The TCP domain is a plant-specific DNA binding domain found in proteins from a diverse array of species, including the cycloidea (cyc) and teosinte branched1 (tb1) gene products and the PCF1 and PCF2 proteins. To understand the role in transcriptional regulation of proteins with this domain, we have analysed the DNA binding and dimerization specificity of the TCP protein family using rice PCF proteins, and further evaluated potential targets for the TCP protein. The seven PCF members including five newly isolated proteins, were able to be grouped into two classes, I and II, based on sequence similarity in the TCP domain. Random binding site selection experiments and electrophoretic mobility shift assays (EMSAs) revealed the consensus DNA binding sequences of these two classes to be distinct but overlapping; GGNCCCAC for class I and GTGGNCCC for class II. The TB1 protein from maize, which belongs to class II, had the same specificity as the rice class II proteins, suggesting the conservation of binding specificity between TCP domains from different species. The yeast 2-hybrid assay and EMSA revealed that these proteins tend to form a homodimer or a heterodimer between members of the same class. We searched predicted 5' flanking sequences of Arabidopsis genes for the consensus binding sequences and found that the consensus sites are distributed in the genome at a considerably lower frequency. We further analysed eight promoters containing the class I consensus TCP sites. The transcriptional activities of six promoters were decreased by a mutation of the TCP binding site, which is consistent with the observation that the class I TCP site can confer transactivation function on a heterologous promoter. These results suggest that the two classes of TCP protein are distinct in DNA binding specificity and transcriptional regulation.