Housekeeping genes are often used as references when quantifying the relative abundance of transcripts of interest, because it is assumed that they are stably expressed across tissues and developmental stages. Standard housekeeping genes are targeted particularly in organisms where there is no detailed information on gene expression profiles. Here, the validity of using the two widely accepted housekeeping genes, 18S rRNA and beta-actin, as reference genes to normalize real-time RT-PCR gene expression data from the Kuruma shrimp, Marsupenaeus japonicus, was tested. Expression patterns of two target genes in a diverse sample set of embryonic, larval, post-larval and gonad mRNAs were quantified using relative and absolute real-time RT-PCR procedures. Comparison of these approaches revealed significant differences (P<0.0001) in transcript level profiles between the relative and absolute procedures for both target genes. When 18S rRNA was used as a reference, target gene expression was more similar to that of the absolute method than when beta-actin was used as a reference. Variability between the relative and absolute procedures occurred for a greater percentage of the embryonic stages compared to later developmental stages. This study indicates that the use of 18S rRNA and beta-actin for studying relative gene expression patterns in Kuruma shrimp embryonic, larval, post-larval and gonad samples will give significantly variable results, and illustrates the proposition that housekeeping genes are not necessarily appropriate references for real-time RT-PCR data normalization. Until suitable reference genes are characterized, gene expression experiments using the studied Kuruma shrimp tissues of different morphological developmental stages should use absolute quantification procedures.