The time required in judging if two probes are on the same curve increased monotonically with the separation of the probes along that curve. This process is called “curve tracing” (Jolicoeur, Ullman & Mackay, 1986). In this study we examined whether curve tracing would occur on a three-dimensionally presented curve with depth variation. By comparing the performance on depth varying and no depth varying stimuli, we examined the properties of three-dimensional curve tracing. The mean RT on three-dimensional stimuli increased monotonically as the distance between two probes increased, which indicates that curve tracing also occurs on three-dimensional stimuli. The mean RT on three-dimensional stimuli was longer than that on two-dimensional stimuli. Our results suggest that three-dimensional structure of the stimuli caused additional costs on curve tracing.