In this paper, some issues governing the block-wise performance of convolutional interleavers used in turbo codes are presented. Two different constructions of convolutional interleaver differing by the position of stuff bits in the interleaved data block are considered here. The performance assessment is based on the contribution of each weight to the overall code performance. For the given turbo code and each utilized interleaver, weight contribution is computed to finalize the code behavior in different signal to noise ratios. Simulations have been performed to verify the conducted analysis.