The Brain’s Mechanism for Color Perception: An Attractor-Based Framework

29 min readApr 25, 2026

From phototransduction to cortical dynamics, color emerges as a stabilized pattern of neural activity rather than a decoded signal

Abstract

Color perception is typically framed as the encoding of wavelength information by the visual system. This essay presents an alternative, dynamical account. External stimuli are first transduced into electrochemical signals, which act as perturbations to an already active neural system. Through distributed processing, opponent interactions, and oscillatory coordination, these perturbations are transformed and stabilized into attractor states. Color is therefore not read out from the input, but emerges as a structured, metastable pattern of activity across interacting neural populations. Evidence from color illusions demonstrates that identical stimuli can lead to different percepts depending on system constraints, supporting the view that perception depends on attractor selection rather than input alone. This framework links molecular mechanisms to system-level dynamics and reframes perception as the stabilization of activity within a high-dimensional neural system.

1. Introduction — The Problem of Color

Color appears to us as one of the most immediate and stable features of experience. A canvas filled with reds, blues, greens, and yellows presents itself as a coherent, richly differentiated whole. Each color is distinct, yet seamlessly integrated into a single perceptual scene. This stability is so natural that it invites a simple assumption: that the brain reads out color directly from the world.

However, this assumption is difficult to sustain under closer examination.

The physical input to the visual system consists of light of different wavelengths. Yet this input is:

noisy, varying with lighting conditions, shadows, and reflections
ambiguous, as different wavelength combinations can produce similar retinal responses
context-dependent, with surrounding colours and inferred illumination altering perception

The same physical stimulus can lead to different perceived colours under different conditions, while different stimuli can produce indistinguishable percepts. These observations suggest that color is not a straightforward readout of wavelength information.

This creates a fundamental problem:

How does the brain preserve distinct colours while integrating them into a unified percept?

Any adequate account must explain two seemingly opposing requirements:

differentiation — the ability to maintain multiple distinct colours simultaneously
integration — the ability to bind these colours into a coherent whole

If the brain simply combined signals, one would expect colours to blur into an average. If it kept them entirely separate, one would expect fragmentation rather than unity. Yet perception achieves both: multiple colours are experienced as distinct elements within a single, stable scene.

This tension points to a limitation in the conventional framing of perception as a process of encoding and decoding information. It assumes that the relevant object is the signal itself, and that perception consists of reading out its content.

An alternative perspective is required.

In this essay, color perception is reframed within a dynamical systems framework. External input is not treated as a message to be decoded, but as a perturbation to an already active system. The resulting percept corresponds to the state into which the system stabilizes.

Color is not read out — it is stabilized.

This shift allows the problem to be approached in terms of system dynamics rather than signal representation, providing a pathway to explain how differentiation and integration can coexist within a single perceptual state.

2. Attractors in Dynamical Systems

The concept of an attractor originates in the study of dynamical systems, a branch of mathematics and physics concerned with how systems evolve over time. Rather than focusing only on a system’s instantaneous state, dynamical systems theory examines the trajectories that systems follow through a space of possible configurations, often referred to as state space.

In this framework, any system — whether physical, biological, or neural — can be described by a set of variables whose values change over time. Each possible configuration of these variables corresponds to a point in state space. As the system evolves, its state traces out a trajectory through this space.

An attractor is a region of state space toward which the system tends to evolve. Regardless of where the system begins (within some range of initial conditions), its trajectory converges toward this region over time.

Two properties define an attractor:

Convergence: trajectories move toward a particular region or pattern
Stability under perturbation: once within this region, small disturbances do not cause the system to leave it; instead, the system reorganizes and returns to the same pattern

This second property is critical. An attractor is not simply a point the system passes through — it is a configuration the system actively maintains despite ongoing fluctuations.

Several types of attractors are commonly described:

Fixed point attractors: the system settles into a single stable state
Limit cycles: the system exhibits stable, repeating oscillatory behavior
Chaotic (or strange) attractors: the system follows complex, bounded but non-repeating trajectories that remain structured over time

Although these differ in form, they share a central feature: they describe patterns of behavior, not static locations.

This leads to the key conceptual shift:

Attractors describe how systems behave over time, not just where they are at a given moment.

They capture the tendency of a system to settle into structured, stable patterns of activity, and to maintain those patterns in the face of perturbation. This makes them a natural framework for understanding systems — such as the brain — whose function depends on the continuous stabilization of dynamic states.

3. Attractors in the Brain

The framework of attractors becomes particularly powerful when applied to the brain. Rather than viewing the brain as a system that passively receives inputs and produces outputs, it can be understood as a high-dimensional dynamical system whose activity is continuously evolving.

At any given moment, the brain is not inactive. Neural populations across cortical and subcortical regions are already engaged in ongoing patterns of activity. These populations — groups of neurons with shared functional roles — interact through complex networks of excitatory and inhibitory connections. The state of the brain at any instant can therefore be described as a point in a very high-dimensional space defined by the activity of these interacting components.

Within this space, the brain does not move arbitrarily. Its activity tends to organize into structured patterns that can be maintained over time. These patterns correspond to attractor states.

In the neural context, an attractor is not a single neuron firing or a localized signal. It is a distributed, coordinated pattern of activity across multiple neural populations. Such a pattern is characterized by:

which populations are active
how strongly they are engaged
how they interact with one another
how their activity is coordinated in time

Once the system enters such a configuration, it exhibits stability. Small perturbations — whether from external stimuli or internal fluctuations — do not immediately disrupt the pattern. Instead, the system reorganizes locally to maintain the overall structure.

This provides a natural way to understand perception.

External inputs do not create activity from nothing. They perturb an already active system. The brain responds by evolving its activity until it reaches a configuration that can be stabilized. The resulting percept is not a direct encoding of the input, but the state into which the system settles.

Perception corresponds to the brain settling into an attractor state.

Different percepts correspond to different attractor configurations, and changes in perception correspond to transitions between them. In this way, the variability and stability of experience can be understood as properties of the underlying dynamical system rather than as features of a static encoding process.

4. From Molecule to Mind: Electrochemical Origin of Perturbation

The dynamical description of perception raises a fundamental mechanistic question: what, physically, constitutes a perturbation in the brain?

Perturbations do not arise in abstraction. They originate in the biophysical processes by which external stimuli are transformed into neural activity. At their core, these processes are electrochemical, linking molecular events to large-scale network dynamics.

Phototransduction: From Light to Electrical Signal

Color perception begins with the interaction between light and photoreceptors in the retina. When photons strike photopigments within cone cells, they trigger a biochemical cascade involving G-protein–mediated signaling pathways. This cascade leads to the closure of ion channels and a change in the membrane potential of the cell.

The result is not a symbolic representation of color, but a graded electrical signal whose magnitude reflects the pattern of incoming light across different wavelengths.

This transformation illustrates a key principle:

External stimuli are not encoded as messages; they are converted into changes in electrochemical state.

The Electrochemical Loop

Once generated, these initial signals propagate through the nervous system via a recursive interaction between chemical and electrical processes.

Within neurons, signals are carried as changes in membrane potential and propagated as action potentials (electrical)
At synapses, these electrical signals trigger the release of neurotransmitters, which diffuse across synaptic clefts and bind to receptors on downstream neurons (chemical)
These chemical interactions, in turn, alter the membrane potentials of recipient neurons, generating new electrical activity

This cycle — electrical propagation followed by chemical transmission — repeats across successive layers of the network.

Neural signaling is not purely electrical or chemical, but an ongoing electrochemical process.

Amplification Across Networks

The initial perturbation generated at the receptor level is typically small. Its impact depends on how it is amplified and distributed across neural systems.

Amplification occurs through:

recurrent connectivity within cortical circuits
convergence and divergence of synaptic inputs
modulation by neuromodulatory systems and ongoing activity

As signals propagate, they are not simply transmitted unchanged. They are reshaped by the structure of the network and by the current state of activity. The perturbation becomes a distributed pattern of electrochemical activity spanning multiple neural populations.

This pattern interacts with ongoing dynamics, influencing the trajectory of the system through its state space.

Perturbations as Drivers of Dynamics

Within the attractor framework, these electrochemical cascades constitute the mechanism by which perturbations enter the system.

A stimulus does not impose a predefined representation. Instead, it introduces a structured disturbance into an already active network. The system responds by reorganizing its activity until it reaches a configuration that can be stabilized.

The outcome depends on:

the structure and magnitude of the perturbation
the current state of the system
the constraints imposed by interacting neural subsystems

This leads directly to the core formulation:

Inputs are electrochemical perturbations, not encoded messages.

From this perspective, perception is not the decoding of information, but the stabilization of activity following perturbation. The transition from molecule to mind is therefore continuous: local biochemical events initiate cascades that propagate through networks and ultimately shape global patterns of neural activity.

5. The Biological Basis of Color Signals

To understand how color perception arises, it is necessary to begin with the physiological mechanisms by which light is first detected. At this initial stage, the system does not represent color as a discrete property, but as a pattern of activity across specialized receptor populations.

Cone Photoreceptors and Spectral Sensitivity

Color vision in humans is mediated by three classes of cone photoreceptors located in the retina:

S-cones (short-wavelength), most sensitive to shorter wavelengths of light (perceived as blue)
M-cones (medium-wavelength), most sensitive to intermediate wavelengths (green)
L-cones (long-wavelength), most sensitive to longer wavelengths (red)

Each cone type contains a different photopigment (opsin), which determines its spectral sensitivity. Importantly, these sensitivities are broad and overlapping, not sharply distinct. A given wavelength of light will typically activate more than one cone type to varying degrees.

For example:

light in the “yellow” range activates both L- and M-cones
light in the “blue” range strongly activates S-cones but may also weakly activate others

This means that no single cone type corresponds uniquely to a specific color.

Color as a Pattern of Activation

Because of this overlap, the visual system cannot determine color from the activity of any single receptor. Instead, color is specified by the relative pattern of activation across the three cone types.

At any point in the visual field, the incoming light generates a triplet of signals:

S-cone response
M-cone response
L-cone response

The combination of these responses defines the initial color signal at that location.

This leads to a fundamental principle:

Color begins as a pattern across receptor populations, not as a labeled signal.

Absence of Dedicated “Color Channels”

At this early stage, there are no neurons that correspond to “red,” “blue,” or “green” as discrete entities. The system does not yet assign semantic or categorical meaning to the input. Instead, it maintains a distributed representation in which different wavelengths produce different patterns of activation.

This has two important consequences:

Ambiguity: different spectral inputs can produce similar activation patterns (metamers), meaning the system must interpret rather than directly read out color
Relational encoding: color information is inherently comparative, depending on differences between cone responses rather than absolute values

From Local Signals to System-Level Dynamics

These receptor-level patterns constitute the initial conditions for further processing. They provide the substrate upon which subsequent transformations — such as opponent processing and cortical integration — operate.

Within the dynamical framework developed in this essay, these patterns represent the initial electrochemical perturbations introduced into the system. Their structure shapes how activity propagates and ultimately how attractor states form.

Color perception begins not with discrete signals, but with structured differences across receptor populations that seed the system’s subsequent dynamics.

This establishes the biological foundation upon which the higher-level mechanisms of color perception are built.

6. Opponent Processing: Encoding Difference

The pattern of activity across cone photoreceptors provides the raw substrate for color vision, but it is not yet in a form that the brain can use directly. The next critical transformation is the computation of differences between these signals. This is achieved through a process known as opponent processing.

Opponent Channels

Rather than preserving the three cone signals independently, the visual system reorganizes them into opponent channels. The two principal channels are:

Red–green opponency: comparing the relative activity of L-cones and M-cones
Blue–yellow opponency: comparing S-cone activity against the combined activity of L- and M-cones

In these channels, neurons are not tuned to absolute values, but to contrasts. A neuron in the red–green channel, for example, may increase its activity when L-cone input exceeds M-cone input and decrease its activity in the opposite case.

This means that the system is explicitly representing:

“more red than green”
“more green than red”
“more blue than yellow”
“more yellow than blue”

rather than encoding any color in isolation.

Mechanism: Lateral Inhibition

These comparisons are implemented through lateral inhibition, a fundamental neural mechanism in which the activity of one pathway suppresses the activity of another.

In the retina and early visual pathways:

signals from different cone types are combined
inhibitory interactions subtract one signal from another
the resulting neurons respond to differences between inputs

This process sharpens distinctions and enhances contrast. It also reduces redundancy, allowing the system to efficiently represent variations in the visual input.

From Absolute Signals to Relational Structure

The transformation from cone signals to opponent channels marks a fundamental shift:

from absolute activation (how much each receptor responds)
to relational encoding (how signals differ from one another)

At this stage, color is no longer tied to individual receptors, but to patterns of difference across populations.

The brain does not encode color as a set of independent values, but as structured differences between signals.

Implications for Perception

This difference-based encoding has several important consequences:

it allows the system to remain sensitive to contrast across a wide range of lighting conditions
it provides a compact and efficient representation of color information
it sets the stage for higher-level processes that integrate color with spatial structure, context, and meaning

Within the attractor framework, opponent processing defines the structure of the perturbation entering the system. Instead of raw receptor signals, the system receives a pattern of differences that will be propagated, amplified, and stabilized through network dynamics.

Color perception is therefore grounded in the encoding of differences, not absolute values, providing the relational structure necessary for the formation of stable perceptual states.

7. Distributed Cortical Representation of Color

The transformation of color signals does not end with opponent processing in the retina and early visual pathways. As signals propagate into the cortex, they become embedded within a broader system that integrates multiple features simultaneously. At this stage, color is no longer processed in isolation but as part of a distributed, interacting network.

Cortical Pathways and Regions

Color-related activity is observed across multiple regions of the visual cortex, including:

V1 (primary visual cortex): initial cortical processing, where basic features such as orientation, spatial frequency, and color opponency are represented
V2: intermediate processing, where signals begin to be integrated across larger regions and feature combinations become more complex
V4: strongly associated with color perception, particularly in maintaining color constancy across varying lighting conditions

While V4 plays an important role in color processing, it is not a standalone “color center.” Instead, it participates in a network of regions that collectively contribute to the perception of color.

Feature Integration

As signals move through these cortical areas, they are combined with other visual features. Neurons become responsive not only to color, but to combinations such as:

color + edge orientation
color + motion
color + spatial position

For example, a neuron may respond preferentially to a red vertical edge moving in a particular direction, rather than to “red” alone. This reflects a fundamental principle of cortical organization:

Features are encoded in combination, not in isolation.

This integration allows the brain to construct coherent representations of objects and scenes, in which color is one attribute among many.

Absence of an Isolated “Color Module”

Although certain regions show stronger involvement in color processing, there is no single, isolated module that exclusively represents color. Instead:

color information is distributed across multiple cortical areas
different populations contribute different aspects of the percept
interactions between regions shape the final representation

This distributed architecture means that color perception depends on the coordinated activity of many neural populations rather than on a single locus.

Implications for the Attractor Framework

Within the dynamical framework developed in this essay, this distributed organization has a clear implication.

Color is not represented as a localized signal that can be read out directly. Instead, it emerges from the collective activity of interacting populations embedded within a larger network.

These populations:

carry different components of the signal
interact through structured connectivity
are influenced by both local processing and global context

The resulting percept corresponds to a stable configuration across this distributed system.

Color is represented across distributed, interacting neural populations, whose coordinated activity gives rise to the perceptual experience.

This distributed representation provides the substrate upon which attractor states form, allowing multiple features — including color — to be integrated into a coherent and stable percept.

8. Temporal Coordination and Oscillatory Structure

The spatial and population-based organization of color signals explains how different features are represented across the brain. However, this alone does not explain how multiple signals are coordinated without collapsing into one another. For this, a temporal dimension is required.

Neural activity is not only distributed across space — it is also organized in time. A central feature of this temporal organization is the presence of neural oscillations, rhythmic fluctuations in the excitability of neural populations that can be observed across multiple scales of brain activity.

Neural Oscillations

Neural populations often exhibit oscillatory dynamics across a range of frequencies. Of particular relevance for perception is activity in the gamma band (approximately 30–80 Hz), which has been strongly associated with sensory processing, feature integration, and conscious awareness.

An oscillation can be understood as a repeating cycle consisting of:

a phase of increased excitability, during which neurons are more likely to fire
a phase of decreased excitability, during which firing is suppressed

These cycles are generated through interactions between excitatory and inhibitory neurons and can synchronize activity across distributed regions.

Phase Relationships

Within an oscillatory cycle, not all neurons fire at the same moment. Instead, different neural populations can occupy distinct phase relationships within the same rhythm.

This means that:

some populations fire earlier in the cycle
others fire later
some may align closely in time (synchronous)
others maintain consistent offsets (out of phase)

These phase relationships are not fixed labels but emerge dynamically from the interaction of network structure, connectivity, and ongoing activity.

Synchrony and Separation

Oscillatory dynamics provide a mechanism for achieving both:

synchrony, allowing distributed populations to be coordinated within a shared temporal framework
separation, allowing different signals to remain distinct through differences in timing

If all neural populations were perfectly synchronized, their signals would collapse into a uniform pattern. Conversely, if there were no shared temporal structure, activity would remain fragmented and uncoordinated.

The brain operates between these extremes. Neural populations participate in a common oscillatory framework while maintaining distinct phase relationships that preserve differences between signals.

Temporal Structure in the Attractor Framework

Within the attractor framework, these temporal relationships are not incidental — they are part of the structure that defines an attractor state.

An attractor is characterized not only by:

which populations are active
how strongly they are engaged

but also by:

how their activity is coordinated in time

This temporal coordination allows multiple features — such as different colors — to coexist within a single stable configuration without being averaged or merged.

Key Formulation

Differences are preserved because populations occupy distinct temporal relationships within shared oscillatory dynamics.

This provides a mechanistic solution to the problem introduced earlier. The brain does not preserve differences by isolating signals completely, nor does it integrate them by merging their values. Instead, it organizes them within a structured temporal framework that allows coordination without loss of distinction.

In this way, oscillatory dynamics provide the temporal scaffold upon which stable, differentiated perceptual states can be formed.

9. Attractor Formation in Color Perception

The preceding sections describe the components of color processing: receptor-level patterns, opponent encoding of differences, distributed cortical representation, and temporal coordination through oscillatory dynamics. The final step is to understand how these components come together to produce a stable percept.

At any given moment, multiple neural populations are active simultaneously. These include populations responding to:

different wavelength patterns (red, green, blue, etc.)
different spatial locations across the visual field
additional features such as edges, motion, and contrast

These populations do not operate independently. They interact within a structured network that imposes constraints on how activity can be organized.

Spatiotopic Constraints

One class of constraint arises from the organization of the visual cortex itself. In early visual areas, activity preserves spatial relationships:

neighboring points in the visual field map to neighboring neural populations
edges, boundaries, and spatial continuity are maintained

These spatiotopic constraints ensure that color signals remain anchored to their correct locations within the scene.

Distributed Constraints

A second class of constraint arises from distributed systems that integrate information beyond immediate sensory input. These include:

memory-related processes
contextual interpretation
attentional modulation
expectations about illumination and environment

These distributed constraints influence how signals are weighted, grouped, and interpreted. They allow the system to maintain color constancy, resolve ambiguity, and integrate color into meaningful percepts.

Interaction of Constraints

Color perception emerges from the interaction of these constraint systems.

Spatiotopic systems preserve the structure of the input
Distributed systems reshape that structure according to context and prior knowledge

The resulting activity is not a simple combination of signals. It is a negotiated configuration in which multiple constraints are satisfied simultaneously.

Within this configuration:

different color-related populations remain active
their relative differences are preserved
their interactions are stabilized over time

Stabilization into an Attractor

Through ongoing interactions, the system evolves toward a configuration that can be maintained despite continuous perturbation. This configuration is an attractor state.

In this state:

multiple colors are represented simultaneously
their spatial relationships are preserved
their differences are maintained
their activity is coordinated across the network

This provides a direct account of how perception achieves both differentiation and integration.

Key Statement

A color percept is a stabilized attractor in which differences between neural populations are preserved and coordinated.

Color is therefore not the output of a decoding process. It is the form taken by neural activity when it settles into a stable configuration under the combined influence of sensory input, network structure, and contextual constraints.

This attractor-based view explains how multiple colors can be experienced as distinct elements within a unified perceptual scene, without being reduced to a single averaged signal or fragmented into unrelated components.

10. Perturbation, Absorption, and Attractor Switching

If a color percept corresponds to a stabilized attractor, the next question is how that state is maintained and how it changes. The answer lies in how the system responds to perturbations — the electrochemical disturbances introduced by ongoing sensory input and internal activity.

At any moment, the brain occupies a particular attractor configuration. This configuration reflects a balance of constraints across interacting neural systems. Incoming stimuli do not create a new state from nothing; they perturb this existing configuration.

The effect of a perturbation depends on how it interacts with the constraints shaping the current state. Three regimes can be distinguished:

Absorption

When perturbations are small or compatible with the existing configuration, they are absorbed.

local neural activity adjusts
the overall pattern is preserved
the system returns to the same attractor

Perceptually, this corresponds to stability:

minor changes in lighting
small variations in color intensity
background fluctuations

do not alter the perceived scene.

Deformation

Get Bill Giannakopoulos’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

When perturbations are moderate, they alter the configuration without destabilizing it.

the attractor is reshaped
relative activity patterns shift
the system remains within the same basin of attraction

Perceptually, this corresponds to gradual change:

attention shifting toward a particular color
subtle changes in shading or contrast
adaptation over time

The percept evolves, but its overall structure remains intact.

Transition (Attractor Switching)

When perturbations are large or incompatible with the current configuration, the system can no longer maintain stability.

constraints cannot be jointly satisfied
the current attractor becomes unstable
activity reorganizes into a new configuration

This results in a transition to a different attractor.

Perceptually, this corresponds to qualitative change:

a shift in how a color is perceived under altered illumination
sudden reinterpretation of ambiguous stimuli
changes observed in color illusions

Mechanistic Interpretation

These regimes are not determined solely by the magnitude of the input, but by its compatibility with the current constraint structure. A small but conflicting perturbation can trigger a transition, while a larger but compatible one may be absorbed.

This reflects a fundamental principle:

Stability depends on the ability of the system to accommodate perturbations within its current configuration.

When this is no longer possible, a new configuration is required.

Key Statement

Perception is stabilization after perturbation.

Color perception is therefore not a static readout of sensory input. It is a dynamic process in which the brain continuously absorbs, reshapes, or reorganizes activity in response to perturbations. Stable percepts correspond to attractor states that can accommodate ongoing input, while changes in perception reflect transitions between such states.

11. Color Illusions as Evidence

The attractor-based account of color perception makes a clear prediction: if perception corresponds to the stabilization of activity under interacting constraints, then identical sensory inputs should, under different conditions, give rise to different percepts. Color illusions provide direct empirical support for this claim.

Two well-known examples illustrate this principle.

Checker shadow illusion

In this illusion, two squares on a checkerboard — one in a shadow and one outside it — can have identical physical luminance. Despite this, the square within the shadow is perceived as lighter.

The retinal input at the level of photoreceptors is effectively the same for both squares. However, the broader visual context introduces constraints:

the presence of a shadow
surrounding contrast relationships
assumptions about illumination

The brain incorporates these factors and stabilizes a percept that interprets the shadowed square as reflecting more light than it physically does.

The Dress

In this case, a single image led observers to report two distinct percepts:

blue and black
or white and gold

The physical stimulus — the pixel values reaching the retina — was identical for all observers. The difference arose from how the brain interpreted the lighting conditions of the scene.

Some observers stabilized an attractor consistent with a shadowed environment, leading to one color interpretation. Others stabilized an attractor consistent with bright illumination, leading to another.

Same Input, Different Percepts

These examples demonstrate a crucial point:

the sensory input alone does not determine perception
the same input can lead to different stabilized states

This variability cannot be explained by a simple encoding model in which specific inputs map directly to specific percepts. Instead, it reflects the influence of multiple interacting constraints:

spatiotopic constraints from the visual scene
distributed constraints from context, expectation, and prior experience

These constraints shape how the system evolves and which attractor state is ultimately stabilized.

Context as Constraint

In both illusions, context plays a decisive role. It does not merely modify the interpretation of a fixed signal; it reshapes the space of possible attractors.

in the checker shadow illusion, context alters perceived brightness
in the dress illusion, context alters perceived color

In each case, the system stabilizes a configuration that is internally consistent given the imposed constraints, even if it deviates from the physical properties of the stimulus.

Key Insight

Color illusions show that perception depends on attractor selection, not input alone.

They provide concrete evidence that perceptual states are not determined by the stimulus in isolation, but by the interaction between input and the system’s internal organization. Identical perturbations can lead to different outcomes because the system may settle into different attractor states under different constraint conditions.

Implication for the Framework

Within the attractor framework, color illusions are not anomalies to be explained away. They are expected outcomes of a system that stabilizes activity under competing constraints.

They demonstrate that:

perception is context-dependent
multiple stable configurations are possible for the same input
the brain selects among these configurations through dynamical processes

In this way, color illusions provide strong empirical support for the view that perception is a process of stabilization rather than a direct readout of sensory information.

12. Why Colors Don’t Collapse Into One

A central paradox in color perception is how multiple color signals can be linked into a unified percept without collapsing into a single average. If the brain simply combined inputs in a linear manner, one would expect blending and loss of distinction. Yet in experience, colors remain clearly differentiated while still forming a coherent whole.

The resolution to this paradox lies in how the system is organized.

Distributed Signals

Color information is not represented in a single channel or location. It is distributed across multiple neural populations:

different cone-derived signals
opponent channels
cortical populations tuned to combinations of features

Each population maintains its own activity. A signal corresponding to “red” does not overwrite one corresponding to “green,” nor are they merged into a single scalar value. The components of the system remain physically and functionally distinct.

No Summation into a Single Value

The brain does not bind color by summing signals into one output. Instead:

multiple populations are active simultaneously
their activity is not collapsed into a single variable
no single neuron or region represents the entire percept

This avoids the averaging problem. The system does not reduce multiple signals into a common value; it preserves their separateness at the level of representation.

Stabilization of Relationships

What changes is not the identity of the signals, but how they are related.

Through interactions mediated by connectivity, timing, and constraint structure:

neural populations become coordinated
their relative activity levels stabilize
their temporal relationships become consistent

These relationships define the attractor state. The system settles into a configuration in which:

multiple signals coexist
their differences are maintained
their interactions are structured and stable over time

Unity arises not from merging signals, but from stabilizing how they relate to one another.

Resolution of the Paradox

The paradox dissolves once the assumption of summation is abandoned.

Colors do not collapse because they are not combined into a single value
They remain distinct because they are carried by separate populations
They are unified because their relationships are stabilized within a coherent dynamical state

This allows the system to achieve both differentiation and integration simultaneously.

Key Statement

The brain binds colors by organizing relationships, not averaging signals.

Color perception is therefore the result of a distributed system in which multiple components retain their identity while participating in a coordinated pattern. The percept corresponds to the structure of this pattern, not to the value of any single signal.

13. Clinical and Physiological Implications

Framing color perception as the stabilization of attractor states is not only conceptually useful — it aligns with well-described clinical and physiological observations. Disturbances of perception can be understood as shifts in the underlying dynamical regime of the brain, particularly in how neural populations are coupled and constrained.

Excess Synchrony: Loss of Differentiation

When coupling between neural populations becomes excessively strong, activity can become highly synchronized.

In this regime:

multiple populations lose their distinct dynamics
timing relationships collapse into uniformity
the system occupies a low-dimensional, overly constrained state

Clinically, this is observed in conditions such as generalized seizures, where widespread synchronization is associated with impaired or absent conscious awareness.

From the perspective of color perception:

distinct signals can no longer be maintained
differentiation between features is lost
the system collapses toward uniform activity

This represents a failure of the system to preserve differences within the attractor.

Weak Coupling: Fragmentation

At the opposite extreme, when coupling is too weak, neural populations fail to coordinate effectively.

In this regime:

activity remains local and transient
interactions between populations are insufficient
no stable global pattern can be maintained

This is seen in states such as:

deep sleep
general anesthesia
severe diffuse brain dysfunction

Perceptually, this corresponds to fragmentation:

signals are not integrated into a coherent percept
stable attractor states fail to form

Here, the system cannot sustain the coordinated relationships required for unified perception.

Dysregulated Perception: Hallucinations

Between these extremes, disturbances in how constraints are applied can lead to altered perceptual states, including hallucinations.

In such cases:

internally generated activity can dominate
distributed constraints (memory, expectation, salience) outweigh sensory input
the system stabilizes attractors that are weakly anchored to external stimuli

This can result in:

percepts that are vivid but not stimulus-driven
altered color experiences
context-inappropriate interpretations

From an attractor perspective, hallucinations represent stabilization of internally driven configurations under abnormal constraint balance.

Constraint Balance and Stability

Across these conditions, a consistent principle emerges. Stable perception depends on a balance between:

sufficient coupling to coordinate distributed activity
sufficient differentiation to preserve distinct signals
appropriate weighting of sensory and contextual constraints

Disruption in any of these dimensions alters the system’s ability to form and maintain stable attractor states.

Key Idea

Stable perception depends on balanced coupling and interaction between constraint systems.

Color perception — and perception more broadly — requires the brain to operate within a constrained dynamical regime. Too much coupling leads to collapse; too little leads to fragmentation; imbalance between constraints leads to distortion.

Clinical states provide concrete evidence that perception is not simply a function of input, but of how the system organizes and stabilizes activity under varying physiological conditions.

14. Implications for Neuroscience

The account developed here, using color perception as a concrete example, has broader implications for how perception and brain function are understood.

From Encoding to Stabilization

The traditional view in neuroscience treats perception as the encoding and decoding of information. Sensory inputs are assumed to be transformed into neural representations, which are then read out as experience.

The attractor framework suggests a different formulation:

perception is not the extraction of information from signals
it is the stabilization of activity within a dynamical system

Inputs do not determine percepts directly. They perturb an ongoing system, and the resulting percept corresponds to the attractor state into which the system settles.

Perception is dynamical stabilization, not signal decoding.

Limits of the Coding Metaphor

The coding metaphor implies:

fixed mappings between input and output
stable representations that can, in principle, be read out
independence from system state

However, phenomena such as:

context-dependent perception
variability in response to identical stimuli
color illusions

demonstrate that these assumptions do not hold.

If perception depends on the state of the system and the interaction of multiple constraints, then a simple encoding framework is insufficient to capture its dynamics.

A Shift in What Must Be Measured

If perception corresponds to attractor states, then the relevant variables are not just signals, but relationships.

Understanding brain function requires studying:

coupling
how strongly neural populations influence one another
how this changes across states
timing
phase relationships and oscillatory coordination
how temporal structure organizes activity
attractor transitions
how the system moves between states
what perturbations trigger switching
how stability is gained and lost

These variables define the structure of the dynamical landscape in which perception occurs.

Toward a Dynamical Neuroscience

This perspective suggests a shift in emphasis:

from identifying localized representations
to characterizing distributed patterns of interaction
from static mappings
to evolving trajectories in state space

Rather than asking “what does this neuron encode?”, the more relevant questions become:

what patterns of activity are stable?
how are these patterns coordinated across the system?
how does the system transition between them?

Generalization Beyond Color

Although this essay focuses on color perception, the same principles apply more broadly:

sensory perception
cognition
memory
internally generated states

In each case, experience corresponds to the stabilization of distributed activity under interacting constraints.

Summary

The attractor framework reframes the study of perception and brain function:

perception is a dynamical process
representations are states, not codes
stability and transition are central

By focusing on coupling, timing, and attractor dynamics, neuroscience can move toward a more complete account of how the brain generates coherent experience from complex and variable inputs.

15. Conclusion

The account developed in this essay reframes color perception from the ground up. What begins as photons interacting with molecular photopigments evolves, through a cascade of electrochemical transformations, into large-scale patterns of coordinated neural activity. At no stage does the system construct a simple code to be read out. Instead, it continuously reorganizes itself in response to perturbation.

Color, in this view, is not a property extracted from the world, nor a signal transmitted unchanged through the visual system. It is the form that neural activity takes when it becomes stabilized under constraint.

This perspective resolves several longstanding problems simultaneously.

First, it explains how multiple colors remain distinct without collapsing into an average. Signals are never reduced to a single value; they are maintained across distributed populations whose relationships are stabilized rather than merged.

Second, it explains how unity emerges without loss of differentiation. The percept is not constructed by combining signals, but by organizing their interactions into a coherent pattern.

Third, it accounts for context dependence and variability, as demonstrated by color illusions. The same input can lead to different percepts because the system can stabilize into different attractor states depending on how constraints are balanced.

Fourth, it connects mechanism to experience across scales. Molecular events at the receptor level propagate through electrochemical loops, are reshaped by network structure, coordinated through oscillatory timing, and ultimately stabilized as system-level attractors.

Taken together, these points support a central claim:

Perception is not the decoding of information, but the stabilization of activity within a dynamical system.

Color perception becomes a specific instance of a more general principle: the brain generates experience by settling into structured, metastable states that preserve differences while maintaining coherence. These states are not imposed by the stimulus alone, but emerge from the interaction between input, internal dynamics, and constraint systems distributed across the brain.

This has broader implications for neuroscience. It suggests that understanding perception requires moving beyond the search for static representations and toward the study of:

how neural populations are coupled
how their activity is coordinated in time
how stable patterns form and transition

The relevant object of study is not the signal, but the state of the system.

Within this framework, color is not something the brain reads from the world. It is something the brain becomes — a dynamically stabilized configuration that reflects both the structure of the input and the organization of the system responding to it.

The brain does not encode color. It stabilizes a world in which color can appear.

Appendix: High-Dimensional Nonlinear Systems with Partial Synchrony and Structured Coupling

To formalize the dynamical picture developed throughout this essay, we require a mathematical framework capable of capturing a key property of perception: the ability to maintain distinct components while simultaneously organizing them into a coherent whole. This demands a system in which separation is preserved and integration is achieved through structured interaction rather than collapse.

The most appropriate framework is that of high-dimensional nonlinear dynamical systems with partial synchrony and structured coupling.

High-Dimensional State Space

The brain is not adequately described by a small number of variables. At any given moment, its state is defined by the activity of a large number of interacting neural populations. Formally, this can be represented as a vector:

where each component corresponds to the activity of a distinct population — for example, those involved in processing specific colors, spatial features, or contextual influences.

Crucially, this representation preserves dimensional separation. Each component retains its identity within the overall system.

Nonlinear Dynamics

The evolution of the system over time is governed by nonlinear interactions:

These equations express that the change in each component depends on the state of many others. The system is therefore not additive or reducible to a simple sum of inputs. Instead, it exhibits:

amplification or suppression depending on context
sensitivity to initial conditions
emergence of stable patterns through interaction

Nonlinearity is essential for allowing structured configurations to form without collapsing distinct components into a single value.

Structured Coupling

The interactions between components are not random. They are constrained by the underlying architecture of the system:

synaptic connectivity
corticocortical and corticothalamic pathways
functional specialization across regions

This structure can be represented by a coupling matrix which determines how strongly each component influences another. The pattern of this coupling imposes constraints on which configurations are possible and which are stable.

Through structured coupling:

certain combinations of activity are reinforced
others are suppressed
stable configurations emerge that reflect both input and internal organization

Partial Synchrony

A central requirement for perception is that components are coordinated without becoming identical. This is captured by the concept of partial synchrony.

In such systems:

components become phase-locked or coordinated
but maintain consistent phase offsets

This allows:

integration through shared temporal structure
separation through preserved differences in phase and amplitude

Attractors as Structured Configurations

Within this framework, attractors are not single points or uniform states. They are structured configurations across multiple dimensions:

each component maintains its own level of activity
relationships between components are stabilized
temporal coordination is preserved

These attractors occupy regions of state space rather than isolated points and can support variation within a constrained structure. They are often metastable, persisting for a time before transitioning under perturbation.

Integration Without Collapse

This framework provides a precise mathematical realization of the principle developed throughout the essay:

separation is preserved because components occupy distinct dimensions and retain independent dynamics
integration is achieved because their interactions constrain them into stable relationships

The system does not average signals or merge them into a single value. Instead, it stabilizes a pattern in which multiple components coexist in a coordinated but differentiated manner.

Summary

A high-dimensional nonlinear system with structured coupling and partial synchrony provides the minimal mathematical framework required to capture the dynamics of perception.

Within such a system:

multiple neural populations remain active and distinct
their interactions are constrained by network structure
their activity is coordinated in time
stable attractor states emerge that preserve differences while maintaining coherence

Attractor states in the brain are best understood as structured, high-dimensional configurations in which coordinated dynamics preserve separation while enforcing relational stability.

This framework bridges the gap between the mechanistic processes of neural activity and the phenomenological experience of unified yet differentiated perception.