As described in Infinitive Clauses, infinitive clauses are analyzed with a supersense if and only if they serve as a purpose adjunct, or in certain purpose-related constructions (result; complement of entity-referring indefinite pronoun; commercial service; that which something is good or bad for, or sufficient or excessive for). The special label `i is reserved for all other uses of infinitival to, as well as for whenever it introduces the subject of an infinitive clause.1

Infinitivals warranting `i include:

  • I want to meet you. [complement of control verb] 001

  • I would_like to try the fish. [would_like is a polite alternative to want] 002

  • It seems to be broken. [complement of raising verb] 003

  • You have an opportunity to succeed. [complement of noun] 004

  • I’m ready to leave. [complement of adjective] 005

  • I’m glad to hear you’re engaged! [complement of emotion adjective] 006

  • These new keys are expensive to copy. [tough-movement] 008

  • My plan is to eat at noon. [infinitival as NP] 009

  • It’s impossible to get an appointment. [infinitival as NP, with cleft] 010

  • I know how to lead. [complement of wh-word] 011

  • I have nothing to hide. [complement of indefinite pronoun] 012

  • I have something to do. [complement of indefinite pronoun that doesn't refer to an entity] 015

Multiword auxiliaries—such as quasi-modals have_to ‘must’, ought_to ‘should’, etc., as well as have_yet_to—subsume the infinitival to, so no label on to is required:

  • You have_to choose a date.

Whenever for introduces a subject of an infinitival clause, the for token is labeled `i (regardless of whether to receives a semantic label; see Infinitive Clauses):


  1. Essentially, our position is that these uses of infinitivals are more like syntactically core elements (subject, object) than obliques, and thus should be excluded from semantic annotation under the present scheme. ↩

Usages (2)

Metadata