This article introduces the concept of law-encoding diagrams (LEDs) an
d presents the argument that they have had a role in scientific discov
ery that has not been previously recognized. An LED is a representatio
n that correctly encodes the underlying relations of a law, or a syste
m of simultaneous laws, in the structure Of a diagram by the means of
geometric, topological, and spatial constraints, such that the instant
iation of a particular diagram represents a single instance of the phe
nomena or a particular case of the law(s). Examples of LEDs in the his
tory of science are discussed, and the benefits of using LEDs in disco
very are considered. LEDs are distinguished from other forms of diagra
mmatic representation. Previous work on the computational modeling of
diagrammatic law induction is reinterpreted in terms of the search for
diagrammatic constraints of LEDs. A general characterization of the r
ole of LEDs in discovery is considered, and a framework for classifyin
g processes of discovery based on LEDs is proposed.