Analysis of transcript representation on gene microarrays requires microgram amounts of total RNA or DNA. Without amplification, such amounts are obtainable only from millions of cells. However, it may be desirable to determine transcript representation in few or even single cells in aspiration biopsies, rare population subsets isolated by cell sorting or laser capture, or micromanipulated single cells. Nucleic-acid amplification methods could be used in these cases, but it is difficult to amplify different transcripts in a sample without distorting quantitative relationships between them. Linear isothermal RNA amplification has been used to amplify as little as 10 ng of total cellular RNA, corresponding to the amount obtainable from thousands of cells, while still preserving the original abundance relationships. However, the available procedures require multiple steps, are labor intensive and time consuming, and have not been shown to preserve abundance information from smaller starting amounts. Exponential amplification, on the other hand, is a relatively simple technology, but is generally considered to bias abundance relationships unacceptably. These constraints have placed beyond current reach the secure and routine application of microarray analysis to single or small numbers of cells. Here we describe results obtained with a rapid and highly optimized global reverse transcription#150;PCR (RT-PCR) procedure. Contrary to prevalent expectations, the exponential approach preserves abundance relationships through amplification as high as 3 x 10(11)-fold. Further, it reduces by a million-fold the input amount of RNA needed for microarray analysis, and yields reproducible results from the picogram range of total RNA obtainable from single cells.