Multivariate data analysis (MVA) has been used as an aid in the analys
is and interpretation of C-13 NMR spectra in the solid state. The goal
of this study was to investigate the effect of some important instrum
ental parameters and calculation strategies on the outcome of the mult
ivariate data analysis. The samples used were two peat forming plants,
Sphagnum fuscum and Carer rostrata, incubated in four different redox
environments. It was found that normalising each NMR spectrum to a co
nstant area should be avoided. Using non-normalised data we get a slig
htly better class separation and the peaks in the 'subspectra' are sha
rpened. Depending on the relative size of interesting variation one sh
ould be careful when choosing the number of variables, i.e. number of
data points characterising each spectrum. The line broadening techniqu
e should be used with great care in order not to obscure the informati
on. We also suggest the use of the free induction decay (FID)/MVA dire
ctly for classification purposes. This is a new approach to analyse th
e output data from NMR measurements.