Metaphor and analogy have been suggested to be at the core of cognition [8]. Conceptual blending, a theory for concept invention, is build on similar mechanisms with the difference that two source domains are merged into a novel domain [3,4]. Investigations into concept invention have lead to an increased interest in the physical body‘s sensorimotor experiences that appears to play a central role in both metaphors and conceptual blending. In embodied/grounded cognition, concepts, including the most abstract ones, are understood to be directly related to these experiences [13,3]. Expressions such as “burning with passion”, “heartbreak” and “life is a roller-coaster” all display bodily experiences.

While embodied theories explain parts of language and conceptual thought, there is limited understanding what these experiences correspond to cognitively. One theory to make these experiences more concrete is the theory of image schemas [9.12]. Understood as conceptual building blocks, they model spatial relationships between objects, an observer, and the environment. Commonly mentioned image schemas are ‘above’, ‘containment’, ‘support’ and ‘path’-following.

Image schemas develop in infancy and become more fine-tuned during cognitive development to adapt to the environment [14,15]. Through analogy they help to build expectations of novel situations (e. g. if ‘support’ has been learned from “tables ‘support’ plates”, “desks ‘support‘ books” can be inferred) and it is in the disruption of expectations that novel image schema concepts are learned and new understandings comes to formed (e. g. how water does not always ‘support’ objects).

Many metaphors utilize these image schemas by instead of transferring attributes they transfer the image schematic relationships [10]. An example is how ‘above’ and up/down schemas explain status and social hierarchies (“She is above my league”, “a career ladder” and “falling from grace”). This phenomenon is also used in the arts. For example musical-pitch is often visualized as a vertical axis [1,2].

Our current work investigates how these image schematic skeletons can be used in computational concept invention in formal approaches to conceptual blending [11], trying to bridge the gap between human language comprehension and computational language production (see [6,7]).