This paper describes experiments on evacuation using escalators with 50 subjects and intends to clarify aspects of group evacuation by escalator and quantify the basic properties for the evacuation planning with escalators. In total 25 experiments on different conditions were conducted using 50m and 22m long escalators in a large conventional facility in downtown Tokyo; walking on a still escalator or a moving escalator, walking by a solo or in a group, various configurations of subjects. Basic evacuation behavior properties such as upward walking velocity and flow coefficient at the escalator entrance are obtained for still and upward running escalators, and the following walking behaviors are clarified; (1) the walking velocity is constant even throughout a 50m long escalator lane, (2) the walking velocity of a solo pedestrian on an escalator running at 0.5m/s is reduced by 0.03~0.10m/s from that on a still escalator, (3) the walking velocity of individual in a group on an escalator is strongly affected by the existence of slowly walking pedestrians, (4) the walking velocity of a solo pedestrian on a still escalator is virtually identical with that on a stairway, and (5) the flow coefficient seems to be dictated by the moving velocity of pedestrians on the escalator lane. The test results were applied to evacuation simulation for an actual very deep subway station, if the escalators can be utilized to the maximum, evacuation time could be shortened greatly; in this report, evacuees finish evacuation in 54% time of normal case; using only stairway. These studies suggest notable potential benefit of the emergency use of escalators moving toward exit, while the current transportation authorities only anticipate stopped escalators used at the event of a disaster.