Volcanic tremor has attracted considerable attention by seismologists because of its potential value as a tool for forecasting eruptions and better understanding the physical processes that occur inside active volcanoes. However, unlike tectonic earthquakes where the dominant source process is brittle failure of rock, the driving mechanism of tremor seems to involve complex interactions of magmatic fluids with the surrounding bedrock. These interactions are responsible for the following distinct characteristics found in volcanic tremor recorded at many volcanoes worldwide: (a) the onset of tremor may be emergent or impulsive, with its amplitude showing in many cases a direct relationship to the volcanic activity; (b) in the frequency domain the spectra consist of a series of sharp peaks in the band 0.1^7 Hz, representing either a fundamental frequency and its harmonics, or a random distribution, while quite often they exhibit temporal variations in their content; (c) the depth of the source can vary considerably from one volcano to another in the range of a few hundred metres to 40 km; (d) tremor may occur prior to and/or after eruptions with a duration that ranges from several minutes to several days or months. The methods used to study tremor include spectral analysis using both the Fast Fourier Transform and the Maximum Entropy Method, polarisation analysis of the wavefield and methods that make use of array data to deduce the backazimuth and type of the seismic waves as well as the location of the source. Visual and/or recorded acoustic observations of the ongoing volcanic activity have assisted in many cases to further constrain proposed physical mechanisms for the generation of tremor. The models suggested as possible sources of tremor can be grouped as follows: (a) fluid-flow-induced oscillations of conduits transporting magmatic fluids; (b) excitation and resonance of fluid-filled cracks; (c) bubble growth or collapse due to hydrothermal boiling of groundwater; (d) a variety of models involving the oscillations of magma bodies with different geometries. It has been proposed by many authors that the source of tremor is not unique and may differ from one volcano to another, a fact that adds more difficulty in the source modelling efforts. As data quality, computer power and speed are improving, it may be possible in the near future to decipher and accurately model tremor source processes at different volcanic environments. ß 2002 Elsevier Science B.V. All rights reserved.