The sensorimotor approach proposes that perception is constituted by the mastery of lawful sensorimotor regularities or sensorimotor contingencies (SMCs), which depend on specific bodily characteristics and on actions possibilities that the environment enables and constrains. Sensory substitution devices (SSDs) provide the user information about the world typically corresponding to one sensory modality through the stimulation of another modality. We investigate how perception emerges in novice adult participants equipped with vision-to-auditory SSDs while solving a simple geometrical shape recognition task. In particular, we examine the distinction between apparatus-related SMCs (those originating mostly in properties of the perceptual system) and object-related SMCs (those mostly connected with the perceptual task). We study the sensorimotor strategies employed by participants in three experiments with three different SSDs: a minimalist head-mounted SSD, a traditional, also head-mounted SSD (the vOICe) and an enhanced, hand-held echolocation device. Motor activity and fist-person data are registered and analyzed. Results show that participants are able to quickly learn the necessary skills to distinguish geometric shapes. Comparing the sensorimotor strategies utilized with each SSD we identify differential features of the sensorimotor patterns attributable mostly to the device, which account for the emergence of apparatus-based SMCs. These relate to differences in sweeping strategies between SSDs. We identify, also, components related to the emergence of object-related SMCs. These relate mostly to exploratory movements around the border of a shape. The study provides empirical support for SMC theory and discusses considerations about the nature of perception in sensory substitution.