Behaviour, morphology and the division of labour in two soldier-producing aphids

Transcription

1 ANIMAL BEHAVIOUR, 2001, 62, doi: /anbe , available online at on Behaviour, morphology and the division of labour in two soldier-producing aphids ALEXANDER W. SHINGLETON & WILLIAM A. FOSTER Department of Zoology, University of Cambridge (Received 25 July 2000; initial acceptance 2 November 2000; final acceptance 5 June 2001; MS. number: 6649R) In social insects the division of labour will reflect the trade-off between behavioural flexibility and physical specialization of workers. We investigated this trade-off in two social aphids: Pseudoregma sundanica (Van der Goot) and P. nicolaiae (Takahashi) (Hormaphididae: Cerataphidini). They are sibling species that produce dimorphic first-instar larvae of soldiers and nonsoldiers. Defence is ostensibly the task of soldiers, but P. sundanica is also defended by tending ants. We determined the extent to which physical castes are developed in both species by measuring the lengths of various body parts of soldier and nonsoldier first instars. We also measured the defensive behaviour of soldiers and nonsoldiers by confronting them with aggressive stimuli. The soldiers and nonsoldiers of P. nicolaiae were morphologically and behaviourally more similar than those of P. sundanica, and the nonsoldiers showed greater behavioural plasticity. This suggests a more flexible division of labour in P. nicolaiae, with nonsoldiers being recruited to defensive tasks. Across both castes first-instar larvae were also more aggressive in P. nicolaiae than in P. sundanica. We relate these findings to the differences in ant tending in the two species. The data suggest that the caste structure of the social aphids is more complex than expected, and is perhaps more similar to that of the social Hymenoptera The Association for the Study of Animal Behaviour The division of labour by allocation of tasks amongst worker members of a colony (polyethism) is an essential component of the ecological success of the social insects (Wilson 1987). In some social insects, this allocation of tasks is plastic, with workers performing a range of tasks throughout their lifetime (e.g. Dew & Michener 1981; Seeley 1982; Calabi & Traniello 1989; Gordon 1989). The division of labour is therefore not fixed and a colony can rapidly allocate individuals to tasks as required, in response to environmental change (Bourke & Franks 1995). In others, workers are more specialized, maybe conducting only one or two tasks (e.g. Wilson 1980; Wetterer 1994; Stapley 1999). This behavioural specialization allows workers to increase efficiency by developing morphologies associated with the tasks they perform (Oster & Wilson 1978). The result is a division of labour based on physical castes. Response to environmental change is, however, limited to the relatively slow process of large-scale adjustments in the caste profile of a colony (Tofts & Franks 1992). The selective pressure to increase efficiency by developing specialized physical castes will therefore be mitigated by the selective pressure to decrease response times by retaining behavioural Correspondence: A. W. Shingleton, Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, U.K. ( aws25@cam.ac.uk). flexibility amongst workers. This trade-off may explain why worker polymorphism is seen in only 15% of ant species and in none of the bees or wasps (Wheeler 1991). The degree to which a species of social insect requires flexibility in task allocation will ultimately depend on its ecological context. We may therefore expect to find variation between species in the flexibility of task allocation, and hence variability in the degree to which physical castes are developed. The horned-soldier aphids (Aphidoidea: Hormaphididae: Cerataphidini) apparently represent a system with an inflexible division of labour and morphologically distinct physical castes. On their secondary host they produce two types of first-instar larvae: sterile soldiers, which do not moult to the next instar, and fertile nonsoldiers (Stern & Foster 1996). Defence is ostensibly the concern of the soldiers. They show two forms of defensive behaviour: waving of the hindlegs, possibly to ward off flying parasitoids, and attacking, which involves grasping potential predators with the forelegs often followed by thrusting of the frontal horns (Aoki & Miyazaki 1978; Sakata & Ito 1991; Schütze & Maschwitz 1991). The nonsoldiers are apparently not involved in defence (Stern & Foster 1996). Such a discrete division of labour relates to an equally discrete difference in morphology. Soldiers are larger than nonsoldiers, have longer /01/ $35.00/ The Association for the Study of Animal Behaviour

2 672 ANIMAL BEHAVIOUR, 62, 4 hindlimbs, elongated frontal horns, and more heavily sclerotized forelimbs (Noordam 1991). Consequently soldiers are better adapted to performing defensive activities than nonsoldiers, and in numerous species have been shown to attack and kill predators (e.g. Aoki & Miyazaki 1978; Schütze & Maschwitz 1991; Arakaki 1992), so that colonies with soldiers are less susceptible to predation than those without (Foster 1990). The division of reproductive and defensive labour in the horned-soldier aphids means that the ability of an aphid colony to adjust allocation to either task is apparently limited to adjusting the caste ratios of the colony as a whole (Shingleton & Foster 2000). Selection on a colony to respond more rapidly to changing conditions will push for a reduction in behavioural, and hence physical, specialization. Results from a number of studies have hinted that the division of labour in the social aphids may not be as inflexible as first appears. There have been reports of nonsoldiers engaging in legwaving activity in P. bambucicola (Ohara 1985) and attacking allospecific individuals in P. alexanderi (Aoki et al. 1981) and predators in Ceratovacuna cerbera (Aoki et al. 1999). A species of Pseudoregma similar to P. bambucicola ( P. near bambucicola ) has lost its soldier caste altogether but has attacking nonsoldier first instars. These larvae have taken on some morphological characteristics in common with the soldiers of P. bambucicola (Stern et al. 1997b). In general, as the defensive division of labour between soldiers and nonsoldiers becomes less discrete and more flexible the morphological distinctions between the castes should become less discrete also. We investigated the relationship between behaviour, morphology and caste in two closely related Pseudoregma species: P. sundanica and P. nicolaiae. Both species produce ostensibly sterile soldiers; however, P. sundanica is also defended by tending ants. These feed on the sweet honeydew produced by the aphids, which is the waste product of the aphids sugar-rich but amino-acid-poor diet of plant sap. This and other ecological differences may have influenced the defensive division of labour of P. sundanica relative to P. nicolaiae and we investigated whether this is reflected in the morphology of the physical castes in both species. Study Species METHODS We conducted the study at two sites in Peninsular Malaysia between January and March 1999: (1) P. sundanica were studied in the Ulu Gombak Valley, 30 km northeast of Kuala Lumpur; and (2) P. nicolaiae were studied in the Cameron Highlands, 200 km north of Kuala Lumpur, 50 km east of the Straits of Malacca, >1300 m above sea level. The aphids P. sundanica and P. nicolaiae feed on a wide range of ginger species (Zingiberacae), their secondary host, although on different parts of the plant: P. sundanica feeds on the stem and the base of the leaves, whilst P. nicolaiae feeds along the leaf midrib. Pseudoregma sundanica is found at lower altitudes than P. nicolaiae, and is obligatorily tended by ants. Pseudoregma nicolaiae is only rarely visited by ants, possibly because ants are uncommon at the altitudes at which it is found. Pseudoregma sundanica typically forms larger colonies than P. nicolaiae (P. sundanica: mean colony size SE= , N=150; P. nicolaiae: 51 7, N=150, Shingleton 2001). Apart from these obvious ecological differences the two species are morphologically similar, and phylogenetically closely related (Stern et al. 1997a). Both produce a morphologically distinct soldier caste. These soldiers show the defensive behaviours of leg waving and attacking, although these behaviours may not be limited to the soldier caste. Therefore, there are potentially four types of first-instar aphids: (1) soldier defenders; (2) soldier nondefenders; (3) nonsoldier defenders; and (4) nonsoldier nondefenders. Identification of Defensive Behaviour We classified individual aphids according to whether they were leg wavers and/or attackers. We induced leg waving by blowing lightly with our mouths across aphids feeding on their host plant. Aphids that responded to the first direct air blow were classified as leg wavers. We tested attacking behaviour by placing forceps smeared with the remains of a conspecific aphid immediately in front of an individual (the haemolymph of conspecific individuals stimulates attacking behaviour in P. sundanica, Schütze & Maschwitz 1991). The aphid would then either grasp at the forceps or back away. We left the forceps in front of the aphid until it showed a response, but once an aphid had grasped them or backed away it was not retested. All aphids were tested on their host plants. Stern et al. (1997b) found a similar forceps assay to be a reliable predictor of attacking behaviour in P. near bambucicola. We did the following experiment to see if the same was true in P. sundanica and P. nicolaiae. We subjected previously undisturbed aphids to the forceps assay and then placed groups of six attacking or six nonattacking larvae in a small plastic box (79 47 mm and 20 mm deep) along with a predator, either a single chrysopid, syrphid or lycaenid larva. (Predators were uncommon so it was not possible to use the same type of predator in all of the tests.) The crushed remains of a conspecific individual were also added to the box, to stimulate attacking behaviour. Over a period of 10 min, we observed the aphids and recorded the proportion that grasped the predator with their forelegs. Because the experiment was dependent on finding suitable predators, which were rare, we were only able to do it four times for forceps-attacking and nonattacking P. sundanica, five times for forceps-attacking P. nicolaiae, and three times for forceps-nonattacking P. nicolaiae. Caste Morphology and Behaviour We examined the morphology and behaviour of first-instar larvae from populations of both P. sundanica and P. nicolaiae. A population constitutes all the aphids living on a single host plant. Individuals from

3 SHINGLETON & FOSTER: DIVISION OF LABOUR IN APHIDS 673 Figure 1. Illustration of a Pseudoregma soldier with the measurements made on each individual. 15 P. sundanica and eight P. nicolaiae populations were initially classified as leg wavers or nonwavers and then tested to see if they also showed attack behaviour. We tested ca. 30% of the first instars in each population, ranging between 5 and 32 individuals. The tested aphids were then placed in 80% alcohol to kill them and for storage. They were then cleared, stained and mounted as described by Blackman & Eastop (1994). We captured digitized images of the aphids and analysed them with NIH Image (developed at the U.S. National Institutes of Health and available on the Internet at rsb.info.nih.gov/nih-image/). Figure 1 shows the measurements we made. For the morphometric analysis we initially divided individuals of each species into soldier and nonsoldier castes, with a dissecting microscope. Four P. nicolaiae and three P. sundanica individuals could not be safely classified by eye and were excluded from further analysis. We then analysed the influence of species and caste (soldier or nonsoldier) on morphology, with a multivariate GLM (general linear model), and subsequent univariate GLMs on each of the 11 dependent measurements, as well as body size (body length width). A Dunn S{idák correction was applied to the results of the 12 univariate GLMs to adjust for inflation of type I error, reducing the significance level of each test to 0.43% (Tabachnik & Fidell 1996). Subsequent multivariate GLMs investigated the relationship between behaviour and morphology in nonsoldier P. sundanica and P. nicolaiae first instars. All GLMs controlled statistically for interpopulation differences as well as overall body size (body length width), except for the univariate GLM on body size. Consequently, any variations in morphology between the morphotypes arose through relative differences in the dimensions of the body parts rather than an overall difference in size. For all the analyses we used SPSS We examined all the mounted aphids for the presence of the next-instar claws. These are the first chitinized morphological features to appear as larvae prepare to moult (Stern et al. 1997b), and so their presence or absence can be used as a guide to the age of a larva. The presence of a next-instar claw also suggests that a larva is not sterile, but can moult through to the second instar and on to adulthood. For the behavioural analysis, we examined the relationship between tendency to attack and species, caste and leg-waving behaviour. Individuals were grouped according to species, caste and leg-waving behaviour for each population. The proportion of attacking individuals in each group was compared with a single binomial GLM, equivalent to a logistic regression, with the statistical program R (available at We used similar binomial GLMs to analyse other aspects of attacking behaviour. To determine whether defensive behaviour varied within individuals we tested and then retested P. nicolaiae larvae, with the forceps assay. After initial testing of all the individuals within a population, all defensive individuals were moved to another leaf on the same host, marked with a dusting of fluorescent powder and left for 24 h before being retested. The fluorescent powder ensured that we could identify defensive individuals even if they moved around the host plant during the 24 h prior to retesting. We repeated this for seven populations, and the number of individuals transferred varied between four and 40 depending on the population size. All the measurement data were log transformed prior to analysis. This ensured that that for parametric tests, the results of evaluation of assumptions of normality, homogeneity of variance covariance matrices, linearity and colinearity were satisfactory. The model is specified for each test. All tests were two tailed. RESULTS Identification of Defensive Behaviour Almost all individuals repeated their behaviour towards the forceps when confronted with a real predator. On average 90% of the individuals that attacked the forceps also attacked the predator (P. sundanica: X SE= %, N=4 trials; P. nicolaiae: %, N=5 trials). All of the individuals that fled the forceps also fled the

4 674 ANIMAL BEHAVIOUR, 62, 4 P. nicolaiae soldier P. sundanica soldier 1 mm P. nicolaiae nonsoldier P. sundanica nonsoldier Figure 2. Soldier and nonsoldier P. nicolaiae and P. sundanica. predator (P. sundanica: N=4 trials; P. nicolaiae: N=3 trials). The correlation between behaviour towards the forceps and behaviour towards a real predator was highly significant (binomial GLM: behaviour towards predator= species+behaviour towards forceps, F behaviour towards forceps 1, 13=172, P<0.0001). Caste Morphology and Behaviour We collected and identified 65 soldier and 126 nonsoldier P. nicolaiae and 74 soldier and 116 nonsoldier P. sundanica. Figure 2 illustrates castes of both species. We remeasured all 11 body parts of seven individuals to estimate measurement error, which was between 0.2 and 3.1% of the mean value for a body part (Table 1). The first multivariate GLM analysis indicated that soldiers were morphologically distinct from nonsoldiers, and P. sundanica from P. nicolaiae, when controlling for body size and variation between populations (Table 2). The univariate GLMs confirmed that soldiers had longer frontal horns, longer and thicker forelimbs, longer hindlimbs, more elongated bodies and were larger than nonsoldiers (Table 1). Across both castes P. nicolaiae first instars had longer limbs, more elongated bodies and were

5 SHINGLETON & FOSTER: DIVISION OF LABOUR IN APHIDS 675 Table 1. Output of GLM analyses of 11 body measurements, controlling for body size and differences between populations, and of GLM analysis of body size, controlling for differences between populations Group means (mm or mm 2 ) Univariate tests: source of variance Variable Measurement error (%) P. nicolaiae P. sundanica Nonsoldier Soldier Nonsoldier Soldier Caste Species Caste species F 1,353 F 1,353 F 1,353 Horn length ± ± ± ± ** ** Last rostral segment length ± ± ± ± ** ** 14.72** Forefemur length ± ± ± ± ** ** 33.48** Maximum forefemur width ± ± ± ± ** 13.60** 2.37 Minimum forefemur width ± ± ± ± ** ** 0.80 Foretibia length ± ± ± ± ** ** 67.10** Second femur width ± ± ± ± ** Hindfemur length ± ± ± ± ** ** 63.96** Hindtibia length ± ± ± ± ** ** 87.52** Body length ± ± ± ± ** ** 32.25** Body width ± ± ± ± ** ** 33.30** Body size (length width) 0.618± ± ± ± * 60.23** 20.59** Means are given±se. For body size the dfs of the F ratios are 1,354. *P<0.0043; **P<0.001.

6 676 ANIMAL BEHAVIOUR, 62, 4 Table 2. Output of multivariate GLM analysis of 11 body measurements (morphology=population+body size+caste+species+caste species) Source of variance Wilks lambda (Λ) F 11,316 P Caste < Species < Caste species < larger than P. sundanica first instars (Table 1). However, there was a significant interaction between the effects of caste and species on morphology (Table 2). This suggests that the morphological differences between castes were less in one species than in the other. Inspection of the group means and the output of the univariate GLMs (Table 1) reveal that this was indeed the case. For certain traits (rostrum length, limb length, body shape and body size, see Table 1) P. nicolaiae soldiers and nonsoldiers were more similar than P. sundanica soldiers and nonsoldiers. Overall, the two P. nicolaiae castes were morphologically more similar than the two P. sundanica castes (Mahalanobis distance: P. nicolaiae soldier and nonsoldier=21.141, P. sundanica soldier and nonsoldier= , Manly 1986). In both species, all soldiers were leg wavers. However, of the 126 nonsoldier P. nicolaiae 51 were also leg wavers. Similarly, 36 of the 116 nonsoldier P. sundanica were leg wavers. These leg wavers were not simply wrongly classified as soldiers, as in both P. sundanica and P. nicolaiae they were morphologically distinct from soldiers (multivariate GLM: morphology=population+body size+caste; P. sundanica: Wilks lambda Λ caste 11, 76 =0.154, P<0.0001; P. nicolaiae: Wilks lambda Λ caste 11, 87 =0.115, P<0.0001). In P. nicolaiae there was no difference between the morphologies of nonsoldier leg wavers and nonwavers (multivariate GLM: morphology=population+body size+ behaviour; Wilks lambda Λ behaviour 11, 97 =0.938, P= 0.842) but there was in P. sundanica (multivariate GLM: morphology=population+body size+behaviour; Wilks lambda Λ behaviour 11, 83 =0.743, P=0.004) in which nonsoldier leg wavers had significantly longer fore- and hindlimbs than nonwavers (univariate GLMs: body part=population+body size+behaviour; F behaviour 1, 92 > , P<0.0043, for fore- and hindtibia and femur lengths). No soldier was found with evidence of a next-instar claw, suggesting soldiers are sterile. The number of individuals with chitinized next-instar claws did not differ significantly between defending and non-defending Percentage of attackers Soldiers P. sundanica P. nicolaiae nonsoldiers in either species (chi-square: P. sundanica: χ 2 1<0.001, P=0.982; P. nicolaiae: χ 2 1=2.51, P=0.113; Table 3). This suggests no difference in age between defending and nondefending nonsoldiers, although a small sample size and lack of power may have hidden any effect. The tendency to attack depended on species, caste and whether an individual was a leg waver or not, with a significant three-way interaction between the effects of all three factors on attack behaviour (binomial GLM: attacking behaviour=species caste leg-waving behaviour, F interaction 5, 39 =23.056, P<0.0001). Figure 3 shows the mean percentage of attacking individuals according to species, caste and leg-waving behaviour for the populations sampled. Overall, P. nicolaiae first instars were more likely to attack than P. sundanica first instars and leg wavers more likely to attack than nonwavers. In P. sundanica attack behaviour was almost exclusively restricted to the soldier caste. Only four nonsoldiers showed attacking behaviour, and these were all leg wavers from the same population. In contrast P. nicolaiae nonsoldiers often showed attack behaviour, and defensive behaviour in soldier and leg-waving nonsoldier P. nicolaiae was almost identical. Therefore in behaviour, as with morphology, P. nicolaiae soldiers and nonsoldiers appeared to be more similar than P. sundanica soldiers and nonsoldiers. There was some variability in defensive behaviour within individuals tested then retested with the forceps assay, after 24 h. Of the seven populations tested, all the defensive individuals from one population disappeared before retesting. In the remaining six populations, the proportion of first instars that retained their attacking 8 6 Nonsoldier leg wavers 8 8 Nonsoldier nonwavers Figure 3. The mean±se percentage of attackers amongst soldiers, nonsoldier leg wavers and nonsoldier nonwavers from populations of P. sundanica and P. nicolaiae. N=number of populations used to calculate the mean percentages. Table 3. Number of individuals with and without next-instar claws in P. sundanica and P. nicolaiae nonsoldier first-instar larvae P. nicolaiae P. sundanica Type of nonsoldier With claws Without claws With claws Without claws Leg waver Nonwaver

7 SHINGLETON & FOSTER: DIVISION OF LABOUR IN APHIDS 677 behaviour was significantly lower in nonsoldiers than in soldiers (percentage retaining attacking behaviour: soldiers: 100%; nonsoldiers: X SE= %: binomial GLM: attack behaviour=caste, F caste 1, 10 =10.59, P=0.009). This suggests that nonsoldiers change their defence behaviour more than soldiers. DISCUSSION Our results establish that the caste structure of social aphids can be unexpectedly intricate and hint at a complexity comparable with that of some of the more familiar social Hymenoptera. In both aphid species, the division of labour initially appeared to be based on physical castes, with a simple relationship between behaviour and morphology. Soldiers were more likely to leg-wave and attack than nonsoldiers, and had elongated limbs and horns and thickened forelimbs. There were, however, clear behavioural and morphological differences between the caste structures of P. nicolaiae and P. sundanica. First, the division of labour was less distinct between the castes of P. nicolaiae: morphologically and behaviourally P. nicolaiae soldiers and nonsoldiers were more similar than P. sundanica soldiers and nonsoldiers. Second, both P. nicolaiae first-instar castes were more aggressive than their P. sundanica counterparts: P. nicolaiae soldiers and nonsoldiers were more likely to attack and were larger with longer limbs. The differences in caste distinction between P. sundanica and P. nicolaiae appear to relate to differences in the trade-off between physical specialization and behavioural plasticity. In general, a colony of social insects will show greater flexibility in the tasks it performs if the division of labour within it is not fixed but individuals are able to change their behaviour over time (Oster & Wilson 1978; Bourke & Franks 1995). Such flexibility allows a colony to cope more rapidly with fluctuating environments and corresponding changes in its requirements (Tofts & Franks 1992; Gordon 1996; Wakano et al. 1998). However, as individuals take up a range of tasks, morphological adaptations to just one of those tasks will be selected against. The efficiency of having polymorphic specialists is therefore replaced by the flexibility of having generalists, and the physical distinctions between castes should be reduced and ultimately removed. This appears to be the case in P. nicolaiae relative to P. sundanica. In P. nicolaiae, soldiers and nonsoldiers are behaviourally and morphologically more similar than in P. sundanica. A complete loss of the physical caste system is seen in P. near bambucicola, as described by Stern et al. (1997b). These aphids are very closely related to soldier-producing P. bambucicola but have monomorphic first instars, some of which show defensive behaviour. All of the first instars have defensive morphology relative to the nonsoldier first instars of P. bambucicola. The defensive division of labour in P. near bambucicola and in P. nicolaiae may be one that is flexible amongst individuals and more similar to that of the social Hymenoptera. Here individuals tend to change the tasks they perform with time, a phenomenon termed temporal polyethism. In many Hymenoptera this temporal polyethism is correlated with age (Dew & Michener 1981; Seeley 1982; Badertscher et al. 1983; Hölldobler & Wilson 1990). In P. near bambucicola there is some evidence that attacking nonsoldiers are older than nonattacking nonsoldiers (Stern et al. 1997b). We found no evidence of this, however, in P. nicolaiae. The defensive division of labour amongst nonsoldiers may be based on the more general foraging for work algorithm (Tofts & Franks 1992; Sendova-Franks & Franks 1993; Tofts 1993). In this scheme, individuals are recruited to tasks as required, with the division of labour driven more by the environment than either age or physical caste (although certain individuals may still show a tendency to perform one task over another). We may therefore expect nonsoldier P. nicolaiae to become more likely to attack under conditions of high predation. In contrast, the division of labour in P. sundanica seems to be one that is built on inflexible physical castes, with populations primarily adjusting their levels of self defence by adjusting the number of soldiers they produce (Shingleton & Foster 2000). Even leg-waving behaviour, whilst being evident in nonsoldier first instars, seems to be associated with certain physical characteristics. Leg-waving nonsoldiers had longer limbs than nonwavers, suggesting the existence of a third physical caste specialized only for leg waving. The aphid P. alexanderi also possesses three firstinstar castes with both major and minor soldiers (Aoki & Miyazaki 1978), although the behavioural characteristics of these castes have not been well studied. There are two aspects of the ecology of P. sundanica and P. nicolaiae that may explain why P. nicolaiae requires greater behavioural flexibility. Firstly, P. nicolaiae typically forms smaller populations than P. sundanica. Ant species with smaller colony sizes are generally thought to be less likely to show physical polyethisms (Oster & Wilson 1978; Sudd & Franks 1987). This is because in large colonies the majority of tasks in the complete repertoire of tasks are likely to be present at any one time. In small colonies random fluctuations in a colony s requirements may make some tasks, and hence some specialized castes, redundant. This would make behavioural flexibility amongst workers selectively advantageous (Bourke & Franks 1995). This argument may be applied to the horned-soldier aphids. If smaller populations show greater fluctuations in predation levels than large populations, this would select for behavioural plasticity in P. nicolaiae relative to P. sundanica. There is some evidence that this may be true. In P. bambucicola large populations are more likely to have a predator present than small populations (Shibao 1998) and therefore will usually have need for specialized soldiers. For small populations, long periods without predator attack reduce the selective advantage of such specialization. The second ecological factor that may favour behavioural flexibility in P. nicolaiae relates to ant tending. Pseudoregma sundanica is obligatorily tended by ants, which defend the aphids from natural predators (Schütze & Maschwitz 1991). Ant tending appears to be the primary form of defence in this species, and soldier production is inversely related to the level of tending they

8 678 ANIMAL BEHAVIOUR, 62, 4 receive (Shingleton & Foster 2000). Because ants actively remove many predators (Schütze & Maschwitz 1991), they may act to suppress fluctuations in predation levels, and reduce the selective pressure to maintain behavioural flexibility amongst castes. In contrast, P. nicolaiae, without the protection of ants, may need to adjust levels of self-defence more rapidly by recruiting nonsoldiers to defensive tasks. Ant tending may also account for the second phenomenon observed in our data; that of increased aggressiveness in both castes of P. nicolaiae relative to P. sundanica. Pseudoregma nicolaiae is rarely visited by ants and must rely on self-defence to prevent predation. It has aggressive soldiers and attacking nonsoldiers. In contrast, P. sundanica shows a relative deinvestment in self-defence in favour of ant defence. Attacking behaviour is largely restricted to the soldiers, which are less aggressive and only attack predators not first removed by ants (Schütze & Maschwitz 1991). The hypothesis that P. nicolaiae is more reliant on self-defence is further supported by patterns of soldier production in both species. In P. sundanica, soldiers are produced only in large populations, typically of more than 70 individuals, when the number of ants per aphid is low. In contrast P. nicolaiae begins to produce soldiers in populations as small as 10 individuals (unpublished data). A reliance on self-defence in P. nicolaiae may also make behavioural flexibility among nonsoldiers selectively advantageous, for reasons other than responsiveness to environmental change. The data suggest that soldiers are sterile and soldier production will therefore be reproductively costly for an aphid clone. Having fertile nonsoldier attackers may allow P. nicolaiae to reduce the reproductive cost of self-defence. In summary, the differences in caste behaviour and morphology appear to relate to differences in ecology, particularly in their association with ants. In general, the data suggest that the caste structure of the social aphids may be less well defined than first appears. The division of labour between physical castes, and consequently the morphological distinctiveness of those castes, may vary considerably between species, in relation to ecological conditions. Acknowledgments We thank C. Braendle, C. Klingenberg, N. Pike, D. Stern and two anonymous referees for their discussions and for critical comments on the manuscript. We thank L. Kirton and U. Maschwitz for their assistance during our stay in Malaysia. We thank the Government of Malaysia for permission to conduct research in Malaysia (U.P.E. Ref. 40/200/19 SJ.627), and the University of Malaysia, Kuala Lumpur for the use of their field station. A. Shingleton is supported by the B.B.S.R.C. References Aoki, S. & Miyazaki, M Notes on the pseudoscorpion-like larvae of Pseudoregma alexanderi (Homoptera, Aphidoidea). Kontyû, 46, Aoki, S., Akimoto, S. & Yamane, S Observations on Pseudoregma alexanderi (Homoptera, Pemphigidae), an aphid species producing pseudoscorpion-like soldiers on bamboos. Kontyû, 49, Aoki, S., Kurosu, U., Shin, K. Y. & Choe, J. C A new soldier-producing species of Ceratovacuna (Homoptera: Aphididae, Hormaphidinae) on Sasa spp. from Japan and Korea. Entomological Science, 2, Arakaki, N Shortened longevity of soldiers of the bamboo aphid Pseudoregma koshunensis (Takahashi) (Homoptera, Aphididae) due to attack behavior. Journal of Ethology, 10, Badertscher, S., Gerber, C. & Leuthold, R. H Polyethism in food supply and processing in termite colonies of Macrotermes subhyalinus (Isoptera). Behavioral Ecology and Sociobiology, 12, Blackman, R. L. & Eastop, V. F Aphids on the World s Trees. Wallingford, Oxon: CAB International. Bourke, A. F. G. & Franks, N. R Social Evolution in Ants. Princeton, New Jersey: Princeton University Press. Calabi, P. & Traniello, J. F. A Behavioural flexibility in age castes of the ant Pheidole dentata. Journal of Insect Behavior, 2, Dew, E. H. & Michener, C. D Division of labor among workers of Polistes metricus (Hymenoptera, Vespidae): laboratory foraging activities. Insectes Sociaux, 28, Foster, W. A Experimental evidence for effective and altruistic colony defense against natural predators by soldiers of the gall-forming aphid Pemphigus spyrothecae (Hemiptera, Pemphigidae). Behavioral Ecology and Sociobiology, 27, Gordon, D. M Dynamics of task switching in harvester ants. Animal Behaviour, 38, Gordon, D. M The organization of work in social insect colonies. Nature, 380, Hölldobler, B. & Wilson, E. O The Ants. Cambridge, Massachusetts: The Belknap Press of Harvard University Press. Manly, B. F. J Multivariate Statistical Methods: a Primer. London: Chapman & Hall. Noordam, D Hormaphidinae from Java (Homoptera: Aphididae). Leiden: National Natuurhistorisch Museum. Ohara, K Observations on the predator-prey relationship between Pseudoregma bambucicola (Homoptera, Pemphigidae) and Metasyrphus confrater (Diptera, Syrphidae), with special reference to the behaviour of the aphid soldier. Esakia, 23, Oster, G. F. & Wilson, E. O Caste and Ecology in the Social Insects. Princeton, New Jersey: Princeton University Press. Sakata, K. & Ito, Y Life-history characteristics and behavior of the bamboo aphid, Pseudoregma bambucicola (Hemiptera, Pemphigidae), having sterile soldiers. Insectes Sociaux, 38, Schütze, M. & Maschwitz, U Enemy recognition and defense within trophobiotic associations with ants by the soldier caste of Pseudoregma sundanica (Homoptera, Aphidoidea). Entomologia Generalis, 16, Seeley, D. S Adaptive significance of the age polyethism schedule in honey bee colonies. Behavioral Ecology and Sociobiology, 11, Sendova-Franks, A. & Franks, N. R Task allocation in ant colonies within variable environments (a study of temporal polyethism: experimental). Bulletin of Mathematical Biology, 55, Shibao, H Social structure and the defensive role of soldiers in a eusocial bamboo aphid, Pseudoregma bambucicola (Homoptera: Aphididae): a test of the defence-optimization hypothesis. Researches on Population Ecology, 40, Shingleton, A. W The evolution of ant aphid interactions. Ph.D. thesis, University of Cambridge.

9 SHINGLETON & FOSTER: DIVISION OF LABOUR IN APHIDS 679 Shingleton, A. W. & Foster, W. A Ant-tending influences soldier production in a social aphid. Proceedings of the Royal Society of London, Series B, 267, Stapley, L Physical worker castes in colonies of an acacia-ant (Crematogaster nigriceps) correlated with an intra-colonial division of defensive behaviour. Insectes Sociaux, 46, Stern, D. L. & Foster, W. A The evolution of soldiers in aphids. Biological Reviews of the Cambridge Philosophical Society, 71, Stern, D. L., Aoki, S. & Kurosu, U. 1997a. Determining aphid taxonomic affinities and life cycles with molecular data: a case study of the tribe Cerataphidini (Hormaphididae:Aphidoidea: Hemiptera). Systematic Entomology, 22, Stern, D. L., Whitfield, J. A. & Foster, W. A. 1997b. Behavior and morphology of monomorphic soldiers from the aphid genus Pseudoregma (Cerataphidini, Hormaphididae): implications for the evolution of morphological castes in social aphids. Insectes Sociaux, 44, Sudd, J. H. & Franks, N. R The Behavioral Ecology of Ants. New York: Chapman & Hall. Tabachnik, B. G. & Fidell, L. S Using Multivariate Statistics. New York: Harper Collins. Tofts, C Algorithms for task allocation in ants (a study of temporal polyethism: theory). Bulletin of Mathematical Biology, 55, Tofts, C. & Franks, N. R Doing the right thing: ants, honeybees and naked mole-rats. Trends in Ecology and Evolution, 7, Wakano, J. Y., Nakata, K. & Yamamura, N Dynamic model of optimal age polyethism in social insects under stable and fluctuating environments. Journal of Theoretical Biology, 193, Wetterer, J. K Forager polymorphism, size-matching, and load delivery in the leaf-cutting ant, Atta cephalotes. Ecological Entomology, 19, Wheeler, D. E The developmental basis of worker caste polymorphism in ants. American Naturalist, 138, Wilson, E. O Caste and division of labor in leaf-cutter ants (Hymenoptera: Formicidae: Atta), II: The ergonomic optimization of leaf cutting. Behavioral Ecology and Sociobiology, 7, Wilson, E. O The causes of ecological success: the case of the ants. Journal of Animal Ecology, 56, 1 9.