In their 1992 essay ‘The image of objectivity’, and again in Objectivity (2007), Lorraine Daston and Peter Galison describe the development of ‘mechanical objectivity’. Nineteenth-century scientists, they argue, pursued ‘truth-to-nature’ by enlisting ‘self-registering instruments, cameras, wax molds, and a host of other devices […] with the aim of freeing images from human interference’. This emphasis on self-recording devices and the morals of machinery, important as it is, tends to focus our attention away from the often messy and convoluted means of image reproduction – by lithograph, hand-coloured engraving or photomechanical process, and often involving steps that seem sharply at odds with narratives of increasing standardization and scientific restraint. This essay draws on the Science Museum’s pictorial collections in order to look again at the construction of objectivity, this time from the point of view of making and reproducing images. Case studies are presented of the Luke Howard collection of cloud drawings and James Nasmyth’s lunar photographs, suggesting that scientists were more flexible in their approach to depictions of the truth than has previously been supposed, and that ‘manufactured’ may be a better term than ‘mechanical’ when we talk of objectivity in the nineteenth century. But this is also a reflexive story, about the collections of the Science Museum – an institution whose own history is, I argue in conclusion, particularly tied up with issues of accuracy, depiction and genre. These are brought together in the consideration of ‘atmosphere’ – a term as important for the historian of science as for the exhibition curator.