Long-distance agreement (LDA) refers to agreement between a functional head in the extended projection of one lexical head with an argument of another lexical head. LDA is found in a variety of languages, including Indo-Aryan (Hindi/Urdu, Kashmiri, Kutchi), Basque, Tsez, Chukotko-Kamchatkan, and native North American languages (Innu-aimûn, Passamaquoddy). All instances of LDA described in the literature share a number of properties. First, LDA uses the same primitives as local agreement. Every element that may control LDA may also trigger local agreement. The morphological forms of LDA are also found in local agreement. Second, LDA is generally optional. This is all the more relevant because local agreement is always obligatory in these languages. Third, a verb can agree with an element in a lower clause, but not with one in a higher clause.

Virtually all studies on LDA converge in that they treat it as local in some sense. The following general lines of analysis have been proposed. First, the LDA controller is (possibly covertly) moved into the matrix clause and triggers agreement from there. Second, the LDA trigger moves to the edge of the embedded clause and agrees from there. Third, LDA depends on whether or not a locality boundary (i.e., phase head) is projected in the embedded clause (restructuring). Fourth, the LDA controller triggers local agreement with a functional head in the embedded clause. A matrix head subsequently agrees with the embedded functional head. Fifth, a silent pronoun coreferent with the embedded object is inserted into the matrix clause and triggers local agreement there. These lines of analysis are not mutually exclusive.

LDA has been argued to provide evidence against Upward Agree. It furthermore is crucial for the general question of syntactic locality. It also raises the general problem of optionality in syntax, in particular since the optionality of agreement appears to correlate with distance.