Minimal similarity in songs suggests limited exchange between humpback whales (Megaptera novaeangliae) in the southern Indian Ocean

Abstract

Comparing humpback whale song from different breeding assemblages can reveal similarities in song due to acoustically interacting males, and therefore indirectly test whether males from different breeding sites are mixing. Northern Hemisphere song comparisons illustrated that whales within ocean basins share similar songs and are subpopulations within a larger population, whereas whales in different ocean basins are isolated populations and therefore do not share songs. During the 2006 breeding season, recordings were collected in Madagascar and Western Australia, and were compared visually plus aurally. Both regions shared one theme, whereas each region had four and six private themes, respectively. This study had a substantially low number of shared themes. The co-occurrence of one theme was interpreted as an indication of limited exchange between these breeding assemblages, and we speculate that limited song similarity is due to inter-oceanic interactions. Male(s) from an Indian Ocean breeding group could be exposed to novel song when they geographically overlap, and acoustically interact, with males from a different ocean basin. Novel song could induce rapid temporal changes as new song content is incorporated, thereby minimizing song similarities between that breeding group and other Indian Ocean breeding groups that were not exposed to the novel song.

Humpback whale (Megaptera novaeangliae) song is a male acoustic display that can be heard on winter breeding grounds, migratory routes, and summer feeding grounds (Tyack 1981, Darling et al. 1983, Glockner 1983, Tyack and Whitehead 1983, Mattila et al. 1987, McSweeney et al. 1989, Clapham and Mattila 1990, Cato 1991, Clark and Clapham 2004). Song structure is a complex arrangement of sounds that are repeated in a cyclical pattern and can be categorized into an ascending hierarchical series of units, subphrases, phrases, and themes (Payne and McVay 1971, Frumhoff 1983, Payne et al. 1983). Within a breeding area all male humpback whales sing similar song content (i.e., same phrase types and therefore same themes) which is characteristic to that particular breeding season (Winn and Winn 1978). Humpback whale song progressively changes during the breeding season (Payne et al. 1983, Payne and Guinee 1983, Payne and Payne 1985, Cerchio et al. 2001, Erikson et al. 2005). At the same time males are constantly updating their song so that, at any one time during a breeding season, an individual's song more closely resembles the song of whales around it than its own songs from previous years or months (Guinee et al. 1983). Such a synchronous change in the song of male humpback whales, strongly suggests that whales are acquiring these changes through vocal learning, and therefore similarities in song content are culturally transmitted between individuals in acoustic contact while geographically overlapping (Payne and Guinee 1983, Payne and Payne 1985, Noad et al. 2000, Cerchio et al. 2001). It is this characteristic of song that allowed researchers to use song comparisons as a tool to infer an exchange of male(s) between different breeding assemblages (Winn et al. 1981, Payne and Guinee 1983, Cato 1991, Gill et al. 1995, Helweg et al. 1998, Cerchio et al. 2001, Darling and Sousa-Lima 2005).

There are three proposed hypotheses concerning the mechanism by which cultural transmission, via vocal learning between males from different breeding assemblages, may occur at some point along the migratory cycle (Payne and Guinee 1983). One, mixing may occur on the feeding grounds or during migration. Two, mixing may occur in subsequent winters if some males migrate to different breeding grounds between breeding seasons. Three, mixing may occur within a breeding season if some males visit different breeding grounds within a single season.

In the Northern Hemisphere, intra-oceanic song comparisons demonstrated that males within the same ocean basin sing a similar song (i.e., songs from different breeding sites contain all the same themes and thematic order) that undergoes synchronous changes within a breeding season and over several years. Therefore, the authors concluded that these whales are a single population with each breeding site constituting different subpopulations (Winn et al. 1981, Payne and Guinee 1983, Cerchio et al. 2001). In contrast, inter-oceanic song comparisons demonstrated that breeding assemblages from different ocean basins and/or different hemispheres are geographically isolated populations whose songs do not share any themes. Therefore, their songs have the same hierarchical structure but contain very different content (Winn et al. 1981, Payne and Guinee 1983).

The majority of intra-oceanic song comparisons conducted in the Southern Hemisphere have agreed with the intra-oceanic Northern Hemisphere song comparisons reviewed above, mainly that breeding assemblages within an ocean basin sing similar songs. A song comparison of Eastern Australia and New Caledonia, breeding sites within the South Pacific, found a complete overlap in song themes during a single breeding season and the authors cited this song similarity as evidence of migratory exchange between these two subpopulations (Gill et al. 1995). In the South Atlantic, song from breeding assemblages in Brazil and Gabon contained all the same themes during two breeding seasons; the authors cited this as evidence of acoustic interaction between these two breeding assemblages but made no definitive conclusions on population structure (Darling and Sousa-Lima 2005).

Initial inter-oceanic song comparisons of the Indian Ocean (Western Australia) and the South Pacific (Eastern Australia) breeding assemblages did not find any shared themes, and thus complemented the results of inter-oceanic song comparisons conducted in the Northern Hemisphere (i.e., that inter-oceanic whales sing very different songs) (Cato 1991, Dawbin and Eyre 1991). Cato (1991) suggested that geographical barriers between breeding grounds and migration routes, rather than the separation distance, created the isolation necessary for these two breeding groups to sing different songs, even though mixing may have occurred on the feeding grounds. Dawbin and Eyre (1991) came to a similar conclusion, citing the difference in song between Eastern and Western Australia as evidence that these two breeding assemblages are isolated from one another.

There are some song comparisons from the Southern Hemisphere, both inter- and intra-ocean comparisons, which contradict the Northern Hemisphere song comparisons reviewed above. A second song comparison of South Pacific breeding sites, which included song from Eastern Australia, New Caledonia, and Tonga, found seven shared themes out of a total of nine themes; therefore there was not a complete overlap in themes as predicted by the Northern Hemisphere song comparisons (Helweg et al. 1998). A more recent study involving Eastern and Western Australian whales found a complete replacement of Eastern Australia song with Western Australia song within a span of 2 yr (Noad et al. 2000). This transformation in song was presumably due to the introduction of Western Australian song into the Eastern population by Western Australian male(s), indicating an exchange of individual(s) between two populations that were considered isolated based on the Cato (1991) and Dawbin and Eyre (1991) studies. Another inter-oceanic exception involved breeding assemblages from Madagascar (Indian Ocean) and Gabon (South Atlantic). In 2003 song recorded at these two breeding sites contained all the same themes sung in the same thematic order (Razafindrakoto et al. 2009).

In this study, humpback whale songs recorded in the coastal waters of Madagascar and Western Australia were analyzed to determine the number of shared theme(s) between these two breeding sites. The amount of shared song content was used to make inferences about the degree to which these two breeding assemblages acoustically interact, via an exchange of male(s) from these different southern Indian Ocean breeding sites. The hypotheses tested and interpretation of these data were based on the assumption that a greater amount of shared song content infers a greater degree of exchange between the breeding sites being compared, and vice versa a smaller amount of shared song content infers a lesser degree of exchange. We believe this assumption is supported by song comparisons conducted in the Northern Hemisphere (i.e.,Winn et al. 1981, Payne and Guinee 1983, Cerchio et al. 2001).

We had three alternative hypotheses based on the results of song comparisons reviewed above. The leading hypothesis was based on Northern Hemisphere intra-oceanic song comparisons; if males from Madagascar and Western Australia have acoustic interactions that result in complete song sharing (i.e., these two regions share all themes and thematic order), then we would infer that these two breeding assemblages may be subpopulations within a single southern Indian Ocean population or metapopulation. Based on the Helweg et al. (1998) and Noad et al. (2000) studies, the second hypothesis was a slight modification of the first; if males from Madagascar and Western Australia have acoustic interactions that result in some shared theme(s), then we would conclude that these two breeding assemblages have limited exchange indicating some connectivity between Madagascar and Western Australia whales. Conversely, the third hypothesis was based on the results of Northern Hemisphere inter-oceanic song comparisons; if males from Madagascar and Western Australia do not acoustically interact, then song from these two regions would not share any themes inferring these two breeding assemblages are completely isolated populations.

Methods

Humpback whale song was recorded in the southern Indian Ocean at five sites during the 2006 winter breeding season (Fig. 1). Song was collected from July to September at two research sites in Madagascar: Antongil Bay and Anakao. In Antongil Bay, focal animal recordings were collected using custom-built hydrophones with preamplifiers connected to a Sony PCM-M1 or TCD-D100 digital audio tape (DAT) deck, using a sampling rate of 44.1 kHz. The system frequency responses for the Sony PCM-M1 and TCD-D100 were flat ± 3 dB from 50 to 17,000 Hz. In Anakao focal animal recordings were collected using a custom-built hydrophone and preamplifier connected to a Marantz PMD671 solid state, stereo digital recorder using a sampling rate of 44.1 kHz. The system frequency response for the Marantz was flat ± 0.5 dB from 20 to 20,000 Hz.

Figure 1. The research sites for Madagascar and Western Australia; Madagascar song was recorded in Antongil Bay and Anakao. Western Australia song was recorded in Perth, Exmouth Gulf, and Pender Bay.

In Western Australia song was collected from September to October at three research sites: Pender Bay, Perth, and Exmouth Gulf. In Pender Bay and Perth, focal animal recordings were collected using a sampling rate of 44.1 kHz with equipment similar to the type used in the Antongil Bay recordings. The Exmouth Gulf data were collected using a sea noise logger (http://www.cmst.curtin.edu.au/products) that consisted of a Hi Tek HTIU90 hydrophone external to a housing containing a sea noise data logger. The sea noise logger was deployed at 9 m in 15 m of water, and left unattended while recording 6 min and 48 s of sound every 15 min. A section with a high signal-to-noise ratio of humpback whale song was supplied for analysis. This section totaled 4 h. The Exmouth Gulf data were recorded using a sampling rate of 12 kHz which had a flat, calibrated frequency response from 20 Hz to 5 kHz.

In Madagascar individuals were confirmed using photo-identification of individually unique pigmentation patterns found on the underside of flukes, and/or the shape and unique scarring pattern of dorsal fins (Katona et al. 1979). In Antongil Bay one individual was recorded twice, with a week between recordings (individual AB04) (Table 1). There was no observable difference in the singer's themes (i.e., phrase structure and patterning) and the thematic order of its songs, so the recordings were combined. There were four Madagascar recordings with no accompanying photographs, three in Antongil Bay and the Anakao recording. Because of its location, it is highly unlikely that the Anakao recording was an individual also recorded in Antongil Bay. Three unidentified singers from Antongil Bay could potentially be resampled individuals. The probability of rerecording identified individuals was 0.09 (one resampled individual out of 11 independent samples). Because the calculated probability of resampling individuals is low, and may be even lower due to differences in whale behavior, the unidentified singers were not considered a significant source of bias (Cerchio et al. 2001).

Table 1.  Madagascar and Western Australia song recordings. In Madagascar a total of 15 individuals were recorded. Individual AB04 was recorded twice; the two recordings are denoted as AB04a and AB04b. In Western Australia a total of four individuals, including the Exmouth Gulf data logger recordings as one “individual,” were recorded.

Location	Date	Individuals	Duration (h:min:s)	Song cycles	Song fragments
Antongil Bay	20 July 2006	AB01	1:31:00	16	3
Antongil Bay	21 July2006	AB02	0:43:21	6	2
Antongil Bay	24 July 2006	AB03	1:06:00	5	4
Antongil Bay	26 July 2006	AB04a	1:05:00	3	2
Antongil Bay	2 August 2006	AB04b	2:08:00	5	4
Antongil Bay	23 August 2006	AB05	1:27:48	6	2
Antongil Bay	24 August 2006	AB06	0:25:13	2	2
Antongil Bay	26 August 2006	AB07	0:53:07	1	2
Antongil Bay	26 August 2006	AB08	0:47:57	2	2
Antongil Bay	27 August 2006	AB09	1:12:07	4	2
Antongil Bay	29 August 2006	AB10	0:42:24	2	2
Antongil Bay	29 August 2006	AB11	0:28:42	2	2
Antongil Bay	29 August 2006	AB12	0:12:28	1	2
Antongil Bay	1 September 2006	AB13	0:29:04	1	2
Antongil Bay	4 September 2006	AB14	0:25:09	1	2
Anakao	17 September 2006	AN01	0:46:00	1	2
Madagascar totals	13 d	15	14:23:20	58	37
Pender Bay	9 September 2006	PB01	0:55:00	2	2
Perth	17 September 2006	PE01	0:14:00	0	1
Perth	25 September 2006	PE02	0:28:00	1	1
Exmouth Gulf	17 October 2006	ExDatalogger	2:16:20	0	25
Western Australia totals	4 d	4	3:53:20	3	29

In Western Australia individuals were not confirmed using photo-identification, but because the recordings were greater than 24 h apart and/or from different locations it was assumed that each focal recording represented different individual whales. Due to the recording methodology used in Exmouth Gulf the number of actual individuals recorded by the data logger during the 4 h deployment period is unknown. There was no sufficient way to determine the number of individuals from the recordings; therefore the best course of action was to treat the recordings as one “individual.” This was preferable to dividing the data logger recordings into several “individuals,” which could potentially create nonindependent samples if consecutive recordings were actually from one individual.

To determine the qualitative song structure, printed spectrograms of recordings from Madagascar (Antongil Bay and Anakao) and Western Australia (Perth and Pender Bay) were made using Avisoft Saslab Bioacoustic software (http://www.avisoft.com), with an analysis band range of 0–8000 Hz (sampled at 16 kHz) and FFT size of 512 points. The phrase types and the sequence of themes were identified by visual inspection of the printed spectrograms, and by visual and aural inspection using Raven 1.2.1 Bioacoustic Software (Bioacoustic Research Program, Cornell Lab of Ornithology, Ithaca, NY) with an analysis band range of 0–8000 Hz, FFT of 512 and a 90% overlap. The Exmouth Gulf recordings were not printed out, because of the sampling scheme used to create these recordings. Instead Raven 1.2.1 was used to visually and aurally inspect these recordings using an analysis band range of 0–3000 Hz, an FFT size of 512, and a 90% overlap. The first author (AM), in consultation with the second author (SC), identified the phrase types and themes for Madagascar first and then Western Australia. The Western Australia themes were labeled to reflect any overlap with the Madagascar 2006 song. Thus, one theme found in both regions retained the name used in the Madagascar analysis, whereas, themes unique to Western Australia were given names sequential to the last theme identified in the Madagascar song. For each individual, using Raven 1.2.1, the theme durations were measured to calculate the theme proportions.

Results

In Madagascar a total of 15 individuals were recorded generating 16 recordings for a total of 14.38 h of recordings, 58 song cycles, and 37 song fragments for analysis (Table 1). Fourteen individuals were recorded in Antongil Bay and one was recorded in Anakao. In Western Australia a total of four individuals, including the Exmouth Gulf data logger recordings as one “individual,” were recorded, generating 22 recordings, 5.88 h of recordings, 3 song cycles, and 29 song fragments for analysis (Table 1). One individual was recorded in Pender bay, two were recorded in Perth, and the fourth “individual” was recorded in Exmouth Gulf.

Madagascar and Western Australia Private Themes

The Madagascar and Western Australia 2006 songs had a total of ten private themes, meaning these themes were only found in either Madagascar or Western Australia, but not in song from both regions. Madagascar 2006 song consisted of four private themes (A, C, D, and E), each with its associated phrase type of the same label, and three transitional phrases (BA, CD, and DC) (Fig. 2). The Western Australia 2006 song consisted of six private themes (BF, F, G, H, I, and J), each with its associated phrase type of the same label, and two phrase type variants in Theme G (Phrase G1 and Phrase G2) (Fig. 3).

Figure 2. Madagascar had four private themes: A, E, C, and D plus three transitional phrases: BA, CD, and DC. Spectrograms were made using Raven 1.2.1 with an FFT of 512 and an overlap of 90%.

Figure 3. Western Australia had seven private themes: BF, F, G, H, I, and J. Spectrograms were made using Raven 1.2.1 with an FFT of 512 and an overlap of 90%.

There are three types of themes found in humpback whale song: static, shifting, and unpatterned (Payne and Payne 1985). Madagascar Themes A, C, D, and E and Western Australia Themes BF, F, H, I, and J were static themes, meaning the phrase structure remained constant with each repetition (Fig. 2, 3). Western Australia Theme G was a shifting theme, meaning the units underwent a transformation with each successive phrase repetition (Fig. 3).

Between two themes there may be a transitional phrase that combines subphrases from the preceding and subsequent themes (Payne and McVay 1971, Payne and Payne 1985). The Madagascar 2006 song contained three transitional phrases: Phrase BA, Phrase CD, and Phrase DC (Fig. 2). Transitional Phrase BA is a combination of subphrases from phrase types found in Themes A and B, with the first subphrase taken from Phrase B1 and the second subphrase from Phrase A. The transitional Phrase CD consisted of the first subphrase from Phrase C, and the second subphrase from Phrase D. Transitional Phrase DC is the opposite combination, with the first subphrase coming from the Phrase D and the second subphrase from Phrase C. These two transitional phrases were always found between Themes C and D.

Western Australia Theme BF is a combination of Western Australia Theme B and Theme F (Fig. 3). Theme BF consisted of two or more repetitions of Phrase BF. Transitional phrases are only sung once between the two themes whose subphrases are combined to make the transitional phrase, therefore Theme BF was considered a combination theme rather than a transitional phrase (Payne and McVay 1971, Payne and Payne 1985).

In both regions, there were dominant themes that were sung for >20% of the song, and there were minor themes that were sung for <20% of the song (Fig 4A, B). The dominant themes in the Madagascar song were Themes D and A, and the minor themes were Themes E and C. In the Western Australia song the dominant themes were Themes G and I, and the minor themes were Themes BF, F, H, and J.

Figure 4. The relative theme representations for Madagascar (A) and Western Australia (B) 2006 songs; dominant themes were sung for >20% of the song and minor themes were sung for <20% of the song. The dominant Madagascar themes were Themes B, A, and D, and the minor themes were Themes E and C. Western Australia dominant themes were Themes B, G, and I and the minor themes were Themes BF, F, H, and J. Theme B was a dominant theme in both regions.

Madagascar and Western Australia Shared Theme

Theme B was the only theme found in both Western Australia and Madagascar during the 2006 breeding season, and was a dominant theme in both regions (Fig. 4, 5). Theme B from the Madagascar 2006 song and Theme B from the Western Australia 2006 song are considered the same theme, because the two themes had similar acoustical characteristics placed in the same positions in the three phrase type variants that comprise Theme B. In both regions Theme B was a shifting theme with phrases that undergo a similar transformation in units, although with small-scale exceptions to their overall similarity.

Figure 5. Shifting Theme B from Madagascar (right) and Western Australia (left); Phrases B1, B2, and B3 are phrase type variants that represent the unit transformation. Theme B from both regions was considered the same theme, because of similar acoustical characteristics placed in the same positions, and similar unit transformation. Spectrograms were made using Raven 1.2.1 with an FFT of 512 and an overlap of 90%.

In Madagascar and Western Australia, Theme B was comprised of three phrase type variants: Phrases B1, B2, and B3 that reflect three stages of unit transformation within Theme B. In the Madagascar song, the first subphrase of Phrase B1 had a single, short (1–2 s) broadband amplitude modulated (AM) unit that was typically repeated three times before gradually merging to form one long (>5 s) broadband AM unit in Phrase B2. In Phrase B3 this unit has transformed into a harmonic unit. In the second subphrase of Phrase B1 there are typically two high frequency ascending units that progressively curved over during Phrase B2 to become a series of four ascending then level units in Phrase B3.

In Western Australia the first subphrase of Phrase B1 had a single ascending unit rather than the broadband unit used in Madagascar. But like the Madagascar song this unit was repeated, typically six times, before gradually merging to form one harmonic unit in Phrase B2 and Phrase B3. Similar to Madagascar, the second subphrase of Western Australia Phrase B1 typically had two high frequency ascending units that progressively curved over during Phrase B2. However, these units transformed into a series of four ascending then descending units in Phrase B3, rather than the ascending then level units found in Madagascar Phrase B3.

Discussion

Based on the majority of intra-ocean song comparisons conducted so far, one would expect to find a complete sharing of song between Madagascar and Western Australia 2006 songs (i.e., song from both regions have all the same themes and thematic order). Contrary to the expected, our study found that the two regions had only one theme in common, Theme B. Between Madagascar and Western Australia there were a total of eleven themes, and ten of these themes were unique to either region; Madagascar had four: A, E, C, and D and Western Australia had six: BF, F, G, H, I, and J. There was no overlap in content or structure for the phrase types found in these unique themes.

The song differences between Madagascar and Western Australia would imply that these two assemblages are predominantly isolated from one another throughout the migratory cycle and may constitute separate populations. However, the presence of Theme B as a dominant theme in both songs, along with the similarity in phrase structure and unit transformation within Theme B, can be taken as evidence of cultural transmission between Madagascar and Western Australia. This type of cultural transmission can only occur via vocal learning between individuals that are in acoustic contact while geographically overlapping. Therefore, the shared theme between Madagascar and Western Australia is an indication of some connectivity between these breeding assemblages.

A quantitative measure of the exchange needed for complete song sharing vs. limited song sharing is still unknown, and whether an exchange of males occurs during the feeding season, on migration routes, or when males switch breeding grounds between years is unclear. Therefore, the following interpretations of the limited song similarity found in this study are speculative, and based upon the current understanding of song dynamics from the existing literature. Northern Hemisphere song comparisons along with various studies involving Southern Hemisphere humpback whales, in conjunction with this study, can be used to formulate conjectures about the location of cultural transmission, the degree of exchange, and a population structure that would result in limited song similarity. We propose the following interpretations with the hope that these ideas will spur further research into the relationship between song and population structure within the Southern Hemisphere.

Interpretations of Limited Song Similarity

One possible location for cultural transmission of song content between Madagascar and Western Australia males is the Antarctic feeding grounds. A genetic analysis of whales in different Antarctic feeding areas suggested that whales from Madagascar and Western Australia could be mixing on the feeding grounds (Loo et al. 2007). Singing is heard on Northern Hemisphere summer feeding grounds, and therefore presumably also occurs in Southern Hemisphere feeding grounds (Mattila et al. 1987, McSweeney et al. 1989, Clark and Clapham 2004). One speculation is that some number of males from Madagascar and Western Australia could potentially be in acoustic contact during the feeding season and during this contact cultural transmission in song content would occur. However, there is currently no direct evidence that this is the point during the migration cycle when cultural transmission of song content occurs between the Madagascar and Western Australia breeding assemblages. Further studies are necessary to confirm this supposition and to determine the amount of song content, if any, which could be transmitted between males on a feeding ground.

We interpret the minimal song similarity found between Madagascar and Western Australia songs as resulting from a limited degree of exchange between these breeding groups. Northern Hemisphere song comparisons (i.e.,Winn et al. 1981, Payne and Guinee 1983, Cerchio et al. 2001) have demonstrated that subpopulations within the Atlantic or Pacific share the same song due to an exchange of males between these subpopulations, but between the Atlantic and Pacific oceans males are completely isolated and therefore sing completely different songs. From these song comparisons we believe it is reasonable to assume that the level of song similarity found between different breeding sites is reflective of the degree of exchange between those breeding sites. If this assumption were correct then extensive song sharing would infer a relatively larger degree of exchange between breeding sites, than a minimal amount of song sharing which would imply a relatively smaller degree of mixing.

The shared content between Madagascar and Western Australia 2006 songs was unusually low (Table 2). In the Northern Hemisphere, the South Atlantic, and the South Pacific, within each ocean basin, geographically separated breeding assemblages either completely or almost completely shared song (Winn et al. 1981, Payne and Guinee 1983, Gill et al. 1995, Helweg et al. 1998, Cerchio et al. 2001, Darling and Sousa-Lima 2005). In contrast, relative to these other populations, males in the southern Indian Ocean are apparently, based on this (one season) comparison, not interacting with one another to the same degree, as indicated by the relatively low level of shared song content (i.e., only one shared theme out of 11). Therefore, the shared theme between Madagascar and Western Australia suggests a limited exchange between these breeding sites. A limited exchange of males implies a complex population structure that would be difficult to define with a single year song comparison.

Table 2.  Summary of intra-ocean song comparisons conducted in the Northern and Southern Hemispheres. Of all the intra-ocean song comparisons conducted to date this study found the least amount of shared song content.

Northern Hemisphere song comparisons
North Pacific			North Atlantic
Breeding sites	Shared all themes (Yes/No)	Authors	Breeding sites	Shared all themes (Yes/No)	Authors
Mexico vs. Hawaii	Yes	Winn et al. (1981)	West Indies vs. Cape Verde	Yes	Winn et al. (1981)
Mexico vs. Hawaii	Yes	Payne and Guinee (1983)
Mexico vs. Hawaii	Yes	Cerchio et al. (2001)

Southern Hemisphere song comparisons
South Pacific			South Atlantic
Breeding Sites	Shared all themes (Yes/No)	Authors	Breeding Sites	Shared all themes (Yes/No)	Authors
New Caledonia vs. East Australia	Yes	Gill et al. (1995)	Gabon vs. Brazil	Yes	Darling and Sousa-Lima (2005)
New Caledonia vs. East Australia vs. Tonga	No (7 out of 9)	Helweg et al. (1998)

One plausible scenario that would create relatively low levels of shared song between Madagascar and Western Australia, when compared to other intra-ocean song comparisons, is that males from the southern Indian Ocean are acoustically interacting more extensively with males from different ocean basins (i.e., with the South Pacific in the Eastern Indian Ocean and with the South Atlantic in the Western Indian Ocean). With only 1 yr of data, it is impossible to know whether this is a stable state or variable across years. Yearly fluctuations in the acoustic interaction between inter- and intra-oceanic breeding groups, and therefore yearly fluctuations in song similarity, could be the result of a dynamic population structure. Southern Hemisphere males from both inter- and intra-oceanic breeding groups may be acoustically interacting to varying degrees, potentially creating years when inter-oceanic song comparisons would find a greater overlap in song content than intra-oceanic comparison, and vice versa. In certain years males from breeding sites on the coasts of Africa (i.e., Madagascar and Gabon) or males from the Australian coasts (i.e., Eastern and Western Australia) may be interacting with one another to a greater extent than males from breeding sites within the Indian Ocean. In other years males from Madagascar and Western Australia may be interacting with each other rather than with males from Gabon or Eastern Australia. If the level of exchange varies from year to year then there may be years in which Madagascar and Western Australia share very little in terms of song content, compared to years with a greater overlap in song content.

Inter-oceanic interactions are possible in the Southern Hemisphere, unlike the Northern Hemisphere where humpbacks in the North Pacific and North Atlantic are completely isolated from one another due to land mass barriers. A greater, or equal, amount of exchange between inter-oceanic breeding assemblages rather than intra-oceanic assemblages could potentially expose males from Madagascar and/or Western Australia to novel song from the South Atlantic or Pacific, respectively. The introduction of a novel song type could induce rapid temporal changes as the new song content is incorporated, causing a dramatic shift in song (e.g.,Noad et al. 2000). A dramatic shift in song at one breeding site could reduce the amount of song similarities found when that breeding site is compared to other breeding sites within the same ocean basin that were not exposed to the novel song type. Two phenomena would be evident if inter-oceanic interactions were occurring in the southern Indian Ocean: (1) inter-oceanic migration of males and (2) rapid temporal changes in song between breeding seasons.

Evidence for inter-oceanic migration can be found in both the Madagascar and Western Australian breeding groups. Chittleborough (1965) reported evidence of migration from the Pacific Ocean to the Indian Ocean, when two whales originally marked with discovery tags off the coast of Eastern Australia were harvested off the coast of Western Australia. Pomilla and Rosenbaum (2005) documented an inter-oceanic migration, of one male, from the Indian Ocean to the Atlantic Ocean. Noad et al. (2000) observation of a complete replacement of Eastern Australian song with Western Australian song also suggests an inter-oceanic migration of adult male(s) from the Indian Ocean to the Pacific Ocean (the opposite direction of Chittleborough 1965). Song sharing between breeding sites in Madagascar and Gabon indicates that males from these breeding groups, in different ocean basins, are in acoustic contact at some point in the migratory cycle (Razafindrakoto et al. 2009).

Whales in Western Australia have exhibited unusually rapid temporal changes in song. Dawbin and Eyre (1991) tracked changes in theme composition within the Western Australian song over a span of 5 yr; the 1986 and 1987 songs had only one theme in common, and the 1987 and 1988 songs had no overlapping themes. However, between 1988 and 1990 the songs had the same four themes. Thus, it appears that during some years the Western Australian song may undergo a year of rapid temporal change in song and in other years exhibit a more progressive temporal change in song.

The quantity of exchange, the likely path and direction of exchange, and how it alters from year to year could not be determined by this study. Further studies identifying temporal and geographical variation in song from multiple ocean basins over multiple years would likely identify in more detail the mechanisms of exchange between inter- and intra-oceanic breeding assemblages in the Southern Hemisphere. Such studies may clarify if whales within the southern Indian Ocean are consistently more isolated from one another than whales within other ocean basins, or whether there is a dynamic relationship between whales in Madagascar and Western Australia that varies across years.

Footnotes

• 1

  Current affiliation: Cetacean Ecology and Acoustics Laboratory, School of Veterinary Science, University of Queensland Gatton, Queensland 4343, Australia.

• 2

  Current affiliation: Animal Biology Department, Antananarivo University, Antananarivo, Madagascar.

Acknowledgments

We are grateful to the Wildlife Conservation Society in New York, Wildlife Conservation Society Madagascar Country Program, Madagascar Masoala National Park, and the American Museum of Natural History for the institutional support provided in both the field and laboratory. We are also grateful to the Madagascar field crew, in particular Matt Leslie, Norbert Andrianarivelo, and the Malagasy mariners: Maro José, Raoul Jaozandry, Yves, Philbert, and Sandra Veloharisoa for their support during the Madagascar 2006 field season. We would also like to thank three anonymous reviewers for their comments, which greatly improved this manuscript.