Gender identification using acoustic analysis in birds without external sexual dimorphism

Acoustic terms and definitions

Sexing monomorphic birds by voice is mainly based on the analysis of computer images of vocalizations (Fig. 1). Spectrograms reflect changes of call frequency characteristic over time and power spectra, reflecting temporal changes in amplitude. The visual analysis of spectrograms and power spectra is much more reliable and unbiased in comparison to sexing by ear, so it is applicable to a broader range of bird species.

Applicability of sexing by voice

Sexing by voice is primarily applicable for bird species for which this method unambiguously classifies the given individual as a male or a female, i.e., it is 100 % reliable. At non-overlapping values of one or a few acoustical variables between sexes (e.g. when male calls are always higher in pitch than female calls, without exclusions), individuals may be sexed with 100 % reliability. Just visual inspection of spectrogram without measuring the acoustic variables would be sufficient in these cases (e.g. Volodin et al. 2009). However, when acoustic variables overlap between sexes, some individuals may be sexed incorrectly. In this case, the percentage of correct classification to sex may be calculated with discriminate function analysis (DFA). DFA calculates the percentage of correct assignment to sex based on the set of measured acoustic variables introduced into the analysis. Given that the number of groups for classifying sex with DFA is only two (males and females), we consider that a <70 % correct classification would make sexing by voice impossible, 70–80 %—low reliability of sexing, 81–90 %—high but imperfect reliability of sexing and 91–100 %—perfect reliability of sexing.

Order Apterygiformes

In the Little Spotted Kiwi (Apteryx owenii), loud whistle calls differ substantially between the sexes (Digby et al. 2013). The average fundamental frequency of males was 1.5 times higher than those of females (2800 and 1800 Hz, respectively). Although these birds have some dimorphism of body size, with males slightly lighter in weight than females (1.2 and 1.4 kg, respectively), distant sexing by far-propagating whistle calls is important for the census of this species in the wild.

Order Anseriformes

Among Anseriformes, sex differences in calls were studied in detail in four species, in which sex differences could not be based on behavior, appearance, nest building, incubation and parental care. Whistling ducks were named after their loud whistle calls, produced by both sexes during the year (Volodin et al. 2005a, 2009). In White-faced Whistling Ducks (Dendrocygna viduata), the maximum fundamental frequency of male loud whistles was always substantially lower than in females (Fig. 2). The values were non-overlapping between sexes, being <4500 Hz in males and >5300 Hz in females. After short training, sexes could be easily discriminated by ear (Volodin et al. 2003).

In Fulvous Whistling Ducks (D. bicolor), loud male whistles were always lower in fundamental frequency, than female calls (Fig. 3). In males, the fundamental frequency was <2100 Hz, whereas in females it was >2800 Hz. It is noteworthy that in Cuban Whistling Ducks (D. arborea), loud whistles by males were higher in fundamental frequency than those of females; in addition, male calls often contained a second fundamental frequency in their spectra, which resulted in biphonation (Fig. 4). In males, this fundamental frequency was >2600 Hz, whereas in females it was <2550 Hz. In loud whistles of Red-billed Whistling Ducks (D. autumnalis), the ranges of male and female fundamental frequency overlapped, although the duration of notes at the end part of calls was always longer in males (>0.13 s) than in females (<0.12 s) (Fig. 5). In addition, male calls contained biphonation, which was never found in female calls. In all four species of whistling ducks, the non-overlapping values of call variables provided the possibility of error free sexing by a single loud whistle (Volodin et al. 2009). As well as spontaneous vocal emission, playbacks of species-specific calls of whistling ducks evoked immediate vocal responses of loud whistles that revealed sex differences (Volodin et al. 2005b).

No sex difference was found in two-syllable calls of the Red-breasted Goose (Branta ruficollis) (Fig. 6). The percentage of correctly classifying calls to sex with DFA (87 %) did not differ from random values (79 %) (Volodin et al. 2008). However, in pair duets, produced during triumph-ceremony displays, the maximum fundamental frequency of female calls was always lower than in male calls; in addition, in duets, females always produced series of two-syllable calls, whereas males produced series of one-syllable calls (Fig. 6) (Volodin 1990b). The definitive triumph-ceremony (for describing the triumph-ceremony see e.g. Radesäter 1974; Volodin 1990a) is displayed only by mates of heterosexual pairs, whereas in homosexual male–male pairs only one of the males vocalizes, whereas the second male is silent (Volodin 1990c). Thus, birds of this species may be sexed by their parties in duets.

In another waterfowl species, the White-fronted Goose (Anser albifrons), female calls during triumph-ceremony were also lower in fundamental frequency compared to male calls (Krechmar 2003). At the same time, it has been reported that in duets of the Horned Screamer (Anhima cornuta), vocalization is lower and harsher in the male than in the female (Gill et al. 1974). The scarcity of data precludes any general conclusions concerning the structures of male and female duet parties in Anseriformes.

Order Galliformes

The Willow Ptarmigan (Lagopus lagopus) represents a single monogamous species of the Tetraonidae family, where mates share parental care; the male guards the territory and the female guides the chicks (Martin and Cooke 1987). This species is practically monomorphic, males being slightly larger than females and differing slightly from females in coloration during spring. Vocal repertoire comprises 11 call types, produced by both sexes (Martin et al. 1995). Male calls are distinguished by their strong amplitude modulation, which results in wideband spectra and harsh sounds. Female calls are distinguished by their tonal structure. These differences are revealed in all call types, in particular in the “krrow” call, emitted by both sexes during territorial conflicts of pairs and during vocal interactions between parents and chicks.

Order Podicipediformes

In the Black-necked Grebe (Podiceps nigricollis), nesting in dense vegetation, sex dimorphism is expressed in advertising calls (Nuechterlein and Buitron 1992). Male advertising calls were lower at the start and maximum fundamental frequency and were longer compared to female advertising calls. Although the dominant frequency values did overlap between sexes, playback experiments of advertising calls to bachelor males have demonstrated that they responded to female calls and ignored male calls.

Order Columbiformes

In Orange-bellied Fruit-doves (Ptilinopus iozonus), male coo calls were always lower in frequency and longer than female coo calls (Baptista and Gaunt 1997). These vocal features were used for mating formation of captive pairs for conservation purposes, since this insular species has disappeared in its natural settings. As a result of breeding success, this species has been released back into the wild. In addition, the lower-frequency male calls compared to female calls have been reported for some other doves (other five species of genus Ptilinopus, as well as Columba livia, C. squamosa, Zenaida graysoni, Tutur tympanistra, Treron vernans, Phapitreron leucotis) (Baptista and Gaunt 1997); however, measuring acoustic variables has not been conducted for these species.

In the Eurasian Collared Dove (Streptopelia decaocto), male coo calls were also at lower-frequency compared to female coos; however, sexing by this variable was unreliable due to slightly overlapping values between the sexes (Ballintijn and ten Cate 1997). Nevertheless, the birds could be sexed with 100 % reliability by the number of coos per sequence (from 6.3 to 11.3 in males and from 3.3 to 4.6 coos in females).

Order Caprimulgiformes

Among different calls of vocal repertoire of the Marbled Frogmouth (Podargus ocellatus), sex-related differences were found only in the gobble call, which both sexes produce throughout the year at territorial conflicts (Smith and Jones 1997). The gobble calls consist of low-frequency elements, repeated with intervals less than one tenth of a second. Male calls contained more elements compared to female calls (5.8 compared to 4.3 on average); their elements were shorter and were of a noticeably lower-frequency than in females, although the ranges of frequency and duration overlapped between sexes.

Order Cuculiformes

In studying the breeding system of the Pheasant Coucal (Centropus phasianinus), sex differences were investigated in their far-distant descending “whoops” calls and “scale” calls, produced separately or in duets (Maurer et al. 2008). Although males are lighter in weight than females (300 and 400 g, respectively), both sexes are colored similarly. Male calls were significantly higher-pitched than female calls for both studied call types. Although these authors did not provide the overall ranges of values, the average values of the dominant frequency of male calls were 1.5 times higher than in females, in both the descending whoops call (456 and 339 Hz, respectively) and in the scale call (503 and 314 Hz, respectively).

Order Gruiformes

Among Gruiformes, the possibility of reliable sexing by crowing calls has been reported for a Rallidae family species, the Purple Swamphen (Porphyrio porphyrio) (Clapperton 1983). The crowing call represents the territorial call, emitted by males and females at any time of day and night. Male calls are somewhat lower in fundamental frequency and purely tonal, whereas in females these calls are higher and modulated in amplitude at their end part. Due to these peculiarities in the structure, female calls sound more harsh and guttural. After training, these differences allow sexing by ear in the field (Clapperton 1983). Although males are somewhat larger than females, identifying sexes by morphometry was not possible (Craig et al. 1980).

In another Rallidae species, the American Coot (Fulica americana), the fundamental frequency of male calls exceeded twice that of female calls (Cosens 1981). Independently from a call type, the maximum fundamental frequency was always >1000 Hz in male calls and <1000 Hz in female calls, which allowed for identification of bird sexes by ear. In this species, external sexual dimorphism is lacking, although males, on average, are 7 % heavier than females. Probably, sex differences in calls result from differences in sizes and shape of male and female syringes (Gullion 1950), given that females are larger than males in this species.

For the Gruidae family, the problem of sexing is very important, for many cranes are endangered and require intensive captive breeding for their conservation. Sexing of the endangered Whooping Crane (Grus americana) is based on the maximum fundamental frequency of the guard call, exceeding 1100 Hz in females (on average, 1115 Hz) and <1000 Hz in males (on average, 946 Hz) (Carlson and Trost 1992). The value of correct classification to sex with DFA for 141 males and 118 females was 99 %, with only 3 of the 259 birds sexed incorrectly.

For White-naped Cranes (Grus vipio), guard calls do not allow reliable sexing (Bragina and Beme 2013). Although the fundamental frequency of female calls (on average, 1170 Hz) was noticeably higher than in male calls (on average, 790 Hz), their ranges substantially overlapped between sexes (790–1350 Hz for females and 600–1110 Hz for males, respectively) (Fig. 7). However, the sex of White-naped Cranes could be reliably identified by their duet parties, consisting of repeated syllables (Archibald 1976; Swengel 1996). Each duet syllable always started with a long male call followed with one to three short female calls (Fig. 7).

For Siberian Cranes (Grus leucogeranus), reliable acoustic indices of sex were not found at any call type (Bragina and Beme 2007). Although average values of fundamental frequency were higher in females in all call types, the range of frequencies substantially overlapped between sexes, precluding reliable sexual identification of any particular bird. The best discriminating ability was found in trills (Fig. 8), in which the fundamental frequency of females was not lower than 810 Hz (average 950 Hz) whereas for males, their fundamental frequency was not higher than 860 Hz (average 700 Hz). In addition, within each of nine pairs, the fundamental frequency of females was always higher than that of males. At the same time, male and female duet parties were indistinguishable, for both male and female produce one long call per each duet syllable (Fig. 8) (Swengel 1996; Bragina and Beme 2010). In duets, the average value of their maximum fundamental frequency was higher in females (1100 Hz) than in males (890 Hz); however the range of values overlapped between sexes (Bragina and Beme 2010).

In Red-crowned Cranes (Grus japonensis), as in other crane species, the fundamental frequency of the female in the duet call (1180 Hz) was higher on average than that of the male (1050 Hz), but bird sex could not be identified by this feature alone due to the broad overlap of their frequencies (Klenova et al. 2008). In addition, in eight of 10 breeding pairs, the fundamental frequency was always higher in female calls than in male calls, whereas in two other pairs it was always lower in females than in males. All the same, sex could be reliably assessed during duets. Duets consisted of repeated syllables, starting with a long (0.46 s) male call followed with one to four short (0.24 s) female calls (Fig. 9).

Order Procellariiformes

Many species of the order Procellariiformes are nocturnal. Their calls are important for recognizing sex. Strong vocal sex differences are typical. In the Leach’s Storm-petrel (Oceanodroma leucorhoa), sex differences were found only in one of three call types, i.e., the chatter call (Taoka et al. 1989a). Male and female chatter calls consist of syllables, differing in patterns of frequency modulation (Fig. 10). Sex-specific differences were found in the central syllables of these calls. The maximum fundamental frequency of the central syllable was always higher than 800 Hz in males (on average, 990 Hz) and lower than 750 Hz in females (on average, 650 Hz). The non-overlapping values between sexes allowed 100 % reliable sexing by voice. Birds sitting in their nesting burrows showed sex-specific responses to playbacks of chatter calls, with males selectively responding to male calls and females selectively responding to female calls.

For Swinhoe’s Storm-petrel (Oceanodroma monorhis), sex differences were found in the flight call, but its acoustic structure is similar to the chatter of the Leach’s Storm-petrel (Taoka et al. 1989b; Taoka and Okumura 1990). Male and female calls differed by their rhythmic structure, which allowed sexes to be distinguished by ear. Birds sitting in their nesting burrows showed sex-specific responses to playbacks of chatter calls, with males selectively responding to male calls and females selectively responding to female calls.

In the Madeiran Storm-petrel (Oceanodroma castro), male calls were tonal and melodious whereas the female calls always contained noise and sounded harsh. The differences were sufficiently strong to allow the sexes to be distinguished by ear (James and Robertson 1985a).

In the Fork-tailed Storm-petrel (Oceanodroma furcated), males produce a call which is lacking in females (Simons 1981). This is a high-frequency whistle, emitted by males when they were near the entrance of a nest burrow or just in the burrow interacting with a female.

The burrow calls of the male Greater Shearwater (Puffinus gravis) were higher in frequency and longer compared to female calls (Brooke 1988). In the Little Shearwater (Puffinus assimilis) (James and Robertson 1985b), male burrow and flight calls were higher in frequency and sounded clearer compared to female calls, which were lower in frequency, more noisy and sounded gnashing. These differences allowed sexing by ear.

In the Manx Shearwater (Puffinus puffinus), as in previously mentioned species of Procellariformes, male calls were higher in frequency and more clearly tonal, whereas female calls were lower in frequency and noisier (Brooke 1978). The male Yelkouan Shearwater (Puffinus yelkouan) also vocalizes at a higher frequency than females, so that birds may be sexed with 100 % reliability using a single acoustic variable (Bourgeois et al. 2007; Cure et al. 2009, 2011). Near their nest burrows, the birds emit calls consisting of two notes, of which one is produced at inspiration and another at expiration. At inspiration, males produce a noisy note, whereas females produce a tonal note. Conversely, at expiration, males produce the tonal note, whereas females produce the noisy note. In male calls, the maximum fundamental frequency of the tonal note is always >740 Hz (on average, 885 Hz) whereas in females <530 Hz (on average, 472 Hz).

Calls of the Balearic Shearwater (Puffinus mauretanicus) are very similar to those of the Yelkouan Shearwater. As in the Yelkouan Shearwater, the maximum fundamental frequency of the tonal note was significantly higher in males than in females (on average, 776 and 430 Hz, respectively) and ranges of values did not overlap between sexes (Cure et al. 2010).

Calls of Cory’s Shearwater (Calonectris diomedea) differed strongly by their shape from two previous species and consisted of four notes (Bretagnolle and Lequette 1990; Cure et al. 2009). Still, male calls again were significantly higher in fundamental frequency than calls of females. The end tonal note was the most convenient for measuring the maximum fundamental frequency: in male calls, it was always >640 Hz (on average, 739 Hz) whereas in female calls, it was always <360 Hz (on average, 291 Hz).

As well, in the Blue Petrel (Halobaena caerulea), male calls were higher in fundamental frequency than female calls (Bretagnolle 1990; Bretagnolle and Genevois 1997). Additionally, the rhythmic structure of calls differed strongly between sexes, since female calls included short notes that were lacking in males. A comparison of male and female calls from 11 populations confirmed the consistency of the overall pattern of male and female calls in this species.

Male and female prions: Thin-billed Prions (Pachyptila belcheri), Broad-billed Prions (P. vittata), Antarctic Prions (P. desolata), Fairy Prions (P. turtur) and Salvin’s Prion (P. salvini), emit noisy calls from burrows, consisting of a few notes (Bretagnolle 1990; Bretagnolle et al. 1990; Genevois and Bretagnolle 1995). Sex differences were found in the temporal structure of their calls. In all five species, females always started calling with a long note, followed by short notes, whereas males started calling with a short note. These differences allowed sexing by ear.

Among Procellariiformes, there are several monomorphic species without remarkable vocal differences between the sexes. For instance, Bulwer’s Petrel (Bulweria bulwerii) never calls during flight or while approaching a nesting colony. The vocal repertoire of this species consists of non-harmonic clicks, produced singly (single call), by two (double call) or in series (repeat call) (James and Robertson 1985c). In these calls, sex differences were poorly expressed. As a rule, females produced clicks at a higher tempo than males, but these differences were not significant.

Also, in Snow Petrels (Pagodroma nivea), no significant differences were found between male and female calls (Barbraud et al. 2000). Although male calls were somewhat lower in fundamental frequency, the range of values overlapped substantially between sexes. As in the Bulwer’s petrel, females produced their rhythmic calls at a higher tempo than males.

Order Ciconiiformes

In the Oriental Stork (Ciconia boyciana), the values of dominant frequency of bill clatters (Fig. 11) were non-overlapping between sexes: they were <670 Hz in males and >703 Hz in females, that is, they were always lower in males than in females, allowing 100 % reliable sexing. In addition, the dominant frequency is negatively correlated with the bill length of the storks (Eda-Fujiwara et al. 2004).

Order Charadriiformes

In Black-legged Kittiwakes (Rissa tridactyla), sex differences were found in their long calls, produced by birds flying over their nests (Aubin et al. 2007). Female calls were somewhat lower in pitch compared to male calls (on average, 540 and 630 Hz, respectively) and were separated with longer intervals between three call parts. Although the values of acoustic variables did overlap between sexes, DFA allowed discrimination between male and female calls with 100 % reliability. However, another study of Black-legged Kittiwakes in the same research group, conducted on another population, provided inconsistent data, with female calls being slightly higher-pitched than male calls (on average, 640 and 620 Hz, respectively) (Mulard et al. 2009). For this population, DFA showed only 69 % correct assignment of calls to sex. Consequently it remains unclear whether vocal cues to sex really exist in Black-legged Kittiwakes (Mulard et al. 2009).

No sex differences were found in the “kyow” and “kek” alarm calls of European Herring Gulls (Larus argentatus) (Hardouin et al. 2013). Although males are somewhat larger than females, the fundamental frequency did not differ between sexes in either call types.

No sex differences were found in the three loud call types of the Brown Skua (Catharacta Antarctica): loud calls, contact calls and alarm calls (Janicke et al. 2007). The value of correct assignment of calls to sex with DFA varied from only 60–73 % between call types.

For Parakeet Auklets (Cyclorhynchus psitaculla), in which parental care of both partners is critically important for survival of the single chick, sex differences were analyzed in their self-advertising whinny calls, representing a series of trills (Fig. 12) (Klenova et al. 2012). The trills were longer in males than in females (on average, 0.39 and 0.29 s, respectively) and contained more notes per trill in males than in females (on average, 16.6 and 10.7 notes per trill). However the values did overlap between sexes and the percentage of correct assignment of calls to sex with DFA was 94 %.

However, in the closely related Crested Auklets (Aethia cristatella), no sex differences were found in their self-advertising trumpet calls (Fig. 13) (Klenova et al. 2012). Frequency values of all measured variables strongly overlapped between sexes. Percentage of correct assignment to sex (72 %) did not differ from random values (69 %).

Order Strigiformes

The fundamental frequency of “bounce” of the Eastern Screech-owl (Otus asio), emitted in response to playbacks of species-specific calls, was lower in males than in females (720 and 820 Hz, respectively) (Cavanagh and Ritchison 1987).

The bounce songs of Western Screech-owls (Otus kennicottii), emitted in response to playbacks of male calls, were lower in fundamental frequency in males than in females (Herting and Belthoff 2001). However the best diagnostic parameter was the dominant frequency, with non-overlapping ranges of values between sexes: always lower 725 Hz in males and always higher 725 Hz in females.

In Barred Owls (Strix varia), male calls were lower in fundamental frequency and longer than female calls (Odom and Mennill 2010). The maximum fundamental frequency of “inspection” calls was 590 Hz in males and 640 Hz in females, lasting 0.73 s in males and 1.20 s in females. The maximum fundamental frequency of one-phrased hoots was 540 Hz in males and 590 Hz in females, lasting 0.38 s in males and 0.76 s in females. The percentage of correct assignment of calls with cluster-analysis was 91 %.

Hoot calls by male African Wood Owls (Strix woodfordii) were lower in fundamental frequency compared to hoot calls of females (Steyn and Scott 1972). The 12 year long study of their population structure of this species was based on these findings (Delport et al. 2002).

Hoot calls (“u-hu” calls) of the male Eagle Owl (Bubo bubo) are substantially lower in fundamental frequency than hoot calls of females (Grava et al. 2008). The maximum fundamental frequency of hoots was 390 Hz in males and 550 Hz in females. These differences are sufficiently strong to allow reliable discrimination between sexes by ear.

Order Coraciiformes

In the Green Woodhoopoe (Phoeniculus purpureus), “kek” calls were twice as low in fundamental frequency in males as in females (Radford 2004). The values of fundamental frequency did not overlap between sexes; these were lower than 600 Hz in males and higher than 700 Hz in females, which ensured 100 % reliable sexing. Although longer bills and slightly different feather coloration of males provide some indices to sex in this species, voice-based sexing is possible from a distance, which is important to a census of males and females in nature or in large enclosures.

Sex dimorphism was lacking in “provisioning” calls of the European Bee-eater (Merops apiaster), emitted when approaching the nest burrow (Lessells et al. 1995). In this species, males and females are of the same size, but mature males are usually a little brighter in coloration than females, noticeable when observed together (Lessells et al. 1995).

Order Falconiformes

Alarm calls of the White-tailed Hawk (Buteo albicaudatus) represent a rapid sequence of notes. The maximum fundamental frequency of the longest first note retains its frequency or increases it to the second note in males, but decreases in females (Farquahar 1993). This feature allows reliable discrimination between sexes. As well, this species has sex dimorphism of body size: the wings are longer and the body mass larger in females than in males.

Order Psittaciformes

For six closely related species of Poicephalus parrots, sex differences were investigated in their distress calls (Venuto et al. 2001). Two of the six species have external sexual dimorphism, whereas four others were sexed using laparoscopy. Distress calls of all six species represented harmonically rich, weakly modulated frequency calls, ranging between 165 and 205 Hz in fundamental frequency. Female calls were longer and higher in dominant frequency compared to male calls; however, values overlapped between sexes.

In Kea Parrots (Nestor notabilis), the fundamental frequency of mew and screech calls was somewhat lower in males than in females and male calls were longer than female calls (Schwing et al. 2012). However, the frequency of the variables overlapped considerably and the differences that were found were insufficient for reliable sexing.

Order Passeriformes

Among monomorphic passerines, fair sex differences have been reported for free-ranging White-rumped Munias (Lonchura striata phaethontoptila) from Taiwan (Mizuta et al. 2003). Distant male calls of this species contain only one note, whereas female calls contain three or more notes (on average 3.67 notes per call). This allowed 100 % reliable sexing even by ear in the field. The 100 % reliability of sexing by voice was confirmed by genetic sexing, whereas the morphometry-based sexing was only 87 % reliable.