For screening to be successful in lung cancer, programmes need to identify a precursor of fatal disease that is highly treatable. This precursor is currently presumed to be early-stage lung cancer, and thus our current efforts in lung-cancer screening have focused on finding, and treating, early-stage lung cancer. In other words, our current natural-history model of lung cancer is based on the notion that for a period before a lung cancer becomes advanced and fatal, it is localised and treatable. Empirical findings emerging from studies of lung-cancer screening suggest this notion to be incorrect. These studies all suggest that early-stage histologically confirmed lung cancers identified by screening are not precursors of advanced or incurable disease. If true, these findings have potentially important implications for the development and assessment of new lung-cancer screening approaches.