This review focusses attention on a ragged edge of our knowledge of self-motion perception, where understanding ends but there are experimental results to indicate that present approaches to analysis are inadequate. Although self-motion perception displays processes of "top-down" construction, it is typically analyzed as if it is nothing more than a deformation of the stimulus, using a "bottom-up" and input/output approach beginning with the transduction of the stimulus. Analysis often focusses on the extent to which passive transduction of the movement stimulus is accurate. Some perceptual processes that deform or transform the stimulus arise from the way known properties of sensory receptors contribute to perceptual accuracy or inaccuracy. However, further constructive processes in self-motion perception that involve discrete transformations are not well understood. We introduce constructive perception with a linguistic example which displays familiar discrete properties, then look closely at self-motion perception. Examples of self-motion perception begin with cases in which constructive processes transform particular properties of the stimulus. These transformations allow the nervous system to compose whole percepts of movement; that is, self-motion perception acts at a whole-movement level of analysis, rather than passively transducing individual cues. These whole-movement percepts may be paradoxical. In addition, a single stimulus may give rise to multiple perceptions. After reviewing self-motion perception studies, we discuss research methods for delineating principles of the constructed perception of self-motion. The habit of viewing self-motion illusions only as continuous deformations of the stimulus may be blinding the field to other perceptual phenomena, including those best characterized using the mathematics of discrete transformations or mathematical relationships relating sensory modalities in novel, sometimes discrete ways. Analysis of experiments such as these is required to mathematically formalize elements of self-motion perception, the transformations they may undergo, consistency principles, and logical structure underlying multiplicity of perceptions. Such analysis will lead to perceptual rules analogous to those recognized in visual perception.