Glass eel movements and shelter use were studied under experimental conditions in small indoor channels and in ponds. Shelters densities (2 stems of artificial plant, 9 stems) were controlled in indoor channels stocked with 55 ind./m2. Glass eel densities were controlled in ponds (1,6 - 6 9 and 12 ind./m2). Experiments lasted for two months and half in channels, and one month and a half in ponds. In indoor channels, glass eels highly looked for shelters after 5-6 days. During this phase, which lasted for 1 month and half, individuals were very gregarious. Thereafter, fish became aggressive and stopped using shelters. In ponds, emigration at the outlet was almost null. At the inlet, 3 phases were observed : the first one (15 days) with many captures, the second one (1 month) without any emigration and the third one, just before the end of the experiment, with few catches. The emigration rate was independent from initial densities, but increased in ponds with no hydrophyte development. The emigration at the inlet occurred for only a short period in the high-density ponds (first 5-6 days), whereas it lasted longer in the low-density ponds (15-20 days). After four weeks without captures, growth appeared high, heterogeneous and associated to low survival rates in the ponds with initial densities less than 10 ind./m2. At higher densities, growth was lower, but homogeneous and associated with good survival rates. That suggests the development of agonistic behaviour or cannibalism in the lowest density ponds linked with a faster growth of some individuals. In high-density ponds, glass eels would exhibit a gregarious behaviour, leading to high survival and homogeneous but low growth. In these ponds the development of agonistic behaviour would be delayed possibly until fish would have reached a certain size and/or large interindividual differences in size have appeared.