The outstanding bit error rate (BER) performance of turbo codes (iterative decoding of parallel concatenated convolutional codes with interleavers) has been reported, but the computational complexity and the decoding delay from the iterative computations have been problems in many ways. This paper presents a method that can be used to reduce the decoding computational complexity in the iterative decoding of turbo codes. With reliability information, soft outputs of a decoder at the (i)-th iteration, we find that some bits in a block need more iterations than others. If the bits don``t require further iterations, the reliability information of the bits at the (i)-th iteration are reused in the next iteration processing and the computation processes of the bits are terminated. That is, using bit-oriented iteration method, we can reduce the decoding computational complexity without degrading BER performance. Assuming an additive white Gaussian noise (AWGN) channel with Eb/N0 of 2.5 dB, about 55% of computations are required as compared with those in conventional method using a fixed number 8 of iterations to decode the turbo codes. The reduction of computational complexity implies some reduction in power consumption at the mobile unit.