In an effort to minimize uncertainties seen in bench scale tests, the sources of variability in fire test data are investigated. An earlier study on poly(aryl ether ether ketone) (PEEK) has showed that the fire performance parameters of this thermoplastic change noticeably when exposed to moisture prior to test. The present research is a follow-up study where a series of cone calorimetry tests are conducted on conditioned PEEK specimens to analyze the previously reported ignition time scatter and its relation to surface bubble formation. Pyrolysis modeling is, subsequently, carried out, to relate the moisture content with the model parameters and to explain the possible physical mechanisms that cause the difference in ignition times between wet and dry samples. (C) 2011 The Combustion Institute. Published by Elsevier Inc. All rights reserved.