Biclustering techniques aim at extracting significant subsets of genes and conditions from microarray gene expression data. This kind of algorithms is mainly based on two key aspects: the way in which they deal with gene similarity across the experimental conditions, that determines the quality of biclusters; and the heuristic or search strategy used for exploring the search space. A measure that is often adopted for establishing the quality of biclusters is the mean squared residue. This measure has been successfully used in many approaches. However, It has been recently proven that the mean squared residue fails to recognize some kind of biclusters as quality biclusters, mainly due to the difficulty of detecting scaling patterns In data. In this work, we propose a novel measure for trying to overcome this drawback. This measure is based on the area between two curves. Such curves are built from the maximum and minimum standardized expression values exhibited for each experimental condition. In order to test the proposed measure, we have incorporated it into a multiobjective evolutionary algorithm. Experimental results confirm the effectiveness of our approach. The combination of the measure we propose with the mean squared residue yields results that would not have been obtained if only the mean squared residue had been used. © 2007 IEEE.