contrast, vocalizations in the other categories were composed of fewer note types (again, this is related to the duration of the vocalizations, with false alarms and aggressive calls often having one or a few notes). False alarms in air and aggressive calls were composed disproportionately of ‘‘crack’’ notes (significantly different from true alarms: for false alarms in air, generalized linear model, Wald statistic z ¼ 12.87, Dunn–Sˇida´k adjusted P , 0.001; for aggressive calls, Wald statistic z ¼ 3.61, Dunn–Sˇida´k adjusted P ¼ 0.017), a category that reached 63% of all notes for false alarms in air. False alarms on the perch were composed disproportionately of ‘‘high-pitched’’ notes (significantly different from true alarms: Wald statistic z ¼ 4.76, Dunn–Sˇida´k adjusted P , 0.001). True alarms had more ‘‘upslurs’’ than false alarms in air (Wald statistic z ¼ 3.65, Dunn–Sˇida´k adjusted P ¼ 0.014) and more ‘‘multinote’’ calls than false alarms in air (Wald statistic z ¼ 4.48, Dunn–Sˇida´k adjusted P , 0.001) or false alarms on the perch (Wald statistic z ¼ 6.01, Dunn–Sˇida´k adjusted P , 0.001). True alarms had more ‘‘downslurs’’ than false alarms in air (Wald statistic z ¼ 4.10, Dunn–Sˇida´k adjusted P , 0.001), and false alarms on the perch also had more notes of this type than false alarms in
Table 2 Differences between the 4 categories of calls in their Acoustic characteristics Acoustic characteristic
True alarms
False alarm on perch
False alarm in air
Aggressive
Percent alarm note Duration (s) Notes per first s Notes per first s, if at least 1 s long Peak frequency Low frequency High frequency Frequency bandwidth
75.7 139.9 3.82 3.82
6 6 6 6
14.9 (17)a 160.8 (9)a 0.95 (17)a 0.95 (17)a
87.5 2.1 2.28 3.16
6 6 6 6
19.9 (113)ab 5.2 (113)b 1.29 (113)b 1.33 (43)ab
91.6 2.4 2.16 2.74
6 6 6 6
17.2 (96)b 9.3 (96)b 1.31 (96)b 1.32 (27)b
87.8 1.1 2.67 3.67
6 6 6 6
20.4 (33)ab 1.0 (33)b 1.71 (33)b 1.23 (12)ab
3271 2343 5250 2907
6 6 6 6
1329 1059 1548 1522
(15)a (15)a (15)a (15)a
3291 2221 4647 2426
6 6 6 6
1455 1263 1132 1189
(15)a (15)a (15)ab (15)ab
2855 1776 4195 2419
6 6 6 6
1181(15)a 798 (15)a 1168 (15)ab 1211 (15)ab
2651 2098 3444 1345
6 6 6 6
1194 (15)a 1166 (15)a 1126 (15)b 842 (15)b
Categories that have different letters were significantly different from each other for that characteristic. Sample sizes in parentheses.
Satischandra et al.
•
Assessing ‘‘false alarms’’
5
their frequency bandwidth (F3,56 ¼ 4.43, P , 0.001). True alarms had significantly higher frequencies and wider bandwidths than aggressive calls (Tukey HSD multiple comparisons, P , 0.001), with the other categories being intermediate. Foraging success was higher after false alarm calls than during other kinds of foraging observations. With the data pooled across locations, foraging success was higher after false alarms on the perch than after silence or nonalarm vocalizations, and higher still for false alarms in the air. This same trend was seen in 6 of 7 sites and was significant (by Fisher’s Exact tests) in 5 sites (Table 3). Playback experiment The alarm response of Orange-billed Babblers to the vocalizations of Greater Racket-tailed Drongos was highest for true alarms, intermediate for false alarms in air, and nonexistent for the control sounds (of either long or short barbet calls, Figure 3). The response to true alarms was significantly higher than to false alarms in air (ordinal logistic regression, loglikelihood v21 ¼ 5.44, P ¼ 0.02), and the response to false alarms in air was significantly higher than to the control (loglikelihood v21 ¼ 11.87, P , 0.001). Figure 2 The percentage of notes of the 4 categories of vocalizations (true alarms, false alarms on the perch, false alarms in air, and aggressive calls) that were of the different kinds of alarm note types (either drongo species-specific alarm notes or mimicked alarm notes). The average percentage for that category of vocalization is shown (n ¼ 17 true alarms, n ¼ 113 false alarms on the perch, n ¼ 96 false alarms in air, and n ¼ 33 aggressive calls). The note types are ordered by their percentage in true alarms. The abbreviations for species-specific note types are as in Figure 1; mimicked note types included renditions of the Crested Serpent Eagle (EA, Spilornis cheela) and the alarm calls of jungle squirrels (SQ, Funambulus sp.).
air (Wald statistic z ¼ 3.94, Dunn–Sˇida´k adjusted P , 0.001). In general, the pattern was for false alarms on the perch to be most similar to true alarms, followed by aggressive calls, followed by false alarms in the air (see Figure 2). As mentioned previously, false alarms were shorter than true alarms. True alarms were prolonged and lasted up to 5 min; in contrast, false alarms were frequently less than 1 s long (Table 2). The differences between the duration of the different categories were highly significant (Kruskal–Wallis test, v23 ¼ 24.00, P , 0.001), with multiple comparisons indicating that true alarms were longer than any of the 3 other categories (Mann–Whitney U tests, W , 32, P , 0.001). False alarms were also less intense than true alarms in the number of notes in the first second. True alarms had more notes in the first second than did any of the other categories (ANOVA, F3,255 ¼ 8.29, P , 0.001; Tukey HSD multiple comparisons, P , 0.015). To some extent this was due to the differences in duration as many of the false alarms or aggressive calls ended before 1 s was up. But even when we limited the analysis to those calls that were at least 1 s long, there was still a significant difference between the 4 different types of calls in the number of notes per second (ANOVA, F3,96 ¼ 3.54, P , 0.020). In particular, false alarms in air had fewer notes per second than true alarms (Tukey HSD multiple comparisons, P ¼ 0.018). The categories also differed in some measurements of the frequency of the first second of the call. The categories did not differ in their peak (F3,56 ¼ 0.62, P . 0.40) or low frequencies (F3,56 ¼ 0.54, P . 0.45) but did differ in the highfrequency reading (F3,56 ¼ 5.51, P , 0.002) and consequently
DISCUSSION This study demonstrates that false alarms, defined in a way similar to landmark paper of Munn (1986), as alarm-like calls made when no predator was present and followed by foraging, are produced by Greater Racket-tailed Drongos in the Sri Lankan mixed-species flock system. These false alarms use the same set of alarm notes that are found in natural responses to actual predators. Yet the relative frequencies of these note types differ between false and true alarms, with false alarms, especially those made while the bird is already in the air, resembling aggressive vocalizations. In addition, like aggressive calls, false alarms are much shorter than true alarms and are intermediate between true alarms and aggressive calls in their frequency. We argue below that the similarities between alarm and aggressive calls make it difficult to resolve the question of whether drongos are using deception to manipulate other species, and thus it may be inappropriate to label these vocalizations ‘‘false’’ alarms. It is clear, however, that drongos benefit from producing alarm-like calls in the absence of predators and in the presence of other flocking species, as other species exhibit escape responses following playback of false alarms, and drongos increase their foraging success after uttering them. From this study and other recordings made over the previous 10 years, we have gathered a unique data set of 15 attacks by predators on mixed-species flocks. The drongo alarm calls in these recordings are usually prolonged, often lasting several minutes. The reason for such a prolonged response is that the stimulus usually remains close by: Accipiter hawks perch within flocks, circle overhead, or make repeated fly-throughs. In a previous study (Goodale and Kotagama 2005), we elicited alarm calls from mixed-species flock participants by throwing a stick to the side of flocks. The elicited calls were similar to the false alarms described here in that they were much shorter than true alarms although they used the same call types. Thus, false alarms that are the result of inaccurate detection (mistaking nonthreatening stimuli for threatening ones, see also Haftorn 2000; Beauchamp and Ruxton 2007) are similar to false alarms that are the result of inaccurate production (the production of an alarm signal in the absence of a stimuli, e.g., Munn 1986). In this study, we also gathered data on aggressive vocalizations produced by drongos in interactions with conspecifics (see below for discussion of aggression with heterospecifics).
6
Figure 3 The response to playback of 3 treatments: true alarms, false alarms in air, and a control sound (barbet calls). Movement could either be a flight or a jump (a movement in which both feet left the substrate, without wing motion).
Although drongos do not hold exclusive territories, they do on occasion chase other drongos out of mixed-species flocks. These chases, in which birds closely (less than a meter) approach each other, do not last long as the bird that is chased flies out and away from the flock. From our observations, we believe the chasing bird was the one that was calling, although because the interactions occurred quickly, this should be investigated more fully in future studies. The important finding of these observations was that aggressive calls were composed of the same note types as true alarms, though they lasted a much shorter duration, and were significantly lower in frequency. Here, we compare true alarms and aggressive calls with 2 types of false alarm calls: those made on the perch shortly before a foraging attempt (false alarm on the perch) and those made during the foraging attempt itself (false alarm in air). In conceiving this project, we planned only on recording false alarms on the perch but noticed at the beginning of the fieldwork that false alarms in air were also common. In comparing false alarms to true alarms and aggressive calls, the most striking difference is the duration of the calls: false alarms calls (of both types) are similar to aggressive calls in being much shorter than true alarms. However, it should be noted that this result is somewhat definitional because false alarms were defined as vocalizations followed by foraging and thus ended when the bird returned to the perch (in contrast, during true alarms, birds stayed on the perch for long durations of time). In other characteristics, false alarms on the perch tended to be more similar to true alarms than false alarms in air. For example, false alarms in air had a higher percentage of crack notes than true alarms and a lower percentage of downslur notes. Aggressive calls show the same pattern; in contrast, false alarms on the perch are statistically indistinguishable in these variables from true alarms. False alarms in air were also of lower frequency than false alarms on the perch, although these differences were not statistically significant; aggressive calls were of significantly lower frequency than true alarms. From the behavioral point of view, similarities between false alarm calls in air and aggressive calls make sense: false alarms in air were often made as drongos approached birds of other species, much in the same way that drongos made aggressive calls when chasing other drongos. Drongos are interspecifically
Behavioral Ecology
aggressive, although interspecific interactions are not as obvious as chases between drongos, as often the other species simply moves away from the area the drongos is moving into without being chased. In opportunistic observations of aggression made during a previous study (Kotagama and Goodale 2004), 15/16 aggressive interactions included drongos, with the drongos being the clear aggressor in 8 observations (Goodale E, unpublished data). Furthermore, drongos directly stole food from other species in about 4% of observed foraging attempts (Satischandra et al. 2007). What do these data imply about whether other species can distinguish between true and false alarms and whether drongos are practicing deception in producing such calls? Our data cannot rule out the possibility that the calls we defined as false alarms are in fact signals of aggressive intent, with a meaning equivalent to ‘‘get out of the way,’’ rather than any deceptive message. Nor can our data rule out the possibility that drongos are indeed producing deceptive calls similar to alarm calls when predators are absent in order to startle other birds into dropping prey. We suspect that a complex blend of these 2 behaviors—deceptive calls and aggressive calls—is occurring. For example, crack notes are particularly frequent in false alarms in air and aggressive calls, suggesting false alarm in air calls may be aggressive. High-pitch notes, in contrast, are usually the first note of a true alarm (12 of 16 predator attacks in which the start of the alarm was recorded began with such notes) and were used with a high degree in false alarms on the perch. Such notes would presumably have high signal value and might be most effective in startling other species. It is also interesting that while some drongos used false alarms in air primarily, other drongos used predominantly false alarms on the perch, as if drongos could chose the strategy they employed (see Table 1). Yet any linkage between false alarms in air and aggression and between false alarms on the perch and deceptive calls is overly simplistic because the categories of true alarms, false alarms, and aggressive calls all grade into each other in their acoustical characteristics. This gradient, we believe, may be due to drongos experiencing a mix of aggression and alarm in many situations. For example, drongos produce the seemingly aggressive crack calls even in true alarms. Perhaps drongos are signaling aggression in those situations, for drongos are well known to be aggressive toward predators, frequently mobbing them (Henry 1971; Nash and Nash 1985; Ali and Ripley 1987; Nijman 2004). Regardless of the meaning encoded in these false alarm calls, drongos profit by making them. Our observational data showed that drongos clearly experience greater foraging success after making such calls. The results of our playback experiment suggest a possible mechanism by which drongos attain greater foraging efficiency: false alarms startle other species of birds, which may allow drongos to capture insects the other birds had caught or were about to catch. It should be noted that only 3 of the 15 randomly selected playback exemplars of false alarms in air consisted solely of crack calls; hence, the subjects’ response to false alarms was not simply a response to this apparently aggressive call type. It would be interesting to know which alarm note types or acoustic characteristics influenced the response but unfortunately the sample size of 7 positive responses to false alarms was too small to fully explore this question; future playback experiments should address this issue. The fact that the response to false alarms in air was lower than response to true alarms suggests that other species of birds may be able to distinguish between false alarms and true alarms, perhaps cuing in on some of the acoustic characteristics such as intensity (notes per second) that differ between them. Yet because a failure to respond to an actual predator might be fatal and because false alarms fall within the range
Satischandra et al.
•
Assessing ‘‘false alarms’’
7
Table 3 The influence of drongo false alarms on success rate, broken down by area Foraging after false alarm
Foraging after silence or nonalarm
Location
Successful
Not
%
Successful
Not
%
Fisher’s Exact test (2-tailed)
Sinhagala Pitakalle Research Center Deniya to Heendola Wathurawa Murakalle Deniyaya
14 28 32 32 17 23 9
5 3 11 11 4 8 4
74 90 74 74 81 74 69
32 58 58 54 54 44 20
51 32 51 56 46 36 8
39 64 53 59 54 55 71
0.009 0.010 0.018 0.006 0.028 0.084 1
Of the 7 sites analyzed, 6 sites (all except Deniyaya) displayed the tendency for foraging attempts after false alarms to be more successful than foraging attempts after nonalarm vocalizations or silence, and for 5 of the sites the relationship was significant. Some observations were unclear as to whether foraging was successful or not and are not included in the analysis. Sites order by the significance of Fisher’s Exact test.
of variation for all the measured variables except duration, other species of birds may continue to respond even when a substantial proportion of alarms are false (Munn 1986). If birds waited for a few seconds, they would obtain information that could distinguish false alarms from true alarms with a high degree of confidence but such waiting could be very costly if the alarm was true. In conclusion, we believe that our study provides a useful cautionary result for those planning future work on false alarms. Previous studies of false alarms (Munn 1986; Møller 1988) have interpreted the result that other species respond to playback of alarm calls to conclude that receivers can not distinguish between true and false alarm calls. Yet such a result does not necessarily mean that signalers are making deceptive calls if the calls also encode aggressive intent. Although drongos may be an exceptional case due to their aggressiveness toward predators, the species that Munn (1986) described in Peru are ecologically similar to drongos in many respects (Munn 1984). Hence, we recommend that future researchers investigate aggressiveness as a potential confounding variable and carefully compare the categories of calls acoustically. FUNDING National Science Foundation (USA), through an International Research Postdoctoral Fellowship (IRFP grant 0601909 to E.G.). This study followed the ethical guidelines of the Animal Behavior Society and the laws of Sri Lanka. We thank P. Ashoka Jayarathna, A. G. Kirteratna, and Dulan R. Vidanapathirana for their help in the field, and Prof Mahinda Rupasinghe for encouragement. Thanks to Guy Beauchamp, Bruce Byers, Amanda Ridley, and 2 anonymous reviewers for helping us improve the manuscript. We are grateful to the Sri Lanka Forest Department for permission to work in the Sinharaja World Heritage Reserve, and the Rainforest Ecolodge for permission to work near Deniyaya. We are also grateful for a grant to S.H.K.S that allowed him to present this work at the 2008 ISBE meeting at Cornell University.
REFERENCES Ali S, Ripley SD. 1987. Compact handbook of the birds of India and Pakistan, together with those of Bangladesh, Nepal, Bhutan and Sri Lanka. 2nd ed. New Delhi (India): Oxford University Press. Beauchamp G, Ruxton GD. 2007. False alarms and the evolution of antipredator vigilance. Anim Behav. 74:1199–1206. Caro TM. 1995. Pursuit-deterrence revisited. Trends Ecol Evol. 10: 500–503. Charnov EL, Krebs JR. 1975. The evolution of alarm calls: altruism or manipulation? Am Nat. 109:107–112.
Chen C-C, Hsieh H. 2002. Composition and foraging behaviour of mixed-species flocks led by the grey-cheeked Fulvetta in Fushan Experimental Forest, Taiwan. Ibis. 144:317–330. Cheney DL, Seyfarth RM. 1991. Truth and deception in animal communication. In: Ristau CA, Griffin DR, editors. Cognitive ethology: the minds of other animals. Hillsdale (NJ): Lea Lawrence Erlbaum Associates. p. 127–151. Crawley MC. 2002. Statistical computing. New York (NY): Wiley and Sons. Evans DM, Ruxton GD, Ruxton DA. 2004. Do false alarm anti-predatory flushes provide a foraging benefit to subdominant species? Biologia. 59:675–678. Gaddis P. 1980. Mixed flocks, accipiters, and antipredator behavior. Condor. 82:348–349. Goodale E, Kotagama SW. 2005. Alarm calling in Sri Lankan mixedspecies bird flocks. Auk. 122:108–120. Goodale E, Kotagama SW. 2006. Context-dependent mimicry by a passerine bird. Proc R Soc Lond B Biol Sci. 273:875–880. Goodale E, Kotagama SW. 2008. Response to conspecific and heterospecific alarm calls in mixed-species bird flocks of a Sri Lankan rainforest. Behav Ecol. 19:887–894. Gunatilleke CVS, Gunatilleke IAUN, Ethugala AUK, Esufali S. 2004. Ecology of Sinharaja rain forest and the forest dynamics plot in Sri Lanka’s world heritage site. Colombo (Sri Lanka): Wildlife Heritage Trust. Haftorn S. 2000. Contexts and possible functions of alarm calling in the Willow Tit, Parus Montanus; the principle of ‘better safe than sorry’. Behaviour. 137:437–449. Henry GM. 1971. A guide to the birds of Ceylon. 2nd ed. London: Oxford University Press. Herremans M, Herremans-Tonnoeyr D. 1997. Social foraging of the forktailed drongo Dicrurus adsimilis: beater effect or kleptoparasitism? Bird Behavior. 12:41–45. King DI, Rappole JH. 2001. Kleptoparasitism of Laughingthrushes by Greater Racket-tailed Drongos in Burma (Myanmar). Forktail. 17:121–122. Kotagama SW, Goodale E. 2004. The composition and spatial organization of mixed-species flocks in a Sri Lankan rainforest. Forktail. 20:63–70. Maynard Smith J. 1965. The evolution of alarm calls. Am Nat. 99: 59–63. Møller AP. 1988. False alarm calls as a means of resource usurpation in the Great Tit Parus major. Ethology. 79:25–30. Morse DH. 1970. Ecological aspects of some mixed-species foraging flocks of birds. Ecol Monogr. 40:119–168. Munn CA III. 1984. The behavioral ecology of mixed-species bird flocks in Amazonian Peru [PhD thesis]. Princeton (NJ): Princeton University. Munn CA. 1986. Birds that ‘cry wolf’. Nature. 391:143–145. Nash AD, Nash SV. 1985. An extreme example of aggression displayed by Greater Racket-tailed Drongo Dicrurus paradiseus. Kukila. 2:7. Nijman V. 2004. Seasonal variation in naturally occurring mobbing behaviour of drongos (Dicruridae) towards two avian predators. Ethol Ecol Evol. 16:25–32. Podos J. 2001. Correlated evolution of morphology and vocal signal structure in Darwin’s finches. Nature. 409:185–188.
8
Rasmussen PC, Anderton JC. 2005. Birds of South Asia: the Ripley guide. Barcelona (Spain): Lynx Editions. Remsen JV Jr., Robinson SK. 1990. A classiciation scheme of foraging behaviour of birds in terrestrial habitats. Study Avian Biol. 13:144–160. R Development Core Team. 2008. R: A language and environment for statistical computing. Vienna (Austria): R Foundation for Statistical Computing. Ridley AR, Child MF, Bell MBV. 2007. Interspecific audience effects on the alarm-calling behaviour of a kleptoparasitic bird. Biol Lett. 3: 589–591. Ridley AR, Raihani NJ. 2007. Facultative response to a kleptoparasite by the cooperatively breeding pied babbler. Behav Ecol. 18: 324–330. Rowell JT, Ellner SP, Reeve HK. 2006. Why animals lie: why dishonesty and belief can coexist in a signalling system. Am Nat. 168: E180–E204.
Behavioral Ecology
Satischandra SHK, Kudavidanage EP, Kotagama SW, Goodale E. 2007. The benefits of joining mixed-species flocks for a sentinel nuclear species, the Greater Racket-tailed Drongo Dicrurus paradiseus. Forktail. 23:145–148. Semple S, McComb K. 1996. Behavioural deception. Trends Ecol Evol. 11:434–437. Sherman PW. 1977. Nepotism and the evolution of alarm calls. Science. 197:1246–1253. Sokal RR, Rohlf FJ. 1995. Biometry: the principles and practice of statistics in biological research. 3rd ed. New York: W.H. Freeman and Company. Sullivan KA. 1984. Information exploitation by downy woodpeckers in mixed-species flocks. Behaviour. 91:294–311. Trivedi P. 2006. Ecology and conservation of avifauna of some forested areas in Gujarat, India. [PhD theisis]. Rajkot (India): Saurashtra University. de Zoysa ND, Raheem R. 1987. Sinharaja, a rain forest in Sri Lanka. Colombo (Sri Lanka): March for Conservation.
