When the car moves forward and its image retreats
In an imagined scene, a father and son watch a short clip of a car on a straight road. Following its body, the child points to a wheel: “Why does it go backwards when the car goes forwards?” The father replays it. The metal spokes seem to slow, stop briefly, then resume in the other direction. Nobody pressed rewind, and the car never stopped. What changed was the relationship between movement and what successive images could show of it.
This small question opens a problem larger than a wheel: does a recording show the whole event or what it left in a series of samples? How does an image being accurate at every preserved instant differ from the correctness of an inference about their succession? Each image may faithfully record the wheel’s position when captured, while their combination suggests a direction that never occurred. A misleading final understanding requires no lie within an individual image.
We begin with the video effect, where a clear numerical example explains the ambiguity, then cautiously consider a related phenomenon reported during direct viewing in continuous light. Similar appearances do not establish identical causes. Between them lies an opportunity to understand imaging tools, limits to comparisons with the brain, and responsibility for statements about what we saw. This explanatory article is neither an eyesight test nor evidence of illness, and does not reduce perception science to one visual trick.
Opening questions
- How can accurate images suggest inaccurate movement?
- Why do repeated spokes make direction ambiguous?
- Does the video effect mean the brain operates like a camera?
- What should we know about a recording before making it evidence?
1. The interval between images is not an emptiness in the world
The wheel keeps moving between images, but the recording supplies no independent description of every position crossed. Video captures successive frames at a particular rate, while the camera collects light over each frame’s exposure time. These differ: images per second and the duration contributing light to each image. We may first explain succession as though images were extremely brief moments, then add exposure duration. That simplifies the explanatory model rather than fully describing every camera.
Imagine drawing a circle with one mark near its edge on successive notebook pages. The mark begins at the top; on the next page it lies slightly left. Knowing nothing else, we might read a short leftward movement. Yet it could have travelled almost a full turn rightward to that location. The images alone do not reveal its whole path; they provide start and end positions while the route needs further knowledge.
“Is the image clear?” is therefore insufficient. Also ask whether the samples’ number and timing suit the movement described. Spatial clarity reveals edges and spokes without guaranteeing preservation of their movement through time. A frame can sharply show metal details while the sequence remains sparse enough to mislead about direction. More pixels alone do not necessarily fill temporal gaps. Improving one aspect of measurement does not automatically repair every deficiency.
Patrice Koehl’s teaching material at the University of California, Davis explains how the interval between frames affects interpretation, using a clock hand. This is sampling continuous change, not a special defect of car wheels. Original material
“the discrete data are often not enough to capture the physics” — Patrice Koehl, original English excerpt supplied in the source metadata. Source
This makes neither digital data useless nor video incapable of providing evidence. Measurement’s suitability depends on the question. One image may establish a car’s colour but not a part’s rotation direction. A sequence sufficient to track the car may not establish thin, repeated spokes’ speed. Each claim needs an appropriate amount and kind of information; claims do not gain equal force merely by coming from one file.
2. When a spoke resembles its neighbour
Similar wheel parts increase the problem. A unique red mark can be tracked individually, but identical spokes may not reveal which one appeared before. The spoke at the top in the next frame may be its neighbour taking its place. Repetition makes distinct positions visually equivalent and leaves several possible stories linking images.
Here is our educational calculation, not a measurement from a human experiment. Assume an ideal wheel with twelve equally spaced spokes. A full turn is 360 degrees, so neighbouring spokes are 30 degrees apart. Assume a camera records 30 frames per second and the wheel turns 29 degrees between images. Ignore blur, changing perspective and small spoke differences to isolate repetition’s relationship with sampling.
After a real 29-degree forward movement, the spoke pattern lies very close to its previous position. Why? Moving this repeated pattern a full 30 degrees puts each spoke where its neighbour was, leaving the overall shape apparently unchanged. We moved one degree less. The new position can therefore also be described as the old pattern shifted backwards one degree. Both descriptions match pattern locations, but one follows actual physical parts while the other links similar parts differently.
Real movement is 29 degrees multiplied by 30 frames: 870 degrees per second. If a viewer follows the short backward link, apparent movement is minus one degree per frame, or 30 degrees per second in reverse. No motor suddenly reverses in this calculation. The ambiguity arises because images contain no independent identity for each spoke, and a large movement of a repeating pattern can leave samples resembling a different small movement.
Change just one number. A full 30 degrees between frames would repeat the same pattern in each image, possibly appearing stationary. At 31 degrees, the pattern advances one degree beyond its repeating location and may appear slow in the correct direction. These cases do not require the wheel ever to have physically stopped. They show appearance changing with rotation relative to the interval between images. The result belongs to the assumed model, not a ready rule for measuring any wheel’s speed by eye alone.
Check the geometry without running a car or approaching machinery. Draw twelve radial lines on transparent paper over a matching drawing, then rotate the upper sheet 29 degrees. The lines approach their previous locations from the other side. This checks positions geometrically, not how every viewer responds to video. Distinguishing proof of pattern similarity from measurement of perceptual impressions prevents assigning psychological findings to a paper exercise that never measured them.
3. Frame rate is not exposure time
Images closer together in time can reduce ambiguity under suitable conditions, but guarantee nothing across every speed and pattern. Shortening exposure also does not increase frame count. The example may remain at 30 frames per second while each image becomes sharper through shorter light collection. That gives clearer spoke edges at selected instants, not new instants between them. This matters when understanding phone-camera settings.
The computational-photography chapter hosted by MIT’s Computer Science and Artificial Intelligence Laboratory explains motion blur and temporal sampling. An image collects light within a time window, so a moving part’s image may spread across several positions in one frame. This integration can reduce some rapid temporal variation, at the cost of blur. Chapter
“A photograph is not an instant” — Original English supplied in the source metadata from “Motion blur and temporal sampling.” Source
A useful paradox follows: the sharpest image alone does not always tell us most about motion. Sharpness supplies spatial detail, while spreading within a frame may reveal something about the path during exposure. Blur is not always better; it serves one function and loses other information. The tradeoff concerns the measurement aim and recording conditions. Photographing text differs from photographing a moving spoke; making a beautiful image differs from measuring speed reliably.
Separate apparent direction from shape distortions in some recordings. A bent fan blade or apparently stretched part does not necessarily have the same explanation as the backward wheel. Strangeness alone cannot group them under one name. This article concerns motion ambiguity from successive samples and repeated patterns; sensor readout distortions and camera engineering need separate analysis for the device and clip.
When comparing recordings, note what actually changed. Capture frame rate or playback speed? An original file or a transmitted, recompressed or edited copy? Did wheel speed change too? These methodological questions do not accuse every edited copy of fabrication. Several changing factors prevent confident attribution to one. Good explanation begins by establishing viewing conditions, not collecting technical names.
4. Why is “the brain is a camera” insufficient?
In 1996, Dale Purves and colleagues reported a similar effect in continuous light and proposed an explanation involving successive visual episodes. Only the original PubMed abstract was examined here, so we attribute no unverified sample or procedural details to it. Original abstract
“a similar illusion is also seen in continuous light” — Purves and colleagues, original English supplied in the source metadata. This describes their reported phenomenon, not consensus on its explanation. Source
The distinction matters: video is known to be sampled by a recording device. Direct vision does not justify assuming a camera inside the head merely because appearances resemble each other. A hypothesis needs testable predictions and comparisons with alternatives. Two felt effects may resemble each other while arising differently. A shared name helps inquiry but misleads if treated as proof of matching mechanisms.
Pakarian and Yasamy reported a different result in 2003: their participants did not report the effect under the continuous-light sources used, although it appeared with flicker. This is the finding described in their abstract, not a verdict that all other reports are false. Abstract
“WWI cannot be taken as evidence for discreteness of conscious visual perception” — Pakarian and Yasamy, original English conclusion supplied in the source metadata. Source
In 2005, VanRullen, Reddy and Koch studied varied patterns and conditions, found an association with attention and offered an attention-dependent sampling model. Procedures show small samples, participation by some authors and a period of familiarisation with seeing the effect. This cannot become a fixed brain frame rate for everyone. Model numbers explain particular data; they are no human technical specification. Original research
In 2008, Kline and Eagleman presented evidence against a particular temporal-sampling account, including independence of some motion reversals and their appearance in nonperiodic patterns. They proposed adaptation and competition between motion signals. Their research also presents explanatory limits, so we do not use it to declare the debate over. Research
A serious objection: does this complicate a simple phenomenon?
A reader may say the spoke example suffices and the debate distracts from the answer. That is fair if we restrict ourselves to video. But “This proves the eye captures a limited number of pictures” makes a broader claim, requiring distinction. Simplify the calculation, but do not delete the condition making it valid. Good brevity brings an idea closer; omitting its limits changes the idea itself.
A balanced position neither equates every explanation nor declares knowledge impossible. We have a clear account of a discrete recording model and a different experimental debate about perception in continuous light. Place each claim within its scope and identify its supporting evidence. Readers can retain practical understanding without acquiring a belief larger than the sources allow.
What does recording teach us about knowledge?
The wheel offers a lesson in observation versus inference. A precise observation might be: “The spoke pattern appeared at these positions across frames.” “The physical wheel rotated backwards” adds a path, cause and direction. Inference is not wrong simply for being inference: understanding needs connections between information. Error begins when we forget that the connection itself can be examined and changed by further information.
Imagine two witnesses watching the same clip. One says the wheel reversed; the other says recording made it look so. Both may honestly describe their experience, but sincere feeling does not settle an event’s explanation. Instead of accusing deceit, examine recording conditions and knowledge of the car’s motion and pattern. We can respect an observer while correcting the judgment drawn. Separating a speaker’s dignity from their explanation’s correctness is useful.
Nor does the illusion justify universal distrust of the senses. If every observation were worthless, we could not discover the ambiguity or compare model and images. Correction uses additional observations, tools and independent knowledge. The eye involved in the first impression can notice differences between recordings; the mind that linked images one way can revise that link. Fallibility does not negate knowledge’s value; it makes its conditions worth attention.
Productive doubt also differs from doubt that accepts no answer. Productive doubt seeks obtainable information: recording rate, pattern shape or light source. Rejecting every piece of evidence merely because error is possible leaves no route to self-correction. The wheel’s lesson is choosing an appropriate test and calibrating confidence, not suspending judgment forever. Evidence sufficient for a limited question allows an answer while other questions remain open.
In the light of the Quran and Islamic wisdom
Allah says: TRANSLATION OF THE MEANING · COMPLETE VERSE 17:36Do not follow[1] that of which you have no knowledge. Indeed, the hearing, the sight, and the heart, all of them will be called to account.Translation notes: [1] i.e., do not say or assume.Rowwad Translation Center · QuranEnc · v1.0.20 [Al-Isra, 17:36]. The verse continues with responsibility for hearing, sight and the heart. Al-Saadi explains the prohibition through careful verification in speech and action, connecting it with responsibility for using one’s faculties. Verse and commentary
“Rather, verify carefully everything you say and do.” — Abd al-Rahman al-Saadi, commentary on Al-Isra 17:36; English rendering of the original Arabic commentary. Source
Our proposed connection is editorial reflection: describing what we saw requires distinguishing the received image from knowledge of its explanation. The verse discusses no frame rates and yields no sampling theory or experimental evidence of neural mechanisms. Its ethical meaning exceeds our example; we invoke it to verify before making impressions firm reports. Respect for the text means assigning it no unstated claim; respect for science means not using it to decorate an unexamined assertion.
More precise language need not be strained: “The wheel appears to reverse in the recording” is better supported than “It actually reversed” when all we have is the clip. The first acknowledges experience while preserving the gap between viewing and judgment. Knowing the conditions may later support a stronger explanation. Verification is positive work: research, compare and correct, rather than merely hesitate or pass astonishment along.
Practical applications: examining an astonishing clip
Specify the question first. Do you want apparent direction, actual speed, or whether the clip was played backwards? These differ. Sampling may explain the first without supplying enough for the second; other objects’ movements may offer clues to the third. Explaining one part does not certify every detail of the image.
Then list knowns and unknowns. You know there is an image sequence and repeating pattern, but may not know capture rate, exposure time or wheel speed. Do not fill gaps with imaginary numbers. The article’s numerical example explains possible ambiguity, not that your clip used those numbers. An example clarifies a relationship; actual measurement needs case-specific data.
For safe learning, use drawings or prepared images instead of approaching a moving wheel or exposed fan. Arrange paper copies of a repeating circle at known angles and ask what movements could connect them. Understanding the calculation needs neither flashing light nor prolonged staring. The aim is to show multiple motions fitting discrete positions, not produce the strongest illusion or test anyone’s tolerance.
Share the explanation with its limits: the example concerns discrete recording of a repeated pattern; continuous-light phenomena have separate literature and debate. Avoid “Science proved the brain records so many frames” when a source discusses a particular model. An explanation’s strength lies in leaving the question clear after astonishment fades, not in its most exciting sentence.
Try changing one condition in the drawing and predicting before looking. If one spoke becomes distinctive, what changes in identity ambiguity? If spokes remain identical but angular gaps between images shrink, do you expect the same account? Keep prediction separate from result, then compare positions. The example becomes practice in revisable questioning rather than a memorised story ending.
A teacher might initially supply only two images and ask students for different connecting routes. Some choose the shortest turn, others add a whole revolution or suggest the opposite direction. A geometric task should assess each proposal against positions rather than strangeness. Add an intermediate image and ask which possibilities are now excluded. This establishes no particular perceptual mechanism, but shows alternatives narrowing with new information.
Daily disagreements over a scene may need the same skill. Ask precisely what someone noticed, distinguish description from interpretation, then seek information separating possibilities. This is a thinking proposal, not a psychological law derived from a wheel. Ethically, it permits correction without ridicule: a shared starting point can lead along different paths to conclusions.
When needed information is unavailable, a limited answer remains useful. The model establishes the possibility of apparent reversal without physical reversal; identifying what occurred in a particular clip requires its conditions. This distinguishes demonstrating possibility from identifying a case, rather than evading it. Exaggeration often begins when an attractive possible explanation is immediately treated as the only one.
Conclusion: an image does not carry the whole path
The father returns to the imagined clip without trying to persuade his son that his experience never happened. “Yes, it looks backwards. Let us ask how the images making it look that way were made.” He draws a circle with similar spokes and two views separated by known movement. Astonishment shifts from disagreement about the eye to inquiry into its information.
A wheel need not reverse for its pattern to reverse in a recording. Repeating parts and sampled positions allowing ambiguous links can suffice. Yet this does not turn the brain into a fixed-rate camera. The wider lesson is choosing questions and answer limits well: trust what evidence supports, stop where it does not, and preserve room for revision between scene and knowledge.
Closing questions
- What distinguishes describing what appeared in a clip from describing what actually happened?
- What missing information would you request before calculating a filmed wheel’s speed?
- Do you know another example of accurate samples supporting a misleading inference?
- How can you explain astonishment without turning analogy into scientific fact?