CD8 T-cell infiltration of metastatic melanoma may be a useful biomarker for prediction of prognosis and response to therapy. The heterogeneous distribution of CD8 T cells within a single tumor, and across different tumors within a single patient, may complicate quantification of infiltration. However, the impact of heterogeneity has not been quantified sufficiently. To address this, we have assessed intratumoral heterogeneity of CD8 T-cell counts, as well as intertumoral heterogeneity for synchronous and metachronous metastases. In a tissue microarray containing 189 melanoma metastases from 147 patients, the density of CD8 T cells per sample was determined by immunohistochemistry. The mean density and coefficient of variation were calculated for each tumor and the rates of discordant values were determined. CD8 counts varied widely among different core samples of the same tumors (average coefficient of variation=0.77, 95% confidence interval: 0.70-0.85), with discordance occurring in 40% of tumors. CD8 densities were similar among pairs of simultaneous tumors; however, significant changes in CD8 densities were observed among 35 pairs of metachronous tumors. CD8 T-cell density is not well represented by a single 1 mm diameter sample. Differences in CD8 T-cell counts, observed in clinical trials, from pretreatment to post-treatment specimens may be explained by the spatial and temporal heterogeneity of CD8 distribution, especially if the assessed samples are small (i.e. 1 mm). A sufficiently large biopsy of one of several synchronous tumors may be representative of CD8 T-cell infiltration of a patient's disease.