Does the hippocampus build scenes or just link things together?
Imagining a 3D scene and imagining the same objects on a flat grid engaged different parts of the hippocampus, suggesting it contains several processing circuits rather than doing one job.
Source
Differentiable Processing of Objects, Associations, and Scenes within the Hippocampus
Study at a glance
- Design
- Human experiment — Within-subject high-resolution fMRI of six mental-imagery tasks; key contrast is imagining three objects on a flat 2D grid (array) versus on a 3D grid (scene), matched for associative demands.
- N
- N=30 · 30 healthy right-handed adults scanned; eye-tracking data were available for 22 and memory-test data were missing for seven; a separate group of 10 rated the stimuli beforehand.
- Population
- Healthy right-handed young adults (20 women, mean age 24) in London
- Outcome
- BOLD activity in the hippocampus and surrounding medial temporal and visual cortex (task partial least squares); vividness, eye movements and surprise memory as control measures
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Thirty adults lay in a high-resolution fMRI scanner with their eyes open on a blank screen and, while hearing descriptions of three objects, imagined them either one at a time in the centre, arranged on a flat 2D grid (an 'array'), or arranged on a 3D grid that felt like a scene. Control tasks included imagining a fixation cross or empty 2D or 3D grids. Because array and scene tasks needed the same binding of objects and space, the contrast between them isolates scene construction from associative processing.
What they found
Scene construction preferentially activated a small anterior medial part of the hippocampus (the pre/parasubiculum) along with parahippocampal, retrosplenial and posterior cingulate cortex. Array construction instead engaged the posterior hippocampus, entorhinal and perirhinal cortex and early visual areas, and imagining single objects activated an anterior lateral hippocampal region near prosubiculum/CA1. Imagining an empty 3D grid did not activate the hippocampus more than a 2D grid. Eye movements, vividness, difficulty and later memory did not differ between the array and scene tasks, so these could not explain the brain differences.
The limits
What it doesn't show
Mental imagery cannot be observed directly, so compliance was inferred from catch trials, eye tracking, ratings and drawings. The fMRI resolution still cannot reliably separate hippocampal subfields, and the posterior hippocampus finding was part of a larger visual-cortex cluster. The object-only task was not matched to the others and produced more vivid imagery and better memory. The authors note that other untested processes, such as imagining objects at a distance, might differ between the scene and array tasks.
Key terms
- Relational (associative) theory
- The view that the hippocampus mainly binds arbitrary relations between elements of an experience, whether or not they are spatial.
- Scene construction theory
- The view that a key hippocampal function is building internal models of the world as spatially coherent 3D scenes.
- Pre/parasubiculum
- Subregions in the anterior medial hippocampus that this and earlier studies link to scene processing.
- Perirhinal cortex
- A medial temporal lobe region next to the hippocampus strongly associated with processing objects.
- Partial least squares (PLS)
- A multivariate fMRI analysis that finds distributed patterns of activity differing between task conditions, tested with permutation and bootstrap statistics.
Flashcards
0 of 12 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 contrast was of primary interest in this study?
Common questions
Why compare arrays with scenes instead of scenes with rest?
Arrays and scenes both require binding three objects to a space, so the only difference is whether the result feels like a 3D scene; comparing them removes associative processing as an explanation.
How did the researchers know people were really imagining what they were told to?
They used catch trials, in-scan eye tracking, vividness ratings, post-scan drawings and debriefing questions, which together showed participants followed instructions and that the array and scene tasks were matched.
Does this mean relational theory is wrong?
Not exactly; the authors argue that both kinds of processing happen in different hippocampal circuits, so theories focusing on only one function are incomplete.
More on Memory systems