How do the hippocampus and prefrontal cortex build imagined scenes?
When people imagined new scenes, the ventromedial prefrontal cortex and hippocampus synchronised their slow rhythms, and the prefrontal region appeared to drive the hippocampus.
Source
The Neural Dynamics of Novel Scene Imagery
Study at a glance
- Design
- Human experiment — Within-subject MEG experiment: imagine novel scenes vs single objects vs counting baseline, cued by spoken words, with beamformer source power, theta coherence and dynamic causal modelling.
- N
- N=22 · 22 native English-speaking adults; one was dropped from the remembered-vs-forgotten coherence analysis for recognising every scene word.
- Population
- Healthy young adults (mean age 27) at University College London
- Outcome
- Theta power in the hippocampus, hippocampal-vmPFC theta coherence (scenes vs objects; remembered vs forgotten), and direction of effective connectivity
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Twenty-two adults lay in an MEG scanner with eyes closed and, after hearing a word, spent 3 seconds either imagining a novel 3D scene (e.g. 'jungle'), imagining a single object floating on a white background, or counting in threes as a baseline. They rated how detailed their imagery was, and afterwards took a surprise recognition test for the words. The authors localised theta-band (4-8 Hz) activity, measured coherence between the anterior hippocampus and the rest of the brain, and used dynamic causal modelling to compare a model where the hippocampus drives vmPFC with one where vmPFC drives the hippocampus.
What they found
Both kinds of imagery reduced theta power in the left anterior hippocampus to a similar degree compared with counting, so scenes and objects did not differ in local power. Scenes, however, produced greater theta coherence between the anterior hippocampus and vmPFC (and with fusiform and parahippocampal cortex), and this coherence was higher for scenes later remembered than for those forgotten. The model with vmPFC driving the hippocampus won with a group probability of 97.66%, supported by 11 individuals versus 2 for the reverse, and vmPFC theta changes began earlier than hippocampal ones.
The limits
What it doesn't show
The key vmPFC coherence effect only appeared at a lenient uncorrected threshold justified by a prior hypothesis, so it is weaker evidence than a corrected whole-brain result. Half the participants showed no conclusive evidence for either direction of influence, and dynamic causal modelling only compared two pre-specified models. MEG localisation of deep structures like the hippocampus is imprecise, and imagery quality was self-rated. The link to episodic memory rests on an incidental recognition test, and the high-versus-low detail difference only approached significance.
Key terms
- Scene construction theory
- The proposal that the hippocampus builds spatially coherent mental scenes, supporting both remembering the past and imagining new events.
- Theta coherence
- The degree to which two regions' 4-8 Hz oscillations stay in a consistent phase relationship, taken as a sign they are communicating.
- Magnetoencephalography (MEG)
- Recording the tiny magnetic fields produced by neural currents, giving millisecond timing of brain activity.
- Dynamic causal modelling (DCM)
- A method that fits competing models of directed influence between regions to data and uses Bayesian comparison to pick the best.
- Beamformer
- A spatial filter that estimates the activity at a particular brain location from sensor data while suppressing other sources.
- Ventromedial prefrontal cortex (vmPFC)
- The lower middle part of the frontal lobe, linked to schemas, valuation and memory, and needed for normal scene imagination.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Which comparison isolated scene construction in this study?
Common questions
Why compare scenes with objects rather than with rest?
Objects also involve vivid imagery but not a 3D space, so the comparison isolates the building of a spatial scene; rest was avoided because it triggers spontaneous activity in the same regions.
Could scenes simply be harder, explaining the extra coherence?
The authors argue not: coherence tended to be higher for richly detailed scenes rather than low-detail ones, and objects were actually remembered better than scenes.
Does this mean the hippocampus is not important for imagination?
No. The hippocampus was active for both kinds of imagery; the finding is that vmPFC appears to steer it, perhaps by selecting relevant schema elements.
More on Memory systems