In 2015-16 researchers from the University of Melbourne and Museums Victoria undertook a collaborative project which sought to visualise archival data from the museum as a means to investigate the structure and context of the Dorothy Howard Collection. This article introduces Dorothy Howard's work, which is part of the internationally significant Australian Children's Folklore Collection, and looks at the project's processes and outcomes, including the initial visualisations produced. In doing so, the authors highlight the data-intensive nature of such work, suggest the potential of visualisations to reveal collection structures, and outline possibilities for future projects and collaborations. You put down on a piece of paper as many dots as you please. You must not do this just any way or you will be wrong. You must go straight up and down and straight across so that the dots will be in squares. Then you take turns joining lines between the dots. (Frank Syaranamual, 10 years old, describing a game called 'Dots' , 1954) 1 Large collections in archives and museums may hold many thousands of items, 2 none of which exist in isolation. If each item is taken as a dot, or node, they are joined by complex networks of connections-to other items as well as to things such as people, organisations, concepts and places-with these relationships accumulating through time. Some items are explicitly joined, their content referring to other items or entities, while others are linked to context within or beyond their collections in ways which help users understand their meaning. However, these networks are often not readily visible to users-whether they be internal staff of the organisation holding the collection, or external users exploring content. Archival description tends to emphasise hierarchical links between items, aggregates (series, collection) and an often-constrained set of provenance entities; 3 and museum description tends toward individual item-level descriptions, loosely coupled to donor information or to like items via common characteristics and classification structures. Furthermore, even where