The research is designed to illuminate the psycholinguistic processes involved in implementing long-distance dependencies during language production. Such dependencies constitute a considerable theoretical challenge, since the abstract structural knowledge required to reliably format the discontinuous elements does not straightforwardly mesh with the articulatory requirement of incremental, word-by-word production. The dependencies to be explored are, primarily, number agreement between subjects and verbs, and, secondarily, number agreement between subjects and subject anaphors. Because verb agreement in English may call on a range of information about number from the nonlinguistic message, from the lexicon, from the morphology, and from other sources, it offers many opportunities to examine the transmission and interaction of information during language production. Conversely, because it exhibits strict syntactic conditioning , it offers opportunities to observe constraints on the transmission and interaction of information. Predictions about the nature of information transmission and the types of constraints likely to operate are derived from an approach to production developed by Garrett (1975, 1988) and elaborated by others (Bock, 1987a; Lapointe, 1985; Lapointe and Dell, 1989; Levelt, 1989). The experiments employ agreement-error elicitation tasks inn tandem with time pronunciation tasks to explore the nature of the number information used in the agreement operation, the nature of subject identification and maintenance operations, and the processing demands of agreement implementation.