Optical nephelometry is the measurement of the angular distribution of light scattered from a particle suspension. Experimental nephelometers confirm the predictions of optical models and their readings are inverted to determine properties of unknown suspensions. Single scattering models, which assume a single particle interaction prior to detection, are used to model tenuous suspensions in the nephelometer. Multiple scattering models can be used to obtain higher-order solutions, but lack generality. Any given method addresses some subset of possible problems, e.g. tenuous or dense suspensions, small or large particles. This thesis explores the feasibility of using empirical models to extrapolate the single scattering approach in a non-linear manner, improving the generality of a multiple-scattering description. Initially, single scattering (Mie) theory for spherical particles is presented and extended to polydispersions of particles and to spectral scattering. The principle of integrating the single scattering result over a finite scattering volume is examined as a precursor to modelling the actual nephelometer. A low-cost, PC controlled scanning nephelometer is developed with a 0.9° resolution and ±150° range and a small (-25ml) volume sample cell. The photodiode detector has a numerical aperture of 0.079, providing, for most angles, a scattering volume with length 10mm and cross-section determined by the HeNe laser source ('-1mrn 2). The optics of the air/glass/water interfaces and of single and first-order multiple scattering over the scattering volume are modelled. These models are found to predict the scattering footprints observed in tenuous suspensions of spherical latex particles. Experimental data are obtained from tenuous to relatively dense (5% by volume) suspensions of latex spheres over a size range of 54nm to 14tm. These data are compared with single and first-order multiple scattering and their form and dependencies are considered. They are used to train an empirical neural (multi-layer perceptron) model of the multiple scattering based on particle characteristics and on the scattering footprint of the individual particles. This non-linear extrapolation of the single scattering model is applied to the nephelometer, improving the generality over a purely theoretical multiple scattering approach. The trained neural model is used, initially, to investigate some of the empirical characteristics of the multiple scattering process.