1. The Neuroscience of Non-Literal Language
Understanding how the brain processes figurative language has been a central challenge in cognitive neuroscience for three decades. Early 'two-stage' models proposed that comprehenders first process literal meaning, detect an anomaly, and only then engage inferential processes to recover figurative meaning. This account has been substantially revised in light of neuroimaging evidence showing that figurative and literal language processing are not neatly sequential.
Conceptual Blending Theory (CBT), proposed by Fauconnier and Turner, offers a more dynamic model. Rather than viewing metaphor as the mapping between two existing structures, CBT proposes that novel meanings emerge through the integration of elements from multiple input mental spaces into a new 'blended space.' This blending process, we hypothesised, should be neurally distinct from both literal language and conventional metaphor processing.
2. Experimental Design and Stimuli
We constructed 180 sentence pairs: 60 novel metaphors ('Her argument was a labyrinth without an exit'), 60 conventional metaphors ('He attacked every weak point in her argument'), and 60 literal paraphrases ('Her argument was difficult to follow and offered no clear solution'). Stimuli were pre-tested with an independent sample to verify novelty and comprehensibility ratings.
Participants were scanned while reading sentence pairs and making acceptability judgements. Task design balanced cognitive engagement across conditions to isolate processing differences rather than difficulty effects. All participants showed equivalent reaction times across conditions, confirming successful experimental control.
3. Key Neural Results
The whole-brain contrast of novel > conventional metaphor revealed significant activation in three regions: bilateral inferior frontal gyrus (IFG), left posterior superior temporal sulcus (pSTS), and right anterior temporal lobe (ATL).
The right ATL finding is particularly significant. This region has been associated in previous work with semantic combinatoriality and distant semantic associations — precisely the cognitive operations that CBT predicts should be required for blend construction. The degree of right ATL activation correlated significantly with behavioural ratings of metaphor 'creativity' (r = 0.51, p < 0.001), suggesting this region specifically supports the generativity of the blending process.
Bilateral IFG activation, typically associated with controlled semantic retrieval, was specific to novel metaphors. This pattern is consistent with the greater cognitive effort required to establish connections between distant semantic domains — a process that should diminish as metaphors conventionalise.
4. Connectivity and Network Analysis
Psychophysiological interaction (PPI) analysis revealed that during novel metaphor processing, IFG showed significantly increased functional connectivity with right ATL and posterior middle temporal gyrus (pMTG) relative to conventional metaphor processing. We term this the 'blend integration triad.'
This network profile is consistent with a model in which IFG provides controlled access to semantic memory, pMTG contributes combinatorial semantic processing, and right ATL performs the cross-domain integration that constitutes blend construction. The three nodes show mutual enhancement during novel metaphor processing but not during conventional metaphor or literal language processing, providing network-level evidence for the qualitative distinction between these processing modes.
5. Implications for Metaphor Theory and Clinical Applications
These findings have several important implications. Theoretically, they provide the first direct neural evidence for a blend-specific network distinct from conventional metaphor processing, supporting CBT's claim that novel and conventional metaphors are cognitively — and, we now show, neurally — distinct.
For clinical practice, the identification of a specific neural network for conceptual blending opens new possibilities for assessing and potentially remediating deficits in figurative language understanding, which are prominent in schizophrenia spectrum disorders, autism spectrum conditions, and acquired language disorders following right hemisphere stroke. Understanding the neural architecture of blending is a necessary precondition for targeted intervention.