Protein production involves a series of stochastic chemical steps. One consequence of this fact is that the copy number of any given protein varies substantially from cell to cell, even within isogenic populations. Recent experiments have measured this variation for thousands of different proteins, revealing a linear relationship between variance and mean level of expression for much of the proteome. This simple relationship is frequently thought to arise from the random production and degradation of mRNAs, but several lines of evidence suggest that infrequent gene activation events also bear responsibility. In support of the latter hypothesis, single-molecule experiments have demonstrated that mRNA transcripts are often produced in large bursts. Moreover, the temporal pattern of these bursts appears to be correlated for chromosomally proximal genes, suggesting the existence of an upstream player.