Dioptric power expressed in the familiar three-component form of sphere, cylinder, and axis is unsuited to mathematical and statistical treatments; there is a particular class of power that cannot be represented in the familiar form; and it is possible that sphere, cylinder, and axis will prove inadequate in future clinical and research applications in optometry and ophthalmology. Dioptric power expressed as the four-component dioptric power matrix, however, overcomes these shortcomings. The intention in this paper is to provide a definitive statement on the nature, function, and mathematical representation of dioptric power in terms of the matrix and within the limitations of paraxial or linear optics. The approach is universal in the sense that its point of departure is not power of the familiar form (that is, of thin systems) but of systems in general (thick or thin). Familiar types of power are then seen within the context of power in general. Dioptric power is defined, for systems that may be thick and astigmatic, in terms of the ray transfer matrix. A functional definition is presented for dioptric power and its components: it defines the additive contribution of incident position to emergent direction of a ray passing through the system. For systems that are thin (or thin-equivalent) it becomes possible to describe an alternative and more familiar function; for such systems dioptric power can be regarded as the increase in reduced surface curvature of a wavefront brought about by the system as the wavefront passes through it. The curvital and torsional components of the power are explored in some detail. Dioptric power, at its most general, defines a four-dimensional inner product space called dioptric power space. The familiar types of power define a three-dimensional subspace called symmetric dioptric power space. For completeness a one-dimensional antisymmetric power space is also defined: it is orthogonal in four dimensions to symmetric dioptric power space. Various bases are defined for the spaces as are coordinate vectors with respect to them. Vectorial representations of power in the literature apply only to thin systems and are not obviously generalizable to systems in general. They are shown to be merely different coordinate representations of the same subspace, the space of symmetric powers. Some of the uses and disadvantages of the different representations are described. None of the coordinate vectors fully represent, by themselves, the essential character of dioptric power. Their use is limited to applications, such as finding a mean, where addition and scalar multiplication are involved. The full character of power is represented by the dioptric power matrix; it is in this form that power is appropriate for all mathematical relationships.