The performance of alternative normalization strategies. The performances of different strategies for normalizing a 50 ng Brain array-section for comparison to a 250 ng UHRR array-section are illustrated. For the analysis-group bead-types, with a 250 ng Brain array-section as a reference, we determine the squared differences between our prediction of log2-intensity and the reference and illustrate the cumulative distribution of those differences. The first prediction uses a basic quantile-style normalization where the 50 ng Brain array-section is transformed to take the distribution of intensities seen on the 250 ng UHRR array-section. A more complicated prediction making use of a 50 ng UHRR section is also illustrated. For reference we give the agreement between replicate 250 ng Brain array-sections, representing the gold-standard that could be achieved by any method.