We consider the problem of joint detection and decoding of low-density parity-check (LDPC) coded signals over partial response (PR) channels. A method to graphically represent the constraints imposed by the channel and the code on the channel output sequence is introduced. This enables the design of a detector and decoder that estimates a posteriori probabilities of noiseless channel output symbols rather than binary channel inputs. By running the sum-product algorithm (SPA) on this graph, a joint decoder is obtained that is shown to perform significantly better than the turbo-equalizer, at the cost of increased computational complexity.