The present study addressed whether visual information about the width of an aperture, obtained at a distance, would be sufficient to guide walking through the aperture without collision. For this purpose, we asked twelve young participants to walk while holding a 66-cm horizontal bar (bar length needs to be considered in order to perceive space necessary for crossing) and pass through an aperture without vision from 3 m in front of the aperture. Participants performed the tasks under each of four visual conditions, which differed in how vision was available: observation for 1.5 s while standing (static vision), observation during two forward steps and stopping (dynamic vision), observation during two forward steps and not stopping (dynamic vision with nonstop walking), and full vision. The results showed that, for narrow apertures (the widths were 0.8 and 1.0 times the bar length), the rate of collision without vision was about 40-50 %. This was mainly due to the maladaptive planning of body rotation. For the aperture 1.0 times the bar length, the percentage of trials with no body rotation was high, suggesting that at least some participants underestimated the space necessary for crossing. The location at which maximum body rotation occurred became farther from the obstacle, which may have been related to decreased movement speed. The availability of dynamic visual sampling during two forward steps did not contribute to improving collision avoidance. These results suggest that, while fundamental locomotor patterns are maintained even without online vision, both the underestimation of space required for crossing and the lack of fine-tuning of behavior prior to crossing increased collision rates.