We set out to clone Bax-specific CD8+ T cells from peripheral blood samples of patients with primary chronic lymphocytic leukaemia. A number of clones were generated using a Bax peptide pool and their T-cell epitope was mapped to two peptides sharing a common 9-amino-acid sequence (LLSYFGTPT), restricted by HLA-A*0201. However, when these T-cell clones were tested against highly purified syntheses (> 95%) of the same peptide sequence, there was no functional response. Subsequent mass spectrometric analysis and HPLC fractionation suggested that the active component in the original crude peptide preparations (77% pure) was a peptide with a tert-butyl (tBu) modification of the tyrosine residue. This was confirmed by modification of the inactive wild-type sequence to generate functionally active peptides. Computer modelling of peptide:HLA-A*0201 structures predicted that the tBu modification would not affect interactions between peptide residues and the HLA binding site. However, these models did predict that the tBu modification of tyrosine would result in an extension of the side chain out of the peptide-binding groove up towards the T-cell receptor. This modified product formed < 1% of the original P603 crude peptide preparation and < 0·05% of the original 23-peptide mixture used for T-cell stimulation. The data presented here, illustrate the potential for chemical modifications to change the immunogenicity of synthetic peptides, and highlight the exquisite capacity of T-cell receptors to discriminate between structurally similar peptide sequences. Furthermore, this study highlights potential pitfalls associated with the use of synthetic peptides for the monitoring and modulating of human immune responses.