Anna K. Lindholm, Department of Zoology, University of British Columbia, Vancouver, BC V6T 1Z4, Canada.
1. Reed warblers are not known to be parasitized by cuckoos in Wales, and are rarely parasitized in some English populations, despite the presence of cuckoos. Nine hypotheses are proposed to explain regional differences in parasitism frequencies. They are tested against national and international data from the BTO, RSPB and EURING, and data from four study sites.
2. The frequency of parasitism of reed warbler populations was found to be negatively correlated with habitat patch size. Thus the frequency of parasitism experienced by reed warblers and the number of cuckoos parasitizing a reed warbler population do not increase in direct proportion to host abundance.
3. Population sizes of two unparasitized (Llangorse Lake and Oxwich) and one rarely parasitized population of reed warblers (Pannel Valley) are neither too large nor too small to be suitable for parasitism by cuckoos.
4. Cuckoos are not excluded from these populations by lack of perches in the habitat, because of a lack of prey in the immediate area, or by territorial competition with cuckoos of other host preferences.
5. There has been no decline in frequency of parasitism of reed warblers in Britain since 1932. Thus, reed warbler cuckoos are probably not declining in numbers and contracting their range away from these three geographically marginal sites.
6. Cuckoos parasitizing reed warbler populations at Llangorse Lake, Oxwich and Pannel Valley would achieve reproductive success equal to or higher than that attained with other local hosts, so that any host switches away from reed warblers would be unprofitable.
7. Reed warbler cuckoos may have dispersed to all or most of these study populations in the past. However, the cuckoo populations would have been small and highly vulnerable to extinction. Cycles of increases, declines and extinctions of local populations could explain the variance in parasitism frequencies and number of cuckoos often observed.
8. The most important factors considered in this study that affect the presence and number of cuckoos at a site appear to be competition between females, habitat patch size and structure, rates of dispersal and demographic stochasticity.
In recent decades, metapopulation theory has been used to explain the absence of a particular species from patches of suitable habitat (Levins 1970; Hanski 1991). If rates of colonization of particular patches are exceeded by rates of extinction, the patches will be occupied only sporadically. This paper examines the absence of brood parasitism by the cuckoo (Cuculus canorus Linnaeus) at some populations of a major host species in Britain, the reed warbler [Acrocephalus scirpaceus (Hermann)], in the light of local ecological conditions and in the context of metapopulation theory.
Individual female cuckoos are host-specific (Chance 1940; Nakamura & Miyazawa 1997). They lay their eggs in nests of passerine birds and then leave all parental care to the nest owners (hosts). The basis of host fidelity is not known, but is believed to involve the young cuckoo imprinting on the host that reared it, or on the habitat it experienced immediately after fledging (de Brooke & Davies 1991; Moksnes & Røskaft 1995).
Female cuckoos lay distinct egg morphs which mimic the appearance of eggs of common host species (Davies & de Brooke 1988). Each female is believed to lay eggs of similar appearance throughout her life (Chance 1940; Wyllie 1981). Since egg colour is genetically determined in domestic fowl [Gallus gallus (Linnaeus)] and in village weavers [Ploceus cucullatus (Müller)] (Punnett 1933a; Collias 1993), cuckoos laying different egg morphs presumably differ genetically. Although in domestic fowl an autosomal locus for egg colour has been found (Stevens 1991), in cuckoos genes for egg colour and patterning may be located on the female-specific W chromosome (Punnett 1933b; Jensen 1966). Such sex-linkage would allow the preservation down the matriline of favourable combinations of genes for egg mimicry, regardless of the male genotype.
If cuckoos return to lay eggs in nests of the host species from which they hatched, host egg mimicry could evolve in different cuckoo matrilines. However, analyses of microsatellite variation in mitochondrial genomes of cuckoo chicks reared by different host species have, as yet, provided no evidence for host-specific female races. The failure to detect differences in microsatellite markers may reflect the rapid evolution of different egg morphs (Gibbs, Brooke & Davies 1996).
As different host species are found in different habitats, female cuckoos of different host preferences are likely to occur as metapopulations. This is particularly true of those cuckoos parasitizing reed warblers, the preferred host in reedbeds (de Brooke & Davies 1987), since reedbeds in Britain occur in discrete fragmented patches. This habitat ‘patchiness’ for reed warbler cuckoos suggests that their populations will fluctuate according to the metapopulation paradigm. Some suitable habitat patches, particularly isolated ones, are likely to be uninhabited by cuckoos. Demographic stochasticity and constraints in patterns of dispersal may explain the absence of parasitism from isolated populations, but other ecological or evolutionary factors may be relevant. Nine nonmutually exclusive hypotheses for local absences of parasitism of reed warbler populations in Britain are examined in detail.
1. Inadequate sampling led to a failure to detect parasitism.
2. Cuckoos may be avoiding the sites because of competition for territories or because the sites lack suitable foraging areas. Cuckoos have a specialized diet, rich in hairy caterpillars (Wyllie 1981).
3. The host population sizes are too small to attract cuckoos. Lack (1963) hypothesized that cuckoos would use reed warblers as hosts ‘only in areas where [the reed warbler] is numerous, and not in areas where it is scarce.’
4. The sites are unsuitable for cuckoos because the habitat patch size is too large and/or lacks perches. Habitat fragmentation, which leads to a higher proportion of habitat area near edges, has been invoked as a factor promoting brood parasitism in both cuckoos and brown-headed cowbirds [Molothrus ater (Boddaert)] (Paton 1994; Robinson et al. 1995; Schulze-Hagen, Pleines & Sennert 1996). Cuckoos are believed to locate host nests by watching from vantage points (Chance 1940; Øien et al. 1996).
5. Currently unparasitized populations of reed warblers may have evolved such high frequencies of egg rejection through heavy parasitism in the past that reed warbler cuckoos went extinct at these sites (Takasu et al. 1993). As costs to egg rejection in the absence of parasitism in reed warblers appear to be small (Davies & Brooke 1988), the rate of loss of genes for rejection is expected to be slow (Davies & Brooke 1989b; Takasu et al. 1993).
6. Unparasitized populations have higher quality alternative hosts nearby, so that all local cuckoos turned to other host species at those sites.
7. Use of reed warblers as hosts is declining. If parasitism of reed warblers is declining overall, then isolated populations towards the edge of the species’ range might be the first ones to become free of parasitism.
8. Reed warbler cuckoos have not yet dispersed to areas with unparasitized populations of reed warblers.
9. Extinction rates of isolated populations of cuckoos are higher than colonization rates. Currently unparasitized populations of reed warblers may have been parasitized in the past but subsequently freed from parasitism by the extinction of the local population of reed warbler cuckoos.
Each of these hypotheses is tested using data either collected directly from the field, provided by databanks, or available in the literature. Each will be discussed in sequence, and data and conclusions from earlier hypotheses will be incorporated into models derived from later hypotheses.
Methods and field sites
Detailed observations were made at four sites (Fig. 1).
1. Llangorse Lake, in Breconshire, Powys, Wales, is a large (153 ha), shallow, eutrophic lake of natural origin fringed by 7 ha of reedbed [Phragmites australis (Cav.) Trin. ex Steud.] in a 2–70 m band along 3 km of shore (Tyler 1983; Francis 1992). Alder [Alnus glutinosa (L.) Gaertner] woodland and/or sedge (Carex spp.) meadows form the inland border of the reedbed. About 200 pairs of reed warblers nest at the lake. Fieldwork was carried out on 135 days in May to July of 1993 and 1994.
2. Oxwich National Nature Reserve, in the Gower Peninsula, Glamorgan, Wales, encompasses 38 ha of Phragmites reedbed (Painter, Smith & Gilbert 1995) lined by woodland. The reed warbler population numbers ≈225 pairs (1993 estimate by D. Thomas, personal communication). The site was visited on 9 days in June and July 1993.
3. Pannel Valley Nature Reserve at Rye Bay in East Sussex, England, contains 100–150 pairs of reed warblers in ≈13 ha of Phragmites and mixed reedbed in 45 ha of grassland, flooded fields and dykes in a river valley bottom. Fieldwork was conducted on 31 days in June and July 1995.
4. Wicken Fen Nature Reserve, Cambridgeshire, England, sustains 45 ha of Phragmites reedbed (Painter et al. 1995) in swamp and dykes and 200–600 pairs of reed warblers (1995 estimates by N.B. Davies and P. Bircham, personal communications). The site was only visited twice in 1993, but reed warblers have been studied previously in the reserve and along nearby waterways (Green & Davies 1973; Davies & Brooke 1988; Duckworth 1990).
Rates of parasitism of reed warblers by cuckoos were calculated using the nest record card collection at the British Trust for Ornithology (BTO) in Thetford for the period 1932–96. All records of nests containing eggs or young were included. Cuckoo eggs are easy to detect in reed warbler nests because of their large size and are therefore unlikely to be mistaken by any nest recorder.
Simulation models and logistic and Poisson regressions were performed using S-Plus for Windows Version 3·3 (Venables & Ripley 1994). An empirically derived scale parameter was used in each regression to account for overdispersion in the data (Wilson & Grenfell 1997).
In this paper female cuckoos are classified, not into gentes based on egg coloration, but by host preference. Thus any female cuckoo that lays an egg into the nest of a reed warbler is termed a reed warbler cuckoo.
Results and discussion
1. inadequate annual sampling led to a failure to detect parasitism
No direct evidence of parasitism was found at any field site, as no cuckoo eggs, nestlings or fledglings were detected during direct nest searches. The reedbeds at Llangorse Lake and Pannel Valley were extensively searched, therefore reed warblers at these populations are unlikely to have been parasitized during the study period. However, only a small proportion of the reedbed at Oxwich and Wicken Fen was searched. These data were therefore supplemented with evidence from previous field studies, BTO nest record cards and nestling ringing records from the Llangorse, Rye Bay (at Pannel Valley) and Wicken Fen ringing groups (Table 1). Where multiple sources of data were available for a year, only one was used, with priority being given to data from field studies, and then to nest records, to prevent nests from being counted twice. The overall or mean frequency of parasitism was calculated as the total number of parasitized nests divided by the total number of observed nests. An alternative method would be to calculate the mean of the parasitism frequency for each year, but this would introduce the bias of giving equal weight to unequal sample sizes.
Table 1. Annual and mean frequencies of nest parasitism at four reed warbler populations as estimated from field studies (F), BTO nest records (N) and nestling ringing records (R). Ringing data are expressed as the number of cuckoos ringed in reed warbler nests/total number of reed warbler nests from which nestlings were ringed
As a test of whether sample sizes were large enough to conclude that the true mean frequencies of parasitism at Llangorse Lake, Oxwich and Pannel Valley are indeed low, the 95% confidence intervals of the estimates for each population were compared with the overall frequency for Britain. Variation in the observed parasitism frequencies is caused by binomial sampling variability, and possibly also by year to year changes in the underlying mean. Both types of variation were estimated from a 33-year data series from Chew Valley Lake, Avon, England (Figs 1 and 2). A model was fitted with normally distributed annual variation (s) in logit parasitism frequency about a mean (k) (Collett 1991, pp. 205–215), with each year considered as an independent datum, using EGRET (Statistics and Epidemiology Research Corporation, Seattle, USA), giving estimates of k and s of −4·45 and 1·40, respectively. The estimate of annual variation at Chew was applied to models for each population to incorporate a realistic degree of annual variation. For each of Llangorse Lake and Oxwich, k was estimated iteratively as the value for which the probability of observing no cuckoos was 0·05 in simulations of counts of parasitism in a data series of the actual numbers of nests checked each year (Table 1). For example, for Oxwich, a large number of 3-year series were generated, each containing a sequence of 8, 96 and 15 observed nests. The mean parasitism frequency was increased (by increasing k) until only 5% of the simulated series had no parasitized nests in any of the 3 years. Simulations were then performed using the value of s from Chew and the estimated value of k, giving a mean value which is the 95% upper confidence limit. At both sites, the lower limits are the observed frequencies of zero. Upper confidence limits for Pannel Valley and Wicken Fen were obtained in the same way, except that k was increased until only 2·5% of the simulated series had the actual observed counts of parasitism or less. For lower confidence limits, k was decreased until only 2·5% of the simulated series had the observed counts of parasitism or more.
The mean frequency of parasitism of reed warblers in Britain of 6·20% (n = 10 528 reed warbler nest records from 1932 to 1996) is greater than the upper 95% confidence limit for Llangorse Lake and Pannel Valley but falls within the 95% confidence interval for Oxwich and Wicken Fen (Table 1). Thus the true mean frequencies of parasitism at Pannel Valley and Llangorse Lake are low relative to the national mean. However, more data would be needed from Oxwich to permit the same conclusion to be drawn.
As further evidence of the frequency of parasitism of the study populations, accounts by naturalists were used. At Llangorse Lake, descriptions of the bird life include reed warblers, but not cuckoos (Armitage & Ley 1869; Phillips 1882, 1901; Francis 1992). Cuckoos in the county are scarce in valley bottoms, but common in gardens, farmland and moorland, and are only known to parasitize dunnocks [Prunella modularis (Linnaeus)], blackbirds [Turdus merula (Linnaeus)] and meadow pipits [Anthus pratensis (Linnaeus)] (Phillips 1882, 1899; Phillips 1948; Massey 1976; Peers & Shrubb 1990; author's personal observation). At Oxwich marsh, no parasitism of reed warblers is known despite frequent visits and regular mist-netting of reed warblers during the breeding season for the last two decades (D. Thomas, personal communication). Cuckoos are thought to parasitize meadow pipits nesting on nearby sand dunes (Grenfell & Thomas 1982; Bibby & Thomas 1985). At Pannel Valley, no cuckoo eggs or young have been observed within the reserve, despite frequent site visits since 1986 (S. Rumsey, personal communication) and a study of 66 sedge warbler [Acrocephalus schoenobaenus (Linnaeus)] nests in the reserve from 1993 to 1995 (C. Papazoglou, personal communication). One record exists for the surrounding area (Table 1). The main cuckoo host appears to be the dunnock, in adjacent farmland (four records from BTO nest record cards and S. Rumsey, personal communication). Parasitized reed warblers have also been found 20 km away, at Dungeness (Riddiford 1986). Finally, Wicken Fen is situated on remains of the most extensive fenlands in England (Bircham, Rathmell & Jordan 1994). Early this century reed warblers were abundant and cuckoos very common, particularly in the fens (Evans 1904; Lack 1934). The main host of the cuckoo in the county is probably the reed warbler (Bircham 1989). Thus, cuckoos are likely to have had a long history of parasitic interactions with reed warblers in this area.
In summary, the concordance between observed frequencies of parasitism, statistical analyses and historical evidence suggests that sampling error is unlikely to explain the low frequencies of parasitism of reed warblers detected at Llangorse Lake, Oxwich and Pannel Valley. The Wicken Fen population is currently parasitized and may have been parasitized for many centuries. In each of the four counties in which the field sites are found, cuckoos have been described as widespread or abundant (des Forges & Harber 1963; Heathcote, Griffin & Salmon 1967; Peers & Shrubb 1990; Bircham et al. 1994).
2. the competition for territories is too great or there is a lack of prey
Female territory holders try to prevent other female cuckoos from laying eggs in their territories (Chance 1940; Riddiford 1986). However, no cuckoos were resident in the woodland edge of the reedbed of Llangorse Lake (a cuckoo was detected once in 97 fieldwork days in May and June over 2 years), and none was seen in the woodland edge of Oxwich marsh (over 7 days). At Pannel Valley, two or three pairs of cuckoos were resident in woodland on the margins of the valley and were seen on 10/29 fieldwork days in May and June and heard nearly every day. As no parasitized reed warbler nests were found, the female cuckoos were apparently not reed warbler cuckoos. Therefore, intraspecific competition for breeding territories might prevent reed warbler cuckoos from breeding at Pannel Valley, but probably is not a factor at Oxwich and Llangorse Lake.
The reedbeds in Wales may harbour lower densities of the hairy caterpillars upon which cuckoos feed than elsewhere, as none were observed in the reedbeds of Llangorse Lake and Oxwich. However, radio-telemetry studies have revealed that cuckoos use separate foraging and breeding areas, with a maximum recorded distance between the two from 4 km to 23 km (Wyllie 1981; Dröscher 1988; Nakamura & Miyazawa 1997). As male cuckoos were often heard calling from nearby habitat at both unparasitized sites in Wales (28/97 days at Llangorse Lake, 3/7 days at Oxwich), and the carrying power of cuckoo song is 1–1·5 km in good conditions (Wyllie 1981), it seems unlikely that any nest parasitism of reed warblers would be constrained by distance to foraging sites. At Pannel Valley, cuckoos frequently foraged near the reedbed on the caterpillar of the drinker moth [Euthrix potatoria (Linnaeus)], which was abundant. Thus, lack of prey is unlikely to explain infrequent parasitism at Pannel Valley, and there is no evidence that differences in prey abundance or quality contribute to the absence of cuckoos from other sites.
3. the host populations are too small to attract cuckoos
Are the unparasitized populations too small to support cuckoo breeding territories?
Female cuckoos have parasitized reed warblers in sites that held less than 20 nests (Wyllie 1981; Dröscher 1988). In 6 years of study at a large site in Cambridgeshire, three, six and eight cuckoos laid at a site that contained 100–200 nests (Wyllie 1981). Thus, insufficient host nests cannot explain the absence of cuckoos from the large populations of reed warblers at Llangorse Lake, Oxwich and Pannel Valley.
Are reed warblers parasitized more heavily where they are more abundant?
As a test of Lack's (1963) hypothesis that reed warblers are parasitized more heavily where they are numerous, parasitism frequency was regressed on an index of abundance of reed warblers, for 100-km squares of the National Grid in mainland Britain. The parasitism frequency per grid square was determined from BTO nest record cards. The measure of abundance per grid square was the mean frequency of occurrence (proportion of visited 2 × 2 km tetrads in a 10-km square in which breeding birds were found) in the nationwide census of breeding birds from 1988 to 1991 (Gibbons, Reid & Chapman 1993). Frequency of parasitism was not significantly related to mean host abundance (Poisson regression: t = −0·65, d.f. = 21, P = 0·52) and was only marginally related to total abundance, measured as the number of tetrads occupied by reed warblers per grid square (Poisson regression: t = 1·92, d.f. = 21, P = 0·07). Thus no clear support was found for Lack's hypothesis.
The analysis should not be seriously biased by differences in the efforts of nest recorders across Britain, as the number of reported nest records is significantly positively correlated with the mean frequency index (Spearman rank correlation: rs = 0·45, n = 28, P = 0·02) and the total number of tetrads occupied by reed warblers (rs = 0·88, n = 28, P < 0·001), over National Grid squares.
4. habitat patch size may be unsuitable and/or perches are lacking
Reedbed area and frequency of parasitism
The relationship between habitat patch size and parasitism frequencies of reed warblers in Britain was examined (Fig. 3). The areas of reedbeds were obtained from the most recent inventory of British reedbeds (Painter et al. 1995) and log transformed. Reedbeds from which 50 or more reed warbler nest records were available were included (total n = 4884 nests). Reedbed area was significantly negatively related to odds of parasitism (t = −3·46, d.f. = 18, P < 0·005). For every tenfold increase in reedbed area, the odds of parasitism are likely to increase by a factor of 0·056 (95% confidence interval: 0·010–0·321), or equivalently, to drop by nearly 95%.
Because of the greater area to edge ratio, large reedbeds are likely to contain a greater proportion of nests located far from an edge than small reedbeds. Consequently, cuckoos may be less able to find host nests in large reedbeds, resulting in lower parasitism frequencies. In the Czech Republic, reed warblers prefer to nest far from wooded edges, where parasitism frequencies are lower (Øien et al. 1996).
At Llangorse Lake, Oxwich, Pannel Valley and Wicken Fen, trees are found near the edge of much of the reedbed, and occasionally occur within it. From measurements at Llangorse Lake, the mean distance (±se) from 30 randomly selected nests to nearest perch (tree branch) which would provide a clear view from above the reeds was 10·2 ± 1·5 m, only slightly more than the mean distance from trees to nests of parasitized reed warblers at ponds in the Czech Republic, 8·7 m (Øien et al. 1996). Thus the reedbed at Llangorse Lake does not lack adequate perch sites. If distance from the woodland edge limits the number of nests that cuckoos can parasitize, then a gradient of parasitism rather than none at all would be predicted from the woodland edge to the reedbed interior of Oxwich, Llangorse Lake and Pannel Valley. This scenario was found in ponds with extensive reedbeds in Poland, where parasitism by cuckoos accounted for 0–5% of nest failures of reed warblers (Dyrcz 1981). Parasitized nests were recorded only near bushes or trees.
5. cuckoos would have poor breeding success because high frequencies of rejection of cuckoo eggs evolved through past interactions with reed warblers
As mimetic model eggs are very rarely rejected by reed warblers at any site (Davies & Brooke 1988; Lindholm & Thomas 1999), frequencies of rejection of non-mimetic model cuckoo eggs (see Davies & Brooke 1988 for methodology) were used to compare current frequencies of egg rejection at field sites. Rejection frequencies were significantly higher at Wicken Fen than at the other sites (vs. Pannel Valley, Llangorse Lake & Oxwich, Lindholm & Thomas 1999). Reed warblers of the unparasitized and rarely parasitized populations therefore did not demonstrate as highly evolved defences against cuckoo parasitism as those in the frequently parasitized population. Thus cuckoos parasitizing reed warblers at sites other than Wicken Fen would be less rather than more likely to suffer egg rejection. The possibility that host defences in the past were different and thereby influenced the distribution of cuckoos is unlikely as long as the costs of maintaining egg rejection in the absence of parasitism are small, in which case the rate of loss of genes for rejection would be expected to be slow (Davies & Brooke 1988; Takasu et al. 1993).
6. sites where reed warbler populations were not parasitized had higher quality alternative hosts nearby, so all local cuckoos turned to other host species at those sites
Reed warblers are good hosts for cuckoos, giving higher hatching and fledging success of cuckoo eggs than alternative hosts such as meadow pipits, dunnocks and robins [Erithacus rubecula (Linnaeus)] (de Brooke & Davies 1987). Reed warblers nest semicolonially, at higher densities than meadow pipits, pied wagtails (Motacilla alba Linnaeus), robins and dunnocks (Sharrock 1976). Moreover, nests of reed warblers are easily located by sight by humans, and therefore may also be easier for visually searching cuckoos to locate than nests of alternative hosts.
Frequencies of rejection of unlike model eggs at Pannel Valley, Llangorse Lake and Oxwich are equal to or lower than rejection frequencies of all other main host species of the cuckoo in Britain in areas where they have been experimentally parasitized (Table 2). A cuckoo which switched hosts from any other species to reed warblers at Llangorse Lake, Oxwich or Pannel Valley is likely to experience equal or higher reproductive success. If cuckoos did readily switch between hosts and habitats, then reed warblers should be much more widely parasitized than they are, given their locally high densities and quality as hosts, measured by acceptance of cuckoo eggs and ability to rear cuckoo young (de Brooke & Davies 1987).
Table 2. Comparison of the frequency of rejection of unlike model cuckoo eggs in the reed warbler at Llangorse Lake (11/82)* and Pannel Valley (15/48)* with other main host species of the cuckoo in Britain (de Brooke & Davies 1987)
7. the use of reed warblers by cuckoos is declining
The annual rate of parasitism of reed warblers in Britain (Fig. 4) shows no significant change over time (logistic regression of odds of parasitism on year: t = 1·24, d.f. = 54, P = 0·22). This result contradicts that of de Brooke & Davies 1987), who reported an increasing rate of parasitism of reed warblers since 1932. Their analysis relied upon Lack's (1963) incorrect published analysis of BTO nest record cards. Over the period 1932–61, Lack reported 11 parasitized reed warbler nests; 29 records now exist. As the abundance of reed warblers in Britain is stable (Gibbons et al. 1993), there is no evidence of a decline in numbers of reed warbler cuckoos, which suggests that their range is not contracting (Lawton 1993), and thereby freeing populations of reed warblers from parasitism.
8. reed warbler cuckoos have not yet dispersed to the unparasitized sites
Having considered alternative hypotheses, dispersal distances will now be estimated and models of extinction used to investigate the possibility that low dispersal rates of cuckoos and high extinction probabilities of cuckoo populations explain the patterns of parasitism of reed warblers in Britain. A major assumption is that cuckoos parasitize only one host species throughout their lives. Therefore the evidence for host fidelity is first examined.
Does host-switching occur?
Studies of female cuckoos by radio-telemetry have shown host species fidelity in 26/27 females within a season, with only one record of an egg (probably) laid in a nest of a second host species (Wyllie 1981; Dröscher 1988; Nakamura & Miyazawa 1997)*. Host-switching is therefore rare. Previous studies, based on indirect evidence, have suggested that host-switching might be more frequent, given records of parasitism of hosts which do not successfully rear cuckoo young (Glue & Morgan 1972; Moksnes & Røskaft 1995), records of parasitism of reed warblers by cuckoos laying non-mimetic eggs (Moksnes et al. 1993), recent changes in host preference of cuckoos (Nakamura 1990) and the frequent occurrence in museum collections of clutches of host eggs with poorly matching cuckoo eggs (Moksnes & Røskaft 1995). However, once a switch in host preference has been made, cuckoos may lay non-mimetic eggs for many generations, depending on the strength of selection pressure exerted by host rejection of odd eggs. Thus the frequency of detection of non-mimetic eggs may not reflect the frequency of host-switching events. In Japan, two main hosts of the cuckoo (azure-winged magpies [Cyanopica cyana (Pallas)] and great reed warblers [Acrocephalus arundinaceus (Linnaeus)]) are parasitized by non-mimetic eggs (Matsuda & Uchida 1990; Nakamura 1990; Lotem, Nakamura & Zahavi 1995), despite strict host fidelity (Nakamura & Miyazawa 1997). Thus, cuckoos that parasitize reed warblers would be expected to maintain a preference for reed warblers throughout their lifetimes.
Are reed warbler cuckoos found near the unparasitized populations?
Despite the absence of parasitism of reed warblers, cuckoos were present in the vicinity of the unparasitized populations (see hypothesis 2). As no parasitism of reed warbler nests has been recorded in Wales (Fig. 5), these cuckoos were unlikely to have been reed warbler cuckoos. The host preferences of cuckoos in Wales are meadow pipits, tree pipits [Anthus trivialis (Linnaeus)], pied wagtails, grey wagtails (M. cinerea Tunstall) and, most frequently, dunnocks.
Distribution of reed warbler cuckoos
If female cuckoos with different host preferences exist as distinct lineages, then the map of locations of parasitized reed warbler nests (Fig. 6) may be considered as a distribution map of reed warbler cuckoos. Reed warbler cuckoos appear to be disproportionately represented in south central England, even in areas where reed warblers are not particularly abundant (Fig. 6 cf. Gibbons et al. 1993, p. 334). The observed distribution could be explained by natal philopatry in cuckoos, leading to the perpetuation of parasitism in particular regions. However, the available data may be insufficient to assess the distribution of cuckoos along the East Anglian coast, as most of the coastal East Anglian reedbeds are now protected as nature reserves which prohibit nest searches (G. Tyler, personal communication).
To estimate natal dispersal in the cuckoo, recapture and recovery data for cuckoos ringed as nestlings were obtained from the European Union for Bird Ringing (EURING) for 13 European ringing schemes. The risk of considering recoveries of cuckoos still on migration as breeding recoveries was minimized by only including records for May and June for cuckoos in their third calendar year or older, but only June for yearling (in their second calendar year) cuckoos. Yearlings may still be migrating in May (Seel 1977). Most adult cuckoos leave their breeding sites in July (Wyllie 1981). Of 16 recoveries of nestlings during breeding months of subsequent years, 19% were recovered at their natal site (distance = 0 km) and 75% were recovered within 40 km of their natal site (Fig. 7). One nestling was recovered as a third-year adult in early May 368 km south (direction 179°) of its natal site, which leaves open the possibility that it was still on migration. Hence there is little evidence for extensive natal dispersal in cuckoos, although the data set is small.
The evidence for breeding dispersal in cuckoos was also examined. Using the criteria of ringing and recovery in May and June, and excluding two birds which were found dead more than 1000 km to the south, 25% (n = 12) of adults returned to their ringing sites, while a total of 50% were recovered within 20 km of the ringing site. Breeding philopatry has also been demonstrated in long-term behavioural studies of individual female cuckoos (Chance 1940; Blaise 1965; Wyllie 1981). No evidence was found of sex differences in breeding dispersal distances between four females and three males (Mann–Whitney U-test, U = 4, U′ = 8, P = 0·48).
Dispersal to Wales
As isolation increases, the probability of colonization decreases (Diamond, Gilpin & Mayr 1976; Harrison, Murphy & Ehrlich 1988). Thus the dispersal of a reed warbler cuckoo from England to a site in Wales may be a rare event. However, the distribution of observed natal dispersal distances suggests that cuckoos occasionally disperse 50–100 km (Fig. 7), which is within the distance between Llangorse Lake or Oxwich and nearest parasitized reed warbler nests (Fig. 6). Moreover, small populations of reed warblers along the south coast of Wales (Gibbons et al. 1993) could act as ‘stepping stones’ for cuckoo colonization from England (MacArthur & Wilson 1967). Also relevant are the ages of the reed warbler populations. Reed warblers were already found at the lake 130 years ago (Armitage & Ley 1869), but they may have colonized much earlier. Reeds were found at the base of the artificial island which was constructed in the lake in around ad 890 (Dumbleton 1870; Redknap & Lane 1994), suggesting the presence of breeding habitat for reed warblers about 1100 years ago. Thus despite a presumably small annual probability of colonization, reed warblers may have been breeding at the lake long enough to render probable their discovery by reed warbler cuckoos. The Oxwich population is younger, as reed warblers first bred at Oxwich in the mid-1920s (Paterson et al. 1898; Paterson et al. 1925), and a large population of reed warblers was established by 1968 (Woods 1969). However, reed warblers have been breeding at other sites in the county since at least 1802, although population sizes were generally small (Heathcote et al. 1967; Hurford & Lansdown 1995). Thus the degree of isolation of sites and the ages of the reed warbler populations appear an inadequate explanation for the apparent complete absence of reed warbler cuckoos from Wales.
In Ireland, reed warblers are also unparasitized, although other Irish species are used as cuckoo hosts. Reed warblers are recent colonists, having been first observed in the 1980s (Sealy, O’Halloran & Smiddy 1997). As reed warbler cuckoos would be expected to disperse from England as far as Ireland very rarely, limits to dispersal can explain the current absence of parasitism of reed warblers. Irish reed warblers are likely to remain free from parasitism for some time.
9. cuckoo populations have short persistence times in isolated areas
Once a female cuckoo disperses to an unparasitized patch of hosts, such as at Llangorse Lake, how likely is the extinction of the population she founds?
Isolated patches of habitat can be considered islands (MacArthur 1972). For cuckoos, patchily distributed host populations represent islands. Using the demographic parameters of average recruitment rate b, and average rate of death d, for cuckoos, and an estimated equilibrium population size K, the probability of extinction of a population founded by one female at Llangorse Lake is estimated below by using both deterministic and stochastic models. It is assumed that she will mate with a male cuckoo from nearby farmland, moorland or gardens, although it is not known whether cuckoos mate assortatively according to species of foster-parent or habitat type in which they were reared*. Fidelity of host use down the matriline, a constant host population size, constant environmental conditions and no further immigration are also assumed.
As this study is only interested in persistence of reed warbler cuckoos, a population of one female cuckoo is considered extant, but one male cuckoo as extinct, because he can mate only with females of other host preferences and his offspring will not be reed warbler cuckoos. Therefore, only the size of the female population will be considered.
The average number of female recruits b produced by a female cuckoo at Llangorse per year is estimated as b = (1/2)ne (1 − j) hfsr, where: ne is the average number of eggs laid per female per breeding season; j is the frequency of rejection of reed warbler cuckoo eggs by reed warblers; h is the hatching success of cuckoo eggs in reed warbler nests; f is the fledging success of cuckoo nestlings from hatch in reed warbler nests; s is the survivorship of a fledgling until breeding age; and r is a natal philopatry index, equal to one minus the probability of dispersal.
Three values for ne were considered. Using his own data and previous studies of cuckoos parasitizing different species, Wyllie (1981) found that mean ne = 9·2. The mean value from studies of parasitism of reed warblers only is 5·8 ± 0·37 using data from von Warncke & Wittenberg (1958), Marbot (1959), Blaise (1965), Gehringer (1979), Wyllie (1981) and Davies & Brooke (1988). Both the present study and that of Wyllie excluded records of cuckoos laying only one or two eggs. Blaise (1965) estimated, however, that only 2% of cuckoo eggs were laid in a reedbed other than the one that a female usually parasitized, suggesting that cuckoos rarely lay at more than one site. Therefore, records of one or two eggs should be included in the totals, yielding ne = 4·1 ± 0·34 for reed warbler cuckoos. Records of few eggs laid could be attributed to first-year cuckoos, which arrive back on breeding grounds later than older birds (Seel 1977), and possibly lay few eggs in their first year. A second possibility is that cuckoos which are competitively excluded from breeding territories lay few eggs. The rejection frequency of reed warbler cuckoo eggs j = 0 (Lindholm & Thomas 1999; see also Davies & Brooke 1988) and hf = 0·319 (de Brooke & Davies 1987). The present study calculated, from EURING ringing recovery records, that 66 of 120 ringed nestlings were recovered dead in their first calendar year, giving a maximum rate of survivorship s of nestlings from ringing age through to the end of the first calendar year of 0·45. First-year cuckoos are assumed to return to Llangorse Lake with a probability r of 0·75, as this is the proportion of recovered nestlings which returned to within 40 km of their natal site.
The rate of adult mortality d in cuckoos was estimated by life table analysis of ringing recovery data of nestlings (Table 3; Haldane 1955). As subadult birds are likely to have higher mortality than adult birds, mortality data from the first calendar year of life were discarded. The rate of annual mortality was calculated as the number of recoveries divided by the number of recovery years (Table 3; d = 38/105 = 0·362). This analysis relies on a constant rate of annual mortality in cuckoos. The present study tested whether mortality is age-dependent in adults by comparing the difference between the expected number of recoveries based on an annual rate of mortality of 0·362 and the observed number of recoveries. The hypothesis of age-independent mortality was not rejected (χ28 = 6·60, P > 0·5), despite the inflation of χ2 values which results from the use of small frequencies (Zar 1984). Other assumptions of this analysis are a constant annual reporting rate and no ring loss (Anderson & Burnham 1987). Although life table analysis has been criticized as an oversimplified model of mortality, there is agreement that it can provide an estimate of the average annual rate of adult mortality (Anderson & Burnham 1987; Baillie & Green 1987; North 1987; Peach, Thompson & Coulson 1994).
Table 3. Life table analysis of cuckoo ringing recoveries. (Ringing data were supplied by EURING)
The average rate of recruitment of female offspring per year (b) multiplied by the average life-expectancy of adult cuckoos (d−1 = 2·76 years) provides an estimate of the reproductive rate R0, the average number of female offspring which survive to maturity produced per adult female parasite in her lifetime (Renshaw 1991; Grenfell & Dobson 1995). At recruitment rates resulting from ne = 4·1 and 5·8, R0 < 1 (Fig. 8). On average, females laying mean clutch sizes of less than 6·7 eggs per year will not replace themselves in the population, which will be doomed to extinction deterministically unless there is continued immigration.
This result may suggest that the lower estimates of ne are too low. However, a calculation of the value of ne required to balance mortality suggests that these estimates are reasonable. At equilibrium, R0 = 1. Assuming that reed warbler cuckoo numbers are currently stable (see hypothesis 7), recruitment rates should balance mortality rates. The mean number of eggs that a female reed warbler cuckoo must lay annually to replace herself can then be estimated with the equation 1/2ne (1 − j) hfs = d. Taking the frequency of rejection j as 0, as above, or as 0·19 as at the Wicken Fen population of reed warblers (Davies & Brooke 1988) and solving, ne = 5·0 and 6·2. These relatively small clutch sizes produce recruitment rates which balance mortality because natal dispersal is no longer relevant.
A population founded by a female laying an average of 9·2 eggs per year will not go extinct deterministically but may still go extinct stochastically. This is especially likely if the equilibrium population size K is small (MacArthur & Wilson 1967; Toft & Schoener 1983; Pimm, Jones & Diamond 1988). Therefore the next question is how many breeding female cuckoos the reed warblers of Llangorse Lake would support.
For cuckoos, the limiting resource is likely to be the availability of host nests. Data from the following studies, which had a common goal of finding all the cuckoo eggs laid at the study site, were used to determine the relationship between numbers of breeding cuckoos and host abundance and parasitism frequency: Diesselhorst (1955), von Warncke & Wittenberg 1958), Marbot (1959), Groebbels (1960), Blaise (1965), Gehringer (1979) and Wyllie (1981). The number of female cuckoos that breed in a reedbed increases with host nest abundance, but with declining slope (Fig. 9), so that the rate of parasitism does not increase with increasing number of females for each host patch (Fig. 10; logistic regression of odds of parasitism on number of females: t = 0·86, d.f. = 35, P = 0·40).
At least two factors may account for this relationship. First, female cuckoos appear to compete for host nests, as multiple parasitism of reed warblers occurs, although at low frequencies (2·1%, Blaise 1965; 3·4%, Wyllie 1981; 13·6%, Davies & Brooke 1988; 9·5%, Øien et al. 1996) and predation of cuckoo eggs by rival females is thought to occur (Gehringer 1979; Davies & Brooke 1988). Therefore, at small sites, with a relatively high ratio of cuckoos to reed warbler nests (Fig. 9), more eggs may be wasted through female–female competition, resulting in a lower rate of nest parasitism per female. Second, sites with more host nests are likely to have a larger reedbed area, with a higher proportion of nests that are inaccessible to cuckoos (Fig. 3). Because of the rarity of host switching, and lack of increase in parasitism rate with increasing number of females (Fig. 10), an equilibrium breeding population size of female cuckoos at a given reedbed should exist.
Assuming that the mean number of females N using each host population in Fig. 9 was at equilibrium, the regression equation can be used to predict an equilibrium population size K of female cuckoos at a site, as it provides a good fit to the observed number of females vs. number of host nests available. The sites from which the data were collected were rivers, canals, small lakes and gravel pits, providing smaller rather than larger blocks of reedbed. The habitat characteristics of these sites appear to be comparable to those of the reedbed at Llangorse Lake. At Llangorse, with 200 reed warbler nests available each year, the expected equilibrium cuckoo population size K, is five females, from the regression equation of Fig. 9.
Empirical studies suggest that the survival chances for a population where K = 5 are poor (Toft & Schoener 1983; Pimm et al. 1988; Berger 1990). On small British islands, bird species that had a mean population size of 2·1–5·0 breeding pairs went locally extinct in 3·5 years, on average (Thomas 1990 for data from Pimm et al. 1988). Considering data for cuckoos only, at 7/9 small islands the local cuckoo population went extinct, with a mean time to extinction of 2·4 years (data combined from Lack 1969 and Pimm et al. 1988).
A stochastic model (Renshaw 1991) is employed here to estimate the probabilities of extinction from random mortality for a population arising from a single colonizing female at Llangorse Lake: N(t + 1) = Nt + B − D, where: Nt is the number of adult females in the population at year t;B is the number of females recruiting to the population in year (t + 1); and D is the number of adult females that die between t and (t + 1).
Recruitment of females depends on both the number of eggs laid, and the survival of offspring until the next year. Each female in each year lays a random number of eggs which are female, drawn from a Poisson distribution of λ = ne/2. Each egg has a random probability of recruiting to the breeding population of (1 − j) hfsr = 0·108. Adult females have a random probability of death of 0·362. Simplifying assumptions are that sex ratios of offspring are unbiased, that recruitment and death rates are constant from the second year of life, that adult mortality occurs after the end of the breeding season, and that breeding philopatry is perfect in female cuckoos at population sizes less than or equal to K, which is here considered as a ceiling to population size.
Population extinction within 5 years was the outcome of half of 1000 runs of the model for all three estimates of fertility (Fig. 11). In most simulations, the population size never reached five; this only occurred in 4%, 13% and 37% of runs of 50 years duration, where ne = 4·1, 5·8 and 9·2, respectively.
The extinction of a population of reed warbler cuckoos is predicted deterministically and empirically and by a stochastic model. A population at Llangorse Lake would be likely to persist for only a few years, purely through stochastic demographic effects. If environmental variation were also considered (such as variation in numbers of available host nests or in mortality), then the risk of extinction would be greater (Goodman 1987). Thus the failure to detect cuckoo parasitism of reed warblers at Llangorse Lake could result from rapid turnover of reed warbler cuckoo populations followed by long intervals to subsequent colonizations.
Generalizing from this result, cuckoo populations which are isolated are likely to go extinct before new immigrants appear, leading to periods of occupation and absence. Populations with close neighbours which can provide a source of immigrants are more likely to persist. A possible model of the reed warbler cuckoo metapopulation is shown in Fig. 12. The core area is hypothesized to overlap with areas of ancient reed swamp in East Anglia (Darby 1940; Bircham et al. 1994).
Prediction from the model
Annual variation in numbers of cuckoos and parasitism frequencies has been widely noted (Power 1866; Walpole-Bond 1938; Wyllie 1981; Diamond 1984; Pimm et al. 1988; Fig. 2; Table 1 for Wicken Fen). Such variation may be explained by the model of cycles of colonization and extinction of cuckoos at habitat patches. If a host population at an isolated habitat patch were monitored over many years, the cycles should be apparent in the form of temporal autocorrelation of parasitism frequencies.
The data for annual frequencies of parasitism at Chew Valley Lake (Fig. 2) were used to test for temporal autocorrelation using a sampled randomization test (Efron & Tibshirani 1993). The first-order autocorrelation coefficient (Brillinger 1975) was computed from the time series (n = 33) and from each of 500 bootstrap replications of the time series. The observed autocorrelation coefficient of 0·366 was larger than the upper 2·5% critical value (0·320) of the distribution of coefficients from randomized samples. Thus in the Chew time series, a year in which no parasitism was detected is significantly more likely to be followed by another year without parasitism than by a parasitized year, upholding the prediction from the model. This result suggests that an explicit model of frequencies of cuckoo parasitism must consider not only random sampling variation and spatial variation, but also temporally autocorrelated annual variation.
Ecological factors such as intraspecific competition, habitat patch size, habitat quality, host availability, host quality, and dispersal patterns of cuckoos are likely to affect the distribution of cuckoos in Britain, but cannot account for the absence of parasitism in Wales, its rarity in Sussex and regularity in Cambridgeshire. A cuckoo metapopulation model is able to explain the observed geographical pattern of parasitism of reed warblers. The following empirical data supported the model: (i) reed warblers in Ireland are not parasitized; (ii) small island populations of cuckoos go extinct quickly; and (iii) parasitism frequencies of reed warblers at an isolated habitat patch are temporally autocorrelated. As the model is supported, cuckoos may well have been present in the past at Llangorse Lake and Pannel Valley, and possibly Oxwich, but local extinctions followed one or more times, such that no cuckoos are found at these sites today.
I thank Robert Thomas for assistance in the field. The Brecon Beacons National Park, Oxwich National Nature Reserve, Stephen Rumsey and Wicken Fen Nature Reserve provided access to the field sites. The Llangorse, Chew Valley, Rye Bay and Wicken Fen Ringing Groups kindly provided me with their unpublished data. I thank the BTO for access to reed warbler frequency data and the national nest record card collection, the RSPB for access to national reedbed data and EURING for access to ringing data on cuckoos. I am most grateful to Mike Brooke for helpful discussions and to Neal Alexander and Bärbel Finkenstädt for analytical help. Nick Davies, Mike Brooke, Bryan Grenfell, David Noble, Rhys Green and Chris Thomas provided comments which improved the manuscript. I was supported financially by the Association of Commonwealth Universities.