Post-mortem analysis - the process of tracing software failure to source code - is an important means for maintenance engineers and regulatory reviewers for establishing the cause of an error. Historically, static slicing techniques have been used for aiding post-mortem fault analysis. However, the slices obtained in this manner can often be too large and may not give a clear understanding of the code when dealing with complex reactive systems. We propose using model abstraction in conjunction with slicing to ameliorate the problem of understanding large slices. Combining slicing with abstraction provides the analyst with an integrated cognition model, leading to a better understanding of the code, and consequently more efficient error analysis. We formalize this concept through the notion of abstraction-driven slicing, and use it to develop CAdS, an automated tool to aid postmortem error detection in C programs using abstraction-driven static slicing. We list our experiences with CAdS and illustrate how it can be used to reduce effort involved in the postmortem analysis process.