Three different precipitation regionalizations based on 80 stations in Catalonia (north-eastern Spain), with a recording period over the last 50 years, are produced. The first classification is achieved by using monthly average precipitation and some concepts of mathematical information theory applied to precipitation frequency histograms. The clustering procedure is based on an ‘average linkage’ algorithm. The results are discussed in terms of orographic features and rainfall regime and compared with the other two classifications, which are briefly described. One of them is obtained from only monthly average values and the other is based directly on frequency histograms of precipitation. Several reasons lead us to accept the first one, based on monthly average data and information theory, as the best method. Another important aspect of this paper is the definition of the optimal interval amplitude when dealing with precipitation frequency histograms.