Polar codes are a class of capacity-achieving codes for the binary-input discrete memoryless channels (B-DMCs). However, when applied in channels with intersymbol interference (ISI), the codes may perform poorly with BCJR equalization and conventional decoding methods. To deal with the ISI problem, in this paper a new joint successive cancellation (SC) decoding algorithm is proposed for polar codes in ISI channels, which combines the equalization and conventional decoding. The initialization information of the decoding method is the likelihood functions of ISI codeword symbols rather than the codeword symbols. The decoding adopts recursion formulas like conventional SC decoding and is without iterations. This is in contrast to the conventional iterative algorithm which performs iterations between the equalizer and decoder. In addition, the proposed SC trellis decoding can be easily extended to list decoding which can further improve the performance. Simulation shows that the proposed scheme significantly outperforms the conventional decoding schemes in ISI channels.