This paper starts out from the observation that the semantic labels of ‘fact’ and ‘action’ (Lees 1968) that are traditionally used in the description of English gerundive nominalizations (e.g. I worry that my posing the question defines me as a depressive ) cannot distinguish gerundives from so-called action - ing nouns (e.g. Saddam's targeting of Israel ) and also fail to capture the more subtle semantic distinctions within the system of gerundive nominalization. The descriptive analysis that is presented tries to move beyond the action/fact dichotomy, is firmly grounded in the nominal-constructional properties of the system and covers all subtypes of gerundive nominalization. Building on Schachter (1976), it argues that gerundive - ing nominalizations have shifted from the representational semantics of action - ing nouns to a more schematic, constructional semantics: gerundives, it is suggested, nominalize either a type or kind of process, with no subject implied, or they nominalize an instance of a process, characterized by the (clausal-)constructional link which it implies between the process type and a subject (Davidse 1997, Heyvaert 2003). The system of gerundive nominalization exploits all nominal-constructional options which it has within the structure of the NP to encode the absence or presence of this subject. Gerundives thus opt for non-specific, generic reference to encode the name of a type or class of process (with no subject implied) or to encode an instance with generic reference. They use specific (definite or indefinite) reference to designate a specific, non-generic instance. The subject is either included in the nominalized clausal unit (and in the oblique case), or it is encoded through nominal means: in the form of bare definite reference of the gerundive NP (signalling control by the matrix clause or anaphoric/exophoric reference with the co(n)text), or in the form of a possessive/genitive determiner. The analysis that is proposed sheds new light on previously undifferentiated categories of gerundive nominalization (such as the control type of bare gerundives), as well as points to interesting resemblances between gerundive nominalization and diachronic changes in other systems of deverbal nominalization. Crucially, it also allows for a better understanding of the semantics of the system and its differences with action - ing nominals, whose nominal-constructional options do not revolve around encoding the presence or absence of a subject but instead allow for a wide range of modifying elements and determiners.