Introduction to Behavioral Neuroscience · Vision
Unsolved Questions in Visual Perception
On this page 9 sections
In 30 seconds
You now know a great deal about vision: the retina computes contrast, V1 detects oriented edges, the ventral stream recognizes objects, and the dorsal stream guides action. This topic turns the telescope around: what do we not know? How do separate features (color, shape, motion) get bound into one object? How do we recognize a chair across every angle, size, and lighting? Why do we see things that are not physically there (illusions), and why does the blind spot not look like a hole? And the deepest one: what makes a visual experience conscious?
Studying unsolved questions is not an afterthought — it is where the interesting science lives. Each open question has competing hypotheses, clever experiments, and real clinical and technological stakes (brain–computer interfaces, prosthetic vision, machine vision).
Why this matters
- It frames the whole chapter. The solved parts of vision are steps toward answering these questions; knowing what is unsolved keeps you from overclaiming what the earlier models explain.
- Illusions and Filling-in The brain completing missing visual information (e.g., blind spot) Full entry → are diagnostic tools. They reveal the assumptions baked into the visual system. Understanding them improves scientific reasoning and explains everyday experiences (why the blind spot is invisible, why the moon looks huge on the horizon).
- Blindsight Residual visual ability without awareness after V1 damage Full entry → and related phenomena have clinical weight. They show that vision can operate without awareness — informing rehabilitation, brain–computer interfaces, and how we think about patients with cortical damage.
- It is a model for how science progresses. Competing theories (bottom-up feature detection vs. Predictive coding Theory that the brain predicts input and processes prediction error Full entry →; grandmother cells vs. population codes) are testable in principle, and experiments decide. That is exactly the kind of reasoning exams and research demand.
The college version
Core Concepts
The binding problem
Features are processed in parallel: color in one area, motion in another, shape in a third. So how does the brain glue "red," "round," and "rolling toward me" into a single coherent apple? This is the Binding problem The open question of how separate features become one object Full entry →. Proposed solutions include:
- Synchronous firing: neurons processing the same object fire in temporal synchrony, binding features by timing.
- Convergence: downstream neurons (e.g., in IT cortex) pool the feature-specific outputs, binding by anatomy.
- Attention: focusing attention on a location selects which features get bound (suggested by binding errors — in rapid displays, people occasionally report a red square when shown a red circle and a blue square, as if features were miscombined).
No account fully explains binding, and it remains one of the deepest open questions in perception.
Invariance: recognizing objects across changes
You recognize your friend's face in profile, in shadow, at three feet or thirty. A camera, by contrast, sees every one of those as a different image. How does the visual system achieve Invariance Recognizing objects despite changes in size, angle, lighting Full entry → (stable recognition across changes in size, viewpoint, lighting, and occlusion)? Competing frameworks:
- View-based recognition: the system stores many views of an object and matches the closest one.
- Structural description: the system encodes an object's parts and their spatial relations (e.g., face = two eyes above a nose above a mouth), which is more robust to viewpoint changes.
- Population codes vs. sparse codes: whether objects are represented by many broadly tuned neurons or a few highly selective "grandmother cells." Evidence suggests a middle ground: sparse, partially selective populations (e.g., face patches with neurons responding to particular faces).
The neural code: what the spikes mean
Even the "language" of the visual system is debated. Some neurons clearly encode stimulus features in their firing rate (more spikes = stronger feature). But rate cannot easily explain the speed of recognition (humans recognize images in a few hundred milliseconds). Alternatives include temporal codes (the precise timing of spikes carries information) and population codes (the pattern across many neurons matters, not any single cell). Modern recordings support a combination, but the "real" neural code — if there is a single one — is unresolved.
Unconscious vision: blindsight and the limits of awareness
Patients with damage to V1 are classically blind in the corresponding field — yet some (blindsight) can still guess the location, motion, or even the emotion of stimuli they deny seeing. This shows that some visual processing reaches action and emotion systems without awareness, presumably via pathways that bypass V1 (e.g., through the superior colliculus and pulvinar). Blindsight raises the question of what, precisely, makes a visual signal conscious — the same question behind the search for the neural correlates of consciousness (NCC).
Filling-in and illusions: perception as inference
You have a blind spot where the optic nerve exits, yet you never see a hole. Your brain fills in the missing region from the surrounding context. Similarly, color perception fills in across the visual field, and illusions like the Kanizsa triangle create edges where none exist. These phenomena suggest perception is inference: the brain combines sensory evidence with prior expectations and constructs the most plausible scene. This is the core of the predictive coding framework, in which the brain constantly predicts its sensory input and processes mainly the prediction error. Predictive coding is a leading (but not settled) explanation for illusions, fill-in, and the speed of perception.
Consciousness and the "hard problem"
The most famous unsolved question: why do some visual processes come with a subjective experience — the redness of red — at all? Philosophers call this the "Hard problem Why subjective experience exists at all Full entry →" of consciousness, distinct from the "easy" problems of explaining behavior and function. Neuroscience has made progress on the correlates of consciousness (which brain states accompany awareness) but has no accepted theory of why experience exists. Competing views (global workspace theory, integrated information theory, predictive processing accounts) all make testable predictions — a genuinely open, active research frontier.
Common Confusions
| Do Not Confuse | With | Difference |
|---|---|---|
| The blind spot and a scotoma | Each other | The blind spot is a normal anatomical gap (optic nerve head) filled in by the brain; a scotoma is a pathological field loss (e.g., from V1 damage) |
| "Unsolved" meaning "nothing is known" | "Unsolved" meaning "competing accounts exist" | There are many well-tested findings and several live hypotheses; the explanations are unsettled |
| Filling-in being a failure | Filling-in being an efficient strategy | The brain completes missing input because it usually matches reality — a feature, not a bug |
| Blindsight being "seeing without eyes" | Blindsight being vision without awareness | The eyes and early processing work; conscious perception is missing after V1 damage |
| Predictive coding being settled theory | Predictive coding being a leading hypothesis | It explains many phenomena but remains actively debated and tested |
| The neural code being one thing | There being candidate codes | Rate, timing, and population codes all carry information; the question is which the brain actually uses |
| Consciousness research being purely philosophical | It having empirical correlates | NCC research measures brain states accompanying experience; only the existence of experience is the "hard" part |
Eli explains
The same idea, in plain words
Explain it like I’m 10
Your brain is like a detective who never gets the whole story at once — it gets a red clue, a round clue, and a moving clue, and has to figure out they belong to the same apple. It also guesses ahead: when part of the picture is missing (like your blind spot), it calmly fills in what it expects to be there. Scientists know a lot about the clues, but they still don't know exactly how the detective puts them together — or why putting them together feels like seeing.
Worked example
Close your left eye and stare at the dot on the left of this mark with your right eye: • ◦ — the circle on the right disappears when its image lands on your optic nerve head, where there are no photoreceptors. Now the interesting part: you do not see a dark hole. The brain fills the region using the texture and color of the surrounding background — in this case, it "invents" white where the circle was. The same constructive process explains why a gray patch on a red field takes on a greenish tint (simultaneous contrast) and why the Kanizsa triangle's edges appear.
Now scale the idea up: if perception is always this constructive, then everyday vision is a running set of inferences, not a faithful recording. That is why a confident eyewitness can be wrong, why change blindness lets us miss obvious alterations, and why two people can look at the same scene and "see" different things. The lesson for studying perception: your visual experience is the brain's best hypothesis, updated by evidence — never the raw image itself.
Key takeaways
- Binding problem: how parallel features (color, shape, motion) become one object — candidate mechanisms: synchrony, convergence, attention.
- Invariance: recognition across viewpoint/size/lighting; competing accounts: view-based vs. structural description vs. population codes.
- Neural code debate: rate codes vs. temporal codes vs. population codes; recognition is fast (hundreds of ms), which rate codes struggle to explain.
- Blindsight: V1 damage → no awareness, yet residual ability to locate/respond to stimuli via pathways that bypass V1.
- Filling-in: the blind spot is invisible because the brain completes the scene — perception is constructive, not a camera.
- Predictive coding: the brain predicts input and processes prediction error; a leading framework for illusions and fill-in (not settled).
- NCC / hard problem: correlates of consciousness are studied; why experience exists at all remains unresolved.
- Exam takeaway: distinguish what is known (feature maps, streams, receptive fields) from what is debated (binding, code, consciousness) — a common essay/SAQ distinction.
Check yourself
6 review questions from the chapter. Try each one, then open the answer.
State the binding problem in one or two sentences, and name two candidate solutions.
Show answer
Features such as color, shape, and motion are processed in separate brain areas, so how do they become one unified object percept? Candidate solutions include synchronous firing of feature-selective neurons, convergence onto downstream object neurons, and attention-based feature selection.
What is blindsight, and why is it evidence for unconscious visual processing?
Show answer
Patients with V1 damage deny seeing stimuli in the affected field yet can still guess their location, direction, or emotional expression. Because some processing continues (via pathways that bypass V1), vision can operate without awareness — so awareness is not required for all visual function.
Why does the blind spot not appear as a hole, and what does this reveal about perception?
Show answer
The brain fills in the missing region using surrounding texture and color, and you never notice the gap. This reveals that perception is constructive/inferential — the brain completes and predicts the scene rather than passively recording it.
What is the difference between the "easy" problems and the "hard problem" of consciousness?
Show answer
The "easy" problems are explaining how the brain performs functions (detecting edges, binding features). The hard problem is explaining why any of this is accompanied by subjective experience — why there is "something it is like" to see. Neuroscience addresses the easy problems and the correlates; the hard problem remains philosophically and empirically open.
What is predictive coding, and what phenomenon does it explain well?
Show answer
Predictive coding says the brain constantly predicts its sensory input and processes mainly the difference between prediction and input (prediction error). It explains illusions, filling-in, the speed of perception, and why context shapes what we see.
Why do rate codes alone struggle to explain how fast humans recognize objects?
Show answer
Recognition happens in a few hundred milliseconds — too fast for many neurons to integrate and average enough spikes into a rate. Precise spike timing and population patterns can carry information much faster, which motivates temporal and population-code accounts.
Study tools & related lessonsKey vocabulary · Related
Key vocabulary
- Binding problem
- The open question of how separate features become one object
- Invariance
- Recognizing objects despite changes in size, angle, lighting
- Population code
- Information carried by the pattern of activity across many neurons
- Blindsight
- Residual visual ability without awareness after V1 damage
- Filling-in
- The brain completing missing visual information (e.g., blind spot)
- Predictive coding
- Theory that the brain predicts input and processes prediction error
- Neural correlate of consciousness (NCC)
- Brain activity that tracks what is consciously experienced
- Hard problem
- Why subjective experience exists at all
- Grandmother cell
- A hypothetical neuron responding to one specific concept
- Illusion
- A percept that differs from physical stimulus
Sources & references
This lesson was adapted from the open educational references above; their licenses and attributions are preserved. See Copyright & Licensing.
Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.