Previous attempts to measure localization bias around a right-angle turn CL-trajectory) have found either no spatial bias off the trajectory (Whitney, Cavanagh, & Murakami, 2000) or a bias, in different experiments, both 'inside' and 'outside' the trajectory (Nieman, Sheth, & Shimojo, 2010). However, Eagleman and Sejnowski (2007) presented data showing that the perceived location of a brief feature on two moving stimuli could be predicted from the vector sum of their directions after the feature appeared. Such a vector sum with an L-trajectory could predict that the perceived position before the turn should be biased 'sideways' off the trajectory, in the direction of the final motion. With stimuli that particularly facilitated accurate vernier judgments, and measuring bias via the flash-lag illusion, this is indeed what we observed. Our data thus favour Eagleman and Sejnowski's (2007) supposition. Further, the bias occurred before the change in direction, rather than after it, supporting the contention that it is motion after a point being sampled that affects its perception (Bachmann et al., 2003; Eagleman & Sejnowski, 2007; Krekelberg & Lappe, 2000; Nieman, Sheth, & Shimojo, 2010). (C) 2013 Elsevier Ltd. All rights reserved.