Song diversification and complexity in canaries and seedeaters (Serinus spp.)

 
CONTINUE READING
Biological Journal of the Linnean Society, 2007, 92, 183–194. With 4 figures

Song diversification and complexity in canaries and
seedeaters (Serinus spp.)
GONÇALO C. CARDOSO* and PAULO GAMA MOTA
Departamento de Antropologia, Universidade de Coimbra, 3000-056 Coimbra, Portugal

Received 6 February 2006; accepted for publication 20 November 2006

Bird song is a composite trait that shows great variation in syntax and phonology between taxa, which hampers
the precise quantification of complexity. This should be more feasible for closely-related species that share aspects
of song organization. We conducted a comparative study of song and syllable characteristics in the largest fringillid
genus, Serinus (canaries and seedeaters), with two objectives: (1) to describe and quantify song complexity in the
genus and (2) to assess evolutionary dynamics of song, in particular whether interspecific song diversification
evolved along intraspecific lines of covariation between syllable characteristics. We document a pattern of intense,
very labile evolution of song in the genus. Well-known vocal constraints such as the relationship between frequency
and body size did not appear to limit this evolution, and intraspecific trade-offs between different aspects of syllable
complexity did not constrain interspecific diversification. Factor analysis returned an axis of variation that
comprehensively expresses song complexity in the genus, and is defined by song speed, repertoire size, and also
various aspects of syllable phonology. The species body size is not related to this axis of complexity, but to a minor
axis that may indicate better respiratory capacity of larger birds. We discuss the pattern of song evolution in
Serinus spp. in the context of the evolutionary dynamics of sexually selected traits. © 2007 The Linnean Society
of London, Biological Journal of the Linnean Society, 2007, 92, 183–194.

ADDITIONAL KEYWORDS: birdsong – evolutionary diversification – sexual selection.

                   INTRODUCTION                                  take into account the variability in other aspects of
                                                                 phonological complexity. This is important because
Bird song exhibits an enormous interspecific variety
                                                                 such syntactical and phonological aspects of song
of syntactical and phonological characteristics, which
                                                                 complexity may be sexually selected (Vehrencamp,
greatly confounds attempts to quantify complexity.
                                                                 2000; Gil & Gahr, 2002). Although important results
Most comparative studies measure song complexity
                                                                 have been obtained comparing simple repertoire mea-
by the size of repertoires because this aspect of com-
                                                                 surements (DeVoogd et al., 1993; Garamszegi, Møller
plexity is often an important sexually selected trait
                                                                 & Erritzøe, 2003), there are cases where the statisti-
(Searcy & Yasukawa, 1996; Searcy & Nowicki, 2000).
                                                                 cal sensitivity in analyses of song is inferior to that
However, repertoires are not comparable between
                                                                 with simpler traits, such as body size or ornamenta-
different types of song organization or syntax
                                                                 tion (e.g. Garamszegi & Møller, 2004), suggesting that
(Kroodsma, 1982), a problem that has not been solved
                                                                 a better quantification of song complexity may be
despite several attempts (Thompson, LeDoux &
                                                                 needed.
Moody, 1994; Garamszegi et al., 2005) and in relation
                                                                    Because of the above problems, it will be difficult to
to which comparative studies have at best made rea-
                                                                 find a relatively complete measurement of song com-
sonable compromises (Gil & Gahr, 2002). Another
                                                                 plexity with a broad range of taxonomic applicability.
problem is that measures of repertoire size do not
                                                                 A more precise and detailed quantification of song
                                                                 complexity may be possible within groups of closely-
                                                                 related species because aspects of syntax and phonol-
*Corresponding author. Current address: Department of
Ecology and Evolution, 1101 East 57th Street, Chicago,           ogy should be shared or vary in directly comparable
IL 60637, USA. E-mail: goncaloc@ci.uc.pt                         ways. The thoroughness and detail of measures

© 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194                   183
184    G. C. CARDOSO and P. G. MOTA

obtained by comparing fine song characteristics in            considerable practical interest in order to frame the
related species may overcompensate for their limited          large volume of work on this model species.
taxonomic applicability. However, detailed compara-
tive studies of song within restricted taxa of birds
                                                                         MATERIAL AND METHODS
have been little explored to date (Price & Lanyon,
2002, 2004; Päckert et al., 2003).                                               SONG   RECORDINGS
   In the present study, we characterize and quantify         We sampled song recordings of the 27 Serinus species
interspecific axes of song complexity across 27 species       available at the National Sound Archive of the British
of the genus Serinus by means of multivariate analy-          Library, London. All recordings were from wild birds
sis of a comprehensive set of song and syllable mea-          and were made in their natural habitats. When more
surements. We also document the pattern of evolution          than five tapes were available per species, the best
of song in the genus, and investigate whether inter-          five were selected based on sound quality. For the
specific song diversification evolved according to            species Serinus serinus, we used three large good-
intraspecific lines of covariation of syllable character-     quality recordings of our own instead. In recordings
istics. It is of particular interest whether negative         containing more than ten songs, we selected the
intraspecific correlations between different aspects of       sequence of ten songs with the best sound quality for
syllable complexity, which can denote trade-offs on           analysis. The focus of our analyses is on the phono-
vocal physiology (Podos, 1997), are replicated in the         logical and syntactical differences between species,
interspecific pattern of covariation. If inter- and           which are much wider than the fine cultural variation
intraspecific patterns are similar, then song diversi-        among conspecific individuals, and are therefore well
fication could result largely from different species          represented with the sample sizes that we obtained.
occupying different positions within a pre-existing           For the number of songs and syllables analysed per
continuum of phonological possibilities. By contrast, if      species, see the Supplementary Material (Fig. S1).
covariation of syllable characteristics between species          All recordings were digitized to 22 050 Hz mono
does not correspond to the intraspecific pattern, then        files and subsequent song measurements were per-
species diversified evolving novel, unpredictable com-        formed on spectrograms with a resolution of 43 Hz
binations of syllable characteristics, possibly overcom-      and 1.45 ms and a lower cut-off frequency of 1 kHz
ing existing trade-offs within species.                       using the software Avisoft-SASLab Pro (Specht, 1999–
   The genus Serinus is the most speciose in the              2002).
family Fringillidae, and is composed of small-bodied
finches, mostly inhabiting Africa, and generally
known as canaries and seedeaters (Clement, Harris &                          DEFINITION   OF SYLLABLES
Davies, 1993). The genus is broadly homogeneous in            We used syllables as the basic unit for spectrographic
ecology (Clement et al., 1993): most birds are granivo-       measurements. Syllables are more or less complex
rous, occupy open habitats, gather in relatively large        temporal aggregations of song elements and are
wandering, nonmigratory flocks (often of mixed-               defined in the literature depending on the species
species) during the nonbreeding season and in                 song syntax (Thompson et al., 1994): in the case of
smaller flocks during reproduction; they are socially         repeated or trilled syntax, syllables are defined by a
monogamous and, except for the rarer species, nest in         syntactical criterion (the repeated units that form
a loose semicolonial pattern. Species in this genus do        trills); in the case of nonrepeated syntax, syllables are
not have repertoires of song-types but, instead, have         defined by a temporal criterion (the units separated
songs with variable lengths and different proportions         by a certain threshold of interval duration). The two
of syllable sequences and repetitions (Güttinger,             syntaxes coexist in most Serinus species and we used
1985; Leitner, Voigt & Gahr, 2001; Mota & Cardoso,            the two criteria accordingly, as described below.
2001). In the species of Serinus studied to date, song           We first obtained provisional temporal units using
has an important role in the reproductive biology,            the Avisoft automatic measurements tool. Amplitude
namely in female stimulation (Leboucher et al., 1998;         thresholds were adjusted for each recording to attain
Mota, 1999; Gil et al., 2004; Mota & Depraz, 2004).           the best sound to interval discrimination. To detect all
One of these species, Serinus canaria, is an exten-           intersyllable intervals and also to ignore small traces
sively used model system in studies of song learning          of background noise, the temporal sensitivity was set
(Mundinger, 1995; Gardner, Naef & Nottebohm,                  at 10 ms because this is the smallest duration of
2005), sexual selection (e.g. Drăgănoiu, Nagle &            intersyllable intervals in S. serinus (Mota & Cardoso,
Kreutzer, 2002), neuroethology (Nottebohm, Notte-             2001), which is the fastest singing species of the
bohm & Crane, 1986; Spencer et al., 2005), and                genus (i.e. with the smallest average duration of
physiology of song (Suthers et al., 2004). Thus, under-       intersyllable intervals; see Supplementary material,
standing the evolution of song in the genus is of             Fig. S1).

                       © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
SONG EVOLUTION IN SERINUS SPP.               185

   In the portions of song with repeated syntax, we            maximum amplitude; FR_DOMIN), frequency range
applied the syntactical criterion by concatenating the         (FR_RANGE), number of intrasyllable elements
provisional units when these were subdividing the              (defined as the smallest temporally separated traces
repeated syllable. A few harsh sounding very fast              in spectrograms; N_ELEM), number of inflections
repetitions, often referred to as buzzes, were excluded        within the syllable (i.e. the number of times a rising
from further analyses because these are likely to be           frequency modulation is followed by a descending one
produced by a different phonatory mechanism from               or vice versa; N_INFLEX), proportion of the syllable
regular syllable repetitions (Brackenbury, 1978).              with two voices (P_2VOICE), with harmonics
These comprised only 3.3% of cases (maximum in                 (P_HARM) and with buzzes (P_BUZZ; proportions
a single species, 15%) and were always characterized           were coded into five classes: none, approximately one-
by an average duration of the repeated unit or an              quarter, half, three-quarters, and all).
average duration of intervals outlying the species                In the case of repeated syllables, only one set of
distribution.                                                  measurements was used per trill: duration of the
   For the nonrepeated portions of song, we plotted            syllables and intersyllable intervals was averaged
the relative abundance of within song interval dura-           across the trill. All the other measurements were
tions for each species. Plots were made without using          taken from the second syllable in the trill because the
arbitrary categorization of interval durations. For            first sometimes was fused with a previous syllable.
each species, we ordered all intervals by their dura-          We also noted the number of repetitions (RS_N_REP)
tion and then calculated, for every interval, the              and calculated the repetition rate (RS_RATE) of each
difference of durations between the two intervals              trill.
separated from it by 5% of the sample size in that                At the level of songs, we measured duration
ordered sequence of intervals. We calculated differ-           (SNG_DUR), proportion of sound (summed duration
ences at 10% of sample size to buffer the curve                of syllables/duration of song; P_SOUND), total
against fortuitous irregularities. Low differences indi-       number of syllables (SNG_SYLL), and proportion
cate a high relative abundance of intervals with               of nonrepeated syllables (counting only once the
similar durations, and high differences indicate that          repeated syllables in a trill; P_NONREP). We also
similar durations are scarce. In 19 species out of the         calculated the proportion of syllables shared between
25 with sufficiently large samples to draw the plot, we        every pair of consecutive songs; this proportion was
could detect either a depression of abundance or an            calculated relative to the smaller song in the pair. The
abrupt halt in the diminishing of abundance at the             average proportion of shared syllables was then used
region of short interval durations. This denotes the           to compute indices of syllable repertoire size and of
juxtaposition of two partly overlapping distributions          serial consistency for each species. The index of rep-
of duration: the longer intersyllable intervals and the        ertoire size (REPERT) is the average number of syl-
shorter intrasyllable intervals (for a detailed example        lables in songs (counting syllable repetitions as one)
on a Serinus species, see Mota & Cardoso, 2001:                multiplied by 1 + (1 - average proportion of shared
fig. 2). These points were used as thresholds to               syllables). This is an approximation of how many
delimitate syllables: the provisional units separated          different syllables are sung in a short bout of songs, a
by intervals shorter than these thresholds were con-           type of measurement proposed to be more biologically
catenated. The threshold values averaged 0.459 of the          relevant than absolute repertoire size per se (Bell
average intersyllable interval duration for each of            et al., 2004; Garamszegi et al., 2005). By serial
these 19 species; the standard deviation of these              consistency, we mean the similarity of syllable
threshold values relatively to average intersyllable           order within sequences. This index (CONSIST) is
duration was remarkably small (0.087), so that we              the average proportion of shared syllables/average
could extrapolate the criterion to the other species by        number of shared series (a shared series is an intact
iterative calculation. Spectrograms in Figure 4 and            sequence of syllables common to both songs; a series
the Supplementary Material (Fig. S2) illustrate the            is disrupted only by changes in sequential order, and
outcome of the criteria for all species.                       not by repetition of the same syllable).
                                                                  The species average values for these 17 syllable and
                                                               song measurements is provided in the Supplementary
                    MEASUREMENTS                               Material (Fig. S1).
We took nine measurements from spectrograms and
amplitude spectra that comprehensively describe
syllable complexity. These were: syllable duration                             EVOLUTIONARY     LABILITY
(DURATION), average duration of previous and fol-              We assessed the significance of the phylogenetic
lowing intersyllable intervals (INTERVAL), dominant            signal of song measurements by calculating the
frequency (the frequency at which the syllable has             Z-values associated with Moran’s I (Gittleman & Kot,

© 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
186    G. C. CARDOSO and P. G. MOTA

Figure 1. Phylogenetic tree of the subset of Serinus spp. used in the present study that are comprised by the molecular
phylogeny of Arnaiz-Villena et al. (1999). Dashed lines indicate separation of time intervals for testing phylogenetic signal
with Moran’s I (see text).

1990) at three time intervals of the genus phylogeny            cance varies approximately linearly with T; T-values
(Fig. 1, dashed lines). Here and in all other phyloge-          measure the strength of the correlation (Zar, 1996)
netically based analyses, we used the molecular phy-            but do not incorporate the effect of the degrees of
logeny of Arnaiz-Villena et al. (1999), which includes          freedom, which differ between species. Thus, logarith-
16 species of our data set. We also assessed the                mization of significances allows using simple averages
phylogenetic signal of body length (midpoint from               and, at the same time, corrects for the variation in
ranges in Clement et al., 1993) as a reference for              degrees of freedom. The logarithms were signed so
comparison.                                                     that positive and negative correlations have opposing
                                                                effects in the calculation of the average.
                                                                   Negative associations between different aspects of
    SIZE   DEPENDENCE OF SONG CHARACTERISTICS                   syllable complexity may denote vocal trade-offs (e.g.
We assessed the interspecific relationship of syllable          shorter syllables having less inflections), whereas
and song characteristics with body length by multiple           positive associations should denote mostly causal
regression. We also separately regressed through the            relations (e.g. harmonics causing a wider frequency
origin independent contrasts of song characteristics            range). Note that we treat song speed, rather than
on contrasts of body length (Felsenstein, 1985) and             slowness, as an aspect of complexity (see ‘Classifica-
then corrected for multiple comparisons. Contrasts              tion of song complexity’ below). Therefore, correla-
were calculated with the program CONTRAST from                  tions involving the duration of syllables or intervals
the package PHYLIP, version 3.6 (Felsenstein, 2000).            are interpreted as if inverting their sign.
                                                                   We assessed whether the intra- and interspecific
                                                                patterns of correlation of syllable characteristics coin-
      INTRA-   AND INTERSPECIFIC COVARIATION OF                 cide, both for the positive and negative associations.
               SYLLABLE CHARACTERISTICS                         Interspecific correlations describe lines of evolution-
In each species, we calculated the pairwise Pearson             ary divergence, and were calculated using species
correlation coefficients between each pair of the nine          mean values.
syllable characteristics. We corrected for syllable
sharing between songs by calculating the degrees
of freedom of these correlations using an adjusted                       CLASSIFICATION     OF SONG COMPLEXITY
N equal to original N ¥ (1 - proportion of syllable             We ran a principal component analysis (PCA) on the
sharing). To assess whether some of these correlations          species average song characteristics and extracted the
are robust genus wide, we averaged the signed loga-             principal component (PC) scores. We also ran a
rithms of significances across species. Over a wide             similar PCA on the specific components of phyloge-
range of degrees of freedom, the logarithm of signifi-          netic autoregressions (Cheverud, Dow & Leutenegger,

                        © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
SONG EVOLUTION IN SERINUS SPP.                 187

1985) to check for possible phylogenetic effects on the
PCA classification. Phylogenetic autoregressions were
conducted using the software COMPARE, version 4.4
(Martins, 1999).
   Throughout, P < 0.05 (two-tailed) was considered
statistically significant and Bonferroni correc-
tions were applied to multiple comparisons where
appropriate.

                       RESULTS
               EVOLUTIONARY      LABILITY
Only two out of 17 syllable and song measurements
showed significant positive phylogenetic signal
(Z > 1.96; Gittleman & Kot, 1990). These were syl-
lable duration (Z = 2.357) and the number of inflec-
tions per syllable (Z = 2.706), both at the middle time
interval of the phylogeny (Fig. 1). Thus, in general,
song characteristics exhibit a pattern of great evolu-
tionary lability. By contrast, body length did show a
significant phylogenetic signal at the middle time
interval (Z = 2.384).                                          Figure 2. Simple correlation plot illustrating the absence
                                                               of relation between body length and dominant frequency of
                                                               songs in Serinus spp. The large and high frequency outlier
    SIZE   DEPENDENCE OF SONG CHARACTERISTICS                  is Serinus burtoni.
The multiple regression model of song characteristics
on body length was nonsignificant (adjusted
R2 = 0.073, P = 0.448, N = 27) and none of the 17 song              INTRA-   AND INTERSPECIFIC COVARIATION OF
and syllable characteristics varied significantly with                       SYLLABLE CHARACTERISTICS
body length (all |T| < 1.858, all P > 0.096, N = 27).          Nine intraspecific correlations of syllable characteris-
The relationship of body size with frequency of songs          tics were robust genus wide (Table 1) (i.e. their signed
in passerines is well documented (Wallschläger, 1980;          logarithms of significance averaged across the 27
Ryan & Brenowitz, 1985), so we also tested for the             species remained significant at the Bonferroni cor-
existence of this relationship with a simple correla-          rected criteria of significance). The first five correla-
tion of dominant frequency and body length (Fig. 2).           tions in Table 1 are negative associations between
The correlation is not significant and showed a signal         different aspects of syllable complexity, and represent
contrary to prediction (rp = 0.147, P = 0.463, N = 27).        likely trade-offs (e.g. shorter syllables must be
The positive correlation coefficient was due to an             simpler). The last four correlations are positive asso-
outlier (Serinus burtoni, the largest species in the           ciations of syllable characteristics with putative
genus and the second with higher dominant fre-                 causal factors.
quency; Fig. 2); removing it inverted the signal of the           Due to the much smaller sample size of species
coefficient but it remained nonsignificant (rp = -0.153,       than of syllables, only one interspecific correlation
P = 0.455, N = 26).                                            remained significant after Bonferroni correction
   Analyses with independent contrasts showed two              (Table 1). For comparison, the interspecific coeffi-
significant relationships with body length that,               cients that correspond to the significant intraspecific
however, did not remain significant after Bonferroni           coefficients are also shown in Table 1. The intra- and
corrections [average duration of syllables (standard-          interspecific patterns of correlation differ in impor-
ized regression coefficient, Bst = 0.670, P = 0.005,           tant aspects: only one of the five likely intraspecific
N = 15) and serial consistency index (Bst = 0.596,             vocal trade-offs has a nonzero correlation coefficient
P = 0.015, N = 15; Bonferroni corrected level of signifi-      between species (syllable duration and number of
cance = 0.003)].                                               inflections). By contrast, three of the four positive
   The effect of body size on the frequency and other          associations with putative causal factors (the three
characteristics of song was, thus, unimportant in the          involving frequency range) have similar intra- and
genus Serinus. Therefore, we used species averages             interspecific correlation coefficients.
uncorrected for body size in the subsequent interspe-             The same is observed when considering the total-
cific analyses.                                                ity of the correlation coefficients: the correlation

© 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
188    G. C. CARDOSO and P. G. MOTA

Table 1. Intraspecific significant correlations of syllable characteristics compared with the corresponding interspecific
correlations of average syllable characteristics

                                                Intraspecific correlations                    Interspecific correlations

Syllable characteristics                        Average rs           Average P*               rs                 P†

DURATION               N_ELEM                    0.333               0.312 ¥ 10-3             -0.047             0.817
DURATION               N_INFLEX                  0.354               0.003 ¥ 10-3              0.518             0.006
DURATION               P_BUZZ                    0.204               0.297 ¥ 10-3              0.079             0.695
DURATION               FR_RANGE                  0.311               0.045 ¥ 10-3             -0.032             0.873
P_BUZZ                 N_INFLEX                 -0.150               0.167 ¥ 10-3             -0.061             0.759
N_INFLEX               N_ELEM                    0.387               1.089 ¥ 10-3              0.005             0.980
FR_RANGE               FR_DOMIN                  0.185               0.070 ¥ 10-3              0.272             0.16
FR_RANGE               N_ELEM                    0.238               0.182 ¥ 10-3              0.393             0.042
FR_RANGE               P_HARM                    0.197               0.732 ¥ 10-3              0.599             0.951 ¥ 10-3

*Average of signed logarithms of significance across the 27 species (see Material and methods), back-transformed to
significance values. The Bonferroni corrected criterion of significance for 36 correlations (pairwise combinations of nine
syllable measurements) is 1.39 ¥ 10-3.
†Bonferroni corrected criterion of significance is 1.39 ¥ 10-3.
For abbreviations, see Material and methods.

                                                                  lation coefficients differ significantly (T-test for com-
                                                                  paring two slopes, Zar, 1996; T = 3.714, P = 0.001,
                                                                  d.f. = 32). Thus, the interspecific pattern of covaria-
                                                                  tion in syllable characteristics agrees with the
                                                                  intraspecific pattern with respect to positive associa-
                                                                  tions of syllable characteristics, but not with respect
                                                                  to putative trade-offs in complexity.

                                                                             CLASSIFICATION   OF SONG COMPLEXITY
                                                                  The PCA on the species syllable and song character-
                                                                  istics resulted in four significant PCs (i.e. with eigen-
                                                                  values above random expectations by the broken-stick
                                                                  criterion; Jackson, 1993). Eigenvalues of these four
                                                                  components were 3.977, 2.909, 2.298, and 1.987,
                                                                  which correspond to proportions of total variance of
                                                                  0.234, 0.171, 0.135, and 0.117, respectively. Phyloge-
                                                                  netic effects did not influence PCA results because the
                                                                  significant PCs of a PCA on the specific components of
                                                                  phylogenetic autoregressions are characterized by
Figure 3. Correlation plot of the average intraspecific           loadings highly correlated with those of a PCA on
pairwise correlation coefficients of syllable characteristics     simple species averages of the same 16 species set (all
and the corresponding interspecific coefficients. Pairwise        rp > 0.996, all P < 0.001, all N = 17).
correlation coefficients corresponding to positive intraspe-         The first PC is characterized mainly by short syl-
cific associations between aspects of complexity (white)          lable and interval durations, fast repetition rate of
and negative associations (solid). Separate regression lines      the repeated syllables, many syllables per song, and
are shown for each group.                                         large repertoires; most measures of syllable complex-
                                                                  ity, high dominant frequency, and frequency range
                                                                  also have positive loadings on this PC (Table 2). The
between the intra- and interspecific coefficients was             second PC is characterized by long syllables, with
significant for positive associations (rp = 0.555,                many inflections, large proportion of sound in songs,
P = 0.014, N = 19), but not for negative ones                     and few syllable repetitions per trill (Table 2). The
(rp = 0.151, P = 0.563, N = 17, Fig. 3). The two corre-           third PC is characterized by preponderance of

                           © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
SONG EVOLUTION IN SERINUS SPP.                   189

Table 2. Loadings of song characteristics on the significant principal components (PC) of an interspecies principal
component analysis (i.e. correlation coefficients of song characteristics with PC scores); song characteristics are listed by
decreasing strength of correlation with each component

First component                 Second component                Third component                  Fourth component

Characteristic     Loading      Characteristic     Loading      Characteristic      Loading      Characteristic     Loading

RS_RATE             0.830       P_SOUND             0.650       P_NONREP            -0.741       P_BUZZ             -0.647
SNG_SYLL            0.805       N_INFLEX            0.646       SNG_DUR              0.670       REPERT              0.561
REPERT              0.735       DURATION            0.641       P_SOUND             -0.541       SNG_DUR             0.491
INTERVAL           -0.616       RS_N_REP           -0.640       INTERVAL             0.507       P_HARM             -0.463
P_2VOICE            0.566       CONSIST             0.558       SNG_SYLL             0.406       P_NONREP            0.459
DURATION           -0.533       P_HARM              0.505       FR_DOMIN            -0.402       FR_RANGE           -0.458
N_ELEM              0.515       FR_RANGE            0.505       DURATION             0.357       RS_N_REP           -0.309
FR_RANGE            0.497       P_BUZZ              0.363       P_2VOICE             0.274       SNG_SYLL            0.289
FR_DOMIN            0.427       SNG_DUR             0.336       N_ELEM               0.236       N_INFLEX            0.227
CONSIST            -0.408       N_ELEM              0.298       RS_N_REP             0.207       DURATION            0.212
SNG_DUR             0.372       P_2VOICE           -0.195       RS_RATE             -0.198       N_ELEM             -0.199
P_HARM              0.259       INTERVAL           -0.130       P_BUZZ               0.191       FR_DOMIN           -0.126
N_INFLEX           -0.132       FR_DOMIN           -0.126       FR_RANGE            -0.159       P_2VOICE           -0.121
P_BUZZ              0.123       RS_RATE            -0.125       CONSIST             -0.121       INTERVAL            0.118
P_SOUND             0.118       SNG_SYLL            0.091       P_HARM               0.081       P_SOUND             0.100
P_NONREP           -0.028       REPERT             -0.036       REPERT               0.024       RS_RATE             0.067
RS_N_REP           -0.022       P_NONREP           -0.006       N_INFLEX             0.001       CONSIST             0.053

For abbreviations, see Material and methods.

repeated syntax, long songs, and long intersyllable                                   DISCUSSION
intervals (Table 2). The fourth PC is difficult to inter-
pret as an axis of song complexity because it has
                                                                                 SONG    DIVERSIFICATION

many negative weights of syllable complexity mea-               The song characteristics of Serinus spp. show a very
sures (Table 2); it appears to be mainly residual from          labile pattern of evolution. The topography of the
the previous components.                                        phylogenetic tree, with long relative lengths of the
   It is thus necessary to consider three axes of varia-        terminal branches (Fig. 1), can contribute to their low
tion in order to conveniently encompass the variation           phylogenetic signal, but it should not be the sole
in Serinus spp. song complexity. In Table 3, we give            explanation because body length did show a signifi-
the scores of the Serinus spp. at the first three PCs.          cant phylogenetic signal.
Figure 4 shows representative song spectrograms of                 This agrees with the expectation that behavioural
the two species with higher scores at each of the first         and secondary sexual traits exhibit high evolutionary
three PCs; spectrograms of the other Serinus spp.               lability. This expectation results from the frequent
analysed are given in the Supplementary Material                phenotypic plasticity of behaviour (West-Eberhard,
(Fig. S2).                                                      1989; Gittleman et al., 1996; but see also de Queiroz
   PC scores did not show significant positive phylo-           & Wimberger, 1993), and the instability and frequent
genetic signal, at any time interval (Fig. 1; all               reversal in the direction of sexual selection (Cuervo &
Z < 1.96; Gittleman & Kot, 1990). Although song char-           Møller, 1999; Omland & Lanyon, 2000; Wiens, 2001).
acteristics were not individually related to body size          Labile evolution is often found for most characteris-
(see above), the scores of the second PC did correlate          tics of bird song (Irwin, 1990, 1996; van Buskirk,
significantly with body length (rp = 0.487, P = 0.010,          1997), but this is not universal: there are cases of
N = 27, significant after Bonferroni correction; with           consistent phylogenetic signal not only in vocaliza-
PC1 and PC3 both rp < 0.246, both P > 0.216, N = 27).           tions of nonpasserines (McCraken & Sheldon, 1997),
The same result is obtained with independent con-               which are not learned, nor likely to be sexually
trasts of PC scores and body length (PC2: Bst = 0.628           selected, but also on passerine groups with complex
P = 0.009, N = 15, significant after Bonferroni correc-         learned song likely to be important in sexual selection
tion; PC1 and PC3: both Bst < 0.371, both P > 0.157,            (Price & Lanyon, 2002; Päckert et al., 2003). There-
N = 15).                                                        fore, the pattern found in Serinus spp. is informative,

© 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
190       G. C. CARDOSO and P. G. MOTA

Table 3. Species scores on the three significant and interpretable principal component s of song complexity; species are
listed by decreasing score at each component

First component                             Second component                           Third component

Species                       Score         Species                      Score         Species                      Score

Serinus pusillus               2.087        Serinus menelli               1.843        Serinus thibetanus            1.765
Serinus canicollis             1.926        Serinus albogularis           1.491        Serinus leucopterus           1.735
Serinus citrinelloides         1.922        Serinus sulphuratus           1.100        Serinus menelli               1.520
Serinus serinus                1.781        Serinus alario                1.065        Serinus canaria               1.441
Serinus leucopygius            0.716        Serinus canicollis            0.654        Serinus gularis               1.408
Serinus canaria                0.572        Serinus dorsostriatus         0.591        Serinus rufobrunneus          1.126
Serinus burtoni                0.472        Serinus serinus               0.562        Serinus alario                0.667
Serinus flaviventris           0.315        Serinus flaviventris          0.552        Serinus pusillus              0.501
Serinus leucopterus            0.291        Serinus gularis               0.492        Serinus burtoni               0.350
Serinus thibetanus             0.230        Serinus atrogularis           0.477        Serinus capistratus           0.254
Serinus capistratus           -0.151        Serinus striolatus            0.446        Serinus citrinelloides        0.186
Serinus sulphuratus           -0.165        Serinus leucopterus           0.380        Serinus esthereae             0.025
Serinus koliensis             -0.168        Serinus canaria               0.104        Serinus mozambicus           -0.114
Serinus gularis               -0.200        Serinus totta                 0.091        Serinus leucopygius          -0.181
Serinus striolatus            -0.234        Serinus pusillus             -0.062        Serinus sulphuratus          -0.271
Serinus esthereae             -0.326        Serinus mozambicus           -0.162        Serinus flaviventris         -0.326
Serinus atrogularis           -0.379        Serinus leucopygius          -0.201        Serinus canicollis           -0.488
Serinus dorsostriatus         -0.455        Serinus burtoni              -0.230        Serinus striolatus           -0.546
Serinus albogularis           -0.456        Serinus rotschildi           -0.313        Serinus totta                -0.579
Serinus mozambicus            -0.526        Serinus tristriatus          -0.423        Serinus koliensis            -0.589
Serinus rotschildi            -0.606        Serinus capistratus          -0.597        Serinus atrogularis          -0.641
Serinus alario                -0.682        Serinus xantholaema          -0.600        Serinus rotschildi           -0.774
Serinus tristriatus           -0.724        Serinus citrinelloides       -0.658        Serinus albogularis          -0.895
Serinus totta                 -0.761        Serinus rufobrunneus         -0.742        Serinus xantholaema          -1.069
Serinus menelli               -1.364        Serinus thibetanus           -0.931        Serinus tristriatus          -1.219
Serinus xantholaema           -1.515        Serinus koliensis            -1.873        Serinus serinus              -1.501
Serinus rufobrunneus          -1.601        Serinus esthereae            -3.055        Serinus dorsostriatus        -1.783

suggesting that song evolution has been particularly               Intra- and interspecific covariation of syllable char-
intense in this genus.                                          acteristics correspond only partly, so that overall evo-
   Intense evolution of song in Serinus spp. is also            lutionary diversification in the genus could not have
apparent in the clear dissociation of all song charac-          been predicted from the within-species covariation of
teristics and body size. These included temporal                syllable characteristics. The intra- and interspecific
characteristics that have been related to body size             correspondence resulted essentially from basic acous-
(Suthers & Goller, 1997; Podos, 2001) and dominant              tic phenomena and not from putative trade-offs of
frequency. Dominant frequency is especially relevant            syllable complexity. In particular, five genus-wide
because its association with body size is well                  robust trade-offs were detected within species, four of
documented in birds (Wallschläger, 1980; Ryan &                 which involving syllable speed (i.e. shorter syllables
Brenowitz, 1985; Wiley, 1991; Tubaro & Mahler, 1998;            being simpler). Of these, all but one had near-zero
Seddon, 2005) and results from simple size-dependent            interspecific correlation coefficients. Thus, species
efficiency of the vocal organ (Bradbury & Vehren-               that evolved fast songs did not merely reduce syllable
camp, 1998). Body size had a minor influence on the             complexity or partition existing syllables into smaller
song structure because it was correlated only with the          ones, but truly evolved fast and complex syllables.
second, and not the first, component from a factorial           The length of the syllable repertoire was suggested as
analysis of song characteristics. This correlation sug-         a measure of song complexity that could overcome the
gests that larger species, which have a more volumi-            comparability problem between song syntaxes (Irwin,
nous respiratory system, are able to produce songs              2000; Gil & Gahr, 2002); however, when evolution of
characterized by a combination of greater sound to              song speed without loss of complexity occurs, such as
silence ratio and longer syllables.                             in this genus, this measure is not appropriate and

                         © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
SONG EVOLUTION IN SERINUS SPP.                  191

Figure 4. Representative spectrograms of songs. A, Serinus pusillus; B, Serinus canicollis; C, Serinus menelli; D, Serinus
albogularis; E, Serinus thibetanus; F, Serinus leucopterus. The species depicted are the two with highest scores at the
first (A, B), second (C, D) and third (E, F) principal components of the principal component analysis of song and
syllable characteristics. Horizontal lines under the spectrograms identify syllables. Each spectrogram
represents the first 3 s of a song; song renditions vary in length.

syllable complexity must be explicitly included in the         variation appear to indicate distinct and largely inde-
classification of song complexity.                             pendent phenomena. For example, the second compo-
   The high evolutionary lability of song characteris-         nent is related to the species body size and is defined
tics, weak effect of body size, and diversification along      by traits, such as long syllables or large proportion of
lines orthogonal to within-species complexity trade-           sound in songs, that could result from better respira-
offs, all indicate that song evolution in Serinus spp.         tory capacities of larger birds.
did not result only from drift, but was often under               Despite the low phylogenetic signal, all species with
selective pressures sufficiently strong to pervasively         high scores in the first component, which character-
overcome evolutionary constraints.                             izes song complexity, are clustered in the same
                                                               lineage (Serinus canicollis, Serinus citrinelloides,
                   SONG COMPLEXITY                             Serinus pusillus and S. serinus). These are inter-
Song complexity in this genus is conveniently                  spersed with other species having median scores
expressed by the first component from the factorial            (Serinus alario, S. canaria and Serinus mozambicus).
analysis of song characteristics because it includes           According to current knowledge on evolutionary
most of the complexity measures taken. Repertoire              dynamics of sexually selected traits, this pattern is
size, a conventional measure of song complexity, is            most easily explained by repeated loss of song com-
important in this axis of variation, but so are other          plexity (Wiens, 2001), which would also explain the
characteristics related to song speed, song frequency,         weak phylogenetic signal remaining. Loss of complex-
and syllable phonology. Although this first compo-             ity was previously suggested for the evolution of song
nent is a comprehensive measure of song complexity,            under domestication in one of these species, S.
two more significant and interpretable components              canaria (Mota & Cardoso, 2001). The pattern that
exist, indicating that song evolution in Serinus               we now document suggests that the same may apply
spp. followed at least as many independent axis of             to some wild species in this lineage. For example,
variation.                                                     S. canaria is an insular species, and insular species
  In other passerine groups, different lineages evolve         often undergo loss of song complexity (Milder &
song in different directions of complexity (Irwin, 2000;       Schreiber, 1989; Baker, 1996; Hamao & Ueda, 2000)
Price & Lanyon, 2004). However, the lack of signifi-           presumably due to relaxation in the intensity of
cant phylogenetic signal of the species principal com-         sexual selection (Griffith, 2000). In this scenario, the
ponent scores indicates that these axes of song                well-documented preference of S. canaria females for
variation do not result from lineage differentiation           fast syllable repetition rates (Vallet & Kreutzer, 1995;
within the genus Serinus. Instead, different axes of           Vallet, Beme & Kreutzer, 1998; Drăgănoiu et al.,

© 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
192    G. C. CARDOSO and P. G. MOTA

2002), which is the strongest defining trait on this             Drăgănoiu TI, Nagle L, Kreutzer M. 2002. Directional
axis of variation, could be an ancestral preference                female preference for an exagerated trait in canary (Serinus
against which song evolved. The measurement of                     canaria) song. Proceedings of the Royal Society of London
sexual and natural selection for song characteristics              Series B, Biological Sciences 269: 2525–2531.
in wild populations of this species could be greatly             Felsenstein J. 1985. Phylogenies and the comparative
informative in this respect.                                       method. American Naturalist 125: 1–15.
                                                                 Felsenstein J. 2000. PHYLIP (Phylogeny inference package),
                                                                   Version 3.6 (alpha). Distributed by the author. Seattle, WA:
                                                                   Department of Genetics, University of Washington.
              ACKNOWLEDGEMENTS                                   Garamszegi LZ, Balsby TJS, Bell BD, Borowiec M, Byers
                                                                   BE, Drăgănoiu T, Eens M, Forstmeier W, Galeotti P,
We thank Richard Ranft, of the National Sound                      Gil D, Gorissen L, Hansen P, Lampe HM, Leitner S,
Archive, for help in the selection of song recordings.             Lontkowski J, Nagle L, Nemeth E, Pinxten R, Rossi J,
This research was funded by the grant PRAXIS XXI/                  Saino N, Tanvez A, Titus R, Török J, Van Duyse E,
BD/19605/99 to G.C.C. and the project POCTI/1999/                  Møller AP. 2005. Estimating the complexity of bird song
BSE/33714 to P.G.M., both from the Fundação para a                 by using capture-recapture approaches from community
Ciência e a Tecnologia.                                            ecology. Behavioral Ecology and Sociobiology 57: 305–317.
                                                                 Garamszegi LZ, Møller AP. 2004. Extrapair paternity and
                                                                   the evolution of bird song. Behavioral Ecology 15: 508–519.
                                                                 Garamszegi LZ, Møller AP, Erritzøe J. 2003. The evolu-
                     REFERENCES                                    tion of immune defense and song complexity in birds. Evo-
Arnaiz-Villena A, Álvarez-Tejado M, Ruíz-del-Valle V,              lution 57: 905–912.
  García-de-la-Torre C, Varela P, Recio MJ, Ferre S,             Gardner TJ, Naef F, Nottebohm F. 2005. Freedom and
  Martinez-Laso J. 1999. Rapid radiation of canaries               rules: the acquisition and reprogramming of a bird’s learned
  (genus Serinus). Molecular Biology and Evolution 16: 2–          song. Science 308: 1046–1049.
  11.                                                            Gil D, Gahr M. 2002. The honesty of bird song: multiple
Baker MC. 1996. Depauperate meme pool of vocal signals in          constraints for multiple traits. Trends in Ecology and Evo-
  an island population of singing honeyeaters. Animal Behav-       lution 17: 133–141.
  iour 51: 853–858.                                              Gil D, Leboucher G, Lacroix A, Cue R, Kreutzer M. 2004.
Bell BD, Borowiec M, Lontkowski J, Pledger S. 2004.                Female canaries produce eggs with greater amounts of
  Short records of marsh warbler (Acrocephalus palustris)          testosterone when exposed to preferred male song.
  song provide indices that correlate with nesting success.        Hormones and Behavior 45: 64–70.
  Journal of Ornithology 145: 8–15.                              Gittleman JL, Anderson CG, Kot M, Luh H-K. 1996.
Brackenbury JH. 1978. A comparison of the origin and               Phylogenetic lability and rates of evolution: a comparison
  temporal arrangement of pulsed sounds in the songs of the        of behavioral, morphological and life history traits. In:
  Grasshopper and Sedge warblers, Locustella naevia and            Martins EP, ed. Phylogenies and the comparative method in
  Acrocephalus schoenobaenus. Journal of Zoology 184: 187–         animal behavior. Oxford: Oxford University Press, 166–205.
  206.                                                           Gittleman JL, Kot M. 1990. Adaptation: statistics and a
Bradbury JW, Vehrencamp SL. 1998. Principles of animal             null model for estimating phylogenetic effects. Systematic
  communication. Sunderland, MA: Sinauer Associates.               Zoology 39: 227–241.
van Buskirk J. 1997. Independent evolution of song struc-        Griffith SC. 2000. High fidelity on islands: a comparative
  ture and note structure in American wood warblers. Pro-          study of extrapair paternity in passerine birds. Behavioral
  ceedings of the Royal Society of London Series B, Biological     Ecology 11: 265–273.
  Sciences 264: 755–761.                                         Güttinger HR. 1985. Consequences of domestication on
Cheverud JM, Dow MM, Leutenegger W. 1985. The quan-                the song structures in the canary. Behaviour 94: 254–278.
  titative assessment of phylogenetic constraints in compara-    Hamao S, Ueda K. 2000. Simplified song in an island popu-
  tive analyses: sexual dimorphism in body weight among            lation of the bush warbler Cettia diphone. Journal of Ethol-
  primates. Evolution 39: 1335–1351.                               ogy 18: 35–57.
Clement P, Harris A, Davies J. 1993. Finches and spar-           Irwin RE. 1990. Directional sexual selection cannot explain
  rows. Princeton, NJ: Princeton University Press.                 variation in song repertoire size in new world Blackbirds
Cuervo JJ, Møller AP. 1999. Evolutionary rates of second-          (Icterinae). Ethology 85: 212–224.
  ary sexual and non-sexual characters among birds. Evolo-       Irwin RE. 1996. The phylogenetic content of avian courtship
  lutionary Ecology 13: 283–303.                                   display and song evolution. In: Martins EP, ed. Phylogenies
DeVoogd TJ, Krebs JR, Healy SD, Purvis A. 1993. Rela-              and the comparative method in animal behavior. Oxford:
  tions between song repertoire size and the volume of brain       Oxford University Press, 234–252.
  nuclei related to song: comparative evolutionary analyses      Irwin DE. 2000. Song variation in an avian ring species.
  amongst oscine birds. Proceedings of the Royal Society of        Evolution 54: 998–1010.
  London Series B, Biological Sciences 254: 75–85.               Jackson DA. 1993. Stopping rules in principal components

                         © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
SONG EVOLUTION IN SERINUS SPP.                      193

  analysis: a comparison of heuristical and statistical           Ryan MJ, Brenowitz EA. 1985. The role of body size,
  approaches. Ecology 74: 2204–2214.                                phylogeny, and ambient noise in the evolution of bird song.
Kroodsma DE. 1982. Song repertoires: problems with their            American Naturalist 126: 87–100.
  definition and use. In: Kroodsma DE, Miller EH, eds. Acous-     Searcy WA, Nowicki S. 2000. Male-male competition and
  tic communication in birds, Vol. 2. New York, NY: Academic        female choice in the evolution of vocal signalling. In:
  Press, 125–146.                                                   Espmark Y, Amundsen T, Rosenqvist G, eds. Animal
Leboucher G, Depraz V, Kreutzer M, Nagle L. 1998. Male              signals: signalling and signal design in animal communi-
  song stimulation of female reproduction in canaries: fea-         cation. Trondheim: Tapir Academic Press, 301–315.
  tures relevant to sexual displays are not relevant to nest      Searcy WA, Yasukawa K. 1996. Song and female choice. In:
  building or egg laying. Ethology 104: 613–624.                    Kroodsma DE, Miller EH, eds. Ecology and evolution of
Leitner S, Voigt C, Gahr M. 2001. Seasonal changes in the           acoustic communication in birds. Ithaca, NY: Cornell
  song pattern of the non-domesticated island canary (Serinus       University Press, 454–473.
  canaria), a field study. Behaviour 138: 885–904.                Seddon N. 2005. Ecological adaptation and species recogni-
Martins EP. 1999. COMPARE 4.4. Distributed by the author.           tion drives vocal evolution in neotropical suboscine birds.
  Bloomington, IN: Indiana University, Department of Biology.       Evolution 59: 200–215.
McCraken KG, Sheldon FH. 1997. Avian vocalizations                Specht R. 1999–2002. Avisoft-Saslab Pro: sound analysis and
  and phylogenetic signal. Proceeding of the National Academy       synthesis laboratory. Berlin: Avisoft.
  of Sciences of the United States of America 96: 3833–3836.      Spencer KA, Buchanan KL, Leitner S, Goldsmith AR,
Milder SL, Schreiber RW. 1989. The vocalizations of the             Catchpole CK. 2005. Parasites affect song complexity and
  Christmas Island warbler, Acrocephalus aequinoctialis, an         neural development in a songbird. Proceedings of the Royal
  island endemic. Ibis 131: 99–111.                                 Society of London Series B, Biological Sciences 272: 2037–
Mota PG. 1999. The functions of song in the serin. Ethology         2043.
  105: 137–148.                                                   Suthers RA, Goller F. 1997. Motor correlates of vocal diver-
Mota PG, Cardoso GC. 2001. Song organization and pat-               sity in songbirds. In: Nolan V Jr, Ketterson ED, Thompson
  terns of variation in the Serin (Serinus serinus). Acta Etho-     CF, eds. Current ornithology, vol. 14. New York. NY: Plenum
  logica 3: 141–150.                                                Press, 235–288.
Mota PG, Depraz V. 2004. A test of the effect of male song        Suthers RA, Vallet E, Tanvez A, Kreutzer M. 2004.
  on female nesting behaviour in the serin (Serinus serinus):       Bilateral song production in domestic canaries. Journal of
  a field playback experiment. Ethology 110: 841–850.               Neurobiology 60: 381–393.
Mundinger PC. 1995. Behaviour-genetic analysis of canary          Thompson NS, LeDoux K, Moody K. 1994. A system for
  song: inter-strain differences in sensory learning, and epi-      describing bird song units. Bioacoustics 5: 267–279.
  genetic rules. Animal Behaviour 50: 1491–1511.                  Tubaro PL, Mahler B. 1998. Acoustic frequencies and body
Nottebohm F, Nottebohm ME, Crane L. 1986. Develop-                  mass in New World doves. Condor 100: 54–61.
  mental and seasonal changes in canary song and their            Vallet E, Beme I, Kreutzer M. 1998. Two-note syllables in
  relation to changes in the anatomy of song-control nuclei.        canary songs elicit high levels of sexual display. Animal
  Behavioral and Neural Biology 46: 445–471.                        Behaviour 55: 291–297.
Omland KE, Lanyon SM. 2000. Reconstructing plumage                Vallet E, Kreutzer M. 1995. Female canaries are sexually
  evolution in orioles (Icterus): repeated convergence and          responsive to special song phrases. Animimal Behaviour 49:
  reversal in patterns. Evolution 54: 2119–2133.                    1603–1610.
Päckert M, Martens J, Kosuch J, Nazarenko AA, Veith               Vehrencamp SL. 2000. Handicap, index, and conventional
  M. 2003. Phylogenetic signal in the song of crests and            signal elements of bird song. In: Espmark Y, Amundsen T,
  kinglets (aves: Regulus). Evolution 57: 616–629.                  Rosenqvist G, eds. Animal signals: signalling and signal
Podos J. 1997. A performance constraint on the evolution of         design in animal communication. Trondheim: Tapir Aca-
  trilled vocalizations in a songbird family (Passeriformes:        demic Press, 277–300.
  Emberizidae). Evolution 51: 537–551.                            Wallschläger D. 1980. Correlation of song frequency
Podos J. 2001. Correlated evolution of morphology and vocal         and body weight in passerine birds. Experientia 36:
  signal structure in Darwin’s finches. Nature 409: 185–            412.
  188.                                                            West-Eberhard MJ. 1989. Phenotypic plasticity and the
Price JJ, Lanyon SM. 2002. Reconstructing the evolution of          origins of diversity. Annual Review of Ecology Evolution and
  complex bird song in the oropendolas. Evolution 56: 1514–         Systematics 20: 249–278.
  1529.                                                           Wiens JJ. 2001. Widespread loss of sexually selected traits:
Price JJ, Lanyon SM. 2004. Patterns of song evolution and           how the peacock lost its spots. Trends in Ecology and
  sexual selection in the oropendolas and caciques. Behavioral      Evolution 16: 517–523.
  Ecology 15: 485–497.                                            Wiley RH. 1991. Associations of song properties with habitats
de Queiroz A, Wimberger PH. 1993. The usefulness of                 for territorial oscine birds of eastern North America. Ameri-
  behavior for phylogeny estimation: levels of homoplasy in         can Naturalist 138: 973–993.
  behavioral and morphological characters. Evolution 47:          Zar JH. 1996. Biostatistical analysis, 3rd edn. Upper Saddle
  46–60.                                                            River, NJ: Prentice Hall.

© 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
194   G. C. CARDOSO and P. G. MOTA

                                      SUPPLEMENTARY MATERIAL
The following material is available for this article online:
Figure S1. Sample sizes and species means of syllable and song measurements for the 27 species of Serinus
spp. used in this study.
Figure S2. Representative spectrograms of songs of Serinus spp. (only the first 3 s of each song are repre-
sented). Horizontal lines under the spectrograms identify syllables.

This material is available as part of the online article from:
http://www.blackwell-synergy.com/doi/abs/10.1111/j.1095-8312.2007.00837.x
(This link will take you to the article abstract).

Please note: Blackwell Publishing are not responsible for the content or functionality of any supplementary
materials supplied by the authors. Any queries (other than missing material) should be directed to the
corresponding author for the article.

                     © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 92, 183–194
You can also read