Microscopy-based high-throughput screens can provide a broad view of biological responses and states at the resolution of single cells. CellProfiler, our open-source software, extracts hundreds of numerical descriptors (morphological features) for every cell in every image. Some phenotypes are readily identifiable; for instance, mitotic arrest can be detected by measuring the intensity of a fluorescent marker for mitosis. Other phenotypes, while apparent upon visual inspection, are much harder to identify computationally. As an example, when signaling pathways related to cell migration are stimulated, T47D breast-cancer cells take on a motile appearance, but this is reflected in a complex pattern of small changes to many different measurements. Quantitative analysis of such complex phenotypes is a focus of our group. Classifiers trained from hand-curated sets of individual cells that exhibit the phenotypes of interest have been successful for a variety of complex phenotypes and assays. Our open-source data-analysis software CellProfiler Analyst supports the training process and allows a researcher to train for most phenotypes in just a few hours. Using this approach, we have successfully scored images in RNA interference and chemical screens for more than two dozen complex phenotypes, many of which were previously intractable because they were complex and lowpenetrance. These screens are uncovering the genetic basis of, and chemical regulators for, various biological processes and diseases including a variety of cancers, leukemias, and infectious diseases. In this talk, I will give an overview of the current capabilities of high-throughput microscopy experiments to quantify subtle cellular phenotypes. I will focus on fluorescent images of cultured cells, but if time allows I will mention whole-organism screens in C. elegans as well as ongoing work to profile and characterize compounds based on their effect on cells.