Summary: Visual programming languages (VPLs) appear to have great promise towards lessening the distinction between programming and problem solving, by making programming more direct, more immediate, and simpler. However, making VPLs suitable for solving large, realistic programming problems often seems to require reintroducing the very complexities that VPLs seek to avoid. This is called the scaling-up problem. This paper discusses the opportunities and obstacles VPLs encounter when attempting to solve the scaling-up problem. An understanding of this problem, its components, and the interrelationships among these components is necessary to ensure that efforts to solve the problem are productive and that these efforts do not compromise the qualities that make VPLs attractive.