Plate tectonics is the governing theory that unifies the Earth Sciences and is unique to Earth. The sinking of lithosphere in subduction zones drives plate tectonics but exactly how and why subduction begins (subduction initiation, SI) remains enigmatic. Most SI models require exploitation of existing lithospheric weaknesses but these are now produced by plate tectonics; so how did the first subduction zones form? One possibility is SI along a plume head-cold lithosphere interface, but no examples have been documented. On the basis of three key observations, we show here that the Late Cretaceous tectonic evolution of Central America, NW South America and the Leeward Antilles is consistent with plume-induced SI (PISI) which nucleated along the southern and western margins of the Caribbean Plate (Caribbean Large Igneous Province, CLIP) at ~. 100. Ma. (1) Trace element chemistry of most 100. Ma and younger units interpreted as CLIP that are exposed along the southern margin of the Caribbean Plate and NW South America, record subduction additions which increased with time beginning at 100. Ma. These 'plume- and arc-related' (PAR) units are distinguishable from: (a) global oceanic plateau basalts (OPB); (b) 140-110. Ma OPB along the western edge of the CLIP; and (c) post-100. Ma OPB in the northernmost CLIP. Whereas the older OPB are compositionally identical to global OPB, the younger northerly CLIP units are similar to oceanic island basalts and record lower degrees of partial melting than PAR units exposed along the southern and western peripheries of the CLIP. Both the older and the younger northerly CLIP units lack evidence of subduction modification. (2) There is no known hiatus between CLIP and younger arc units, suggesting continuous tectono-magmatic evolution from plume to arc. (3) Generation of the CLIP and earliest, overlying and crosscutting arc units overlaps in time, space, chemical and isotopic compositions; both units are consistent with derivation from Galapagos Plume-like mantle which became increasingly subduction-modified with time. These observations illustrate that formation of the CLIP and earliest arc volcanism reflects partial melting of the same hybrid plume-subduction-modified source in a single, rapidly evolving tectonic environment. The scale of Late Cretaceous PISI in the SW Caribbean realm is consistent with the expected large scale of lithospheric collapse as seen for other SI examples, extending some 1400. km from southern Costa Rica-Panama to western Colombia and 1700. km from Ecuador to the Leeward Antilles (Aruba and Curaçao). This first documented example of PISI may be relevant to the start of plate tectonics which may have started when subduction began around Precambrian plume heads. Establishment of lava chemostratigraphy designed to establish the composition of early SI magmatic successions might provide keys for deciphering whether SI occurred along a collapsed transform (like the IBM arc; early MORB-like lavas) or around a mantle plume (like the Caribbean; early plateau-like lavas). Recognition of this first example of PISI further suggests that interaction of a sufficiently large plume head with sufficiently dense oceanic lithosphere in Precambrian time may have triggered the modern regime of plate tectonics.