Thus far I’ve been interested in predictive coding’s hierarchical inference, but cross-domain mapping allows horizontal, analogic inference as well. As Lakoff (1992) notes, much of the language we use to talk about abstract domains is derived from the figurative extension of language used to discuss analogous physical-world situations. For example, in the analogy love is a journey, physical journeys, in which a subject crosses a distance of land or space over a not-insignificant period of time, are used as a metaphor for romantic love (a dead-end street; a long and bumpy road; can’t turn back now; at a crossroads; spinning the wheels; off-track; on the rocks; taking control of the wheel; bailing out).
More importantly for our purposes, Lakoff argues that we reason about abstract and metaphysical phenomena through inferential work based in analogous physical phenomena. In other words, the inferential patterns used to reason about travel and journeys are deployed when reasoning about romantic relationships; the solutions to getting figuratively “stuck” are analogized versions of the solutions to getting literally stuck (e.g. in the mud, etc). The process ostensibly happens in correspondence with the radial proximity of given situations, that is, weakly analogous cases will have less influence than strongly analogous cases. And in the case of predictive error minimization and processing, we can imagine that the analogous situations might engage a common set of distributions, or priors, especially when one knowledge domain is weaker and thus its predictions high-error.