Pseudo-partitives are strings of the form [N1 – of – N2] in which N1 denotes a quantity or amount of whatever it is that N2 denotes and in which N2 is a bare nominal. Such strings come in two types, depending on whether the combination shares the number value of N1 or N2. The first type can be analyzed along familiar lines, but the second one is a hard nut to crack. The article presents existing treatments, showing that they all involve departures from independently motivated principles. As an alternative we propose an analysis that is cast in the Typed Feature Structure notation of Head-driven Phrase Structure Grammar. It handles both types of pseudo-partitives, arguing that N1 and the preposition of are complement-selecting heads if the number value is shared with N1, while N1 and the preposition are head-selecting functors if the number value is shared with N2. The switch from head to functor status is characteristic of grammaticalization and is shown to affect pseudo-partitives with a quantifier noun or a collection noun as N1, but not pseudo-partitives with a measure noun or collection noun. Examples and quantitative data are extracted from COCA.

This paper explores the interaction of regularity and idiosyncracy in the formation of nominals. It treats both nominals whose formation is highly regular, such as red box, and nominals whose formation is rather idiosyncratic, such as the Big Mess Construction (bmc; so good a bargain) and the Binominal Noun Phrase Construction (bnpc; her nitwit of a husband). Both the bmc and the bnpc conform to productive patterns, but the proper place of those patterns in the grammar as a whole is not easy to identify. To rise to the challenge, we build on recent developments in Head-driven Phrase Structure Grammar and the more formally inclined strands of Construction Grammar. Taking a cue from the treatment of clausal constructions in Ginzburg & Sag (2000), we develop a bi-dimensional hierarchy of nominal phrase types, in which the regular nominals inherit their properties from independently motivated higher types, while the idiosyncratic nominals are characterized by a mixture of inherited and inherent properties. The resulting treatment is sufficiently flexible to deal with the subtle interaction between the regular and the idiosyncratic, and sufficiently rigorous to be falsifiable. It is also compared with alternative treatments.