The objectives of this thesis were (i) to investigate the main sources and paths of noise on modern utility size wind turbines; (ii) to explore methods of reducing the noise; (iii) to assess our current ability to accurately predict and measure wind turbine noise. The accomplishment of these objectives would enable quieter wind turbines to be developed and allow them to be located near residential dwellings with greater confidence that the noise would not be a nuisance. A comprehensive review of the current literature was carried out and the findings were used as a basis for the investigative work conducted. It was found that wind turbine noise could be classed as either aerodynamically produced noise or mechanically produced noise. Aerodynamically produced noise on wind turbines arises mainly from the interaction of the flow over the blade with the surrounding air. Mechanically produced noise arises from a number of sources such as the gearbox, generator and hydraulic pumps. The noise can be radiated directly from the noisy component (airborne) and / or transferred through the structure of the turbine and radiated elsewhere (structure-borne) such as the tower. The prototype Windflow 500 wind turbine near to Christchurch was used for the majority of the investigative work carried out, and to assess the predictions made. The main radiators of noise from the turbine were identified as the blades (86 – 90% of the total sound power), the tower (initially 8 – 12% but later reduced to ~4% of the total sound power), and the nacelle cladding (1% of the total sound power). A prominent tone in the sound power spectrum from the turbine was observed in the 315 Hz 1/3 octave band. This was shown to be predominantly caused by gear meshing in the second stage of the gearbox at 311 Hz. The presence of the tone was significant because under commonly used standards a tonal penalty would be applied to the measured sound pressure level from the turbine to account for the extra annoyance caused by the tone. This in turn would mean that any potential wind farms would need to be sited further from residential dwellings than would otherwise be necessary in order to comply with noise regulations. Investigations were carried out that addressed the noise radiated from each of the main contributors outlined above. The sound power level radiated from the tower was found to be effectively reduced by attaching rubber tiles at strategic locations inside the tower. Noise radiated from the nacelle was reduced with a combination of acoustic insulation and acoustic absorption inside the nacelle. An investigation into the gearbox noise was also carried out. Attempts to reduce the tonal noise caused by gear meshing were made with little success but the investigation provided a good basis upon which to conduct further work. Preliminary investigations into both structure-borne and aerodynamically generated blade noise were carried out. The structure-borne blade noise investigation showed that the blades readily vibrated at a range of frequencies, the result being that structurally transmitted noise radiated from the blades was likely to be present at high levels. Research showed that the structure-borne noise radiated from the blades could be significantly reduced by partially filling the internal cavity of the blades with foam. The investigation of aerodynamically produced noise was carried out on a section of Windflow 500 blade in the low noise wind tunnel at the University of Canterbury. The tests showed that the blade generated noise at a range of frequencies including those in the 315 Hz 1/3 octave band. This suggested that the tonal noise measured from the blades was not only due to structurally transmitted noise from the gearbox but was also contributed to by aerodynamically produced noise. It was found that the noise from the blade section could be reduced by up to 4.5 dB at certain frequencies by attaching serrated strips to the trailing edge of the aerofoil. Empirical equations for prediction of wind turbine sound power levels were evaluated and found to be in good agreement with measured data. It was found that accurate spectral predictions of the sound power level were much more difficult. However given spectral data for a turbine, it was found that accurate predictions of the noise propagation from the turbine could be made, taking into account meteorological effects and the effect of complex topography. It was found that the CONCAWE propagation model was well suited to the prediction of noise propagation from wind turbines because of its superior handling of meteorological effects. In an investigation carried out which modelled the Gebbies Pass site of the Windflow 500 it was found that the CONCAWE model could predict sound pressure levels from the turbine to within 2 dB at distances of up to 1400 m. Further work in the area of wind turbine noise should be focused on the reduction of blade noise. This is especially relevant to the Windflow 500 since blade noise was found to be by far the largest contributor to total noise radiated from the turbine. Acoustic treatments elsewhere would therefore produce only small reductions in the total sound power emitted by the turbine.