To understand the logics behind the honeybee’s anti-collision system, we filmed bees (Apis Mellifera) flying through a wide outdoor flight tunnel (0.95m-wide, 3m-long). We observed that bees do not centre systematically, in contrast with previous observations made in a narrower, 0.12m wide corridor (Srinivasan et al., 1991). Bees may instead follow either wall (Fig. A,B), depending on the entrance (EL or ER) and reward (RL or RR) locations. The ‘optic flow balance’ hypothesis (Srinivasan et al., 1991) does not account for this wall following behaviour. The bee’s sideway motion is well accounted for by an optic flow based feedback loop that we called an optic flow regulator (Ruffier and Franceschini Rob. Aut. Syst. 2005; Serres et al., IEEE Biorob 2006). This scheme would require the bee to measure neither its forward speed nor its distance to the walls.