Constraining argument structure in nominalizations: The case of English -er

Abstract

The article contributes to the typology of structural factors constraining argument realization in nominalizations, focussing on English -er-nominals. It reappraises conclusions of earlier studies on when -er-nominals allow argument realization. Derivations disallowing argument linking are treated as semantically and structurally parallel to nominal compounds, and their argument-structural behaviour is attributed to a generalization that non-heads of basegenerated complex heads cannot realize arguments, with principled exceptions. In argumentrealizing -er-nominals, some speakers allow the full range of argument structures permitted by head movement analyses, while less liberal varieties require a (lexicalist-inspired but syntactically implemented) analysis where -er is an Agent-realizing affix that selects V°, forcing arguments of V to merge above affixation as arguments of nouns, which is only possible for PP and of-insertion arguments.Keywords: nominalization, argument structure, compounds, affixation, eventiveness IntroductionA central problem concerning the interfaces between syntax, semantics and morphology is that category-changing processes disallow the realization (inheritance) of the arguments of the related verb to varying degrees, as seen in (1)- (3).(1) a. an amputee ( (*) of a leg) Nominalizations with -er like those in (1) have attracted considerable attention as a revealing problem for theories of argument structure (e.g. Levin & Rappaport 1988, van Hout & Roeper 1998, Alexiadou & Schäfer 2010, Borer 2013, Roy & Soare 2014, but I argue that not all empirical generalizations in previous work are correct and that data like (1) require explanations which do not figure in previous accounts of nominalizations.One innovation proposed here concerns nominalizations like (1b) which block all (phrasal) arguments of their parent verbs. After describing problems with existing accounts of these, I argue that of-phrases are excluded for the same reasons as those in compounds like tuning device (*of guitars). The generalization is that (phrasal) arguments of base-generated complex heads cannot be realized, with certain exceptions which are given principled explanations.Another claim made here concerns uses of -er like (1a) which do allow inheritance of arguments. Non-lexicalist studies standardly assume that such nominalizations involve head movement (e.g. Alexiadou & Schäfer (2010), Baker & Vinokurova (2009), Borer (2013, Bowers (2011), Harley (2008, Roy & Soare (2014)). Such approaches share two assumptions: (i) the nominalizing affix is a syntactic head which selects a phrasal projection containing V and its arguments (VP in (4a)), and (ii) V undergoes head movement to the nominalizer to form a derived nominal such as washing, washer. (Both assumptions hold regardless of differences between accounts regarding the size of the constituent initially selected by the affix, and how of-marking of V's argument is implemented.)By contrast, lexicalist studies standardly capture argument inheritance phenomena using str...