An image forming unit having image bearing body, on which a latent image is formed, and an intermediate transfer belt, onto which a toner image obtained by developing the latent image formed on the image bearing body is transferred. A primary transfer member transfers the toner image onto the intermediate transfer belt, and a plurality of rolls, across which the intermediate transfer belt is stretched, apply tension to the intermediate transfer belt. A secondary transfer backup roll, which is one of the plurality of rolls, backs up the secondary transfer by a secondary transfer roll. A pair of side frames pivotally support the primary transfer member and the plurality of rolls. A positioner, with which a contact member rotatably fitted at both ends of a shaft of the secondary transfer roll is brought into contact, is disposed at each side frame in the vicinity of the secondary transfer backup roll.