This paper provides empirical evidence in favour of Marti Girbau’s (2003, 2010) quantifier phrase analysis (QP of N) of the partitive construction, which is based on the analysis of verbal agreement in a large data set from the BNC (British National Corpus) and COCA (Corpus of Contemporary American English). The competing binominal N of NP analysis, advocated by some, predicts that there is number agreement with the first nominal element. In order to account for verbal agreement along these lines, two further principles, that of proximal agreement and possibly that of notional agreement, are needed. The agreement patterns with the partitive construction as observed in the corpus can be explained in a more elegant way when the NP is analysed in terms of a quantifier phrase followed by a nominal head (QP of N): verb number is predicted in 95.74% of the cases (compared to 54.24% when the binominal analysis is adopted), agreement in the remaining 4.26% exceptional cases falling out of further semantic and lexical considerations.