Site attachment in young birds has sometimes been considered to take place during a sensitive phase as the result of an imprinting-like process. We conducted two four-year-long experiments on winter-site attachment in young migratory dark-eyed juncos (Junco hyemalis). In both, we caught birds over a range of dates, displaced them to large outdoor enclosures, and released them near the end of winter. In the first experiment, release was at the capture site and all returns next winter were to that site. The return rate did not vary with date of capture and confinement; therefore attachment could have been completed either before capture or after release. The absence of returns to the confinement site suggests that attachment to a place requires that the individual live free there. In the second experiment, release was at the confinement site, and many subjects remained there, probably until they migrated northward. Returns next winter were to both the capture site (the majority) and the confinement-release site. The fact that a few juncos formed a bond to the confinement-release site in this experiment reinforces the view that living free is necessary for site attachment. Individuals caught and confined earlier in winter tended to be less likely to return than those caught later. The distribution of returns according to date of capture indicates that attachment was completed at different dates in different subjects, which is not consistent with imprinting during a clearly defined sensitive period. However, young juncos hatch and migrate at various dates. Therefore our experiment does not rule out the existence of a brief sensitive phase for site attachment, its dates varying among individuals because age and/or timing of migration vary.