Proc. Nati. Acad. Sci. USA Vol. 85, pp. 21%-2199, April 1988 Ecology On the advantage of being different: Nest predation and the coexistence of bird species (community structure/density-dependent predation/resource partitioning/search images) THOMAS E. MARTIN Department of Zoology, Arizona State University, Tempe, AZ 85287-1501 Communicated by Jared M. Diamond, December 23, 1987

ABSTRACT A long-standing debate in ecology centers on tors will increase for individual species with the cumulative identifying the processes that determine which species coexist density ofall similar species (ref. 17; also see refs. 23 and 24). in a local community. Partitioning of resources, where species Thus, such behavioral responses by the predators, if they differ in resource use, is often thought to reflect the primary exist, can favor coexistence of species that use different role of competition in determining coexistence of species. How- nesting sites (partitioning of resources, where nest sites are ever, in theory predation can favor similar patterns. This the resource) to minimize cumulative density effects (17). theory premises that predators increase their search intensity Here, I report experiments that support the premise that with increasing density ofprey. One set ofexperiments reported predator search intensity increases with the frequency of here supports this premise based on predators that search for occupied nest sites (reward rate). I then show that, for a bird nests. A second set of experiments documents that preda. given density of nests, nest predation is reduced when those tion rates are lower when nest sites are partitioned among nests are placed in sites that differ (partitioned nesting space) different sites than when the same number ofnests are placed in than when placed in similar sites (unpartitioned space). similar sites. Moreover, predation rates on experimental nests are more similar to rates on real nests when experimental nests STUDY AREA AND METHODS are partitioned among different sites. These results provide support for a hypothesis that nest predation is a process that can Study Area. Experiments were conducted in four mixed- favor coexistence ofbird species that partition resources, where conifer drainages in central Arizona at -2300-m elevation. nest sites are the resources. These drainages were of the same vegetation type and general location as other sites on which I have been studying nest predation (25, 26). Nest predators were identified using Local processes determine the number and types of species cameras outfitted with infrared light beams that trigger the that coexist in a local'community, within regional and histor- camera when the beam is broken. Ten cameras were set up ical constraints on availability of species (1). Competition is at artificial nests that were in the same drainages as experi- an example ofa local process. In fact, resource partitioning- mental nests but which were not included in the results i.e., differences among coexisting species in their use of presented here. Preliminary results based on 11 pictures resources-is commonly observed and often thought to re- indicated that the nest predators were red squirrels (Tamia- flect the primary role of competition in determining coexist- sciurus hudsonicus) and gray-neck chipmunks (Tamias cine- ence of species (2-5). However, the importance of competi- reicollis). These results coincide with my impressions on tion has been hotly debated, creating a need to examine these sites from observations of birds chasing these two alternative processes (6-9). Predation is an alternative proc- predators. However, other predators are also on the sites ess that can affect coexistence of species by mediating com- [e.g., long-tailed weasels (Mustelafrenata) and Steller's jays petitive interactions or by eliminating species (10-14). More (Cyanocitta stelleri)] and undoubtedly account for some nest importantly, theory suggests that predation can also favor losses. resource partitioning similar to competition (15-17). Thus, Experiments used artificial wicker nests baited with quail this theory provides an alternative explanation for a common (Coturnix coturnix) eggs. Artificial nests have been success- pattern (i.e., resource partitioning). Here, I test whether nest fully used in other studies to examine predation rates (27, predators exhibit the behaviors that can favor resource parti- 28). Moreover, I previously tested potential biases associ- tioning among coexisting bird species. ated with using artificial nests in the habitats studied here Nest predation is probably an important agent of natural and found that nests that simulate the appearance and selection for birds because such predation commonly is a position of real nests can elicit predation responses that primary limit on reproductive succes's (18). However, the reflect real trends (29). effect of predators depends on their searching behavior (17). Experiment 1. This experiment tested the response of Predators, in searching for bird nests, must examine some predators to reward frequency by modifying the frequency potential nest sites that are not occupied by eggs (no of egg-occupied nests in three treatments. In all treatments a reward). Increased reward frequency should reinforce constant number of nests (seven) was placed in small white search behavior. Consequently, both search intensity and firs (Abies concolor) in 10-m-diameter circles (referred to as the proportion of nests lost to predators should increase with clumps). Ten clumps were used in each treatment and placed the frequency or density of occupied nest sites (reward rate). about 25 m apart. Treatments were separated by 100 m. One, Moreover, predators can enhance their search efficiency by three, and all seven nests in each clump contained eggs in specializing on prey types and learning search images treatments one, two, and three, respectively. Thus, this de- (19-22). If predators do not discriminate among species with sign used 70 nests per treatment with 10, 30, and 70 nests similar nest sites, then the proportion of nests lost to preda- containing eggs in treatments one, two, and three, respectively. Experiment 2. This experiment tested the effect of nest were The publication costs of this article were defrayed in part by page charge space partitioning on predation rates. Four species payment. This article must therefore be hereby marked "advertisement" simulated. One nest type simulated an orange-crowned in accordance with 18 U.S.C. §1734 solely to indicate this fact. warbler (Vermivora celata) and was placed in a small de-

Downloaded by guest on September 27, 2021 2196 Ecology: Martin Proc. Natl. Acad. Sci. USA 85 (1988) 2197

pression in the ground under the stem of a deciduous shrub (usually big-toothed maple-Acer grandidentatum) (termed ground species). The second nest type simulated a hermit thrush (Catharus guttatus); it was covered with moss and placed about 1 m above ground in a small white fir (termed fir species). The third type was unmodified and placed 1 m above ground in maple [termed maple (1m) species], simu- lating a MacGillivray's warbler (Oporornis tolmiei). The fourth type was placed 3 m above ground in maple [termed maple (3m) species] simulating a black-headed grosbeak (Pheucticus melanocephalus). Nests were placed at =10-m intervals along two parallel transects (long continuous strips of sample area) that were a minimum of 25 m apart; exact distance between transects -J and among nests on a transect varied depending on the availability of a suitable nest site. The same spacing was 0 used in both of two treatments (to be described) to maintain constant density between treatments. Both treatments were included in each of three spatial replicates. Treatments were 6 9 12 15 separated by 100 m in each replicate, and spatial replicates DAYS EXPOSED were separated by 3-5 km. Experiments in all three spatial replicates were initiated at the end of May when egg-laying FIG. 1. Predation rate as a function of the ratio of egg-occupied nests to total nests in a clump. The three treatments include one, activity peaks. The experiments were then in late repeated three, and seven nests containing eggs in each clump. (a) Percentage June when many birds were renesting. This design used 80 of clumps remaining without any nests in the clump losing eggs to nests per spatial replicate (40/treatment) or 240 nests per predators. (b) Percentage of intact nests remaining (no egg loss to temporal replicate, with 480 nests used overall. predators). Treatment one simulated a four-species assemblage (par- titioned nesting sites). The four nest types were alternated arguably reflect an increasing probability of random encoun- sequentially along the two transects in each replicate, using ters of nests across treatments. However, the risk to indi- 40 nests per replicate. Treatment two simulated a single- vidual nests is shown by the percentage of intact nests (no species assemblage (unpartitioned sites); 20 nests of one egg loss) that were remaining at each nest check; the results species were placed along the transects followed by a 100-m show that even individual nests had a greater rate of preda- buffer and then a second set of 20 nests of a different species tion (F = 6.08, P < 0.02) with increases in reward frequency for a total of 40 nests in this treatment. This design, which (Fig. lb). only tested two single species per temporal replicate, was Experiment 2. Predation rates were greater (F = 515.89, P used to maintain constant sample size (40 nests) between < 0.03) in the single-species treatment across all spatial and treatments and because of time and space constraints on temporal replicates (Fig. 2). The higher predation rates could putting out additional nests. The ground species was tested Temporal Replicate 1 Temporal Replicate 2 in both temporal replicates to test for any temporal biases. 0-O Multiple species The other single species tested in the first and second 1 00*@-4 Single species temporal replicates were the fir z and maple (1m) species, 0 respectively. Thus, three of the four species were tested as single species. Statistical Tests. Predation rates on artificial nests were 0~ measured by examining loss of eggs from nests every 3 days

during 15 days of exposure to predators. Differences among 0 treatments were examined by comparing the percentage of m nests remaining without having lost any eggs to predators. These data were arcsine-transformed and analyzed by one-, z two-, or three-way, repeated measures, analysis of variance (ANOVA). One-way analyses were used for the first exper- iment. Two-way analyses were used in the second experi- ment for those comparisons that did not include temporal z replications. Otherwise, three-way analyses were used. Two- and three-way interactions with treatment effects were not significant (P > 0.10 in all cases). Lii 0( 0 Predation rates on real nests of the four species being z simulated were measured by examining loss of eggs every LA- io10 O W 3-4 days and using the Mayfield method, which measures w 4 L1Z'~O_ the proportion of nests that are lost to predators per day (30, 0~ 0 !O . 31). Differences in these daily mortality rates between arti- 2 0L.,4-- - ficial and real nests were tested with the z test (31). 0 h 0 3 6 9 12 15 0 3 6 9 12 1 RESULTS DAYS EXPOSED TO PREDATORS Experiment 1. Predation rates increased (F - 10.07, P < FIG. 2. Percentage of nests that lose no eggs to predators. Each with increased reward when based on the 0.01) frequency column is a temporal replicate, and each row is a spatial replicate. percentage of clumps remaining without any nests losing The multiple species treatment represents the combined predation eggs to predators (Fig. la). Given that the number of nests at rate for four species; the single-species treatment combines two sets risk increased across treatments, then such results may of single species in each temporal replicate. Downloaded by guest on September 27, 2021 2198 Ecology: Martin Proc. Natl. Acad. Sci. USA 85 (1988)

a Ground Species b Fir Species Maple (1 m) Species Temporal Replicate 1 Temporal Replicate 1 Temporal Replicate 2 1 0--O 1 -@ z 00t Multiple species 0op- Multiple species Single species 0 0!R w 0A 0-0- 0-- I 0 I-3: z 801 0K z 601 \ w l:: I- 20 C) -- w O+-~ . . =- * t z 1 0 LL 0 00.~ ol *l- 0 1-.~ z w 4 '1~ w 2 ol \X0-0-0-o oI EL 3 6 9 12 15 3 6 9 12 15 3 6 9 12 15 0 3 6 9 12 15 DAYS EXPOSED TO PREDATORS DAYS EXPOSED TO PREDATORS

FIG. 3. Percentage of nests that lose no eggs to predators. Each column is a temporal replicate, and each row is a spatial replicate. (a) Data in each treatment are based only on the nests that simulated the ground species in all spatial and temporal replicates. (b) Data from both treatments are based only on the nests that simulated the fir species for the first temporal replicate and the maple (im) species for the second temporal replicate. have occurred because the species used in this treatment have containing eggs. These results complement other data show- generally higher predation rates than those in the multiple- ing that predation rates on real hermit thrush nests are lower species sample; this hypothesis is unlikely because three of when greater numbers of white firs (unoccupied potential the four species were tested in the single-species treatment. nest sites) surround the nests (26). Thus, data from both real However, comparison of predation rates for the individual and artificial nests support the premise that predators modify species used in both treatments allows direct examination of their search behavior in response to the frequency of occu- this potential bias. Such comparisons show that predation pied nests that they encounter. rates are unequivocally greater in the single-species treatment The second set of experiments documents that predators = < (Fig. 3); predation rates were higher (F 30.06, P 0.04) for can exhibit the behaviors necessary to favor partitioning of the ground-nesting species in both temporal replicates (Fig. nesting space; predation rates were reduced when nests 3a) and for the fir species (F = 52.43, P < 0.02) and maple were partitioned among different sites. These results do not (im) species (F = 18.83, P < 0.05) in temporal replicates 1 and 2, respectively (Fig. 3b). Moreover, predation rates in the seem to be an abnormal response of predators to artificially nest multiple-species treatment were much more similar to preda- high densities given that predation rates in the multiple- tion rates on real nests than in the single-species treatment species treatment were similar to predation rates on real (Table 1). Predation rates can differ among years (unpub- nests (Table 1). The lower predation rates in the multiple- lished data), but such fluctuations should not influence the species treatment as compared with single-species treatment general pattern documented here because the results are may arise because the four nest types were so different that based on controlled treatments. different predator species specialized on each nest type; these predators may then have simply responded to the different DISCUSSION densities of individual nest types in the two treatments. The first experiment documents that predators can increase Photographs of predators at the two nest types (fir and their searching intensity with increasing frequency of nests ground) monitored with cameras suggest this to be an unlikely Table 1. Mean ± variance in predation rates, measured as daily mortality rates, for the artificial nests in the multiple- and single-species treatments and for real nests of the four species simulated in the treatments Daily mortality rate, nests lost to predators per da-y Multiple-species Single-species treatment Real nests treatment Ground 0.075 ± 0.0001 * 0.027 ± 0.0002 (17)t ** 0.210 ± 0.0004 Maple (im) 0.050 ± 0.0001 0.047 ± 0.0002 (22) ** 0.168 ± 0.0004 Fir 0.068 ± 0.0001 0.081 ± 0.0005 (19) * 0.154 ± 0.0004 Maple (3m) 0.035 ± 0.0001 0.012 ± 0.0001 (8) Predation rates on real nests are based only on nests monitored during the same season as experiments (1987) and only during the egg stage. *, P < 0.02 and **, P < 0.001 in comparisons between adjacent numbers. tSample size in parentheses, which equals the number of nests observed. Downloaded by guest on September 27, 2021 Ecology: Martin Proc. Natl. Acad. Sci. USA 85 (1988) 2199

explanation; both red squirrels and chipmunks were detected already differ due to differences in their individual evolution- taking eggs at both nest types (unpublished data). In addition, ary histories (17). Such effects emphasize the importance of observations suggest that these same predators also take eggs considering historical and regional processes in the structur- from the other two nest types (personal observation). The ing of communities (1, 17). lower predation rate in the multiple-species treatment may instead occur because multiple prey types inhibit develop- I thank J. Connell, J. Diamond, D. Levey, M. Douglas, T. A. ment of search images and thereby reduce foraging efficiency Markow, M. A. Newton, R. E. Ricklefs, J. J. Roper, T. W. (17, 21, 22). Separation of these two possibilites will require Schoener, and D. Simberloff for helpful comments on earlier drafts. K. Donohue, M. Godwin, T. McCarthey, and J. Soliday provided more and additional None- photographic data experiments. able assistance in putting out and monitoring the artificial nests. This theless, these experimental results show that there can indeed work was supported by Whitehall Foundation, Inc., and National be an advantage to being different; nest predation is signifi- Science Foundation (BSR-8614598). cantly reduced when nest sites differ. The two major predator species, or their sibling species, 1. Ricklefs, R. E. (1987) Science 235, 167-171. are widespread throughout forested North America (32), 2. Schoener, T. (1974) Science 185, 27-39. providing a basis for expecting the behavior documented 3. Diamond, J. (1978) Am. Sci. 66, 322-331. here to be widespread. Moreover, because nest predation is 4. Schoener, T. (1983) Am. Nat. 122, 240-285. commonly the primary source of nesting failure in avian 5. Martin, T. (1986) Curr. Ornithol. 4, 181-210. systems (18), the predation behavior documented here may 6. Wiens, J. (1977) Am. Sci. 65, 590-597. 7. Connell, J. (1983) Am. Nat. 122, 661-696. represent a process that commonly favors coexistence of 8. Simberloff, D. (1983) Am. Nat. 122, 626-635. species that partition nesting sites. Some evidence does 9. Strong, D. R., Jr., Simberloff, D., Abele, L. G. & Thistle, indeed indicate that partitioning of nesting space may be A. B., eds. (1984) Ecological Communities (Princeton Univ. common (17). Such patterns are supported by analogous Press, Princeton, NJ). arguments and evidence that predation may favor coexis- 10. Connell, J. (1975) in Ecology and Evolution of Communities, tence of species that differ in appearances or escape behav- eds. Cody, M. & Diamond, J. (Harvard Univ. Press, Cam- iors (23, 24). bridge, MA), pp. 460-490. 11. Sih, A., Crowley, P., McPee, M., Petranka, J. & Strohmeier, Other processes clearly may be acting concurrently with J. (1985) Annu. Rev. Ecol. Syst. 16, 269-311. predation, and the relative importance of these processes 12. Pacala, S. & Roughgarden, J. (1984) Oecologia 64, 160-162. will vary over time and among habitats and areas. However, 13. Savidge, J. A. (1987) Ecology 68, 660-668. the results documented here suggest that predation may 14. Schoener, T. W. & Spiller, D. A. (1987) Science 236, 949-952. provide an alternative explanation for some patterns of 15. Holt, R. D. (1977) Theor. Popul. Biol. 12, 197-229. resource partitioning. Moreover, previous work on preda- 16. Holt, R. D. (1984) Am. Nat. 124, 377-406. tion has focused on the loss of independent juveniles and/or 17. Martin, T. (1988) Evol. Ecol. 2, 11-24. 18. Ricklefs, R. E. (1969) Smithson. Contrib. Zool. 9, 1-48. adults, even though reproductive success is an important 19. Tinbergen, N. (1960) Arch. Neer. Zool. 13, 265-336. component of fitness. This study emphasizes that predation 20. Croze, H. (1970) Z. Tierpsychol. 5, 1-85. on eggs and dependent young can constitute another impor- 21. Persson, L. (1985) Oecologia 67, 338-341. tant level of selection. These points are underscored by 22. Lewis, A. C. (1986) Science 232, 863-865. demonstrating them on birds because predation effects have 23. Rand, A. S. (1967) Atas Simp. Sobre Biota Amazonica 5, been ignored in birds more than any other taxonomic group 73-83. (11). Indeed, studies of birds provided much of the original 24. Ricklefs, R. E. & O'Rourke, K. (1975) Evolution 29, 313-324. impetus for the long-held view that resource partitioning 25. Martin, T. (1988) Ecology 68, 74-84. induced by competition was a primary and widespread cause 26. Martin, T. & Roper, J. (1988) Condor 90, 51-57. of some that birds 27. Wilcove, D. S. (1985) Ecology 66, 1211-1214. species coexistence. Finally, data show 28. Loiselle, B. A. & Hoppes, W. G. (1983) Condor 85, 93-95. are relatively invariant in their nesting heights among geo- 29. Martin, T. (1987) Condor 89, 925-928. graphic regions (17). This evidence indicates that nest site 30. Mayfield, H. (1975) Wilson Bull. 87, 456-466. differences among coexisting species do not necessarily 31. Hensler, G. L. & Nichols, J. D. (1981) Wilson Bull. 93, 42-53. reflect local coevolution. Instead, the differences may reflect 32. Hall, E. R. (1981) The Mammals of North America (Wiley, selection for coexistence of species with nest sites that New York). Downloaded by guest on September 27, 2021