Brain Region Analysis — Metaphor vs. Literal

The orbitofrontal cortex (OFC) is the dominant metaphor-predictive region in our classifier, but this is not a well-established finding in the figurative language literature. Meta-analyses of metaphor processing (Rapp et al., 2012) identify left temporal, bilateral inferior frontal, and medial frontal regions — not the OFC — as the canonical metaphor network. The OFC is primarily known for reward evaluation, outcome prediction, and flexible value computation. Our interpretation is that metaphorical language may carry richer emotional and evaluative content than literal language, engaging the OFC's evaluation machinery. This is a plausible but speculative explanation — the OFC's role in metaphor processing has not been directly established.

The left ITG is part of the ventral language stream and supports semantic processing. Meta-analyses of figurative language processing (Rapp et al., 2012) identify the left temporal lobe as a key region for metaphor comprehension. The ITG specifically supports the retrieval and integration of distant semantic associations — mapping meaning from one conceptual domain to another.

The SFG/mPFC is implicated in processing novel metaphors. Subramaniam et al. (2013) showed that novel unfamiliar metaphors elicit significant activation in the medial prefrontal cortex, posterior cingulate cortex, and right inferior parietal lobe. The mPFC likely mediates attention and cognitive control for integrating unusual semantic associations. Since our stimuli include creative metaphors ("His anger was a volcano about to erupt"), this activation is expected.

The temporal pole is a "semantic hub" that integrates conceptual features across modalities (Patterson et al., 2007). TMS studies by Pobric et al. (2007) confirmed that disrupting the temporal pole impairs semantic processing broadly. Patients with semantic dementia (which causes temporal pole atrophy) lose conceptual knowledge across all modalities. While the temporal pole is well-established for general semantic integration, its specific role in metaphor (as opposed to semantics broadly) is our interpretation — metaphor requires binding distant conceptual domains, which is the type of cross-domain integration the temporal pole supports.

Motor and somatosensory regions. One explanation is embodied cognition — many metaphors reference bodily experience ("grasping an idea," "heavy heart"), which may activate sensorimotor simulations. Desai et al. (2011) showed that action-related metaphors activate motor regions, though less strongly than literal action sentences. However, the large weight share seems disproportionate and may also reflect TRIBE v2's TTS pipeline producing different articulatory patterns.

This is by far the strongest literal-predictive region, accounting for nearly half of all literal classification weight. The occipital pole is primary visual cortex. This is highly consistent with the literature — concrete, literal language activates visual cortex more strongly than abstract or figurative language (Binder et al., 2005). When processing "The cat sat on the mat," the brain generates stronger visual imagery than when processing "Time is a thief." Visual and linguistic semantic representations are aligned at the border of visual cortex (Popham et al., 2021).

Supports scene construction and contextual/spatial processing (Epstein & Kanwisher, 1998). Literal sentences describe concrete, spatially situated events, which would engage scene-level representations more than abstract metaphors.

Involved in spatial attention and visuospatial processing. Literal sentences with concrete spatial content would engage dorsal stream spatial processing more than figurative language.