The Persistence of Retinotopic Coordinates in High-Level Vision

22 concepts15 papers1 proposal

When we look around, our eyes dart from point to point several times a second. Yet, our perception of the world remains perfectly stable. The classic textbook model of the visual system offers an elegant explanation: early visual areas map the world exactly as it hits the retina (an eye-centered, or retinotopic coordinate frame), but as information flows up the visual hierarchy, the brain transforms these inputs into abstract, position-invariant, or world-centered (spatiotopic) representations. In this view, high-level areas care about what an object is, not where it lands on the retina.

However, a growing body of neuroimaging and electrophysiological evidence flips this classic model on its head. Rather than discarding the retinal frame, the brain stubbornly maintains retinotopic coordinates across nearly the entire visual hierarchy—stretching deep into regions responsible for object recognition, motion processing, scene layout, and even memory.

To understand why the brain retains this eye-centered architecture, we must look at how these maps are organized and the computational advantages they provide.

The Myth of the Spatiotopic Shift

If the brain constructed a stable, world-centered map of our surroundings, neurons in high-level visual streams should fire based on where an object sits in the room, regardless of where our eyes are pointing.

Gardner et al. (2008) directly tested this hypothesis using functional magnetic resonance imaging (fMRI). By tracking neural responses as subjects shifted their gaze, the authors demonstrated that maps of visual space in human occipital cortex remain strictly anchored to the retina, showing no shift toward world-centered coordinates.

This eye-centered fidelity persists even during natural, active vision. Sharma et al. (2024) recorded from the ventral visual stream—the pathway specialized for object recognition—during free exploration of natural scenes. They found that neurons respond exclusively to stimuli currently hitting the retina. The object-recognition pathway does not bother to transform visual inputs into a stable, world-centered coordinate system; it processes the world frame-by-frame, relative to the eye.

Retinotopy Throughout the Visual Hierarchy

Because the brain does not abandon the retinal coordinate frame, structured spatial maps can be found in high-level cortical regions once thought to be completely non-spatial or purely category-selective.

Object and Face Processing

Traditional models assumed that regions like the lateral occipital cortex or face-selective areas are position-invariant. However, Larsson (2006) identified two distinct retinotopic maps, LO1 and LO2, within the human lateral occipital cortex, demonstrating that shape recognition remains spatially organized. Furthermore, Hemond et al. (2007) showed that even highly specialized face- and object-selective cortices retain a robust preference for contralateral stimuli (objects in the opposite visual field), rather than discarding spatial location in favor of pure abstraction.

Scene Perception and Navigation

Recognizing a scene layout requires understanding space, but even here, the brain anchors its representations to the eye. Livingstone (2017) discovered two new visual field maps (OTS1 and OTS2) within the scene-selective inferior temporal cortex of macaques, proving that advanced scene processing is built upon a structured retinotopic scaffold. While areas like the occipital place area (OPA) rapidly encode spatial layout independently of low-level textures, as shown by Henriksson et al. (2019), this spatial mapping remains tied to visual coordinates.

Remarkably, this organization extends beyond traditional visual boundaries. High-resolution mapping by Benson et al. (2018) and wide-field imaging in mice by Zhuang et al. (2017) reveal that retinotopic maps extend into the retrosplenial cortex (RSC)—a region critical for memory and navigation—as well as parietal areas. Esteva et al. (2015) emphasize that mapping these precise retinotopic boundaries in the RSC is key to understanding how the brain bridges immediate vision with spatial memory.

Motion and Parietal Integration

In the dorsal stream (the "where" or "how" pathway), spatial localization is paramount. Huk et al. (2002) separated the human motion-processing complex into areas MT and MST, showing they preserve a highly organized topographic map of the contralateral visual space. Higher up in the posterior parietal cortex, Rima et al. (2020) discovered a structured "cloverleaf" cluster of four visual maps sharing a central foveal representation. Even in the ventral intraparietal area (VIP), an area heavily involved in multisensory integration, Chen et al. (2014) found that the vast majority of neurons exhibit eye-centered receptive fields rather than head- or world-centered ones.

[Retinal Input] ──> [Early Visual (V1/V2)] ──> [High-Level Ventral (LO/IT)] & [Dorsal (MT/VIP)]
      │                     │                               │                           │
      └─────────────── Retinotopic Coordinate Frame Maintained Throughout ──────────────┘

Why It Works: The Logic of Eye-Centered Maps

Why would the brain choose to maintain an eye-centered coordinate system everywhere, forcing itself to deal with a shifting visual input every time the eyes move? The answer lies in computational efficiency and architectural elegance.

1. Spatial Stability via Gain Fields

Instead of physically moving or remapping a visual coordinate frame—which would require an immense, biologically expensive recalculation of millions of neural connections—the brain achieves spatial stability by layering information.

Merriam et al. (2013) demonstrated that gaze direction modulates the signal strength (or gain) of visual neurons without changing their retinotopic receptive fields. By multiplying eye-centered visual inputs with eye-position signals, the brain creates "gain fields." Downstream areas can read out these layered signals to compute where an object is in the world, without the cortex ever needing to construct a separate, explicit world-centered map.

2. Wiring Efficiency and Connectivity

Maintaining a shared coordinate frame across different brain regions makes communication highly efficient. Honey et al. (2015) discovered that functional connectivity across the visual cortex is fundamentally organized by eccentricity (how far a point is from the center of gaze). Areas that represent the center of the gaze connect preferentially to other foveal areas, while peripheral representations connect to peripheral ones. By keeping a uniform retinotopic architecture, the brain can use simple, localized wiring rules to link distant specialized areas.

3. Multi-Feature Integration

A persistent spatial map allows the brain to easily bind different visual features together. For instance, Zhang et al. (2023) showed that spatial frequency maps in areas V2 and V4 are systematically arranged alongside orientation and color maps. Because all these features are mapped onto the same underlying retinotopic grid, the brain can instantly integrate the color, orientation, scale, and location of an object simply by looking at which columns of cortex are co-activated.

The brain does not solve the problem of a moving eye by transforming vision into an abstract canvas. Instead, it preserves the retinal map at every stage of processing, using gaze modulation and structured topography to extract stability, meaning, and memory directly from the moving eye.

Key Foundational Work

Two Retinotopic Visual Areas in Human Lateral Occipital Cortex (2006) — contributed 1 concept elements to this direction.
Retinotopy and Functional Subdivision of Human Areas MT and MST (2002) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_retinotopic_coordinate_mapping_2fd814da,
  title = {retinotopic coordinate mapping},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/retinotopic-coordinate-mapping/rdr_2fd814da852b42c3baa991d661babbe7},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_retinotopic_coordinate_mapping_2fd814da_v1,
  title = {retinotopic coordinate mapping},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/retinotopic-coordinate-mapping/rdr_2fd814da852b42c3baa991d661babbe7/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake