The variation of an asteroid’s reduced visual magnitude (i.e. adjusted to unit heliocentric and geocentric distance) with solar phase angle, or disk-integrated phase curve, provides useful information about its physical and compositional properties, and is helpful in planning future astronomical observations. Its behaviour is typically explained in terms of the optical scattering properties of a rough particulate surface present on the asteroid observed. Without a comprehensive theoretical formulation for the complex scattering by such surfaces, empirical models of the phase curve behaviour, such as the widely adopted H and G magnitude system, do not include any assumptions about the physical processes involved. Typically, any given asteroid is assumed to have a single set of H and G values that describe its phase curve. However, the observed magnitude at a given phase angle is a result of both the scattering properties of the surface material and the viewing aspect geometry of the body. The phase curves for irregularly shaped bodies are typically defined using either the maximum or rotationally-averaged mean (if the light-curve is measured) resulting in small differences in H and G, as evidenced by the well-documented amplitude-phase effect. Here we show via physical models of phase curve behaviour, and with multi-epoch SuperWASP observations of main-belt asteroids, how an asteroid’s phase curve is affected by both its scattering properties and shape. In particular, the phase curves of irregular and/or elongated asteroids can vary substantially between epochs, and even between pre- and post-opposition observations, and cannot be represented by a single set of H and G values. Only asteroids with spherical shapes show phase curves that are consistently similar between epochs. Many asteroids have default G parameters (perhaps constrained by spectral type) or values derived from limited phase angle coverage at a single apparition. Our results have implications for brightness predictions for observation planning (particularly for NEAs, which may have extreme shapes and be observed at large phase angles where shape effects are significant). They are also relevant for radiometric diameter determination (where the phase integral links the geometric and Bond albedos), and statistical studies of asteroid properties.