The reliable detection of concealed substances in sealed opaque plastic and coloured glass containers, with low falsealarm rate, is a problem in numerous areas of security. For example, in aviation security, there is no reliable methodology for alarm resolution of substances with high chemical specificity unless the substances are in optically transparent containers. We present a recently developed method called Spatially Offset Raman Spectroscopy (SORS) which enables the discrimination of the Raman spectrum of the content substance from the Raman spectrum of the container material with no prior knowledge of either, thereby allowing unambiguous identification of the container contents. The method is effective with coloured plastic containers that are several millimetres thick and which are not see-through to the eye and also for coloured glass bottles. Such cases do not typically yield to conventional backscatter Raman spectroscopy (or have poor false-alarm rates) since the content signal is often overwhelmed by the signal from the container, which may in addition have a strong interfering fluorescence background. SORS measurement can be performed in a few seconds by shining a laser light onto the container and then measuring the Raman signal at the excitation point but also at one or more offset positions. Each measurement has different relative orthogonal contributions from the container and contents Raman spectra, so that, with no prior knowledge, the pure spectra of both the container and contents can be extracted - either by scaled subtraction or via multivariate statistical methods. The content spectrum can then be compared to a reference library of pure materials to give a threat evaluation with a confidence level not compromised by interfering signals originating from the container wall. In this paper, we describe the methods and their optimization, and characterize their performance in practical screening applications. The study shows that there is frequently a well-defined optimum spatial offset that maximizes the signal to noise ratio (SNR) of the resultant SORS spectrum and that this optimum can vary greatly depending on content and container material. It is also shown for the first time that, for a fixed total acquisition time available, a very high fraction of this time should be spent acquiring the offset spectrum. For common samples, the best results were obtained where the offset measurement was acquired for 20x longer than the zero offset position.© (2010) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.