Idan Landau (BGU)

A Selectional Criterion for Adjunct Control


Nonfinite adjuncts display a non-uniform control distribution: While all adjuncts accept control by the local matrix subject (Obligatory Control, OC), only some accept other controllers (non-obligatory control, NOC). For example, the rationale clause in (1a) allows NOC but the stimulus clause in (1b) does not.

(1) a.  Maryi has made up her mind. Billj would present the speakers
    [in order PROi to give himj the opportunity to practise their names].
  b.  Maryi giggled. Billj smiled [PROj/*i to see heri/*himj in underwear].

The question which adjuncts fall in which category, and why, has rarely been addressed. Following Landau 2015, control operates via predication (a property-denoting clause) or logophoric anchoring (a propositional clause). The (possibly null) prepositional head of Strict OC adjuncts (as in (1b)) s-selects a property, while that of alternating OC/NOC adjuncts (as in (1a)) s-selects either a property or a proposition. This selectional distinction is independently detectable by testing whether the adjunct accepts a lexical subject, providing us with a reliable predictor of its control behavior. In this talk, I examine 10 different types of adjuncts in English and demonstrate how this system derives their control patterns. It is further shown that purely configurational theories, that posit complementarity between OC and NOC, are empirically inadequate. Finally, I address the question of why the predicative variant of nonfinite adjuncts is available by default (within and across languages), whereas the propositional variant is not. The explanation hinges on the principle of Economy of Projection, which favors the smaller, predicative variant over the propositional one. The dual analysis of adjunct control offers insights into puzzling language-internal facts as well as typological generalizations, so far unrelated in the theory of control.