Behavioral
Ecology
The official journal of the
ISBE
International Society for Behavioral Ecology
Behavioral Ecology (2014), 25(6), 1380–1394. doi:10.1093/beheco/aru136
Original Article
Reciprocal territorial responses of parapatric
African sunbirds: species-level asymmetry and
intraspecific geographic variation
Jay P. McEntee
Museum of Vertebrate Zoology and Department of Integrative Biology, 3101 Valley Life Sciences
Building, University of California, Berkeley, Berkeley, CA 94720, USA
Received 18 February 2014; revised 2 July 2014; accepted 3 July 2014; Advance Access publication 21 August 2014.
Territorial competition regularly occurs between ecologically similar species with substantial divergence in territorial signals. Strong
responses to heterospecific signals can result either from the conservation of broad response functions or from species discrimination
with competitor recognition. The African sunbird sibling species Nectarinia moreaui and Nectarinia fuelleborni share similar niches
and an extremely narrow parapatric boundary. These species sing strikingly different songs, yet have only subtly different morphologies. Familiarity with sibling species is only possible in the small contact zone where their ranges abut. A series of simulated territorial
intrusion experiments reveal asymmetry in response frequency to heterospecific songs: N. moreaui respond less frequently to N. fuelleborni than to conspecific songs, whereas N. fuelleborni respond with an approximately equal frequency to both N. moreaui and conspecific songs. However, there is strong geographic variation in heterospecific song response among N. fuelleborni populations, with
one allopatric population exhibiting stronger responses to conspecifics than to N. moreaui. These results indicate that N. fuelleborni
populations can retain broad response functions absent contact with N. moreaui, such that competitor recognition is not necessary
to explain strong territorial responses to N. moreaui. Strong geographic variation within N. fuelleborni, however, implies that spatial
replication of allopatric/sympatric treatments should benefit future territorial response experiments. Lastly, a multimodal experiment in
the contact zone area reveals that, following initial approaches to song signals, response differences are not greater to conspecific
signals. This result suggests that associative learning may result in high heterospecific aggression via competitor recognition, even in
N. moreaui, where syntopy occurs.
Key words: bird song, Eastern Afromontane, heterospecific aggression, Nectariniidae, range boundaries, social signals.
Introduction
When recently diverged animals interact, intense competition for
resources can occur (Mayr 1942; Lack 1944), including competition
for mates if hybridization is attempted (Groening and Hochkirch
2008). Such competition is mediated by social signals, which are
often divergent during secondary contact regardless of whether the
interacting forms have diverged in ecological niche (West-Eberhard
1983). Although social signal divergence has been characterized
across a wide swath of diversity, behavioral response to such signals
among closely related competitors is less well understood (Grether
et al. 2009; Peiman and Robinson 2010; Pasch et al. 2013). When
substantial signal divergence characterizes interacting species and
competition between them is great, each faces the dual challenge
of correctly associating the signals with the competitive threats they
Address correspondence to Jay P. McEntee, who is now at Department of
Ecology and Evolutionary Biology, University of Arizona, 1041 E. Lowell
Street, Tucson, AZ 85721, USA. E-mail: [email protected].
© The Author 2014. Published by Oxford University Press on behalf of
the International Society for Behavioral Ecology. All rights reserved. For
permissions, please e-mail: [email protected]
represent (Grether 2011), and of calibrating aggressive responses
under a potentially greater uncertainty of outcomes than with conspecifics (CS) (Peiman and Robinson 2010).
Parapatry, or mutually shared range boundaries, may bring this
challenge into focus better than any other circumstance. Parapatric
boundaries generally indicate high ecological similarity (Bull 1991),
yet a very small proportion of each species interacts with the competing heterospecifics (HS). This rarity of interspecific interaction
limits the development of HS categorization via associative learning
(Irwin and Price 1999; Martin and Martin 2001; Jankowski et al.
2010; Svensson et al. 2010), the evolution of heritable HS categorization (Wellenreuther et al. 2010; Grether 2011), character displacement forms of selection (Brown and Wilson 1956; Grether et al.
2009), and hybridization (Patton 1993). However, response behaviors during interspecific interactions may have major fitness consequences; for example, if an individual fails to exhibit aggression
toward a competitive threat. Moreover, such behaviors are ecologically important at a larger scale, as they can impact whether mutual
McEntee • Territorial response of parapatric sunbirds
range boundaries remain as boundaries (Jankowski et al. 2010), and
whether their position remains geographically stable (Pearson and
Rohwer 2000). As such, the evolution of responses to parapatric HS
is a matter of interest to both evolutionary biologists and ecologists.
Territorial animals provide an opportunity to study competitor
perception and response via field manipulations known as simulated territorial intrusion experiments (or playback experiments
in the case of auditory signals). Territoriality, when individuals or
groups defend resources over a limited area, is mediated by social
signals (Noble 1939; Nice 1941). Territorial defense is adaptive
when investment in defense activities is commensurate with marginal fitness benefits of an attempt at intruder exclusion (Hinsch
and Komdeur 2010). Intraspecific territoriality is extremely common among animals, and the evolution of perception and response
to territorial competitors may be driven primarily by intrapopulation or intraspecific interactions. Because of the great importance
of CS interactions in determining fitness, it is expected that signals and signal perception should coevolve tightly: Sensory functions which determine territorial responses should approximately
describe the limits of CS signal variation, that is, species recognition (Emlen 1972; Ryan and Rand 1993; Ord and Stamps 2009).
Species recognition, as indicated by greater territorial responses to
CS versus non-CS, is indeed the dominant pattern in meta-analyses
of simulated intrusion experiments (Ord and Stamps 2009; Ord
et al. 2011). Meanwhile, the evolution of HS threat recognition and
territorial aggression can occur via a number of pathways: as a byproduct of the coevolution of CS signals and recognition (Murray
1971, 1976, 1981); by adaptation resulting from interactions with
HS (i.e., HS competitor recognition—Orians and Willson 1964;
Cody 1969; Grether et al. 2009; Peiman and Robinson 2010); as a
by-product of selection on unrelated traits (via trait correlation or
linkage—Peiman and Robinson 2010); and by genetic or cultural
drift. Heterospecific territorial responses can also develop as a result
of learning, in which case heritable, species-specific HS responses
may not be observed (Irwin and Price 1999; West-Eberhard 2003).
Insight into the evolution of territorial response to HS can be
gained by examining spatial variation in HS aggression with respect
to the presence or absence of a particular HS (Peiman and Robinson
2010). For example, comparisons of territorial responses to HS in
allopatric and sympatric populations can be informative in assessing the degree to which HS aggression represents treatment as a CS
(i.e., recognition errors: Murray 1976) versus aggression aimed at HS
after successful discrimination. Some studies have found responses of
a greater magnitude to HS in sympatry versus allopatry, suggesting
1381
that aggression is triggered by sympatric HS that represent territorial
threats. (fish: Sabo and Pauley 1997; salamanders: Nishikawa 1987;
birds: Rohwer 1973; Catchpole 1978; Reed 1982; Catchpole and
Leisler 1986; Sorjonen 1986; Prescott 1987; Baker 1991; Gil 1997;
Sedlacek et al. 2006; Tobias and Seddon 2009). This pattern could
indicate either a broadening of the psycho-perceptual concept of CS
or a discrimination between CS and HS, with HS competitor recognition. Conversely, at least 5 studies (Gill and Murray 1972; Morrison
1982; Nuechterlein and Buitron 1998; Amezquita et al. 2006;
Anderson and Grether 2010) have found greater responses to HS in
allopatry than in sympatry. The similarity of the signals between species pairs suggests that individuals in allopatric populations are treating signals that are unfamiliar, or at least not characteristic of their
breeding environments, as CS in the absence of the opportunity to
associate the signals with additional aspects of the HS biology (see
also Nelson 1989; Ord and Stamps 2009). Response shifts in either
direction in birds are likely to be frequently facilitated by associative
learning (Prescott 1987; Gil 1997; Price 2008), and under narrow
parapatry such associative learning can be limited to the very narrow
band where interactions are frequent (Jankowski et al. 2010).
Here, male responses to simulated territorial intrusions of
HS and CS in the ecologically similar species Moreau’s Sunbird
Nectarinia moreaui and Füelleborn’s Sunbird Nectarinia fuelleborni are
examined (Bowie et al. 2004). These birds share a recently discovered, extremely narrow parapatric boundary in southern Tanzania
(estimated cline width < 5 km; McEntee JP, Penalba JV, Werema C,
Mulungu E, Mbilinyi M, Bowie RCK, in preparation) with limited
evident hybridization and no introgression (McEntee JP, Penalba
JV, Werema C, Mulungu E, Mbilinyi M, Bowie RCK, in preparation). The 2 species exhibit extremely different song phenotypes
(Figure 1a,b, audio files in Supplementary Data), but only subtly
different plumage phenotypes and morphology (Bowie et al. 2004;
Figure 1c,d and Supplementary Figure S1). Their strong song differences are maintained in adjacent populations despite the possibility for HS copying introduced by song learning (McEntee JP,
Penalba JV, Werema C, Mulungu E, Mbilinyi M, Bowie RCK,
in preparation), as sunbirds are oscine songbirds that presumably
develop their complex songs through the learning process known
from this clade (Brenowitz and Beecher 2005). The 2 species are
similar in bioclimatic niche (McEntee JP, Penalba JV, Werema C,
Mulungu E, Mbilinyi M, Bowie RCK, in preparation) and autecology as indicated through natural history observation, suggesting they are ecological competitors in parapatry (Mayr 1963; Bull
1991) in addition to competing, to some degree, for mates.
Figure 1
Representative sonograms of (a) Nectarinia moreaui and (b) Nectarinia fuelleborni territorial songs. The difference in duration is typical, as N. fuelleborni’s songs
average 2.5–3 times as long. The greater exploration of frequency space by N. fuelleborni should also be noted. Depictions of (c) N. moreaui male (l) and female
(r) and (d) N. fuelleborni male and female show the subtle plumage and morphometric distinctions between species. Male N. moreaui possess pale yellow feathers
on the side of the breast adjacent to the red breast band, whereas N. fuelleborni males share only the yellow pectoral tufts. Nectarinia moreaui males average larger
with longer bills. See also photographs in Supplementary Figure S1.
Behavioral Ecology
1382
The results of 2 types of simulated territorial intrusion experiments that investigate understudied aspects of social signal perception are reported: multimodal signaling and geographic
variation in response to a recently diverged sibling species. The
first experiment tests whether multimodal signals encode species
identity in the 2 species. Combined mount and song presentations
in the vicinity of the parapatric boundary are used to test whether
territorial N. moreaui and N. fuelleborni males distinguish between
HS and CS (see Table 1). The very high probability of response
to N. moreaui’s song by N. fuelleborni males in this first experiment
raised the prospect that N. fuelleborni males near the parapatric
boundary either 1) recognize N. moreaui as a competitor despite
the minimal level of sympatry, and therefore interaction, between
the species; or 2) respond aggressively to N. moreaui at a species
level as a consequence of either the failure to distinguish between
acoustic signals of each species or greater general aggression. The
second set of experiments (song-only experiments) investigates
these possibilities by examining geographic variation in response
to HS in N. fuelleborni (see Table 1). Symmetry in responses
between species is also tested by a similar examination of HS
response in N. moreaui. To further understand the spatial variation
in relative HS versus CS responses, the results of the 2 experiments are synthesized by examining effect sizes of CS versus HS
songs on response probability for the 3 N. moreaui and 4 N. fuelleborni populations tested.
Methods
Natural history of focal taxa
Nectarinia fuelleborni (née Cinnyris mediocris fuelleborni) and N. moreaui
(née Cinnyris moreaui or Cinnyris mediocris moreaui) are nectar-feeding
sunbirds (Nectariniidae) that occupy montane forest and forest edge
habitats in eastern and southeastern Africa. These taxa are sibling
species, and may be sister species, depending on their relationship
with the third member (Nectarinia loveridgei) of their 3-taxon clade
(Bowie et al. 2004). Both species have “sky-island” distributions,
where component populations occur as geographical isolates separated by inhospitable lowlands. Their breeding distributions range
from ~1400 to 2400 m elevation, but both species are elevational
migrants that descend to lower elevations during the nonbreeding season where suitable habitat is available Nectarinia moreaui is a
Tanzanian endemic which is geographically restricted to the central Eastern Arc Mountains, a chain of ancient, isolated mountain
blocks created through punctuated fault-block action beginning in
the early Tertiary (30 million years ago) and subsiding during the
Miocene (6 million years ago, Lovett 1993). Within the Eastern
Arc Mountains, N. moreaui is limited to the Nguru, Ukaguru, Wota,
Rubeho, and northeast Udzungwa Mountains (Bowie et al. 2004).
Nectarinia fuelleborni (including N. fuelleborni bensoni) has a broader latitudinal distribution, spanning the mosaic of volcanic and igneous
mountains ringing Lake Malawi, from southern Tanzania to southern Malawi/central Mozambique. The distributions of N. fuelleborni and N. moreaui abut in the northeast Udzungwa Mountains
(Figure 2).
The 2 species exhibit subtle morphometric differences. N. moreaui
males average slightly larger, with slightly longer bills (Bowie et al.
2004; McEntee JP, unpublished data). All plumage differences are
subtle, but there are several visible differences. Nectarinia fuelleborni
males possess yellow pectoral tufts only, whereas N. moreaui individuals have yellow pectoral tufts and a patch of yellow feathers
inward on the breast from the pectoral tufts (Bowie et al. 2004).
The red breast band of N. moreaui appears more orange-red than
the more grenadine band of N. fuelleborni, whereas N. fuelleborni’s red
band averages taller. The rump of N. fuelleborni is iridescent green,
whereas the rump of N. moreaui is dull green.
Males sing starkly different territorial songs. N. fuelleborni sings
songs that average 2.5 times as long in duration. The peak frequency range of N. fuelleborni’s songs is 1.6 times as broad as in N.
moreaui’s songs. Gaps between elements in N. fuelleborni’s songs
are 3.4 times as long as gaps those in N. moreaui’s songs (McEntee
JP, Penalba JV, Werema C, Mulungu E, Mbilinyi M, Bowie RCK,
in preparation; see Figure 1). These songs appear to have the dual
functionality generally attributed to passerine bird songs: mate
attraction (intersexual) and territorial signaling (intrasexual—
Catchpole and Slater 2008). Extensive field observations have indicated that singing behavior is seasonal. Territorial singing in the
area of the study commences in May and June, at the beginning
of the breeding season, which lasts into September. Both species
exhibit social monogamy, and territorial singing activity drops precipitously when the brood of a male’s social partner hatch. Males
cease singing when they begin provisioning young in the nest. All
playback experiments are carried out during periods of high singing rates at the population level, with breeding activity confirmed
by the discovery of active nests during the course of the field
experiments for all 7 studied populations. Breeding-season observations have revealed that males using acoustic signals within the territory of another male elicit strong territorial responses, including
Table 1
A priori hypotheses and predictions for both experiments
Hypotheses
Experiment 1: multimodal experiment Divergent signal phenotypes encode species
identity (species recognition)
Song plays a stronger role than morphology
(including plumage) in species recognition
Experiments 2a, 2b: geographic
Nectarinia fuelleborni is generally more aggressive
variation in song response
than Nectarinia moreaui
Asymmetric discrimination: N. fuelleborni does
not discriminate by song (recognition error),
but N. moreaui discriminates
Aggressive responses to HS are induced by
competition (competitor recognition), not
recognition error
Predictions
Greater response probability and magnitude to CS signals
Response magnitude: CS song/CS mount > CS song/
HS mount > HS song/CS mount > HS song/HS mount
Nectarinia fuelleborni exhibits strong territorial responses to
both N. moreaui and Nectarinia olivacea; N. moreaui exhibits
weak responses to both HS
Nectarinia fuelleborni exhibits similar aggression toward
N. moreaui and CS songs in populations near and distant
to N. moreaui; N. moreaui exhibits stronger CS response
Response probability and magnitude to HS signals are
greater in contact zone populations
McEntee • Territorial response of parapatric sunbirds
1383
Simulated territorial intrusion experiments
Figure 2
Map of the study area. Colors in pie charts indicate frequencies of
species from genetic samples of populations (McEntee JP, Penalba JV,
Werema C, Mulungu E, Mbilinyi M, Bowie RCK, in preparation). Purple
represents Nectarinia fuelleborni, and green represents Nectarinia moreaui.
Territorial intrusion experiments were conducted in 2009 (Experiment 1,
multimodal) at Ikokoto and Image; and in 2010 (Experiments 2a and 2b,
song-only) in the remaining 5 populations. It should be noted that there
are 2 “arms” of the contact zone area: the Image–Ikokoto contact area
and the Ndundulu–Nyumbanitu contact area. In the limited area where
sympatry occurs, subject species identity was confirmed by plumage and/
or song.
vigorous chases, which at times are preceded by soft songs (Searcy
and Nowicki 2006; Hof and Hazlett 2010; Templeton et al. 2012).
The 2 species share a remarkably narrow parapatric boundary. The position of this boundary between species is surprising,
because it occurs within the Udzungwa highlands, unassociated
with any known ecological cline and midway between 2 biogeographic breaks for montane forest taxa: the Makambako Gap
(Lovett et al. 2001) and Great Ruaha River gorge (Fjeldså et al.
2010). The geographic position and narrow width of the boundary strongly suggest that the 2 species mutually exclude one
another from a suitable habitat (West-Eberhard 1983); that is,
their abundances are coregulated. There is one known location
(within Nyumbanitu forest, see Figure 1) where males of both
species hold adjacent territories. Interspecific aggression has
been observed at this location, suggesting interspecific territoriality. However, it is unknown whether boundaries of adjacent territories are stable among HS. Molecular evidence indicates that
hybrids are produced at a low frequency in the contact zone, and
that individuals of mixed ancestry are at low frequency in the
study populations in the contact zone (McEntee JP, Penalba JV,
Werema C, Mulungu E, Mbilinyi M, Bowie RCK, in preparation). Furthermore, an examination of bioclimatic niches of the
2 species indicates that they possess extremely similar niches
(McEntee JP, Penalba JV, Werema C, Mulungu E, Mbilinyi
M, Bowie RCK, in preparation): They are effectively ecological replacements across space in montane forests (Mayr 1963;
Bull 1991; Price 1998). Though both species thrive in forests
and forest edges, within these environments they are generalists in microhabitat preferences for both foraging and nest sites
(McEntee JP, personal observation).
Experiment 1: Song playback/mount presentation
The first experiment was carried out in 2009 at the forest localities
locally known as Ikokoto (7.69°S, 36.14°E; elevation ~1790–1910
m) and Image (7.41°S, 36.15°E; elevation ~2000–2260 m), in the
northeast Udzungwa Mountains, Tanzania. Ikokoto is dominated
by N. fuelleborni, whereas Image is dominated by N. moreaui (see
McEntee JP, Penalba JV, Werema C, Mulungu E, Mbilinyi M,
Bowie RCK, in preparation; Werema et al. 2013). A mount presentation/song playback design, with the species of the mount
and song fully crossed, was employed, and it was similar to the
design used by Uy et al. (2009). Seven N. moreaui and 8 N. fuelleborni tracks with ~60 consecutive seconds of high signal-to-noise
ratio bout singing were chosen for playback from digital recordings
made by J.P.M. in 2008–2009 (16 bits, 44.1 kHz sampling). Noise
below 1.5 kHz and above 11 kHz was then removed by bandpass
filtering (Raven Pro 1.3, Bioacoustics Research Program 2008) for
all recordings. Each song playback sequence began with 90 s of
silence, followed by 2 min of song, and then 5 min of silence. The
2-min song period was created by repeating the ~1-min bout initially chosen. As in Uy et al. (2009), 2 male mounts were used to
represent each species. The limited number of mounts may introduce a pseudo-replication issue (Kroodsma 1989), but because there
are multiple discrete and quantitative plumage differences between
the 2 species (Bowie et al. 2004), a small number of mounts may
adequately represent species-level differences. Pseudo-replication
was further controlled for by including mount identity as a random
effect in the general linear mixed modeling approach used in the
response duration analysis for this experiment (see Duration models). From the 15 song tracks and 4 mounts, 32 stimulus combinations were created to represent 4 treatments: N. fuelleborni song ×
N. fuelleborni mount, N. fuelleborni song × N. moreaui mount, N. moreaui
mount × N. fuelleborni song, and N. moreaui mount × N. moreaui song.
Each subject was tested in a single trial, where it responded to a
single stimulus type. The order of presentation of the 32 stimulus combinations was randomized. Two trials were discarded (one
because the sex of the responding bird could not be confirmed, and
one because of habituation concerns).
I sought subjects for this experiment by walking forest roads,
trails, and forest edges, while listening for vocalizing male sunbirds. Males of both species vocalize throughout daylight hours,
but I avoided performing trials early in the morning when ambient temperatures were low, as I discovered in preliminary trials that
birds rarely responded in these conditions. Experiments were, thus,
initiated between 0933 and 1705, and were run from 15 July to 1
August 2009. When a vocalizing male was found, the speaker and
mount were placed at a distance ~10–30 m either from the individual or from a perch recently used by the subject for vocalization.
The speaker (Anchor Minivox, Anchor Audio) and mp3 player (3rd
generation Apple iPod Nano, Apple Inc.) used for song playback
were placed on the ground, and mounts were affixed in an upright
position to vegetation above the speaker (height 0.5–2 m). As much
as possible, I performed trials with subjects who were either singing
or had sung immediately before the trial was initiated. Overall, 48
of the 62 subjects retained for analyses sang within the 5 min before
trial initiation. The remaining subjects exhibited behaviors indicative of territory defense, that is, extensive calling, chasing another
individual, close male–female interaction, or exclusively using an
area with multiple nectar sources during the pretrial period. Three
observers were positioned in different directions from the speaker
Behavioral Ecology
1384
and mount, with one (J.P.M.) recording observations of subject
responses onto a digital voice recorder. A second observer made an
audio recording of the trial using a Sennheiser ME67/K6 power
module microphone combination and a Marantz PMD670 solidstate digital recorder. To check distance estimates made during the
trial, distances to perches used during approaches were measured
with a tape. Transcriptions of the recorded observations were compared with audio recordings of the playback trials before analysis as
a secondary check on the timing of events during trials. The time
period from the start of the song playback through the end of the
5-min silent period following the song stimulus, a total of 7 min,
was taken as the response period for Experiment 1. It should be
noted that the response period is shorter (4 min) in Experiment 2.
Experiments 2a and b: geographic variation in relative
species response
The second set of experiments were designed to build on and complement the results of the first experiment. The strongest result of
the first experiment was an asymmetry in response to HS songs:
N. moreaui exhibited responses which were consistent with species recognition, whereas N. fuelleborni responded with an equal or
greater frequency to HS than to CS songs. The great differences in
song between the species suggested the possibility of HS competitor
recognition (discrimination between species, but high aggression)
in N. fuelleborni, despite the very low global levels of sympatry of
the 2 species and the rarity of N. moreaui at Ikokoto. Alternatively,
the strong response by N. fuelleborni could be a result of recognition function conservatism with permissive recognition, even as
N. moreaui exhibits strong discrimination suggestive of a narrow recognition function. Song phenotypes and phylogenetic relationships
of sunbirds closely related to the focal species pair suggest that their
common ancestor exhibited songs which were more similar in finescale spectrotemporal structure to N. moreaui (McEntee JP, Penalba
JV, Werema C, Mulungu E, Mbilinyi M, Bowie RCK, in preparation), suggesting the possibility that N. fuelleborni has retained
response to ancestral song forms. The song-only experiment was
designed to take advantage of the geography of the 2 species’
respective distributions, which enables sampling of populations at
distances great enough from the HS such that there is no possibility
that HS aggression is locally maintained by ongoing interspecific
interactions. Furthermore, the presence of multiple contact zone
populations was taken advantage of in order to spatially replicate
playback experiments near the parapatric boundary. For N. fuelleborni, it was predicted that under HS competitor recognition with
species discrimination, strong responses to N. moreaui would be
found only near the parapatric boundary. However, under recognition function conservatism, all N. fuelleborni populations would fail
to discriminate between CS and HS songs (see Table 1). Spatial
replication of populations distant from contact is rarely performed,
and was desirable in this case not only as it is generally desirable
in experimental design but also because it was possible that HS
response could decrease with distance from the parapatric boundary in N. fuelleborni.
To examine geographic variation in territorial response to HS
versus CS songs, I performed a set of 2 song-only territorial intrusion experiments with spatial replication in 1) N. moreaui (2 populations, see Figure 2) and 2) N. fuelleborni (3 populations). Localities
for N. moreaui experiments were Ndundulu forest (7.77°S, 36.49°E;
elevation 1800–2250 m) and Mafwemiro forest (6.94°S, 36.59°E;
elevation ~1850–2100 m); localities for N. fuelleborni experiments
were Mt. Rungwe (9.16°S, 33.61°E; elevation ~1600–1900 m),
Mufindi (8.6°S, 35.3°E; elevation ~1840–1950), and Nyumbanitu
forest (7.81°S, 36.39°E; elevation ~1550–2000 m). The experiments were carried out between 9 June and 20 July 2010. As in
Experiment 1, I avoided carrying out experiments when the ambient temperature was especially low early in the morning. Active
nests were found in all 5 experimental populations during the
course of the experiments. I employed a repeated-measures design
similar to that used by Martin and Martin (2001), where each subject was given the opportunity to respond to 3 different stimulus
types: CS song, HS song, and a control song. Nectarinia olivacea was
chosen for the control song species, because it is a forest-dependent
sunbird that occurs syntopically with the 2 focal species across
their distributions, but is distantly related to both focal species and
sings extremely dissimilar songs. Aggressive responses to N. olivacea
should, therefore, result exclusively from attempts to exclude an
ecological competitor, and should represent neither competition
for mates nor misdirected aggression with failure to distinguish
between CS and N. olivacea song. In contrast to other negative controls such as white noise, I expected a priori that some individuals would respond aggressively to the control stimulus. I argue that,
because the goal of this study is to understand variation in response
to CS and HS, the inclusion of an ubiquitous ecological competitor
as a control improves the informativeness of the experiment versus
the use of a negative control (e.g., white noise).
Playback tracks consisted of 60 s of silence and, subsequently,
a 2-min stimulus period followed by 3 min of silence. The 3-min
silent period was chosen as a compromise between the likelihood
that an individual might leave the playback area during an overly
long silent period, and that the response to the previous stimulus
might be continuing if the silent period was too short. I examined
responses from the multimodal experiment as a guide to determine an appropriate response period. For duration measurements
calculated afterward, the responses of individuals who started
any stimulus period within 8 m were calculated from when they
moved toward the speaker. Six playback tracks were produced, representing all possible orders of stimulus types, to be used in both
experiments. To produce these tracks, we selected songs as in the
multimodal experiment. For this experiment, N. moreaui and N. fuelleborni recordings were bandpass filtered at 2.5–10 kHz. Nectarinia
olivacea songs were lower in frequency and were bandpass filtered
at 1.5–10 kHz. The 6 tracks incorporated a total of 6 N. fuelleborni,
6 N. moreaui, and 4 N. olivacea recordings. Tracks were selected from
among the highest quality recordings to maximize geographic sampling in all species. Outside of indicated differences, trials were
performed as in the multimodal experiment. In the song-only
experiments, I sought to perform 18 trials for each population.
Some subjects did not respond to any of the 3 stimuli, and these
subjects were not included in analyses. Responses were recorded
for the 4-min period beginning with the start of each 2-min song
stimulus period and ending at 2 min after, unlike in the multimodal
experiment where the response period was 7 min.
Analyses
I analyzed playback responses in 2 parts (see Jankowski et al.
2010): 1) a binary response variable indicating approach or failure
to approach within 8 m and 2) the duration of approach within
8 m. All analyses were performed using Bayesian Markov chain
Monte Carlo (MCMC) generalized linear mixed models (GLMMs),
using the mcmcGLMM package (Hadfield 2010) for R (R Core
Team 2012). This package uses either Metropolis-Hastings or
slice sampled (Damien et al. 1999) MCMC simulations to develop
McEntee • Territorial response of parapatric sunbirds
posterior probability distributions for model parameters. The first
analysis for both experiments tested which factors were responsible for the binary response variable approach versus no approach
(i.e., logistic regression), interpreted as indicating whether the song
stimulus was initially perceived as a potential territorial threat. The
duration of approach within 8 m was interpreted as an indicator of
total territorial aggression, as it was correlated with subjective scores
of aggression based on the array of behaviors exhibited during
responses. These behaviors were too diverse in form among individuals to reduce in dimensionality using multivariate techniques
(e.g., principal components analysis), which assume normality for
all input variables; that is, the distributions of individual response
variables (e.g., time spent singing, number of calls, fly-bys) were
non-normal, and could not be transformed to normality because
they tended toward zero-inflation or bimodality. Specifications and
coding for all models can be found in Appendix A. Priors were
specified so as to be proper, but close to flat or weakly informative.
Alternate values for prior distribution parameters were checked
to ensure that posterior distributions and the statistical inferences
made from these distributions did not vary qualitatively with
prior specification. To assess statistical significance between treatments, I calculated the 95% highest posterior probability density
(95% HPD) intervals for contrasts between treatment levels from
the MCMC posterior probability distributions. Overall, 95% HPD
interval contrasts that did not include 0 were interpreted as statistically significant differences.
Binary (approach vs. no-approach) models
In the great majority of Experiment 1 trials, subjects did not have
an unobstructed view of the mount when the trial started, or when
the song playback started. Trials were conducted in forests, forest
edges, and forest gaps, and because of this environment and the fixing of the mount to standing vegetation, subjects had to approach
and maneuver to find the mount. Even in the small number of
instances when the mount was visible to a subject before song playback began, approaches were never instigated by the presence of
the mount without song playback. Thus, the initial approach to
within 8 m during Experiment 1 was unlikely to be affected by the
mount species. To accommodate the possibility, however, an initial GLMM with logit link was fit to the binary approach versus
no-approach response variable. This model included mount species (HS versus CS), song species (HS versus CS), subject species
(N. fuelleborni or N. moreaui), and their interactions as fixed effects,
and the specimen and track identities as random effects, with
separate random-effect intercepts for each subject species. I ran
the MCMC chains for 550 000 iterations with a 50 000 iteration
burn-in, and thinned the posterior sample by a factor of 10, resulting in 50 000 posterior distribution samples from the chain. To
check convergence of the MCMC simulations, I examined effective sample size estimates and visualized the posterior probability
distributions of parameters. Second, I fit a similar model, but without mount effects. Inclusion of mount species did not substantially
improve the model fit (Δ Deviance Information Criterion = 0.66),
and the parameter estimates for the effect of the HS mount were
close to 0. Because few subjects were able to see the mount before
approach, and as a model with ‘mount species’ as a parameter
does not substantially improve fit, statistical inferences of treatment effects were made from the model without the mount species
as a fixed effect. Treatment effects were quantified by computing
contrasts from the MCMC chain for the posterior probabilities of
response to HS versus CS for each subject species, with the null
1385
hypothesis that responses were equally probable to HS and CS
songs within each species. Treatment effect estimates and credible
intervals are presented for these comparisons.
For Experiment 2a (N. moreaui subjects, song-only experiments)
and Experiment 2b (N. fuelleborni subjects, song-only experiments),
binomial-family models were developed separately for N. moreaui
and N. fuelleborni as subjects to analyze treatment effects of the 3
stimuli. Because of the repeated-measures design, individual identity was included as a random effect. Identity of the playback track
was included as a random effect to account for stimulus order and
track-specific response variation. The fixed effects include the treatment (song species), subject population, and their interactions. The
details of the model and the specification of the MCMC chain
were otherwise similar to the first model. Posterior means and credible intervals were presented for each treatment level. Effect size
estimates for species recognition (conditional differences in probability of responding to CS versus HS song) were calculated for a
comparison with the results of Experiment 1.
For both binary models, prior distributions were selected to be
minimally informative on the probability scale. A half-Cauchy distribution was used as a prior in the random-effects variances to
restrict sampling space to reasonable values (Gelman 2006), and
parameter expansion (van Dyk and Meng 2001; Gelman 2006)
and slice sampling were employed (Damien et al. 1999) to improve
MCMC sampling efficiency (see Appendix A for code). The residual variance for these models was fixed at 1, as this source of variance cannot be estimated from categorical responses (Hadfield
2010).
Duration models
For subjects that approached the speaker/mount in Experiment
1, the combined effects of song species and mount species on the
intensity of response were tested by fitting a Poisson-family model
to the log link of approach duration within 8 m. Poisson models
were fit with an additive over-dispersion model (Hadfield 2010).
The primary goal of this study was to test the hypothesis that
responses exhibit a pattern which is consistent with species recognition, with the secondary hypothesis that different song treatments
should exhibit greater response differences because they are likely
easier to distinguish at a species level than morphology. This set
of 2 hypotheses results in the prediction that the order of intensity
of responses by treatment is as follows: CS song/CS mount > CS
song/HS mount > HS song/CS mount > HS song/HS mount (see
Table 1; also see Uy et al. 2009). This result would be consistent
with predictions of the species recognition hypothesis where traits
that are perceptually relevant to cognitive distinction have evolved
faster in song than in morphology. As sample size was limited in the
reduced data set (subjects who approached to within 8 m), especially
for N. moreaui responses to HS songs, the data for both species were
pooled and subject species was not a fixed effect. This left the 4
stimulus categories of interest with a sample size range of 7–11 (CS
song/CS mount: 9, CS song/HS mount: 11, HS song/HS mount:
7, HS song/HS mount: 9; total n = 36). As has been recommended
for simulated intrusion experiments with limited exemplar stimuli
(Kroodsma et al. 2001) and as implemented in mixed model frameworks by multiple territorial intrusion studies (Grant and Grant
2002; Newman et al. 2006; Uy et al. 2009), the variation associated
with individual exemplars for mounts (2 exemplars per species) and
for song tracks was controlled for (7 N. moreaui and 8 N. fuelleborni
exemplars) by including the exemplar identities as random effects.
To account for variation between subject species in these random
Behavioral Ecology
1386
effects, the random-effect intercepts were estimated by subject species. Default specifications were followed for the fixed-effect prior
distributions for Poisson-family models, which are normal with a
mean of 0 and a large variance on the logit scale (108). Prior distributions for the random-effects variances and the residual variance (Poisson over-dispersion) are inverse-Wishart priors, with the
variance limit parameter V = 1 and the degree-of-belief parameter
nu = 0.002. The MCMC chains were run and checked as in the
binomial model above, in this case with 140 000 iterations and a
40 000 iteration burn-in.
As responses to N. olivacea were infrequent, secondary analyses
were performed for Experiments 2a and 2b by examining the difference in response duration to CS and HS (i.e., either N. moreaui
or N. fuelleborni, but not N. olivacea) songs for those subjects who
responded to at least 1 of the 3 signals. For these subjects, the duration of responses to HS songs was subtracted from CS songs. This
approach has the advantage of completely accounting for amongsubject variation in response. These difference values were approximately normally distributed for both species (Shapiro–Wilks tests:
N. moreaui, n = 31, P = 0.59; N. fuelleborni, n = 41, P = 0.71), hence
I fit Gaussian-family GLMMs separately for N. moreaui and N. fuelleborni subjects. The single fixed effect in these models was population,
and the single random effect was the stimulus track identification,
which also encompasses variation in order effects. For this model, the
default prior for the mean and variance of fixed effects for Gaussianfamily models in mcmcGLMM and an inverse-Wishart prior for the
random effects and residual variances were used (V = 1, nu = 0.002).
These models yielded similar results. The Gaussian models were run
for 550 000 iterations with a 50 000 iteration burn-in.
Results
Territorial response behaviors
Males of both species responded to simulated territorial intrusions by approaching the speaker and/or mount, rapidly moving
from perch to perch, calling, singing loudly, singing soft songs,
erecting their yellow pectoral tufts, making flyovers (rare), and/
or facing the mount in the multimodal experiment (Table 2:
Table 2
Experiment 1: multimodal signals
Species
moreaui
Treatment
CS song/CS mount
CS song/HS mount
HS song/CS mount
HS song/HS mount
fuelleborni CS song/CS mount
CS song/HS mount
HS song/CS mount
HS song/HS mount
Close
Approach Duration dist
5/7
7/8
2/8
3/8
4/8
4/8
5/7
6/8
113 ± 54a
296 ± 71
113 ± 89
221 ± 193
108 ± 32
230 ± 65
122 ± 74
177 ± 45
4.0 ± 1.4a
3.3 ± 1.0
2.7 ± 1.8
3.8 ± 1.7
3.5 ± 1.6
1.7 ± 1.1
3.9 ± 1.1
1.7 ± 1.1
Song Tufts
2/4a
4/7
0/2
0/3
3/4
4/4
1/5
2/6
1/4a
2/7
0/2
0/3
1/4
2/4
0/5
0/6
Approach: proportion of individuals that approached within 8 m of the
speaker/mount. Duration: time spent within 8 m of the speaker/mount
(s, mean ± SE). Close dist: among only those individuals that made an
approach, the closest perching distance from the mount/speaker (m, mean
± SE). Song: proportion of approaching subjects that sang (territorial
advertisement and soft songs were not differentiated by observer during this
experiment, see Table 3). Tufts: proportion of approaching subjects that
erected their pectoral tufts.
aPost-approach data excluded for one Nectarinia moreaui CS song/CS mount
trial where a second male approached after the intended subject initially
approached.
Experiment 1 responses by treatment, Table 3: Experiments 2a
and 2b responses by treatment and population). As discussed
earlier, responses were highly individualistic. Most subjects did
not exhibit all or even most of these behavioral categories, such
that there was marked zero-inflation in the distributions of individual potential response variables for analysis. For this reason,
the duration of approach appeared to represent the best possible
response variable for statistical analysis. No individuals physically attacked the mount during trials used for analysis, although
this occurred once in a pilot trial. Outside of the physical attack
of a mount, the most intense behavioral responses to simulated
intrusion appeared to be soft song of 2 types: the first type of a
similar structure to the long-distance song form used in playback
stimuli and a second form observed in the field only during highintensity, close-proximity intraspecific interactions (with the only
exceptions to this species specificity being the use of the latter
form in interspecific contests between N. fuelleborni and N. moreaui
males in their extremely limited area of breeding-season sympatry, personal observation).
Comparative initial response probabilities to HS/
CS song are asymmetric between species
In the multimodal experiment (Experiment 1), N. moreaui males
approached within 8 m in 87% (n = 15) of trials with CS songs,
and in just 31% (n = 16) of trials with N. fuelleborni songs. Nectarinia
fuelleborni males, however, approached in 62.5% (n = 16) of CS song
trials and in 87% of trials (n = 15) with N. moreaui songs. In the
binary model of Experiment 1, with approach as the response variable, 95% credibility intervals for the difference in log odds ratio of
response to CS versus HS songs does not include 0 for N. moreaui
(consistent with species recognition; Figure 3—population Im) but
does so for N. fuelleborni (i.e., inconsistent with species recognition;
Figure 3—population Ik). Overall, 95% HPD intervals for contrasts
between species in relative probability of response to CS versus HS
do not include 0, indicating differences between the N. moreaui and
N. fuelleborni populations tested.
The species-level asymmetry in initial song response suggested by
the multimodal experiment was supported by the song-only experiment, though the asymmetry found in the second experiment is not
as strong. Including only those trials where subjects approached
during at least one song stimulus, N. moreaui responded during 90%
of CS song trials compared with 58% of N. fuelleborni song trials (n = 31); whereas N. fuelleborni responded during 80.5% of CS
song trials compared with 68% of N. moreaui song trials (Figure 4).
Contrasts between CS and HS stimuli for N. moreaui were significant (95% highest posterior density intervals for differences in log
odds ratios do not include 0), whereas contrasts between CS and
HS stimuli for N. fuelleborni were not significant despite the greater
sample size and increased power to detect differences for N. fuelleborni subjects (95% highest posterior density intervals for differences
in log odds ratios include 0; posterior probabilities were averaged
for each sampled iteration across levels of population to make
contrasts from the model for the song species treatment). Looking
across both experiments, 95% credibility intervals for the effect size
estimate corresponding to degree of species recognition (difference
in log odds ratio of response to CS versus HS songs) include 0 for
3 of the 4 N. fuelleborni populations, and do not include 0 for any
N. moreaui population (Figure 3). Thus, species-level asymmetry is
evident, despite geographic variation within N. fuelleborni in relative
response probabilities to CS versus N. moreaui songs.
McEntee • Territorial response of parapatric sunbirds
1387
Table 3
Experiment 2a (Nectarinia moreaui subjects) and Experiment 2b (Nectarinia fuelleborni): responses to songs
Species, population
moreaui
Mafwemiro
Ndundulu
fuelleborni
Nyumbanitu
Mufindi
Rungwe
Treatment
Duration
Close dist
CS
HS
Olivacea
CS
HS
Olivacea
131 ± 19
48 ± 13
18 ± 10
100 ± 18
65 ± 19
39 ± 14
2 ± 1
10 ± 3
14 ± 3
7 ± 2
12 ± 3
15 ± 3
CS
HS
Olivacea
CS
HS
Olivacea
CS
HS
Olivacea
89 ± 16
96 ± 20
12 ± 8
161 ± 26
67 ± 20
47 ± 20
94 ± 21
77 ± 19
47 ± 23
6 ± 3
9 ± 2
23 ± 3
6 ± 3
9 ± 3
13 ± 3
11 ± 3
12 ± 4
18 ± 4
Voc output
TA song
SF song
Call
Tufts
11 ± 3 (13)
15 ± 9 (13)
3 ± 1 (13)
13 ± 6 (14)
22 ± 6 (14)
9 ± 3 (14)
8/16
4/16
5/16
3/15
8/15
6/15
4/16
1/16
0/16
2/15
0/15
1/15
6/16
5/16
8/16
13/15
10/15
8/15
5/16
1/16
0/16
1/15
0/15
1/15
18 ± 8 (15)
23 ± 9 (15)
18 ± 7 (15)
16 ± 5 (12)
3 ± 1 (12)
26 ± 11 (12)
7 ± 4 (12)
10 ± 7 (12)
6 ± 4 (12)
8/15
5/15
4/15
2/13
0/13
4/13
5/13
4/13
3/13
0/15
3/15
0/15
2/13
0/13
0/13
4/13
1/13
0/13
11/15
10/15
5/15
5/13
9/13
8/13
7/13
8/13
9/13
5/15
2/15
1/15
3/13
2/13
0/13
5/13
0/13
1/13
Contact zone populations are in boldface, and entirely allopatric populations are not. Data are included for all subjects who approached within 8 m during one or
more stimulus response periods (see Methods). Duration: duration of approach within 8 m (s, mean ± SE). Close dist: closest perch distance from the speaker (m, mean
± SE). Voc output: summed duration of all vocalizations within response period (s, mean ± SE; n in parentheses). Note that sample sizes for this variable are reduced,
because noise prevented accurate estimates from some trial recordings. TA song: proportion of subjects exhibiting territorial advertisement song. SF song: proportion
of subjects exhibiting soft song directed towards the speaker. Call: proportion of subjects exhibiting calls. Tufts: proportion of subjects erecting pectoral tufts.
Figure 3
Effect size estimates for the natural log of the log odds of approach versus no approach to HS versus CS songs as estimated from Bayesian MCMC GLMMs.
Positive values indicate greater likelihood of responding to CS versus HS songs. Estimates are shown for both experiments (Experiment 1: multimodal
experiment; Experiments 2a and 2b: song-only experiment, effect size accounts for repeated-measures experimental design). Position 0 km along the x axis
represents the center of the parapatric boundary.
Response duration to multimodal signals
In the multimodal experiment, there was no support for the species recognition hypothesis based on the pattern of approach duration among treatments. Neither N. fuelleborni nor N. moreaui subjects
exhibited a pattern of approach duration among treatments, which
was consistent with the species recognition hypothesis (Figure 5).
Contrary to the prediction based on the species recognition hypothesis (CS song/CS mount > CS song/HS mount > HS song/CS
mount > HS song/HS mount), the modeled means to treatments,
pooled for both species (see Methods), were ordered as follows: CS
song/HS mount > HS song/HS mount > HS song/CS mount
> CS song/CS mount. None of the treatment differences were
significantly different in the pooled analysis (95% highest posterior density intervals for all treatment contrasts include 0). Thus,
though there is support for species recognition of songs in the initial response probability of N. moreaui subjects (see previous paragraph), there is no support for species recognition driving responses
of longer duration to CS signals in the multimodal experiment.
Moreover, though sample sizes did not provide for strong statistical
power for this test, the trends in the data do not suggest that a result
which is consistent with species recognition would emerge with
increased statistical power. Instead, there is a trend in both species
for increased duration of response to HS mounts, which may be
more pronounced in N. fuelleborni subjects.
1388
Behavioral Ecology
This result indicates that there is substantial conservatism of the song
response function in both species, despite strong song divergence and
the significantly different probabilities of response to HS and CS
songs exhibited by N. moreaui (Figures 3 and 4). Contrasts from the
MCMC posterior probability distribution were significant for both
species for HS versus control comparisons (Figure 4; 95% highest
posterior density intervals for differences in the log of response probability did not include 0; posterior probabilities were averaged across
the population fixed effect to make contrasts from the model for this
factor). There is no support for the hypothesis that N. fuelleborni exhibits higher aggression toward HS, generally, than N. moreaui.
Geographic variation in response within species
Figure 4
Percentage of trials where subjects responded by approaching within 8
m to each of the 3 stimulus types in Experiments 2a and 2b (song-only):
CS = conspecific song; HS = heterospecific (either Nectarinia moreaui or
Nectarinia fuelleborni) song; and Control = Nectarinia olivacea song. The
letters above bars indicate statistical significance: Each letter represents a
statistical grouping based on contrasts performed after Bayesian generalized
linear mixed modeling using “approach” or “no-approach” as a binomial
response. The experiment is a repeated-measures experiment, and the
modeling framework takes into account within-individual variation and
variation due to each of the 6 song tracks used (see Methods).
Figure 5
Means ± SE for approach duration within 8 m for Experiment 1. The
dashed line represents the predicted pattern of responses under the a priori
hypothesis that each species would show the strongest responses to CS
signals, assuming that species differences in songs have a greater effect on
receivers than species differences in morphology/plumage.
A sympatric competitor’s song elicits fewer
responses than the parapatric HS
Despite the lack of interaction between N. moreaui and N. fuelleborni for
all but a tiny fraction of individuals within each species, both species
exhibit response probabilities at least twice as great to the parapatric sibling taxon as the ubiquitous ecological competitor N. olivacea.
In Experiments 2a and 2b (song-only), strong geographic variation
is found in relative CS/HS responses within N. fuelleborni, and less
geographic variation is found within N. moreaui. Nectarinia fuelleborni
males at Mufindi were more likely to respond to CS versus HS songs
(P < 0.05, contrasts between CS and HS treatments within populations computed from MCMC posterior probability distribution,
Figure 3: Mu), unlike subjects at Rungwe and Nyumbanitu whose
response probabilities to CS and HS songs were similar (Figure 3: Ny
and Ru). Within N. fuelleborni, between-population contrasts in relative
probabilities of response to HS versus CS (as calculated by log odds)
included 0. Nectarinia fuelleborni males at Mufindi spent longer responding to CS song than to HS song (P < 0.05, within-population contrasts as in binary model), unlike males at Rungwe and Nyumbanitu
(Figure 6). For CS − HS response duration differences between-population contrasts indicate statistically significant differences between
Mufindi and Rungwe, and between Mufindi and Nyumbanitu.
CS−HS response duration differences were similar at Rungwe and
Nyumbanitu. The geographic pattern of variation within N. fuelleborni does not closely fit the hypothesis that higher levels of aggression toward N. moreaui signals result from contact between species, as
Rungwe is distant from any N. moreaui populations but shows responses
of a similar probability and duration to N. moreaui and CS song; that
is, it is a naive population that shows strong responses to N. moreaui.
The results, therefore, fit better with the hypothesis that N. fuelleborni
exhibits less discrimination of CS versus N. moreaui song, but exhibits
stochastic geographic variation (see Discussion).
Nectarinia moreaui responses in the song-only experiment
(Experiment 2a) were not statistically different between populations
(P > 0.05 for contrasts among populations for probability and duration difference models). However, there is a trend in the duration
differences, in that N. moreaui exhibits a slightly greater response
duration to N. fuelleborni at Ndundulu (a contact zone population)
than at Mafwemiro (an allopatric population, Figure 6). Indeed,
the 95% credibility interval for duration differences in response
to CS and HS songs at Ndundulu includes 0, unlike Mafwemiro
where species recognition predicts response differences. This result
suggests that N. moreaui males within Mafwemiro more probably
respond to CS song, but those that respond to N. fuelleborni song
exhibit strong responses. This pattern suggests that longer duration
responses could result from parapatric contact within N. fuelleborni,
consistent with competitor recognition increasing HS aggression.
Discussion
Intraspecific geographic variation in HS responses
This study may provide the first evidence of strong populationlevel variation in territorial response to an HS taxon among
McEntee • Territorial response of parapatric sunbirds
1389
Figure 6
Population means and 95% credibility intervals of individual differences in responses to CS versus HS (CS − HS). Positive values reflect responses which are
consistent with species recognition (greater response to CS), whereas 0 reflects equivalent responses to CS and HS. Position 0 km represents the position of
the parapatric boundary center. Statistical significance (*) was determined at an α-level of 0.05 from MCMC posterior probability distributions of contrasts
(see Methods).
populations distant (allopatric) from that HS. Lack of evidence for
this kind of variation elsewhere may be a consequence of the rarity of experiments with the potential to find it. In reviewing the literature, I found just 4 studies that have presented variation among
populations that are entirely allopatric with the focal HS (Prescott
1987; Pearson and Rohwer 2000; Peiman and Robinson 2007;
Anderson and Grether 2010). Such variation can be assessed in
an additional case where multiple studies have been performed in
the same taxa (Vermivora warblers: Gill and Murray 1972; Crook
1984). None of these studies found strong variation to HS among
allopatric populations of a focal taxon. In the test with the most
power to detect differences, Anderson and Grether (2010) manipulated CS to appear more similar to HS in presentations to multiple allopatric populations of 2 Hetaerina damselfly species, and
in this case it is unclear whether variation in HS response among
such allopatric populations occurs independently of variation in
the overall intensity of response to intruders. Although it would
be optimal to replicate many allopatric populations in future studies, including just a second entirely allopatric N. fuelleborni population has a strong impact on inferences regarding the evolution of
territorial responses in this study. In N. fuelleborni, equally strong
responses to CS and N. moreaui songs are not limited to the parapatric boundary area, and instead are also found distant from the
contact. Had only a single allopatric population been examined,
either increased responses to HS via competitor recognition in the
parapatric zone or universally strong responses to N. moreaui songs
by N. fuelleborni would have been inferred. Instead, a more nuanced
perspective is gained that is indicative of the complex processes
involved in the decay of HS response during allopatric or parapatric speciation.
For populations allopatric with an HS, the generation of variation in response to HS must occur as a byproduct of evolution
of another trait, as a result of stochastic processes (cultural drift,
genetic drift, or genetic draft) or gene flow from populations
where direct selection drives responses. There is no possibility
that the N. fuelleborni populations distant from contact can learn
N. moreaui phenotypes (Grether 2011), and direct selection on
territorial response to N. moreaui cannot occur locally. In N. fuelleborni, increased relative response to HS distant from sympatry as a
result of gene flow is both unlikely because sympatry is so limited
in these species, and, furthermore, refuted by evidence in this study
by the spatial pattern of responses. The relatively weak response to
N. moreaui by N. fuelleborni at Mufindi appears to reflect a more narrow CS response function in that population. A potential explanation lies in the relative abundance of a bird species with a strikingly
similar song to N. fuelleborni, the Yellow-browed Seed-eater Serinus
whytii, which was observed syntopically with N. fuelleborni at Mufindi
but not in the other tested N. fuelleborni populations (e.g., at Ikokoto,
Werema et al. 2013) during these experiments. The 2 species’ songs
are similar enough that one S. whytii individual approached the
speaker during playback of N. fuelleborni’s song during the experiment. However, because S. whytii has similar songs but little niche
overlap with N. fuelleborni, N. fuelleborni males should generally avoid
responding to S. whytii song (and vice versa) where the 2 species
are sympatric. No interspecific aggression was otherwise observed
between N. fuelleborni and S. whytii over a total of 12 days at Mufindi
in 2008 and 2010. N. fuelleborni males may avoid responding to
S. whytii song by generating specific perceptual distinctions between
the 2 species’ songs, which, in turn, cause CS song features to be
weighted differently during CS perception (Nelson 1989). Reduced
responses to N. moreaui at Mufindi could be a by-product of such
a process. This hypothesis could be explored in future territorial
intrusion experiments.
With respect to the evolution of HS response during speciation,
it could be contended that because allopatric populations are not
generally subject to direct selection on their HS responses, geographic variation in HS response is free to accumulate. This variation might often be generated in a pattern that appears stochastic.
Sympatric populations, however, may be comparatively constrained phenotypically because there are likely to be direct fitness
consequences of response. A few studies of territorial response
to HS have used spatial replication of allopatric populations, so
evidence for similarly stochastic-appearing spatial variation awaits
future studies.
1390
Species-level asymmetry in HS response
Asymmetry in reciprocal response to divergent signals has been
discovered across a broad diversity of organisms (e.g., Kaneshiro
1976; Barth and Schmitt 1991; Svensson et al. 2006), and especially birds (Robinson and Terborgh 1995; Colbeck et al. 2010;
Dingle et al. 2010). Here, evidence is provided for asymmetric
discrimination probability in recently diverged, parapatrically
distributed sunbird species with strong divergence in songs. What
explains such asymmetry in this case? An important first consideration is the degree of divergence in signal evolution versus
perception. Although song divergence between these 2 species
is measurably great (e.g., >3-fold difference in mean duration of
pauses between song elements, McEntee JP, Penalba JV, Werema
C, Mulungu E, Mbilinyi M, Bowie RCK, in preparation), relatively high probabilities of response to HS occur in both species
(Figure 4). Even in distant allopatric populations where contact
with the sibling species HS is not possible, response probabilities
to the closely related allopatric HS are far greater than to the sympatric competitor N. olivacea. This suggests that divergent signals
are treated as CS on the outer edge of the distribution of CS signal space by a substantial proportion of subjects of both species
(Phelps et al. 2006). This interpretation is further supported by evidence of hybridization at the parapatric boundary (McEntee JP,
Penalba JV, Werema C, Mulungu E, Mbilinyi M, Bowie RCK, in
preparation): Individuals sometimes treat the sibling taxon as CS
in mating interactions, and such recognition errors should be more
common in territorial interactions where broader CS recognition
functions are expected (Seddon and Tobias 2010). Thus, though
species-level designations are used for N. fuelleborni and N. moreaui,
the 2 species can and do regularly perceive each other as CS. This
result contrasts with previous studies where aggressive responses
are inferred to result purely from HS competitor recognition without the possibility of recognition errors (Robinson and Terborgh
1995; Grether et al. 2009; Matyjasiak 2005; Jankowski et al. 2010).
Furthermore, though recognition errors are more commonly
exhibited by N. fuelleborni in the dataset, N. moreaui individuals are
also error prone, unlike in other studies where HS responses are
rare or very weak for one lineage (e.g., Gray-breasted Wood-wrens
Henicorhina leucophrys, Dingle et al. 2010).
One explanation for the observed territorial response asymmetry between 2 species (or divergent populations) is disparity in
aggression; for example, in this system that N. fuelleborni is more
aggressive than N. moreaui at a species level. In a series of studies, Rohwer and colleagues (Pearson and Rohwer 2000; Krosby
and Rohwer 2009, 2010) have documented aggression disparity between Townsend’s Warblers Setophaga townsendi and Hermit
Warblers Setophaga occidentalis in their hybrid zone, and further
have provided evidence that aggression disparity has caused this
zone to move across space as Townsend’s Warblers replace Hermit
Warblers. Here, I refute a general aggression disparity between
N. moreaui and N. fuelleborni with 2 lines of evidence. First, responses
to the ubiquitously sympatric competitor N. olivacea’s song occurred
with a similar probability in both N. moreaui (29%) and N. fuelleborni (22%), suggesting similar instead of disparate levels of aggression toward a shared competitor. Moreover, for greater aggression
to account for N. fuelleborni’s greater HS response, N. fuelleborni
might be expected to exhibit greater CS aggression (Peiman and
Robinson 2010). In both studies presented here, there is a trend
for N. moreaui to approach CS song with a higher probability than
N. fuelleborni, contrary to the prediction from the aggression disparity hypothesis.
Behavioral Ecology
An alternate explanation for response asymmetry is based on
perceptual asymmetries which result from divergent coevolution of signals and signal response functions. Dingle et al. (2010)
suggest that territorial response asymmetry between parapatric
subspecies of H. leucophrys could be attributed to frequency bandwidth differences between subspecies. These authors propose
that responses are more aggressive to signals that have a broader
frequency range when one taxon’s phenotype’s frequency range
envelops another’s. In their study, subspecies H. hilaris’s song
occupies a narrower frequency bandwidth than H. leucophrys,
and H. hilaris responds aggressively to the heterotypic H. leucophrys, whereas the converse is not true. I propose that a perceptual
asymmetry model fits the data in this study, but the combined
pattern of song frequency overlap and responses is opposite that
found in Dingle et al. 2010. In this study, N. fuelleborni uses a
broader range of frequencies than N. moreaui, with the frequency
band employed by N. moreaui subsumed within that used by N. fuelleborni (Figure 1). I hypothesize that N. moreaui’s more narrowly
circumscribed song phenotype corresponds with a concomitantly
narrow territorial response function. The frequency range of
N. fuelleborni’s song falls outside of N. moreaui’s range, and, therefore, may not be interpreted as CS as frequently as the reverse.
Meanwhile, N. fuelleborni’s songs span broadly across the frequency
spectrum, whereas N. moreaui’s songs fall entirely within the peak
frequency boundaries of N. fuelleborni’s song, such that N. moreaui’s
song could be perceived to be within the range of N. fuelleborni’s
song limits. The pattern of greater variation within N. fuelleborni’s
song as compared with N. moreaui’s song is not restricted to the frequency spectrum: More generally, N. fuelleborni’s songs appear to
have greater spectro-temporal variation. The individual elements
which comprise N. fuelleborni’s songs vary greatly in terms of spectro-temporal properties, including even the incorporation of elements of HS songs through mimicry (e.g., of Cinnamon Bracken
Warbler Bradypterus cinnamomeus, unpublished data). The elements
of N. moreaui’s song are more narrowly circumscribed by comparison. N. fuelleborni’s songs can be truncated at times, so that they are
within the range of N. moreaui’s songs; whereas N. moreaui’s songs
rarely reach the mean duration of N. fuelleborni’s songs. I hypothesize that the territorial song response function of these species
tracks their mutual song divergence such that N. moreaui’s song
response function permits less variation (i.e., it is more narrow)
than that of N. fuelleborni, and correspondingly, N. moreaui less frequently commits recognition errors when presented with N. fuelleborni’s song than the reverse.
Eco-evolutionary consequences of male–male behavioral
interactions
The circumstances of N. moreaui–N. fuelleborni parapatry suggest
that their mutual range boundary could be moving in space. The
boundary is not associated with an apparent ecotone that could
stabilize the boundary’s geographical position, which suggests
that asymmetric interspecific interactions could shift the boundary
(Barton and Hewitt 1985). If competitive asymmetry exists between
species, one species will move along the spine of the Udzungwa
Mountains to the detriment of the other. Aggressive asymmetries
in sibling or sister species are known from a wide swath of diversity,
and contribute to the geographic movement of the Setophaga warbler hybrid zone (Pearson and Rohwer 2000; Krosby and Rohwer
2009, 2010) and the local replacement of Sialia mexicana by Sialia
currucoides in western North America (Duckworth and Badyaev
2007). Could the asymmetry in HS song response found in this
McEntee • Territorial response of parapatric sunbirds
experiment contribute to interference competition asymmetries
between species that might then influence the movement of the
species boundary?
The multimodal experiment in this study gives the most comprehensive indication of how territorial interactions between these species are likely to play out (Hebets and Papaj 2005). The initially low
probability of HS song response by N. moreaui subjects gives way to
intense aggressive responses to all combinations of song and mount
species. The especially strong responses by N. moreaui to N. fuelleborni’s
song/N. fuelleborni mount combinations could indicate that low levels
of HS song response would not translate to a failure of N. moreaui to
respond aggressively to N. fuelleborni intruders. Given the prevalence
of associative learning of competitors in birds (Rohwer 1973; Martin
and Martin 2001), it is unlikely that territorial N. moreaui males that
exhibit an initial failure to respond to N. fuelleborni song would continue to outright ignore singing N. fuelleborni males after encountering
them within their territories. If associative learning of N. fuelleborni
songs by N. moreaui is efficient enough, the disparity in initial probability of HS response between species may not result in a competitive
asymmetry. However, the disparity between response probabilities
to HS and CS songs in N. moreaui (Figures 3 and 4) suggests that a
male N. moreaui competing for territory simultaneously with both
HS and CS males is more likely to be compelled to respond to CS
song rather than to HS song, and, therefore, may exert more total
effort and/or time interacting with its CS rather than with its HS
competitors. There is no indication that such a bias would occur in
N. fuelleborni males, in which both the probability and duration of HS
responses are similar to CS responses. Hence, the results presented
in this study suggest that N. fuelleborni males may more easily occupy
parts of N. moreaui territories than the converse, which could facilitate
the colonization of N. moreaui-dominated areas by N. fuelleborni if associative learning does not equalize the responses.
Conclusions
Though the number of studies is not great, reciprocal signal
response experiments in divergent groups tend to reveal some
degree of asymmetry between taxa or populations. The significance of this asymmetry has not yet become thoroughly understood, but there are likely to be both ecological and evolutionary
consequences to this neuro-ecological phenomenon (Grether 2011),
especially during secondary contact. A lesson from this study is that
spatial replication could be critical to making inferences regarding
the evolution of signal response from geographic variation. I suggest that stochastic or stochastic-appearing variation (from cryptic
but deterministic processes) is especially likely among populations
without direct selection of the trait of interest: Attempting to estimate the level of geographic variation from such processes may
give new insights into the nature of evolution of signal responses.
Finally, though the topic was little addressed within this study,
simulated territorial intrusion experiments are frequently used to
predict whether there is a basis for discrimination that would provide a path to pre-zygotic reproductive isolation in divergent groups.
In this experiment, apparent species recognition by N. moreaui might
suggest that pre-zygotic reproductive isolation is likely. However,
parapatry provides the natural experiment to test this hypothesis,
and hybridization involving female N. moreaui is evident (individuals
of mixed ancestry have N. moreaui mtDNA haplotypes, McEntee JP,
Penalba JV, Werema C, Mulungu E, Mbilinyi M, Bowie RCK, in
preparation). As evident in other studies (e.g., Seddon and Tobias
2010), simulated territorial intrusion experiments on male territory
1391
holders in birds may not yield much information regarding the
likelihood of pre-zygotic reproductive isolation. Though inferences
from such experiments are generally cautious with regard to what
might actually happen in secondary contact, I suggest that territorial response is a rich topic in and of itself, and that inferences from
simulated territorial intrusion experiments might best be restricted
to this aspect of organismal biology.
Supplementary Material
Supplementary material can be found at http://www.beheco.
oxfordjournals.org/
Funding
Funding for field efforts was generously provided by Charles
Koford and Louise Kellogg Grants from the Museum of Vertebrate
Zoology (UC Berkeley), a Beim Summer Research Award from the
Department of Integrative Biology (UC Berkeley), a UC Berkeley
Center for African Studies Andrew and Mary Thompson Rocca
Scholarship, a National Geographic Society/Waitt Foundation
Grant, the American Ornithologists’ Union, two Exploration Fund
grants from The Explorers’ Club, and National Science Foundation
Doctoral Dissertation Improvement Grant Award #1011702.
Support for writing was provided by a Joseph Maillard Fellowship
from the Museum of Vertebrate Zoology.
E. Mulungu and M. Mbilinyi provided exceptional research assistance in all
field research components of this study. C. Werema (University of Dar-essalaam) contributed toward sound recording and participated in pilot experiments. Field logistical support and site recommendations were provided by
N. and L. Baker, D. Moyer, N. Cordeiro, S. Sheridan-Johnson, and Unilever
Tea Tanzania Ltd. K. Rudolph provided feedback on experimental design
and manuscript preparation. The R.C.K. Bowie lab, “Endler Reading
Group”, and Chelsea Specht (UC Berkeley), and the Kimball-Braun and
S.K. Robinson labs (University of Florida) provided valuable discussion
regarding the interpretation of results. This research was performed as a
part of J.P.M.’s dissertation research at UC Berkeley, which was improved
throughout by a dissertation committee comprising R. Bowie, C. Moritz,
and F. Theunissen. The manuscript was improved by comments from T. Day
and 2 anonymous reviewers. Permissions for field research were furnished by
the Tanzania Wildlife Research Institute (TAWIRI), Commission for Science
and Technology (COSTECH, Tanzania), Forestry and Beekeeping Division
(Tanzania), and the village government of Ikokoto, Tanzania.
Handling editor: Marc Thery
Appendix A Bayesian GLMM modeling
details
Code used to run models in R (package MCMCglmm, Hadfield
2010):
2009 binary models:
With mount species as a fixed effect:
priorR<-list(R=list(V=1,fix=1),G=list(G1=list(V=diag(2),
nu=1,
alpha.mu=c(0,0), alpha.V=100*diag(2)), G2=list(V=diag(2), nu=1,
alpha.mu=c(0,0), alpha.V=100*diag(2))))
bin2009_withmount<-MCMCglmm(respond ~ song*mount*subject.
sp.-1, random = ~idh(subject.sp.):song_ID + idh(subject.
sp.):Spec_ID, data=data, nitt=550000, thin=10, burnin=50000,
verbose=FALSE, prior=priorR, family=“categorical”, DIC=TRUE,
slice=TRUE, saveX=TRUE)
Without mount species as a fixed effect:
priorR<-list(R=list(V=1,fix=1),G=list(G1=list(V=1, nu=1, alpha.
mu=0,alpha.V=100)))
Behavioral Ecology
1392
Table A1
Bayesian GLMM
Year
Subject species
Model family
Fixed effects
Random effects
DIC*
2009
moreaui, fuelleborni
Categorical (binary)
Playback track
82.2
2010
moreaui
Categorical (binary)
2010
fuelleborni
Categorical (binary)
2009
moreaui, fuelleborni
Poisson
2010
2010
moreaui
fuelleborni
Gaussian
Gaussian
Song species, subject species, song
species × subject species
Treatment, population, |
treatment × population
Treatment, population, treatment ×
population
Mount species, song species, mount
species × song species
Population
Population
Playback track, territory
100.9
Playback track, territory
141.3
Specimen ID, playback track
320.5
Playback track
Playback track
365
491.5
*DIC = Deviance Information Criterion.
bin2009_nomount
<-MCMCglmm(respond~song*subject.
sp.-1,random=~song_ID,
data=data,
nitt=130000,
thin=10,
burnin=30000,
verbose=FALSE,
prior=priorR,
family=“categorical”, DIC=TRUE,slice=TRUE,saveX=TRUE)
2010 N. fuelleborni binary:
bin2010prior_fuel< list(B=list(mu=c(rep(0,9), V=diag(9)*(1+pi^2/3)),
R=list(V=diag(1),fix=1),
G=list(G1=list(V=1,nu=1,alpha.mu=0,
alpha.V=100), G2 = list(V=1,nu=1,alpha.mu=0,alpha.v=100)))
bin2010_fuel<-MCMCglmm(respond
~
stim_class*pop-1,
random=~stim_id + territory, data=fuel, prior=bin2010prior_
fuel, nitt=550000, thin=10, burnin=50000, verbose=FALSE,
family=“categorical”,
DIC=TRUE,
slice=TRUE,
saveX=TRUE,pr=TRUE)
2010 N. moreaui binary:
bin2010prior_more<-list(B=list(mu=c(rep(0,6),
V=diag(6)*(1+pi^2/3)),
R=list(V=diag(1),fix=1),
G = l i s t ( G 1 = l i s t ( V = 1 , n u = 1 , a l p h a . mu = 0 , a l p h a . V = 1 0 0 ) ,
G2 = list(V=1,nu=1,alpha.mu=0,alpha.v=100)))
bin2010_popfixedmore<-MCMCglmm(respond
~
stim_
class*pop-1, random=~stim_id + territory, data=more,
prior=bin2010prior_ more, nitt=130000, thin=10, burnin=30000,
verbose=FALSE, family=“categorical”, DIC=TRUE, slice=TRUE,
saveX=TRUE,pr=TRUE)
2009 Poisson:
prior_4stims <- list(R=list(V=1,nu=0.002), G = list(G1=list(V=dia
g(2),nu=0.002), G2=list(V=diag(2),nu=0.002)))
poisson_2009<-MCMCglmm(Dur8m7min
~
mount*song,
random=~idh(subject.sp.):Spec_ID
+
idh(subject.sp.):song_
ID, data=data, prior=prior_4stims, nitt=140000, thin=10,
burnin=40000, verbose=FALSE, family = “poisson”, DIC=TRUE,
saveX=TRUE, pl=TRUE, pr=TRUE)
2010 N. fuelleborni, Gaussian:
prior <- list(G = list(G1 = list(V = 1, nu = 0.002)), R = list(V = 1,
nu = 0.002))
MCMCglmm(csminushs ~ pop - 1, random=~stim_id,
data=subset(data2,subj_sp==“f ”),
nitt=550000,thin=10,bur
nin=50000,verbose=FALSE,family=“gaussian”,DIC=TRUE,
saveX=TRUE, prior = prior)
2010 N. moreaui, Gaussian:
prior <- list(G = list(G1 = list(V = 1, nu = 0.002)), R = list(V = 1,
nu = 0.002))
gauss_m<-MCMCglmm(csminushs ~ pop-1, random=~stim_id,
data=m, nitt=550000, thin = 10, burnin=50000, verbose=FALSE,
family = “gaussian”, DIC=TRUE, saveX=TRUE)
References
Amezquita A, Hoedl W, Lima AP, Castellanos L, Erdtmann L, De Araujo
MC. 2006. Masking interference and the evolution of the acoustic communication system in the Amazonian dendrobatid frog Allobates femoralis.
Evolution. 60:1874–1887.
Anderson CN, Grether GF. 2010. Character displacement in the fighting
colours of Hetaerina damselflies. Proc R Soc Lond B. 277:3669–3675.
Baker MC. 1991. Response of male Indigo and Lazuli Buntings and
their hybrids to song playback in allopatric and sympatric populations.
Behaviour. 119:225–242.
Barth FG, Schmitt A. 1991. Species recognition and species isolation in
Wandering Spiders (Cupiennius Spp, Ctenidae). Behav Ecol Sociobiol.
29:333–339.
Barton NH, Hewitt GM. 1985. Analysis of hybrid zones. Annu Rev Ecol
Syst. 16:113–148.
Bioacoustics Research Program. 2008. Raven Pro: interactive sound analysis software (version 1.3). Ithaca (NY): Cornell Lab of Ornithology.
Bowie RCK, Fjeldsa J, Hackett SJ, Crowe TM. 2004. Systematics and biogeography of double-collared sunbirds from the Eastern Arc Mountains,
Tanzania. Auk. 121:660–681.
Brenowitz E, Beecher M. 2005. Song learning in birds: diversity and plasticity, opportunities and challenges. Trends Neurosci. 28:127–132.
Brown WL, Wilson EO. 1956. Character displacement. Syst Zool. 5:49–64.
Bull C. 1991. Ecology of parapatric distributions. Annu Rev Ecol Syst.
22:19–36.
Catchpole CK. 1978. Interspecific territoriality and competition in
Acrocephalus warblers as revealed by playback experiments in areas of
sympatry and allopatry. Anim Behav. 26:1072–1080.
Catchpole CK, Leisler B. 1986. Interspecific territorialism in Reed Warblers
Acrocephalus scirpaceus: a local effect revealed by playback experiments.
Anim Behav. 34:299–300.
Catchpole CK, Slater PJB. 2008. Bird song. Cambridge (UK): Cambridge
University Press.
Cody ML. 1969. Convergent characteristics in sympatric species—a possible
relation to interspecific competition and aggression. Condor. 71:223–239.
Colbeck GJ, Sillett TS, Webster MS. 2010. Asymmetric discrimination of
geographical variation in song in a migratory passerine. Anim Behav.
80:311–318.
Crook JR. 1984. Song variation and species discrimination in Blue-Winged
Warblers. Wilson Bull. 96:91–99.
Damien P, Wakefield J, Walker S. 1999. Gibbs sampling for Bayesian nonconjugate and hierarchical models by using auxiliary variables. J R Stat
Soc B. 61:331–344.
Dingle C, Poelstra JW, Halfwerk W, Brinkhuizen DM, Slabbekoorn H.
2010. Asymmetric response patterns to subspecies-specific song differences in allopatry and parapatry in the Gray-Breasted Wood-Wren.
Evolution. 64:3537–3548.
Duckworth RA, Badyaev AV. 2007. Coupling of dispersal and aggression
facilitates the rapid range expansion of a passerine bird. Proc Natl Acad
Sci USA. 104:15017–15022.
van Dyk D, Meng X. 2001. The art of data augmentation. J Comput
Graph Stat. 10:1–50.
McEntee • Territorial response of parapatric sunbirds
Emlen ST. 1972. Experimental analysis of parameters of bird song eliciting
species recognition. Behaviour. 41:130–171.
Fjeldså J, Kiure J, Doggart N, Hansen LA, Perkin A. 2010. Distribution of
highland forest birds across a potential dispersal barrier in the Eastern
Arc Mountains of Tanzania. Steenstrupia. 32:1–43.
Gelman A. 2006. Prior distributions for variance parameters in hierarchical models (Comment on an Article by Browne and Draper). Bayesian Anal. 1:515–533.
Gil D. 1997. Increased response of the short-toed treecreeper Certhia
brachydactyla in sympatry to the playback of the song of the common
treecreeper C. familiaris. Ethology. 103:632–641.
Gill FB, Murray BG Jr. 1972. Discrimination behavior and hybridization of the Blue-winged and Golden-winged Warblers. Evolution.
26:283–293.
Grant BR, Grant PR. 2002. Simulating secondary contact in allopatric
speciation: an empirical test of premating isolation. Biol J Linn Soc.
76:545–556.
Grether GF. 2011. The neuroecology of competitor recognition. Integr
Comp Biol. 51:807–818.
Grether GF, Losin N, Anderson CN, Okamoto K. 2009. The role of interspecific interference competition in character displacement and the evolution of competitor recognition. Biol Rev. 84:617–635.
Groening J, Hochkirch A. 2008. Reproductive interference between animal
species. Q Rev Biol. 83:257–282.
Hadfield JD. 2010. MCMC methods for multi-response generalized linear
mixed models: the MCMCglmm R package. J Stat Softw. 33(2):1–22.
Hebets E, Papaj D. 2005. Complex signal function: developing a framework
of testable hypotheses. Behav Ecol Sociobiol. 57:197–214.
Hinsch M, Komdeur J. 2010. Defence, intrusion and the evolutionary stability of territoriality. J Theor Biol. 266:606–613.
Hof D, Hazlett N. 2010. Low-amplitude song predicts attack in a North
American wood warbler. Anim Behav. 80:821–828.
Irwin DE, Price T. 1999. Sexual imprinting, learning and speciation.
Heredity. 82:347–354.
Jankowski JE, Robinson SK, Levey DJ. 2010. Squeezed at the top: interspecific aggression may constrain elevational ranges in tropical birds.
Ecology. 91:1877–1884.
Kaneshiro K. 1976. Ethological isolation and phylogeny in Planitibia subgroup of Hawaiian Drosophila. Evolution. 30:740–745.
Kroodsma DE. 1989. Suggested experimental designs for song playbacks.
Anim Behav. 37:600–609.
Kroodsma DE, Byers BE, Goodale E, Johnson S, Liu WC. 2001.
Pseudoreplication in playback experiments, revisited a decade later. Anim
Behav. 61:1029–1033.
Krosby M, Rohwer S. 2009. A 2000 km genetic wake yields evidence for
northern glacial refugia and hybrid zone movement in a pair of songbirds. Proc R Soc Lond B. 276:615–621.
Krosby M, Rohwer S. 2010. Ongoing movement of the Hermit Warbler X
Townsend’s Warbler hybrid zone. PLoS ONE. 5:e14164.
Lack D. 1944. Ecological aspects of species-formation in passerine birds.
Ibis. 86:260–286.
Lovett JC. 1993. Climatic history and forest distribution in eastern Africa.
In: Lovett JC, Wasser SK, editors. Biogeography and ecology of the rain
forests of Eastern Africa. Cambridge (UK): Cambridge University Press.
Lovett JC, Clarke GP, Moore R, Morrey GH. 2001. Elevational distribution of restricted range forest tree taxa in eastern Tanzania. Biodivers
Conserv. 10:541–550.
Martin P, Martin T. 2001. Behavioral interactions between coexisting species:
song playback experiments with wood warblers. Ecology. 82:207–218.
Matyjasiak P. 2005. Birds associate species-specific acoustic and visual cues: recognition of heterospecific rivals by male blackcaps. Behav Ecol. 16:467–471.
Mayr E. 1942. Systematics and the origin of species. New York: Columbia
University Press.
Mayr E. 1963. Animal species and evolution. Cambridge (MA): Belknap
Press of Harvard University Press.
Morrison ML. 1982. The structure of western warbler assemblages:
ecomorphological analysis of the Black-Throated Gray and Hermit
Warblers. Auk. 99:503–513.
Murray B. 1971. Ecological consequences of interspecific territorial behavior in birds. Ecology. 52:414–423.
Murray B. 1976. Critique of interspecific territoriality and character convergence. Condor. 78:518–525.
Murray B. 1981. The origins of adaptive interspecific territorialism. Biol
Rev Camb Philos Soc. 56:1–22.
Nelson DA. 1989. The importance of invariant and distinctive features in
species recognition of bird song. Condor. 91:120–130.
1393
Newman MM, Yeh PJ, Price TD. 2006. Reduced territorial responses in
dark-eyed juncos following population establishment in a climatically
mild environment. Anim Behav. 71:893–899.
Nice MM. 1941. The role of territory in bird life. Am Midland Nat.
26:441–487.
Nishikawa KC. 1987. Interspecific aggressive behavior in salamanders species: specific interference or misidentification? Anim Behav.
35:263–270.
Noble GK. 1939. The role of dominance in the social life of birds. Auk. 56
(3):263–273.
Nuechterlein GL, Buitron D. 1998. Interspecific mate choice by late-courting male western grebes. Behav Ecol. 9:313–321.
Ord TJ, King L, Young AR. 2011. Contrasting theory with the empirical
data of species recognition. Evolution. 65:2572–2591.
Ord TJ, Stamps JA. 2009. Species identity cues in animal communication.
Am Nat. 174:585–593.
Orians G, Willson M. 1964. Interspecific territories of birds. Ecology.
45:736–745.
Pasch B, Bolker BM, Phelps SM. 2013. Interspecific dominance via vocal
interactions mediates altitudinal zonation in Neotropical singing mice.
Am Nat. 182:E161–E173.
Patton JL. 1993. Hybridization and hybrid zones in pocket gophers
(Rodentia, Geomyidae). In: Harrison RG, editor. Hybrid zones and the
evolutionary process. Oxford: Oxford University Press. p. 290–308.
Pearson SF, Rohwer S. 2000. Asymmetries in male aggression across an
avian hybrid zone. Behav Ecol. 11:93–101.
Peiman KS, Robinson BW. 2007. Heterospecific aggression and adaptive divergence in brook stickleback (Culaea inconstans). Evolution.
61:1327–1338.
Peiman KS, Robinson BW. 2010. Ecology and evolution of resource-related
heterospecific aggression. Q Rev Biol. 85:133–158.
Phelps S, Rand A, Ryan M. 2006. A cognitive framework for mate choice
and species recognition. Am Nat. 167:28–42.
Prescott DRC. 1987. Territorial responses to song playback in allopatric and
sympatric populations of Alder Empidonax alnorum and Willow Empidonax
traillii Flycatchers. Wilson Bull. 99:611–619.
Price T. 1998. Sexual selection and natural selection in bird speciation. Phil
Trans R Soc Lond. 353:251–260.
Price T. 2008. Speciation in birds. Greenwood Village (CO): Roberts and
Company.
R Core Team. 2012. R: a language and environment for statistical computing. Vienna (Austria): R Foundation for Statistical Computing.
Reed TM. 1982. Interspecific territoriality in the Chaffinch Fringilla
coelebs and Great Tit Parus major on islands and the mainland of
Scotland, UK: Playback and removal experiments. Anim Behav.
30:171–181.
Robinson S, Terborgh J. 1995. Interspecific aggression and habitat selection
by Amazonian birds. J Anim Ecol. 64:1–11.
Rohwer SA. 1973. Significance of sympatry to behavior and evolution of
Great Plains Meadowlarks. Evolution. 27:44–57.
Ryan MJ, Rand AS. 1993. Species recognition and sexual selection as a unitary problem in animal communication. Evolution. 47:647–657.
Sabo JL, Pauley GB. 1997. Competition between stream-dwelling cutthroat trout (Oncorhynchus clarki) and coho salmon (Oncorhynchus kisutch):
effects of relative size and population origin. Can J Fish Aquat Sci.
54:2609–2617.
Searcy WA, Nowicki S. 2006. Signal interception and the use of soft song in
aggressive interactions. Ethology. 112:865–872.
Seddon N, Tobias JA. 2010. Character displacement from the receiver’s
perspective: species and mate recognition despite convergent signals in
suboscine birds. Proc R S Lond B. 277:2475–2483.
Sedlacek O, Cikanova B, Fuchs R. 2006. Heterospecific rival recognition in
the Black Redstart (Phoenicurus ochruros). Ornis Fenn. 83:153–161.
Sorjonen J. 1986. Mixed singing and interspecific territoriality consequences of secondary contact of two ecologically and morphologically
similar nightingale species in Europe. Ornis Scand. 17:53–67.
Svensson EI, Eroukhmanoff F, Friberg M. 2006. Effects of natural and sexual selection on adaptive population divergence and premating isolation
in a damselfly. Evolution. 60:1242–1253.
Svensson EI, Eroukhmanoff F, Karlsson K, Runemark A, Brodin A. 2010. A
role for learning in population divergence of mate preferences. Evolution.
64:3101–3113.
Templeton CN, Akcay C, Campbell SE, Beecher MD. 2012. Soft song is a
reliable signal of aggressive intent in song sparrows. Behav Ecol Sociobiol.
66:1503–1509.
1394
Tobias JA, Seddon N. 2009. Signal design and perception in Hypocnemis Antbirds:
evidence for convergent evolution via social selection. Evolution. 63:3168–3189.
Uy JA, Moyle RG, Filardi CE, Presgraves D. 2009. Plumage and song differences mediate species recognition between incipient flycatcher species
of the Solomon Islands. Evolution. 63:153–164.
Wellenreuther M, Vercken E, Svensson EI. 2010. A role for ecology in male
mate discrimination of immigrant females in Calopteryx damselflies? Biol J
Linn Soc. 100:506–518.
Behavioral Ecology
Werema C, McEntee JP, Mulungu E, Mbilinyi M. 2013. Preliminary observations of the birds of Ikokoto Forest, Udzungwa Mountains, Tanzania.
Scopus. 32:19–26.
West-Eberhard MJ. 1983. Sexual selection, social competition, and speciation. Q Rev Biol. 58:155–183.
West-Eberhard MJ. 2003. Developmental plasticity and evolution. Oxford:
Oxford University Press.