P ERSPECTIVES cur on length scales far below those that are tiny, with diameters on the order of 10–19 m, will provide enhanced sensitivity to the pu- currently accessible. The most promising and evaporate explosively after only 10–27 s. tative microscopic black holes (12, 13). The approach is to look not for small effects at Today’s particle colliders are not suffi- Large Hadron Collider, currently under con- relatively large length scales, but for large ciently energetic to produce microscopic struction in Geneva, will provide an even effects at the smallest possible length black holes. However, ultrahigh-energy higher sensitivity to large extra dimensions. scales, where gravity is predicted to be cosmic rays have been observed to collide If no anomalous effects are seen in these strong. These probes are equally powerful with particles in Earth’s atmosphere with ambitious projects, the possibility of large for any n. For low n, they are superseded by center-of-mass energies that are 100 times extra dimensions will be excluded. If seen those discussed above, but for large n, they those available at human-made colliders. and confirmed, however, these effects will provide the leading experimental tests. The ultrahigh-energy neutrinos that are ex- provide the first evidence for strong gravity Perhaps the most remarkable possibility pected to accompany these cosmic rays and a radically new view of spacetime. for testing large n has been the realization may create microscopic black holes. that if gravity is strong at 10–19 m, tiny black Although these black holes are extremely References holes may form in high-energy particle col- short-lived and hence impossible to detect 1. N. Arkani-Hamed, S. Dimopoulos, G. R. Dvali, Phys. Lett. B 429, 263 (1998). lisions (5–8). The formation of a black hole directly, their explosive evaporations pro- 2. E. G. Adelberger et al., http://arXiv.org/abs/hep- is expected when a large mass or, equiva- duce events with unusual properties (7, 8). ex/0202008 (2002). lently, a large energy is concentrated in a The fact that no such events have been ob- 3. S. Cullen, M. Perelstein, Phys. Rev. Lett. 83, 268 (1999). small volume (9, 10). In the conventional 3D served so far places strong constraints on 4. L. J. Hall, D. R. Smith, Phys. Rev. D 60, 085008 (1999). 5. S. B. Giddings, S. Thomas, Phys. Rev. D 65, 056010 world, gravity is so weak that the required large extra dimensions, but does not yet ex- (2002). energy density is never achieved in observ- clude these scenarios altogether (11). 6. S. Dimopoulos, G. Landsberg, Phys. Rev. Lett. 87, able particle collisions. However, if large ex- The search for large extra dimensions 161602 (2001). 7. J. L. Feng, A. D. Shapere, Phys. Rev. Lett. 88, 021303 tra dimensions exist and gravity is intrinsi- will intensify. The currently operating (2001). cally strong, very high energy particles occa- Antarctic Muon and Neutrino Detector 8. L. A. Anchordoqui, H. Goldberg, Phys. Rev. D 65, sionally pass close enough to each other to Array and its successor IceCube are kilome- 047502 (2002). 9. D. M. Eardley, S. B. Giddings, Phys. Rev. D 66, 044011 trigger gravitational collapse, forming mi- ter-scale cosmic neutrino detectors buried (2002). croscopic black holes. Like conventional deep in the Antarctic ice. The Auger 10. H.Yoshino,Y. Nambu, Phys. Rev. D 66, 065004 (2002). black holes, these black holes are expected to Observatory, consisting of water Cerenkov 11. L. A. Anchordoqui et al., http://arXiv.org/abs/hep- emit “Hawking radiation,” which leads to the detectors covering a 3000-km2 area in the ph/0307228 (2003). 12. M. Kowalski, A. Ringwald, H. Tu, Phys. Lett. B 529,1 evaporation of the black holes. In contrast to high desert of Argentina, will also begin op- (2002). the astrophysical variety, however, they are eration in 2 to 3 years. These large projects 13. J. Alvarez-Muniz et al., Phys. Rev. D 65, 124015 (2002).

ECOLOGY specialist predator hypothesis, which postu- lates that small populations undergo periodic fluctuations in numbers in response Stranglers to by a specialized predator (4). This hypothesis has taken center stage be- and Cycles cause the fundamental theory of predator- prey interactions—encapsulated in the wor- Peter J. Hudson and Ottar N. Bjørnstad thy Lotka-Volterra model—predicts cycles in prey and predator abundance. Hence, it is or more than 80 years, population ecol- (Dicrostonyx groelandicus) in northeastern natural to consider that a predator (or some ogists have been preoccupied with the Greenland and describe how these dynam- other specialist consumer) is the crucial play- Frise and fall in population numbers ics are affected by predators. The mathemat- er in the cyclic dynamics of small mammal among small mammal , but they still ical model that the investigators develop il- populations. At a more detailed level, theory cannot agree on the reasons for these cyclic lustrates how the cyclic fluctuations of col- predicts that interactions between a special- variations in abundance. The controversy lared are driven by predation by ized predator and its main prey—such as the arises from three central questions: What the lemming specialist, the , and then stoat’s predation of collared lemmings— are the ecological mechanisms that generate are molded (when lemming populations should result in cycles in which the peak in fluctuations in these cycles? Are these reach high densities) by three generalist predator numbers lags behind that of its prey mechanisms common to all cyclic popula- predators: the , the snowy , and by one-quarter of a cycle (4). This prediction tions? Does understanding of these mecha- the long-tailed (see the figure). The is beautifully borne out by the Gilg et al. nisms allow us to explain why some popula- new work answers the first question and study (1). Indeed, this is one of those rare in- tions are cyclic whereas others are not? The provides key insights into the third question. stances when appears to reflect basic debate has been so heated among small The saying “Lemmings cycle—unless theory—a textbook case. mammal researchers that other ecologists they don’t” (2) embodies the enigma of One important feature of the specialist jokingly refer to them as the “vole stran- cyclic fluctuations in many lemming and predator hypothesis is that a second stabiliz- glers.” On page 866 of this issue, Gilg et al. vole populations inhabiting boreal and arctic ing effect is needed at high lemming densi- (1) present their long-term field study of the ecosystems. The is an ex- ties to slow down the growth rate of the prey cyclic dynamics of collared lemmings cellent example: Some populations exhibit and allow the specialist predator to catch up violent and periodic fluctuations in their and drive prey abundance downward (5). numbers, whereas others exhibit no clear sta- The collared lemming is, again, a wonderful P. J. Hudson is in the Department of Biology and O. N. tistical pattern (3). The “vole stranglers” have illustration. The cyclic fluctuations in lem- Bjørnstad is in the Departments of Entomology and Biology, Pennsylvania State University, University come up with many hypotheses to account ming populations in northeastern Greenland Park, PA 16802, USA. E-mail: [email protected] for this paradox. A favorite is the so-called appear to result from the tension between the

www.sciencemag.org SCIENCE VOL 302 31 OCTOBER 2003 797 P ERSPECTIVES destabilizing force of the specialist predator AB and the stabilizing effects of three generalist predators. The three generalist predators fo- cus their predation on the lemmings only when populations of these reach high densities. There is such a tight relationship between predator and prey that, as indeed the authors argue, these population fluctuations can be understood without having to invoke food availability, competition, or social inter- actions. In contrast, the noncyclic lemming populations in Arctic Canada appear to be CD trapped in a “generalist predator pit” where a complex guild of predators prohibits any in- creases in the lemming population (3). Taken together, such biogeographic comparisons provide critical insights into how the conse- quences of predator-prey (or, more generally, consumer-resource) interactions are affected by embedding in diverse food webs—that is, how predator-prey interactions are influ- enced by further consumer or competitive in- teractions (6, 7). A choice of predator. The cyclic dynamics of collared lemming populations are determined by a range Of course, there is a difference between of predators. In northeastern Greenland, the collared lemming (A) undergoes cyclic fluctuations in abun- consistency of models and biological ground dance as a consequence of predation by the lemming specialist, the stoat (B). However, lemming popu- truth. The next step in the study of the col- lations could escape control by the stoat if it were not for the stabilizing predation of three generalist lared lemmings is to use experimental ma- predators: the (C), the (D), and the long-tailed skua (not shown). nipulation of the rates of lemming predation to test the hypothesis mooted by Gilg and ed by the Popperian approach of hypothesis predator-prey ecosystems in the world—the colleagues [e.g., (8, 9)]. Such experimental falsification and with one experimental falsi- Serengeti plains of Tanzania in East manipulations will provide a test of the theo- fication have “thrown a general hypothesis Africa—we start to see some interesting pat- ry and will reveal how to further refine the out with local idiosyncrasies.” A clear illus- terns. A recent study (16) provides evidence theoretical model. Sadly, ecologists rarely tration of geographic differences regarding that the abundance of small in have the resources to “grasp the nettle” and causes of population cycles is provided by tropical Africa is determined principally go for such large-scale experiments. The is- northern European grouse. Experimental through top-down predation, whereas the sue for field biologists has been a tradeoff studies have identified parasites (an extreme larger herbivores are regulated by limited re- between replication of treatment and suffi- specialist) (9, 13) and territorial behavior (14) sources in a bottom-up process. We are now cient scale to ensure successful manipulation as forces that destabilize host abundance in well on the road to explaining biogeographic of the mechanism. Population-level experi- the British Isles. Yet there is also evidence variations in abundance and dynamics ments need to ensure that such manipula- that predation by the specialist gyr through species interactions and embedding tions result in effective changes in the puta- drives the cyclic fluctuations of grouse in in ecosystems, the sort of questions Elton tive ecological mechanism controlling cyclic Iceland (rock ptarmigan) (15). To understand was asking some 80 years ago when he first dynamics. It could be argued that “appropri- the cycles of boreal and arctic lemmings, we described the cyclic dynamics of small ate-scaled” field experiments are logistically sorely need more field experiments and an mammal populations. difficult and may have fiscal constraints that analysis that quantifies the changes in preda- compromise the goal of effective treatment tor-prey interactions resulting from such field References and appropriate data replication. There is manipulations. An ultimate meta-analysis of 1. O. Gilg, I. Hanski, B. Sittler, Science 302, 866 (2003). 2. I. Hanski, Trends Ecol. Evol. 2, 55 (1987). some truth in this. However, even with low these experiments will resolve the final ques- 3. D. G. Reid, C. J. Krebs, A. Kenney, Ecol. Monogr. 67,89 data replication, we can test such experi- tion about lemming cycles: Does the same (1997). ments against models and distinguish be- mechanism account for all cyclic lemming 4. W. M. Murdoch, C. Briggs, R. M. Nisbet, Consumer Resource Dynamics (Princeton Univ. Press, Princeton, tween competing hypotheses. After all, true populations, and if not, why not? NJ, 2003). independence of study sites is virtually im- These complexities aside, Gilg et al. (1) 5. I. Hanski, L. Hansson, H. Henttonen, J. Anim. Ecol. 60, 353 (1991). possible, as, according to the ecologist demonstrate how a simple (“few-species”) 6. E. McCauley, R. M. Nisbet, W. W. Murdoch, A. M. De Stephen Ellner, “nature itself is just one un- clockwork predator-prey interaction results Roos, W. S. C. Gurney, Nature 402, 653 (1999). replicated realization of a large stochastic in spectacular lemming cycles in northeast- 7. O. N. Bjørnstad, N. C. Stenseth, T. Saitoh, O. C. Lingjære, Res. Popul. Ecol. 40, 77 (1998). process.” ern Greenland, and how dynamics are 8. C. J. Krebs et al., Science 269, 1112 (1995). Elegant manipulations of predation have regulated in a top-down manner. Because 9. P. J. Hudson, A. P. Dobson, D. Newborn, Science 282, been undertaken in other cyclic rodent popu- most rodent species are found in widely di- 2256 (1998). 10. E. Korpimaki, K. Norrdahl, Ecology 79, 2448 (1998). lations but have resulted in divergent conclu- vergent environments and are embedded in 11. J. Sundell, Oikos 101, 416 (2003). sions (10–12). These experiments illustrate complex food webs involving competitors 12. I. M. Graham, X. Lambin, J. Anim. Ecol. 71, 946 (2002). 13. P. J. Hudson et al., Philos. Trans. R. Soc. 357, the scientific complexities that have faced the and resources, extrapolating the Gilg et al. 1259 (2002). “vole stranglers” for decades. The theories findings to other cyclic rodent populations is 14. F. Mougeot, S. M. Redpath, F. Leckie, P. J. Hudson, are sound and developed, the hypotheses are probably unwise. However, if we turn from Nature 421, 737 (2003). 15. O. Nielsen, J. Anim. Ecol. 68, 1034 (1999). elegant, and the predictions are clean. At the apparent simplicity of high arctic ecosys- 16. A. R. E. Sinclair, S. Mduma, J. Brashares, Nature 425,

times, however, researchers have been tempt- tems to arguably one of the most complex 288 (2003). CREDIT: OLIVIER GILG & BRIGITTE SABARD

798 31 OCTOBER 2003 VOL 302 SCIENCE www.sciencemag.org