We conclude that differences in popula- tion variability were affected by the deter- ministic characteristics of the population dy- namics as well as the stochastic factors (Fig. 3, C and D). These differences were in turn related to the position of the species along slow-fast continuum of life history variation (3, 5, 6 ) as expressed by r 1 (Fig. 3B). The relative contribution of the parameters de- scribing the expected dynamics and stochas- tic factors to the population fluctuations is in solitary birds likely to be closely related to the type of demographic process, i.e., wheth- er it is recruitment-driven or survival-restrict- ed (Fig. 4). Fluctuations in the size of recruit- ment-driven populations are more strongly influenced by environmental stochasticity than survival-restricted populations. Conse- quently, reliable projections of avian popula- tions will require precise estimates and mod- eling of stochastic as well as deterministic components of the dynamics. On the other hand, estimating the form of density depen- dence will be more important for predicting population fluctuations of long-lived species with high values of . Reliable estimates of the environmental stochasticity (11) as well as precise estimates of the carrying capacity (21) will require access to long-term time series with small sampling errors (29). References and Notes 1. R. M. May, Philos. Trans. R. Soc. London 351, 1951 (1999). 2. R. C. Lewontin, in The Genetics of Colonizing Species, H. G. Baker, G. L. Stebbins, Eds. (Academic Press, New York, 1965), pp. 77–91. 3. E. R. Pianka, Am. Nat. 104, 592 (1970). 4. D. E. L. Promislow, P. H. Harvey, J. Zool. 220, 417 (1990). 5. E. L. Charnov, Life History Invariants (Oxford Univ. Press, Oxford, 1993). 6. B.-E. Sæther, Ø. Bakke, Ecology 81, 642 (2000). 7. T. R. E. Southwood, J. Anim. Ecol. 46, 337 (1977). 8. C. W. Fowler, Ecology 62, 602 (1981). 9.  , Evol. Ecol. 2, 197 (1988). 10. M. E. Gilpin, F. J. Ayala, Proc. Natl. Acad. Sci. U.S.A. 70, 3590 (1973). 11. B.-E. Sæther et al., Am. Nat. 151, 441 (1998). 12. O. Diserud, S. Engen, Am. Nat. 155, 497 (2000). 13. Supplementary material is available on Science On- line at www.sciencemag.org/cgi/content/full/295/ 5562/2070/DC1. 14. We based our analyses on time series of populations of solitary bird species that have been censused for 15 or more years with no significant linear trend in popula- tion size with time and no population estimate of less than 10 pairs. To reduce the bias in the parameters because of large sampling errors in population esti- mates, we only included time series that were based on direct nest counts (e.g., hole nesters) or on the presence of a large number of color-ringed individuals. The pop- ulation parameters were estimated by maximum like- lihood methods. We modeled fluctuations in the size of the logarithm of the population fluctuations, X = ln N, where N is the population size at time t. Let X = ln(N +N) - ln (N) and  e 2 be the environmental stochasticity (23). We assume large enough population sizes to ignore any effects of demographic stochasticity. The distribution of X, conditional on N, is assumed to be normal with mean m(x) = r m [1 - (N/K)  ]t and variance  e 2 t . Here K is the carrying capacity, r m is the mean specific growth rate, and  describes the type of density regulation in the theta-logistic function (10) m(x). The strength of the density regulation at K is then (21)  = r m = r 1 /(1 - K - ), where r 1 is the specific growth rate when N = 1. Thus, strong density regulation occurs at K when the population growth rate is high and/or for large values of . The parameters were estimated by maximum likelihood. The expected change in the logarithm of population size may be written as E(X) = r 1 {1 - [(e X - 1)/(e ln(K) - 1)]} =(X, K, ) for  0 and E(X) = r 1 [1 - (X/ln K)] for = 0(21, 23). The parameters are estimated by maximizing the likeli- hood function L(K,, e 2 ) =  i=1 n f X i+1 X i ;K,, e 2 ), where X 0 = K, and X i+1 = X i + Xi . The function f(X i+1 X i ; K, ,  e 2 ) is the normal distribution with mean X t +(X i , K, ) and variance  e 2 . The uncertainty in the parameters was determined by simulating repeated data sets from the model, with the maximum likelihood estimates of the parameters, and then calculating bootstrap repli- cates from each simulation. Reliable estimates of r 1 are often difficult to obtain in stationary time series be- cause the populations most of the time are found fluctuating around K (24). We therefore estimated this parameter by a standard Leslie matrix model (6 ) using available demographic information. We entered into the model maximum fecundity rate (number of off- spring produced to independence per female) and the lowest age specific mortality rate recorded during the study period or in any age class. The estimates of r 1 obtained in this way were in nine bird species (25 ) closely correlated to bias-corrected values of r 1 ob- tained from time-series analyses (correlation coeffi- cient = 0.73, P  0.05, n = 9). 15. We assume that the parameters (, ln  e 2 ) are binor- mally distributed among the species and that the estimates of these parameters are approximately binormally distributed for each species with param- eters determined by the results of the parametric bootstrapping (14 ). Each pair of estimates (, ln  e 2 ) will then have bivariate normal distributions. Esti- mates of the binormal variation among species are found by maximum likelihood. 16. C. W. Fowler, Curr. Mammol. 1, 401 (1987). 17. T. Royama, Analytical Population Dynamics (Cam- bridge Univ. Press, Cambridge, 1992). 18. A problem in analyses of time series of population fluctuations is that the coefficient of variation in- creases with the census period (26, 27 ) because of autocorrelations due to age structure effects, delayed density dependence, or autocorrelation in the envi- ronmental noise (28, 29 ). However, from the diffu- sion approximation for the theta-logistic model, we can calculate the variance of the stationary distribu- tion (12). If we assume the demographic variance  d 2 to be much less than the product of the carrying capacity K and environmental stochasticity  e 2 , we can compute the variance of the stationary distribu- tion of population sizes  N 2 = K 2 [(+ 2)/]/[(+ 1)/] 2/ (/), where = 2r 1 / e 2 (1 - K - ) - 1, r 1 is the growth rate when N = 1, and  denotes the gamma function (12). Thus,  N 2 can be used to com- pare the variability in time series of different lengths. However, in the present data set, a close correlation was found between  N 2 and CV (correlation coeffi- cient = 0.94, P  0.001, n = 13). 19. I. Newton, Population Limitation in Birds (Academic Press, San Diego, 1998). 20. P. Arcese, J. N. M. Smith, W. M. Hochachka, C. M. Rogers, D. Ludwig, Ecology 73, 805 (1992). 21. B.-E. Sæther, S. Engen, R. Lande, P. Arcese, J. N. M. Smith, Proc. R. Soc. London Ser. B 267, 621 (2000). 22. R. Lande et al., Am. Nat., in press. 23. S. Engen, Ø. Bakke, A. Islam, Biometrics 54, 840 (1998). 24. S. Aanes, S. Engen, B.-E. Sæther, T. Willebrand, V. Marcstro ¨m, Ecol. Appl., 12, 281 (2002). 25. S. Engen, B.-E. Sæther unpublished data. 26. S. L. Pimm, The Balance of Nature? (Univ. of Chicago Press, Chicago, 1991). 27. A. Arin ˜o, S. L. Pimm, Evol. Ecol. 9, 429 (1995). 28. T. Coulson et al., Science 292, 1528 (2001). 29. O. N. Bjørnstad, B. T. Grenfell, Science 293, 638 (2001). 30. We thank H. Weimerskirch and A. Dhondt for access to unpublished data and R. Lande and R. Aanes for critically reading a previous draft. Supported by the Research Council of Norway, the EU Commission ( project METABIRD), and a grant from Programme International Cooperation Scientifique (CNRS) to B.-E.S. 7 December 2001; accepted 21 January 2002 The RAR1 Interactor SGT1, an Essential Component of R Gene–Triggered Disease Resistance Cristina Azevedo, 1 * Ari Sadanandom, 1 * Katsumi Kitagawa, 2 Andreas Freialdenhoven, 3 Ken Shirasu, 1 † Paul Schulze-Lefert 1,3 Plant disease resistance (R) genes trigger innate immune responses upon patho- gen attack. RAR1 is an early convergence point in a signaling pathway engaged by multiple R genes. Here, we show that RAR1 interacts with plant orthologs of the yeast protein SGT1, an essential regulator in the cell cycle. Silencing the barley gene Sgt1 reveals its role in R gene–triggered, Rar1-dependent disease resistance. SGT1 associates with SKP1 and CUL1, subunits of the SCF (Skp1- Cullin–F-box) ubiquitin ligase complex. Furthermore, the RAR1-SGT1 complex also interacts with two COP9 signalosome components. The interactions among RAR1, SGT1, SCF, and signalosome subunits indicate a link between disease resistance and ubiquitination. Plant disease resistance (R) genes are key com- ponents in pathogen perception; R genes acti- vate a battery of defense reactions, collectively called the hypersensitive response (HR) (1). A number of R genes from various plant species have been isolated and characterized in detail (2). Although different R genes confer resis- tance to a variety of pests, including bacteria, viruses, fungi, nematodes, and insects, many R gene products share common structural mod- ules such as a nucleotide-binding site and a leucine-rich repeat domain. The structural sim- R EPORTS www.sciencemag.org SCIENCE VOL 295 15 MARCH 2002 2073