BENGT KARLSSON and CHRISTER WIKLUND. Department of Zoology, University of Stockholm, S Stockholm, Sweden

Transcription

1 Ecology , Butterfly life history and temperature adaptations; dry Blackwell Publishing, Ltd. open habitats select for increased fecundity and longevity BENGT KARLSSON and CHRISTER WIKLUND Department of Zoology, University of Stockholm, S Stockholm, Sweden Summary 1. Evidence suggests that changes of temperature-related performance curves can trigger a selective response in life-history traits. Hence, it should be expected that insects adapted to different temperature regimes should exhibit optimal performance at the temperature to which they are adapted. 2. To test this idea we investigated how fecundity and longevity are influenced by ambient temperatures in a set of satyrine butterflies adapted to live in dry open landscapes or in closed forest landscapes, respectively, by keeping egg-laying adult females at five different constant temperatures ranging between 20 and 40 C. 3. We studied four species, two of which are confined to dry and hot open habitats, namely the grayling (Hipparchia semele) and the small heath (Coenonympha pamphilus), and two of which are shade dwelling, namely the ringlet (Aphantopus hyperantus) and the speckled wood butterfly (Pararge aegeria). 4. As predicted, the results showed that lifetime fecundity exhibited bell-shaped curves in relation to temperature with the open landscape group peaking at a higher temperature, 30 C, compared with the shade-dwelling group that peaked at 25 C. Longevity decreased with increasing temperatures among all species, but the open landscape living species survived better at higher temperatures. Moreover, although the magnitude of reproductive effort measured as lifetime egg mass did not differ between the two ecological groups, lifetime fecundity did with open landscape species laying more and smaller eggs than the shade-dwelling species. 5. This difference in life-history character traits suggests either that dry and relatively warm open habitats open life-history opportunities in terms of higher fecundity and longevity that remain closed to butterflies adapted to cooler temperatures, or that life in dry open habitats actively selects for higher fecundity and survival as a result of increased offspring mortality. Key-words: fitness, Lepidoptera, life history, Nymphalidae, reproduction, temperature. Ecology (2005) 74, doi: /j x Ecological Society Introduction Butterflies occupy various habitats from the tropics to the Arctic with a wide range of ambient temperatures that affect their ability to cope with the environment. Body temperature is a crucial factor that affects reproductive performance (e.g. Stern & Smith 1960; Gossard & Jones 1977; Kingsolver 1983a,b; Shreeve 1984; Watt 1992), and typical for many organisms is that there is an intermediate single optimum. As the body temperature Correspondence: Bengt Karlsson, Department of Zoology, University of Stockholm, S Stockholm, Sweden. Tel.: ; Fax: ; bengt.karlsson@zoologi.su.se departs from the optimum, performance declines gradually from optimum (Huey & Kingsolver 1989). The shape of the performance curve is typically asymmetric and skewed towards lower temperatures. Although butterflies are ectothermic, the range of operating body temperatures is much more narrow (Heinrich 1993). However, the optimal body temperature, i.e. the temperature that generates the highest reproductive output, might differ depending on differences in life history and on ecological factors (e.g. Chai & Srygley 1990; Srygley & Chai 1990; Dudley 1991; Van Dyck & Matthysen 1998). Karlsson & Van Dyck (unpublished) found that habitat fragmentation affects temperature performance curves in different populations of a satyrid butterfly. The population living in a woodland

2 100 B. Karlsson & C. Wiklund landscape showed higher fecundity under relatively cold temperatures and the population in a more open agricultural landscape did better under relatively warm temperatures. Interestingly there is also evidence that changes of the temperature performance curves can trigger a selective response of life-history traits. Partridge et al. (1995) found in a study on Drosophila melanogaster that females reared under two different temperature regimes for 4 years showed differences in longevity and fecundity. Females of each selection regime showed higher longevity and fecundity than females from the other selection regime when they were tested at the experimental temperature at which they had evolved. In line with the above experiments it would be expected that insects adapted to different habitats should exhibit optimal performance at the environmental conditions to which they are adapted. To test this idea we investigated how fecundity and longevity are influenced by ambient temperature in satyrine butterflies adapted to relatively hot open landscapes and more shaded woodland landscapes, respectively. In total we studied four Swedish satyrine butterflies, two of which are classified as open landscape species, the grayling (Hipparchia semele L.) and the small heath (Coenonympha pamphilus L.), and two of which are classified as woodland species, the ringlet (Aphantopus hyperantus L.) and the speckled wood (Pararge aegeria L.). H. semele is an extreme sunshine lover confined to dry and hot habitats (Henriksen & Kreutzer 1982; Emmet & Heath 1989; Thomas & Lewington 1991). C. pamphilus also occurs in open areas, e.g. fields, along roadsides, mountain slopes, and meadows (Henriksen & Kreutzer 1982). A. hyperantus thrives in damp and sheltered areas, such as open wet forest lands and woodland edges (Henriksen & Kreutzer 1982; Emmet & Heath 1989; Thomas & Lewington 1991). P. aegeria is in Sweden a typical forest butterfly, confined in the north parts in coniferous and mixed-deciduous forests, and in the southernmost parts in deciduous forests (Henriksen & Kreutzer 1982). However, although these butterflies should be expected to perform optimally at the temperatures to which they are adapted, it is feasible that adaptations to warmer temperatures open opportunities that are closed to species living under cooler circumstances. In line with this scenario Wiklund, Karlsson & Forsberg (1987) and Wiklund & Karlsson (1988) suggested that the observation that satyrines living in open landscapes lay smaller eggs and exhibit more pronounced femalebiased sexual size dimorphism, could be seen as adaptations to allow higher fecundity made possible by a thermophilic life-style. Here, we have tested this hypothesis by investigating whether there is a positive interaction between fecundity and temperature so that the satyrines from open landscapes allocate their reproductive expenditure into more and smaller eggs compared with the shade-dwelling, woodland living, satyrines. Materials and methods Eggs of H. semele, C. pamphilus, A. hyperantus, and P. aegeria L. were taken from wild-caught females. H. semele and P. aegeria originated from south Sweden, whereas C. pamphilus and A. hyperantus originated from the Stockholm area. Eclosing larvae were reared in climatic chambers in the laboratory throughout the larval period on their natural food plants. In total, 80 females were reared in the laboratory. In addition, 15 very young females of H. semele, three females of A. hyperantus, and two females of C. pamphilus were caught in the field and brought to the laboratory. Thus, a total of 100 females were used in the experiment (Table 1). The females were weighed on a Sauter AR 1014 balance on the day of capture, or weighed on the day of eclosion in the laboratory. Newly eclosed females were allowed to mate before they were transferred to environmental cabinets. All females were kept individually in 0 5-L plastic jars together with a tuft of the grass Poa annua throughout their lives. The environmental cabinets used had constant temperatures of 20, 25, 30, 35, and 40 C, respectively, and with a 6 : 18, L : D cycle. H. semele had the highest body mass with a mean of mg ± 4 17 SE and C. pamphilus the smallest with a mean of 41 6 mg ± P. aegeria weighed 81 5 mg ± 2 08 and A. hyperantus 66 3 mg ± Female mass for each species did not differ significantly among the temperature regimes: H. semele, F 4,24 = 2 62, P = 0 062; A. hyperantus, F 4,20 = 1 15, P = 0 362; C. pamphilus, F 4,11 = 0 330, P = 0 852; P. aegeria, F 4,25 = 0 889, P = The number of eggs laid was assessed every day and females were allowed to feed on 20% sucrose solution after each day s egg laying. The eggs were weighed on a Cahn 28 automatic electrobalance. All ANCOVAs were done with general linear model procedure of STATISTICA. In these analyses female body mass at eclosion or mass at capture was used as covariate, and species effects nested within landscapes. Results When comparing the first day of egg laying, P. aegeria and A. hyperantus laid the highest number of eggs at 30 C, whereas C. pamphilus and H. semele laid most eggs during the first day of egg laying at 35 C and at 40 C, respectively (Table 1). P. aegeria had the highest lifetime egg production at 25 C. H. semele and A. hyperantus had their maximum at 30 C, whereas C. pamphilus laid the highest number of eggs at 35 C. C. pamphilus and especially H. semele managed to lay eggs at 40 C, a temperature immediately lethal to most females of P. aegeria and A. hyperantus. Temperature also affected number of egg-laying days and life span (Table 1). All species except H. semele showed a decreased life span and a shorter egg-laying period with increasing temperature. Categorizing the investigated species into open living and shade-dwelling (sensu Wiklund et al. 1987) showed that females in these groups reacted differently to

3 101 Butterfly life history and temperature adaptations Table 1. Female mass, lifetime number off eggs laid, lifetime total egg mass, maximum number of eggs laid during 1 day, number off egg-laying days, and life span for all species in all temperature regimes Species Hipparchia semele Coenonympha pamphilus Aphantopus hyperantus Pararge aegeria Temperature ( C) No. of eggs laid day 1 Total no. of eggs Total egg mass (mg) Daily maximum no. of eggs laid Days of egg laying Life span (days) (1 1) 69 8 (46 6) 18 7 (12 0) 17 3 (10 9) 8 7 (4 1) 15 2 (3 8) (10 8) (161 3) 59 3 (46 5) 44 3 (23 7) 10 5 (4 1) 15 0 (5 9) (5 1) (181 5) 77 7 (52 7) 70 7 (23 6) 11 3 (10 0) 14 0 (9 1) (4 2) (88 4) 37 9 (23 8) 46 7 (27 0) 8 2 (4 2) 11 2 (4 4) (19 5) 63 2 (55 2) 16 0 (14 4) 44 0 (43 0) 3 6 (1 1) 6 8 (2 4) 5 F P (0 3) 47 0 (68 4) 10 2 (14 5) 8 7 (9 8) 13 3 (2 5) 21 0 (1 7) (2 9) (80 0) 26 5 (19 5) 25 0 (9 8) 11 0 (1 0) 12 7 (1 5) (3 2) (40 1) 30 8 (8 4) 30 0 (15 4) 12 7 (3 8) 14 3 (4 2) (9 2) (7 6) 40 2 (2 1) 53 0 (7 9) 8 0 (2 0) 10 0 (3 0) (7 1) 17 8 (28 6) 3 9 (6 2) 9 5 (14 2) 1 2 (1 5) 5 2 (2 7) 4 F P (1 3) 92 6 (75 3) 20 6 (16 5) 13 4 (8 3) 11 8 (5 4) 17 8 (4 8) (5 2) (82 6) 30 1 (18 0) 24 8 (14 7) 12 2 (7 5) 13 6 (7 2) (10 4) (122 3) 31 7 (25 5) 35 0 (21 2) 6 4 (3 8) 7 4 (3 6) (6 3) 73 6 (30 0) 17 3 (7 3) 37 2 (18 1) 5 0 (2 9) 6 8 (3 8) (2 2) 2 2 (4 9) 0 6 (1 3) 2 2 (4 9) 0 2 (0 4) 2 8 (1 1) 5 F P (7 2) 91 7 (61 7) 36 7 (23 6) 24 1 (15 6) 11 4 (4 5) 15 7 (4 8) (5 2) (60 2) 56 8 (22 4) 38 9 (12 0) 8 4 (3 8) 11 3 (4 9) (7 2) 88 9 (40 0) 39 5 (17 0) 41 0 (19 9) 7 0 (2 4) 9 6 (3 3) (7 0) 23 7 (14 8) 10 8 (6 9) 21 7 (15 1) 1 3 (0 8) 5 7 (1 6) (0 0) 0 0 (0 0) 0 0 (0 0) 0 0 (0 0) 0 0 (0 0) 1 3 (0 6) 3 F P n Values are mean ± SE. F-values from ANOVA. Fig. 1. Total lifetime egg production measured as total number of eggs in five different temperatures when controlling for female body size. Filled circles (mean ± SE) indicate woodland species and open circles (mean ± SE) indicate open landscape species. ANCOVA (female body mass as covariate), effect of temperature: F 4,87 = 10 86, P < 0 001, effect of landscape: F 1,87 = 4 41, P = 0 038, landscape temperature interaction: F 4,87 = 2 83, P = Fig. 2. Total lifetime egg production measured as total egg mass in five different temperatures when controlling for female body size. Filled circles (mean ± SE) indicate woodland species and open circles (mean ± SE) indicate open landscape species. ANCOVA (female body mass as covariate), effect of temperature: F 4,87 = 10 35, P < 0 001, effect of landscape: F 1,87 = 0 56, NS, landscape temperature interaction: F 4,87 = 2 70, P = temperature treatment, revealed by the significant interaction between temperature and landscape ( Figs 1 3). Open landscape species had a significantly higher lifetime egg production, measured both as total number of eggs (Fig. 1) and total egg mass produced (Fig. 2), at higher temperatures. At low temperatures this relationship was reversed. This pattern was already pronounced when comparing the first day of egg laying (Fig. 3). When comparing the maximum number of eggs and maximum egg mass that was laid during a single day there were differences between the woodland landscapes and open landscape butterflies. Although both groups had a peak in their maximum egg production at 30 C,

4 102 B. Karlsson & C. Wiklund Fig. 3. The number of eggs laid day 1 in five different temperatures when controlling for female body size. Filled circles (mean ± SE) indicate woodland species and open circles (mean ± SE) indicate open landscape species. ANCOVA (female body mass as covariate), effect of temperature: F 4,87 = 4 83, P = , effect of landscape: F 1,87 = 0 76, NS, landscape temperature interaction: F 4,87 = 2 86, P = Fig. 4. Mean egg weight measured as total egg mass divided by the total number of eggs laid. Filled circles (mean ± SE) indicate woodland species and open circles (mean ± SE) indicate open landscape species. ANCOVA (female body mass as covariate), effect of temperature: F 4,79 = 1 39, NS, effect of landscape: F 1,79 = 113 3, P < 0 001, landscape temperature interaction: F 4,78 = 2 07, NS. there was a significant interaction between temperature and landscape (F 4,87 = 2 58, P = 0 043), and open landscape species tended to have a higher maximum fecundity at higher temperatures and woodland species a higher maximum fecundity at lower temperatures. There was no difference in magnitude of reproductive effort, measured as lifetime egg mass, between the open landscape and the woodland groups (Fig. 2, F 1,87 = 0 56, NS). However, there was a significant difference in number of eggs laid (Fig. 1, F 1,87 = 4 41, P = 0 038). This implies that the two groups differ how they allocate their reproductive effort, with open landscape species laying a larger number of smaller eggs compared with the woodland species. This is also reflected when comparing the mean egg weight for each group (Fig. 4). Both open landscape and woodland species showed a decrease in life span with increasing temperature (Fig. 5). Also the number of egg-laying days decreased with increasing temperature (F 4,87 = 16 32, P < 0 001). However, open landscape species had on average a longer life span than shade-dwelling species (Fig. 5), and also the egg-laying period was longer among the open landscape living species (F 1,87 = 5 93, P = 0 017). Discussion The investigated satyrines differed in several respects in relation to temperature. As we predicted, the group of satyrines that were classified as open landscape living by Wiklund et al. (1987), namely C. pamphilus and H. semele had their highest egg production under relatively high temperatures ( Table 1, Figs 1 3), whereas the shade-dwelling species, namely A. hyperantus and P. aegeria peaked their egg production under relatively low temperatures. Moreover, the shade-dwelling species produced very few eggs at high temperatures and most of the females in this group laid no eggs at all at 40 C (Fig. 1). Fig. 5. The relationship between temperature and life span in five different temperatures when controlling for female body size. Filled circles (mean ± SE) indicate woodland species and open circles (mean ± SE) indicate open landscape species. ANCOVA (female body mass as covariate), effect of temperature: F 4,87 = 21 41, P < 0 001, effect of landscape: F 1,87 = 10 09, P = 0 002, landscape temperature interaction: F 4,87 = 0 88, NS. Already on the very first day of egg laying there was a significant difference in how the butterflies were affected by temperature. The open landscape species showed a rapid and strong response to high temperatures and laid significantly more eggs compared with the woodland butterflies during these conditions ( Table 1, Fig. 3). Open landscape living satyrines survived better, and had a longer egg-laying period at high temperatures compared with shade-dwelling satyrines. We interpret these differences in longevity and reproductive period as further support for the idea that these species are better adapted to high temperatures. Although lifetime reproductive success and survival often are appropriate measures of fitness, the maximum egg-laying rate may be a more valuable measure of fitness in temperate areas (Cole 1954). In our study, the pattern of maximum egg-laying rate showed a

5 103 Butterfly life history and temperature adaptations similar pattern as total egg production. The significant interaction between group and temperature showed that open landscape species outperformed shadedwelling species at high temperatures and vice versa. Thus, the investigated species differ in their optimal body temperature for reproduction and longevity and this optimum is correlated to habitat differences between the species. We have demonstrated that, although there is no difference in magnitude of reproductive effort between shade-dwelling and thermophilic species, the satyrines adapted to higher temperatures allocate their reproductive expenditure into a larger number of smaller eggs (Figs 1 4). Why do the open landscape satyrines lay smaller eggs, or, why do shade-dwelling satyrines lay larger eggs? There are two alternative possible explanations. There is an obvious trade-off between number and size of eggs, and eggs should be expected to be so large as to maximize the number and quality of a female s offspring. In butterflies (as in most insects) there is a positive association between egg size and the size of the newly hatched larva, and in butterflies the size of the larval head capsule has been viewed as important with a large head capsule being necessitated in species where the larvae feed on tough food low nitrogen plants (Braby 1994). However, this factor seems unlikely to explain why woodland satyrines lay larger eggs. First, two of the woodland species P. aegeria and A. hyperantus typically feed on fairly softleaved nutrient-rich grasses such as Dactylis glomerata and Poa annua, whereas the two open landscape satyrines typically feed on Festuca ovina, which is considerably coarser. Hence, the idea that the coarseness of the food plant underlies the larger egg size laid by the shadedwelling species receives little support. Secondly, previous tests of size-related fitness advantages associated with large egg size has yielded little evidence that larval performance is strongly coupled to egg size (Wiklund & Persson 1983; Karlsson & Wiklund 1984, 1985). The alternative hypothesis, that smaller egg size has been selected for because of fitness advantage associated with the higher fecundity, could be explained either because the higher temperatures experienced by the open landscape species makes it possible for females to lay more eggs in response to higher bioenergetic capacity associated with adaptations to higher temperatures (cf. Huey & Kingsolver 1989), or because dry open habitats select for higher fecundity by some other mechanism presumably associated with lower larval survival (in turn driven by predation or physical factors). At the present time we have no data to separate between these two hypotheses, and the observation that femalebiased sexual size dimorphism is higher among the two thermophilic species could be equally well explained by both mechanisms. Butterflies can, in spite of variation in ambient temperatures, adjust their body temperature by several behavioural adaptations, e.g. by seeking warm sunny microhabitats, and orientate body and wings to catch sunshine (e.g. Clench 1966; Watt 1968; Kingsolver 1983a; Shreeve 1984; Heinrich 1993). This will make it possible for an individual butterfly to keep a body temperature that is optimal for flight and reproductive performance, and to avoid overheating. Thus, in nature we expect that butterfly females will thermoregulate behaviourally to adjust their body temperature corresponding to the temperatures with the highest egg production observed in the laboratory. However, as such behaviours are in conflict with other demands, e.g. predator avoidance (Shreeve 1992), feeding or egg laying (Kingsolver 1983b), an optimal body temperature (in respect to reproductive success) close to the ambient temperature in the preferred habitat will thus always be of adaptive value for the individual butterfly. As the current study is of a comparative nature, shared ancestry can be a confounding effect (Harvey & Pagel 1991), and ideally we should reduce this effect by comparing species on a phylogeny based context. However, this is difficult as the position of A. hyperantus in the current phylogeny of satyrid butterflies is uncertain (Martin, Gilles & Descimon 2000). We can say, however, that the investigated species in each landscape are not monophyletic groups, i.e. H. semele is not grouped with C. pamphilus and P. aegeria is not grouped with A. hyperantus. This will reduce the possible interaction of shared ancestry when we compare the two sets of species. Furthermore, comparative studies are correlative by their nature and assume that investigated adaptations are a result of selection starting at an intraspecific level. Thus, as support to this comparative study, the same temperature fecundity patterns were found on an intraspecific level in two Belgian populations of P. aegeria, one population living in an open landscape and the other living in a woodland landscape (Karlsson & Van Dyck, unpublished). Thus, both studies support the idea that habitat preference may lead to adaptive differences in temperature-related life-history traits. References Braby, M.F. (1994) The significance of egg size variation in butterflies in relation to hostplant quality. Oikos, 71, Chai, P. & Srygley, R.B. (1990) Predation and the flight, morphology, and temperature on neotropical rain-forest butterflies. American Naturalist, 135, Clench, H.K. (1966) Behavioral thermoregulation in butterflies. Ecology, 47, Cole, L.C. (1954) The population consequences of life history phenomena. Quarterly Journal of Biology, 29, Dudley, R. (1991) Thermoregulation in unpalatable danaine butterflies. Functional Ecology, 5, Emmet, A.M. & Heath, J. eds., (1989) The Moths and Butterflies of Great Britain and Ireland, Vol. 7. P. 1, Hesperiidae Nymphalidae: the Butterflies. Harley Books, Oxford. Gossard, T.W. & Jones, R.E. (1977) The effects of age and weather on egg-laying in Pieris rapae. Journal of Applied Ecology, 14, Harvey, P.H. & Pagel, M.D. (1991) The Comparative Method in Evolutionary Biology. Oxford Series in Ecology and Evolution. Oxford University Press.

6 104 B. Karlsson & C. Wiklund Heinrich, B. (1993) The Hot-blooded Insects: Strategies and Mechanisms of Thermoregulation. Springer, Berlin. Henriksen, H.J. & Kreutzer, I.B. (1982) The Butterflies of Scandinavia in Nature. Skandinavisk Bogforlag, Odense. Huey, R.B. & Kingsolver, J.G. (1989) Evolution of thermal sensitivity of ectotherms. Trends in Ecology and Evolution, 4, Karlsson, B. & Wiklund, C. (1984) Egg weight variation and lack of correlation between egg weight and offspring fitness in the wall brown butterfly, Lasiommata megera. Oikos, 43, Karlsson, B. & Wiklund, C. (1985) Egg weight variation and starvation endurance in five satyrid butterflies. Ecological Entomology, 10, Kingsolver, J.G. (1983a) Thermoregulation and flight in Colias butterflies: elevational patterns and mechanistic limitations. Ecology, 64, Kingsolver, J.G. (1983b) Ecological significance of flight activity in Colias butterflies: implications for reproductive strategy and population structure. Ecology, 64, Martin, J.F., Gilles, A. & Descimon, H. (2000) Molecular phylogeny and evolutionary molecular phylogeny and evolutionary patterns of the European satyrids (Lepidoptera: Satyridae) as revealed by mitochondrial gene sequences. Molecular Phylogenetics and Evolution, 15, Partridge, L., Barrie, B., Barton, N.H., Fowler, K. & French, V. (1995) Rapid laboratory evolution of adult life-history traits in Drosophila melanogaster in response to temperature. Evolution, 49, Shreeve, T.G. (1984) Habitat selection, mate location, and microclimatic constraints on the activity of the speckled wood butterfly, Pararge aegeria. Oikos, 42, Shreeve, T.G. (1992) The development of butterfly settling posture: the role of predators, climate, hostplant-habitat and phylogeny. Biological Journal of the Linnean Society, 45, Srygley, R.B. & Chai, P. (1990) Predation and the elevation of thoracic temperature in brightly colored Neotropical butterflies. American Naturalist, 135, Stern, V.M. & Smith, R.F. (1960) Factors affecting egg production and oviposition in populations of Colias philodice eurytheme Boisduval (Lepidoptera: Pieridae). Hilgardia, 29, Thomas, J. & Lewington, R. (1991) The Butterflies of Britain and Ireland. Dorling Kindersley Ltd, London. Van Dyck, H. & Matthysen, E. (1998) Thermoregulatory differences between phenotypes in the speckled wood butterfly: hot perchers and cold patrollers? Oecologia, 114, Watt, W.B. (1968) Adaptive significance of pigment polymorphism in Colias butterflies. I. Variation of melanin pigment in relation to thermoregulation. Evolution, 22, Watt, W.B. (1992) Eggs, enzymes, and evolution-natural genetic variants change insect fecundity. Proceedings of the National Academy of Sciences USA, 89, Wiklund, C. & Karlsson, B. (1988) Sexual size dimorphism in relation to fecundity in some Swedish satyrid butterflies. American Naturalist, 131, Wiklund, C. & Persson, A. (1983) Fecundity, egg weight variation and its relation to offspring fitness in the speckled wood butterfly (Pararge aegeria), or why don t butterfly females lay more eggs? Oikos, 40, Wiklund, C., Karlsson, B. & Forsberg, J. (1987) Adaptive versus constraint explanations for egg-to-body size relationships in 2 butterfly families. American Naturalist, 130, Received 29 January 2004; accepted 22 June 2004