Dispersal is a key process in metapopulations, as migrants genetically connect populations and enable the colonization of empty habitat patches. Sub-populations may diVer in their numerical contribution of migrants within a metapopulation. This has strong implications on evolutionary and ecological dynamics and has led to two diVerent hypotheses about the Daphnia metapopulation studied here: the assessment by some authors is that subpopulations contribute equally to the production of migrants, while others have postulated long-lived core populations in large "mainland" habitat patches as the dominant source of migrants. We have studied the resting and dispersal stage (ephippium) in a natural Daphnia metapopulation and in mesocosm experiments, and tested for eVects of habitat size and summer desiccation. We found that a 1000-fold increase in rock pool volume resulted on average in only in a 2.8-fold increase in ephippium production. Mesocosm experiments conWrmed these results: a 1000-fold increase of the mesocosms' volume resulted in a 7.2-fold increase in ephippium production. Additionally, we showed that ephippium production did not depend on the initial population size. Thus, populations in small pools may contribute only marginal fewer potential migrants in the whole metapopulation than populations in large pools. In a second mesocosm experiment we found that summer desiccation, which is a typical occurrence in small pools, is not detrimental for the populations. Daphnia hatched out of ephippia that were produced earlier within the same season and built up viable populations again. The substantial production of ephippia by populations in small pools suggests that these populations might be important for both the dynamics and global stability of metapopulations.