Estimating the three-dimensional (3-D) structure of the environment is a principal function of the visual system, facilitating key tasks such as segmentation, object recognition, material perception and the control of movement. Despite the central importance of 3-D estimation, we still know relatively little about the functional roles of different cortical areas in processing depth. In particular, how do responses to depth signals in different cortical areas relate to our perception of 3-D shape? This presentation will review recent human brain imaging work that has sought to test the functional roles of different portions of the dorsal and ventral visual streams in supporting 3-D vision. The general strategy for this fMRI work is to (a) use parametric stimulus manipulations that can be cross-referenced to psychophysical- and electrophysiological- studies; (b) use high-resolution data acquisition to obtain fMRI signals that are closer to the neural representation; (c) examine information related to depth at the level of individual voxels and patterns of voxels using advanced statistical analysis methods. I will start by describing work that addresses the representation of depth from binocular disparity. In particular, I will describe studies that have used fMRI decoding approaches to test for cortical representations of binocular disparity that are compatible with the observer’s perception of depth from disparity. This work makes use of a low-level stimulus manipulation (reversing the contrast polarity of stimuli viewed by the two eyes) to study stimuli in which depth structure is-, or is not-, clearly visible. This technique reveals disparity representations that are compatible with perceptual judgments in higher portions of the ventral stream, and intermediate and higher portions of the dorsal stream. In contrast, responses in early visual areas and intermediate ventral areas contain information that is dissociated from perception. Further, using parametric manipulations, the work demonstrates important differences between the disparity representations in higher portions of the dorsal and ventral streams. Specifically, higher dorsal areas contain information that relates to the particular disparity that is viewed by the participant; in contrast, higher ventral areas show responses that relate to disparity sign (near vs. far) rather than disparity magnitude. This suggests a difference between dorsal and ventral representations related to metric vs. configural depth. The second portion of the talk will address the question of how information from different depth cues (disparity and relative motion) is synthesised to support 3-D perception. Behavioural tests show humans combine depth cues near-optimally, a feat that could depend on: (i) discriminating the outputs from cue-specific mechanisms, or (ii) fusing signals into a common representation. I will describe fMRI studies that combine established psychophysical methods, high-resolution imaging and advanced analysis methods to address the processing of individual depth cues and their combination. I will describe approaches needed to identify cue fusion, and highlight different representational codes in dorsal vs. ventral cortical areas. The work suggests that area V3B/KO in the dorsal stream plays an important role in representing depth from different cues. Activity in this area may underlie improved performance when different cues are combined to represent 3-D depth. The findings from both portions of the talk suggest perceptually-relevant 3-D representations in both dorsal and ventral visual areas. The empirical approach outlined in the talk provides a means of dissociating these representations to reveal the network of functional computations that support 3-D perception.