It is shown that the diluted two-dimensional central-force problem belongs to a new class of percolation problems. Geometric properties such as the fractal dimension of the backbone, the correlation-length exponent, and the connectivity are completely different from those of previously studied percolation problems. Explicit calculations of the backbone and the construction of an algorithm which identifies the infinite rigid cluster clearly demonstrate the absence of singly connected bonds, the overwhelming importance of loops, and the long-range nature of the rigidity.PACS numbers: 64.60. Cn, 05.70.Jk, 63.70. + h The percolation model 1 has, for a number of years, been a source of insight into many diverse physical phenomena. 2 Recently, it has been employed to understand elastic properties of tenuous media such as gels, 3 ' 4 sinters, 5 or even certain glasses. 6,7 It was originally suggested by de Gennes 3 in the polymer context that the critical properties of the elastic moduli at the percolation threshold would be the same as those of the conductivity. It has become evident, however, that this is not generally true, 8 " 10 one of the reasons lying in the different tensorial character of the problem.Recent work has concentrated on two models of elasticity percolation, the bond-bending model 9,11 and the central-force model. 8,12 It has become clear that they have different critical elastic behavior but the full nature of this difference and its underlying causes have not yet been generally appreciated. In this Letter we address this issue: We present the first analysis of the geometry of the percolating rigid clusters for the two-dimensional central-force model. While the backbone geometry of the bond-bending universality class is the same as that of ordinary percolation, 9 we show here that for the central-force model, the geometry is both quantitatively and qualitatively different. We compute the backbone fractal dimension 13,14 to exhibit a quantitative difference and present graphic and algorithmic evidence which show qualitative differences such as the absence of singly connected bonds, the importance of loops, and the essentially long-range nature of elastic connectivity.The best-understood model for elasticity percolation is the so-called bond-bending model. 9,11 That model is the simplest realization of a general class of models 15 where geometric connection implies elastic connection. For these models the threshold, p c , and the percolation critical exponents £ p , v p , y p , and a p are identical to those of the equivalent geometric problem. "Dynamic" exponents, however, depend on the problem: Namely, the exponent /which governs the elastic moduli is different from the exponent t which governs the conductivity. Nevertheless, the basic physics which controls these exponents can be readily understood in both cases from the "nodes-linksblobs" picture 14 of percolation clusters. In fact, it is fair to say that in the vast majority of percolation problems encountered to date, even continuum percolation,...