VISUAL PODCAST (need to revise a bit)

The visual system: core organisation and logic

The eye as the first stage of the brain

Vision does not begin in cortex. A large amount of processing already happens in the eye, which is why the retina is considered neural tissue rather than a simple camera sensor.

Light enters through the cornea and lens, which together focus the image onto the retina, a thin neural layer at the back of the eye. Where the optic nerve exits the eye there are no photoreceptors, producing the natural blind spot.

Retinal circuitry and photoreceptors

The retina contains a layered circuit:

  • Photoreceptors (rods and cones) detect light.

  • Bipolar cells relay signals.

  • Ganglion cells send output to the brain via the optic nerve.

This layered structure already performs substantial computation, including contrast enhancement and data compression, before signals ever reach cortex.

There are two main photoreceptor types:

Rods

  • Extremely light sensitive.

  • Operate mainly in low-light (night) conditions.

  • Do not support colour vision.

  • Dominant in peripheral vision.

Cones

  • Less light sensitive.

  • Support high-acuity and colour vision.

  • Three types, sensitive to long (red), medium (green), and short (blue) wavelengths.

  • Colour perception arises from comparisons between cone types, not from any single cone alone (e.g., red vs green, blue vs yellow).

The fovea and central vision

At the centre of the retina is the fovea, where:

  • Photoreceptors are densely packed.

  • Only cones are present.

  • Visual acuity is highest.

Everything you fixate directly is projected onto the fovea. This explains a classic night-vision phenomenon: faint stars are easier to see when looked at slightly off-centre, because rods dominate outside the fovea and are more sensitive in low light.

Clinical relevance: in conditions where cones are non-functional (e.g., achromatopsia), fixation creates a functional “hole” in vision because the fovea cannot signal.

From eye to brain: hemifield organisation

Visual information is organised by visual field, not by eye.

  • The left visual field projects to the right hemisphere.

  • The right visual field projects to the left hemisphere.

This is achieved by partial crossing at the optic chiasm:

  • Fibres from the nasal retina cross to the opposite hemisphere.

  • Fibres from the temporal retina remain on the same side.

This crossing preserves a clean separation of left vs right visual space in the brain.

Lateral geniculate nucleus (LGN)

After the optic chiasm, signals reach the LGN in the thalamus.

Key features:

  • Acts as a relay, not a passive waypoint.

  • Has six distinct layers.

  • Input from each eye remains segregated.

  • Different layers carry different types of information (e.g., fine detail vs motion-related signals).

This segregation ensures that eye-specific and pathway-specific information remains organised before cortex.

Primary visual cortex (V1): ordered inputs

Signals travel from the LGN to primary visual cortex (V1) via the optic radiations.

In V1:

  • Input arrives mainly in layer 4.

  • Information from the two eyes remains separated in ocular dominance columns.

  • V1 has a six-layer cortical structure, mirroring the highly organised input it receives.

Because of these alternating eye-specific stripes, V1 is also called striate cortex.

Retinotopic maps

V1 contains a retinotopic map, meaning neighbouring points on the retina project to neighbouring points in cortex.

Two key dimensions are preserved:

  • Eccentricity: centre of gaze is over-represented; peripheral vision occupies less cortical space.

  • Polar angle: upper vs lower and left vs right visual field locations map systematically across cortex.

This mapping means the visual world is laid out across cortex like a distorted but continuous projection.

Vision beyond V1: distributed mapping

The brain contains many copies of the visual field, not just one.

Multiple visual areas beyond V1 each maintain retinotopic organisation, forming a highly structured visual scaffold across the brain. These maps are used to:

  • Define boundaries between visual areas.

  • Support specialised processing while preserving spatial layout.

There is no “little observer” inside the brain looking at these maps. Instead, they allow visual information to remain usable for perception, attention, eye movements, and action.

Two broad visual pathways

From early visual cortex, processing broadly diverges into two interacting streams:

Ventral pathway (occipito-temporal)

  • Processes colour, texture, shape, and object identity.

  • Often described as the “what” pathway.

  • Closely linked to memory and recognition.

Dorsal pathway (occipito-parietal)

  • Processes spatial location, motion, and visuomotor transformations.

  • Often described as the “where/how” pathway.

  • Critical for guiding action (e.g., reaching, grasping).

Important correction: these streams are not independent. There is extensive interaction between them, especially during development.

Development of vision and how it can go wrong (and why cognition cares)

Why vision affects “non-visual” development

Visual impairment in infancy doesn’t only reduce visual detail; it changes access to information in the world. That can cascade into broader developmental delays. A common pressure point is social development, because faces, gaze, and subtle expressions are a huge part of early learning—if those signals are hard to read, social learning gets harder.

A useful framing: some behaviours that look similar on the surface (e.g., reduced social cue use) can come from different underlying causes—either atypical social-cognitive processing, or simply reduced access to visual social information.

What’s already organised before birth

Eye-to-brain wiring is strongly pre-specified

Key layout decisions are largely set up prenatally:

  • The optic chiasm crossing pattern (which fibres cross vs stay).

  • Retinotopic organisation (neighbourhood structure preserved from retina → brain).

This organisation is guided by chemical gradients that help axons choose paths and maintain orderly maps.

When this goes wrong: albinism as a wiring example

In albinism, early retinal patterning and signalling are altered, which can disrupt normal guidance. Result: abnormal crossing at the optic chiasm.

Big consequence:

  • Normally: left visual field → right hemisphere, right visual field → left hemisphere.

  • With abnormal crossing: representations can become overlaid, so one hemisphere may receive information corresponding to both sides of space.

Yet people don’t usually experience double vision, implying the brain applies compensations (suppression, reformatting, or downstream separation). This is an extreme example of neuroplasticity adapting to a highly unusual input format.

Experience before eyes open: retinal waves

Retinal waves as “training data”

Even before true seeing, the retina produces systematic waves of activity. These create structured, correlated signals that help the developing system wire itself up sensibly.

Why they matter: building eye-specific organisation

These waves support eye-specific separation in early relay structures (thalamus/LGN). If retinal waves are disrupted, that layered/segregated organisation becomes less well formed.

So it’s not “genes vs experience.” It’s:
chemical scaffolding + internally generated patterned activity + later external visual input.

Retinotopy persists even without sight

Even in congenital blindness, V1 can still show organisation that resembles central vs peripheral structure in resting brain activity. Meaning: a lot of the brain’s visual-map layout is laid down early and is robust, even without typical visual experience.

Why this early structure matters for cognition

The infant brain is not a blank slate. Early visual wiring imposes a highly organised scaffold (retinotopic mapping and structured connectivity) across cortex.

Working idea: because later learning builds on this scaffold, it can influence where cognitive specialisations develop and how efficiently different functions emerge. That’s the bridge from “wiring of vision” to “development of the mind.”

Photoreceptors mature slowly: why early vision is blurry

Photoreceptors are immature at birth

Photoreceptors (especially cones) are structurally immature and take years to become adult-like (around ~4 years is often used as a rough benchmark). A key limiting factor is the outer segment: when it’s short, light capture is inefficient.

Why acuity is poor early on

Immature (“stubby”) photoreceptors are larger and packed in a way that reduces sampling resolution, so fine detail is hard to resolve.

Acuity improves dramatically across infancy and early childhood (e.g., big changes from early months through the first couple of years). That matters for cognition because degraded early input doesn’t just remove detail—it changes the quality of information feeding systems for objects, faces, and spatial learning.

Measuring children’s vision is harder than it sounds

Adult standards don’t transfer neatly to infants/young children. “Normal” vision varies by age and test, so interpreting children’s performance (in school or clinic) requires age-appropriate assumptions.

Contrast sensitivity function: more informative than acuity alone

Acuity focuses on the smallest resolvable detail (high spatial frequency). Contrast sensitivity measures detectability across a range of spatial frequencies:

  • typically worst at very low and very high frequencies

  • best in the middle range

  • develops across childhood

Practical implication: a child might seem “okay” on an acuity-like measure but still struggle in low-contrast, cluttered real scenes.

Coarse-to-fine development: bug or feature?

A modern computational angle: early low-resolution vision might be useful, not just a limitation.

When an artificial system is trained first on blurred inputs and later on sharper inputs, it can become more robust across different blur levels than systems trained in other orders. Working interpretation: starting coarse may force learning of stable global structure before the system locks onto fragile fine detail.

Amblyopia (“lazy eye”): a canonical critical-period example

What it is and why it happens

Amblyopia is common (around 2–5%) and arises when early input is unbalanced between the eyes, such as:

  • strabismus (misalignment)

  • anisometropia (one eye chronically out of focus)

  • deprivation (e.g., cataract)

Often comes with reduced binocularity / stereopsis.

What happens in the brain

With monocular deprivation early in life, V1 shifts responsiveness and territory toward the stronger/open eye. The “binocular middle ground” becomes underdeveloped.

Why timing matters: “reverse” plasticity

During a sensitive window, ocular dominance can shift quickly if you change which eye is deprived. That’s the basis for patching: penalise the dominant eye to strengthen the weaker one. The catch is timing—plasticity is much more limited later.

Beyond balancing acuity: restoring cooperation

Balancing acuity isn’t always the same as restoring true binocular integration. Newer approaches try to keep both eyes engaged while rebalancing input, aiming to support cooperative functions like depth processing.

Binocular depth perception: a fast-developing window

What stereopsis requires

Stereo depth uses binocular disparity (small left-right differences) to infer depth. A noticeable minority of people have poor stereopsis, even if they can use other depth cues (motion, texture, perspective).

When it emerges

Stereo sensitivity shows a rapid developmental “switch-on” over a short window (roughly 2.5–5 months), not a slow gradual climb. That makes early alignment/balance especially important.

What triggers onset: experience vs maturation

Comparing preterm and full-term infants suggests onset is driven largely by postnatal visual experience (timed from birth), rather than simply unfolding on a gestational biological timer.

Take-home synthesis (development → cognition → risk)

Early visual development combines:

  • built-in chemical scaffolding (maps, crossings)

  • internally generated activity (retinal waves) refining organisation

  • postnatal input tuning acuity and binocular integration

This scaffold supports later cognitive specialisation, but it also creates sensitive windows. If input is degraded or imbalanced during those windows, the system can reorganise in ways that become harder to fully reverse—affecting learning and social development, not just visual clarity.

Colour perception develops slowly (and baby visuals can be accidentally terrible)

People often assume babies can’t see colour. Not true. Babies can detect colour contrast from birth, but it’s limited and low-quality, so what’s “visually obvious” to adults often isn’t to them.

  • Early colour ability: red–green contrast is available early, but overall colour vision is weak.

  • Blue–yellow sensitivity comes online later (around ~2 months).

  • Colour “pop-out” (colour grabbing attention instantly in clutter) emerges later (around ~5 months), likely because broader contrast sensitivity is still developing.

Practical implication: designing baby-appropriate visuals

  • Colour images often have lower luminance contrast than black-and-white, so babies may still prefer high-contrast patterns.

  • But not all black-and-white is helpful: high-frequency, busy monochrome patterns can create distracting edges that hide object boundaries. They may look “stimulating” to adults, but they can be genuinely hard for babies to parse as shapes/objects.

  • Pastel, low-saturation palettes may look stylish to adults but can function like “weak grey soup” to babies because colour contrast sensitivity is still immature.

  • High-saturation, high-contrast “circus colours” are often more perceptually accessible early on.

Colour constancy: the brain’s lighting-correction trick

Adults experience colours as relatively stable across lighting. That stability is an achievement: the brain uses context to infer illumination and “discount” it.

  • Colour constancy = recognising that an object’s colour stays the same even when lighting changes (bluish shade vs yellowish indoor light).

  • Young children are less reliable at this, so the same object under different lighting can look like different colours to them.

  • This has knock-on effects: colour becomes a less reliable sorting cue early in development.

Link to language (developmental coupling)

  • Children can often discriminate colours before they can label them reliably.

  • Colour constancy development is correlated with language development timing, suggesting language may help stabilise colour categories (not proven as the only cause, but it fits the pattern).

“Poor colour early” might be a feature, not a bug

There’s a recurring theme: early perceptual limitations might guide learning in a helpful order.

Evidence from late sight restoration (cataract removal)

People who had cataracts removed relatively late can show:

  • surprisingly good basic colour discrimination soon after treatment,

  • but they may rely on colour too much for object recognition compared with typical controls.

Interpretation: if you start learning objects with atypically strong/available colour cues, you might over-weight colour instead of learning robust shape/structure.

Parallel evidence from AI models

Artificial vision systems trained in a “more typical” order (less colour / greyscale-like first, then colour) can end up more robust when colour is removed or changes. Same developmental logic: early constraints may prevent overfitting to colour and push learning toward stable structure.

Object perception: early abilities, but “real-world robustness” takes years

Early shape discrimination

Soon after birth, babies can tell apart basic shapes. By ~3 months, they begin forming visual categories (e.g., treating multiple cats as “same kind of thing”).

But early categorisation is heavily visual-statistics-driven, not conceptual.

  • Example pattern: after seeing many cats, a dog stands out as “new,” but after seeing many dogs, a cat may not stand out as strongly—because the visual variability of dogs is larger and can “contain” the cat space. Early categorisation tracks the geometry of visual features.

Object permanence

Understanding that objects persist when hidden (“still there behind the occluder”) emerges around 8–9 months in classic behavioural measures (with some evidence that eye-movement measures can show earlier sensitivity).

The long haul: objects in clutter/noise

Recognising objects when they’re degraded, ambiguous, or embedded in clutter develops slowly, continuing for years.

  • Two-tone (“Mooney”) images are a great example: once you know the object, you can’t unsee it—but young children often can’t reliably extract the structure and get pulled around by misleading edges.

  • This kind of perceptual grouping is also hard for many AI systems, hinting it’s a genuinely difficult computational problem.

Face perception: early preferences, late expertise

Newborn “face preference” isn’t full face understanding

Newborns tend to prefer face-like patterns, but this can be explained by low-level biases (top-heavy configurations). They also like things like car fronts for similar reasons.

Familiar faces emerge early, fine-grained face skills emerge late

  • By ~4 months, infants start showing preferences for familiar face categories (e.g., caregiver-related familiarity).

  • But subtle, adult-like face abilities (recognising individuals across changes, dealing with crowding/clutter, sensitivity to configuration changes) continue developing into later childhood.

Brain-wise, face-selective responses (e.g., regions like the fusiform face area) tend to increase in size/strength with age, and measurement can be variable in younger children.

Why specialised visual regions end up where they are

A neat organising principle: high-level specialisations seem to “sit on top of” low-level visual biases.

  • Regions involved in faces tend to be located near cortex biased toward central (foveal) vision—consistent with the fact that we typically fixate faces.

  • Regions involved in places/scenes tend to be located near cortex biased toward peripheral vision—consistent with scenes spanning the visual periphery.

This suggests viewing behaviour and retinal statistics help shape where category selectivity develops.

Experience shapes specialisation (Pokémon is the funniest evidence)

Expertise can create or strengthen category-selective patches.

  • People with intense experience visually discriminating a category presented in a consistent way (e.g., small, foveal items in games) can show specialised cortical responses for that category.

Translation: the cortex is opportunistic—it will allocate real estate to what you train it on, especially if the training has consistent visual demands.

But specialisation is not purely visual: it can emerge without sight

A mind-bending finding: category gradients (e.g., living vs non-living distinctions in ventral regions) can appear even in people blind from birth when they process sounds of animals/tools.

That implies some combination of:

  • built-in organisational constraints,

  • cross-modal connectivity,

  • and non-visual learning signals

…helps structure these “visual” regions. Vision shapes them, but doesn’t fully invent them from scratch.

Take-home synthesis: how development interacts with cognition (and failure modes)

  1. Early limits (poor colour stability, low contrast sensitivity, immature grouping) shape what information is available for learning.

  2. The brain builds representations that match the statistics of the input it gets (and the actions it performs—fixation patterns matter).

  3. When input is atypical (late sight restoration, unusual early deprivation, altered experiences), perception can still develop—but the system may weight cues differently (e.g., over-reliance on colour), which can affect robustness and downstream cognition (social recognition, learning in clutter, categorisation).

What “visual function” includes

Vision isn’t one ability. It can be broken into different “senses”:

  • Light sense (basic detection)

  • Form sense (detail/shape)

  • Colour sense

  • Depth sense

  • Motion sense

Visual acuity (VA) is a form sense measure: the sharpness/clarity of vision—how well someone can discriminate fine detail. It’s the most commonly used clinical index of visual function.

Visual acuity in practice: why children are hard to test well

Adult-style charts don’t translate cleanly

Adult acuity is typically measured with letter charts (e.g., Snellen/LogMAR style formats). The problem is that this becomes messy fast in children because:

  • many children can’t read yet,

  • charts are visually and cognitively demanding (knowing where to look, understanding the task, staying engaged),

  • performance can reflect attention/confusion as much as vision.

Preferential-looking methods for pre-verbal children

For infants/toddlers, a common approach is “do they look at it?”

  • A clinician presents striped gratings (or similar patterns) and judges whether the child’s gaze indicates detection.

  • This is clever, but it’s also fragile: it depends on the examiner’s skill, and it’s easy to accidentally change viewing distance.

Why distance matters: if the card/screen is closer, the pattern becomes bigger on the retina, so it’s easier to see. Keeping distance constant while also watching gaze is genuinely hard, which adds noise to the measurement.

Test variability is a real clinical problem

Even with training, acuity measures in young children can be variable. That matters because:

  • small changes can be clinically meaningful (progression or treatment effects),

  • but unreliable measures make small changes hard to detect confidently.

Different tests aren’t always comparable

There are multiple child-friendly formats:

  • modified letter tests (e.g., simplified layouts)

  • picture-based tests for pre-literate children

  • isolated symbols vs crowded arrays

Because these tests differ in format and difficulty, they’re often not directly interchangeable, which complicates tracking development over time or comparing results across clinics.

Why this matters: sensitive periods and knock-on effects

Measuring child vision well is important because treatable visual problems can cascade into learning, cognitive, and social development (less effective access to information, less efficient learning signals).

A key developmental claim here:

  • the sensitive (“critical/plastic”) period for acuity development spans roughly birth to ~8 years

  • some change may continue beyond that, but treatment is far more effective when delivered within this window

Translation: early detection isn’t just helpful—it can be the difference between “fixable” and “stuck with it.”

Newer approaches: making acuity measurement more objective

Eye-tracking based detection (research approach)

Instead of relying on examiner judgement, an eye-tracker can quantify:

  • where the child is looking

  • whether they orient toward a grating presented slightly off fixation

Logic: if the pattern is detectable, gaze should shift toward it. This reduces subjectivity, but it’s not yet standard in everyday clinics.

Tablet-based interactive tests (more scalable)

An emerging practical approach uses child-friendly tasks on tablets:

  • patterns (“targets”) appear on-screen

  • toddlers can interact (e.g., touch/drag/pop the target)

Advantages:

  • engaging for children

  • can approximate contrast sensitivity or acuity-like measures

  • tablets can use the built-in camera to estimate viewing distance, helping keep retinal image size consistent (a core technical headache in child testing)

Brain-based methods: electrophysiology for objective detection

Steady-state visually evoked potentials (SSVEPs)

This is the “your brain can’t help but sync” method:

  • show a flickering pattern at a fixed frequency (e.g., ~15 Hz black/white reversal)

  • if the visual system detects it, the brain response shows a clear signal at that frequency

This is useful because it can provide an objective readout even when:

  • the child can’t do instructions well,

  • behavioural responses are inconsistent.

Gene therapy case: why objective tests matter (and what they show)

The clinical context

A severe early-onset retinal condition can lead to rapid deterioration—sometimes to near-total blindness by early childhood. That creates a narrow window where treating early is crucial.

What objective measures show after treatment

Using objective tools like:

  • EEG/SSVEP responses to fine patterns

  • child-friendly grating-detection games (showing better detection in the treated eye)

…you can demonstrate that:

  • the treated eye supports stronger pattern detection

  • functional vision improves (not perfect, but meaningfully better)

What “meaningfully better” looks like

The practical impact isn’t subtle: the difference can be between profound darkness by around early childhood versus being able to access colours, letters, and faces to some extent.

And the reason this section exists at all: if you want therapies to be adopted widely, you need reliable, objective outcome measures, not just “it seems better.”

Take-home picture

Child vision assessment is hard because it’s a moving target: development is rapid, behaviour is unreliable, and classic adult tools don’t fit. The field is shifting toward objective, scalable measures (eye tracking, tablet-based tasks, EEG), because treatment decisions—and proof that treatments work—depend on measurement quality.

Visual Development — Revision Notes

1. Ocular and neural structures of the visual system

Eye and retina

  • Cornea + lens focus light onto the retina.

  • Retina is neural tissue, not a camera sensor.

  • Photoreceptors:

    • Rods: very light-sensitive, night vision, no colour, peripheral vision.

    • Cones: colour and high acuity; three types (L/M/S ≈ red/green/blue).

  • Fovea:

    • High cone density, highest acuity.

    • No rods → poor night vision at fixation.

  • Retinal circuitry (photoreceptors → bipolar → ganglion cells) already performs substantial computation (contrast, compression).

Post-retinal pathways

  • Optic nerve → optic chiasm:

    • Nasal retinal fibres cross; temporal fibres stay ipsilateral.

    • Preserves left visual field → right hemisphere, and vice versa.

  • Lateral geniculate nucleus (LGN):

    • Thalamic relay with layered, eye-specific input.

  • Primary visual cortex (V1):

    • Input mainly to layer 4.

    • Ocular dominance columns (left/right eye kept separate).

    • Retinotopic maps preserve spatial layout of the visual field.

Beyond V1

  • Multiple retinotopic maps across cortex.

  • Two interacting pathways:

    • Ventral (“what”): colour, shape, object identity.

    • Dorsal (“where/how”): spatial processing, action, visuomotor control.

  • High-level specialisations (faces, places, objects) build on low-level visual biases (foveal vs peripheral).

2. Developmental trajectories of main visual processes

Prenatal and early postnatal organisation

  • Eye-to-brain wiring is strongly pre-specified by chemical guidance cues.

  • Retinal waves (before eye opening):

    • Provide structured activity.

    • Refine retinotopy and eye-specific segregation.

  • Retinotopic organisation persists even in congenital blindness → early scaffold is robust.

Visual acuity and contrast

  • Photoreceptors are immature at birth (short outer segments).

  • Early vision is low acuity and low contrast sensitivity.

  • Acuity improves rapidly in infancy, continues into early childhood.

  • Sensitive (“critical”) period for acuity: roughly birth–8 years.

Colour perception

  • Colour perception present from birth but very poor.

  • Red–green sensitivity develops earlier than blue–yellow.

  • Colour pop-out and reliable colour use emerge later.

  • Colour constancy (stable colour across lighting) develops slowly and is linked to language development.

  • Early weak colour vision may scaffold more robust object recognition.

Binocular vision and depth

  • Inputs from the two eyes initially segregated.

  • Stereopsis (binocular depth) emerges rapidly around 2.5–5 months.

  • Driven mainly by postnatal visual experience, not gestational age.

  • Sensitive to misalignment or deprivation during this short window.

Object and face perception

  • Newborns discriminate basic shapes.

  • By ~3 months: visual categorisation based on feature statistics.

  • Object permanence emerges around 8–9 months.

  • Recognition in clutter/noise develops slowly (years).

  • Face preference at birth is low-level (configuration-based).

  • Adult-like face recognition and face-selective brain regions develop into late childhood.

When development goes wrong

  • Amblyopia: unbalanced eye input → cortical reorganisation.

  • Critical periods determine reversibility.

  • Late or abnormal input (e.g. cataracts, retinal disease) alters cue weighting (e.g. over-reliance on colour).

3. Assessing visual development

Behavioural measures

  • Visual acuity (form sense):

    • Adult charts (Snellen / LogMAR) unsuitable for many children.

  • Preferential looking:

    • Examiner judges gaze toward patterns.

    • Useful but subjective and variable.

  • Picture / symbol tests:

    • For pre-literate children.

    • Different formats not always comparable.

Limitations of traditional testing

  • High variability in young children.

  • Small but meaningful changes can be missed.

  • Performance reflects attention and understanding as well as vision.

More objective and modern approaches

  • Eye-tracking:

    • Measures orienting responses to visual targets.

    • Reduces examiner bias.

  • Tablet-based interactive tests:

    • Child-friendly, engaging.

    • Can estimate viewing distance via camera.

  • Contrast sensitivity function (CSF):

    • Measures vision across spatial frequencies, not just fine detail.

    • More informative than acuity alone.

Brain-based measures

  • Steady-state visually evoked potentials (SSVEPs):

    • Flickering stimuli evoke frequency-locked brain responses.

    • Objective, works even when behavioural responses are unreliable.

Clinical relevance

  • Early detection is critical due to sensitive periods.

  • Objective measures are essential for evaluating interventions (e.g. patching, binocular therapies, gene therapy).

  • Improvements can be functional and life-changing even if vision is not “normal”.