Experimental evidence that distinct song phrases in the Grey-cheeked Fulvetta Alcippe morrisonia permit species and local dialect recognition

Transcription

1 Ibis (2013), 1, Experimental evidence that distinct song phrases in the Grey-cheeked Fulvetta Alcippe morrisonia permit species and local dialect recognition BAO-SEN SIE, 1 * SI-SIUNG LIANG, 2 SIAO-WEI YUAN 3 & CAO-CIE CEN 1 1 Department of Biomedical Science and Environmental Biology, Kaohsiung Medical University, 0 Shihchuan 1st Road, Kaohsiung, 807, Taiwan 2 Department of Biotechnology, National Kaohsiung Normal University, 62 Sanchung Road, Yanchao Township, Kaohsiung, 824, Taiwan 3 School of Forestry and Resource Conservation, National Taiwan University, No.1, Sec. 4, Roosevelt Road, Taipei, 6, Taiwan Birds with small song repertoires have a limited number of song types which may serve multiple functions. The Grey-cheeked Fulvetta Alcippe morrisonia is a non-migratory species of lowland forests in Taiwan. Its song consists of two distinct phrases: a whistled phrase and a harmonic one. Each individual usually sings only one type of whistled phrase, and the geographical patterns of songs can be distinguished by the motif of this phrase. We proposed a dual adaptation hypothesis for the functions of these two phrases. Playback experiments including six sound stimuli (familiar whistled phrases, familiar harmonic phrases familiar complete songs, foreign whistled phrases, foreign harmonic phrases and foreign complete songs) were conducted at 12 sites in the Shoushan Nature Park of Kaohsiung, Taiwan. Grey-cheeked Fulvettas came closer to the playback speaker and increased counter-singing more in response to the familiar whistled phrase (local dialect) than to a foreign whistled phrase (distant dialect). owever, birds responded equally to the playback of harmonic stimuli from familiar and foreign sources. We suggest that in this species, whistled phrases are used for local recognition and harmonic phrases are used for species recognition in short-range communication. Keywords: communication, experiment, repertoire, Taiwan. Adaptation to function and environmental context have been identified as the primary drivers of the evolution of bird song acoustics (Kroodsma & Miller 1982, Boncoraglio & Saino 2007). One important function of song is to enable birds to recognize individuals of their own species (Catchpole & Slater 199). Features such as syntax, element structure and frequency, which vary little and are relatively constant among individuals of the same species, have been shown to be the most important cues for species recognition in several bird species (Becker 1982). For example, frequency changes are important in species recognition among Field Sparrows Spizella pusilla (Nelson *Corresponding author. bsshieh@kmu.edu.tw 1988). Songs are also used for territory defence and mate attraction, and both of these functions may require the recognition of individuals and local populations (Catchpole & Slater 199). The simultaneous need for both species specificity and individual variability places potentially conflicting demands on the evolution of bird songs (Marler 1960). For species with medium to large song repertoires (i.e song types) (Stoddard et al. 1991), different song types may be used to meet the respective needs of species specificity and individual variability. Geographically widely distributed song types are typically those that function in species recognition (Becker 1982). For birds with small repertoires (i.e. only one or two song types), songs with different functions involve different

2 Grey-cheeked Fulvetta song recognition 33 patterning and organization instead of distinct song types. For example, the Chestnut-sided Warbler Dendroica pensylvanica has two song patterns: accented, which contains a distinctive ending, and unaccented, which does not (Kroodsma et al. 1989). The unaccented song is used primarily in territorial defence, and the accented song is used for mate attraction. Local dialect recognition is known to play a key role in prezygotic reproductive isolation (Baker 1982, Danner et al. 2011). Variability in songs is required for local dialect recognition. For songbirds with a small repertoire, a local dialect can be recognized in various parts of a song (e.g. in the terminal trill of the song in the Puget Sound White-crowned Sparrow Zonotrichia leucophrys pugetensis) (Nelson & Poesel 2007) or in acoustic cues with different degrees of individual variation in the song (Skierczyński & Osiejuk 20). In addition to its function, environmental features such as habitat structure affect signal transmission and thus have profound effects on the evolution of acoustic structures in bird songs (Morton 197). Birds use songs not only for longrange communication but also for short-range communication; thus, short-range songs are expected to differ acoustically from long-range songs (Titus 1998). Wiley and Richards (1982) proposed that tonal, frequency-modulated signals are optimal for long-range communication, and that wide-spectrum sounds with sharp amplitude modulation and limited frequency modulation are optimal for short-range communication. The Grey-cheeked Fulvetta Alcippe morrisonia morrisonia is a non-migratory subspecies endemic to Taiwan (Collar & Robson 2007). It is found commonly in lowland forests (Chou et al. 1998) and forages in mixed-species flocks (Chen & sieh 2002). The Grey-cheeked Fulvetta is monogamous and sexually monomorphic, and both sexes participate in parental care (Lin 1996, Kuo 2000). A typical song comprises two distinct phrases: the whistled phrase and the harmonic phrase. The spectrogram structure of the whistled phrase is tonal and whistle-like, whereas the harmonic phrase has a broad frequency range with many harmonics (Fig. 1). All the songs of this species contain the whistled phrase, but not all of them include the harmonic phrase. Grey-cheeked Fulvettas usually perform a single song type based on the whistled phrase (Shieh 2004), and this species is therefore considered a small-repertoire species. We propose a dual adaptation hypothesis to explain the evolution of two distinct phrases in the songs of the Grey-cheeked Fulvetta as a result of adaptation to function and environment. From FT1 W FT2 W W FT3 W FT s s s s FT W FT6 W FT7 W FT8 W s s s s FT9 W FT W FT11 W FT12 W OT W s s s s s Figure 1. Sources of sound stimuli used for the playback experiments: FT1 12 were collected from 12 different sites in the study park; OT was collected from an area with a foreign dialect in Shanping. The whistled phrase is labelled with W and the harmonic phrase with on the spectrogram. FT4, FT and FT6 have identical syllable combinations in the whistled phrase.

3 34 B.-S. Shieh et al. the functional adaptation perspective, we predict that the whistled phrase is used for local recognition, due to its consistency within individuals and its geographical variability (Shieh 2004), and the harmonic phrase is used for species recognition, due to its consistency across populations in different locations (B.-S. Shieh, pers. obs.). We therefore predict stronger individual response to playback of the local whistled phrase than to the playback of whistled phrases from different dialects, but similar responses to harmonic phrases irrespective of local or foreign sources. In addition, according to Wiley and Richards s (1982) proposition for environmental adaptation, we predict that whistled phrases are used for long-range communication because their tonal structure transmits efficiently across long distances, and harmonic phrases are used for short-range communication due to their wide spectrum structure. That is, from the environmental adaptation perspective, the two distinctive phrases of this species song are used for long-range and short-range communication, respectively. Therefore, we also predict that the Grey-cheeked Fulvetta will respond to short-range signals (i.e. harmonic phrases) with short-range signals, and respond to long-range signals (i.e. whistled phrases) with long-range signals. Furthermore, the Grey-cheeked Fulvetta should come closer to the speaker during the playback of the harmonic phrase and counter-sing from a longer distance during the playback of the whistled phrase. METODS Study area and playback recordings We studied a Grey-cheeked Fulvetta population in the Shoushan Nature Park (22 38 N, E) of Kaohsiung City in southern Taiwan. The Shoushan Nature Park covers an area of approximately 00 ha and reaches an altitude of 36 m above sea level. The vegetation of the study area is characterized as evergreen rainforest. All of the sounds used in the playback experiments were recorded in the study park and in Shanping (22 8 N, E), which is located approximately 80 km from the study park. The sounds were recorded between 07:00 and :00 h in 200 using a Denon Portable IC Recorder (DN- F20R) equipped with a Sennheiser ME67 unidirectional microphone. For the experiment, we selected songs with good qualities of both whistled and harmonic phrases from 12 different sites (all at least 0 m apart) in the study park and used one of the most common song types from Shanping (recorded in 2004) as the source of our foreign sound stimuli. Because the Grey-cheeked Fulvetta defends no obvious territorial boundaries and because the nests of different pairs can be as close as m (Lin 1996), to ensure that the recordings from the 12 sites were from 12 different birds only one visible singing individual was recorded at each site, and it was followed until it flew out of recording range. From each site, we selected only one highquality recording of a source song with three syllables in the harmonic phrase. Recordings of the source songs were digitized and filtered at 1 using Avisoft-SASLab Pro software (R. Specht, Berlin, Germany). The spectrographic cross-correlation (SPCC) coefficients were computed to investigate the similarities between source songs using Avisoft- CORRELATOR (sampling rate = 2200 z, FFT = 12, frequency tolerance deviation = 0) (Specht 20). SPCC coefficients were calculated as the two spectrograms shifted incrementally past each other along the time axis and ranged from 1 to +1. A value of +1 means that the two spectrograms were identical. igher values of SPCC coefficients indicate a greater similarity between the two sound spectrograms. This method was used because comparisons are difficult to make using parametric methods (Cortopassi & Bradbury 2000). Recordings of the 13 source songs (12 from the study park and one from the Shanping area) were prepared as 20-s playback files and categorized as having one of three types of sound stimuli: whistled phrases (W), harmonic phrases () or complete songs (T). The 20-s sound stimuli of the separate whistled and harmonic phrases consisted of evenly spaced phrases, and the 20-s sounds of the complete songs consisted of five whistles and five harmonic phrases to maintain periods of silence comparable to the whistled and harmonic playbacks (Fig. 2). Playback procedures We used a Denon Portable IC Recorder (DN- F20R) connected to a wireless speaker (MIPRO MA-1, frequency response z ± 3 db) for the playback experiments. The speaker was placed on a tree approximately 1 2 m above the ground. Playback trials were conducted between 06:00 and 09:00 h, a period correspond-

4 Grey-cheeked Fulvetta song recognition 3 a) b) c) d) e) f) 1 s 1 s 1 s 1 s 1 s 1 s Figure 2. Examples of sound stimuli used for the playback experiments at site 1: (a) familiar whistled phrase (FW); (b) familiar harmonic phrase (F); (c) familiar complete song with both whistled and harmonic phrases (FT); (d) foreign whistled phrase (OW); (e) foreign harmonic phrase (O); and (f) foreign complete song with both whistled and harmonic phrases (OT). ing to the peak in the natural song activity of birds. After setting up the equipment, we waited until there was no disturbance and at least one Grey-cheeked Fulvetta was nearby. Sounds were played back at a sound-pressure level of approximately 6 db at m from the speaker. Playback levels were adjusted to approximate the natural levels of a singing bird, depending on the habitat and wind conditions. We broadcast two sets of sound stimuli at 12 sites in the study park. The familiar set of sound stimuli was played back at the site in which it was recorded, and the foreign set of sounds originating from the Shanping area was played at all 12 sites. Each set of sound stimuli included three types of sound stimuli. Thus, each site received six sound stimuli: (1) familiar whistled phrases (FW), (2) familiar harmonic phrases (F), (3) familiar, complete songs (FT), (4) foreign whistled phrases (OW), () foreign harmonic phrases (O) and (6) foreign complete songs (OT) (Fig. 2). We used complete songs as a positive control for our playback experiments using only whistled or harmonic phrases. In each trial, the order of the sound stimuli was initially randomized within each set (familiar or foreign) and then counterbalanced so that each order occurred once in a trial. For example, the played order of the sound stimuli in a trial could be FW-F-FT, F-FT-FW and FT-FW-F. During each triplet, the first sound stimulus was broadcast three times with silent intervals of 40, 40 and 160 s. After a period of at least 2 min of silence (i.e. with no songs heard nearby), the second sound stimulus was broadcast. Thus, in a single trial, each sound stimulus was played three times in the first, second and third place of the threepart sequence and was thus played a total of nine times. Because one trial lasted at least 6 min and the singing activity in the study park decreased sharply after 09:00 h, we performed only two trials per day. The two trials for each day were conducted at two separate sites to avoid habituation effects. Trials with familiar and foreign sets of sound stimuli were conducted at each site in the late breeding season (May July) of 200 and another two trials with the same familiar and foreign sets of sound stimuli in the early breeding season (March April) of Thus, playback experiments were repeated at two different times of the breeding season to increase the number of response samples while examining the effect of the time of the breeding season on the results. Measurement of singing and approach response We considered a site to be our sampling unit, and all singing or approach responses of Greycheeked Fulvettas near the playback speaker at the site were recorded. Aggressive responses

5 36 B.-S. Shieh et al. were not considered as they occurred only rarely and call responses were not considered because they often occurred irrespective of playback. We used the site instead of the individual as the sampling unit for two reasons. First, we did not identify individuals that sang or approached the playback area. Secondly, we expected that individuals at the same site might respond similarly to the same sound stimuli. owever, responding adults at the same site may often be different individuals due to the high annual mortality rate of adult Grey-cheeked Fulvettas (8.7% for second-year adults; u 1999). All song responses were recorded and categorized as whistled (W), harmonic () or complete songs (T) based on spectrograms (Fig. 1). We also calculated the following two measures for singing and approach responses to the sound stimuli in a trial: (1) proximity score and (2) occurrence percentage, as follows. We marked the horizontal distance from the speaker at and m. Using these marks, we estimated the distance between the speaker and a bird that was either singing or approaching. The singing proximity score was given a value of 1 when songs were heard within a m radius of the speaker during the playback period and the silent interval that followed, 2 when within a 20-m radius, 3 when within a -m radius and 4 when within a -m radius. Vegetation structure limited visibility so that approach behaviour could only be certainly recorded within m of the speaker. The approach proximity score was thus 1 when a bird approached within m of the speaker and 2 when it came within m. We used only the closest response to calculate proximity scores because birds might respond more than once during the playback period and the following silent interval. We then summed the nine singing (or approach) proximity scores from the nine playbacks of the same sound stimulus in a single trial to calculate the final singing (or approach) proximity scores and used these in further statistical analyses. The final proximity scores were natural logarithm-transformed. The occurrence percentage of singing (or approach) during playbacks was calculated as the number of playbacks with the singing (or approach) events divided by the total number of playbacks in a trial (9). The occurrence percentage was arcsine-transformed for further analyses. Statistical analyses Statistical analyses of responses to playbacks were based on a repeated measures approach using linear mixed models (LMM). The type of sound stimulus (whistled, harmonic or complete), familiarity of the sound stimulus (familiar or foreign), and the time of the breeding season (early or late) were entered as fixed factors in the models, and intra-site variation was taken into account by including the variable site as a random factor. Model selection was based on backward selection from a full model, including all two- and three-way interactions. The differences between pairs of stimuli were analysed using a least significant difference post hoc comparison of the estimated marginal means. We examined the association between response sound types and the type of sound stimulus, and whether this association depends on the time of the breeding season and familiarity of the sound stimulus using a hierarchical log-linear analysis procedure (SPSS Inc. 2003). We performed model selection based on a backward elimination procedure with the probability of removal set at 0.1. In other words, we started with a saturated model, and each interaction was tested for significance by deleting it from the model; these interaction terms were removed from the model if the significance probability for the change in chi-square for each effect was greater than 0.1, until only significant terms remained. All analyses were conducted with SPSS version RESULTS Song characteristics The 12 source songs from the 12 sites in the study park could be classified into song types based on the whistled syllables; three of the 12 source songs had identical syllable combinations in the whistled phrase (Fig. 1). The SPCC coefficients of the whistled phrases in the song types ranged from to 0.434, whereas the SPCC coefficients between the whistled phrases in the foreign song and in the song types in the study park ranged from to Amongst harmonic phrases, the two most similar source songs had an SPCC coefficient of (Fig. 1), but with this exception the coefficients ranged from to 0.38 within the study park, whereas the SPCC coefficients between the harmonic phrases of the

6 Grey-cheeked Fulvetta song recognition 37 foreign song and the phrases of the songs from the study park ranged from to The SPCC coefficients between the whistled phrases of the foreign and familiar songs were significantly greater than those between the harmonic phrases of the foreign and familiar songs (Wilcoxon signed rank test, n = 12, Z = 2.83, P = 0.00). Singing response to playback The final LMM revealed that time of breeding season and familiarity of sound stimulus had significant effects on the singing proximity score (time: F 1,137 = 20.19, P < 0.001; familiarity: F 1,137 = 27.17, P < 0.001) and on occurrence percentage (time: F 1,137 = 2.11, P < 0.001; familiarity: F 1,137 = 20.9, P < 0.001). Playbacks conducted earlier in the season and of familiar song provoked more and closer singing responses than playbacks later in the breeding season and those of foreign song (Table 1). In addition, the two measures of singing responses were significantly influenced by the interaction between familiarity and type of Table 1. Differences of least-square means in the final linear mixed model for singing response. Difference Estimate a ± SE t 137 P Proximity score Early vs. Late 0.70 ± <0.001** Familiar vs. Foreign 1.08 ± <0.001** armonic vs. Total 0.69 ± * Whistled vs. Total 0.04 ± armonic vs. Total 0.21 ± Whistled vs. Total 0.16 ± Occurrence percentage Early vs. Late.8 ± <0.001** Familiar vs. Foreign 12.1 ± ** armonic vs. Total. ± ** Whistled vs. Total 3.0 ± armonic vs. Total 1.9 ± Whistled vs. Total 2.2 ± Early: early in the breeding season; Late: late in the breeding season; Familiar: sound stimulus from the same site; Foreign: sound stimulus from a distant population; armonic: harmonic phrase; Whistled: whistled phrase; Total: complete song. a Estimates of transformed data. **P < 0.01; *P < 0.0. sound stimulus (LMM, proximity score: F 4,137 = 2.3, P = 0.044; occurrence percentage: F 4,137 = 3.96, P = 0.00). Specifically, harmonic phrases provoked significantly less singing, and responses occurred further from the playback speaker than in the whistled and complete songs, but only during playbacks of familiar sound stimuli (Table 1). In addition, familiar harmonic phrases provoked singing responses similar to the response to the foreign harmonic phrases, but familiar whistled phrases significantly increased singing responses compared with the foreign whistled phrases (Fig. 3). The final log-linear model indicated a two-way interaction (familiarity*type of sound stimulus: df = 2, likelihood-ratio chi-square change = 24.46, P < 0.001) and a three-way interaction (time of the breeding season*familiarity*response sound type: df = 2, likelihood-ratio chi-square change =.4, P = 0.063) (Goodness-of-fit test statistics for the final log-linear model: likelihood-ratio chi square = 12.6, df = 20, P = 0.892; Pearson chisquare = 11.71, df = 20, P = 0.926). The threeway interaction demonstrated that the relationship between response sound type and familiarity was not the same at different times of the breeding season (Fig. 4). Furthermore, response sound types did not have a significant interaction with the type of sound stimulus in the final log-linear model, indicating that the Grey-cheeked Fulvetta responded with similar sound types regardless of the type of sound stimulus. Approach response to playback Approach scores differed significantly between early and late in the breeding seasons (LMM, F 1,137 =.67, P = 0.019) and between familiar and foreign sound stimuli (LMM, F 1,137 = 20.33, P < 0.001). Type of sound stimulus interacted significantly with familiarity (LMM, F 4,137 = 2.94, P = 0.023). Specifically, in playback of foreign sound stimuli, harmonic phrases provoked significantly closer approaches than whistled and complete songs, whereas playback of all three types of familiar sound stimuli provoked similar approach responses (Table 2, Fig. 3). In addition, familiar and foreign harmonic phrases provoked similar approach responses, whereas familiar whistled phrases provoked significantly closer approaches than did foreign whistled phrases (Fig. 3). The final linear mixed model for approach occurrence only included the main effects of the

7 38 B.-S. Shieh et al. Table 2. Differences of least-square means in the final linear mixed model for approach response. Difference Estimate a ± SE t 137 P Proximity score Early vs. Late 0.28 ± * Familiar vs. Foreign 0.3 ± * armonic vs. Total 0. ± Whistled vs. Total 0.13 ± armonic vs. Total 0.4 ± * Whistled vs. Total 0.23 ± Occurrence percentage t 141 Early vs. Late 3.0 ± * Familiar vs. Foreign.3 ± < 0.001** Early: early in the breeding season; Late: late in the breeding season; Familiar: sound stimulus from the same site; Foreign: sound stimulus from a distant population; armonic: harmonic phrase; Whistled: whistled phrase; Total: complete song. a Estimates of transformed data. **P < 0.01; *P < 0.0. time of the breeding season (F 1,141 =.43, P = 0.021) and familiarity (F 1,141 = 16.64, P < 0.001) and did not include interaction effects. Playbacks early in the breeding season provoked more approaches than those late in the breeding season, and playbacks of familiar sound stimuli provoked more approaches than playbacks of foreign sound stimuli (Table 2). (a) (b) Ln (Sining proximity score) Arcsine (Singing occurrence percentage) Familiar Foreign * * armonic phrase Whistled phrase Complete song Familiar Foreign Playback type * armonic phrase Whistled phrase Complete song Playback type * DISCUSSION Whistled phrases for local recognition in both long- and short-range communication The whistled phrase is the initial phrase in Greycheeked Fulvetta song and consists of three to five syllables with distinguishable motifs that show significant geographical variation (Shieh 2004). The foreign stimuli used in the study originated approximately 80 km from the study population and possess a combination of whistled syllables that are quite different from that of the study population (the SPCC coefficients ranged from to 0.377). The differential response to the familiar and foreign whistled phrases, consistent through the breeding season, demonstrated that Grey-cheeked Fulvettas were able to use the whistled phrase to discriminate between the local dialect and the (c) Ln (Approaching proximity score) Familiar Foreign * * armonic phrase Whistled phrase Complete song Playback type Figure 3. Estimated marginal means ± se of transformed data for singing and approach responses in regarding with type of sound stimulus and familarity during playback trials. *P < 0.0.

8 Grey-cheeked Fulvetta song recognition 39 Frequency Time of breeding season and sources of stimuli Figure 4. Frequency of the three types of response sounds after playbacks of familiar or foreign sound stimuli during different times (early or late) of the breeding season. T, complete song with both whistled and harmonic phrases; W, whistled phrase;, harmonic phrase. distant dialect. Furthermore, we suggest that the discrimination between local whistled phrases and foreign whistled phrases results from familiarity rather than non-similarity of acoustic structures. Given that the SPCC coefficients between two whistled phrases in the study park ranged from to 0.434, it is not unusual for neighbouring birds to sing very different whistled phrases. That is, Grey-cheeked Fulvettas recognize dissimilar whistled phrases and counter-sing in response to those that are familiar (i.e. those of territorial neighbours). A lower singing response to foreign whistled phrases is associated with their unfamiliarity rather than the dissimilarity of their acoustic structures. Grey-cheeked Fulvettas can thus recognize familiar local whistled phrases even though there are high levels of variations in those local whistled phrases. The Grey-cheeked Fulvetta s use of the initial, whistled phrase for local recognition differs from that of other small-repertoire species, such as the Puget Sound White-crowned Sparrow, in which the terminal trill is used for dialect identity (Baptista 1977). Furthermore, whereas the Puget Sound White-crowned Sparrow uses the note complex in the introductory phrase and the trill in the terminal phrase as cues for individual and local dialect identity, respectively (Nelson & Poesel 2007), we suggest that the whistled phrase the introductory part of the Grey-cheeked Fulvetta s song may identify both local dialect and individual. The whistled phrase in the song of the Greycheeked Fulvetta is an effective cue for individual identity because it varies among individuals and is highly consistent with the stereotypy of each individual (Shieh 2004). To verify our suggestion, further playback designs should test for individual recognition, and an analysis of the stability of individual cues across seasons (Wegrzyn et al. 2009) should be conducted. If the Grey-cheeked Fulvetta uses whistled phrases for long-range communication and harmonic phrases for short-range communication, we predicted that the birds would respond to short-range signals (i.e. harmonic phrases) with short-range signals, and to long-range signals (i.e. whistled phrases) with long-range signals. owever, we found no significant relationship between the type of sound stimulus and the response sound types in playback experiments. Grey-cheeked Fulvettas were more likely to sing songs with whistled phrases regardless of the type of sound stimuli. Furthermore, when the stimuli were from familiar sources, Grey-cheeked Fulvettas responded with whistled phrases at close proximity. These results demonstrated that whistled phrases were used both as long- and short-range signals. armonic phrases for species recognition in short-range communication The results indicated that Grey-cheeked Fulvettas did not discriminate between foreign and familiar harmonic phrases. This lack of discrimination was not due to acoustic similarities, given that the SPCC coefficients between the harmonic phrases of foreign and familiar song stimuli were often smaller than those for the whistled phrase ( ). owever, it is possible that because the SPCC analyses were conducted by cross-correlating the time frequency spectrograms of two sounds and included all discernible structural features of both sound spectrograms, the correlation values of two harmonically rich sounds were more sensitive to overtone content and duration than the tonal sounds (Khanna et al. 1997). Certainly, the familiar and foreign harmonic phrases are similar when viewed as spectrograms (or as heard by

9 40 B.-S. Shieh et al. the human ear), and these tonal differences might explain why two harmonic phrases that seem similar have smaller SPCC coefficients than two whistled phrases that sound different. Greycheeked Fulvetta individuals came close to the playback speaker in response to both the foreign and the familiar harmonic stimuli, whereas they did not do so in response to the foreign whistled phrase. This provides evidence that the Grey-cheeked Fulvetta recognizes foreign harmonic phrases. If harmonic phrases are used for species recognition, we would expect that the structures of the harmonic phrases would be relatively consistent across different dialect areas in Taiwan as well as among different subspecies in other countries. Although we did not record songs from other subspecies of the Grey-cheeked Fulvetta, one song sample (recorded by David Farrow and accessible at from a subspecies in Thailand showed similar harmonic syllables in the final part of the song based on spectrograms. Additionally, Collar and Robson (2007) reported that songs of the Grey-cheeked Fulvetta in all distribution areas normally end with two to three buzzy eerh sounds, which are identified as harmonic phrases in our study. The consistency of harmonic phrases across populations may serve for both recognition of own species and distinction from congeneric, sympatric species. For example, the harmonic syllables make the songs of the Grey-cheeked Fulvetta easily distinguishable from songs of closely related species such as the Gould s Fulvetta Alcippe brunnea, which is usually found in the same habitats in Taiwan. The recognition of species-specific acoustic cues plays an important role in the competition for ecological resources among sympatric species of birds (Martin & Martin 2001, Matyjasiak 2004). In the present study, we used a fixed number of syllables (three) in the harmonic phrase during the playback experiments. Further field observations are needed regarding the variation of the number of syllables in the harmonic phrase and their signalling contexts (Shieh 2004), as these data could verify the function of the length of the harmonic phrase in the Greycheeked Fulvetta. In the playback experiments, we used complete songs with both whistled and harmonic phrases as a positive control, and we found that harmonic phrases provoked significantly closer approaches than complete songs when the sound stimuli were from foreign sources. This finding indicated that Grey-cheeked Fulvettas approached the speaker because they recognized the foreign harmonic phrases rather than because they were simply attracted by sounds; therefore, foreign harmonic phrases but not foreign whistled phrases provoked closer approaches. Moreover, our results revealed that the Grey-cheeked Fulvetta came close to the speaker during playback of harmonic stimuli from both familiar and foreign sources. This finding demonstrated that the Grey-cheeked Fulvetta used harmonic phrases for short-range signals. In conclusion, our results suggest that the whistled and harmonic phrases in the song of the Grey-cheeked Fulvetta are used for local and species recognition, respectively; the harmonic phrase in the final part of the song is primarily used in short-range communication and is not mutually exclusive with the function of species recognition. We thank the administration of the Shanping Forest Ecological Garden of the Taiwan Forestry Research Institute for its support during the study. We are very grateful to J. Wilson, L. Pichegru and anonymous reviewers for their valuable comments on earlier drafts of the manuscript. This work was supported by the Taiwan National Science Council (grant no. NSC B ). REFERENCES Baker, M.C Vocal dialect recognition and population genetic consequences. Am. Zool. 22: Baptista, L.F Geographic variation in song and dialects of the Puget Sound White-crowned Sparrow. Condor 79: Becker, P The coding of species-specific characteristics in bird sounds. In Kroodsama, D.E. & Miller, E.. (eds) Acoustic Communication in Birds, Vol. 1: New York: Academic Press. Boncoraglio, G. & Saino, N abitat structure and the evolution of bird song: a meta-analysis of evidence for the acoustic adaptation hypothesis. Funct. Ecol. 21: Catchpole, C.K. & Slater, P.J.B Bird Song: Biological Themes and Variations. New York: Cambridge Univeristy Press. Chen, C.-C. & sieh, F Composition and foraging behavior of mixed-species flocks led by the Grey-cheeked Fulvetta in Fushan Experimental Forest, Taiwan. Ibis 144: Chou, L.S., Chen, C.C. & Loh, S.W Diet analysis of the Grey-cheeked Fulvetta (Alcippe morrisonia) at Fushan Experimental Forest in Taiwan. Acta Zool. Taiwanica 9: Collar, N.J. & Robson, C Family Timaliidae (Babblers). In del oyo, J., Elliott, A. & Christie, D.A. (eds) andbook of the Birds of the World, Vol. 12: Barcelona: Lynx Edicions.

10 Grey-cheeked Fulvetta song recognition 41 Cortopassi, K.A. & Bradbury, J.W The comparison of harmonically rich sounds using spectrographic crosscorrelation and principal coordinates analysis. Bioacoustics 11: Danner, J.E., Danner, R.M., Bonier, F., Martin, P.R., Small, T.W. & Moore, I.T Female, but not male, tropical sparrows respond more strongly to the local song dialect: implications for population divergence. Am. Nat. 178: u, M.-F Age identification and population dynamic of Grey-cheeked Fulvetta (Alcippe morrisonia) at Fushan Experimental Forest. Master s thesis, National Taiwan University (in Chinese with English abstract). Khanna,., Gaunt, S.L.L. & McCallum, D.A Digital spectrographic cross-correlation: tests of sensitivity. Bioacoustics 7: Kroodsma, D.E. & Miller, E Acoustic Communication in Birds, Vol. 1. New York: Academic Press. Kroodsma, D.E., Bereson, R.C., Byers, B.E. & Minear, E Use of song types by the Chestnut-sided Warbler: evidence for both intra- and inter-sexual functions. Can. J. Zool. 67: Kuo, W.-W The sex difference and breeding ecology of Grey-cheeked Fulvetta (Alcippe morrisonia) at Fushan Experiment Forest. Master s Thesis, National Taiwan University (in Chinese with English abstract). Lin, R.-S The breeding and flock ecology of Greycheeked Fulvetta (Alcippe morrisonia). Master s thesis, National Taiwan University (in Chinese with English abstract). Marler, P Bird songs and mate selection. In Tavolga, W.N. (ed.) Animal Sounds and Communication: Washington, DC: A.I.B.S. Symposium Proceedings. Martin, P.R. & Martin, T.E Behavioral interactions between coexisting species: song playback experiments with wood warblers. Ecology 82: Matyjasiak, P Birds associate species-specific acoustic and visual cues: recognition of heterospecific rivals by male blackcaps. Behav. Ecol. 16: Morton, E.S Ecological sources of selection on avian sounds. Am. Nat. 9: Nelson, D.A Feature weighting in species song recognition by the Field Sparrow, Spizella pusilla. Behaviour 6: Nelson, D.A. & Poesel, A Segregation of information in a complex acoustic signal: individual and dialect identity in white-crowned sparrow song. Anim. Behav. 74: Shieh, B.-S Song structure and microgeographic variation in a population of the Grey-cheeked Fulvetta (Alcippe morrisonia) at Shoushan Nature Park of southern Taiwan. Zool. Stud. 43: Skierczyński, M. & Osiejuk, T.S. 20. Sharing song types within local dialect does not hinder neighbour-stranger recognition in Ortolan Bunting Emberiza hortulana males. Behaviour 147: Specht, R. 20. Avisoft-CORRELATOR v.3.1. Berlin: Avisoft Bioacoustics. SPSS Inc SPSS Advanced Models Chicago, IL: SPSS Inc. Stoddard, P.K., Beecher, M.D., orning, C.L. & Campbell, S.E Recognition of individual neighbors by song in the song sparrow, a species with song repertoires. Behav. Ecol. Sociobiol. 29: Titus, R.C Short-range and long-range songs: use of two acoustically distinct song classes by dark-eyed Juncos. Auk 11: Wegrzyn, E., Leniowski, K. & Osiejuk, T.S Introduce yourself at the beginning possible identification function of the initial part of the song in the Great Reed Warbler Acrocephalus arundinaceus. Ornis Fenn. 86: Wiley, R.. & Richards, D.G Adaptations for acoustic communication in birds: sound transmission and signal detection. In Kroodsma, D.E., Miller, E.. & Ouellet,. (eds). Acoustic Communicaiton in Birds, Vol. 1: New York: Academic Press. Received July 2011; revision accepted 3 October Associate Editor: Lorien Pichegru.