An image forming apparatus including an intermediate transfer device capable of making a change in a degree of expansion and contraction of an intermediate transfer belt slow is provided. A position and a moving range of a cleaning roll are set in order to make a rear surface contact position on a rear surface of the intermediate transfer belt relative to the contact position where the cleaning roll comes in contact with the intermediate transfer belt, away from a cleaning backup roll while an intermediate transfer belt cleaner rocks and moves to a contact position of the cleaner. Because of the intermediate transfer belt being not sandwiched between the cleaning and backup rolls, the change in the degree of expansion and contraction of the intermediate transfer belt is made slow, and the intermediate transfer belt is kept to have the degree of expansion and contraction within a predetermined range.