This paper presents the findings of part of ACARP Project C9003 (Green & Ward, 2002) which examined the possibility of deriving quantitative guidelines for the use of sonic velocity logs in the identification of potential goaf delamination planes, with a view to improving predictive capability for the delineation of heavy roof conditions. The downhole sonic velocity log is widely used for the interpretation of overburden strata into geomechanical units and for identifying thick or strong sandstone layers in the main overburden. It can also depict discrete weaker horizons that can act as goaf delamination planes within such layers as high transit time (low velocity) spikes. However, delineation of these planes becomes subjective if the contrast between the peak velocity and the background velocity diminishes. In terms of predicting goafing behaviour the question then arises as to whether the potential for bed separation can be predicted on the basis of the sonic velocity contrast alone. There were no quantitative guidelines for using the sonic velocity log for this purpose. The research carried out under C9003 was thus directed at establishing the value of the sonic log for the delineation of potential goaf delamination planes. The proposal was to systematically test sonic log responses for potential separation planes against monitored goafing behaviour from a number of mine sites, with the objective of deriving quantitative guidelines for the identification of goaf delamination planes from the sonic velocity log.