A triple junction is the point where the boundaries of three tectonic plates meet. At the triple junction each of the three boundaries will be one of three types – a ridge (R), trench (T) or transform fault (F) – and triple junctions can be described according to the types of plate margin that meet at them (e.g. fault–fault–trench, ridge–ridge–ridge, or abbreviated F-F-T, R-R-R). Of the ten possible types of triple junctions only a few are stable through time (stable in this context means that the geometrical configuration of the triple junction will not change through geologic time). The meeting of four or more plates is also theoretically possible, but junctions will only exist instantaneously.

The first scientific paper detailing the triple-junction concept was published in 1969 by Dan McKenzie and W. Jason Morgan. The term had traditionally been used for the intersection of three divergent boundaries or spreading ridges. These three divergent boundaries ideally meet at near 120° angles.

In plate tectonics theory during the breakup of a continent, three divergent boundaries form, radiating out from a central point (the triple junction). One of these divergent plate boundaries fails (see aulacogen) and the other two continue spreading to form an ocean. The opening of the south Atlantic Ocean started at the south of the South American and African continents, reaching a triple junction in the present Gulf of Guinea, from where it continued to the west. The NE-trending Benue Trough is the failed arm of this junction.

In the years since, the term triple-junction has come to refer to any point where three tectonic plates meet.

The properties of triple junctions are most easily understood from the purely kinematic point of view where the plates are rigid and moving over the surface of the Earth. No knowledge of the Earth's interior or the geological details of the crust are then needed. Another useful simplification is that the kinematics of triple junctions on a flat Earth are essentially the same as those on the surface of a sphere. On a sphere, plate motions are described as relative rotations about Euler poles (see Plate reconstruction), and the relative motion at every point along a plate boundary can be calculated from this rotation. But the area around a triple junction is small enough (relative to the size of the sphere) and (usually) far enough from the pole of rotation, that the relative motion across a boundary can be assumed to be constant along that boundary. Thus, analysis of triple junctions can usually be done on a flat surface with motions defined by vectors.

Triple junctions may be described and their stability assessed without use of the geological details but simply by defining the properties of the ridges, trenches and transform faults involved, making some simplifying assumptions and applying simple velocity calculations. This assessment can generalise to most actual triple junction settings provided the assumptions and definitions broadly apply to the real Earth.

A stable junction is one at which the geometry of the junction is retained with time as the plates involved move. This places restrictions on relative velocities and plate boundary orientation. An unstable triple junction will change with time, either to become another form of triple junction (RRF junctions easily evolve to FFR junctions), will change geometry or are simply not feasible (as in the case of FFF junctions). The inherent instability of an FFF junction is believed to have caused the formation of the Pacific plate about 190 million years ago.

By assuming that plates are rigid and that the Earth is spherical, Leonhard Euler's theorem of motion on a sphere can be used to reduce the stability assessment to determining boundaries and relative motions of the interacting plates. The rigid assumption holds very well in the case of oceanic crust, and the radius of the Earth at the equator and poles only varies by a factor of roughly one part in 300 so the Earth approximates very well to a sphere.