Determining how people perceive distance is a central issue in the study of visual perception. Whilst near-distance perception has been extensively researched, far-distance perception has received little attention. We review the literature and illustrate key points with new data, focusing on analysis of available information/cues, degree of accuracy, measurement procedures, perceptual/cognitive influences, and geometry of perceived space. Analysis of available cues suggests that under natural viewing there is sufficient information for veridical distance perception. However, distance judgements are inaccurate and vary with mode of measurement (eg absolute vs relative judgements). Inaccuracy includes compression of distance. This has been modelled with the use of different geometries—most commonly power functions. Exponents often average 1.0 but this conceals considerable individual variation (in our data individual exponents ranged from 0.5 to over 1.0). Further, even for averaged exponents values vary between 0.8 and 1.25, as a function of viewing conditions, experimental method, and of the relative contribution of cognitive and perceptual factors. Evidence suggests that distance is encoded at an ordinal level, but for many practical tasks the final judgement must be metric (eg range finding) and this transformation is error-prone (in our data numerical estimates could be a factor of 10 out). Further, many natural judgements require perception of the full layout of the scene (including exocentric distances) rather than the more commonly investigated perception of egocentric distance. Evidence suggests that training based on practice with feedback produces some improvement in accuracy, but this is highly context-specific.