From a corpus consisting of free text comments submitted to the website RateMyProfessors.com, adjectives describing people have been identified and clustered using Principal Components Analysis. This follows the Meaning Extraction method used by Chung & Pennebaker (2007). The outcome is a set of seven factors, each of which is interpretable as representing a dimension of meaning along which individual professors have been evaluated. These dimensions are in turn discussed using Martin & White’s (2005) parameters of Judgement and Appreciation, and Coffin’s (2002) concept of evaluative voices. It is argued that contributors to RateMyProfessors.com have available three distinct voices: ‘novice intellectual’, ‘consumer’, and ‘subordinate’. The paper demonstrates how a bottom-up, statistical technique may be used to provide the initial data for identifying evaluative parameters. It raises the possibility that such parameters may be specific to individual discourse communities. It therefore offers a complementary and problematizing alternative to top-down, researcher-driven taxonomies of evaluative meaning.