Behavioral Ecology and Sociobiology

, Volume 61, Issue 8, pp 1237–1251

Feeding over the 24-h cycle: dietary flexibility of cathemeral collared lemurs (Eulemur collaris)

Authors

    • Dipartimento di Biologia Unita’ di AntropologiaUniversità degli Studi di Pisa
  • An Bollen
    • UNEP ROAP, UN Building
  • Silvana M. Borgognini-Tarli
    • Dipartimento di Biologia Unita’ di AntropologiaUniversità degli Studi di Pisa
  • Joerg U. Ganzhorn
    • Department of Animal Ecology and Conservation
Original Paper

DOI: 10.1007/s00265-007-0354-x

Cite this article as:
Donati, G., Bollen, A., Borgognini-Tarli, S.M. et al. Behav Ecol Sociobiol (2007) 61: 1237. doi:10.1007/s00265-007-0354-x

Abstract

Animals show specific morphological, physiological and behavioural adaptations to diurnal or nocturnal activity. Cathemeral species, i.e. animals with activities distributed over the 24-h period, have to compromise between these specific adaptations. The driving evolutionary forces and the proximate costs and benefits of cathemerality are still poorly understood. Our goal was to evaluate the role of predator avoidance, food availability and diet quality in shaping cathemeral activity of arboreal mammals using a lemur species as an example. For this, two groups of collared lemurs, Eulemur collaris, were studied for 14 months in the littoral forest of southeastern Madagascar. Data on feeding behaviour were collected during all-day and all-night follows by direct observation. A phenological transect containing 78 plant species was established and monitored every 2 weeks to evaluate food availability during the study period. Characteristics of food items and animal nutritional intake were determined via biochemical analyses. The ratio of diurnal to nocturnal feeding was used as response variable in the analyses. The effects of abiotic environmental variables were removed statistically before the analyses of the biotic variables. We found that diurnal feeding lasted longer during the hot–wet season (December–February), whereas nocturnal feeding peaked during the hot–dry and cool–wet seasons (March–August). Although the lemurs foraged mostly in lower forest strata during daylight and used emergent trees preferably at night, the variables which measured animal exposure to birds of prey failed to predict the variation of the ratio of diurnal/nocturnal feeding. Ripe fruit availability and fiber intake are the two variables which best predicted the annual variation of the lemur diurnality. The data indicate that feeding over the whole 24-h cycle is advantageous during lean periods when animals have a fibre-rich, low-quality diet.

Keywords

CathemeralityFeeding strategiesLemur ecologyNutritional intakeEulemur collaris

Introduction

Most animal species are active either at night or during the day. Nocturnality or diurnality are linked to anatomical and physiological adaptations with consequences for reproduction, interspecific interactions or feeding strategies of predators and prey (Charles-Dominique 1975; Aschoff et al. 1982; Martin 1990; Halle and Stenseth 2000; Heesy and Ross 2001; Jacobs 2002; Curtis et al. 2006). Daylength or nightlength per se are important ecological constraints, as they set the period within which obligate diurnal or nocturnal animals must perform their essential behaviours (Dunbar 1988, 1992; Halle and Stenseth 2000; Hill et al. 2003). For example, in exclusively diurnal or nocturnal species, seasonal variations in daylength have been demonstrated to represent a significant constraint, as they restrict the length of the activity period (Hill et al. 2003). Thus, diurnal and nocturnal animals are forced to fulfil their foraging requirements in a limited time. However, cathemeral species, i.e. animals active over the 24-h cycle (Tattersall 1987), evade the above constraints. Cathemerality is ubiquitous across mammalian taxa (Curtis and Rasmussen 2006), and in some orders, e.g. artiodactyls and rodents, this activity has been recorded in most species (Nowak 1991; van Schaik and Griffiths 1996; Halle and Stenseth 2000). The benefits of this flexible activity are supposed to be substantial, given the costs of being exposed to very different light conditions, temperatures or predators. Nevertheless, rather little attention has been given to the trade-off of costs and benefits related to an animal decision of scheduling its activities over the 24-h period (Halle and Stenseth 2000; Kappeler and Erkert 2003).

Lemurs, the primates of Madagascar, offer an excellent opportunity to study the relationship between activity patterns and related environmental factors. The monophyletic radiation of lemurs gave rise to diurnal, nocturnal and cathemeral species (Kappeler 1998). This group of mammals is not strictly limited by their phylogenetic heritage, but has the potential to respond to environmental constraints or possibilities by adjusting their activity cycle rather rapidly. The genus Eulemur is of particular interest in this respect. This genus is classified as cathemeral and shows a marked ecological flexibility within and between habitats (Ossi and Kamilar 2006). The Eulemur dietary habits range from a folivorous diet in some western dry forests (Sussman and Tattersall 1976; Tattersall 1979; Tarnaud 2006) to a year-round frugivorous pattern in the eastern rain forests (Overdorff 1993; Vasey 2000; Johnson 2002) passing through a mixed frugivore–folivore diet in other areas of Madagascar (Curtis et al. 1999; Rasmussen 1999; Simmen et al. 2003). The seasonal and environmental dietary variation in Eulemur spp. have been thought to be associated with a high flexibility of their activity cycle which, in turn, seems to have facilitated the wide dispersal of the genus into different habitats (Engqvist and Richard 1991; Wright 1999; Ossi and Kamilar 2006). Analyses of the high variability within and between habitats in relation to environmental factors should help us to better understand causes and effects in the activity variation of cathemeral primates.

Several hypotheses have been formulated to account for possible direct and indirect benefits as well as for the evolutionary significance of a cathemeral feeding strategy (Tattersall and Sussman 1975; Overdorff 1988; Engqvist and Richard 1991; Overdorff and Rasmussen 1995; van Schaik and Kappeler 1996; Andrews and Birkinshaw 1998; Colquhoun 1998; Curtis et al. 1999, 2006; Donati et al. 1999, 2001; Mutschler 1999; Curtis and Rasmussen 2002; Fernandez-Duque 2003; Kappeler and Erkert 2003). Most of these hypotheses are either related to predator avoidance or to diet quality.

According to the predator avoidance hypothesis, lemurs should be more nocturnal when the pressure from diurnal predators, mainly large birds of prey, increases. This would happen when the protection from raptors is seasonally reduced, such as during periods of leaf fall, and/or when lemurs are obliged to visit the emergent forest layers in search of food, where they are more exposed (Overdorff 1988; Andrews and Birkinshaw 1998; Curtis et al. 1999; Donati et al. 1999; Curtis and Rasmussen 2002; Rasmussen 2005). Whereas in Malagasy ecosystems there is no indication for seasonal changes in predation pressure (Karpanty 2006), trees of the dry deciduous forests lose their leaves during the dry season, making arboreal animals more vulnerable to diurnal birds of prey. This should result in reduced diurnal and increased nocturnal activities of lemurs. If this hypothesis would apply, in the evergreen humid forests there should be no dramatic changes in diurnality and nocturnality, as trees provide cover year-round. Also, nocturnal activity should increase when the lemurs feed on emergent plant species, i.e. in the exposed canopy.

The diet hypothesis predicts that cathemerality should be an adaptation to cope with a low-quality diet during food scarcity periods (Engqvist and Richard 1991; Wright 1999; Tarnaud 2006). Repeated food intake over the 24-h cycle should allow continuous digestion and extraction of nutrients from low-quality fibrous food. This idea is supported by the very short food retention time in the gut recorded for Eulemur spp., which would prevent an efficient energy extraction from fibres (Overdorff and Rasmussen 1995; Campbell et al. 2004). If this hypothesis would apply, activity should spread over the 24-h period when food availability and/or diet quality decreases, i.e. fibre contents increase.

The studies on lemurs available until today fall into two categories which both have their advantages but also disadvantages. Most of the conventional studies recorded variables such as dietary seasonality, phenology and other ecologically relevant factors, but then were unable to compile comprehensive activity records for the 24-h cycle by direct observations (reviewed by Curtis and Rasmussen 2002). Because of the potential for quantitative and qualitative differences between diurnal and nocturnal behaviours, activity budgets based on data sets skewed towards only one phase of the 24-h cycle (often the diurnal part) may not reflect the true animal biorhythm. On the other hand, automated long-term activity recordings resulted in activity records of yet unmatched resolution, but contextual variables were not recorded simultaneously, thus making it difficult to put the activity records into an ecological context (Kappeler and Erkert 2003; Erkert and Kappeler 2004).

In the present research, we tried to overcome the above problems by combining direct observations equally distributed between the two phases of the 24-h cycle with simultaneous monitoring of the main ecological variables. We therefore tested the two hypotheses, i.e. predator avoidance and diet quality, with data collected on cathemeral collared lemurs in the evergreen littoral forest of southeastern Madagascar during a 14-month period. Field observations were combined with chemical analyses of macronutrients, fibres and tannins to get quantitative data on food quality.

Materials and methods

Study site and study species

The study was conducted between December 1999 and January 2001 by GD and AB in “S9”, a 377-ha fragment of the littoral forest (24°45′S, 47°11′E) close to the village of Sainte Luce, 50 km north of Fort Dauphin (Tolagnaro), southeastern Madagascar. This area is characterized by a tropical wet climate, with average monthly temperatures of 23°C, annual rainfall of 2,480 mm and no clear-cut dry season (Bollen and Donati 2005). Temperatures were recorded via data loggers, Hobo H8 pro, set up to work at 2-h intervals, whereas rainfall was measured everyday using a rain gauge placed near the camp (see also Donati and Borgognini-Tarli 2006).

During the study period, we distinguished four different climatic segments: hot–wet (December–February), hot–dry (March–May), cool–wet (June–August) and transitional–dry (September–November; Fig. 1). We considered wet a month with a total rainfall higher than 100 mm (Morellato et al. 2000). Littoral forest grows on sandy soils and occurs within 2–3 km from the coast, at an altitude of 0–20 m a.s.l. (Dumetz 1999). Collared lemurs are cat-sized prosimians living in multi-male, multi-female groups ranging from 2 to 17 individuals. Mean body mass was 2.18 ± 0.11 kg (n = 10), and mean body length was 48.2 ± 2.1 cm (n = 10; Donati 2002).
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-007-0354-x/MediaObjects/265_2007_354_Fig1_HTML.gif
Fig. 1

Rainfall and mean temperature at Sainte Luce from January to December 2000. Seasons are indicated on top of the figure and marked by dashed lines: hot–wet (December–February); hot–dry (March–May); cool–wet (June–August); trans–dry (dry transitional season from September to November)

Phenology

To estimate variation in potential food availability, phenological data were recorded for the plant species (n = 78) known to have fruits, flowers or leaves consumed by collared lemurs during the study period, starting in January 2000. Five individuals for each tree species were selected, and diameter at breast height (DBH) was measured in two botanical transects covering 2,320 m × 10 m. Trees were checked for the presence/absence of new leaves, flowers and fruits twice a month (Bollen and Donati 2005). The overall availability was obtained by weighting the presence of a specific phenophase for each species by its mean DBH. As this lemur species relies mainly on ripe fruits, a distinction was made between the availability of ripe and unripe fruits.

Behavioural observations

Two groups of collared lemurs (group A size: 8–13 individuals; group B size: 4–6 individuals) were followed 3 days and three nights each month from 6:00 to 18:00 hours during diurnal and from 18:00 to 06:00 hours during nocturnal sessions. From December 1999 to January 2001 (group A: 13 observation months; group B: eight observation months), a total of 1,716 observation hours equally distributed between the two phases of the day–night cycle were accomplished. Before the observations, most individuals were captured and marked with nylon collars and coloured pendants. One individual per group was radio-collared.

During the day, animal activity was recorded by the instantaneous focal method (Altmann 1974) at 5-min intervals. Focal animals were chosen evenly from all adult individuals in both study groups. Activity, food type consumed and canopy level were registered during observations. Individual identification and classification of behavioural items were difficult at night. Therefore, the auditory group sampling method (Andrews and Birkinshaw 1998) was applied during nocturnal observations, recording the general activity of the entire group based on visual and auditory clues every 5 min. Specific noises were associated with particular feeding activities. At night, when it was not possible to see the animals, food items were identified from fragments falling to the ground and/or from knowledge of the category exploited at a particular plant species based on diurnal observations. The auditory method might overestimate feeding activity, as “feeding” is also used in cases when part of the group may be resting. However, the usual synchrony of collared lemur groups suggests that this approximation is acceptable (Donati 2002).

Dietary overlap (O) between diurnal and nocturnal activity phases was determined via the formula:
$$ O = - {\sum\nolimits_{{\text{i}} = 1}^{\text{i}} {{\text{Si}}} } $$
adding up the percentages of feeding time shared between phases (S) for each food item/species (i) (Struhsaker 1975; Overdorff 1993a).

All feeding trees used by the focal animals for more than 5 s (i.e. feeding patches) were marked during observation sessions and identified with the help of a local expert on a subsequent day. In addition, diameter at breast height, crown volume and height were measured or estimated. A herbarium was prepared including the plants consumed by collared lemurs. The specimens were identified by botanists and subsequently deposited at the Missouri Botanical Garden of Antananarivo (Bollen and Donati 2005).

Most food plant species were visited during both phases of the cathemeral cycle. To differentiate between resources used mostly during the day or at night, plants were categorized as diurnal, nocturnal and 24-h species defined as: “diurnal species”—those with more than 70% of feeding records (instantaneous records of feeding activity) during the day, “nocturnal species”—those with more than 70% of feeding records at night and “24-h species”—those plants whose diurnal and nocturnal records were between 30 and 70%. To test for seasonal changes in exposure to diurnal birds of prey, feeding records were also grouped according to the two forest levels where they occurred: understorey plus main canopy (below 12 m height) versus emergent (above 12 m height). For the analysis, we considered only the diurnal feeding time spent in the emergent layer, as the evaluation of this variable at night was too difficult.

Nutritional analyses

Biochemical analyses were conducted at the Department of Animal Ecology and Conservation of Hamburg University. A total of 112 food samples eaten by collared lemurs, most of which were fruits (74), were analysed (see also Bollen et al. 2005). Because of the potential nutritional variation in time and space, an effort was made to collect food items where and when lemurs were feeding. However, this was often difficult, if not impossible, when animals were feeding in the canopy or at night, so most fruits were collected on the same trees in a subsequent day. Samples were weighed with an electronic balance (fresh weight), dried in an oven for a standard period, weighed again (dry weight), ground to pass a 2-mm sieve and dried again at 50–60°C before the analyses. The lipid content was determined by extraction using petroleum ether, followed by evaporation of the solvent. The amount of crude protein (total nitrogen × 6.25 = crude protein) was determined using the Kjeldahl procedure. As not all nitrogen is bound in proteins and not all proteins are available for digestion, soluble proteins were assessed by BioRad after extraction of the plant material with 0.1 N NaOH for 15 h at room temperature. Soluble carbohydrates and procyanidin (condensed) tannins were extracted with 50% methanol. Concentrations of soluble sugars were determined as the equivalent of galactose after acid hydrolization of the 50% methanol extract. Concentrations of procyanidin tannins were measured as equivalents of quebracho tannin (Oates et al. 1977). Samples were analysed for neutral (NDF) and acid (ADF) detergent fibres (Goering and van Soest 1970; van Soest 1994; modified according to the instructions for use in an “Ankom fibre analyser”). NDF represents all the insoluble fibre (cellulose, hemicellulose and lignin) partly digestible in species with hindgut fermentation. ADF represents the fibre fraction containing cellulose and lignin, which are mostly indigestible for Eulemur spp. Water content was determined by approximation as the difference between fresh and dry weight. A detailed review of the procedures and their biological relevance is provided by Ortmann et al. (2006).

The estimate of nutrient intake for primary and secondary compounds was obtained as the weighted percentage of dry matter per month, with the proportion of feeding records for each food item as the weighted coefficient (Kurland and Gaulin 1987):
$$ {\text{Intake}}:\,{\sum {{\left( {F_{i} \times X_{i} } \right)}} } $$
where Fi is the monthly proportion of feeding records and Xi is the percentage of dry matter of each chemical parameter for the ith item. We estimated animal nutrient intake by using the proportion of feeding records, as it was impossible to quantify the absolute amount of items consumed at night.

Data analyses

The records of feeding activity were weighted by the total number of instantaneous records. Daily average activity frequencies were calculated for each individual lemur during the day and for the whole group at night. Then, data were pooled by month, and daily grand means per month were obtained. As the two groups did not show significant differences in monthly averages of feeding activity, data were pooled (Mann–Whitney U test: z = 1.2, n1 = 13 months for group A, n2 = 8 months for group B, p = 0.238). We considered feeding activity to be nocturnal if it occurred between the end of the astronomical evening twilight (from when the sun is 18° below the horizon) and the beginning of astronomical morning twilight (until when the sun is 18° below the horizon). Diurnal feeding activity included morning and evening twilights.

We used the non-parametric Friedman analysis of variance (ANOVA, and Wilcoxon post-hoc tests) to investigate seasonal differences in feeding activity. Friedman analysis was performed by comparing mean hourly activity. Twilight hours changed over the year seasonally. Therefore, to compare strictly nocturnal with strictly diurnal activity over the year, we restricted the comparisons to 12 diurnal and eight nocturnal hours. Kruskal–Wallis ANOVA was used to compare the three independent sets of nutritional analyses and dimensional characteristics between diurnal, nocturnal and 24-h plant species. A combination of standard and stepwise multiple regression was used to determine the biotic variables that best account for seasonal variations in the relation between diurnal and nocturnal feeding activity. All the variables used for the analysis were tested for normality. Only the dependent variable, i.e. monthly ratio of diurnal/nocturnal feeding, was severely skewed. Therefore, it was log-transformed. As several abiotic factors strongly influence the Eulemur activity (Donati and Borgognini-Tarli 2006), we removed the effect of mean monthly daylength, mean monthly moon phase, mean monthly temperature and rainfall (all averaged on the actual observation days) by regressing the ratio of diurnal/nocturnal feeding on these variables. Then, we used the residuals of the ratio of diurnal/nocturnal activity as the new dependent variable in the stepwise multiple regression to select a subset of predictive independent variables for each biotic factor, i.e. animal exposure, food availability and diet quality. For each factor, we used a set of non-overlapping independent variables to minimize the problem of multicollinearity (e.g. NDF and ADF were not included in the same equation). The variables measured to estimate the effect of animal exposure to diurnal birds of prey (diurnal time spent above 12 m, mean height of feeding trees), food availability (availability of mature fruits, unripe fruits, flowers and young leaves) and diet quality (intake of lipids, soluble proteins, carbohydrates, NDF, tannins and water) entered the analysis in their original form. As a final step, we used the resulting subset of variables to assess simultaneously the influence of each factor on the ratio of diurnal/nocturnal feeding in a standard multiple regression analysis. Multiple regressions were run with monthly data. All tests are two-tailed, with the level of significance set at p = 0.05 for inclusion in the models.

Results

Cathemeral feeding pattern

The two groups of collared lemurs exhibited cathemeral feeding activity with a monthly average of 22.9 ± 4.2% of the total records devoted to feeding, 26.4 ± 6.3% of diurnal and 18.2 ± 9.4% of nocturnal records, respectively. Diurnal feeding activity varied seasonally (Friedman: χ2 = 12.2, n = 12 h, df = 3, p = 0.007). Post-hoc tests indicate that diurnal feeding was significantly higher during the hot–wet season than during the other three seasons (Wilcoxon, n = 12 h: hot–wet vs hot–dry: z = 2.0, p = 0.041; hot–wet vs cool–wet: z = 2.6, p = 0.009; hot–wet vs trans–dry: z = 3.0, p = 0.003), whereas significant differences were not found among the other seasons. Nocturnal feeding showed significant seasonal variations as well (Friedman: χ2 = 20.6, n = 8 h, df = 3, p < 0.001). Nocturnal feeding was higher during the hot–dry and cool–wet seasons than during the hot–wet and transitional–dry seasons (Wilcoxon, n = 8 h: hot–wet vs hot–dry: z = 2.5, p = 0.012; hot–wet vs cool–wet: z = 2.5, p = 0.012; hot–dry vs trans–dry: z = 2.5, p = 0.012; cool–wet vs trans–dry: z = 2.4, p = 0.018; Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-007-0354-x/MediaObjects/265_2007_354_Fig2_HTML.gif
Fig. 2

Annual variation in the proportion of feeding time recorded each month (±SE) during the day and at night. Feeding time is calculated as the number of feeding records (instantaneous every 5 min) on the total number of records. Diurnal feeding includes all the feeding records observed between sunrise and sunset and during the astronomical twilights. Seasons as in Fig. 1

Collared lemurs appeared to be mainly frugivorous year round (74.0 ± 12.8% of the total consumption were ripe fruits and 4.5 ± 4.3% were unripe fruits) during the study period, with minor contributions of flowers (14.0 ± 9.2%), leaves (3.2 ± 2.4% were young leaves and 1.2 ± 1.4% were mature leaves), invertebrates (2.6 ± 2.4%) and other items (0.7 ± 0.6%; Table 1).
Table 1

Proportion of time spent each month feeding on different food categories by collared lemurs from December 1999 to January 2001

Seasons

RF

UF

ML

YL

FL

I

O

Hot–wet

Dec

83.92

4.11

1.99

6.10

1.15

1.90

0.83

Jan

66.95

1.20

1.20

2.89

23.42

4.30

0.04

Feb

58.91

3.65

1.97

2.94

24.83

6.20

1.50

Hot–dry

Mar

72.73

2.52

0.00

3.03

19.70

2.00

0.02

Apr

55.42

2.65

1.76

6.44

30.20

3.50

0.03

May

79.60

4.47

0.00

5.47

9.86

0.30

0.30

Cool–wet

Jun

91.55

5.05

0.34

0.37

2.09

0.60

0.00

Jul

46.69

23.70

5.08

1.66

22.87

0.00

0.00

Aug

77.32

3.45

1.85

8.19

12.79

0.20

0.80

Trans–dry

Sep

79.35

1.00

0.57

2.40

15.24

0.30

1.14

Oct

87.95

2.41

1.20

2.41

2.41

2.40

1.22

Nov

75.70

3.24

0.00

1.04

15.02

3.80

1.20

Hot–wet

Dec

82.97

3.05

0.52

0.44

8.72

2.80

1.50

Jan

76.23

2.50

0.00

1.63

9.84

8.20

1.60

mean

73.95

4.50

1.17

3.21

14.00

2.61

0.71

RF Ripe fruits; UF unripe fruits; ML mature leaves; YL young leaves; FL flowers; I invertebrates; O other. Seasons as in Fig. 1.

From December 1999 to January 2001, collared lemurs exploited food items from a total of 120 plant species from 45 families. During diurnal activity, animals fed on 99 plant species (82.5% of the entire sample), 35 of which (29.2%) were visited exclusively during daylight hours. During nocturnal activity, the study groups fed on 85 plant species (70.8%), 21 of which (17.5%) were visited exclusively at night. In spite of the fact that very few plants were exploited exclusively during the day or at night, there is a considerable quantitative difference between diurnal and nocturnal diets (Table 2). On average, the two menus overlap by 45.0%, with a maximum in June (81.0%) and a minimum in December (12.6%).
Table 2

Scientific name, family, part eaten (frm ripe fruits, fru unripe fruits, flo flowers, yle young leaves), percentage in the diet (%), number of diurnal feeding records (Diur), number of nocturnal feeding records (Noct) and vertical layer (V.l.; C canopy, E emergent) of the plants visited during >1% of total feeding time by collared lemurs from December 1999 to January 2001

Genus and species

Family

Part eaten

Percentage

Diur

Noct

V.l.

Syzigium sp2

Myrtaceae

frm

13.20

210

238

C

Uapaca ferruginea

Euphorbiaceae

frm, fru

9.28

37

278

E

Uapaca littoralis

Euphorbiaceae

frm, fru

7.25

103

143

C

Pandanus dauphinensis

Pandanaceae

frm

4.06

51

87

C

Cynometra cloiselii

Fabaceae

flo, frm, yle

3.98

83

52

C

Olea sp1

Oleaceae

frm

3.86

85

46

C

Eugenia sp2

Myrtaceae

frm, fru

3.76

48

79

E

Ravenala madagascariensis

Streliziaceae

flo

3.45

12

105

C

Canthium variistipule

Rubiaceae

frm, fru, yle

3.33

72

41

C

Dypsis prestoniana

Arecaceae

frm

2.74

42

51

E

Canarium boivinii

Burseraceae

frm

2.39

2

79

E

Vepris elliotii

Rutaceae

frm, fru

2.36

63

17

C

Cinnamosna madagascariensis var namorensis

Canellaceae

frm

1.97

21

46

C

Terminalia fatraea

Combretaceae

frm, flo

1.91

9

56

E

Symphonia sp2

Clusiaceae

flo

1.86

54

9

C

Schizolaena elongata

Sarcolaenaceae

frm

1.83

22

40

E

Homalium albiflorum

Flacourtiaceae

flo

1.77

11

49

C

Sarcolaena multiflora

Sarcolaenaceae

frm, fru, flo

1.53

22

30

C

Acanthostyla aff. Longistylus

Pandanaceae

frm

1.41

15

33

C

Erythroxylum nitidulum

Erythroxylaceae

frm

1.21

0

41

C

Poupartia chapelieri

Anacardiaceae

frm, fru

1.06

20

16

C

Syzigium sp1

Myrtaceae

frm, fru, flo

1.01

8

26

E

Noronhia ovalifolia

Oleaceae

frm, yle

1.00

12

22

C

Predator avoidance: effect of animal exposure

Characteristics of food plants and variation of animal exposure

Following our distinction between diurnal (n = 48), nocturnal (n = 36) and 24-h (n = 31) food plants, we found a significant difference among the heights of the trees visited during the three time categories (Kruskal–Wallis: H = 36.74, n = 115, df = 2, p < 0.001). Diurnal species (6.18 ± 3.39 m) were markedly lower than nocturnal (10.23 ± 2.87 m) and 24-h (10.42 ± 2.71 m) species. Collared lemurs spent on average 14.5 ± 13.8% of their diurnal feeding in the emergent forest layer, i.e. above 12 m. Diurnal feeding above 12 m peaked in the transitional–dry season (40.0% in September) and was minimal during the hot–wet season (2.7% in January 2000).

Feeding activity variation failed to be significantly predicted by the model, including variation in mean height of feeding trees and diurnal feeding spent above 12 m (the two proxy variables used to measure animal exposure). In fact, whereas the residual ratio of diurnal/nocturnal feeding (y) was related to diurnal feeding spent above 12 m (x) via the equation \( y = - 0.622x + 0.195 \), the regression was not significantly different from zero (F = 2.02, p = 0.175, n = 14; R2 = 0.377).

Diet: effect of food availability and diet quality

Seasonal availability

Seasonal flushing, flowering, and fruiting patterns of the 78 plant species included in the diet of collared lemurs are shown in Fig. 3. Overall, the littoral forest exhibits a period of resource abundance during the hot–wet season and a relative scarcity during the rest of the year. Leaf flush is concentrated in the hot–wet season. Flower availability reaches maximum level in October–November. Ripe fruit availability peaks during the hot–wet season (January–February) and has two lean periods in the hot–dry season (March–April) and in the cool–wet season (July–August), with only 6.4% of the plant species having ripe fruits. Unripe fruit availability precedes the ripe fruit peak with a gradual increase from September to January.
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-007-0354-x/MediaObjects/265_2007_354_Fig3_HTML.gif
Fig. 3

Ripe fruit, unripe fruit, new leaf and flower availability at monthly intervals from January 2000 to January 2001 at Sainte Luce. From February to May, data on young leaves are not available. Seasons as in Fig. 1

The residual ratio of diurnal/nocturnal feeding (y) was related significantly only to the availability of ripe fruits (x) via the equation \( y = 0.649x - 0.055\,{\left( {F = 4.12,\,p = 0.036,\,n = 14;\,R^{2} = 0.647} \right)} \). Thus, the lemurs did more diurnal feeding when ripe fruit availability was high and more nocturnal feeding when ripe fruit availability was low.

Nutritional characteristics of food items and animal nutritional intake

As collared lemurs were mostly frugivorous during the study period, the small sample size of flowers, unripe fruits and leaves eaten does not allow statistical tests. Therefore, we considered only ripe fruits for the comparison between nutritional characteristics of the diurnal and the nocturnal food items. Even if nocturnal fruits tended to have higher fibre content than diurnal and 24-h fruits, the three fruit categories did not differ significantly (Table 3).
Table 3

Phytochemical characteristics as a percentage of dry matter of diurnal (>70% of feeding records during the diurnal phase), nocturnal (>70% of feeding records during the nocturnal phase), and 24-h (70%> diurnal or nocturnal records> 30%) fruits eaten by collared lemurs

 

Lipids

CP

SP

SC

CT

NDF

ADF

Watera

Diurnal fruits

2.7

5.4

2.2

19.3

0.12

28.8

23.1

75.0

Quartiles

1.4–4.5

4.6–8.0

1.4–4.4

9.7–36.6

0.00–0.41

23.9–47.3

17.2–33.5

67.7–84.9

n

29

31

31

31

31

28

28

23

Nocturnal fruits

2.3

5.6

2.8

13.3

0.14

43.4

33.9

75.3

Quartiles

1.6–5.3

4.3–7.0

1.6–3.9

5.7–32.1

0.00–0.40

28.0–62.4

17.9–46.0

68.6–79.8

n

28

32

32

32

32

25

25

25

24-h fruits

2.2

5.4

3.8

16.7

0.23

32.6

22.4

76.1

Quartiles

1.2–3.1

3.9–7.0

2.5–5.6

10.0–27.7

0.15–0.68

26.4–41.8

18.1–29.2

63.5–83.9

n

24

24

24

24

24

24

24

20

H

1.31

1.07

4.44

2.02

4.19

4.60

3.97

0.36

P

0.52

0.58

0.12

0.36

0.12

0.10

0.14

0.83

CP crude proteins; SP soluble proteins; SC soluble carbohydrates; CT condensed tannins; NDF neutral detergent fibres; ADF acid detergent fibres

aWater is indicated as a percentage of wet matter. Values are medians (in bold), quartiles and (n) sample size. Statistics are H values based on the Kruskal–Wallis non-parametric ANOVA.

The estimation of monthly nutrient intake indicated that the largest percentage of dry food weight ingested by collared lemurs was represented by fibres (37.3 ± 5.3% NDF during the whole study period). Fibre intake peaked in the hot–dry and transitional–dry season, with a maximum value in September/October 2000 (44.8% NDF and 33.7/34.2% ADF; Table 4). Intake of soluble carbohydrates (18.8 ± 8.1%) peaked in June 2000 (38.3%). Total protein intake (7.3 ± 1.6%) reached the highest level during the transitional–dry season and the hot–wet season (11.9% in January 2001). Soluble protein intake (3.9 ± 0.6%) was highest during the hot–wet and the hot–dry season (4.9% in March 2000). Lipid intake (3.7 ± 0.8%) was highest in the hot–wet season (5.3% in February 2000 and January 2001). Finally, tannin intake (0.6 ± 0.2%) showed monthly fluctuations and a maximum during the cool–wet season (1.0% in June).
Table 4

Lemur nutritional intake as a percentage of the total dry matter in monthly diets

Seasons

Lipids

CP

SP

SC

CT

NDF

ADF

Hot–wet

Dec

2.82

5.50

2.72

25.04

0.26

36.12

25.50

Jan

4.06

6.84

4.32

20.14

0.58

29.51

20.58

Feb

5.32

6.83

4.22

16.38

0.27

33.83

23.24

Hot–dry

Mar

2.89

6.73

4.85

17.63

0.36

41.31

23.28

Apr

3.57

7.48

4.00

22.01

0.41

38.89

23.39

May

2.78

6.60

3.65

29.57

0.90

33.86

24.42

Cool–wet

Jun

3.45

5.48

3.01

38.26

1.04

26.88

21.01

Jul

3.21

5.97

3.62

20.59

0.78

35.50

26.16

Aug

4.49

7.23

3.54

16.63

0.78

41.97

31.82

Trans–dry

Sep

3.37

8.32

4.06

15.57

0.39

44.75

34.15

Oct

3.98

7.01

4.28

8.56

0.32

44.76

33.66

Nov

3.09

7.92

4.69

10.82

0.64

39.74

29.18

Hot–wet

Dec

3.53

8.44

4.55

10.67

0.71

39.56

29.26

Jan

5.32

11.88

3.58

11.35

0.41

35.28

26.24

mean

3.71

7.30

3.93

18.80

0.56

37.28

26.56

CP crude proteins; SP soluble proteins; SC soluble carbohydrates; CT condensed tannins; NDF neutral detergent fibres; ADF acid detergent fibres. Seasons as in Fig. 1

Water accounts for a large proportion (72.5 ± 9.3%) of the fresh weight in all food items. Indeed, water intake from food remained around 70% during the entire study period. The highest water intake from food (84%) was found in May and June.

The stepwise regression indicated that the residual ratio of diurnal/nocturnal feeding (y) was related significantly only to the NDF intake (x) via the equation \( y = - 0.923x + 0.255\,{\left( {F = 11.55,\,p = 0.002,\,n = 14;\,R^{2} = 0.908} \right)} \). Thus, the lemurs did more diurnal feeding when NDF intake was low and more nocturnal feeding when NDF intake was high.

Combined effects

Annual variation in the availability of mature fruits, NDF intake and diurnal time spent above 12 m explained 71% of the variation in the residual ratio of diurnal to nocturnal feeding (Table 5). However, only NDF intake contributed significantly to the multiple regression model. Thus, increased nocturnal feeding activity coincided with the consumption of a more fibrous diet.
Table 5

Multiple regression analysis of annual variation in the residuals of the ratio of diurnal/nocturnal feeding (F = 7.552; p = 0.008, n = 14, R2 = 0.716)

Variables

Beta

t

p

NDF intake

−0.768

−3.021

0.014

Availability of mature fruits

0.276

1.228

0.250

Diurnal feeding above 12 m

0.067

0.223

0.828

NDF neutral detergent fibre

Discussion

Our study showed that collared lemurs use a cathemeral feeding strategy year-round. Diurnal feeding showed a clear seasonal variation with an increase during the hot–wet season, whereas nocturnal feeding peaked during the hot–dry, cool–wet seasons. This activity regime matches the profiles recorded in western Eulemur populations which exhibit a prevalent diurnal activity during the wet season, whereas nocturnal activity is more frequent during the dry period (Curtis et al. 1999; Donati et al. 1999; Rasmussen 1999; Kappeler and Erkert 2003; Tarnaud 2006).

Role of predator avoidance

The predator avoidance hypothesis suggests advantages in visiting food resources at night that might be dangerous to exploit during the day due to animal exposure to diurnal raptor attacks. It has been calculated that diurnal raptor predation removes on average 5.8% per year of Eulemur populations in Malagasy rainforests (Karpanty 2006). Thus, the threat of fatal attacks by diurnal raptors is substantial. A variable, temporal feeding strategy is known to be used by various mammals to avoid or confuse predators (Chiarello 1998; Zielinsky 2000; Colquhoun 2006; Halle 2006). In our study, in spite of the fact that the main food resources exploited by collared lemurs were available both during the day and at night, some differences in their relative utilization during the two phases were evident. As also indicated by previous studies on other Eulemur spp. (Overdorff 1988; Andrews and Birkinshaw 1998; Curtis et al. 1999; Donati et al. 1999; Rasmussen 1999), E. collaris showed a preference for lower forest layers during diurnal feeding and for the upper canopy and emergent trees during nocturnal feeding. This observation apparently supports the hypothesis that cathemerality may function as a strategy to reduce predation by diurnal raptors (Polyboroides radiatus, Accipiter henstii and Buteo brachipterus), in the sense that lemurs visit exposed trees preferably at night (Curtis and Rasmussen 2002; Rasmussen 2005). In line with the predator avoidance hypothesis, the decrease of Eulemur diurnal activity during the dry season observed in the deciduous forest was interpreted as an antipredatory strategy (Curtis et al. 1999; Donati et al. 1999; Rasmussen 1999; Curtis and Rasmussen 2002) to avoid diurnal exposure when vegetation cover is scarce. However, in our study, we observed a similar reduction of the diurnal/nocturnal activity ratio during dry periods in the evergreen littoral forest where vegetation cover is fairly constant. Most importantly, our data showed that after having statistically removed the effects of abiotic variables, the annual variation of feeding time spent in the emergent layer, a proxy of animal exposure to birds of prey, turned out to be a weak predictor of the variation in the ratio of diurnal to nocturnal feeding. This variable becomes non-significant when simultaneously controlled for the variables of food quality and availability. Thus, even if our data indicate that cathemerality of E. collaris may be linked to avoid diurnal exposure, the evidence for an antipredatory strategy as a determinant of the observed activity changes is not clear. In fact, a temporal selection of forest layers is also a general strategy to avoid heat stress which is not just limited to cathemeral lemurs (Pollock 1979). A number of other factors such as visual advantages may also favour the exploitation of lower layers during the day and of the upper canopy at night (Donati et al. 2001; Kappeler and Erkert 2003; Yamashita et al. 2005). Interestingly, also in contrast to the expectation, during the nestling period of Polyboroides (November–December) when these birds should maximize hunting pressure, diurnal feeding of the collared lemurs increased. Recent evidence showing a lack of correlation between predation rates of diurnal raptors, and activity patterns in lemur species is consistent with our conclusion (Powzyk 1997; Karpanty and Goodman 1999; Goodman 2003; Karpanty 2006).

Role of diet: food availability and quality

According to the diet hypothesis, lemurs use cathemeral feeding to maximize the intake of low quality food during periods of resource scarcity by extending their feeding bouts over the 24-h cycle (Engqvist and Richard 1991). This strategy would allow continuous digestion and might represent a solution to compensate for the lack of digestive specialisations in Eulemur spp. To test this hypothesis, it is necessary to demonstrate first an influence of food availability on lemur activity, and then an association between the latter and the quality of the diet.

Given the almost exclusively frugivorous diet of E. collaris at Sainte Luce, the periods of fruit scarcity recorded in our phenology seem to be particularly harsh. Our data indicated that once the effects of abiotic variables had been removed, the availability of ripe fruits was a good predictor for the variation of the diurnal/nocturnal feeding activity ratio of these lemurs. Collared lemurs concentrated their diurnal feeding bouts during periods of ripe fruit abundance, whereas they increased nocturnal feeding during fruit bottlenecks.

As to the quality of the diet, our lemurs did not show dramatic changes in food category consumption, as they remained mainly frugivorous year-round. Furthermore, fruits eaten during the two phases of the 24-h cycle did not show significant differences in nutritional contents. However, our estimation of seasonal variations of nutritional intake showed that lemurs had a higher quality diet when they were mainly diurnal compared to the periods when they were equally active during the day and at night. When the effect of abiotic variables was removed, fibre intake was the best predictor of the annual variation in the ratio of diurnal/nocturnal feeding and remained highly significant once the variables representing animal exposure to birds of prey and food availability were combined in the model (Table 5). Thus, whereas our animals remain frugivores during lean periods, as fruits are more fibrous and less nutritious, lemurs have to expand their feeding activity period. These results support the hypothesis of Engqvist and Richard (1991).

Whereas the link between food availability and activity changes in previous studies on lemurs is rather controversial (Andrews and Birkinshaw 1998; Curtis et al. 1999; Colquhoun 1998; Rasmussen 1999; Kappeler and Erkert 2003), a similar relationship between fibre content in food items and activity changes was recently described for E. fulvus (Tarnaud 2006). In contrast, nutritional intake of E. mongoz does not show any association with its cathemeral activity changes (Curtis et al. 1999).

How primates cope with a fibrous diet

Although large mammals may easily shift from high-quality to low-quality fibrous food, small animals need to find a solution to meet energy requirements (van Soest 1994). According to the Jarman–Bell principle, small mammals such as Eulemur spp. are predicted to have high energy requirements per unit of body mass (Kleiber 1961; Parra 1978; Demment and van Soest 1985) and a relatively small gut capacity (Jarman 1974; Sailer et al. 1985; Lambert 2002). In some primate species, the need for efficient extraction of nutrients and energy and the digestion of fibrous components is achieved by specific anatomic adaptations such as the sacculated stomach of colobines (Chivers 1995) or the enlarged caecum-colon of leaf-eating lemurs (Chivers and Hladik 1980; Martin 1990). However, Eulemur spp. do not possess specialisations of their digestive tract (Hill 1953). On the contrary, they show an extremely rapid food transit time (between 1.5 and 3.2 h according to Overdorff and Rasmussen 1995; Campbell et al. 2004). Holding anatomy and diet constant, shorter transit times result in lower levels of fermentation and reduced energy extraction from fibrous food (Milton 1981, 1998; Lambert 2002). Scaled on body mass (mean transit time in hours/mean body mass in kilograms), the food transit time of brown lemurs is not only shorter than that recorded for specialised, folivorous primates (E. fulvus: ≈0.9; Hapalemur griseus: ≈35.0; Overdorff and Rasmussen 1995; Campbell et al. 2004), but it is short even when compared to frugivore/folivore primates lacking gut specialisations (e.g. Cercopithecus spp., between 3.0 and 8.0; Lambert 2002). In fact, Eulemur food transit time approaches the values recorded for specialised fruit-eaters (Pan troglodytes, 0.5; Lambert 2002; Ateles geoffroyi, 0.6; Milton 1984). In line with the above observations, Eulemur fibre digestion ability is low as compared to other frugivore/folivore primates (Lambert 2002; Campbell et al. 2004).

In addition, Eulemur spp. neither fall into seasonal torpor as other lemurs do, i.e. Cheirogaleus spp. and Microcebus spp. (Ganzhorn et al. 2003), nor they seem to reduce their activity to face periods of low-quality diet (Richard 1977; Milton 1978, 1998; Nash 1998). In contrast, if we evaluate over 12 h the equivalent of the 24-h feeding activity of our study groups, the resulting value, 45.8% of total time, largely exceeds the investment devoted to feeding of most primate species (see also Mutschler 1999 for Hapalemur alaotrensis).

In summary, lacking anatomical adaptations to cope with a fibrous diet Eulemur spp. seem to adopt a “power feeding strategy” that consists in processing a large volume of food per unit time to meet nutritional requirements (Milton 1981, 1998; Simmen et al. 2003). During lean periods, when the quality of diet decreases further, a 12-h temporal window might not be enough to meet these requirements, and lemurs need to extend their feeding over the 24-h period. It is important, however, to remind that we did not evaluate the absolute amount of food ingested because of the poor observation conditions at night. Thus, given the potential error of estimating food intake via feeding time (Zinner 1999), our conclusions have to be taken with caution until more fine-grained data will be available.

Evolutionary aspects

A power feeding strategy over the 24-h to maximize the overall food intake has been observed in many mammals with a limited capacity of fibre digestion. Examples range from large perissodactyls (Janis 1976; Cork and Foley 1991; Nowak 1991) to small rodents (Keys and van Soest 1970; Halle and Stenseth 2000; Halle 2006). Voles, in particular, feed on cellulose rich food with low energy content but they do not possess structured fermentation chambers. Thus, voles are unable to process the ingested food completely during a long rest period but they require a 24-h foraging activity (Halle 2006).

We may therefore ask why a cathemeral feeding strategy is rather rare, if not absent, in primates, but has been evolved among Malagasy prosimians. The constraints represented by Malagasy habitats as well as the physiology of Eulemur spp. may help to understand this issue.

It has been postulated repeatedly that Madagascar is characterized by low plant productivity, irregular phenological cycles and climatic extremes, making this island a harsh place for primates (Dewar and Wallis 1999, Ganzhorn et al. 1999, 2003; Wright 1999; Wright et al. 2005; Bollen and Donati 2005; van Schaik et al. 2005). Lemur species ought to have evolved adaptations to cope with these constraints.

From a physiological point of view, Eulemur spp. appear to be extremely hypometabolic (Daniels 1984; Müller 1985) and able to adjust their energy expenditure seasonally (Pereira 1993; Pereira et al. 1999). In particular, whereas hypometabolism is common in lemurs (Schmid and Ganzhorn 1996; Genoud 2002), E. fulvus has been shown to have a basal metabolic rate which is among the lowest in prosimians, ranging 28–56% of the expected value of the Kleiber equation (Daniels 1984; van Schaik and Kappeler 1996). Thus, if the field metabolic rate is a constant multiple of the basal metabolic rate, as it seems the case in lemurs (Drack et al. 1999), it is possible that a very low metabolism would allow these animals to be active over the 24-h cycle during periods of food shortage. Therefore, a low, seasonally changing metabolism may represent a pre-adaptation of Eulemur spp. to evolve a flexible activity regime as a strategy to deal with lean periods. In line with this argument, it has already been suggested that in arboreal, frugivorous–folivorous mammals, hypometabolism reflects low food digestibility and limited access to energy resources (McNab 1978, 1986; Genoud 2002; but see Harvey et al. 1991; Ross 1992). Cathemerality might thus represent an alternative to hibernation or gut specialisations. The lack of physiological or morphological adaptations might then allow the application of other strategies without inhibiting boundaries. An exception to this scenario is represented by some Hapalemur spp. which are specialised folivores with digestive adaptations and which are also cathemeral (Mutschler 1999; Tan 1999). However, in the latter genus where most populations appear to be diurnal (Overdorff et al. 1997; Tan 1999; Grassi 2001; Donati, personal observation), cathemerality has been recorded only in unusual habitat conditions (Mutschler 1999) or in case of high interspecific competition (Tan 1999).

There is still an open debate whether cathemerality represents an ancestral strategy in lemurs (Tattersall 1982; Curtis and Rasmussen 2002), or it is the result of a non-adaptive disequilibrium from nocturnality to diurnality due to the recent Holocene demise of large diurnal raptors and competitive lemurs in Madagascar (van Schaik and Kappeler 1996; Kappeler and Erkert 2003). Besides the age of emergence of this trait among lemurs, our results, together with a number of recent long-term studies, have accumulated evidence for the adaptive significance of this activity in primates (reviewed by Curtis et al. 2006). New findings on the anatomy of the visual system demonstrated that cathemeral primates posses intermediate, well-distinguished traits between the nocturnal and the diurnal eye (Kay and Kirk 2000; Kirk 2006), which seem not to be indicative of a recent origin. Furthermore, a recent reconstruction of the lemur phylogenetic tree based on large-scale molecular analyses have indicated an ancient common ancestry (≈43 Mya) of the Indridae/Lemuridae families, which include only diurnal and cathemeral genera except for the nocturnal Avahi (Roos et al. 2004). Thus, while there is much debate whether the ancestral primate was nocturnal (Martin 1990; Kappeler 1998; Heesy and Ross 2001) or cathemeral/diurnal (Tan and Li 1999; Tan et al. 2005), it seems parsimonious to hypothesize an old, adaptive origin of cathemerality in lemurs for the time being.

Conclusions

The data support the idea that food availability and quality are important in determining the cathemeral activity of collared lemurs at Sainte Luce. The reduced availability of resources during parts of the year seems to favour extended feeding activity over the 24-h cycle when animals have a fibre-rich diet. Earlier reports have already demonstrated that lemurs show behavioural and physiological adjustments to cope with Malagasy seasonality (Jolly 1966; Morland 1993; Wright 1999; Ganzhorn et al. 2003; Erkert and Kappeler 2004). In evolutionary terms, primates cope with fruit scarcity in different ways (van Schaik et al. 2005). Some switch diets and evolved gut adaptations to support seasonal increase in folivory, and others overcome food scarcity by going into torpor/hibernation. Eulemur spp. seem to have evolved a third strategy, i.e. extending feeding periods without linking themselves to a given section of the day–night cycle. This “third way” appears to be more flexible than the above “solutions”, as it can be modulated according to specific seasonal conditions.

It must be stressed, however, that there is no single explanation for the cathemeral trait. Different constraints seem to act simultaneously and/or in different ways depending on the habitat (Curtis et al. 2006). In our study, we tried to separate the effects of some biotic factors on feeding activity, but the overlapping effects of abiotic factors and other biotic factors should not be underestimated. For example, temperature variations have profound effects on lemur activity, particularly in very seasonal climatic conditions (Curtis et al. 1999; Donati et al. 1999; Mutschler 1999; Fernandez-Duque 2003; Kappeler and Erkert 2003; Donati and Borgognini-Tarli 2006). Biotic factors other than those examined in our study, such as interspecific competition, also seem to play a role in shaping activity of cathemeral lemurs, particularly in habitats where two Eulemur spp. co-occur in sympatry (Rasmussen 1999; Curtis and Rasmussen 2006).

Acknowledgements

This study was carried out under the collaboration agreement between the Departments of Animal Biology and Anthropology of the University of Antananarivo, the Institute of Zoology of Hamburg University and QIT Madagascar Minerals. We thank the Commission Tripartite of the Malagasy Government, the Ministère des Eaux et Forêts, QMM and Missouri Botanical Garden at Antananarivo for their collaboration and permissions to work in Madagascar. In particular, we acknowledge Manon Vincelette, Jean-Baptiste Ramanamanjato and Laurent Randriashipara of the QMM Conservation Team for providing help at various stages of this research. We are grateful to Nicoletta Baldi and Valentina Morelli for providing additional data on E. collaris feeding ecology; many thanks as well to Dauphin Mbola, Givé Sambo, Ramisy Edmond, the local assistants who helped with the collection of behavioural and phenology data. Irene Tomaschewsky helped with plant analyses. We thank Peter Kappeler and the reviewers of BES for their constructive comments on earlier drafts of the manuscript. The first author would like to express a special gratitude to Sabine Groos for her continuous help during the elaboration of this paper. GD was supported by a doctoral grant of the Italian Ministry for Scientific Research (MURST) and the University of Pisa. AB was supported by a grant of the Belgian Fund for Scientific Research, Flanders (FWO).

Copyright information

© Springer-Verlag 2007