This paper considers two discrete time, finite state processes X and Y. In the usual hidden Markov model X modulates the values of Y. However, the values of Y are then i.i.d. given X. In this paper a new model is considered where the Markov chain X modulates the transition probabilities of the second, observed chain Y. This more realistically can represent problems arising in DNA sequencing. Algorithms for all related filters, smoothers and parameter estimations are derived. Versions of the Viterbi algorithms are obtained. (C) 2013 Elsevier Ltd. All rights reserved.