This paper analyses two dominant ways of conceptualising the research agendas in the climate change sphere, the ‘knowledge pyramid’ and the ‘knowledge net’. Using the idea of a ‘certainty trough’ from sociology of science, and employing crop models as an example, it explores the sometimes terse relationship between climate modellers and the climate impacts community. The pressures to develop a more holisitic analysis are discussed, but we argue that much integrated assessment modelling still exhibits an implicit and acultural reductionism, and frequently misconstrues the character and significance of uncertainty as well as the role of analytical knowledge in policy making.