Using tracking radars, we investigated the variability of flight directions of long-distance nocturnal passerine migrants across seasons (spring versus autumn migration) and sites at the southern (56° N) and northern (68° N) ends of the Scandinavian Peninsula (Lund versus Abisko). Whilst most migrants at Lund are on passage to and from breeding sites in Fennoscandia, the majority of the migrants at Abisko are close to their breeding sites, and migration at Abisko thus to a large degree reflects initial departure from breeding sites (autumn) or final approach to breeding destinations (spring). The radar data were used to test predictions about differences in orientation and wind drift effects between adult and juvenile birds (a large proportion of autumn migrants consists of juvenile birds on their first journey), between situations far away from or near the goals and between different phases of migration (initial departure, en route passage, final approach to goal). The concentrations (both total and within-night concentrations) of flight directions differed significantly between seasons as well as sites, with the highest concentration at Lund in spring (mean vector length of track directions, r = 0.79) and lowest at Abisko during spring (r = 0.35). Partial wind drift and partial compensation were recorded at Lund, with a similar effect size in spring and autumn, whilst possible wind drift effects at Abisko were obscured by the large directional scatter at this site. The results from Lund support the prediction that the high proportion of juveniles in autumn contributes to increase the directional scatter during this season, whilst there was no support for predictions of differential wind drift effects between seasons and situations with different goal distances. The most striking and surprising result was the exceedingly large scatter of flight directions at Abisko, particularly in spring. We suggest that such an exaggerated scatter may be associated with final approach orientation, where migrants reach their specific goals from all various directions by final navigation within a more wide-ranging goal region. The larger scatter of autumn flight directions at Abisko compared to Lund may be due to exploratory flights in variable directions being more common at initial departure from breeding sites than later during migratory passage. These surprising results highlight the importance of studying and analysing orientation during final approach to (and initial departure from) migratory goals for understanding the orientation systems of migratory birds.