Animal Conservation (2004) 7, 161–168 C 2004 The Zoological Society of London. Printed in the United Kingdom DOI:10.1017/S1367943004001246
Ecological correlates of the threat of extinction in Neotropical
G. S. Gage1 , M. de L. Brooke1 , M. R. E. Symonds1,2 and D. Wege3
1 Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK
2 School of Tropical Biology, James Cook University, Townsville, Qld 4811, Australia
3 BirdLife International, Wellbrook Court, Girton Road, Cambridge CB3 0NA, UK
(First received 14 October 2002; revised received 3 October 2003; accepted 11 October 2003)
Predicting the threat of extinction aids efﬁcient distribution of conservation resources. This paper utilises a
comparative macroecological approach to investigate the threat of extinction in Neotropical birds. Data on
ecological variables for 1708 species are analysed using stepwise regression to produce minimum adequate
models, ﬁrst using raw species values and then using independent contrasts (to control for phylogenetic
effects). The models differ, suggesting phylogeny has signiﬁcant effects. The raw species analysis reveals that
number of zoogeographical regions occupied, elevational range and utilisation of specialised microhabitats
were negatively associated with threat, while minimum elevation and body mass were positively associated,
whereas the independent contrasts analysis only identiﬁes zoogeographical regions as important. Conﬁning the
analysis to the 582 species restricted to a single zoogeographical region reveals elevational range and number
of habitats occupied to be negatively correlated with threat whether the analysis is based on the raw data or on
independent contrasts. Analysis of four contrasting zoogeographical regions highlights regional variation in the
models. In two Andean regions the threat of extinction declines as the elevation range across which the species
occurs increases. In the presence of substantial human populations on high Andean plateaus, a species with a
greater elevational range may be more likely to persist at some (relatively) unsettled altitudes. In Central South
America, the strongest predictor of threat is minimum elevation of occurrence: species with a lower minimum
are less threatened. The minimum elevation result suggests that lowland species experiencing an ecological
limit to their minimum elevation (min. elevation > 0 m) may be more at risk than those not experiencing such
a limit (min. elevation = 0 m). Finally, in southern Amazonia, where there is little altitudinal variation, the
only weak predictors of threat are body size, larger species being more threatened, and number of habitats,
species occupying more habitats being less threatened. These contrasting results emphasise the importance of
undertaking extinction risk analyses at an appropriate geographical scale. Since the models explained only a
low percentage of total variance in the data, the effects of human-mediated habitat disturbance across a wide
range of habitats may be important.
INTRODUCTION risk in two mammalian clades, carnivores and primates,
in terms of such parameters as population density and
Extinction rates are rapidly increasing and recent rates may
geographical range size. Predicting the threat of extinction
be 100–1000 times those of pre-human times (Raup, 1978;
is of theoretical interest to population biologists and of
Pimm et al., 1995). Future rates may be still higher. Studies
practical importance to both conservation biologists and
of non-random patterns of species extinction (Raup, 1994;
wildlife managers (Pimm, Jones & Diamond, 1988). For
Bennett & Owens, 1997; McKinney, 1997; Russell et al.,
conservation biologists, such ‘knowledge . . . may provide
1998; Purvis et al., 2000a; von Euler, 2001) and of the
operational paradigms and parameters within which
variables associated with the threat of extinction (Gaston
successful conservation strategies can be developed’
& Blackburn, 1995; Bennett & Owens, 1997; Purvis et al.,
(Winker, 1998). For example, it may aid the identiﬁcation
2000b) suggest it may be possible to predict the threat of
of species not yet threatened but likely to become so
extinction. For example, Purvis et al. (2000b) explained
in the future if pre-emptive conservation measures are
nearly 50% of the between-species variation in extinction
not taken. Successful prediction will also allow current
estimates of extinction rates to be revised (Manne,
All correspondence to: M. de L. Brooke. Fax: +44-1223-336676; Brooks & Pimm, 1999). Thus, understanding extinction
E-mail: firstname.lastname@example.org patterns is one step in allowing conservation biologists
162 G. S. GAGE ET AL.
to devise management strategies (Goerck, 1997). This Table 1. Pearson correlation coefﬁcients between the logarithmic-
paper utilises a comparative approach to investigate ally transformed values of the ﬁve continuous variables used in this
how ecological factors relate to the threat of extinction study of 1708 species of Neotropical bird
amongst Neotropical bird species. Individual species will
function as replicates in assessing the association between Minimum Elevational No. of Zoogeographical
ecological variables and the threat of extinction. Elevation Range Habitats Regions
Parker, Stotz & Fitzpatrick (1996) compiled compre-
hensive information on the ecology and zoogeography Body Mass − 0.107∗∗∗ 0.021 0.054∗ 0.203∗∗∗
of Neotropical bird species. Their unparalleled work in Minimum 0.035 − 0.206 − 0.312∗∗∗
this ﬁeld was based on their belief that knowledge of Elevation
bird communities can be used to prioritise effectively the Elevational 0.350∗∗∗ 0.329∗∗∗
allocation of conservation resources and aid the selection Range
No. of 0.362∗∗∗
of sites as conservation targets.
To this end, we examine how variation in the threat
of extinction varies with ecological traits on a macro- ∗
P < 0.05; ∗∗∗ P < 0.001.
ecological scale. Signiﬁcant statistical correlation may
indicate that the ecological trait is associated with
the threat of extinction. More speciﬁcally, we analyse
how the estimated risk of extinction varies with ele- extinction risk uses the IUCN (The World Conservation
vational distribution, geographical distribution by habitat Union) categories developed by Mace & Stuart (1994).
and zoogeographical region, body mass and dependence The criteria used are primarily based on a combination of
on microhabitat or undisturbed edge habitats. The population and range size and their rate of decline. The
relationship between body mass and extinction risk has threat of extinction was scored on a ﬁve point ordinal scale
already been investigated by several authors, whose where: 1 = not currently threatened; 2 = near threatened;
studies variously report positive, negative or no relation- 3 = vulnerable; 4 = endangered; 5 = critical (after Bennett
ship (see Terborgh & Winter, 1980; Diamond, 1984; Pimm & Owens, 1997). This scale is viewed as a continuous
et al., 1988; Soul´ et al., 1988; Laurance, 1991; Gaston &
e variable (after Purvis et al., 2000b). Species not listed
Blackburn, 1995; Bennett & Owens, 1997; Owens & in Hilton-Taylor (2000) were classiﬁed as ‘not currently
Bennett, 2000). threatened’. A total of 1556 species was so classiﬁed,
while 72 species were near-threatened, 49 vulnerable,
21 endangered and 10 critical.
Data on Neotropical bird species were compiled into a The seven predictor variables
database to allow comparative testing of a single dataset
Male Body Mass: data were taken from Dunning (1992).
through the tandem methods of interspeciﬁc analysis using
Where more than one mean body mass was reported the
raw data and using independent contrasts (Felsenstein,
arithmetic mean, weighted according to sample size, was
1985). Ecological data were primarily taken from the
calculated. Although female body mass may be a more
comprehensive reference database of Parker et al. (1996);
accurate reﬂection of a species’ fecundity, the resulting
data on body masses were taken from Dunning (1992).
reduction in sample size meant that the results were
The full dataset is available on request.
potentially subject to sampling effects and, therefore, we
The data were compiled into a single database of
used male body mass.
1708 Neotropical bird species, approximately half of
Minimum Elevation and Elevational Range: elevational
the Neotropical species. For every species, there was
data in Parker et al. (1996) included the minimum and
a response variable and seven predictor variables, each
maximum elevational distribution for all 1708 species.
explained below. A total of 125 species listed as having
The former was Minimum Elevation in our study. Both of
primary habitats of saltwater marshes, coastal sand
Parker et al.’s elevational ﬁelds used the term ‘L’ to refer
beaches/mudﬂats and coastal or pelagic waters was
to low-relief lowland areas. This was recoded as ‘0 m’
excluded because their ecology is so different to that of the
under minimum elevation and ‘500 m’ under maximum
terrestrial majority of species, and the threats they face,
elevation, 500 m being the upper elevational limit of
such as wetland drainage, are correspondingly different.
lowland regions in Parker et al. (1996). The difference
These excluded species comprised some or all of the
between the maximum and minimum elevations of a
members of the orders Procellariiformes, Pelecaniformes,
species was calculated by straightforward subtraction and
Sphenisciformes, Charadiiformes, Ciconiiformes and
is referred to as the Elevational Range of a species. These
two elevation measures are not signiﬁcantly correlated
(Table 1), an important consideration when undertaking
stepwise regression (see Analysis, below). Together, they
The response variable
give a measure of the altitude occupied by a species and
Threat of Extinction: measures of the threat of extinction whether it occurs over a narrower or wider altitudinal
were taken from Hilton-Taylor (2000). This listing of avian range.
Extinction risk in Neotropical birds 163
No. of Habitats and No. of Zoogeographical Regions: 1985). The results provide a practical comparison of
habitats refers to 41 habitat categories based on how the use of interspeciﬁc analyses of raw data and
Neotropical birds appear to distinguish vegetation type independent contrasts (Blackburn & Gaston, 1998) and
(Parker et al., 1996). Zoogeographical regions refers to allow signiﬁcant conclusions to be accepted with greater
the 22 regions of the Neotropics identiﬁed by Parker et al. statistical validity.
(1996) on the basis of major vegetation types and The Comparative Analysis by Independent Contrasts
physiogeographical features. For each species the number (CAIC) program (Purvis & Rambaut, 1995) employs
of regions and habitats occupied is tallied on a continuous Pagel’s version of Felsenstein’s method (Pagel, 1992) to
scale. produce statistically independent contrasts at each node
Microhabitat: species with a critical dependence on of a phylogeny. The threat of extinction was taken as the
a speciﬁc feature of a habitat type are listed as being dependent, or response, variable. For comparative studies
associated with a microhabitat (Parker et al., 1996). taxa must represent equivalent and comparable units and
The eight individual microhabitats (army ant swarms, so a single phylogeny was used to avoid taxonomic
bamboo, ground bromeliads, rocky outcrops or caves, differences. The full Sibley & Ahlquist (1990) phylogeny
streamside, treefalls, vine tangles or bamboo/vine tangles) based on DNA–DNA hybridisation comparisons was used.
are typically restricted to forest areas. In this analysis Data on lengths were not available and so branch lengths
Microhabitat is a dichotomous variable; species either were set under CAIC’s default assumptions to equal
show a dependence upon any microhabitat or they lengths.
do not. While CAIC’s BRUNCH programme may analyse
Edge: edge species listed are those whose primary dichomotomous variables one at a time, we wished
habitat comprises the edges of undisturbed habitat. Note to examine the predictive effect of several variables
that they are distinct from the species that have adapted simultaneously. Moreover, for multiple regression, all
to secondary forests. This variable is dichotomous; species contrasts must be generated within the same analysis and,
are either dependent upon an edge habitat or they are therefore, we had to exclude the dichotomous variables
not. Edge and Microhabitat, from the analysis of independent
contrasts. The remaining six continuous variables (Threat
of Extinction, Body Mass, Minimum Elevation, Eleva-
Analysis tional Range, Habitats and Zoogeographical Regions)
were logarithmically transformed (as mentioned above)
All data were analysed using the Minitab statistical com-
and analysed using CAIC’s CRUNCH programme.
puter package. Continuous variables were logarithmically
Independent contrasts were tested by regression through
transformed before analysis and, once transformed, were
the origin (Garland, Harvey & Ives, 1992). If the
found to be normally distributed. In the presentation of
regression coefﬁcients differed signiﬁcantly from zero,
results, the regression coefﬁcients given are based on the
then the correlated changes between the two characters can
analysis using transformed variables.
be said to have occurred independently of phylogenetic
An aggregated minimum adequate model (MAM) was
association (Cotgreave & Harvey, 1992).
derived using stepwise regression with forward selection
To examine regional variation in the predictors of
and model simpliﬁcation (Zar, 1999). The chosen cut-
extinction risk, habitat, elevational and body mass data
off P value was 0.15. The result is a model representing
were analysed for the species dwelling in four individual
a particular statistical likelihood and should not be
zoogeographical regions, two of high montane habitat
considered as the true or complete model. For instance,
and two of lowland (descriptions were taken from Parker
the inclusion of additional variables or changes to the P
et al., 1996):
value may result in a different model.
The results of stepwise regression may be unreliable Northern Andes: all the montane regions, from the coastal
if the independent variables are themselves highly cordilleras of Venezuela, south to Porculla Pass and the
correlated. For this reason, Table 1 shows the correlation Rio Maranon in Peru.
matrix for the ﬁve continuous variables assessed in our Central Andes: the montane region and associated valleys
study: no correlation coefﬁcients exceed 0.4. from Porculla Pass and the Maranon Valley, Peru, south
Evolutionary relatedness amongst species means that to Tucuman and Catamarca, Argentina and northern
each species is not an independent data point and Atacama, Chile.
that sample sizes and the degrees of freedom available Central South America: the lowland, open habitats
for testing statistical signiﬁcance are artiﬁcially inﬂated stretching from Maranhao east to Rio Grande de
(Felsenstein, 1985; Harvey, 1996). This is due to phy- Norte and south through interior eastern Brazil,
logenetically related species being more similar than eastern Bolivia and Paraguay to Rio Negro in central
would be expected by chance alone and results in Argentina.
elevated type I error rates (false rejection of a null hypo- Amazonia South: the area south of the Rio Maranon, Rio
thesis). In order to account for variation in the degree of Solimoes and Rio Amazonas, west to the base of the
common phylogenetic association the data were assessed, Andes and south and east to the edge of Amazonian
ﬁrst, using interspeciﬁc analysis of raw data and, second, forest in Santa Cruz, Bolivia and Mato Grosso, Goias
using phylogenetically independent contrasts (Felsenstein, and Maranhao, Brazil.
164 G. S. GAGE ET AL.
Table 2. Results of linear regression of extinction risk of Neotropical y-intercept) of the raw data. The resultant model, which
birds on various parameters explains 9.97% of the total variance in extinction risk, is
(for log-transformed variables):
Predictor N t value P value R2 adj. (%)
Threat of Extinction
No. of Zoogeographical 1708 − 11.13 < 0.001 6.7
Regions = 0.567 − 0.088 No. of Zoogeographical Regions
Minimum Elevation 1708 4.54 < 0.001 1.1
Elevational Range 1708 − 8.71 < 0.001 4.2 − 0.070 Elevational Range + 0.023 Body Mass
Body Mass 1708 2.53 0.011 0.3
+ 0.005 Minimum Elevation − 0.051 Microhabitat
No. of Habitats 1708 − 6.72 < 0.001 2.5
Edge 1708 − 2.84 0.005 0.4 − 0.043 Edge
Microhabitat 1708 − 1.24 0.214 0.0
Increasing No. of Zoogeographical Regions and Ele-
All parameter values were logarithmically transformed before vational Range, or dependence upon a Microhabitat or
analysis. possibly an Edge, are all associated with a reduction
in the threat of extinction. Increasing Body Mass and
Minimum Elevation increases the threat of extinction. Of
Following the methods of the cross-continent analysis,
these predictors, No. of Zoogeographical Regions is the
we focused on the ﬁve continuous variables, Minimum
most signiﬁcant. A small geographical range is a major
Elevation, Elevational Range, No. of Zoogeographical
correlate of threat: so much so that its inclusion in the
Regions, No. of Habitats and Body Mass for the regional
model may obscure other correlates.
analyses. These were logarithmically transformed and
Therefore, to explore in more detail predictors of
independent contrasts produced by CAIC were entered
threat for species with modest range size, we repeated
into a stepwise regression as above.
the analysis for the 582 species that are conﬁned to a
single zoogeographical region. In this case, the model,
explaining 8.56% of the variation, included four variables
RESULTS (all P < 0.05). It was:
Non-phylogenetic analysis of raw species values Threat of Extinction
Table 2 shows the results of the ﬁrst step of stepwise = 1.067 − 0.127 Elevational Range
regression, which is equivalent to univariate analysis of
each variable. All variables, except Microhabitat, were − 0.134 No. of Habitats − 0.186 Microhabitat
signiﬁcant at the set P-value of 0.15 and, in fact, at 0.05. + 0.032 Body Mass
The most signiﬁcant variable, No. of Zoogeographical
Regions, was included in the model and the regression Thus, the model showed considerable similarities to
step was repeated with each remaining variable being the trans-continental model, but No. of Habitats, which is
included in the model in turn. Table 3 shows the step- moderately well correlated with No. of Zoogeographical
by-step development of the stepwise regression (with a Regions (see Table 1), now entered the model.
Table 3. The step-by-step development of the regression model using interspeciﬁc analysis to identify correlates of extinction risk in 1708
species of Neotropical bird
Step 1 2 3 4 5 6
Constant 0.1705 0.6392 0.5390 0.5517 0.5660 0.5673
No. Zoogeographical Regions − 0.1044 − 0.0865 − 0.0964 − 0.0882 − 0.0891 − 0.0883
t-value − 11.13∗∗∗ − 8.79∗∗∗ − 9.64∗∗∗ − 8.34∗∗∗ − 8.42∗∗∗ − 8.35∗∗∗
Elevational Range − 0.068 − 0.065 − 0.070 − 0.071 − 0.070
t-value − 5.52∗∗∗ − 5.31∗∗∗ − 5.61∗∗∗ − 5.70∗∗∗ − 5.64∗∗∗
Body Mass 0.0236 0.0241 0.0236 0.0228
t-value 4.79∗∗∗ 4.88∗∗∗ 4.78∗∗∗ 4.61∗∗∗
Minimum Elevation 0.0054 0.0052 0.0051
t-value 2.36∗ 2.29∗ 2.22∗
Microhabitat − 0.051 − 0.051
t-value − 1.95(∗) − 1.98∗
Edge − 0.043
t-value − 1.81(∗)
R2 adj. (%) 6.72 8.30 9.46 9.70 9.85 9.97
All parameter values were logarithmically transformed before analysis, and the regression coefﬁcients refer to the transformed values.
P < 0.1; ∗ P < 0.05; ∗∗ P < 0.01; ∗∗∗ P < 0.001.
Extinction risk in Neotropical birds 165
Table 4. The step-by-step development of the regression model using independent contrasts to identify correlates of extinction risk in
those Neotropical bird species conﬁned to a single zoogeographical region
Step 1 2 3 Species Contrasts
Elevational Range − 0.154 − 0.132 − 0.125 582 182
t-value − 5.74∗∗∗ − 4.76∗∗∗ − 4.49∗∗∗
No. of Habitats − 0.115 − 0.118 582 182
t-value − 3.00∗∗ − 3.10∗∗
Body Mass 0.035 582 182
R2 adj. (%) 5.21 6.50 7.37
All parameter values were logarithmically transformed before analysis, and the regression coefﬁcients refer to the transformed values.
P < 0.05; ∗∗ P < 0.01; ∗∗∗ P < 0.001.
Phylogenetic analysis of independent contrasts
When we undertook the regional analyses within CAIC
When we performed a stepwise regression (through the using the ﬁve continuous variables, No. of Zoogeo-
origin) of the 381 independent contrasts, the resultant graphical Regions appeared as a signiﬁcant variable
model included only a single variable, No. of Zoogeo- (P < 0.05) in every model, along with 0–2 other variables.
graphical Regions, and explained 8.6% of the variation Thus, as was true of the continent-wide analysis, the
around the origin. Increasing No. of Zoogeographical regional analyses indicated that number of regions, a
Regions was associated with reduced threat of extinction surrogate for geographical range size, was an important
(t = − 6.06, P < 0.001). Thus, while No. of Zoogeo- variable. We then repeated the analyses excluding No.
graphical Regions was signiﬁcant in both the raw data of Zoogeographical Regions. The results for the four
and independent contrasts models, Elevational Range, individual regions are shown in Table 5. Within both
Body Mass, Minimum Elevation and Microhabitat were the two primarily montane regions (Northern Andes
signiﬁcant only in the former (where the sample size was and Central Andes), Elevational Range was the most
over four times greater). signiﬁcant predictor variable. The direction of the
Conﬁning the independent contrasts analysis to the effect was similar in both regions: species with greater
582 species occurring in just one region yielded a model elevational ranges were less threatened. Within Central
that explained 7.4% of the variation about the origin South America, Minimum Elevation was the most
(Table 4). With the exception of Microhabitat, the signiﬁcant predictor, but No. of Habitats also entered the
variables included (Elevational Range, No. of Habitats model. Finally, within Amazonia South, Body Mass and
and Body Mass) were the same as appeared in the raw No. of Habitats entered the model, but both were only
data analysis (see above). marginally signiﬁcant. The percentage of variation about
Table 5. Analysis of extinction risk of birds in four zoogeographical regions
Northern Andes† Central Andes‡ Central South America§ Amazonia South
Step 1 2 1 1 2 1 2
Elevational Range − 0.212 − 0.247 − 0.128
t-value − 4.10∗∗∗ − 4.39∗∗∗ − 2.60∗
Minimum Elevation 0.135 0.122
t-value 7.82∗∗∗ 6.80∗∗∗
Body Mass 0.027
No. of Habitats 0.086 − 0.090 − 0.033 − 0.034
t-value 1.53(∗) − 2.12∗ − 1.88(∗) − 1.97(∗)
All parameter values were logarithmically transformed before analysis, and the regression coefﬁcients refer to the transformed values.
P < 0.15; ∗ P < 0.05; ∗∗∗ P < 0.001.
N = 332 species and 120 contrasts.
N = 345 species and 128 contrasts.
N = 346 species and 135 contrasts.
N = 512 species and 166 contrasts.
166 G. S. GAGE ET AL.
the origin explained by these models varied considerably extrinsic factors. However, here we investigated variation
as follows: Northern Andes 12.7%, Central Andes 4.3%, in ecological factors that might predispose Neotropical
Central South America 32.8% and Amazonia South 2.4%. bird species to the threat of extinction. Major determinate
factors, such as human-mediated habitat destruction, lay
outside the scope of the current investigation and their
DISCUSSION effect may have limited the explanatory power of the
analyses, a point emphasised by Owens & Bennett (2000).
Comments on analysis
In fact, despite the inclusion of ﬁve signiﬁcant predictor
Analysis of the raw data illustrates both the importance variables, only 10% of the total variance was explained
and power of stepwise regression and MAM techniques. by the model using the raw data (Table 2). The results
Probably because of the correlation with No. of Zoogeo- of other studies (Gaston & Blackburn, 1995; Bennett &
graphical Regions, the stepwise model did not contain Owens, 1997) indicate that the inclusion of life history and
the variable No. of Habitats, which was signiﬁcant when fecundity data would possibly enhance the explanatory
considered as the sole predictor variable in the univariate power of the model, but no relevant comprehensive data
analysis. sets are available.
The tandem methods of analysis, using raw data and In many cases, the threat of extinction will be sub-
using independent contrasts, produced different models, stantially due to large-scale habitat destruction operating
as might be expected from the fact that both methods over large geographical areas and across various habitats.
are imperfect (c.f. Ricklefs & Starck, 1996). Thus, while The ecological features of a species will then only weakly
No. of Zoogeographical Regions was signiﬁcant in both explain a species’ risk of extinction because, put simply,
the raw data and the independent contrasts models for the a signiﬁcant fraction of species will be at risk. This effect
entire continent, Elevational Range, Body Mass, Minimum of large-scale habitat destruction could be a reason for the
Elevation and Microhabitat were signiﬁcant only in the low percentage of total variance explained by the analysis.
former. The decline in threat of extinction with occupation of
Our independent contrasts analysis is, of course, more distinct zoogeographical regions, identiﬁed by both
dependent on the phylogeny used. Sibley & Ahlquist’s continent-wide models, is intuitively reasonable. As the
(1990) phylogeny is, like any phylogeny, imperfect. In number of zoogeographical regions occupied increases,
particular, its lack of resolution (only 381 contrasts were the geographical area occupied by the species increases,
produced from 1708 species) results in a substantial loss of reducing the species’ exposure to smaller scale local
information (Purvis, Gittleman & Luh, 1994) and this may effects. The results, therefore, mirror those of Purvis
explain why fewer variables were identiﬁed as signﬁcant et al. (2000b) who found that a greater geographical
in the independent contrasts analysis. However, there range among mammals was associated with a reduced
is considerable evidence that using even a moderately threat. However, and perhaps unexpectedly, the number
accurate, if imperfect, phylogeny produces more accurate of habitats occupied by a bird species in our study did
results than using no phylogeny at all (Symonds, not appear as a signiﬁcant predictor of extinction risk
2002). in either multivariate analysis. This may have arisen
The generation of 381 contrasts from 1708 species may because number of habitats is correlated with number of
be due to one of two factors: a lack of resolution or zoogeographical regions (Table 1) and it is the latter that
accuracy in the phylogeny, or an accurate representation is the more important predictor of threat.
of polytomies found in the bird phylogeny. The full This suggestion is reinforced by the analyses of those
Sibley & Ahlquist (1990) phylogeny is based on DNA– species conﬁned to a single region (Table 4). Here it
DNA hybridisations and is thought to reﬂect true bran- emerged that species that were able to utilise more habitats
ching patterns more accurately than other phylogenies were less likely to be threatened.
currently available (Cotgreave & Harvey, 1992). However, The continent-wide raw data model, but not the
the phylogeny (Sibley & Ahlquist, 1990) is relatively independent contrasts model, identiﬁed four further
unresolved below the subfamily level and this may signiﬁcant variables that we now consider in turn, since the
contribute to the redundancy of data. Polytomies do discussion is of direct relevance to the models emerging
exist in the true phylogeny and are limited by issues of from the regional analysis.
speciation and species deﬁnition. Moreover, analysis using Larger body mass was associated with an increased
independent contrasts is fairly robust to the ‘bushiness’ of risk of extinction in the raw data model (Owens &
phylogenetic trees caused by polytomies (Purvis et al., Bennett, 2000). We cannot determine whether this is a
1994). direct or indirect effect. Body mass may correlate with
other variables, which are, in turn, correlated with the
threat of extinction (Pimm et al., 1988; Lawton, 1994;
Comments on extinction risk
Gaston & Blackburn, 1995). For instance, body size may
Extrinsic variables, as opposed to ecological variables, are be negatively correlated with species abundance (Peters,
likely to be signiﬁcant in explaining variance in the threat 1983), which may itself be negatively correlated with the
of extinction; Diamond’s (1984) ‘evil quartet’ of habitat threat of extinction (Pimm et al., 1988; Laurance, 1991).
loss, over-exploitation, introduced species and chains Alternatively body size may be positively correlated with
of extinction provides well-known examples of such susceptibility to environmental perturbation (Lindstedt &
Extinction risk in Neotropical birds 167
Boyce, 1985) and this may be positively correlated with (Table 3). This was also true when the analysis was
the threat of extinction (Pimm, 1991; Lawton, 1994). restricted to the species conﬁned to a single zoogeo-
Other ecological traits that might be related to both graphical region. For technical reasons this effect could
body size and the threat of extinction are foraging strata, not be investigated when phylogenetic effects were
longevity, fecundity and dispersal ability (see Pimm controlled. If there is an effect, it is almost certainly
et al., 1988; Kunin & Gaston, 1993; Lawton, 1994). small and it perhaps arises because at least some
When phylogenetic effects are controlled, the relationship microhabitats (e.g. rocky outcrops, caves) may remain
between male body mass and the threat of extinction is no relatively unaffected by regional habitat damage.
longer signiﬁcant. Within the two montane regions (Central Andes and
The raw data model predicted that increased elevational Northern Andes) the threat of extinction can be predicted
range was correlated with a reduced risk of extinction. A from data on Elevational Range (Table 5). Reduced
species with a larger elevational range has a greater chance Elevational Range increases the threat of extinction,
of avoiding human impact somewhere within that range as in the continent-wide interspeciﬁc analysis. Within
than a species with a lesser range. these mountainous regions human populations are not
In the raw data model, the threat of extinction increased concentrated at lower elevations but inhabit relatively ﬂat
with increasing minimum elevation. If human populations land wherever it might be available. Species with small
are concentrated in low-lying coastal regions, species elevational ranges may be adversely affected by human
with a minimum elevation of 0 m would be expected to populations living on high altitude mountain plateaus,
be more affected by human-mediated disturbance, and whereas species with greater ranges may be able to avoid
thus at a greater threat of extinction, than species with human inﬂuence, at least at some altitudes within their
a higher minimum elevation. In fact the opposite result overall range. Other studies, both in the Andes (Manne
emerged and we offer the following two non-exclusive et al., 1999) and in south-east Asia (Brooks et al.,
explanations. 1999), have commented on the vulnerability of montane
First, following Brown (1984), species may be thought avifaunas in populated regions.
of as being limited by combinations of physical and The models for the lowland regions are different. In
biotic variables. Spatial variation in population density Central South America, Minimum Elevation is signiﬁcant,
reﬂects the probability density distribution of the required possibly for the ecological reasons discussed above, while
variables and, if some sets of environmental variables are an increased No. of Habitats is associated with reduced
independently distributed, then the density of a species threat. In Amazonia South, where the topography is essen-
should decline towards the limits of its range. Whilst tially ﬂat and elevational variables are unlikely to be im-
this theory was used to explain an abundance–range portant, greater Body Mass and fewer Habitats are the two
size relationship it is also applicable to elevational data variables associated with increased threat, albeit weakly.
if elevation is taken as one of the required variables. These disparate results serve to remind us that models
This might be a reasonable assumption if elevation for each individual zoogeographical region may differ
controls a secondary variable, such as limiting the from the model for the entire Neotropical region and
occurrence of speciﬁc habitat features. Species with a highlight the signiﬁcance of detecting trends at the
minimum elevation at or above 500 m experience an appropriate ecological scale. The importance of detecting
ecological limit to their range and are thus limited in their trends at the regional scale should be borne in mind by
response to habitat disturbance, since they are restricted those planning regional conservation policy, despite the
by the probability density distributions of their required temptation to draw on more readily available global or
variables. In addition such species are likely to occur at continent-wide analyses.
lower relative abundances at the lower elevational limit
of their range, reducing their resilience to population
ﬂuctuations. However, species naturally occurring down Acknowledgements
to sea level may not be ecologically limited. Rather this
We would like to thank Andrew Balmford, Chris Elphick,
minimum elevation is set by the obvious topographical
Rhys Green, Oliver Kr¨ ger, Alison Stattersﬁeld and two
constraint: occurrence below sea level is impossible. Thus,
referees for their thoughtful and patient comments and
populations of such species may be more resilient than
Bill Entwistle, Liz Playle and Buffy Robinson for all sorts
those of species with a higher minimum elevation.
Second, it is broadly the case that the amount of land at
a given elevation decreases the higher the elevation. Thus
the area occupied by species with a minimum elevation
at sea level may be greater than the area occupied by
those with higher minimum elevations. This could lead Bennett, P. M. & Owens, I. P. F. (1997). Variation in extinction risk
to smaller population sizes and higher susceptibility to among birds: chance or evolutionary predisposition? Proc. Roy. Soc.
extinction among the latter group. Lond. Ser. B 264: 401–408.
Blackburn, T. M. & Gaston, K. J. (1998). Some methodological issues
Although species associated with a distinctive Micro- in macroecology. Am. Nat. 151: 68–83.
habitat were not more or less at risk of extinction in a .
Brooks, T. M., Pimm, S. L., Kapos, V & Ravilious, C. (1999). Threat
simple regression (Table 2), they appeared to be slightly from deforestation to montane and lowland birds and mammals in
less at risk of extinction in the raw species analysis insular South-east Asia. J. Anim. Ecol. 68: 1061–1078.
168 G. S. GAGE ET AL.
Brown, J. H. (1984). On the relationship between abundance and ecology and conservation: Stotz, D. F., Fitzpatrick, J. W., Parker,
distribution of species. Am. Nat. 124: 255–279. T. A. III & Moskovits, D. K. (Eds). Chicago: University of Chicago
Cotgreave, P. & Harvey, P. H. (1992). Relationships between body size, Press.
abundance and phylogeny in bird communities. Funct. Ecol. 6: 248– Peters, R. H. (1983). The ecological implications of body size.
256. Cambridge: Cambridge University Press.
Cracraft, J. (1982). Geographical differentiation, cladistics, and Pimm, S. L. (1991). The balance of nature? Chicago: University of
vicariance biogeography: reconstruction of the tempo and mode of Chicago Press.
evolution. Am. Zool. 22: 411–424. Pimm, S. L., Jones, H. L. & Diamond, J. (1988). On the risk of extinction.
Diamond, J. M. (1984). “Normal” extinctions of isolated populations. Am. Nat. 132: 757–785.
In Extinctions: 191–246. Nitecki, M. H. (Ed.). Chicago: Chicago Pimm, S. L., Russell, G. S., Gittleman, J. L. & Brooks, T. M. (1995).
University Press. The future of biodiversity. Science 269: 347–350.
Dunning, J. B. (1992). CRC handbook of avian body masses. Boca Purvis, A. & Rambaut, A. (1995). Comparative analyses by independent
Raton, FL: CRC Press. contrasts: an Apple Macintosh application for analysing comparative
Felsenstein, J. (1985). Phylogenies and the comparative method. Am. data. Comp. Applic. Biosci. 11: 247–250.
Nat. 125: 1–15. Purvis, A., Gittleman, J. L. & Luh, H.-K. (1994). Truth or consequences:
Garland, T., Harvey, P. H. & Ives, A. R. (1992). Procedures for the effects of phylogenetic accuracy on two comparative methods. J.
analysis of comparative data using phylogenetically independent Theoret. Biol. 167: 293–300.
contrasts. Syst. Biol. 41: 18–32. Purvis, A., Agapow, P.-M., Gittleman, J. L. & Mace, G. M. (2000a).
Gaston, K. J. & Blackburn, T. M. (1995). Birds, body size and the Nonrandom extinction and the loss of evolutionary history. Science
threat of extinction. Phil. Trans. Roy. Soc. Lond. Ser. B 347: 205– 288: 328–330.
212. Purvis, A., Gittleman, J. L., Cowlishaw, G. & Mace, G. M. (2000b).
Goerck, J. M. (1997). Patterns of rarity in the birds of the Atlantic forest Predicting extinction risk in declining species. Proc. Roy. Soc. Lond.
of Brazil. Conserv. Biol. 11: 112–118. Ser. B 267: 1947–1952.
Harvey, P. H. (1996). Phylogenies for ecologists. J. Anim. Ecol. 65: Raup, D. M. (1978). Cohort analysis of generic survivorship.
225–263. Paleobiology 4: 1–15.
Hilton-Taylor, C. (Compiler) (2000). 2000 IUCN Red List of threatened Raup, D. M. (1994). The role of evolution in extinction. Proc. Natl.
species. Gland and Cambridge: IUCN. Acad. Sci. USA 91: 6758–6763.
Kunin, W. E. & Gaston, K. J. (1993). The biology of rarity: patterns, Ricklefs, R. E. & Starck, J. M. (1996). Applications of phylogenetically
causes and consequences. Trends Ecol. Evol. 8: 298–301. independent contrasts: a mixed progress report. Oikos 77: 167–172.
Laurance, W. F. (1991). Ecological correlates of extinction proneness Russell, G. J., Brooks, T. M., McKinney, M. M. & Anderson, C. G.
in Australian tropical rain forest mammals. Conserv. Biol. 5: 79–89. (1998). Present and future taxonomic selectivity in bird and mammal
Lawton, J. H. (1994). Population dynamics principles. Phil. Trans. Roy. extinctions. Conserv. Biol. 12: 1365–1376.
Soc. Lond. Ser. B 344: 61–68. Sibley, C. G. & Ahlquist, J. E. (1990). Phylogeny and classiﬁcation of
Lindstedt, S. L. & Boyce, M. S. (1985). Seasonality, fasting, endurance, birds: a study in molecular evolution. Yale: Yale University Press.
and body size in mammals. Am. Nat. 125: 873–878. e
Soul´ , M. E., Bolger, D. T., Alberts, A. C., Sauvajot, R., Wright, J.,
Mace, G. M. & Stuart, S. (1994). IUCN Red List categories. Species Sorice, M. & Hill, S. (1988). Reconstructed dynamics of rapid
21–22: 13–24. extinctions of chaparral-requiring birds in urban habitat islands.
Manne, L. L., Brooks, T. M. & Pimm, S. L. (1999). Relative risk of Conserv. Biol. 2: 75–92.
extinction of passerine birds on continents and islands. Nature 399: Symonds, M. R. E. (2002). The effects of topological inaccuracy
258–260. in evolutionary trees on the phylogenetic comparative method of
McKinney, M. L. (1997). Extinction vulnerability and selectivity: independent contrasts. Syst. Biol. 51: 541–553.
combining ecological and paleontological views. Annu. Rev. Ecol. Terborgh, J. & Winter, B. (1980). Some causes of extinction.
Syst. 28: 495–516. In Conservation biology: an evolutionary–ecological perspective:
Owens, I. P. F. & Bennett, P. M. (2000). Ecological basis of extinction e
119–133. Soul´ , M. E. & Wilcox, B. A. (Eds). New York: Sinauer
risk in birds: habitat loss versus human persecution and introduced Associates.
predators. Proc. Natl. Acad. Sci. USA 97: 12144–12148. Von Euler, F. (2001). Selective extinction and rapid loss of evolutionary
Pagel, M. D. (1992). A method for the analysis of comparative data. history in the bird fauna. Proc. Roy. Soc. Lond. Ser. B 268: 127–130.
J. Theoret. Biol. 156: 431–442. Winker, K. (1998). Recent geographical trends in neotropical avian
Parker, T. A. III, Stotz, D. F. & Fitzpatrick, J. W. (1996). Ecological research. Condor 100: 764–768.
and distributional databases. Supplement to Neotropical birds: Zar, J. H. (1999). Biostatistical Analysis. New Jersey: Prentice Hall.