<p>A biased evaluative description represents an individual positively, but at the same time signals that the speaker relies on a prejudiced stereotype about members of a social group to which the individual belongs. Bernstein argues that this category of socially harmful language does not fit neatly into any existing category of harmful speech (such as dog whistles or slurs), nor can it be straightforwardly explained by appealing to linguistic frameworks such as Gricean implicature. This article categorizes biased evaluative descriptions as a sub-category of oppressive praise, as described by Holroyd, and argues that the pragmatic framework presented by Levinson and expanded by Björkholm can be used to linguistically analyze and understand this category of oppressive speech. It is argued that the application of the linguistic framework not only elucidates how biased evaluative descriptions relate to stereotypes and implicit bias. It also elucidates practical issues concerning what is appropriate to do in the conversational settings where biased evaluative descriptions are used and difficulties in combatting their effects through counter-speech.</p>