Time-series of collaboration trends indicated through co-authorships are examined from 1800 to presence in mathematics, logic, and physics. In physics, the share of co-authored papers expands in the second half of the19th century, in mathematics in the first decades of the 20th century, in logic in the second half of the 20th century. Subdisciplines of mathematics, of physics, and areas of logic show large differences in their respective propensities to collaborate. None of the existing explanatory approaches meets this heterogeneity; the most salient feature is a propensitiy to collaborate in fields where theoretical and applied research is combined.