being attacked per day exposed [for further validation of this response, see (33)].
Consistent with our predictions, we found that
predation rates were highest at the equator and
decreased significantly toward the poles (Fig. 1A;
F1,27.8 = 10.28, P = 0.003). For every 1° of latitude
away from the equator, the daily odds of a caterpillar being attacked decreased by 2.7% (odds ratio
0.973, confidence limits 0.959 to 0.987; Fig. 1A).
Thus, at the highest latitude studied (74.3°N;
Zackenberg, Greenland), the daily odds of a caterpillar being attacked by a predator were only 13%
(odds ratio 0.131, confidence limits 0.046 to 0.376)
of the odds at the equator. Predation rates also decreased with increasing elevation (F1,27.1 = 6.35, P =
0.02; Fig. 1D), independent of latitude (i.e., no latitude by elevation interaction; F1,27.8 = 0.70, P = 0.41).
For every 100 m moved upward from sea level, the
daily odds of predation decreased by 6.6% (odds
ratio 0.934, confidence limits 0.884 to 0.987). At the
highest elevation studied (2106 masl, table S1), the
daily odds of predation were 24% of those at sea level
(odds ratio 0.238, confidence limits 0.074 to 0.765).
Notably, higher predation at lower latitudes
and elevations was due to more frequent attacks
by arthropod predators. The daily odds of a caterpillar suffering an arthropod attack decreased by
3.5% for every 1° latitude moved away from the
equator (odds ratio 0.966, confidence limits 0.947
to 0.984, F1,25.1 = 14.11, P < 0.001), as did the odds
of attack marks that could not be attributed to
any specific predator group (odds ratio 0.972,
confidence limits 0.954 to 0.991, F1,24.3 = 9.57, P =
0.005; Fig. 1C). By contrast, we found no evidence
for a gradient in predation by birds or mammals;
the frequencies of attack marks by these predator groups were unrelated to latitude (F1,26.6 =
1.20, P = 0.28 and F1,28.6 = 2.9, P = 0.10, respectively). These latitudinal patterns in predation rate
were mirrored across elevation: The odds of a
caterpillar suffering arthropod attack decreased
by 9.6% for every 100 m moved upward from sea
level (odds ratio 0.904, confidence limits 0.839 to
0.975, F1,26.1 = 7.48, P = 0.01), whereas the odds of
receiving attack marks not attributable to any
specific predator group (F1,21.3 = 0.18, P = 0.68)
or of being attacked by birds (F1,29.3 = 1.86,
P = 0.18) or mammals (F1,25.0 = 0.63, P = 0.44)
were unrelated to elevation (Fig. 1F).
Overall, our study reveals a strong latitudinal
and elevational signature on biotic interaction
strength (i.e., predation rates) across the globe.
In doing so, it provides a clear pattern that can
be used to inform future efforts in this field (3–7 )
and to move beyond the obstacle of contradic-
tory evidence from variable methodologies among
studies conducted at different subsets of latitude
(7, 13). The parallel patterns in predation across
elevation [compare (21)] suggest that the ecolog-
ical factors constraining predation rates are likely
to show concordant latitudinal and elevational
gradients (34). The clarity of our findings offers a
simple lesson: To unmask a global ecological pat-
tern, we may need to apply standardized methods
to specific hypotheses determined a priori, rather
than combine data derived from different meth-
ods a posteriori [compare (8, 13, 34)]. This study
thus illustrates the power of simple, low-cost, glob-
ally distributed experiments [compare (17, 18, 35)].
We found that global gradients in predation
rate were driven by arthropod predators, with no
systematic trend in attack rate by birds or mam-
mals. This latitudinal shift in the relative im-
portance of different predator groups has clear
implications for understanding herbivore evo-
lution, interpreting global patterns of herbivory,
and understanding global community organi-
zation and functioning. In terms of arthropod
herbivore evolution, much theory has been de-
veloped on latitudinal patterns in plant defense
against herbivores, suggesting that if plants at
lower latitudes suffer high herbivory, they need to
evolve stronger defenses [e.g., (10, 11, 15)]. Our find-
ings motivate an analogous theory for defense
deployed by herbivores against their predators.
In the tropics, the fraction of model caterpillars
attacked per day is notably high (Fig. 1), and
attack rates on live prey tend to be even higher
(35). Thus, predation in the real world creates a
very real selection pressure. This leads to the test-
able hypothesis that arthropod herbivores in the
tropics should be better defended than those at
higher latitudes and that these defenses should
target arthropod rather than vertebrate predators
Real herbivores accumulate predation risk over
their development time, which may be shortened
in warmer climates and thus counteract our predictions for the ecological and evolutionary
impact of predation gradients. Although a comprehensive assessment goes beyond the current
study, we analyzed larval development times from
the available literature (table S2), finding a much
lower latitudinal effect on development times
than on predation rates (table S3). Hence, as a
net effect, we expect increased selection pressure on larval herbivores at lower latitudes.
From a plant perspective, the patterns detected
in this study suggest increased per capita predation pressure on plant consumers toward the
tropics and strong differences in the relative
impacts of different predator groups across different regions of the globe (Fig. 1C). This finding
suggests markedly different roles for different
predator groups in regulating herbivore abundance and traits across geographic gradients,
and potential differences in trophic organization
between regions. Whether the patterns revealed
by the current study translate into patterns in
net herbivory is unresolved, particularly considering that our experiments took place in the
understory, whereas most primary production
takes place in the canopy layer. Although seminal
findings suggested latitudinal gradients in herbivory and plant defense (15, 16), recent appraisals of the evidence find less support for these
trends (7, 8, 14). Nonetheless, the lack of clear
patterns in herbivory may be as much a reflection of variable methods as a true lack of a pattern (7, 13, 15, 16, 36). The current study should
stimulate further standardized comparisons of
species interactions, facilitating a clearer view of
these key biological patterns to enlighten our
search for their drivers and consequences.
REFERENCES AND NOTES
1. H. Hillebrand, Am. Nat. 163, 192–211 (2004).
2. J. H. Lawton, Oikos 84, 177–192 (1999).
3. D. W. Schemske, G. G. Mittelbach, H. V. Cornell, J. M. Sobel,
K. Roy, Annu. Rev. Ecol. Evol. Syst. 40, 245–269 (2009).
4. D. S. Gruner et al., Ecol. Lett. 11, 740–755 (2008).
5. H. Hillebrand, J. Phycol. 45, 798–806 (2009).
6. J. H. Lambers, J. S. Clark, B. Beckage, Nature 417, 732–735
7. A. T. Moles, S. P. Bonser, A. G. B. Poore, I. R. Wallis, W. J. Foley,
Funct. Ecol. 25, 380–388 (2011).
8. N. R. Andrew, I. R. Roberts, S. J. Hill, Open J. Ecol. 2, 202–213
9. M. Schleuning et al., Curr. Biol. 22, 1925–1931 (2012).
10. T. Dobzhansky, Am. Sci. 38, 209–221 (1950).
11. R. H. MacArthur, Geographical Ecology: Patterns in the
Distribution of Species (Princeton Univ. Press, 1972).
12. S. C. Pennings, B. R. Silliman, Ecology 86, 2310–2319 (2005).
13. D. N. Anstett, K. A. Nunes, C. Baskett, P. M. Kotanen, Trends
Ecol. Evol. 31, 789–802 (2016).
14. S. Zhang, Y. Zhang, K. Ma, J. Ecol. 104, 1089–1095 (2016).
15. P. D. Coley, T. M. Aide, in Plant–Animal Interactions:
Evolutionary Ecology in Tropical and Temperate Regions,
P. W. Price, T. M. Lewinsohn, G. W. Fernandes, W. W. Benson,
Eds. (Wiley, 1991), pp. 25–49.
16. P. D. Coley, J. A. Barone, Annu. Rev. Ecol. Syst. 27, 305–335
17. L. F. Fraser et al., Front. Ecol. Environ. 11, 147–155 (2013).
18. E. T. Borer et al., Methods Ecol. Evol. 5, 65–73 (2014).
19. P. A. Low, K. Sam, C. McArthur, M. R. C. Posa, D. F. Hochuli,
Entomol. Exp. Appl. 152, 120–126 (2014).
20. K. Tvardikova, V. Novotny, J. Trop. Ecol. 28, 331–341 (2012).
21. K. Sam, B. Koane, V. Novotny, Ecography 38, 293–300 (2015).
22. A. J. Lotka, J. Phys. Chem. 14, 271–274 (1910).
23. V. Volterra, Mem. Acad. Lincei Roma 2, 31–113 (1926).
24. P. A. Abrams, Annu. Rev. Ecol. Syst. 31, 79–105 (2000).
25. J. B. Shurin et al., Ecol. Lett. 5, 785–791 (2002).
26. L. Abdala-Roberts, K. A. Mooney, Ecology 95, 2879–2893
27. J. S. Bale et al., Glob. Change Biol. 8, 1–16 (2002).
28. C. A. Deutsch et al., Proc. Natl. Acad. Sci. U.S.A. 105,
29. E. L. Zvereva, M. Kozlov, Glob. Change Biol. 12, 27–41 (2006).
30. E. D. Currano et al., Proc. Natl. Acad. Sci. U.S.A. 105,
31. P. Wilf, New Phytol. 178, 486–502 (2008).
32. T. Alerstam, A. Hedenstrom, S. Akesson, Oikos 103, 247–260
33. Materials and methods are available as supplementary materials.
34. W. K. Petry et al., Science 353, 69–71 (2016).
35. G. L. Lövei, M. Ferrante, Insect Sci. 10.1111/1744-7917.12405 (2016).
36. X. Moreira et al., Ecol. Lett. 17, 537–546 (2014).
Data are archived at the Dryad digital repository,
doi:10.5061/dryad.j432q. We acknowledge funding from the
Academy of Finland (grants 138346, 276909, and 285803
to T.R.); European Science Foundation (grant 669609 to V.N.);
National Science Foundation (grant OPP 0908502 to A.A.
and L. Gough, grant DGE-1321846 to C.S.N., and grant 1158817
to E.N.); Instituto de Ecología, A.C. (field work grant to W.D.);
Finnish Cultural Foundation, Oskar Öflunds Stiftelse, and
Societas Entomologica Helsingforsiensis (grants to T.H.);
Natural Environment Research Council (NERC) (grant NE/
J011169/1 to O. T.L.); São Paulo Research Foundation (FAPESP)
(grant 14/11676-8 to E.N. and grant 13/23457-6 to J. P. Metzger
for E.N.); Grant Agency of Czech Republic (grant 14-32024P
to K.S., grant 14-04258S to V.N., and grant 14-04258S to A.S.;
National Feasibility Program I (LO1208 “TEWEP” to A.S.);
Estonian Ministry of Education and Research (grant IUT20-33 to T. T.);
and Norwegian Research Council’s Climate Change and Impacts in
Norway (NORKLIMA) program (grant 230607/E10 to V.V.).
Materials and Methods
Tables S1 to S3
12 September 2016; accepted 6 April 2017
744 19 MAY 2017 • VOL 356 ISSUE 6339 sciencemag.org SCIENCE
RESEARCH | REPORTS