The 1891 censuses taken in New South Wales and Tasmania abandoned the long-established practice of grouping working-class occupations into "skilled" and "unskilled" categories. Instead, they were grouped into "Industrial" categories that did not differentiate between grades or degrees of skill. In this paper the sudden disruption to the preceding practice is explained as an effect of the intersection of two histories: the changing meaning of skill, and the history of scientific method. The paper traces the transformations in meanings of skill from an "artisanal" to an "industrial" form and examines how the two central figures in the construction of the 1891 census - the statisticians T.A Coghlan and R.M Johnston - were enmeshed in that history. Coghlan is usually given the more prominent role in accounts of late nineteenth-century statistics, but in this case Johnston's expertise in using scientific method was instrumental in the "deskilling" of the census.