Molecular analysis of isolated single cells is a powerful tool for studying heterogeneity within a population of cells and for clarifying issues of cell origin and clonality. Here, we investigate the applicability of molecular techniques at a single-cell level by using routinely processed archival tissue. An ultraviolet laser in conjunction with a computer-controlled micromanipulator and a microscope were used for the contamination-free isolation of single tumor cells from stained sections of diffuse-type gastric cancer. A total of 1,328 single cells and 654 clusters of 10-30 cells each, taken from specimens of 14 patients, were analyzed for parts of the E-cadherin gene by the polymerase chain reaction (PCR). With increasing length in base pairs (bp) of the amplified fragments, the efficiency of single-cell PCR as measured by the rate of detectable amplification products declined from approximately 25% (156, 213, and 228 bp) to 14% (246 bp) and 11% (264 and 296 bp). For groups of 10-30 cells, a similar effect was seen at a higher level at 33% (246 bp), 31% (264 bp), and 26% (296 bp), respectively. To our knowledge, this is the first report that has studied the outcome of single-cell PCR on a large systematic scale. The average degree of DNA disintegration in paraffin-embedded, stained tissues was estimated to be approximately 100 bp when the aforementioned data were used in a mathematical model. This study provides evidence that in order to obtain reasonable sensitivity with single-cell PCR, short fragments, preferably < 200 bp long, should be used. Furthermore, whenever applicable, pooling of cells of interest may be another favorable option.