Results are presented of a comparison of the amino acid (AA) δ13C values obtained by gas chromatography–combustion–isotope ratio mass spectrometry (GC/C/IRMS) and liquid chromatography–isotope ratio mass spectrometry (LC/IRMS). Although the primary focus was the compound-specific stable carbon isotope analysis of bone collagen AAs, because of its growing application for palaeodietary and palaeoecological reconstruction, the results are relevant to any field where AA δ13C values are required. We compare LC/IRMS with the most up-to-date GC/C/IRMS method using N-acetyl methyl ester (NACME) AA derivatives. This comparison involves the analysis of standard AAs and hydrolysates of archaeological human bone collagen, which have been previously investigated as N-trifluoroacetyl isopropyl esters (TFA/IP). It was observed that, although GC/C/IRMS analyses required less sample, LC/IRMS permitted the analysis of a wider range of AAs, particularly those not amenable to GC analysis (e.g. arginine). Accordingly, reconstructed bulk δ13C values based on LC/IRMS-derived δ13C values were closer to the EA/IRMS-derived δ13C values than those based on GC/C/IRMS values. The analytical errors for LC/IRMS AA δ13C values were lower than GC/C/IRMS determinations. Inconsistencies in the δ13C values of the TFA/IP derivatives compared with the NACME- and LC/IRMS-derived δ13C values suggest inherent problems with the use of TFA/IP derivatives, resulting from: (i) inefficient sample combustion, and/or (ii) differences in the intra-molecular distribution of δ13C values between AAs, which are manifested by incomplete combustion. Close similarities between the NACME AA δ13C values and the LC/IRMS-derived δ13C values suggest that the TFA/IP derivatives should be abandoned for the natural abundance determinations of AA δ13C values. Copyright © 2011 John Wiley & Sons, Ltd.