Most (epi)mutations in cancers are specific to particular tumours or occur at specific stages of development, cell differentiation or tumorigenesis. Simple molecular mechanisms, such as tissue-restricted gene expression, seem to explain these associations only in rare cases. Instead, the specificity of (epi)mutations is probably due to the selection of a restricted spectrum of genetic changes by the cellular environment. In some cases, the resulting functional defects might be constrained to be neither too strong nor too weak for tumour growth to occur; that is, they lie within a 'window' that is permissive for tumorigenesis.