A computational approach is presented for prediction and interpretation of core-level spectra of complex molecules. Applications are presented for several isolated organic molecules, sampling a range of chemical bonding and structural motifs. Comparison with gas phase measurements indicates that spectral lineshapes are accurately reproduced both above and below the ionization potential, without resort to ad hoc broadening. Agreement with experiment is significantly improved upon inclusion of vibrations via molecular dynamics sampling. We isolate and characterize spectral features due to particular electronic transitions enabled by vibrations, noting that even zero-point motion is sufficient in some cases. (C) 2008 Elsevier B.V. All rights reserved.