C-Command & Binding Domains

Binding theory is geometric: an antecedent can be visible to a pronoun or reflexive because of where it sits in the constituency tree.

Mechanistic pages often use "binding" for a pointer from a use site back to a source. Linguistic binding is narrower. It asks which noun phrases are licensed as antecedents for reflexives, pronouns, and names. The basic geometry is c-command: a node c-commands the material under its sister.

Figure 1 · Click a node to flood its c-command domain

1. C-command is a tree relation

A node c-commands its sister and everything below that sister. In "John said Mary praised himself", John c-commands the embedded clause, but Mary c-commands the reflexive inside her local clause. Principle A requires a reflexive to have a local antecedent, so Mary is licensed and John is not.

Figure 2 · Constituency relation versus dependency-depth proxy

2. Binding principles are local tests

Principle A says a reflexive needs a local binder. Principle B says an ordinary pronoun must be free in its local domain. Principle C says a name cannot be bound by a co-referring expression that c-commands it. The principles are not just labels; they change when an antecedent is dragged in or out of the local domain.

Figure 3 · Move the antecedent in or out of the local domain

3. Islands block filler-gap dependencies

Filler-gap dependencies use similar geometry but a different constraint family. The gap can be linked to a displaced filler across ordinary clause boundaries, but not out of islands such as complex noun phrases or adjuncts. An island suite therefore tests whether a model is tracking structural boundaries rather than only linear distance.

Figure 4 · Dragging a gap across an island boundary
Related pages

What next