This corpus-based study investigates the nǐ zhè(ge)/gè + NP constructions in Chinese: nǐ zhège + NP, nǐ zhè + NP and nǐ gè + NP, where nǐ is ‘you’, zhège/zhè a demonstrative phrase and gè a classifier. Drawing on data from the Chinese Web Corpus 2011, we conduct a multiple distinctive collexeme analysis, together with a detailed analysis of co-text, to examine the impolite use of the constructions and their relationships with one another. Our results demonstrate that all three constructions are conventionalized for impoliteness, which contradicts the prevailing view in the literature that (im)politeness is just a matter of context and needs to be evaluated in each situation. Nǐ gè + NP is also shown to differ significantly from the other two constructions in terms of the attracted noun phrases, the proportion of impolite usage, the nature of the impoliteness, and the proportion of address usage. We therefore argue, contrary to some earlier claims, that nǐ gè + NP is an independent construction rather than a reduced form of nǐ zhège + NP. Finally, we examine how the constructions’ components –the second person pronoun, proximal demonstrative, and general classifier– contribute to their impoliteness.