The expression "blends " is often used to refer to a type of data where, very visibly, two or more inputs are partially mapped onto each other and selectively projected to a new mental space in which novel structure can emerge (Fauconnier and Turner 1994, 1998, 2002). Famous examples of such blends are The Buddhist Monk, Regatta, Nixon in France, Complex Numbers, The Image Club. As it turns out, far from being exceptional, marginal, or genre-specific, such blends are all over the place, and especially visible in fields as different as scientific discovery, humor, advertising, or religious rituals. What warranted a new category for this kind of data when we first studied it was that it didn't fit into any of the known mapping schemes, in particular the source– target scheme of metaphor theory as understood at the time, or analogy, or metonymy, or simple framing. Methodologically, the abundance of previously unnoticed (and hence never analyzed) "blending " data suddenly offered a wealth of empirical resources to study with precision the cognitive operations 1 of mapping and integration that made such