- Email: [email protected]
Leaf-cutting ants tease optimal foraging theorists
TREE vol. 8, no. IO, October
to be a vigorous debate between those who favor ecological models for sex, and those who favor mutational models. Whatever the outcome of this debate, it is clear that Hamilton was onto something when he recognized the possibility of periodicities generated by life itself.
References 1 Hamilton, W.D. (1975) 0. Rev. Biol, 50, 175-180 2 Williams,
G.C. (1975) Sex and Evolution, Princeton University Press 3 Levin, D. (1975) Am. Nat, 109,437-451 4 Clarke, B. (1976) in Genetic Aspects of Host-Parasite Relationships (Taylor, A.E.R. and Muller, Ft., eds), pp. 87-103,
Leafxutting AntsTease Optimal Foraging Theorists Alex Kacelnik THE DEVELOPMENT of scientific knowledge is not unlike the process of individual learning: the essence is surprise, or discrepancy between expectation and outcome. Individual animals learn more when experienced events are unexpected. Similarly, unexpected scientific observations promote more progress than those which fit common sense, and common sense in behavioural ecology has to do with fitness. When an animal’s behaviour maximizes its fitness, expectations are satisfied. When it does not, expectations are violated and research is stimulated. This is why behavioural ecologists have paid attention to paradoxical observations such as altruism in social insects, reproductive suppression in socially living mammals, the development of bizarre morphological features in the males of many species, or superparasitism in parasitoids. Failure to comprehend this role for theory as a guide for action has led to two symmetric misunderstandings. Many adaptationists emphasise the success of optimality hypotheses to the detriment of awkward facts that refute them. In the opposite camp, the adaptationist program has been criticized as tautological precisely because empirical mismatches lead to reformulation of models and further proposals for research. Both naive adaptationists and their critics are misguided: the former ought to defend their approach by advertising reasonable models that fail, and their critics ought to accept that given that
Alex Kacelnik is at the Dept of Zoology, University of Oxford, South Parks Road, Oxford, UK OX1 3PS.
346
optimality models do fail and are modified the approach could not be construed as being circular. My perverse delight in the failure of optimality models has been recently rewarded by observations on the foraging behaviour of the leafcutting ants Acromyrmex lundi by Races and Nuriez’. Leaf-cutting ants (190 New World species of the myrmicine tribe Attini) depend on the growth of a fungus that they cultivate in ‘gardens’ supplied with leaf segments brought into the colonies by workers. Adult workers are fundamentally nectar feeders, predators and scavengers, but the larvae feed exclusively on the tips of hyphae, or ‘gongylidia’, which are rich in nutrients -the ants are incapable of extracting these nutrients directly from the leaves. The fungi, in turn, depend on their symbiotic ants for the provision of substrate, for dispersion and for recycling of some of their enzymes2. This symbiotic relationship extends from the basic physiology to behavioural adaptations. The ants’ fitness depends on the growth of the fungi, and this depends on the supply of substrate, so that the foraging behaviour of the ants is likely to be adapted for efficient substrate supply, both in quality and in quantity. Ant workers are central place foragers3, since they deliver resources to the colony’s garden, and they may be expected to maximize rate of resource delivery to the central place. In the case of nonsocial foragers, such as birds supplying their nestlings, this currency has had its However, for eusocial successes4. insects, the problem is deeper and more interesting because the maximizing agent is the colony and not
1993
Blackwell Scientific 5 Jaenike, J. (1978) Evol. Theor. 3,191-194 6 Bell, G. (1982) The Masterpiece of Nature: The Evolution and Generics of Sexuality, University of California Press 7 Jaenike, J. (1993) Evol. Ecol. 7, 103-108 6 Jaenike, J. (1992) Am. Nat. 139, 893-906 9 Hori, M. (1993) Science 260,218-219
necessarily each individual worker. Indeed, NuAez5 and others have shown that honeybees (Apis mellifera) do not maximize gross rate of nectar delivery, because they return to their hives from nondepleting nectar sources with partially filled crops, Nuriez suggested that the bees’ behaviour could be understood by optimizing information transfer: individual bees may sacrifice rate of delivery to reduce intervals between recruiting episodes, which could result in higher overall gains for the colony. However, further analysis6 showed that the bees’ behaviour was also compatible with modified hypotheses about individual worker maximization, leaving the issue unresolved. Leaf-cutting ants may help in rethinking the problem. Leaf-cutting ants usually follow trails towards suitable leaf sources, but some workers (scouts) wander outside the established tracks, and upon the detection of new suitable leaf sources return to the colony laying new pheromone trails. Other workers, if recruited, travel to the new food source and proceed to collect leaf material. Races and NuAez’ used an elegant laboratory technique to examine the information transmitted during recruitment. They first exposed scouts to droplets of scented sugar solution of either of two concentrations, 1% or 10%. Scouts detected these droplets and returned to the colony leaving a recruiting trail. When the recruits arrived, they encountered no nectar but sheets of parafilm embedded in the same scent but containing no sugar. Since all recruits found the same material, differences between the groups must be the result of information transmitted by the scouts. I shall refer to workers recruited by scouts who found the 10% solution as ‘10% workers’ and to workers recruited by scouts who had found 1% solution as ‘1% workers’, but it should remain clear that both groups encountered sugar-free parafilm at the source.
Q 1993, Elsevier Science Publishers
Ltd (UK)
TREE vol. 8, no. IO, October
1993 +
go0
Races and Nunez’ report several effects of scouts’ experience on workers’ behaviour. 10% workers (1) cut smaller sections of parafilm, (2) returned to the colony at a faster pace and (3) displayed more active recruiting behaviour than 1% workers by marking the substrate. When, in another experiment, recruits found pieces of parafilm of a homogeneous size, they still travelled faster if they had been recruited by 10% scouts. Thus, differences in velocity were not caused by differences in burden. Greater velocity did not compensate for the reduction in fragment size, and 10% workers had a lower rate of leaf transport during travel than 1% workers (Fig. 1). Overall rate of delivery, however, depends on the time spent cutting the leaf fragments and processing them in the colony as well as on rate of transport, and this information is not available. These results contradict the most obvious functional hypotheses. Let us examine some possibilities. Delivery rate maximization The strategy leading to maximum rate of leaf delivery is independent of real or expected sugar content. Since 10% and 1% workers behaved differently, they could not have maximized delivery rate in both treatments. They could however have been maximizing a compromise between rate of leaf delivery and some unknown cost correlated with travel velocity. If such a cost exists, workers might pay more of it (i.e. walk faster to reduce travel time) if the value of the leaves is greater. Thus they would achieve a higher delivery rate for higher-quality leaves. Ten-percent workers did travel faster and experienced shorter travel times, even for equal-sized fragments, but why did they carry smaller loads? Delivery rate maximization predicts smaller loads for shorter travel times if fragment size is a concave function of cutting time. Races and Nunez do not report this function, but a fragment’s size is a quadratic function of its radius while cutting time is likely to be proportional to its semi-circumference and thus is linear to its radius. The expected size versus cutting time function is thus convex, the opposite of the required relation. These speculations will hopefully become redundant when the real functions for cutting and colony times versus fragment size are known so that delivery rate can be compared between treatments, but it appears that not even postulating an imaginary cost for velocity could save this hypothesis.
local contrast Workers may detect the contrast between the scouts’ information and the actual value of the resource. Applying this hypothesis to invertebrates is admittedly far-fetched, but it does predict a lower delivery rate by 10% workers. Ten-percent workers find a greater discrepancy between the scouts’ information and the test condition. This greater contrast between ‘expectation’ and experience may enhance detection of the experimental ploy, leading to more reluctant foraging on the part of the 10% recruits. Faster travel from the colony to the source would reflect higher ‘expectation’, while smaller burden and faster return velocity would reflect lower motivation or greater urgency to spend foraging time somewhere else. However, 10% workers showed higher trail marking activity, so that for this hypothesis to be correct they ought to have been laying a negativemessage pheromone rather than a recruiting one. This is unlikely. Just as with rate maximization, this hypothesis is directly testable: differences between treatments should disappear if workers find the same concentration as their own scouts, regardless of the scouts’ experience. Informational transfer Races and Nuriez’ preferred explanation is based on putative recruitment strategies by the workers, and it is also based on a trade-off between individual delivery rate and another currency. The idea is that workers sacrifice individual leaf delivery rate to return earlier to the colony for further recruitment. In contrast with my modified version of the rate hypothesis, it assumes that individual delivery rate is lower for higher-quality leaves, but shorter times outside the colony enhance recruitment leading to higher rate of leaf gathering by the colony. This hypothesis is consistent with the increased velocity and with the reduced leaf size in 10% workers. By cutting smaller loads 10% workers save cutting time and by increasing velocity they save travel time. This hypothesis can also accommodate the more intense trail marking by 10% workers. However, there are still some difficulties: workers of both treatments found equal (worthless) parafilm sheets, and thus the different recruitment behaviour is just an amplification of the differences in the scouts’ findings, with no correction for the poor foraging source encountered by the workers. This would result in a positive feed-
”
Burden Fig. 1. Load transport rate (load mass x velocity) of sugarfree parafilm fragments for individual ants as a function of burden [(ant mass + load massMant mass)]. Filled symbols (0) indicate workers recruited by scouts that had encountered 10% sugar solution and empty symbols (0) workers recruited by scouts who had found 1% sugar solution. Ten-percent workers carry smaller burdens, and, although they walk faster (not shown in the figure), this is enough to generate lower transport rates. From Ref. I, wifh permission.
back process that would lead the colony to progressively concentrate on once good but presently worthless food sources. One should expect workers to correct the information received from the scouts, but other pheromone-related positive feedback phenomena do occur in ants in the context of colony defence2 and if confirmed, this would add another intriguing observation. This hypothesis, the strongest so far, requires confirmation of the reduction of individual delivery rate with resource quality, a prediction that opposes that of the rate hypothesis and that will be tested by the same information. It is satisfying that for the time being there is no winning functional model, although the information transfer hypothesis is in my view the strongest candidate. What is the way ahead? Some critics of the adaptationist stand may reject further optimality tinkering, thinking that animals have been shown to be nonoptimal yet again. Naive adaptationists may brush the difficulties under the carpet by claiming that the artificial nature of the laboratory conditions creates behavioural attefacts. I would argue that a more promising attitude is to rise to the challenge implied in these results of more precise testing of these and other possible optimality models using further detailed measurements of the ants’ behaviour. Many empirical issues remain open. Knowledge of the interactions between leaf size and cutting time and colony time versus leaf size is crucial to calculate individual delivery rates. The time course of the effects of treatment could elucidate the possible positive feedback on recruitment. The control of recruits’ behaviour must be analysed; they could respond to information received in the colony or to the pheromone laid by the scouts in the proximity of the food source. These facts could be incorporated 347
news &Go~~ent
into formal models of individual and colony foraging, so that the putative trade-off between loss of individual performance and increase in recruitment may be assessed. The general message is that future progress depends on linking theory and facts. Neither abstract optimality theorizing nor purely descriptive fact reporting will solve these puzzles, but both are required. The dialectical exchange between pre-
TREE
cise theoretical expectations and real-life phenomena promotes progress, and progress is faster and more fun when models and facts clash.
1 Races, F. and Nuriez, Behav. 45, 135-143
J.A.
(1993)
Aim.
2 Holldobler, 6. and Wilson, E.O. (1990) The Ants, Springer Verlag
FatTimes, Lean Times andCompetition among Predators John A. Wiens CLASSICAL COMPETITION THEORY predicts that coexisting species that share limiting resources should compete. For coexistence to continue, the species should diverge in resource use, reducing niche overlap’,2. This view of nature, which was prevalent in ecology during the 1960s and early 197Os, has been questioned on several counts3,4. The effects of environmental variation on resource levels and competition have received particular attention. If environmental conditions vary (as they do), then resources such as food may be abundant at some times (‘fat’ times) and scarce at others (‘lean’ times). Accordingly, we might expect niche overlap among species to be greater during the fat times than the lean times, when competition is presumably more severe2*3*5. In the early 198Os, Schoeneti tested this expectation by surveying some 30 studies in which changes in niche overlap among coexisting species had been documented between seasons or years. In most cases, overlap was indeed less during the relatively lean period, usually the winter. Does this mean that the species are competing during the lean times but not during the fat times? Not necessarily. Competition is related to resource limitation, and without direct information about resource levels, terms such as ‘fat’ or ‘lean’
John Wiens is at the Dept of Biology and Graduate Degree Program in Ecology, Colorado State University, Fort Collins, CO 80523, USA. 348
tell one little about whether or not a resource-limitation threshold has been passed (Fig. la). Also, niche overlap may not change monotonically with a reduction in resource supplies. Coexisting species may respond opportunistically to certain resources when they are superabundant, specialize as resources become more limiting, but then converge again as resources become extremely scarce, leading to a high-low-high sequence of niche overlaps (Fig. 1 b). A recent report by Jaksic and his colleagues6 at least partially addresses these problems, and shows that the ‘fat-lean’ scenario may be an oversimplification. Jaksic et al. studied the food habits and guild structure of predatory vertebrates over a 4-year period in semi-arid scrub desert north of Santiago, Chile. These predators (four owl species, four falconiform hawk species and two foxes) are well-suited to an investigation of niche dynamics. They are large and conspicuous and habitually use sites such as roosts or dens where their pellets or feces may be collected, and remains of prey in these castings can be identified with some precision’. This is important, for the determination of the trophic guild structure of a community is sensitive to the level to which prey items can be resolved taxonomicallp. In this study, prey remnants in the castings were identified to species for vertebrate prey and to orders for invertebrate prey - not ideal resolution, but better than that in most such studies.
8, no.
10, October
1993
3 Orians, G.H. and Pearson, N.E. (1979) in Analysis of Ecological Systems (Horn, D.J., Stairs, G.R. and Mitchell, PD., eds), pp. 154-177, Ohio University Press 4 Kacelnik, A. (1984) Anim. Behav. 53, 283-299 5 Nutiez, 139-150
References
vol.
J.A. (1982) J. Apic.
Res.
21,
6 Kacelnik, A., Schmid-Hempel, P. and Houston, A.I. (1986) Behav. Ecol, Sociobiol.
19, 19-24
In this system, bird and arthropod prey populations showed the expected pattern of an increase in abundance associated with the flush of vegetation growth at the onset of the breeding period and a decline at the end of this season - breeding and nonbreeding seasons were relatively ‘fat’ and ‘lean’ for consumers of these prey. Small mammal populations, on the other hand, showed no such changes. Instead, they irrupted early in the study, apparently in response to unusually high rainfall and vegetation productiong. Numbers declined dramatically throughout the remainder of the study, and estimated densities at the lowest level were only 7% of those recorded at the abundance peak. For predators on small mammals, then, ‘fat’ and ‘lean’ were expressed on a scale of years, not seasons. Under these conditions, one might anticipate that omnivorous predators should shift diets seasonally, reducing overlap during the nonbreeding season, whereas predators on mammals might emigrate or shift to other prey as the mammal populations crashed, increasing overlap with the omnivores. However, despite the fact that nearly every predator species displayed a different response to the seasonal and yearly changes in food abundance, the basic guild structure of the assemblage was stable. One mammal-eating guild, containing the owls Tflo and Bubo, was always present, and these species exhibited high dietary overlap. During the few survey periods when Tyto was not recorded, Bubo remained isolated in this guild. Neither species showed any dietary shifts in response to the decline in small mammal abundance. A second group (two other owls and the two fox species) formed a tight omnivorous feeding guild based on arthropods and a broad range of
0 1993, Elsevier Science Publishers
Ltd (UK1
to be a vigorous debate between those who favor ecological models for sex, and those who favor mutational models. Whatever the outcome of this debate, it is clear that Hamilton was onto something when he recognized the possibility of periodicities generated by life itself.
References 1 Hamilton, W.D. (1975) 0. Rev. Biol, 50, 175-180 2 Williams,
G.C. (1975) Sex and Evolution, Princeton University Press 3 Levin, D. (1975) Am. Nat, 109,437-451 4 Clarke, B. (1976) in Genetic Aspects of Host-Parasite Relationships (Taylor, A.E.R. and Muller, Ft., eds), pp. 87-103,
Leafxutting AntsTease Optimal Foraging Theorists Alex Kacelnik THE DEVELOPMENT of scientific knowledge is not unlike the process of individual learning: the essence is surprise, or discrepancy between expectation and outcome. Individual animals learn more when experienced events are unexpected. Similarly, unexpected scientific observations promote more progress than those which fit common sense, and common sense in behavioural ecology has to do with fitness. When an animal’s behaviour maximizes its fitness, expectations are satisfied. When it does not, expectations are violated and research is stimulated. This is why behavioural ecologists have paid attention to paradoxical observations such as altruism in social insects, reproductive suppression in socially living mammals, the development of bizarre morphological features in the males of many species, or superparasitism in parasitoids. Failure to comprehend this role for theory as a guide for action has led to two symmetric misunderstandings. Many adaptationists emphasise the success of optimality hypotheses to the detriment of awkward facts that refute them. In the opposite camp, the adaptationist program has been criticized as tautological precisely because empirical mismatches lead to reformulation of models and further proposals for research. Both naive adaptationists and their critics are misguided: the former ought to defend their approach by advertising reasonable models that fail, and their critics ought to accept that given that
Alex Kacelnik is at the Dept of Zoology, University of Oxford, South Parks Road, Oxford, UK OX1 3PS.
346
optimality models do fail and are modified the approach could not be construed as being circular. My perverse delight in the failure of optimality models has been recently rewarded by observations on the foraging behaviour of the leafcutting ants Acromyrmex lundi by Races and Nuriez’. Leaf-cutting ants (190 New World species of the myrmicine tribe Attini) depend on the growth of a fungus that they cultivate in ‘gardens’ supplied with leaf segments brought into the colonies by workers. Adult workers are fundamentally nectar feeders, predators and scavengers, but the larvae feed exclusively on the tips of hyphae, or ‘gongylidia’, which are rich in nutrients -the ants are incapable of extracting these nutrients directly from the leaves. The fungi, in turn, depend on their symbiotic ants for the provision of substrate, for dispersion and for recycling of some of their enzymes2. This symbiotic relationship extends from the basic physiology to behavioural adaptations. The ants’ fitness depends on the growth of the fungi, and this depends on the supply of substrate, so that the foraging behaviour of the ants is likely to be adapted for efficient substrate supply, both in quality and in quantity. Ant workers are central place foragers3, since they deliver resources to the colony’s garden, and they may be expected to maximize rate of resource delivery to the central place. In the case of nonsocial foragers, such as birds supplying their nestlings, this currency has had its However, for eusocial successes4. insects, the problem is deeper and more interesting because the maximizing agent is the colony and not
1993
Blackwell Scientific 5 Jaenike, J. (1978) Evol. Theor. 3,191-194 6 Bell, G. (1982) The Masterpiece of Nature: The Evolution and Generics of Sexuality, University of California Press 7 Jaenike, J. (1993) Evol. Ecol. 7, 103-108 6 Jaenike, J. (1992) Am. Nat. 139, 893-906 9 Hori, M. (1993) Science 260,218-219
necessarily each individual worker. Indeed, NuAez5 and others have shown that honeybees (Apis mellifera) do not maximize gross rate of nectar delivery, because they return to their hives from nondepleting nectar sources with partially filled crops, Nuriez suggested that the bees’ behaviour could be understood by optimizing information transfer: individual bees may sacrifice rate of delivery to reduce intervals between recruiting episodes, which could result in higher overall gains for the colony. However, further analysis6 showed that the bees’ behaviour was also compatible with modified hypotheses about individual worker maximization, leaving the issue unresolved. Leaf-cutting ants may help in rethinking the problem. Leaf-cutting ants usually follow trails towards suitable leaf sources, but some workers (scouts) wander outside the established tracks, and upon the detection of new suitable leaf sources return to the colony laying new pheromone trails. Other workers, if recruited, travel to the new food source and proceed to collect leaf material. Races and NuAez’ used an elegant laboratory technique to examine the information transmitted during recruitment. They first exposed scouts to droplets of scented sugar solution of either of two concentrations, 1% or 10%. Scouts detected these droplets and returned to the colony leaving a recruiting trail. When the recruits arrived, they encountered no nectar but sheets of parafilm embedded in the same scent but containing no sugar. Since all recruits found the same material, differences between the groups must be the result of information transmitted by the scouts. I shall refer to workers recruited by scouts who found the 10% solution as ‘10% workers’ and to workers recruited by scouts who had found 1% solution as ‘1% workers’, but it should remain clear that both groups encountered sugar-free parafilm at the source.
Q 1993, Elsevier Science Publishers
Ltd (UK)
TREE vol. 8, no. IO, October
1993 +
go0
Races and Nunez’ report several effects of scouts’ experience on workers’ behaviour. 10% workers (1) cut smaller sections of parafilm, (2) returned to the colony at a faster pace and (3) displayed more active recruiting behaviour than 1% workers by marking the substrate. When, in another experiment, recruits found pieces of parafilm of a homogeneous size, they still travelled faster if they had been recruited by 10% scouts. Thus, differences in velocity were not caused by differences in burden. Greater velocity did not compensate for the reduction in fragment size, and 10% workers had a lower rate of leaf transport during travel than 1% workers (Fig. 1). Overall rate of delivery, however, depends on the time spent cutting the leaf fragments and processing them in the colony as well as on rate of transport, and this information is not available. These results contradict the most obvious functional hypotheses. Let us examine some possibilities. Delivery rate maximization The strategy leading to maximum rate of leaf delivery is independent of real or expected sugar content. Since 10% and 1% workers behaved differently, they could not have maximized delivery rate in both treatments. They could however have been maximizing a compromise between rate of leaf delivery and some unknown cost correlated with travel velocity. If such a cost exists, workers might pay more of it (i.e. walk faster to reduce travel time) if the value of the leaves is greater. Thus they would achieve a higher delivery rate for higher-quality leaves. Ten-percent workers did travel faster and experienced shorter travel times, even for equal-sized fragments, but why did they carry smaller loads? Delivery rate maximization predicts smaller loads for shorter travel times if fragment size is a concave function of cutting time. Races and Nunez do not report this function, but a fragment’s size is a quadratic function of its radius while cutting time is likely to be proportional to its semi-circumference and thus is linear to its radius. The expected size versus cutting time function is thus convex, the opposite of the required relation. These speculations will hopefully become redundant when the real functions for cutting and colony times versus fragment size are known so that delivery rate can be compared between treatments, but it appears that not even postulating an imaginary cost for velocity could save this hypothesis.
local contrast Workers may detect the contrast between the scouts’ information and the actual value of the resource. Applying this hypothesis to invertebrates is admittedly far-fetched, but it does predict a lower delivery rate by 10% workers. Ten-percent workers find a greater discrepancy between the scouts’ information and the test condition. This greater contrast between ‘expectation’ and experience may enhance detection of the experimental ploy, leading to more reluctant foraging on the part of the 10% recruits. Faster travel from the colony to the source would reflect higher ‘expectation’, while smaller burden and faster return velocity would reflect lower motivation or greater urgency to spend foraging time somewhere else. However, 10% workers showed higher trail marking activity, so that for this hypothesis to be correct they ought to have been laying a negativemessage pheromone rather than a recruiting one. This is unlikely. Just as with rate maximization, this hypothesis is directly testable: differences between treatments should disappear if workers find the same concentration as their own scouts, regardless of the scouts’ experience. Informational transfer Races and Nuriez’ preferred explanation is based on putative recruitment strategies by the workers, and it is also based on a trade-off between individual delivery rate and another currency. The idea is that workers sacrifice individual leaf delivery rate to return earlier to the colony for further recruitment. In contrast with my modified version of the rate hypothesis, it assumes that individual delivery rate is lower for higher-quality leaves, but shorter times outside the colony enhance recruitment leading to higher rate of leaf gathering by the colony. This hypothesis is consistent with the increased velocity and with the reduced leaf size in 10% workers. By cutting smaller loads 10% workers save cutting time and by increasing velocity they save travel time. This hypothesis can also accommodate the more intense trail marking by 10% workers. However, there are still some difficulties: workers of both treatments found equal (worthless) parafilm sheets, and thus the different recruitment behaviour is just an amplification of the differences in the scouts’ findings, with no correction for the poor foraging source encountered by the workers. This would result in a positive feed-
”
Burden Fig. 1. Load transport rate (load mass x velocity) of sugarfree parafilm fragments for individual ants as a function of burden [(ant mass + load massMant mass)]. Filled symbols (0) indicate workers recruited by scouts that had encountered 10% sugar solution and empty symbols (0) workers recruited by scouts who had found 1% sugar solution. Ten-percent workers carry smaller burdens, and, although they walk faster (not shown in the figure), this is enough to generate lower transport rates. From Ref. I, wifh permission.
back process that would lead the colony to progressively concentrate on once good but presently worthless food sources. One should expect workers to correct the information received from the scouts, but other pheromone-related positive feedback phenomena do occur in ants in the context of colony defence2 and if confirmed, this would add another intriguing observation. This hypothesis, the strongest so far, requires confirmation of the reduction of individual delivery rate with resource quality, a prediction that opposes that of the rate hypothesis and that will be tested by the same information. It is satisfying that for the time being there is no winning functional model, although the information transfer hypothesis is in my view the strongest candidate. What is the way ahead? Some critics of the adaptationist stand may reject further optimality tinkering, thinking that animals have been shown to be nonoptimal yet again. Naive adaptationists may brush the difficulties under the carpet by claiming that the artificial nature of the laboratory conditions creates behavioural attefacts. I would argue that a more promising attitude is to rise to the challenge implied in these results of more precise testing of these and other possible optimality models using further detailed measurements of the ants’ behaviour. Many empirical issues remain open. Knowledge of the interactions between leaf size and cutting time and colony time versus leaf size is crucial to calculate individual delivery rates. The time course of the effects of treatment could elucidate the possible positive feedback on recruitment. The control of recruits’ behaviour must be analysed; they could respond to information received in the colony or to the pheromone laid by the scouts in the proximity of the food source. These facts could be incorporated 347
news &Go~~ent
into formal models of individual and colony foraging, so that the putative trade-off between loss of individual performance and increase in recruitment may be assessed. The general message is that future progress depends on linking theory and facts. Neither abstract optimality theorizing nor purely descriptive fact reporting will solve these puzzles, but both are required. The dialectical exchange between pre-
TREE
cise theoretical expectations and real-life phenomena promotes progress, and progress is faster and more fun when models and facts clash.
1 Races, F. and Nuriez, Behav. 45, 135-143
J.A.
(1993)
Aim.
2 Holldobler, 6. and Wilson, E.O. (1990) The Ants, Springer Verlag
FatTimes, Lean Times andCompetition among Predators John A. Wiens CLASSICAL COMPETITION THEORY predicts that coexisting species that share limiting resources should compete. For coexistence to continue, the species should diverge in resource use, reducing niche overlap’,2. This view of nature, which was prevalent in ecology during the 1960s and early 197Os, has been questioned on several counts3,4. The effects of environmental variation on resource levels and competition have received particular attention. If environmental conditions vary (as they do), then resources such as food may be abundant at some times (‘fat’ times) and scarce at others (‘lean’ times). Accordingly, we might expect niche overlap among species to be greater during the fat times than the lean times, when competition is presumably more severe2*3*5. In the early 198Os, Schoeneti tested this expectation by surveying some 30 studies in which changes in niche overlap among coexisting species had been documented between seasons or years. In most cases, overlap was indeed less during the relatively lean period, usually the winter. Does this mean that the species are competing during the lean times but not during the fat times? Not necessarily. Competition is related to resource limitation, and without direct information about resource levels, terms such as ‘fat’ or ‘lean’
John Wiens is at the Dept of Biology and Graduate Degree Program in Ecology, Colorado State University, Fort Collins, CO 80523, USA. 348
tell one little about whether or not a resource-limitation threshold has been passed (Fig. la). Also, niche overlap may not change monotonically with a reduction in resource supplies. Coexisting species may respond opportunistically to certain resources when they are superabundant, specialize as resources become more limiting, but then converge again as resources become extremely scarce, leading to a high-low-high sequence of niche overlaps (Fig. 1 b). A recent report by Jaksic and his colleagues6 at least partially addresses these problems, and shows that the ‘fat-lean’ scenario may be an oversimplification. Jaksic et al. studied the food habits and guild structure of predatory vertebrates over a 4-year period in semi-arid scrub desert north of Santiago, Chile. These predators (four owl species, four falconiform hawk species and two foxes) are well-suited to an investigation of niche dynamics. They are large and conspicuous and habitually use sites such as roosts or dens where their pellets or feces may be collected, and remains of prey in these castings can be identified with some precision’. This is important, for the determination of the trophic guild structure of a community is sensitive to the level to which prey items can be resolved taxonomicallp. In this study, prey remnants in the castings were identified to species for vertebrate prey and to orders for invertebrate prey - not ideal resolution, but better than that in most such studies.
8, no.
10, October
1993
3 Orians, G.H. and Pearson, N.E. (1979) in Analysis of Ecological Systems (Horn, D.J., Stairs, G.R. and Mitchell, PD., eds), pp. 154-177, Ohio University Press 4 Kacelnik, A. (1984) Anim. Behav. 53, 283-299 5 Nutiez, 139-150
References
vol.
J.A. (1982) J. Apic.
Res.
21,
6 Kacelnik, A., Schmid-Hempel, P. and Houston, A.I. (1986) Behav. Ecol, Sociobiol.
19, 19-24
In this system, bird and arthropod prey populations showed the expected pattern of an increase in abundance associated with the flush of vegetation growth at the onset of the breeding period and a decline at the end of this season - breeding and nonbreeding seasons were relatively ‘fat’ and ‘lean’ for consumers of these prey. Small mammal populations, on the other hand, showed no such changes. Instead, they irrupted early in the study, apparently in response to unusually high rainfall and vegetation productiong. Numbers declined dramatically throughout the remainder of the study, and estimated densities at the lowest level were only 7% of those recorded at the abundance peak. For predators on small mammals, then, ‘fat’ and ‘lean’ were expressed on a scale of years, not seasons. Under these conditions, one might anticipate that omnivorous predators should shift diets seasonally, reducing overlap during the nonbreeding season, whereas predators on mammals might emigrate or shift to other prey as the mammal populations crashed, increasing overlap with the omnivores. However, despite the fact that nearly every predator species displayed a different response to the seasonal and yearly changes in food abundance, the basic guild structure of the assemblage was stable. One mammal-eating guild, containing the owls Tflo and Bubo, was always present, and these species exhibited high dietary overlap. During the few survey periods when Tyto was not recorded, Bubo remained isolated in this guild. Neither species showed any dietary shifts in response to the decline in small mammal abundance. A second group (two other owls and the two fox species) formed a tight omnivorous feeding guild based on arthropods and a broad range of
0 1993, Elsevier Science Publishers
Ltd (UK1