Part IV · Chapter 12 · Lecture 9
Motion
Why the brain infers motion rather than detecting it, the four problems motion solves, area MT/V5, and the four kinds of evidence — one from each method family — that motion is coded there.
1. Motion at Marr's three levels
| Level | Motion perception |
|---|---|
| Computational | Distinguish depth relations, recognize objects, grab attention in dangerous situations, interact with the environment (e.g. landing a plane) |
| Algorithmic | Compare visual information — changes in position and changes in contrast — over time to estimate direction and speed; use optic flow to infer our own movement through the scene |
| Implementation | V5, a.k.a. the middle temporal area (MT), with a column structure tuned to specific directions of motion |
2. Motion is inferred, not recorded
Definition
Motion: the visual system infers motion from the changing pattern of light in the
retinal image. There's no "motion receptor" in the retina — only light at places and times.
The proof that it's an inference: you can get the inference without any motion at all.
Apparent motion — two dots that never move
What to notice: nothing moves — one dot switches off, the other switches on. At short
gaps you see one dot jumping smoothly back and forth. Push the gap near 0 and you see two dots
flickering together; push it past ~300 ms and you see two separate dots taking turns. Motion appears only in a
band of timings: the brain's comparison of position over time. (Your own mileage may vary with screen
refresh.)
Apparent motion
An illusion in which stationary objects viewed in quick succession (or next to moving objects) appear to move.
Every film and screen you've watched relies on it. Related: microsaccades — tiny involuntary
eye movements — can make certain static patterns appear to move.
3. Four problems motion solves
1 · Depth relations — motion parallax
As you move, near things sweep across your retina fast and far things slowly (Ch. 11's non-pictorial cue; the praying mantis sways its head to get it).
2 · Recognizing objects — segmentation
A camouflaged animal blends into the background until it moves. Motion groups the parts that move together into one object:
Find the hidden shape (it only exists while moving)
What to notice: every frame is just random dots; pause and the shape is gone. While
playing, the dots inside a region all drift together and a shape pops out. Lower the coherence and the shape
dissolves — the grouping depends on shared motion.
Biological motion
The ability to perceive the movement of a biological entity (a person, an animal) and recognize the activity
— even from a handful of moving dots on the joints.
Point-light walker — 13 dots
What to notice: paused, it's a meaningless dot cloud; moving, it's instantly a person
walking. Turn it upside down and it gets much harder (an inversion effect, like faces). Scramble the starting
positions — each dot keeps its own motion — and the person vanishes: it's the relationships between
motions that matter. Synthetic walker, not motion-capture data.
3 · Grabbing attention
Motion grabs attention so strongly it can silence awareness of static images: motion-induced blindness (moving dots make static foreground targets disappear) and motion silencing (motion suppresses awareness of other changing features like colour).
4 · Interacting with the environment — optic flow
Optic flow
The pattern of apparent motion of objects in a visual scene caused by relative motion between the observer
and the scene. Moving forward → everything expands from a point: the point you're heading toward. Pilots use it
to land. Optic flow also creates vection — the sense of self-motion — which helps maintain
balance (the train-next-to-you illusion).
Optic flow — where are you heading?
What to notice: every dot streams away from one point — the focus of expansion — and
that point is exactly where you're heading. Near dots (bigger) move faster than far ones: motion parallax is
built into the flow field.
4. Evidence that MT/V5 codes motion — one per method
This section is Chapter 8's "convergence across methods" applied to a new function. Learn it as a set:
| Method | Finding |
|---|---|
| 1. Neuropsychology | Stroke in MT → impaired motion perception. Patient LM: sees the world like a strobe light — "When I'm looking at the car first it seems far away, but then when I want to cross the road suddenly the car is very near." Intact colour vision, intact 3D (stereo) vision, intact faces/objects → a selective deficit |
| 2. Single-cell recording | MT neurons are sensitive to motion direction and velocity, and neurons with similar preferences line up in columns |
| 3. fMRI | MT is more activated by optic flow than by dynamic noise (flickering dots with no coherent motion) |
| 4. Behavior | Motion adaptation — fatigue one population of MT direction cells and the opposite population dominates, producing an illusory motion aftereffect |
Fix the slide's term
The slides call LM's condition "Akinesia". The correct term for loss of motion perception is
akinetopsia ("-opsia" = vision). Akinesia means loss of movement — a motor
symptom, closer to Parkinson's. If the exam uses the slide's word, recognize it — but know what it means here.
Why LM is a good patient case
With intact colour, stereo, faces and objects, LM can't be "generally impaired" — the spared abilities are a
dissociation (Ch. 5). And notice the world doesn't go blank for her; she sees snapshots. Position is
still computed; what's lost is the inference that links positions over time.
5. Motion adaptation: the lecture's final question, answered
Why does a stationary scene appear to move in the opposite direction after you stare at motion for a while? Try it first:
Motion aftereffect — stare at the centre for 20 s, then watch it stop
What to notice: after the rings stop, the static pattern seems to expand
— the opposite of the contraction you adapted to. The bars on the right show the mechanism: two opponent
populations of MT cells (inward-tuned and outward-tuned). A static image normally drives both equally, so they
cancel. Adaptation fatigues the inward cells; for a few seconds afterwards the outward cells win.
The answer
MT has direction-tuned cells for opposite directions, and perceived motion depends on the balance
between them. Staring at motion in one direction fatigues the cells tuned to that direction. When
the scene stops, the fatigued cells respond below normal while the opposite-direction cells respond normally —
so the balance tips toward the opposite direction, and a stationary scene appears to move that
way. Same logic as tilt adaptation (Ch. 10).
Slide wording
The L9 slide reads: fatigue the inward-flow cells → outward-flow cells dominate → "we will perceive an
inward optic flow". By its own logic that should say outward: if outward cells dominate,
the static image seems to expand. That's also what the demo above shows.
6. Check yourself
A patient after a stroke sees cars "jump" from far to near, but sees colours, depth and faces normally. The lesion is most likely in…
Why: Selective motion loss with everything else intact = MT (patient LM). V1
damage would cause a blind region, not a motion-only deficit.
After watching a waterfall for a minute, the rocks beside it seem to drift upward. This is because…
Why: You adapted to downward motion, fatiguing downward cells. The
aftereffect is always opposite to the adapting direction.
Which finding comes from single-cell recording rather than fMRI?
Why: Tuning of individual neurons needs single-neuron resolution. The flow
vs noise contrast is the fMRI result; LM is neuropsychology; the aftereffect is behavior.
You're on a stopped train; the train beside you pulls away and you feel yourself moving backward. This is…
Why: Large-field optic flow creates a sense of self-motion — vection.