No model system is perfect, and much has been written about the relevance — or lack thereof — of cancer models to human tumours, but cancer would still be a black box without them. Two articles in this issue trace the development of model systems.

We often describe cancer cells as having unlimited replicative potential, so we might expect them to be amenable to permanent culture. But over 40 years of failed experiments stand between the development of early animal-cell cultures and HeLa, the first continuous human cell line. On page 315, John R. Masters takes us back to the tissue-culture lab of the 1950s and follows 50 years of progress in cancer-cell culture.

'Cancer in a test tube' was a great breakthrough, but a homogeneous culture does not approximate to the complex array of cell types that make up a tumour. As described by Jos Jonkers and Anton Berns on page 251, inbred strains of mice with increased susceptibility to cancer — developed almost 100 years ago — were largely superseded by transgenic and knockout models. But in these genetically modified mice, tumours develop in a background of mutant cells, which is very different from a spontaneous tumour that arises from a single cell in a wild-type background. Jonkers and Berns talk us through the genetic tools that are now available to breed mice in which tumorigenesis more closely mimics spontaneous tumour formation in humans.

Provided that we appreciate their limitations, both types of model can provide deep insights into the nature of human tumours. Not only can we dissect the molecular architecture of the cancer cell to discover new therapeutic targets, but we can also put new therapeutics through rigorous preclinical testing before taking them into clinical trials.