The following piece, containing citations of scientific literature, appeared in the February 2004 issue of MPA News without the citations.

Moving the Discussion About Marine Reserve Science Forward

By Benjamin Halpern, Robert R. Warner, and Steven D. Gaines

It was with great interest and concern that we read the piece by Willis et al. in the last issue of MPA News (MPA News 5:6).  Although we fully support healthy scientific debate, we feel that this piece misrepresents the current state of knowledge about marine reserves.  This is not an issue of differing opinions about what marine reserves can and cannot do; instead it is a question of what the science of marine reserves can and cannot tell us.  We appreciate the call for more rigor in marine reserve monitoring.  As Willis et al. note, some reserve monitoring schemes have been poorly conceived.  Nonetheless, our current state of knowledge is not nearly as dire as they depict.

Methods for evaluating reserve effects

Willis et al. claim we know little about how reserves affect fish species, since few monitoring studies show density increases greater than 100%.  However, posing an arbitrary one-tailed "minimum criterion" is misguided (i.e., we cannot presume only positive effects); species should vary in their response, including some declines (Micheli et al. in press).  More importantly, the Willis et al. approach ignores the power of synthesis.  Meta-analyses of reserve effects have shown that there is a strikingly consistent increase in density, size, and diversity of exploited stocks within reserves (e.g., Halpern 2003).  This is what meta-analyses do: when single studies have limited statistical power, consistency in the direction of changes across multiple studies can uncover real and significant patterns.

Willis et al. also misrepresent the significance of variable outcomes from theoretical models. Variable outcomes arise in theoretical studies because they include different assumptions or processes.  Syntheses of modeling results gain power from variable outcomes.  This variability focuses our attention on critical mechanisms, generates hypotheses to test empirically (e.g., Gerber et al. 2003, Gaines et al. 2003), and provides insight into problems that are experimentally intractable.  Although we are unaware of any management decisions that have been based solely on reserve theory, the future development of these models should enhance management decisions, rather than obscure them.

Inadequate experimental design

How many studies of reserve effects have had perfect experimental designs?  None, as Willis et al. point out.  But enough have had designs that allowed a test for the presence and the potential effects of the five experimental shortcomings Willis et al. identify.  The results we discuss here come from published analyses (Halpern et al. in press; Halpern & Warner 2002):

1.  Insufficient sample replication: Although it is theoretically possible to find a location where multiple reserves could be established to ensure proper replication, it is politically and logistically unlikely.  Meta-analysis is a way to overcome this limitation.

2.  Spatial confounding (reserve sites are always in better areas than control sites): Syntheses of results from Before-After-Control-Impact (BACI) studies show that although initial conditions in reserve and control sites differ, there is no bias or trend in these differences.

3.  Lack of temporal replication: Reserve studies have monitored reserve impacts over years to decades, and these studies show a rapid response for some species and a slower response for others.  On average, however, responses inside reserves typically exceed those from associated control sites.

4.  Lack of treatment (reserve) replication: Most reserves have been established in isolation rather than in networks or multiple reserves.  Recent noteworthy exceptions (e.g., the Channel Islands of California) will provide such treatment replication for future studies.  Until then, meta-analysis is a powerful way to overcome this limitation.

5.  Non-random placement of reserves: Again, BACI studies have shown that control sites are just as often placed in better than in worse locations.  There is no a priori reason to expect a reserve effect to be positive.

Finally, while it is true that "with a sufficiently large sample size, a statistically significant difference between two sites can almost always be obtained due simply to true natural biological variability between the sites," this expectation is two-tailed (the difference could be positive or negative) unless one assumes that reserve sites always start out better than control sites (existing studies suggest this is not true).  Sadly, no study of marine reserves has had the luxury of detecting statistically significant yet trivial effects, since sample sizes are consistently too small.

Lessons for reserve managers

Rather than simply instructing managers to be more rigorous in reserve design, we think it is more useful to point out the potential problems with reserve monitoring, and to suggest practical solutions (Halpern et al. in press).

First, because of the expected export function of reserves, there are no true controls even when habitats vary little among locations; sites outside of reserves may show increases (from export) or decreases (due to displaced fishing effort).  While most existing reference areas have shown increases following the establishment of nearby reserves (Halpern et al. in press), many more studies are needed.  Monitoring designs based upon analyses of gradients in responses as a function of distance from reserves may prove especially insightful.

Second, meta-analyses suggest the level of exploitation, life history, and trophic level of a species can strongly affect its response to protection, and monitoring schemes and expectations must take these expected differences into account.  Arbitrary "minimum criteria for biological significance" ignore basic ecological principles.

Third, power analyses using known natural variation in recruitment and population size in marine species suggest that many effects of marine reserves will be difficult to detect even in the best of circumstances.  This variation must be taken into account in both monitoring design and in setting time limits for marine reserve performance.


If the requirement for an adequate and appropriate experimental design for testing reserve effectiveness is multiple (and identical) reserve and control sites, measured before and after reserve creation, across multiple times, that demonstrate at least a 100% increase in some measure of interest, for all species, as Willis et al. suggest, then we will never have a good empirical test of reserve effects.  Furthermore, reserves are not established as perfectly designed ecological experiments; they arise from political processes with inherent compromises.  Should we then ignore reserves as a conservation and management option?  We argue that this is not only conservative beyond reason, but that it also ignores the power and lessons of syntheses of both data and models. We have learned much about the science of marine reserves in the past decade, and this guides us toward areas that require further research.  Importantly, we currently know enough about marine reserves to justify their use and implementation as one of several conservation and resource management tools.

Ben Halpern
National Center for Ecological Analysis and Synthesis, University of California, Santa Barbara, CA 93106, USA. E-mail:

Bob Warner
Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, CA 93106, USA. E-mail:

Steve Gaines
Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, CA 93106, USA. E-mail:

Gaines, S.D., B. Gaylord, and J.L. Largier. 2003.  Avoiding current oversights in marine reserve design.  Ecological Applications 13: S32-S46.

Gerber, L.R., L.W. Botsford, A. Hastings, H.P. Possingham, S.D. Gaines, S.R. Palumbi, and S. Andelman. 2003.  Population models for marine reserve design: a retrospective and prospective synthesis.  Ecological Applications 13: S47-S64.

Halpern, B.S.  2003.  The impact of marine reserve: do reserves work and does reserve size matter?  Ecological Applications 13: S117-S137.

Halpern, B.S., S.D. Gaines, and R.R. Warner.  2002.  Letter to the editor.  MPA News 4(5): 5.

Halpern, B.S., S.D. Gaines, and R.R. Warner.  In press.  Export of production and the displacement of effort from marine reserves: effects on fisheries and monitoring programs.  Ecological Applications.

Halpern, B.S. and R.R. Warner.  2002.  Marine reserves have rapid and lasting effects.  Ecology Letters 5: 361-366.

Micheli, F., B.S. Halpern, L. Bostford, and R.R. Warner.  In press.  Trajectories and correlates of community change in no-take marine reserves. Ecological Applications.

conference list
submission guidelines
contact us