We undertook a quantitative literature review to assess implicit scaling ch
oices made in the design of enclosed experimental aquatic ecosystems (mesoc
osms). A database was constructed with information on temporal scale, spati
al scale, and other design characteristics for 360 experiments reported in
the literature. We found that key scaling variables such as complete physic
al dimensions were often not reported. Overall mesocosm experiments had a m
edian volume of 1.7 m(3) and median duration of 49 d. Volume and duration v
aried by habitat type, experimental treatment, number of trophic levels inc
luded, and the response variable under investigation. A number of experimen
tal design characteristics varied with mesocosm size. For instance, charact
eristics that decreased with increasing mesocosm size included the number o
f replicates and treatments: and the degree of experimental control over sp
ecies composition and the physical environment. We also found a bias toward
s scaling cylindrical containers for a constant ratio of radius to depth as
size was increased. This design choice inevitably alters surface-area-to-v
olume relationships as well as associated ecological variables such as gas
and light energy flux, benthicpelagic coupling, and the relative dominance
of periphyton growth on mesocosm walls. Our results indicate the need for b
oth 'scale sensitive' experiments, that explicitly consider scale in design
and interpretation of results, and;'multi-scale' experiments that manipula
te temporal and spatial attributes in order to test specific hypotheses reg
arding the ecological impact of modification in scale. Both types of experi
ments are prerequisites for improved mesocosm design and for the systematic
extrapolation of information from experimental ecosystems to nature.