• Using the wind-dispersed plant Mycelis muralis, we examined how landscape fragmentation affects variation in seed traits contributing to dispersal. • Inverse terminal velocity (V(t)¯¹) of field-collected achenes was used as a proxy for individual seed dispersal ability. We related this measure to different metrics of landscape connectivity, at two spatial scales: in a detailed analysis of eight landscapes in Spain and along a latitudinal gradient using 29 landscapes across three European regions. • In the highly patchy Spanish landscapes, seed V(t)¯¹ increased significantly with increasing connectivity. A common garden experiment suggested that differences in V(t)¯¹ may be in part genetically based. The V(t)¯¹ was also found to increase with landscape occupancy, a coarser measure of connectivity, on a much broader (European) scale. Finally, V(t)¯¹ was found to increase along a south–north latitudinal gradient. • Our results for M. muralis are consistent with 'Darwin's wind dispersal hypothesis' that high cost of dispersal may select for lower dispersal ability in fragmented landscapes, as well as with the 'leading edge hypothesis' that most recently colonized populations harbour more dispersive phenotypes.