Maria Triantafyllidou

The

limits

of

species

recognition:

heterospecific

song

learning

in

pied

flycatchers

Maria

Triantafyllidou

Degree project inbiology, Bachelor ofscience, 2016 Examensarbete ibiologi 15 hp tillkandidatexamen, 2016

Biology Education Centre and Department ofEcology and Genetics/Animal Ecology, Uppsala University

1 ABSTRACT

The closely related species pied flycatcher (Ficedula hypoleuca) and collared flycatcher (F. albicollis) co-occur on the Swedish island of Öland, where they compete over similar resources. The majority of male pied flycatchers have been found to incorporate elements of the collared flycatcher song in their repertoire. Given that birdsong is partly inherited and partly learned, the relative contribution of genetic predispositions versus acoustic stimuli varies across different species. The results show that in pied flycatchers, song acquisition is tightly correlated with imprinting, and can therefore be greatly influenced by heterospecific tutors in their surroundings, i.e. male collared flycatchers. I found that pied males are capable of not only memorizing collared song elements, but also producing them with high fidelity. Thus, I infer that pied flycatchers are characterized by a high degree of vocal plasticity.

INTRODUCTION

The importance of sexual signals in speciation

It is largely recognized that sexual signals play a key role in mate recognition as they indicate species identity and mate quality. It has been increasingly appreciated that they also play a significant role in patterns of speciation (Slabbekoorn and Smith 2002, Ritchie 2007, Verzijden et al. 2012).That is linked with the fact that sex traits evolve quickly and are therefore likely to diverge among closely related species, eventually leading to reproductive isolation (Qvarnström et al. 2006, Kraaijeveld et al. 2011).

Sex traits arise by sexual selection (named “intersexual selection” by Charles Darwin), which is the result of male competition for copulation with the opposite sex. Female mating discriminability, as male signaling, is complex and subject to evolution, and thereby there is a well-established tradeoff between female preferences for sexual signals and male signaling traits (Ritchie 1996, Michaelidis et

al. 2006, Ron 2008). Given that diversification is a prerequisite for speciation, divergence either in male sexual signals or in female preferences represent cases in which pre-zygotic reproductive isolation gets promoted and consequently induces speciation (Ritchie 2007, Kraaijeveld et al. 2011, Verzijden et

al. 2012). Signal evolution, thereby, acts as a driver of speciation (Ritchie 2007, Seddon and Tobias

2010).

2

the same time, sexual traits themselves can also be affected by learning. Thus, learned sexual traits and sexually imprinted preferences constitute models useful in understanding the tradeoffs of attracting fe-males while at the same time avoiding interspecific matings.

By having preferences and selecting their mate, females establish reproductive isolation. Keeping those female preferences species-specific is of paramount importance with regard to evolution and speciation. Even though they are normally easily maintained in traits such as plumage, they can easily break down in traits that are learned and acquired. One such trait is birdsong, which is species-specific and functions in species recognition (Emlen 1972). On that basis, maintaining song differences across species is greatly important yet significantly challenging.

Bird song

Song in birds is a unique and intricate trait that has received a lot of scientific attention and has led to various researches and proposed hypotheses over the years. It serves a dual purpose: female attraction with the ultimate goal of copulation, and territory defense. The latter ties back to the former, since the exclusive access to a territory is tightly correlated to female settlement and nesting (Nowicki and Searcy 2004).

Birdsong constitutes a sexual signal and is therefore characterized by great variation not only across species but also across individuals. Given that a song is an elaborate conglomeration of numerous parameters (such as pitch, length, trill rate, etc.), interspecific and intraspecific variation is only to be expected. A major source of variation in birdsong is linked to geographic divergence, i.e. the result of the process of different populations adapting to distinct environments (Slabbekoorn and Smith 2002, Podos and Warren 2007).

Females base their mate choice on those features that are reliably correlated with condition and quality of the male (Nowicki and Searcy 2004). Birdsong is a sexual signal that serves as an indicator of the signaler's quality and is therefore broadly used by females in the evaluation process. Assessment of vocal performance is based upon vocal capabilities which reflect a male's ability to provide benefits to the female. Such benefits can be either direct and associated with better parental care and access to a better territory, or indirect, such as the transfer of robust genes to the offsprings that will increase their chances for survival and reproductive success (Ballentine et al. 2004, Nowicki and Searcy 2004).

3

period for perceiving and storing information. Clayton (1989) performed a series of cross-fostering experiments showing that birds raised by a foster father have the capacity to learn and identically produce elements of their tutor's song despite the heterospecific nature of the song. On that account, bird song is a culturally inherited trait largely dependent on the hatchling's acoustic surroundings.

Nonetheless, it cannot be disregarded that there are certain constraints on what can be learned, that are correlated with species-specific innate predispositions (Kroodsma and Miller 1982, Clayton 1989, Nowicki et al. 2001, Ballentine et al. 2004). In other words, birds inherit a song template from their parents which is later influenced by input from the environment. How much of this template is inherited and what kind of genetic constraints it introduces to the system are subject to discussion (Marler 1997, Beecher et al. 2010). There is a tendency, more scientifically described as “innate predispositions”, for individuals to learn and internalize the song of the species they belong to, i.e. the conspecific song (Marler 1997, Doupe and Kuhl 1999). Baptista (1996) showed that even cross-fostered individuals preferentially learn the song of conspecifics rather than the song of their foster parents. When considered in the context of evolution, it becomes clear that those predispositions have a well-defined evolutionary significance. Conspecific song preference is essential for species recognition which constitutes the main driver of reproductive isolation (Grant and Grant 1997). The degree of maintenance of reproductive isolation, thenceforth, will determine the speciation patterns among different species.

Given that birdsong is the outcome of a learning process during which sexual cues are being imprinted and mating preferences begin to form, various challenges are inevitably introduced into the system. When this intricate imprinting process is based upon parental and thus conspecific tutors, assortative mating is promoted. When, however, random heterospecific individuals influence the learning process, this balance is disturbed (Slabbekoorn and Smith 2002, Verzijden et al. 2012). This is where mixed singing comes into the picture.

Mixed singing

Traits that are at least partly learned, such as birdsong, unlike inherited traits strictly constrained by the genotype, have the potential to be subject to rapid change (Irwin et al. 2001). Change can occur in the form of either divergence, in which case rapid evolution is induced (Verzijden et al. 2012), or con-vergence, i.e. when different species learn from one another (Cody 1973, Helb et al. 1985).

4

or completely learns and produces elements from the song of another species with which it is closely related and lives in close proximity. Such cases are commonly observed in natural systems where relatively recently speciated species are brought into secondary contact and ecological similarity leads to sympatry. With the possibility of introgression ruled out (Haavie et al. 2004, Vokurková et al. 2013), all evidence points towards a case of vocal convergence induced by social interactions.

Selective learning of the conspecific song translates to learning and producing acoustic cues that lie within the structural (frequency range, amount of modulations, etc.) and temporal (note length, tempo, etc.) patterning that characterize the species song. Imitating vocal signals of other species, i.e. mixed singing, is commonly observed in songbirds, yet the mechanisms that underlie it and allow for heterospecific copying remain highly controversial. There is always some degree of constraint as to how feasible it is for a species to learn and be able to efficiently produce other species' song elements and that degree varies among different species (Clayton 1989, Wooley and Moore 2011). By and large, birds will most likely copy those song components that resemble the conspecific ones (Thorpe 1958).

Besides, song features vary in their degree of plasticity. Some tend to be rather plastic and therefore variable while others are more species-conserved. Clayton (1989) showed through a series of cross-fostering experiments, that even though heterospecific copying allows for learning of individual notes, yet song features (e.g. song length, song tempo) are not copied among heterospecifics. Moreover, Reichard and Price (2008) investigated two mimid species (Family Mimidae) living in sympatry and found that even though all kinds of notes are copied between both species, the repetition rate and the frequency range constitute species-specific features that are always retained by the individuals. Physically demanding song features, such as trill and multiple harmonics are also normally species-conserved and cannot be accurately produced by heterospecifics (Clayton 1989, Ballentine et al. 2004).

5

claim and defense. Given that in sympatric populations, resources (e.g. territories) can be limited and competition intensified, it is likely that a need for differentiation in vocal design will emerge. Song convergence, contrary to costly physical aggression, will mitigate interspecific competition and will provide the singer with exclusive access to the territory.

Another, not necessarily mutually exclusive, hypothesis correlates repertoire size with advantageous to the females male qualities, such as breeding experience and body mass. The claim goes as follows: complexity and bigger repertoire size in mixed songs serve as a plus in intraspecific communication (Nowicki and Searcy 2004, Hansen et al. 2010, Vokurková et al. 2013, Wheatcroft 2015). This theory contradicts Nowicki's et al. claim (2001) that natural selection acts against innovations in sex-related traits and alleges that the incorporation of elements from two well distinct songs into one, indicates higher quality and is particularly appealing to females in pursuit of the fittest male to copulate with.

In certain species population studies it has been shown that female recognition is not influenced by song convergence (Seddon and Tobias 2010). In these cases females were still able to discriminate between conspecific and heterospecific males and thus mixed pairing never occurred (Hudson & Price 2014, Wheatcroft 2015). Nevertheless, in other natural systems, as a consequence of mixed singing, the individual attracts to his territory potential mates among which some heterospecific females (Grant and Grant 1997, Secondi et al. 2003, Qvarnström et al. 2006). Heterospecific mating is thereby likely to occur and result in hybridization.

On that account, interspecific discrimination of songs is essential in sympatric bird populations as it promotes premating isolation and minimizes hybridization (Haavie et al. 2004, Qvarnström et al. 2006). Nevertheless, it seems like this discrimination is somewhat weaker between closely related species (Qvarnström et al. 2006, Vokurková et al. 2013), where the uninterrupted pattern of hybridization would normally become progressively eliminated by the mechanism of reinforcement (Haavie et al. 2004). Convergent singing is, notwithstanding, maintained in sympatric populations presumably because the benefits reaped outweigh the evolutionary costs related with the low fitness of the hybrids (Vokurková et al. 2013, Wheatcroft 2015).

6 Flycatchers

The pied flycatcher and the collared flycatcher are believed to have speciated during the Pleistocene glaciations (Sætre et al. 2001, Backström et al. 2013) and have recently been brought into secondary

contact in central and eastern Europe as well as on the Baltic islands of Gotland and Öland (Qvarnström and Bailey 2009). In the latter areas, their breeding ranges largely overlap, whereas in the former ones the habitat segregation is more pronounced.

The Öland flycatcher populations are the youngest (1960's) and therefore constitute an excellent re-search model for evolutionary and hybridization dynamics. Being recently speciated species and newly established populations, the collared and the pied flycatchers exhibit little morphological and ecologi-cal divergence. As a result they often compete over nest sites and territories and occasionally hybridize (Qvarnström et al. 2010). This indicates that the pre-zygotic reproductive barriers are not well-developed and thereby reproductive isolation is regularly overcome.

Demographic researches have proved that the relative abundance of collared flycatchers in the area is significantly higher than that of pied flycatchers (Vallin et al. 2012). In line with this finding is a pa-rameter of major significance: some male pied flycatchers have been observed to incorporate elements of the collared flycatcher song in their own song (Gelter 1987, Haavie et al. 2004), i.e. pied flycatchers can, and very often do, perform mixed singing. Converged song accounts for 2/3 to 3/4 of all pied fly-catchers’ vocal performances (Haavie et al. 2004, Qvarnström et al. 2006). What emerges as a reliable theory is that song learning is strongly influenced by the vocal performances in the surroundings where collared flycatchers have higher densities. Heterospecific-singing pied flycatchers had been exposed to a larger amount of collared songs during the critical for song learning period of their life (Gelter 1987).

On the other hand, there are numerous theories proposing that mixed singing is not a mere by-product of the learning process but rather it is a feature that has been acquired to serve a certain evolutionary purpose: male communication and development of territoriality (Qvarnström et al. 2006, Vokurková et al. 2013, Wheatcroft 2015). Given that collared flycatchers are more successful in interspecific competition (Qvarnström et al. 2010), pied flycatchers might resort to mixed singing so as to improve their territoriality. Emitting a song that is by and large perceived as a conspecific song by the collared flycatchers, pied flycatchers decrease the chances of getting driven off by neighbouring collared males that are in pursuit of a territory.

7

hybridization is estimated to be 30% (Qvarnström et al. 2006). The presence of hybrids indicates that species isolating mechanisms are not fully developed, since they allow for heterospecific mating. Given that hybrids suffer reduction in fitness, it is likely that in the future, maladaptive hybridization will result in development of reproductive barriers, i.e. more pronounced differences in traits used for species recognition (Haavie et al. 2004). As it has been suggested, heterospecific pairing, as a consequence of converged singing, appears to be more likely to occur in the case of secondary contact quickly after speciation (Grant and Grant 1997, Qvarnström et al. 2006, Wheatcroft 2015).

The rate of mixed singing is lower in the central European hybrid zone, where hybridization also occurs less. It has been alleged that pre-zygotic barriers and reproductive isolation have had more time to evolve in the central European hybrid zone (Haavie et al. 2004). In light of the above, the two population zones are currently in different speciation stages: primitive in the island populations and more advanced in the continental populations. In time, similar mechanisms are expected to evolve in the Baltic island flycatcher populations as well (Haavie et al. 2004). Nonetheless, the two hybrid zones differ in a very significant parameter: habitat segregation. Collared and pied flycatchers in central Europe occupy distinctly different habitats for the most part and the overlap of the two species populations is far lower than it is on the Baltic islands (Qvarnström et al. 2010). The underlying idea behind this is that ecological displacement leads to character displacement, i.e. song divergence (Slabbekoorn and Smith 2002).

This study aims at providing a better understanding of the mechanisms that control mixed singing in flycatchers. Previous researches (Gelter 1987, Haavie et al. 2004) have also had a similar aim. What differentiates my research from previous ones is that I focus on notes rather than on whole songs. Considering that it is the individual notes that are being copied by the mixed singing pied flycatchers, I expect to be able to draw reliable conclusions about the mechanisms of mixed singing.

8 METHODS

Study species and sites

Pied and collared flycatchers are closely related bird species, diverged probably less than 1 Ma (Nadachowska-Brzyska et al. 2013), with large similarity in ecology and breeding biology. They are small (12 g) passerine, insectivorous songbirds with similar habitat preferences. Despite those, the overall collared flycatchers breeding range is more southern and eastern, while that of pied flycatchers extends further North. Their morphology is very similar, with both species exhibiting black and white plumage with white forehead patches. The morphological differences can be summarized in the collared flycatchers' prominent white collar, which extends around the neck and the pied flycatchers' smaller forehead patches. Female differentiation in the field is more difficult, but in general female collared flycatchers have larger white patches on their wings and more of an olive tone to their grey-brown color. Both collared and pied flycatchers are long-distance migrants with wintering grounds in sub-Saharan Africa. Breeding populations of the two species span big parts of Europe and co-occurring (sympatric) populations can be found in a large area in central Europe, where they breed at distinct elevations, as well as in more isolated hybrid zones in the Baltic islands of Gotland (since 150 years ago) and Öland (since the 1960's). On the island of Öland (57° 10´ N, 16° 58´ E) collared flycatchers are far more abundant than pied flycatchers and are currently in the process of displacing pied flycatchers from preferred breeding sites.

All the data have been collected on the island of Öland, where there are well established nestbox breeding areas being monitored each spring since 2002. Both pied and collared flycatchers nestbox areas are located in both deciduous, generally occupied by collared flycatchers, and pine or mixed forest, generally occupied by pied flycatchers. In the northern part of Öland, nestbox areas are recently showing a tendency to become species-specific, with pied flycatchers usually being surrounded by other pied flycatchers within poorer habitat sites. Mixed-pairing is relatively rare, constituting less than 6% of all pairs (Qvarnström et al. 2010).

9 Recordings

All of the song recordings were made on the island of Öland within the time span of 2007 to 2015. 18 songs (2 collared, 8 pied, 8 mixed singers) were recorded using DAT and minidisc recorders (Sony TCD-D7, TCD-D10 and Kenwood DMC-J7R), with microphones that were mounted on 55cm parabolic dishes (Classic, Science and PRO II, Telinga, Tobo, Sweden). The rest 18 recordings (10 collared, 4 pied, 4 mixed singers) were made using a digital audio recorder (TASCAM DR-40, TEAC America Inc., USA) and condenser microphone (Sennheiser ME66, GmbH & Co. KG, Germany). All recordings were made in 16-bit, 44.1 kHz sample rate WAV format. All recordings were edited of background noise using RAVEN v1.4 (Bioacoustics Research Program 2011) and contained high-quality recordings of at least five songs. The pied songs were classified as either mixed (converged) or pure songs. This distinction was based on both the recordist's notes and the auditory inspection.

Acoustic analysis

For each singer type (pied, collared, mixed) I analyzed 12 individuals. For collared flycatchers, I measured 10 songs from individual 1 and individual 2, 3 songs from individual 3, and 5 songs from individuals 4 – 12; for pure singing pied flycatchers, 12 songs from individual 1, 9 songs from individual 2, and 5 songs from individuals 3 – 12; for mixed singing pied flycatchers, 7 songs from individual 1, 9 songs from individual 2, and 5 songs from individuals 3 – 12.

The measurements were taken through RAVEN v1.4, using basically the spectrogram view (the waveform was only occasionally viewed so as to infer the distance between 2 notes that appeared to be in close proximity in the spectrogram). For the spectrogram analysis, I used the Hann window function (default), a spectrogram window size of 512, and the time axis was set to 2s while the frequency axis was set to 15000Hz (Figure 2).

The analysis included variables both in note level and in song level. 11 note measurements were taken through Raven: begin time (s), end time (s), low frequency (Hz), high frequency (Hz), Q1 frequency (Hz) (the class value for the 1/4th _{cumulative frequency), Q3 frequency (Hz) (the class value}

for the 3/4th _{cumulative frequency), delta frequency (Hz) (i.e. bandwidth), delta time (s), peak}

10

notes with multiple harmonics, only the fundamental frequency (the lowest frequency in a harmonics series) was measured. For the song-level analysis, the same variables were measured through Raven (begin time, end time, low frequency, high frequency, Q1 frequency, Q3 frequency, delta frequency, delta time, peak frequency, peak time, max power, peak time2). An additional variable, the number of notes per song, was calculated.

In my analysis, notes are defined as song components, which can be composed of a single element or a group of elements, and which are separated by noticeable time intervals in the spectrogram. Several notes make up phrases, i.e. repeatable parts of the song sequence (Figure 1). An individual produces several songs during a vocal performance. Each song, however, varies in the patterns formed by its components (notes).

Figure 1. Pied flycatcher song which includes a commonly repeated phrase and a note consisting of several elements.

11

Figure 2. Spectrograms of collared, mixed singer and pied flycatcher songs. Common element between collared and mixed

song indicated by circle and common element between pied and mixed song indicated by brackets.

12 Statistical analysis

The first step of the analysis was to figure out the significant variables, i.e. the ones that I should base the further analysis on. I constructed a correlation matrix of all the continuous variables: minimum frequency, maximum frequency, Q1 frequency, Q3 frequency, delta frequency, delta time, peak frequency and peak time 2. Thereby, I was able to exclude the variables with a correlation greater than 0.50 and thus narrow it down to the important variables, i.e. those that are weakly correlated. In this way, I summarized variation across singer types with three weakly correlated acoustic measurements: low frequency, delta frequency, delta time.

Using the R software (R Core Team 2012), I created linear mixed models for all the variables of interest, i.e. low frequency, delta frequency, delta time, polyphony, buzz, modulations, direction, for the notes, and tempo, number of notes per song, delta time for the songs. Those models account for non-independence, i.e. for the fact that this analysis encompasses more than one measurement per individual. For the continuous variables (low frequency, delta frequency, delta time, tempo) I used linear models. For the categorical, ordered variables (direction) I used ordinal models. For 'count' variables (modulations, number of notes per song) I used generalized linear poisson models. For binomial variables (buzz, polyphony) I used generalized linear binomial models. From the results that those models yielded, I was able to deduce the rough position of each singer type relatively to the other two in respect to each variable individually. To illustrate this position, I created one chart for each variable.

Up until this part of the analysis, the mixed singer notes were grouped together irrespective of whether they actually were collared or pied-like. Nevertheless, this approach leads to an averaging across collared and pied-like notes. This means that in case I find that mixed notes are closer to pied notes, it might actually be due to a larger proportion of their notes being pied, rather than because they cannot sing collared notes accurately. In order to overcome this problem, I performed a discriminant function analysis (DFA) so as to end up with a reliable and discrete classification of mixed singer notes. The DFA was performed with R software using the package MASS (Venables and Ripley 2002). The discriminant analysis assigned each note in the data to one of the two species regardless of type of note. All note classifications were later tested for their degree of accuracy. The spectrogram variables were used to correctly identify every single note in the dataset.

13

collared notes. Next, using the aggregate function in R software, I calculated the variation in the proportion of collared notes across mixed singing individuals. In addition, I calculated the proportion of collared notes used in a certain order within songs. The last step was to compare across pied-like mixed notes, collared-like mixed notes, pure pied notes and pure collared notes and look at how accurately mixed singers produce collared notes.

RESULTS

Correlation among variables

The correlation table (Table 1) indicates that from the total of 8 continuous variables, the three that account for most of the variability (97%) in the dataset are low frequency, delta frequency and delta time.

Table 1. Correlation matrix of 8 song variables. The variables that account for most of the variability are indicated by

bolded text.

Acoustic differences between singer types and species

14

difference across the singer types is found. As far as the individual notes are concerned, the variables of interest are low frequency, delta frequency, delta time, polyphony, buzz, modulations and direction. All those differ significantly between collared and pied flycatchers. Mixed singing pied flycatchers were shown to be in between the two main singer types for all of the variables, although they were shown to bear an overall greater resemblance to pure singing pied flycatchers.

Polyphony and song length were found to be only insignificantly different between pure and mixed singing pied flycatchers (Figure 3a&3c). On the contrary, mixed singers and collared flycatchers were shown to exhibit great similarity in note delta frequency (Figure 3b). The variable number of notes per song was greatly similar across all three singer types (Figure 3d). Overall however, the rest of the variables (note low frequency, buzz, direction, modulations, note length and song tempo), although significantly different between pure and mixed singing pied individuals, they were still found to exhibit a bigger similarity to pied than to collared flycatchers (Figure 4).

Figure 3. Charts for the variables (a) note polyphony, (b) note delta frequency, (c) song length (i.e. song delta time) and (d)

number of notes per song. For polyphony and song length, the difference between mixed singers and pure singing pied individuals is insignificant. For delta frequency, the difference between mixed singers and collared flycatchers is insignificant. For number of notes per song, the difference across all three singer types is insignificant.

a b _b -5 5 15 25 35 45

collared mixed pied

polyphony (%)

a

collared mixed pied

delta frequency (Hz)

b

collared mixed pied

song length (s)

collared mixed pied

notes per song (n)

d

15

Figure 4. Charts for variables (a) note low frequency, (b) note buzz, (c) note length (i.e. note delta time), (d) note

modulations, (e) note direction and (f) song tempo. For all variables, the difference between mixed singers and collared or pied flycatchers is significant, i.e. mixed singers are in between collared and pied flycatchers.

Collared notes in mixed songs

The DFA assigned the mixed singer notes as either 'collared' or 'pied' with 80% accuracy (the coefficients of the DFA are given in Table 2). 57.7% of the notes were classified as pied notes and 42.2% as collared notes. Haavie et al. (2004) had yielded similar results, with pied-like notes constituting 60% of the total mixed singer notes. The density plot across all three singer types (Figure

a b c 0 1000 2000 3000 4000 5000 6000 7000

collared mixed pied

low frequency (Hz)

a

collared mixed pied

buzz (%)

collared mixed pied

note length (s)

c

collared mixed pied

modulations (n)

d

collared mixed pied

direction (%)

collared mixed pied

song tempo (s)

b

16

5a) shows that while pied notes tend to be positive and collared notes tend to be negative, mixed singer notes appear all over the place, spanning from negative to positive values. Nevertheless, what can also be seen is that a bigger proportion of mixed singer notes lie at the right side of the plot; this proportion represents pied-like notes. This goes along with the results that the linear mixed models yielded, i.e. overall, mixed singers end up in between the two singer types but they tend to be closer to pied than to collared individuals. At the same time, there is a 20% overlap between collared and pure singing pied notes, reflecting that the DFA does not work perfectly.

The density plot that compares the discriminant scores of the mixed singer notes that are classified as 'collared' with the pure collared flycatcher notes (Figure 5b) shows that even among those mixed singer notes that are classified as 'collared', a substantial proportion of the notes tend to be more pied-like than the pure collared notes.

Figure 5. Density plot for (a) collared, mixed singer and pied notes and (b) mixed collared and pure collared notes.

17 Table 2. Coefficients of linear discriminants.

measurement LD1

polyphony -1.817545e-01 modulation 1.833067e-01

buzz -2.902667e-03 direction -2.931788e-01 low freq -2.964476e-04 delta freq 2.801369e-04 delta time -5.352210e+00

peak freq 5.977985e-04 peak time 2 -1.584327e+00

high freq -2.356992e-05 Q1 freq -2.982861e-04 Q3 freq -5.290353e-04

I looked at the proportion of collared notes used by mixed singers and how this proportion varies across individuals so as to infer the proportion of collared notes in the different individuals' songs (Table 3). The results show that some mixed singing individuals sing songs with a big amount of collared notes in them (e.g. individual 1 and 6), whereas some others sing songs that only remotely resemble collared song, i.e. songs that include very few collared notes (e.g. individual 10 and 12).

I then performed the same kind of analysis (aggregate function) across notes (Table 4) to see whether there was some discrete pattern in the order in which mixed singers choose to emit pied-like or collared-like notes. I took into account notes 1 to 11, since only 5 out of 66 songs were composed of more than 11 notes. The results show that the first note of a mixed song is 71% of the times a collared note, i.e. an alarm call, which mixed singers produce with relatively high fidelity. The second note is usually (54%) a collared note too. However, as the song proceeds, pied notes exceed the collared ones in frequency (e.g. the 11th _{note is only 16% of the times a collared note). The post-hoc test on the linear}

model showed that only the first note is significantly more likely to be a collared note than any of the other notes (P < 0.005).

18

note features with high accuracy. Song rate is the only feature that is neither too pied nor too collared-like in mixed songs. Mixed singing pied flycatchers seem to deviate from their conspecific song rate and exhibit song rates intermediate between the pure pied and the collared ones. As far as the low frequency is concerned, mixed pied notes exhibit a lower low frequency than the pure pied notes. In polyphony, mixed pied and pure pied notes are very similar, whereas mixed collared and pure collared ones diverge more. Mixed singers are shown to be perfectly capable of producing notes as accurate in buzz, direction, modulations, low frequency, delta time and delta frequency, as pure collared individuals do.

Table 3. Percentage of pure collared notes used in mixed Table 4. Percentage of pure collared notes correlated with

songs across individuals. the ordering of the notes in a mixed song.

mixed singer individual

percentage of collared notes in mixed songs

1 65% 2 52% 3 37% 4 52% 5 49% 6 71% 7 30% 8 33% 9 36% 10 3% 11 62% 12 7% mixed singer note (ordering of notes in song)

19

Table 5. Post-hoc tests on linear mixed models.

mixed collared – pure collared mixed pied – pure pied estimate std. error z value p

value

estimate std. error z value p value low freq -12.55 90.18 -0.139 0.999 432.34 82.18 5.261 <1e-05 delta freq -182.63 86.59 -2.109 0.147 134.47 78.88 1.705 0.316 delta time 0.005503 0.004825 1.141 0.66 0.007221 0.004397 1.642 0.35 polyp hony -0.2670 0.1722 -1.551 0.401 0.1385 0.1748 0.792 0.855 modul ations -0.01167 0.07088 -0.165 0.998 -0.02202 0.03778 -0.583 0.934 buzz -0.1074 0.1824 -0.589 0.926 2.2648 1.0292 2.201 0.102 directi on 0.13983798 0.1603195 0.8722454 0.8192 -0.05944264 0.1474682 -0.4030880 0.9779 song tempo -0.0197939 0.0052719 -3.755 <0.001 -0.0004214 0.0045103 -0.093 1 DISCUSSION

What I found is that even though the song of mixed singing pied males bears a stronger resemblance to the conspecific song, still the similarities with the song of the heterospecific collared males are striking. Pied flycatchers have the capacity to memorize heterospecific song elements and incorporate them into their song with high accuracy. Thus, I can reliably infer that pied flycatchers exhibit high rates of plasticity in vocal learning and vocal performance, as has been previously suggested (Eriksen

et al. 2009, Vabishchevich and Formozov 2010).

In the following paragraphs I will discuss the implications of the results and compare my findings with results from previous researches. I will also elaborate on the role of social experience and inherited genetic predispositions in song learning. Last, I will discuss the function of mixed song as well as the reasons behind its maintenance in the flycatchers hybrid zone of Öland.

Interpretation of the results

20

species-conserved, i.e. mixed singers emit songs equally long to the pure conspecific songs. This suggests that song length is not a plastic feature. Song duration is always of fixed length and, at least in flycatchers, does not vary between pure singing individuals and individuals that converge their song to heterospecifics.

What is more, mixed singing pied flycatchers were found to sing songs in a rate in between the rates of pure singing individuals and collared individuals. In other words, song tempo in flycatchers constitutes a plastic feature that can and often does diverge from the normal conspecific range. This might be due to the fact that mixed singing individuals choose to start off their song with collared notes between which there is a relatively big time interval (i.e. collared-like rate) and only later on do they incorporate the pied notes and the faster song tempo.

Concerning the number of notes in song, pure pied and collared flycatchers already exhibit highly similar values and thus mixed singing individuals were not expected to diverge much from those values, as was proven. When cross-fostered, however, to tits (Eriksen et al. 2009), males compose songs with more notes, which implies that this feature might also be a plastic one but this cannot be confirmed when comparing species songs with approximately the same number of notes.

As far as the note variables are concerned, the results showed that for the most part, pied males are perfectly capable of adjusting their note features in such a way that the notes become collared-like. Note length, even though considered to be mostly species-conserved (Gelter 1987), is converged to collared values. The same pattern is observed with the note variables buzz, modulations and direction. Low frequency also assumes intermediate values, as was found when pied flycatchers were cross-fostered to tits (Eriksen et al. 2009). Delta frequency was found to largely converge to collared values. This implies that delta frequency constitutes a feature of extra-high plasticity, contrary to other researchers' results (Reichard and Price 2008, Kagawa et al. 2014). On the other hand, mixed singers were found to be almost as polyphonic as pure singing individuals, which suggests that the innate predispositions for polyphony are strict and do not allow for big divergence. This can be explained by the fact that polyphony constitutes a feature that is physically demanding; the production of two simultaneous separate sounds requires that birds vocalize with both sides of their syrinx. Hence, big deviation from the conspecific range could not easily occur.

Comparison with previous researches

21

themselves can often be learned across species, the temporal and structural patterning is always species-conserved (Güttinger 1979, Clayton 1989, Kagawa et al. 2014). Clayton (1989) conducted a series of cross-fostering experiments on finches, concluding that notes can be produced with high accuracy but the temporal pattern as well as the sequence in which the song elements are placed in a song constitute qualities always retained by the individuals. On the contrary, Eriksen's et al. (2009) cross-fostering experiments of pied flycatcher eggs into the nests of blue (Cyanistes caeruleus) and great tits (Parus major) and Johannessen's et al. (2006) cross-fostering experiments of great tits to blue tits, showed that the repertoire composition is also influenced by the heterospecific tutor. Furthermore, song tempo is generally regarded to be species-conserved, even in cross-fostered individuals (Kagawa

et al. 2014). My results lie somewhere in the middle, since song tempo was found to converge to

heterospecific values, while song length values were kept conspecific. It should always, however, be taken into consideration that interspecific cross-fostering and naturally occurring heterospecific copying are two different situations and thus one cannot always make direct comparisons between the two. In both cases, however, song acquisition is critically affected.

Even though trill is present neither in pied nor in collared flycatcher song, we can look at Johannessen's et al. (2006) results of cross-fostering experiments and make analogies with the flycatcher system. Cross-fostered great tits were found to emit trills which were learned from their heterospecific blue tit parents. Considering that buzz is also a note type that is never present in pure pied song but is almost always part of mixed pied song, we can infer that song elements like trill and buzz, can be learned across different species, even if they do not constitute components of the original species-specific repertoire.

Contrary to my findings, numerous researchers (Reichard and Price 2008, Kagawa et al. 2014) who conducted cross-fostering experiments, have claimed that frequency properties remain species-specific. I showed that in flycatchers, those properties are plastic and therefore subject to heterospecific copying.

Vocal plasticity in pied flycatchers

22

proportion of collared notes varies across mixed singer individuals, ranging between 3% and 71%. This suggests that there is significant variability in the level of mixed singing the individuals can perform. Some male pied individuals emit a song that overall resembles collared song to a higher extent than pied song. Some other individuals on the other hand, are classified as mixed singers based only on a minor part of their song repertoire. Those individuals maintain a high proportion of the notes they have inherited from their parents and incorporate in their song no more than a few collared notes or copy no more than a few note features from the heterospecifics.

Mixed singing has never been observed in collared flycatchers. A proposed explanation lies in the fact that pied flycatchers are characterized by higher plasticity and their song-learning predisposition allows for the learning of collared song components (Vabishchevich and Formozov 2010,Eriksen et al. 2011). This plasticity is activated by the high abundance of neighboring collared flycatchers.

When and from whom pied males learn songs

Learning does not always happen from parents but also from other conspecifics in the surroundings. In fact, it is more common that other conspecifics function as tutors when young males learn their species-specific song (Catchpole and Slater 2008). In the studied flycatcher zone, male collared flycatchers might be perceived as song tutors by the young male pied flycatchers. Assuming that the claim that pied flycatchers are open-ended learners (i.e. they have the ability to internalize new song elements throughout their lives) holds, and that this species is also characterized by extensive vocal plasticity (Espmark and Lampe 1993, Vabishchevich and Formozov 2010, Eriksen et al. 2011), male pied individuals are expected to be perfectly capable of learning and producing heterospecific song elements. It is, however, critical to look at the relative time they spend around conspecifics and heterospecifics and examine the different possibilities as to when exactly they imprint acoustic cues.

Pied flycatchers are mostly surrounded by conspecifics during winter as well as migration and therefore, during that time they are almost exclusively exposed to pied song. In addition to that, all individuals have inherited from their parents the 'pied song template' which favours conspecific song learning. On that ground, we can infer that those two independent factors account for the biggest proportion (~60%) of pied notes in the mixed singers' repertoire. Differentiating between social experience and innate predispositions is a challenge with the dataset I have worked with, thus it is not possible to compare their relative significance.

23

therefore assured and the outcome of this exposure is that ~40% of the mixed singer notes are collared notes. That being said, I predict a correlation between the proportion of collared individuals and the proportion of collared notes in mixed songs; pied individuals that are surrounded by more collared flycatchers and thus exposed to more collared songs, will incorporate in their songs more collared song elements. Moreover, the fact that pied flycatchers learn up to ~40% pure collared notes and produce them with high accuracy, despite being much more exposed to conspecific notes, suggests that their inherited conspecific song template imposes only little restricting in their learning and producing capacity and allows for collared notes to be learned and produced with high fidelity.

However, we ought to be cautious when attempting to make such inferences given that the exact periods during which individuals learn songs remain unknown. If the learning period is restricted in the time they spend in the breeding grounds, where they are surrounded by high density collared populations, it would suggest fairly strong innate predispositions for conspecific song learning. If, on the other hand, most of the learning takes place in the wintering grounds in Africa, where pied and collared flycatcher habitats are largely segregated, that would indicate that the pied song template is very flexible and susceptible to external song output by the collared individuals with which they interact only during the breeding season.

What is also generally appreciated is that song activity significantly decreases after pairing. Males sing very little after the nestlings have hatched and substantially less after they have fledged (Espmark and Lampe 1993). Learning of collared song may potentially take place immediately after the arrival at the breeding grounds (Eriksen et al. 2009). Given that both pied and collared flycatchers migrate from Africa to their breeding grounds on Öland, it is also possible that pied flycatchers might learn collared song during migration. That being said, I suggest that future investigations track individuals and look at their acoustic surroundings and vocal experiences in both the breeding and the wintering grounds.

Mixed singing patterning

Part of mixed singers' song (~60%) is always conspecific which means that at a certain extent the genetic auditory template guides their song tutor choice. For a fairly big part of their song (~40%) however, mixed singers seem to defy their innate predispositions and copy elements from collared songs.

24

mixed singers produce lie within the pure pied note range. This could be either due to the fact that the majority of the notes that mixed singers produce are pied notes or because overall the mixed singer notes are indeed more pied-like. To investigate that, I compared the mixed singer notes to pure pied and pure collared notes.

I had predicted that the mixed singer notes would be plotted in two separate places – one in the collared range and one in the pied range, which would suggest that they in fact learn notes from the two pure singer types as discrete units. Nevertheless, a significant amount of those notes ended up in between the two pure singers' ranges. Foremost, this suggests that only occasionally do they produce pure (i.e. identical) collared notes. This pattern can be interpreted by two non-mutually exclusive hypotheses. The first hypothesis suggests that mixed singers do produce collared notes with high fidelity but they are selective in the collared notes they choose to learn and produce, i.e. they might be choosing to learn only those collared notes that exhibit similarities with the pied notes. Alternatively, mixed singers copy certain aspects of collared notes and not the notes themselves. By combining those features and the conspecific features, they create mixed notes, which however resemble pied notes to a bigger extent.

For the majority of the variables, the difference between pure singing and mixed singing pied flycatchers is more or less insignificant. This was expected given that no matter how plastic their song template is and how influenced they are by the other species, they are still, to a certain extent, constrained by their species-specific 'biases'. Nevertheless, when they do produce collared notes, they do so with high fidelity. This indicates that this species is characterized by a high degree of vocal plasticity (Eriksen et al. 2009, Vabishchevich and Formozov 2010). In other words, their innate song template is not strictly constrained by the species-specific temporal and structural song properties but rather appears to be plastic and open to external song output. This finding goes against the popular claim that the choice of song tutor in many species is largely genetically driven and innate biases favour conspecific song learning (Marler 1997, Catchpole and Slater 2008, Hansen et al. 2010). Conspecific imprinting is not the norm in the flycatchers system I examined and the reason behind that might lie in the fact that the two species are still very closely related. Pied and collared flycatchers diverged less than 2 million years ago (Ellegren et al. 2012). This implies that in time, converged singing might become less and less common among pied flycatchers (Haavie et al. 2004).

Function of mixed singing

25

(Helb et al. 1985) or it actually constitutes a mechanism with a distinct adaptive value remains largely controversial (Gelter 1987, Gorissen et al. 2006, Hansen et al. 2010, Vabishchevich and Formozov 2010, Vokurková et al. 2013, Wheatcroft 2015). The analysis revealed a pattern of ordering in the note sequence of mixed songs: Mixed singers tend to make the first part of their song collared-like by putting collared (or collared-like) notes in the beginning of the song, whereas pied notes are put towards the end, often making the last part of the song indistinguishable from the conspecific song. I suggest that there must be some adaptive benefit underlying this pattern. It is possible that this pattern serves as a mitigator of interspecific communication. Given that male pied flycatchers are in general weaker competitors than collared males, mixed singing individuals might choose to emit collared notes in the very first part of their song so as to ensure that they efficiently advertise themselves as soon as they start singing. By more thoroughly defending their territory against collared individuals, mixed singers avoid the risk of being driven off by neighbouring heterospecifics. Given that this pattern is commonly observed in mixed singers, it cannot be overlooked and thus I suggest that future researches attempt to look into it and test this hypothesis by playback experiments.

Maintainance of song convergence

26 CONCLUSION

In this research I compared the singing patterns of 3 singer types by looking at individual notes and their features. Even though partly contradictory to previous findings and claims, the results are of major importance and interest. Although mixed singing pied flycatchers were found to produce a bigger proportion of conspecific notes or notes more similar to conspecific than to collared notes, I showed that when they do produce collared notes, they do so with high fidelity. This accuracy in heterospecific note production is attributed to the species’ plasticity. Mixed singing pied flycatchers are greatly plastic in their singing capacities and not significantly constrained by the biases that their innate conspecific song template imposes on the song learning process. There is, of course, variation across individuals in respect to the proportion of collared notes they incorporate into their repertoire, as well as the accuracy in which they do so. This variation is potentially correlated with the degree of exposure to collared song. Song learning periods in a pied male’s life and interaction with the heterospecifics throughout those periods, are key factors in the mechanisms that control mixed singing.

REFERENCES

Backström N, Sætre GP, Ellegren H. 2013. Inferring the demographic history of European Ficedula flycatcher populations. BMC evolutionary biology 13: 1.

Baker MC, Boylan JT. 1999. Singing behavior, mating associations and reproductive success in a population of hybridizing Lazuli and Indigo Buntings. Condor 1: 493-504.

Ballentine B, Hyman J, Nowicki S. 2004. Vocal performance influences female response to male bird song: an experimental test. Behavioral Ecology 15: 163-8.

Baptista LP. 1996. Nature and its nurturing in avian vocal development. In ‘Ecology and Evolution of Acoustic Communication in Birds’ (Eds DE Kroodsma and EH Miller.) pp. 39–60.

Beecher MD, Brenowitz EA. 2005. Functional aspects of song learning in songbirds. Trends in Ecology & Evolution 20: 143–149.

27 Encyclopedia of Behavioral Neuroscience 164-168.

Bioacoustics Research Program. 2011. Raven pro: interactive sound analysis software. Cornell (NY): The Cornell Lab of Ornithology.

Catchpole C, Slater PJ. 2008. Bird song: biological themes and variations. 2nd _{edition. Cambridge}

University Press, Cambridge.

Clayton NS. 1989. The effects of cross-fostering on selective song learning in estrildid finches. Behaviour 109: 163-74.

Cody ML. 1973. Character convergence. Annual Review of Ecology and Systematics 1: 189-211.

Cynx J, von Rad U. 2001. Immediate and transitory effects of delayed auditory feedback on bird song production. Animal Behaviour 62: 305-12.

Dobkin DS. 1979. Functional and evolutionary relationships of vocal copying phenomena in birds. Zeitschrift für Tierpsychologie 50: 348-363.

Doupe AJ, Kuhl PK. 1999. Birdsong and human speech: common themes and mechanisms. Annual review of neuroscience 22: 567-631.

Ellegren H, Smeds L, Burri R, Olason PI, Backström N, Kawakami T, Künstner A, Mäkinen H, Nadachowska-Brzyska K, Qvarnström A, Uebbing S. 2012. The genomic landscape of species divergence in Ficedula flycatchers. Nature 491: 756-60.

Emlen ST. 1972. An experimental analysis of the parameters of bird song eliciting species recognition. Behaviour 41: 130-71.

28

Eriksen A, Slagsvold T, Lampe HM. 2011. Vocal plasticity–are pied flycatchers, Ficedula hypoleuca, open ended learners?. Ethology 117: 188-98.

Espmark YO, Lampe HM. 1993. Variations in the song of the pied flycatcher within and between breeding seasons. Bioacoustics: The International Journal of Animal Sound and its Recording 5: 33–65.

Güttinger HR. 1979 The integration of learnt and genetically programmed behaviour. Zeitschrift für Tierpsychologie 49: 285-303.

Gelter HP. 1987. Song differences between the pied flycatcher Ficedula hypoleuca, the collared flycatcher F. albicollis, and their hybrids. Ornis Scandinavica 205-15.

Gorissen L, Gorissen M, Eens M. 2006. Heterospecific song matching in two closely related songbirds (Parus major and P. caeruleus): great tits match blue tits but not vice versa. Behavioral Ecology and Sociobiology 60: 260–269.

Grant PR, Grant BR. 1997. Hybridization, sexual imprinting and mate choice. American Naturalist 149: 1–28.

Grether GF, Losin N, Anderson CN, Okamoto K. 2009. The role of interspecific interference competition in character displacement and the evolution of competitor recognition. Biological Reviews 84: 617-35.

Haavie J, Borge T, Bures S, Garamszegi LZ, Lampe H, Moreno J, Qvarnström A, Török J, Sætre G-P. 2004. Flycatcher song in allopatry and sympatry: convergence, divergence and reinforcement. Journal of evolutionary biology 17: 227–237.

Hansen BT, Johannessen LE, Slagsvold T. 2010. Interspecific cross-fostering of great tits (Parus

major) by blue tits (Cyanistes caeruleus) affects inter-and intraspecific communication. Behaviour 147:

413-24.

29

songbirds: a review. Zeitschrift für Tierpsychologie 69: 27–41.

Hudson EJ, Price TD. 2014. Pervasive reinforcement and the role of sexual selection in biological speciation. Journal of Heredity 105: 821-33.

Irwin DE, Irwin JH, Price TD. 2001. Ring species as bridges between microevolution and speciation. Genetica 112: 223-43.

Johannessen LE, Slagsvold T, Hansen BT. 2006. Effects of social rearing conditions on song structure and repertoire size: experimental evidence from the field. Animal Behaviour 72: 83–95.

Kagawa H, Suzuki K, Takahasi M, Okanoya K. 2014. Domestication changes innate constraints for birdsong learning. Behavioural processes 106: 91-7.

Kraaijeveld K, Kraaijeveld-Smit FJ, Maan ME. 2011. Sexual selection and speciation: the comparative evidence revisited. Biological Reviews 86: 367-77.

Kroodsma DE, Miller EH, editors. 1982. Acoustic communication in birds. New York: Academic Press.

Marler P. 1997. Three models of song learning: evidence from behavior. Journal of neurobiology 33: 501-16.

Michaelidis CI, Demary KC, Lewis SM. 2006. Male courtship signals and female signal assessment in

Photinus greeni fireflies. Behavioral Ecology 17: 329-35.

Nadachowska-Brzyska K, Burri R, Olason PI, Kawakami T, Smeds L, Ellegren H. 2013. Demographic divergence history of pied flycatcher and collared flycatcher inferred from whole-genome re-sequencing data. PLoS genetics 9: e1003942.

30

Nowicki S, Searcy WA. 2004. Song function and the evolution of female preferences: why birds sing, why brains matter. Annals of the New York Academy of Sciences 1016: 704–723.

Podos J, Warren PS. 2007. The evolution of geographic variation in birdsong. Advances in the Study of Behavior 37: 403-58.

Qvarnström A, Haavie J, Sæther SA, Eriksson D, Pärt T. 2006. Song similarity predicts hybridization in flycatchers. Journal of evolutionary biology 19: 1202–1209.

Qvarnström A, Bailey RI. 2009. Speciation through evolution of sex-linked genes. Heredity 102: 4-15.

Qvarnström A, Rice AM, Ellegren H. 2010. Speciation in Ficedula flycatchers. Philosophical Transactions of the Royal Society of London B: Biological Sciences 365: 1841-52.

R Core Team. 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.URL: http://www.R-project.org.

Reichard DG, Price JJ. 2008. Species recognition in a vocal mimic: repetition pattern not the only cue used by northern mockingbirds in discriminating songs of conspecifics and brown thrashers. The Wilson Journal of Ornithology 120: 717–724.

Ritchie MG. 1996. The shape of female mating preferences. Proceedings of the National Academy of Sciences 93: 14628-31.

Ritchie MG. 2007. Sexual selection and speciation. Annual Review of Ecology, Evolution, and Systematics 79-102.

Ron SR. 2008. The evolution of female mate choice for complex calls in túngara frogs. Animal Behaviour 76: 1783-94.

31 Molecular Ecology 10: 737-49.

Secondi J, Bretagnolle V, Compagnon C, Faivre B. 2003. Species-specific song convergence in a moving hybrid zone between two passerines. Biological Journal of the Linnean Society 80: 507-517.

Seddon N, Tobias JA. 2010. Character displacement from the receiver’s perspective: species and mate recognition despite convergent signals in suboscine birds. Proceedings of the Royal Society of London B: Biological Sciences 277: 2475–2483.

Slabbekoorn H, Smith TB. 2002. Bird song, ecology and speciation. Philosophical Transactions of the Royal Society B 357: 493–503.

Slabbekoorn H, Smith TB. 2002. Habitat‐dependent song divergence in the little greenbul: An analysis of environmental selection pressures on acoustic signals. Evolution 56: 1849-58.

Thorpe WH. 1958. The learning of song patterns by birds, with reference to the song of the chaffinch

Fringilla coelebs. Ibis 100: 535–570.

Tobias JA, Seddon N. 2009. Signal design and perception in Hypocnemis antbirds: evidence for convergent evolution via social selection. Evolution 63: 3169–3189.

Tobias JA, Cornwallis CK, Derryberry DP, Claramunt S, Brumfield RT, Seddon N. 2014. Species coexistence and the dynamics of phenotypic evolution in adaptive evolution. Nature 506: 359–363.

Vabishchevich AP, Formozov NA. 2010. Song variability in pied flycatchers Ficedula hypoleuca: impact of the sympatry with collared flycatchers F. albicollis. Acta Ornithologica 45: 189-202.

Vallin N, Rice AM, Bailey RI, Husby A, Qvarnström A. 2012. Positive feedback between ecological and reproductive character displacement in a young avian hybrid zone. Evolution 66: 1167-79.

32

Verzijden MN, Ten Cate C, Servedio MR, Kozak GM, Boughman JW, Svensson EI. 2012. The impact of learning on sexual selection and speciation. Trends in Ecology & Evolution 27: 511-9.

Vokurková J, Petrusková T, Reifová R, Kozman A, Mořkovský L, Kipper S, Weiss M, Reif J, Dolata PT, Petrusek A. 2013. The causes and evolutionary consequences of mixed singing in two hybridizing songbird species (Luscinia spp.). PLoS One 8: e60172.

Wheatcroft D. 2015. Repetition rate of calls used in multiple contexts communicates presence of predators to nestlings and adult birds. Animal Behaviour 103: 35-44.

Williams H. 2004. Birdsong and singing behavior. Annals of the New York Academy of Sciences 1016: 1-30.