Lexical bundles have always played an important role in describing the linguistic features of a language. As an English for specific purposes, maritime legal English has its own specialised patterns. However, there is little research focusing on the lexical bundles frequently used in different maritime legal genres. This present paper attempts to investigate frequently used lexical bundles in four maritime legal genres, namely case law, legal documents, legislation and academic legal articles, based on a self-built Maritime English Law Corpus. In this present study, the text dispersion-based keyword analysis proposed by Egbert and Biber (2019) will be used to extract domain-specific lexical bundles, as keywords serve as an efficient indicator of such lexical bundles according to Qi (2019). All 2- to 5-word lexical bundles uniquely used in four maritime legal genres will be extracted using WordSmith Tools 8.0 (Scott, 2020). In order to compare the differences in lexical keyword bundles between the four maritime legal English genres, four-word keyword bundles will be classified into VP-based bundles, NP-based bundles and PP-based bundles according to Chen and Baker's (2010) syntactic classification criteria. Furthermore, these lexical keyword bundles are analysed in terms of semantic preference and semantic prosody. Some examples are given to illustrate how the same lexical bundles are used semantically differently in general English BNC than in maritime English legal texts. The results reveal unique linguistic features specific to maritime legal texts, as well as implications for the study of formulaic language.