RESEARCH ARTICLE Contingent Conclusions: Year of Initiation Influences Ecological Field Experiments, but Temporal Replication is Rare Kurt J. Vaughn 1,2 and Truman P. Young 1 Abstract Interannual variation in experimental field conditions pro- duce variability in the results of experiments monitored over multiple years, termed here “year effects.” When experimental treatments are replicated in separate years, interannual variation may influence treatment effects and produce significant treatment by initiation-year interac- tions. Understanding the frequency and strength of these effects requires initiating identical experiments across years. We conducted a review of literature covering more than 500 experimental articles published in 7 journals between 1966 and 2008. Only 5% of the 276 general eco- logical field studies initiated experiments in multiple years. This rarity was even more evident in the journal Restora- tion Ecology, in which none of the 173 surveyed experi- mental studies initiated experiments in multiple years. In contrast, 48% of the 58 field experiments published in an agronomy journal were replicated across years. We found only 17 studies that tested treatment by initiation-year interactions. Despite their rarity, 76% of these studies found significant interactions between treatment and initia- tion year. We conclude that the results of many ecological field experiments are likely to be contingent on the year in which they are implemented. We discuss the impor- tance of treatment by initiation-year interactions in ecology and restoration, factors that have hindered the inclusion of temporal replication in the past, and some suggestions for the appropriate design and analysis of temporally repli- cated experiments. We argue for more deliberate investiga- tion of temporal contingency in ecological experimentation, especially in the field of restoration ecology, which may be particularly sensitive to treatment by initiation-year interactions. Key words: agronomy, annual phenomena, historical con- tingency, interannual variation, multiple initiation-years, restoration ecology, treatment by initiation-year interac- tion, year effects. Introduction Interannual variations in biotic systems have been widely documented by ecologists, including such dramatic phenom- ena as masting (Kelly 1994), episodic recruitment (Estes & Duggins 1995), irruptions (Predavec 1994) and epidemics (Caceres et al. 2006), as well as more common variations in the demographies (Suarez et al. 2005; Lee et al. 2007) and behaviors (Callaghan et al. 1997; Rotenberry & Wiens 1998) of plants and animals. This interannual biotic variation often has substantial impacts on interspecific interactions (Riginos & Young 2007; Veblen 2008), the structure of communities (Pitt & Heady 1978; Bakker et al. 2003), and the functions of ecosystems (Sala et al. 1988; Knapp & Smith 2001). Despite widespread recognition of the importance of temporal vari- ation to ecological processes (hereafter, “year effects”), few ecological studies are designed to test how year effects interact with experimental results. 1 Department of Plant Sciences, University of California, Davis, CA 95616, U.S.A. 2 Address correspondence to K. J. Vaughn, email kjvaughn@ucdavis.edu © 2010 Society for Ecological Restoration International doi: 10.1111/j.1526-100X.2010.00714.x The outcome of an experiment depends on the conditions under which it is implemented. For example, because uncon- trolled variation in field conditions exists at different experi- mental sites, researchers recognize the importance of testing experimental conclusions across spatial variation. Replicat- ing experiments across multiple sites increases the confidence, generality, and applicability of conclusions (Krebs 1989). By this same logic, testing experimental conclusions across tem- poral variation by initiating experiments in multiple years also increases the validity of conclusions. Interannual variation in uncontrolled experimental field conditions (e.g. precipitation, temperature, or competitor/predator pressures) can drive dif- ferences in results for experiments conducted in different years (Johnson 2002; Young et al. 2005). The fundamental answers to ecological questions may depend on the year in which the question is asked (Walters et al. 1988). For instance, Bakker et al. (2003) initiated iden- tical treatments in three different years in order to test the con- tingency of grassland restoration techniques across interannual variation. A subset of these plots was then resampled 9–11 years after plot establishment (MacDougall et al. 2008). Both studies found that their experimental treatments had profoundly different effects in each of the three initiation SEPTEMBER 2010 Restoration Ecology Vol. 18, No. S1, pp. 59–64 59