This investigation, based on CADIS (Corpus of Academic Discourse compiled at the University of Bergamo, Gotti 2006) and relying on a corpus of legal RAs written in English and published between 1980 and 2010, proposes an integrated model − combining parameters and criteria drawn from different studies (White / Hernandez 1991, Hagan 2004, Hartley 2007, Eyrolle et al. 2008) − for analysing RA titles and assessing the effectiveness of the different possible constructions. This model examines such strategies according to three different levels, namely: the structural level (distinguishing between nominal, compound and full-sentence constructions) on the basis of which it will be possible to measure the degree of informativeness found in titles (Swales 1990, Fortanet 1998, Soler 2007, Jaime Sisó 2009); the sentence level, and more specifically the level of syntactic relations between nominal, verbal and prepositional phrases (distinguishing between topic, bi-head and focus titles), which will provide the basis for measuring the semantic transparency in relation to the varying degree of accurateness and vagueness (Wang/Bai 2007); the lexical level, accounting for the presence of specialized and domain-specific lexis, through which to measure the specificity of titles and their gate-keeping function, in that specialized lexis determines the degree of text accessibility and appeal (Hartley 2005, 2007). The application of such model to our corpus allows us 1) to establish which title typology is most appropriate for legal RAs, 2) discuss its prototypicality in relation to the disciplinary epistemology at the basis legal research, and 3) account for the possible degree of variation in titling practices over time.