We describe the 2dF Galaxy Redshift Survey (2dFGRS), and the current status of the observations. In this exploratory paper, we apply a Principal Component Analysis to a preliminary sample of 5869 galaxy spectra and use the two most significant components to split the sample into five spectral classes. These classes are defined by considering visual classifications of a subset of the 2dF spectra, and also by comparing to high quality spectra of local galaxies. We calculate a luminosity function for each of the different classes and find that later-type galaxies have a fainter characteristic magnitude, and a steeper faint-end slope. For the whole sample we find M*=-19.7 (for Omega=1, H_0=100 km/sec/Mpc), alpha=-1.3, phi*=0.017. For class 1 (`early-type') we find M*=-19.6, alpha=-0.7, while for class 5 (`late-type') we find M*=-19.0, alpha=-1.7. The derived 2dF luminosity functions agree well with other recent luminosity function estimates.

We analyse synthetic galaxy spectra from the evolutionary models of Bruzual&Charlot and Fioc&Rocca-Volmerange using the method of Principal Component Analysis (PCA). We explore synthetic spectra with different ages, star formation histories and metalicities, and identify the Principal Components (PCs) of variance in the spectra due to these different model parameters. The PCA provides a more objective and informative alternative to diagnostics by individual spectral lines. We discuss how the PCs can be used to estimate the input model parameters and explore the impact of noise in this inverse problem. We also discuss how changing the sampling of the ages and other model parameters affects the resulting PCs. Our first two synthetic PCs agree with a similar analysis on observed spectra obtained by Kennicutt and the 2dF redshift survey. We conclude that with a good enough signal-to-noise (S/N>> 10) it is possible to derive age, star formation history and metallicity from observed galaxy spectra using PCA.