A word sense disambiguation (WSD) method based on dependency fitness is proposed to solve the problem of knowledge acquisition bottleneck in the development of WSD techniques. The method achieves automatic knowledge acquisition in WSD by taking full advantage of dependency parsing. First, a large-scale corpus is parsed to obtain dependency cells whose statistics information is utilized to build a dependency knowledge base (DKB); then, the ambiguous sentence is parsed to obtain the dependency constraint set (DCS) of ambiguous words. For each sense of ambiguous word, sense representative words (SRW) are obtained through WordNet. Finally, based on DKB, dependency fitness of all kinds of SRW on DCS is computed to judge the right sense. Evaluation is performed on coarse-grained English all-words task dataset of SemEval 2007. Compared with unsupervised and knowledge-based methods which don't utilize any sense-annotated corpus, the proposed method yields state-of-the-art performance with F1-measure of 74.53%.