The phenolic profile data (phenolic concentrations in seedless dry matter) were analyzed with principal component analysis (PCA), which is a technique for modelling a set of data onto itself (Mardia, Kent & Bibby, 1979). PCA is typically used for measurements which are ill-conditioned and for highly correlated data; as a consequence, their covariance matrix is nearly singular. With PCA, the data arranged in tables can be reduced to a set of new latent variables called principal components (PCs). The loadings of the PC define the direction of greatest variability and the score values represent the projection of each object onto PC. The first PC is the linear combination of the original variables which explains the greatest variability. The second PC has been defined to be orthogonal to the first one and explains the second greatest amount of variability. The analysis proceeds until all PCs are obtained, the number of which is typically much smaller than the number of variables. Since PCA usually deals with relative differences between objects, the average of each variable is first subtracted from each column in the data matrix. This centring brings the origin of the coordinate system to the centre of the data set. To give each variable (column in the table) equal weight in the analysis, it is customary to rescale the data by dividing each column in the data matrix by its standard deviation. This makes the variation same for all variables. For principal component analysis calculations, computer program Statistic for Windows, version 5.0 (StatSoft Inc., Tulsa, OK) was used.