Abstract. -We define and analyse an elementary model for a network of strings with random natural lengths. As this system is expanded, a threshold is reached where some of the strings form a taut string network. Further expansion increases the fraction of taut strings, until eventually all strings are taut and the classical problem of a spring network is recovered. Here we provide an analysis of the properties of the system at and beyond the threshold expansion, relating results to the expectations of constraint theory.Introduction. -We will define and analyse an elementary model which has a close affinity to existing models that represent the compression of randomly packed soft spheres, often invoked as a description of liquid foams [1, 2] and randomly cut elastic strings. It therefore belongs to the subject of "rigidity percolation" [3]. It appears to be distinct from either of its antecedents, and not trivially related to them. It was originally conceived as a variation on the theme of soft disk and sphere packings, as formulated by Liu and others [4][5][6]. That work is mainly founded on Hooke's law interactions which act under compression only (similar to those found for example in the interaction between the spheres in a Newton's cradle [7]). It has proved to be extremely rich in subtle phenomena, such as non-integer indices that relate contact number variations with compression. It occurred to us to simply turn this problem "inside-out", by defining Hooke's law interactions under extension only, and hence a model of elastic strings that are loose under compression. The strings initially connect nearest-neighbour vertices of a regular lattice and are given natural lengths l i that are random variables. Here we will use a uniform distribution for l i . We may ask: when and how does the resulting network become taut as its boundaries are expanded? This is analogous to the "jamming" problem of compressed hard spheres, but simpler in some respects, as we shall see.We have simulated the above system for large numbers of vertices in two dimensions, using periodic boundary conditions. Results for different lattices are quite distinct, depending on their coordination number. Here we will consider the cases of the hexagonal, square and triangular networks, with each vertex having three, four and six neighbours, respectively.