Species–area relationships of primates in tropical forest fragments: a global analysis


A. H. Harcourt, Department of Anthropology and Graduate Group in Ecology, University of California, One Shields Avenue, Davis, CA 95616, USA (fax 530 7528885; e-mail ahharcourt@ucdavis.edu).


  • 1While in general the tropics and large-bodied tropical forest mammals are poorly understood, the effects of fragmentation on tropical forest and the tropical mammalian order of primates are relatively well studied. Nevertheless, no quantitative synthesis exists of the response of primates to habitat fragmentation. We therefore conducted a meta-analysis of the literature on the species–area relationships of primates in forest fragments in order to investigate regularities and differences among continents and sites.
  • 2The sample comprised 136 forest fragments (≤ 100 km2) at 33 sites in four continents (Africa, Asia, Madagascar and South America). We conducted our analysis at three spatial scales: global, continent and site. As richness (number of species) per site varied, we analysed both richness and proportional richness by area.
  • 3Study sites and size of fragments were unequally distributed across the globe. South America had three times as many sites as any other continent, for which we found less than 10 sites each. Fragment size was small in all continents (global median 1·0 km2).
  • 4Despite considerable noise in the data, primate richness and proportional richness generally declined significantly and linearly with fragment area at all spatial scales except in Africa.
  • 5Neither isolation (distance of fragment from main forest block) nor age of fragment was an obvious influence on proportional richness. However, the global median isolation was only 2 km.
  • 6Synthesis and applications. Fragmentation of habitat clearly threatens the survival of primates. However, study of the effects of fragmentation on primates might be directed in the wrong place. Estimates of minimum area requirements for primate species exceed tens of km2, yet most forest fragments studied measure less than 1 km2. Both to elucidate the biology of contrasts between species in susceptibility to fragmentation and to use research sites for associated conservation efforts, it might be better to direct more attention to fragments of a size in which long-term persistence of some species is possible.


This study was an investigation of the species–area relationship in a tropical forest mammalian taxon. The tropics remain far less well known than temperate regions, despite their biodiversity (Janzen 1986; Gaston & Blackburn 1999). For instance, an analysis of rarity in mammals found data for fewer than 25% of the species in three main tropical zoogeographical regions, compared with data on half of the species in three main temperate regions (Yu & Dobson 2000). Our relative ignorance of the tropics is exacerbated by the fact that tropical forest is disappearing faster than temperate forest (Myers 1984; Collins, Sayer & Whitmore 1991; McNeely et al. 1995; World Resources Institute 2000).

While there have been many studies of the effects of fragmentation of forest, most are of birds (Turner 1996; Laurance & Bierregaard 1997; Hill & Hamer 2004). One of the least frequently studied taxa is large mammals. Given that different taxa respond differently to fragmentation even within classes or orders (Boecklen & Simberloff 1986; Turner 1996; Laurance & Bierregaard 1997; Debinski & Holt 2000; Hill & Hamer 2004), the imbalance in the taxonomic representation needs rectifying.

Primates represent one large-bodied, tropical-forest mammalian taxon that has been studied thoroughly, both in general and in relation to its response to fragmentation (Bernstein et al. 1976; Smuts et al. 1987; Rowe 1996; Nowak 1999; Cowlishaw & Dunbar 2000; Marsh 2003); nevertheless, no quantitative meta-analysis has been conducted. We are thus missing a general synthesis on which to base an understanding of the responses of primates to fragmentation.

The aim of our analysis was not simply to show that fragmentation is associated with loss of primate species; it could hardly be otherwise. Rather, it was to quantify the general species–area relationship and to investigate variation in the relationship across continents and sites. Funding for conservation is in short supply and effort needs to be prioritized. The shape and slope of the species–area curve could, perhaps should, affect conservation planning. Do some continents suffer a higher loss of species than do others? Are there fragment sizes above which increasing size makes no difference to species richness (Lomolino 2000, 2002; Williamson, Gaston & Lonsdale 2001)? Are there differences between continents in rate of loss of species with loss of habitat? If so, why?

We have provided the first, comprehensive, quantitative global review of the consequences of fragmentation for primate species richness. We conducted our analysis at three scales: global, continent and site within continent (cf. Hill & Hamer 2004). We concentrated on proportional richness in relation to area, because the usual richness–area regressions are inappropriate for counts with small sample sizes.

Materials and methods

data source

We searched the published literature for studies of primate species in forest fragments (defined here as 100 km2 or less) that provided data on number of primate species in relation to either size of fragment or distance of fragment to adjacent intact forest (defined as greater than 100 km2). Our final sample consisted of 136 fragments at 33 study sites (Table 1; see the Appendix). We omitted sites where only one species occurred in the region and therefore only one was available to be in any fragment, for example at the north edge of species’ ranges in central America and the south edge in Africa (Lawes 1992; Estrada & Coates-Estrada 1996; Estrada et al. 1999; Lawes, Mealin & Piper 1999). We also omitted nine sites of one fragment each, because we could not be sure from the original papers that more than one species was present in the fragments. These nine sites are listed and marked in the Appendix.

Table 1.  Characteristics of forest fragments that contained primates. Area, *significantly different from each other; no. fragments, number of fragments of < 1 km2 and the total number; distance, distance of fragment to main forest block (i.e. block of > 100 km2); –, no data available
RegionArea (km2)No. fragmentsDistance (km)Age (year)No. species
RangeMedian< 1 km2/total (%)RangeMedian (n)RangeMedian (n)RangeMedian
Globe0·01–1001·072/148 (49)0·1–532 (33)1·5–7230 (69)1–73
Africa0·01–510·2*28/43 (65)0·2–72·1 (22)   19–4343 (23)1–73
Asia0·01–680·914/28 (50)12112 (1)  9–4534 (5)1–62
Madagascar0·01–750·410/19 (53)  6–4030 (18)1–73
S. America0·01–1002·0*20/58 (34)0·1–530·2 (10)1·5–7215 (23)1–73

Except for studies in Madagascar, where nocturnal primate species comprise a large proportion of the total (45%), nocturnal species were often not mentioned in the sources. The omission will not markedly change counts of species richness in the Americas, where less than 10% of the total complement of primate species is nocturnal. However, in Asia they comprise 15% of the primate fauna, and in Africa 20%. Thus, the Asian and African data must be considered a sample of the diurnal species. Because censuses usually miss species (Grayson & Livingston 1993), we used long-term studies as our sources whenever possible, and therefore the counts of the diurnal species should be largely complete.

Our analyses were limited to subsets of the total sample, because not all categories of data were available for every site or fragment.


The review considered only fragments in which primates were recorded. With more than one fragment per site, there was the possibility of pseudoreplication/spatial autocorrelation through site effects when sites were combined. We reduced the effect by running analyses that omitted the site per continent with the most fragments with the relevant data (area, isolation, age). With area, we also removed all sites with five or more fragments, but a sample size of only nine fragments remained. The slope was positive but not significant. Additionally, the site-by-site analysis (below) obviated pseudoreplication.

Many forms of species–area analysis have been applied in fragmentation and insularization studies (Tjørve 2003). Ordinary least-squares linear regression of double-log or semi-log data are common. However, ordinary least-squares regression is inappropriate for small sample size, because the data are effectively counts and therefore not a continuous distribution (McCullagh & Nelder 1989). Also, in the case of our data, the residuals (errors) did not approximate a normal distribution, whether logged or not (e.g. globe number species by log10 area, Shapiro–Wilk W= 0·95, n= 136, P= 0·0007). Poisson regression would be suitable in such cases (McCullagh & Nelder 1989). However, because of the very small number of primate species per fragment (never more than seven), we only used correlation tests on counts. These included non-parametric Spearman correlation tests because of the small sample size for Madagascar (n < 20).

We also ran a site-by-site analysis for degree and direction of slope of richness to area. Spearman correlation tests were used for each site. Besides the importance of examination of the effect at the local as well as regional scale, this site-by-site test prevented spatial autocorrelation.

With the global- and continental-level analyses, we ran ordinary least-squares regressions of arcsine square-root proportional richness (the proportion of the full complement of species in the main forest that was in each fragment) against log10 fragment area and isolation. These regressions provided more information than correlation tests, and the use of proportions accounted for the fact that the number of species available to be in a fragment could differ across sites (cf. Telleria, Baquero & Santos 2003). Use of proportions was valid because their values were more nearly continuously distributed than counts, and the distribution of errors in the proportions was not statistically different from a normal distribution (e.g. globe percentage species by log10 area, Shapiro–Wilk W= 0·97, n= 63, P= 0·41).

Because arguments exist for the use of other than double-log plots, and because of suggestions that the species–area relationship is not necessarily a straight line (Lomolino 2000), we also examined semi-log plots and polynomial functions. We obtained no consistent results as to whether or not polynomials provided a closer fit than a linear regression, and neither they nor semi-log plots changed any relationship from being significant to non-significant or vice versa. We therefore do not provide any of the results here.

We removed outliers in statistical analyses to allow investigation of what the majority of the sampled population was doing. In the same way that it is bad practice to claim a significant relationship that is determined by only a minority of the data, so it is bad practice to suggest no relationship when a relationship is obscured by a minority of the data. Hence outlier detection facilities in statistical programs (below).

When accounting for the interaction of influences (area, continent, isolation, age) on proportional richness, we used analysis of covariance and multiple regression. In regression analyses, the r2 value that we report is always the adjusted r2. We used JMP 5·0·1·2 (SAS Institute Inc. 2002) and Siegel (1956) for statistical analysis. We identified outliers with JMP's Mahalanobis outlier analysis. P-values are two-tailed, despite predictions of the direction of effect, but we give significance values up to 0·1.


fragment number, size and isolation

Studies of primates in fragments are distributed unequally across the globe (Table 1). In our sample, more than three times as many sites were in South America as in any of the other continents, for all of which we found less than 10 study sites.

Three-quarters of the fragments in our sample were under 10 km2, and nearly half were less than 1 km2, although the limit for definition of a fragment was 100 km2 (Table 1; see the Appendix). Ninety-five per cent of fragments in Madagascar were less than 10 km2; 65% of fragments in Africa were less than 1 km2; only in South America was the median fragment size over 1 km2 (Table 1; see the Appendix). Fragment size differed significantly across the continents (F1,3 = 4·51, P < 0·01 for log10 area). South America and Africa differed the most from the other regions; a post-hoc Tukey–Kramer HSD test on contrasts from the overall mean showed that they differed significantly (alpha = 0·05) from one another (Abs(Dif)-LSD = 0·23) but not from the other two continents.

With regard to isolation of fragments (measured as distance to the nearest main forest block), sufficient data were available for only Africa and South America (Table 1). South America's median distance of 0·2 km was a tenth of Africa's, but the difference was not significant. Ages differed across the continents because of South America's significantly younger age (F1,3 = 9·5, P < 0·001).

area effects: richness

Globally, and separately in all continents but Africa, the number of primate species found in fragments decreased significantly as fragment size decreased (Table 2). When spatial pseudoreplication was counteracted by omission of the site per continent with the most fragments, the global relationship remained significant, the relationship in Africa became significant, and Asia was reduced to a very small sample size and its relationship lost significance.

Table 2.  Results of Pearson and Spearman correlation tests on log10 number of species by log10 forest fragment area. Results shown for total data (rT) and after removal of outliers (rT−O). n in parentheses shows number of outlier fragments. Spearman correlations (rS) are with outliers omitted.; rS−1, Spearman coefficients when site per continent with most fragments is omitted, in order to reduce spatial autocorrelation. Total n less than Table 1, because sites with only one species detected to be in a region are omitted
Regionn =rT =P <rT − O =P <rS =P <rS−1 = (n)P <
Globe136 (3)0·300·0010·400·00010·320·00010·35 (79)0·01
Africa 41 (1)0·11NS0·22NS0·20NS0·52 (20)0·05
Asia 260·560·010·560·010·580·010·12 (7)NS
Madagascar 18 (1)0·880·00010·950·00010·920·00010·83 (9)0·01
South America 51 (3)0·250·10·360·020·340·050·36 (42)0·05

Considering sites separately, and investigating whether or not within sites it was usually the case that species richness decreased with decreasing size of fragment, 10 of the 11 sites with at least five fragments showed the expected relationship (Fig. 1). The sites with positive correlations were distributed approximately equally among the continents. The fragments at the single site with a negative correlation (in Brazil) were all extremely similar in size (0·09–0·1 km2), and the smallest fragment had the highest number of species (see the Appendix).

Figure 1.

Spearman correlation coefficients (corr. coeffs) of number of species by fragment area per site in the four continents. Circles, values for individual sites; bar, median. (Outliers not omitted, because sample size per site so small.) Binomial test, assuming p=q = 0·5 n, P < 0·02. *Individual sites with significant coefficients.

area effects: proportional richness

As fragment size dropped, so the proportional richness in the fragments dropped significantly (Fig. 2 and Table 3). The result held for the globe, Madagascar and South America, but not for Africa. (Asia could not be analysed because of lack of data). With correction for spatial pseudoreplication, the relationship became insignificant for South America.

Figure 2.

Proportional species richness (% of number in nearby main forest block) by fragment area for the globe and the four continents (double log10). Globe, P < 0·0001; Africa, NS; Asia, n = 1; Madagascar, P = 0·01; South America, P < 0·01. Statistical details in Table 3. Statistics are for arcsine √proportional richness (outliers removed). Log10 percentage richness shown in graph, because log10 values are more immediately comprehensible by most than are arcsine √values.

Table 3.  Results of linear regression and Spearman's correlation analyses of arcsine √proportion of available species by log10 forest fragment area, with available total number taken to be that reported for the adjacent main forest block. Outliers removed before analysis. n in parentheses in first column show number of outlier fragments. rS, Spearman correlation coefficient; rS−1, Spearman coefficients when site per continent with most fragments is omitted in order to reduce spatial autocorrelation. Results shown only if n≥ 5
Regionn =r2 =F =P <SlopeInterceptrS =P <rS−1 = (n)P <
Globe58 (4)0·3938·20·0001 0·150·81 0·620·00010·60 (20)0·001
Africa21 (2)0·060·05NS−0·010·58−0·06NS
Asia 1
Madagascar18 (1)0·4212·50·01 0·140·70 0·670·010·70 (9)0·05
South America230·2011·10·01 0·200·95 0·580·010·50 (16)0·1

The global results showed a significant continent effect because of Madagascar (F1,3 = 4·3, P < 0·01). When Madagascar was omitted, no continent effect existed (F1,2 = 0·8, P > 0·1), and the proportional richness–area relationship remained highly significant (F1,2 = 13·9, P < 0·001). The global result with correction for spatial auto-correlation showed no continent effect (F = 0·8), leaving the global proportional richness–area relationship significant (n = 27, F1,3 = 11·8, P < 0·01).

isolation and ageproportions

Globally, there was no significant effect of isolation on proportional richness (all P > 0·1). A global effect of age was apparent (n = 42, F= 4·1, P < 0·05). However, one site in Africa provided most of the data on both age and distance (Onderdonk & Chapman 2000). When this single site was omitted, no effects were evident. Within continents, either there were insufficient data for analysis (n < 5) or no significant effect was found.

Only two sites with five or more fragments had data on distance to the main forest block; neither showed a significant association of isolation with proportional richness. Four sites provided data on age and proportional richness. In three sites, age did not vary across fragments; in the other, no correlation with age existed. Accounting for area did not change these non-significant results for isolation and age.

With no effects evident for isolation and age alone, incorporation of either measure in a multivariate analysis left relationships between proportional richness and area significant (isolation n= 24, F= 6·35, P < 0·02; age n= 43, F= 11·3, P < 0·01). With corrections for spatial auto-correlation, the global sample size for effects of isolation on the proportional richness–area relationship was too small for analysis. The correction did not change the global result for age (n = 14, Pearson r= 0·60, partial r= 0·68). The sample size was too small for any effects to be investigated within continents.


unequal distribution of studies

Studies of the effects of fragmentation on primates appear to be unequally distributed across the tropics. This inequality is apparently true of not just primates but of all fragment studies. Thus a search on ‘forest fragment’ for the 5 years 1999–2003 with Ovid Biosis Previews (Ovid Technologies 2004) produced more than four times as many references for South American studies as for the next most commonly studied tropical continent (Asia), whether studies of primates were excluded or not.

The disparity might not matter if there were scores of studies in the least-studied continents. However, in all continents but South America we found less than 10 sites studied. A more representative and informative understanding of the response of primates to fragmentation requires greater study in continents other than South America, given that continents differ in their complement of primates (indeed all taxa) and in the biology of those primates (Wallace 1876; Terborgh & van Schaik 1987; Fleagle & Reed 1996; Kappeler & Heymann 1996; Fleagle, Janson & Reed 1999).

a strong influence of area but not isolation or age

A strong effect of area on richness is apparent globally, and also in all continents separately, except for Africa. The area effect is present despite all the noise in the data, especially the confounding factor of differences between sites in total numbers of species in similarly sized fragments. Moreover, the effect is present despite all the noise that we did not include in the data (such as contrasts between sites and fragments in time since fragmentation, intensity of hunting, nature of secondary growth in the fragments, edge effects, and all the other important biological processes that are influenced by and influence the effects of fragmentation). However, the fact that so many of these effects are correlated with area of fragment (Laurance & Cochrane 2001) means that they are to some extent subsumed within the measure of area of fragment.

We tested whether any semi-log or polynomial regressions fitted the data better than linear regressions. They did not, implying that there are no asymptotes in the data, either upper or lower, despite the very small size of most of the fragments (discussed below) (Lomolino 2000, 2002; Williamson, Gaston & Lonsdale 2001). In other words, species continue to be lost at the same rate until the last species goes, and there is no fragment size up to 100 km2 above which the full complement of species is present and safe.

In contrast to area, neither isolation nor age showed any obvious effect on proportional richness, and neither obviously changed the relationship with area. However, distances appear to be reported mainly for fragments close to a main source, meaning that too little variation exists in the data to allow detection of the effect of isolation. Also, some of the fragments were demonstrably not isolated, because individuals of several of the species in Africa, the continent from which most of the isolation data come, are known to have moved between patches and the main forest (Tutin 1999; Onderdonk & Chapman 2000). We do not know of a simple reason why no effect of age of fragment was detected.

tiny fragments

This global analysis confirms the point made previously for the Americas that most forest fragments are extremely small, probably far too small to support any primates in the long term (Chiarello 2000; Cowlishaw & Dunbar 2000; Peres 2001). Thus Peres (2001) calculated about 40 to several hundred km2 as the minimum viable forest area for 10 hunted mammal species in the Amazon, including four primate genera. Similarly, Struhsaker's (1981) analysis of diversity of diurnal primates in forest blocks in east Africa indicates that it is highly unlikely that populations resident in fragments of less than several tens of square kilometres are viable in the long term: of the 11 diurnal species in the region, only five persisted in blocks of less than 100 km2, and only one in less than 50 km2 (in 25 km2). The median fragment area in our sample was just 1 km2.

Primates are not studied in a subset of unusually small fragments. For instance, Peres's (2001) sample of more than 5000 isolated Amazon forest fragments detected in Landsat images indicated that more than 99% are less than 1 km2. Chatelain, Gautier & Spichiger (1996) found that more than 60% of a sample of Africa's Ivory Coast fragments were only 1 ha or smaller, and more than 90% were 10 ha or smaller. Added to the direct threat of the small size of these fragments on the species within them must be the associated effects of small size, such as edge effects and more adverse matrix, which make the fragments effectively smaller than they are (Cowlishaw & Dunbar 2000; Harcourt, Parks & Woodroffe 2001; Laurance & Cochrane 2001; Peres 2001; Parks & Harcourt 2002).

continental differences

Africa stands out as showing no obvious, overall significant effect of fragment area on richness or proportional richness. One reason concerns the suggestion that extant taxa in general are more resistant to extinction than in the past, because only the resistant ones have survived the past changes (Balmford 1996; MacPhee & Marx 1997; McKinney 1997). As Africa has been exposed to the influence of humans for longer than the other continents, particularly South America and Madagascar, it might be the case that African primates are the most resistant (Martin 1984; Channell & Lomolino 2000). However, problems exist with the idea of evolution of resistance to extinction, for instance that the environmental changes experienced in the past are not the same as current ones (McShea 1998). Also, there are probably complex interactions between the past environmental changes that taxa have survived and the scale and intensity of human influence (Jansson & Dynesius 2002).

The other explanation for Africa's statistically flat species–area slope is sampling artefact. Africa had smaller fragments on average than any of the other continents, which perhaps provided too little range of size to detect an obvious species–area relationship (Onderdonk & Chapman 2000). Also, isolation was incomplete in one of the three study sites that provided most of the data (Onderdonk & Chapman 2000).


We have a robust finding that across most of the globe, the number of primate species declines linearly with area of fragment: there is no fragment area above which species are safe (fragments defined as ≤ 100 km2). If the continents differed in the severity of the decline, it might make sense to concentrate conservation effort on continents (and other regions) where decline is steepest. However, the number of fragmentation studies from continents other than South America is so small that before any firm decisions on priority are made, research effort needs to be more equalized across the continents.

Most forest fragments in which research is currently conducted are an order of magnitude too small to save most primate species in the long term, or indeed to save any other large-bodied mammal of equally poor dispersal ability (Brooks & Balmford 1996; Laurance & Bierregaard 1997; Soulé & Sanjayan 1998; Cowlishaw 1999; Chiarello 2000; Peres 2001; Harcourt 2002; Ferraz et al. 2003). One reason for fragment research is to investigate the biology of extinction by making use of contrasts between species in their susceptibility to habitat loss (Brown 1971; Harcourt & Schwartz 2001). Another reason is to use the research programme as a basis for conservation of the species under study. However, if most fragments in which research on primates is occurring are too small to contain any primate species permanently, then not only does no contrast exist on which to base a study of the biology of differential risk, but the conservation efforts will necessarily fail in the long run. Both to elucidate the biology of contrasts between species in susceptibility to fragmentation and in order to use research sites for associated conservation efforts, it might be better to direct more attention to fragments of a size in which long-term persistence of some species is possible.


We thank Shasta Markos for some of the analysis, Mark Grote for expert statistical advice, and Colin Chapman, Harriet Eeley, Jörg Ganzhorn, Mike Lawes, Laura Marsh and two referees for generously detailed commentary that considerably improved the paper.

Supplementary material

The following supplementary material is available for this article online.

Appendix S1. Details of sites and fragments identified for analysis.