- Email: [email protected]
Chaos and evolution
CORRESPONDENCE Finally, although successful biological control programs support the contention that parasitoids can reduce the average density of hosts, how they influence changes in host density thereafter is not so clearlo. No one could deny that parasitoids are ‘important’, but it remains a question if they are sufficient to cause cyclic dynamics of forest caterpillars.
Judith H. Myers Depts of Zoology and Plant Science, Centre for Biodiversity Research, University of British Columbia, Vancouver, Canada V6T 124 References i Berryman, A. (1996) Trends Ecol. ho/. 11, 28-32 2 Berryman, A.A., Millstein, J.A. and Maston, R.R. (1990) in Population Dynamics of Forest Insects (Watt, A.D. et al., eds), pp. 369-380, Intercept 3 Baltensweiler, W. and Fischlin, A. (1988) in Dynamics of Forest hsect Populations: Patterns, Causes, Implications (Berryman, A.A., ed.), pp. 331-351, Plenum Press 4 Baltensweiler, W. (1993) Oecologia 94,62-66 5 Anderson, R.M. and May, R.M. (1980) Science 210,658-661 6 Bowers, R.G., Begon, M. and Hodgkinson, D.E. (1993) Oihos 67,529-538 7 Ginzberg, L.R. and Taneyhill, D.E. (1994) J. Anim. Ecol. 63, 79-92 8 Myers, J.H. and Kukan, B. (1996) Oecologia 103, 475-480 9 Krebs, C.J. et al. (1995) Science 269,1112-1115 10 Roland, J. (1994) J. Anim. Ecol. 63,392-398
Reply from A.A. Berryman Judy Myers accuses me of confusing ‘factors that are “important” with those that are necessary and/or sufficient to cause population cycles’. The question I was asking, however, was a different one. The necessary condition for cycles in any variable is that its dynamics be dominated by delayed negative feedbacki-5, while sufficient conditions are met if the equilibrium point is unstable or if the variable is continuously or periodically disturbed6,7. Many factors can interact with population density to produce delayed negative feedback (e.g. parasitoids, predators, pathogens, food quantity and quality), while others can act as disturbances (e.g. weather, virus epizootics, spray projects). The question I addressed in my paper was: of all the potential factors, which one(s) is the most likely to have been involved in the delayed negative feedback (necessary condition) that produced the observed cyclic dynamics in populations of certain forest Lepidoptera, under the reasonable assumption that the sufficient conditions (environmental disturbances) are always met in nature? I analyzed all the data I could find that showed evidence of 6-12-year cycles, and this led me to conclude that insect parasitoids were the most likely factor to have created the necessary condition (delayed negative feedback) in many (but not all) of the populations. This conclusion was supported by the biological control literature, which suggested that
336
0 1996, Elsevier
Science Ltd
parasitoids often have a strong impact on the dynamics of their prey (a necessary ingredient for a particular feedback loop to dominate*), and the fact that life cycles of specific parasitoids are often synchronized with those of their prey (which could produce the observed second-order effects). I see no reason to change that conclusion. Myers uses Krebs et a/.‘s recent experiments with snowshoe hares9 as a foil to my observation that predator-prey interactions have been receiving more attention recently in explanations for population cycles. Krebs et al. concluded from their interesting experiments that three trophic levels (food-hare-lynx; all predator-prey interactions in the broadest sense) were involved in the cyclical dynamicsgslo. I have no problem at all with this interpretation, particularly as I had anticipated it 15 years previously (see Fig. 3.5 in Ref. 11, which shows the third-order feedback expected from three-trophic-level dynamics, and see pp. 198 and 210 for an explanation; see also Ref. 12 for similar reasoning, and Refs 13 and 14 for an alternative interpretation). Larch budmoth cycles may have a similar interpretation because third-order feedback can also be detected in this time seriesl3,14. Alan
A. Berryman
Depts of Entomology and Natural Resource Sciences, Washington State University, Pullman, WA 99164, USA References 1 Hutchinson, E. (1948) Ann. N.Y. Acad. Sci. 50, 221-246 2 Morris, R.F. (1959) Ecology40, 580-588 3 May, R.M. (1972) Stability and Complexity in Model Ecosystems, Hatvard University Press 4 Benyman, A.A. (1978) Can. Entomol. 110,513-518 5 Berryman, A.A. (1995) in Encyclopedia of Environmental Biology, pp. 291-298, Academic Press 6 Berman, A.A. (1986) Can. Entomol. 118,775-779 7 Berryman, A.A., Millstein, J.A. and Mason, M.M. (1990) in Population Dynamics of Forest insects (Watt, A.D. et a/., eds), pp. 369-380, Intercept 8 Berryman, A.A. (1993) Oikos 68, 183-185 9 Krebs, C.J. et a/. (1995) Science 269,1112-1115 10 Stenseth, N.C. (1995) Science 269,1061-1062 11 Benyman, A.A. (1981) Population Systems: A General Introduction, Plenum Press I.2 Finerty, J.P. (1981) The Population Ecology of Cycles in Small Mammals, Yale University Press I3 Royama, T. (1977) Ecol. Monogr. 47, l-35 14 Royama, T. (1992) Analytical Population Dynamics, Chapman & Hall
Chaos and evolution In commenting on the article on chaos and evolution by Ferri&e and myselfl, Doebeli and Koella* made several points to which I would like to respond. I appreciate Doebeli and Koella calling attention to their recent work. Had they been available before we went to press, Doebeli and Koella’s results would have broadened the scope of our paper. The substantive claim in their letter is that
their recent work shows that selection generally favors equilibria1 population dynamics. They contrast this with the result of Ferriere and Gatto”, who showed that, in a particular model, there are parameter values for which chaotic population dynamics are favored and values for which equilibria are favored. It is not obvious what the point of this contrast is. We made no claim that chaotic dynamics are usually favored by selection -only that there are circumstances under which they are favored, as illustrated by the Ferriere and Gatto3 study. We stated that ‘chaos may be selectively favorable in some circumstances, but cycles or equilibria are surely favored in others. It is time, however, to abandon the prejudice that selection always favors constancy.. .’ In contrast, Doebeli and Koella2 claim that their own work allows them to draw general conclusions about selection acting on population dynamics specifically, that ‘there is a clear tendency for natural selection to favor simple dynamics.’ I think it is premature to draw this conclusion, for several reasons. First, I think the important question is this: under what circumstances does selection favor complex dynamics, and when does it favor equilibria? I believe that asking this is not only more important than asking which type of selective regime is more common; in fact, answering the former question is prerequisite to answering the latter. Second, Doebeli and Koella base this broad conclusion on their own studies of some particular models. They claim that these models are general, implying thereby that the results are general. This claim of generality is surprising, inasmuch as they made no such claim when these results were published. Indeed, they stated4 that ‘it is not at all clear how far our results generalize to more complicated situations, and the general question remains...: how often should we expect complex dynamics to occur in natural populations?’ It is possible that they have changed their opinion since then, but there is nothing about the particular model they used4 that is inherently more general than that used by Ferrihre and Gatto3 (an age-structured Ricker map). I believe Doebeli and Koella had it right the first time, when they stated that they didn’t know how general their results were. I would argue that we can’t yet say how general any results are on the evolution of population dynamics. Different models often yield different results; the question raised by such differences is what factors in the models’ structures or biological assumptions are responsible for the differences. Only after answering this question will we be able to say anything about the generality of theoretical results. In this particular case, it seems likely that the absence of age-structure in the Doebeli-Koella model (and its presence in the FerriBre-Gatto model) is likely to be important, as these authors themselves suggested4. Third, it is not clear that Doebeli and Koella’s work really leads to such different conclusions than those of Ferriere and Gatto. Atthough it is not obvious from reading their letter*, Doebeli and Koella also found4.5 regions of parameter space for which chaotic population dynamics are favored by selection. Since constraints on their demographic parameters typically led the system to evolve to the boundary between stable and complex dynamicsa, it seems odd to call this a ‘clear tendency’ for the evolution of stable equilibria. TREE
vol. II, no. 8August
1996
CORRESPONDENCE Fourth, I wish to point out that it is ultimately an empirical question6 whether selection generally favors one or the other type of dynamics. As noted in our review, there are presently no data that can tell us about the importance of chaos in any of the contexts discussed in our paper. Analyses of models can inform our thinking, and ultimately will be useful in helping to design appropriate experimental studies. But such analyses can’t tell us whether the models themselves, or their special biological assumptions, are adequate only empirical data can do that. Thus, we will have to postpone any firm conclusions about the generaltendencies of selection. I thank B. Kendall and C.K. Kelley for comments.
Gordon
250,
1
1
7
t=0.66, P =0.74
(
38
t=0.72,
P=O.76
l
-1
200 7
t=0.33.
P=O.37
I
8 F -im
150-
p
A. Fox
Dept
of Biology 0116, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0116, USA ([email protected]), and Dept of Biology, San Diego State University, San Diego, CA 92182-0057, USA References 1 FerriBre, R. and Fox, G.A. (1995) Trends Ecol. Evol. 10, 480-485 2 Doebeli, M. and Koella, J. (1996) Trends Ecol. Evol. 11, 220 3 Ferri&re, R. and Gatto, M. (1993) Proc. R. Sot. London Ser. 13251,33-38 4 Doebeli, M. and Koella, J.C. (1995) Proc. R. Sot. London Ser. B 260,119-125 5 Doebeli, M. and Koella, J.C. (1994) Proc. R. Sot. London Ser. B 257,17-23 6 Hastings, A. et al. (1993) Annu. Rev. Ecol. Syst. 24, l-33
Conserving wild dogs Christopher Dye’s recent TREEarticle asked what really happened when the last few African wild dogs (Lycaon pictus) disappeared from the Serengeti. He considered the hypothesis that handling by researchers stressed the dogs and thus increased their vulnerability to disease*, and he concluded that the likelihood of this hypothesis being correct is ‘tiny’. He also noted that arguments rejecting the hypothesis are not ‘watertight’. I agree with both of these conclusions, but two open points can be resolved. First, Dye noted that data on short-term changes in cortisol levels in response to handling3 do not resolve whether handling and carrying a radiocollar are likely to provoke immune suppression. A more conclusive approach is to test for chronic elevation of glucocorticoids in collared animals4. We measured faecal glucocorticoid levels of wild dogs over two years (noninvasively, in the wild)5 and found that corticosterone levels did not differ for collared and uncollared dogs (controlling for rank and sex, partial Fi,21i= 0.91, P= 0.41; see Fig. 1)6. For collared dogs, pre- and post-collaring corticosterone levels did not differ (partial FI,49= 1.01, P= 0.32) (Ref. 6), providing a direct experimental rejection of the stress hypothesis. TREEuol.
II, no. SAu.31st 1996
Collared
Uncollared
Fig. 1. Baseline (unstressed) faecal corticosterone levels did not differ for collared and uncollared African wild dogs (males, black bars; females, unshaded bars; both sexes, shaded bars) in the Selous Game Reserve. Error bars show one standard error.+values are for one tailed t-tests.
Second, Dye notes that Ginsberg et a/.g found that ‘handled animals had consistently higher survival rates, which invites questions about their underlying assumptions’ (p. 188). For the data that I contributed to this analysis, the reason that handled dogs tended to survive better is simple - I intentionally collared dogs that were in better-than-average condition. Identifying what delivered the knockout punch to a declining population of 20 dogs is not really a priority. Far more important is to understand why the Serengeti population fell to such a low density six years before the final blow. The central issue for wild dog conservation is understanding what limits them to low population density even with the best of conditions (e.g. wild dog densities fall between 1/8 and 1/12o of spotted hyena densities across several ecosystems)7. Dye asks ‘why dogs need to be handled at all...much of this [needed] information could be (and is now being) obtained without catching free-ranging dogs’ (p. 189). To my knowledge, there are now three studies focusing on the conservation of wild dogs, and all use radiocollars. Wild dogs are extremely difficult to find and follow, particularly in the woodland habitats where they attain their highest densities. When we began our study in Selous, we spent five months just looking for wild dogs. Radiocollars are essential (1) to gather the demographic data needed to assess the stability of populations and their level of threat, (2) to identify ecological processes (e.g. interspecific competition’or hunting success8) that may limit density, (3) to study the epidemiology and pathology of infectious disease+11, and (4) to identify behavioral and demographic processes affecting survival and reproductionQs13. Progress is being made on all of these fronts. Without radiocollars -which do not chronically stress the dogs or increase their chance of dying-we would have none of this information.
All agree that the effects of handling should be considered when studying endangered species (or any species). Ironically, wild dog studies are still on the spot, even though the data showing that radiocollaring is benign are now more extensive than for any other species on the globe.
Scott Creel Field Research Center for Ecology and Ethology, Rockefeller University, Box 38B RR2, Millbrook, NY 12545, USA References 1 Dye, C. (1996) Trends. Ecol. Evol. 11, 188-189 2 Burrows, R., Hofer, H. and East, M. (1995) Proc. I?. Sot. London Ser. B 262,235-245 3 De Villiers, MS. et al. (1995) Proc. R. Sot. London Ser. B 262,215-220 4 Creel, S. (1992) Nature 360, 633 5 Creel, S., Creel, N.M. and Monfort, S.L. (1996) Nature 379, 212 6 Creel, S., Creel, N.M. and Monfort, S.L. Consew Biol. (in press) 7 Ginsberg, J.R. et al. (1995) Conserv. Biol. 9, 665-674 8 Creel, S. and Creel, N.M. Conserv. Biol. (in press) 9 Creel, S. and Creel, N.M. (1995) Anim. Behav. 50,1325-1339 10 Mills, M.G.L. (1993) OnderstepoortJ. Vet. Res. 60,405-409 11 Van Heerden, J. eta/. (1995) J. S. Afr. Vet. Assoc. 66,18-27 12 Creel, S. et al. (1995) J. Zoo/. London 236, 199-209 13 McNutt, J.W. Anim. Behav. (in press) 14 McNutt, J.W. J. Zoo/. London (in press)
0 1996, Elsevier Science Ltd
337
Judith H. Myers Depts of Zoology and Plant Science, Centre for Biodiversity Research, University of British Columbia, Vancouver, Canada V6T 124 References i Berryman, A. (1996) Trends Ecol. ho/. 11, 28-32 2 Berryman, A.A., Millstein, J.A. and Maston, R.R. (1990) in Population Dynamics of Forest Insects (Watt, A.D. et al., eds), pp. 369-380, Intercept 3 Baltensweiler, W. and Fischlin, A. (1988) in Dynamics of Forest hsect Populations: Patterns, Causes, Implications (Berryman, A.A., ed.), pp. 331-351, Plenum Press 4 Baltensweiler, W. (1993) Oecologia 94,62-66 5 Anderson, R.M. and May, R.M. (1980) Science 210,658-661 6 Bowers, R.G., Begon, M. and Hodgkinson, D.E. (1993) Oihos 67,529-538 7 Ginzberg, L.R. and Taneyhill, D.E. (1994) J. Anim. Ecol. 63, 79-92 8 Myers, J.H. and Kukan, B. (1996) Oecologia 103, 475-480 9 Krebs, C.J. et al. (1995) Science 269,1112-1115 10 Roland, J. (1994) J. Anim. Ecol. 63,392-398
Reply from A.A. Berryman Judy Myers accuses me of confusing ‘factors that are “important” with those that are necessary and/or sufficient to cause population cycles’. The question I was asking, however, was a different one. The necessary condition for cycles in any variable is that its dynamics be dominated by delayed negative feedbacki-5, while sufficient conditions are met if the equilibrium point is unstable or if the variable is continuously or periodically disturbed6,7. Many factors can interact with population density to produce delayed negative feedback (e.g. parasitoids, predators, pathogens, food quantity and quality), while others can act as disturbances (e.g. weather, virus epizootics, spray projects). The question I addressed in my paper was: of all the potential factors, which one(s) is the most likely to have been involved in the delayed negative feedback (necessary condition) that produced the observed cyclic dynamics in populations of certain forest Lepidoptera, under the reasonable assumption that the sufficient conditions (environmental disturbances) are always met in nature? I analyzed all the data I could find that showed evidence of 6-12-year cycles, and this led me to conclude that insect parasitoids were the most likely factor to have created the necessary condition (delayed negative feedback) in many (but not all) of the populations. This conclusion was supported by the biological control literature, which suggested that
336
0 1996, Elsevier
Science Ltd
parasitoids often have a strong impact on the dynamics of their prey (a necessary ingredient for a particular feedback loop to dominate*), and the fact that life cycles of specific parasitoids are often synchronized with those of their prey (which could produce the observed second-order effects). I see no reason to change that conclusion. Myers uses Krebs et a/.‘s recent experiments with snowshoe hares9 as a foil to my observation that predator-prey interactions have been receiving more attention recently in explanations for population cycles. Krebs et al. concluded from their interesting experiments that three trophic levels (food-hare-lynx; all predator-prey interactions in the broadest sense) were involved in the cyclical dynamicsgslo. I have no problem at all with this interpretation, particularly as I had anticipated it 15 years previously (see Fig. 3.5 in Ref. 11, which shows the third-order feedback expected from three-trophic-level dynamics, and see pp. 198 and 210 for an explanation; see also Ref. 12 for similar reasoning, and Refs 13 and 14 for an alternative interpretation). Larch budmoth cycles may have a similar interpretation because third-order feedback can also be detected in this time seriesl3,14. Alan
A. Berryman
Depts of Entomology and Natural Resource Sciences, Washington State University, Pullman, WA 99164, USA References 1 Hutchinson, E. (1948) Ann. N.Y. Acad. Sci. 50, 221-246 2 Morris, R.F. (1959) Ecology40, 580-588 3 May, R.M. (1972) Stability and Complexity in Model Ecosystems, Hatvard University Press 4 Benyman, A.A. (1978) Can. Entomol. 110,513-518 5 Berryman, A.A. (1995) in Encyclopedia of Environmental Biology, pp. 291-298, Academic Press 6 Berman, A.A. (1986) Can. Entomol. 118,775-779 7 Berryman, A.A., Millstein, J.A. and Mason, M.M. (1990) in Population Dynamics of Forest insects (Watt, A.D. et a/., eds), pp. 369-380, Intercept 8 Berryman, A.A. (1993) Oikos 68, 183-185 9 Krebs, C.J. et a/. (1995) Science 269,1112-1115 10 Stenseth, N.C. (1995) Science 269,1061-1062 11 Benyman, A.A. (1981) Population Systems: A General Introduction, Plenum Press I.2 Finerty, J.P. (1981) The Population Ecology of Cycles in Small Mammals, Yale University Press I3 Royama, T. (1977) Ecol. Monogr. 47, l-35 14 Royama, T. (1992) Analytical Population Dynamics, Chapman & Hall
Chaos and evolution In commenting on the article on chaos and evolution by Ferri&e and myselfl, Doebeli and Koella* made several points to which I would like to respond. I appreciate Doebeli and Koella calling attention to their recent work. Had they been available before we went to press, Doebeli and Koella’s results would have broadened the scope of our paper. The substantive claim in their letter is that
their recent work shows that selection generally favors equilibria1 population dynamics. They contrast this with the result of Ferriere and Gatto”, who showed that, in a particular model, there are parameter values for which chaotic population dynamics are favored and values for which equilibria are favored. It is not obvious what the point of this contrast is. We made no claim that chaotic dynamics are usually favored by selection -only that there are circumstances under which they are favored, as illustrated by the Ferriere and Gatto3 study. We stated that ‘chaos may be selectively favorable in some circumstances, but cycles or equilibria are surely favored in others. It is time, however, to abandon the prejudice that selection always favors constancy.. .’ In contrast, Doebeli and Koella2 claim that their own work allows them to draw general conclusions about selection acting on population dynamics specifically, that ‘there is a clear tendency for natural selection to favor simple dynamics.’ I think it is premature to draw this conclusion, for several reasons. First, I think the important question is this: under what circumstances does selection favor complex dynamics, and when does it favor equilibria? I believe that asking this is not only more important than asking which type of selective regime is more common; in fact, answering the former question is prerequisite to answering the latter. Second, Doebeli and Koella base this broad conclusion on their own studies of some particular models. They claim that these models are general, implying thereby that the results are general. This claim of generality is surprising, inasmuch as they made no such claim when these results were published. Indeed, they stated4 that ‘it is not at all clear how far our results generalize to more complicated situations, and the general question remains...: how often should we expect complex dynamics to occur in natural populations?’ It is possible that they have changed their opinion since then, but there is nothing about the particular model they used4 that is inherently more general than that used by Ferrihre and Gatto3 (an age-structured Ricker map). I believe Doebeli and Koella had it right the first time, when they stated that they didn’t know how general their results were. I would argue that we can’t yet say how general any results are on the evolution of population dynamics. Different models often yield different results; the question raised by such differences is what factors in the models’ structures or biological assumptions are responsible for the differences. Only after answering this question will we be able to say anything about the generality of theoretical results. In this particular case, it seems likely that the absence of age-structure in the Doebeli-Koella model (and its presence in the FerriBre-Gatto model) is likely to be important, as these authors themselves suggested4. Third, it is not clear that Doebeli and Koella’s work really leads to such different conclusions than those of Ferriere and Gatto. Atthough it is not obvious from reading their letter*, Doebeli and Koella also found4.5 regions of parameter space for which chaotic population dynamics are favored by selection. Since constraints on their demographic parameters typically led the system to evolve to the boundary between stable and complex dynamicsa, it seems odd to call this a ‘clear tendency’ for the evolution of stable equilibria. TREE
vol. II, no. 8August
1996
CORRESPONDENCE Fourth, I wish to point out that it is ultimately an empirical question6 whether selection generally favors one or the other type of dynamics. As noted in our review, there are presently no data that can tell us about the importance of chaos in any of the contexts discussed in our paper. Analyses of models can inform our thinking, and ultimately will be useful in helping to design appropriate experimental studies. But such analyses can’t tell us whether the models themselves, or their special biological assumptions, are adequate only empirical data can do that. Thus, we will have to postpone any firm conclusions about the generaltendencies of selection. I thank B. Kendall and C.K. Kelley for comments.
Gordon
250,
1
1
7
t=0.66, P =0.74
(
38
t=0.72,
P=O.76
l
-1
200 7
t=0.33.
P=O.37
I
8 F -im
150-
p
A. Fox
Dept
of Biology 0116, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0116, USA ([email protected]), and Dept of Biology, San Diego State University, San Diego, CA 92182-0057, USA References 1 FerriBre, R. and Fox, G.A. (1995) Trends Ecol. Evol. 10, 480-485 2 Doebeli, M. and Koella, J. (1996) Trends Ecol. Evol. 11, 220 3 Ferri&re, R. and Gatto, M. (1993) Proc. R. Sot. London Ser. 13251,33-38 4 Doebeli, M. and Koella, J.C. (1995) Proc. R. Sot. London Ser. B 260,119-125 5 Doebeli, M. and Koella, J.C. (1994) Proc. R. Sot. London Ser. B 257,17-23 6 Hastings, A. et al. (1993) Annu. Rev. Ecol. Syst. 24, l-33
Conserving wild dogs Christopher Dye’s recent TREEarticle asked what really happened when the last few African wild dogs (Lycaon pictus) disappeared from the Serengeti. He considered the hypothesis that handling by researchers stressed the dogs and thus increased their vulnerability to disease*, and he concluded that the likelihood of this hypothesis being correct is ‘tiny’. He also noted that arguments rejecting the hypothesis are not ‘watertight’. I agree with both of these conclusions, but two open points can be resolved. First, Dye noted that data on short-term changes in cortisol levels in response to handling3 do not resolve whether handling and carrying a radiocollar are likely to provoke immune suppression. A more conclusive approach is to test for chronic elevation of glucocorticoids in collared animals4. We measured faecal glucocorticoid levels of wild dogs over two years (noninvasively, in the wild)5 and found that corticosterone levels did not differ for collared and uncollared dogs (controlling for rank and sex, partial Fi,21i= 0.91, P= 0.41; see Fig. 1)6. For collared dogs, pre- and post-collaring corticosterone levels did not differ (partial FI,49= 1.01, P= 0.32) (Ref. 6), providing a direct experimental rejection of the stress hypothesis. TREEuol.
II, no. SAu.31st 1996
Collared
Uncollared
Fig. 1. Baseline (unstressed) faecal corticosterone levels did not differ for collared and uncollared African wild dogs (males, black bars; females, unshaded bars; both sexes, shaded bars) in the Selous Game Reserve. Error bars show one standard error.+values are for one tailed t-tests.
Second, Dye notes that Ginsberg et a/.g found that ‘handled animals had consistently higher survival rates, which invites questions about their underlying assumptions’ (p. 188). For the data that I contributed to this analysis, the reason that handled dogs tended to survive better is simple - I intentionally collared dogs that were in better-than-average condition. Identifying what delivered the knockout punch to a declining population of 20 dogs is not really a priority. Far more important is to understand why the Serengeti population fell to such a low density six years before the final blow. The central issue for wild dog conservation is understanding what limits them to low population density even with the best of conditions (e.g. wild dog densities fall between 1/8 and 1/12o of spotted hyena densities across several ecosystems)7. Dye asks ‘why dogs need to be handled at all...much of this [needed] information could be (and is now being) obtained without catching free-ranging dogs’ (p. 189). To my knowledge, there are now three studies focusing on the conservation of wild dogs, and all use radiocollars. Wild dogs are extremely difficult to find and follow, particularly in the woodland habitats where they attain their highest densities. When we began our study in Selous, we spent five months just looking for wild dogs. Radiocollars are essential (1) to gather the demographic data needed to assess the stability of populations and their level of threat, (2) to identify ecological processes (e.g. interspecific competition’or hunting success8) that may limit density, (3) to study the epidemiology and pathology of infectious disease+11, and (4) to identify behavioral and demographic processes affecting survival and reproductionQs13. Progress is being made on all of these fronts. Without radiocollars -which do not chronically stress the dogs or increase their chance of dying-we would have none of this information.
All agree that the effects of handling should be considered when studying endangered species (or any species). Ironically, wild dog studies are still on the spot, even though the data showing that radiocollaring is benign are now more extensive than for any other species on the globe.
Scott Creel Field Research Center for Ecology and Ethology, Rockefeller University, Box 38B RR2, Millbrook, NY 12545, USA References 1 Dye, C. (1996) Trends. Ecol. Evol. 11, 188-189 2 Burrows, R., Hofer, H. and East, M. (1995) Proc. I?. Sot. London Ser. B 262,235-245 3 De Villiers, MS. et al. (1995) Proc. R. Sot. London Ser. B 262,215-220 4 Creel, S. (1992) Nature 360, 633 5 Creel, S., Creel, N.M. and Monfort, S.L. (1996) Nature 379, 212 6 Creel, S., Creel, N.M. and Monfort, S.L. Consew Biol. (in press) 7 Ginsberg, J.R. et al. (1995) Conserv. Biol. 9, 665-674 8 Creel, S. and Creel, N.M. Conserv. Biol. (in press) 9 Creel, S. and Creel, N.M. (1995) Anim. Behav. 50,1325-1339 10 Mills, M.G.L. (1993) OnderstepoortJ. Vet. Res. 60,405-409 11 Van Heerden, J. eta/. (1995) J. S. Afr. Vet. Assoc. 66,18-27 12 Creel, S. et al. (1995) J. Zoo/. London 236, 199-209 13 McNutt, J.W. Anim. Behav. (in press) 14 McNutt, J.W. J. Zoo/. London (in press)
0 1996, Elsevier Science Ltd
337