31P spectra of rat tumours obtained with surface coils are shown to include skin signals of varying intensity. As reported previously by Stubbs, M., Rodrigues L. M., and Griffiths, J. R., (NMR in Biomedicine 1, 50-55, 1988) three hepatomas (rapidly growing Morris hepatoma 7777 and slow growing 9618A, and the UA hepatoma) had negligible phosphocreatine (PCr) or creatine (Cr) in acid extracts but frequently had PCr signals in surface coil spectra. Prolactinomas and mammary adenocarcinomas, which had significant PCr and Cr in extracts, showed higher PCr/NTP ratios in spectra taken in vivo than in extracts. A phantom for studying skin signals in vivo is described. A glass sphere of typical tumour size (3-4 mL) is implanted subcutaneously in the rat. Variations in skin signal with pulse duration are demonstrated with this phantom. The factors that could contribute to skin artefact in 31P tumour spectra include: (i) the relative concentrations of metabolites in skin and tumour; (ii) the skin thickness, which depends on the implantation site and rat size; (iii) skin invasion by the tumour; (iv) coil design (solenoid coils and Faraday shields are unlikely to eliminate this problem); (v) pulse repetition times; (vi) pulse duration and other NMR parameters. Careful attention to these factors could reduce skin artefacts.