A place to keep looking.
The mechanisms, their histories, and the limits of these models.
The balance
A small weight can have a large say.
A balance tips when the turning effects on its two sides differ. This turning effect, torque, depends on force and its perpendicular distance from the pivot. Equal masses balance at equal distances. A lighter mass can balance a heavier one if it has a longer arm.
A weighing balance compares masses. A lever uses the same turning relationship to change the force needed for a task. The principle is older than its mathematical description; Archimedes studied the equilibrium of levers in antiquity.
Take an equal-arm balance to the Moon and it still compares masses. Both sides feel the same smaller gravitational acceleration.
What this model leaves out
Both masses are 2 kg; the left one is 80 cm from the pivot. The torque readout compares them with the beam level. The settling motion and tilt limits are illustrative, not the response of a calibrated scale.
The compass
It has no idea where you are going.
A compass needle is a small magnet free to turn. It aligns with the local magnetic field, whether that field comes mostly from Earth or from a nearby magnet. Bring a strong magnet close and the compass becomes a very poor navigator.
A compass does not usually point at geographic north. The angle between the two directions is called magnetic declination, and it changes with location and time. NOAA and the British Geological Survey model Earth’s changing field for navigation.
You can make a simple compass from a magnetized needle floating on water. The water gives the needle the freedom to turn.
What this model leaves out
The nearby magnet’s south end faces the compass, so the needle’s north end turns toward it. Earth’s background field and the needle’s vertical tilt are omitted. The field lines are schematic.
The battery
A difference becomes useful when you connect it.
Chemical reactions maintain a voltage between the battery’s terminals. Close the circuit and current can flow, transferring energy to the bulb. Here, the voltage stays fixed: divide it by the resistance to find the current.
Alessandro Volta’s pile, announced in 1800, stacked different metals with electrolyte-soaked material between them. It provided a sustained source of electric current, opening a different kind of experiment from a brief static spark.
A battery is not a little tank of electrons. Its useful store is chemical energy; charge already exists throughout the circuit.
What this model leaves out
The model holds the supply at 3 V. Real cells have internal resistance, finite capacity and chemistry-dependent behavior. This is not an instruction to short a battery.
The tuning forks
The second fork was never struck.
The first fork sends pressure waves through the air. These can set a second fork vibrating. The response is strongest near the second fork’s natural frequency, where repeated pushes reinforce its motion. This is resonance. How well sound passes between the forks, and how quickly they lose energy, matter too.
The invention of the tuning fork is generally credited to musician John Shore in 1711. Later forks became precision tools for studying sound, not just for tuning instruments.
Touch a vibrating fork’s stem to a suitable wooden surface and the sound can become much louder. The larger surface moves more air.
What this model leaves out
Visible vibration is greatly slowed and enlarged. The response curve is an illustrative damped oscillator, not a calibrated acoustic simulation. Sound is optional and starts only when you enable it.
The hammer & anvil
The blow depends on what it meets.
Lifting the hammer gives it gravitational potential energy. During the fall this becomes kinetic energy. At impact, the hammer, workpiece and anvil deform; some energy becomes heat and sound. A sufficiently large stress can permanently change the workpiece’s shape.
An anvil supports the metal so a smith can shape it instead of simply sending it flying. The tools and techniques of metalworking developed in many places, using different metals and sources of heat.
A hammer’s force is not just its weight. How rapidly the moving hammer stops is crucial to the impact.
What this model leaves out
The energy calculation uses a 1 kg hammer. The orange metal shows permanent deformation building up with each blow. Its changing shape is illustrative; temperature, hardening and rebound are not modelled.
The pendulum
Fastest at the bottom. Still for an instant at each end.
Gravity accelerates the bob toward its lowest point. As it descends, gravitational potential energy becomes kinetic energy. It then climbs the opposite side, slowing down. Air resistance and friction gradually take mechanical energy out of the swing.
A longer pendulum swings more slowly. For small angles, one full back-and-forth swing takes approximately 2π√(L/g) seconds. Here L is length in metres; g is gravitational acceleration in metres per second squared. The bob’s mass does not enter the formula. That regularity made pendulums useful for keeping time.
A large swing takes slightly longer than a small one. The familiar period formula is a small-angle approximation, not an exact rule for every swing.
What this model leaves out
The motion follows the pendulum equation with a small energy loss that gradually reduces the swing. The rod is treated as rigid and massless, and the bob as a point mass. The displayed period uses the small-angle approximation.
The wheel
The same world that lets it roll can stop it.
A turning wheel has angular momentum. Without a net external torque its angular momentum stays constant. A brake applies frictional torque, slowing the wheel and converting mechanical energy into thermal energy.
Friction is not always the enemy of motion. Static friction between a driven tire and the ground is what lets a car accelerate without spinning its wheels. Here, a brake demonstrates the other side of the relationship.
When a wheel rolls without slipping, its contact point is momentarily at rest relative to the ground, even while the axle moves forward.
What this model leaves out
This is a wheel on an axle, not a rolling vehicle. Brake pressure controls a simplified resisting torque; a small bearing loss remains even with the brake released.
The seed & soil
Being ready is not the same as being able.
A seed contains an embryo and stored resources. Germination depends on a species-appropriate combination of water, oxygen and temperature; some seeds also need particular light or dormancy-breaking conditions. Water alone is not a universal start button.
At Kew’s seed bank, storing seeds is only part of the work. Researchers test how long they survive, what wakes dormant seeds and which conditions help seedlings establish.
Many seeds can germinate on moist paper. Soil is not always needed for that first step: the seed carries resources of its own.
What this model leaves out
Growth and wilting are accelerated illustrations, not a watering guide. The slider is relative, not a measured soil-moisture percentage. Dry or waterlogged conditions stop growth here; they do not turn a plant back into a seed. Species, drainage and temperature are left out. Excess water can reduce oxygen around roots.
The spark & fuel
Starting a fire is not the same as keeping one.
A spark can supply enough energy to start burning. To keep burning, a fire needs fuel, an oxidizer such as oxygen, and enough heat to sustain the reactions. Remove a required condition and the flame goes out.
The familiar fire triangle is a useful teaching model, not a complete account of combustion chemistry. The fire tetrahedron adds the sustaining chemical chain reaction.
A hot spark can fail to ignite a material. Its temperature is only part of the story; energy, contact time and the fuel’s condition also matter.
What this model leaves out
The fuel burns down only after ignition. Flame size is an illustration, not a prediction of fire behavior. Do not reproduce this experiment with real flames.
The voice & wire
The wire carries a signal. Not what it means to you.
A simple wired telephone turns sound into a changing electrical signal, then back into sound at the receiving end. The signal can weaken along the line, and noise can interfere with it. Understanding the message is something the listener does.
The telephone grew from overlapping experiments and competing claims. Bell received a US telephone patent in 1876, but a useful history also includes other inventors, operators and the networks that made calls possible.
Electrical signals travel through a circuit much faster than the average drift speed of individual electrons. A wire is not a queue of electrons sprinting from speaker to listener.
What this model leaves out
Travel is slowed for visibility. The wave and noise are illustrative, not an audio recording or a model of a particular telephone. No microphone is requested.
The tree & animal
One exchange. More than two participants.
In photosynthesis, plants use light energy to build sugars from carbon dioxide and water, releasing oxygen. Animals use oxygen in cellular respiration and release carbon dioxide. Plants respire too, in daylight and darkness.
Carbon moves through living things, air, soil, water and rocks. NASA observes parts of this cycle from space. The tree and animal shown here belong to a much larger exchange.
Much of a growing tree’s dry mass comes from carbon dioxide in the air, not from material lifted out of the soil.
What this model leaves out
The dots show processes, not quantities of gas. The sunlight slider is relative, not a measured light level. Low light reduces the illustrated photosynthesis; respiration continues. Real rates depend on the species and conditions.
The tide
A shore can belong to the sea twice a day.
Tides arise from differences in gravitational pull across Earth, principally from the Moon and Sun. A simplified equilibrium picture has two tidal bulges, not just a mound beneath the Moon. Earth’s rotation carries places through the changing pattern.
Real coastlines do not behave like the smooth ocean shown here. Basin shape, depth, rotation and local geography modify the timing and height. Some places have one high tide a day, others two or a mixed pattern.
The Sun contributes to tides too. Its alignment with the Moon helps produce the spring-neap cycle; “spring” here does not mean the season.
What this model leaves out
Bulges are dramatically exaggerated. This is the equilibrium-tide idea, not a tide forecast. The Sun, continents and rotating fluid dynamics are omitted.
The fungus & root
Beyond the root, finer threads.
Mycorrhizal fungi live in partnership with plant roots. Their fine threads, called hyphae, can reach water and nutrients beyond the root. Plants supply carbon-rich sugars and fats in return. The costs and benefits depend on the partners and their surroundings.
Kew researchers study these partnerships across habitats and plant groups. Different fungi are better at gathering different resources. A connection alone does not tell us how much each partner gains.
The familiar mushroom is a fruiting structure. Much of a fungus can exist as a network of fine threads, out of sight.
What this model leaves out
The slider extends a drawing, not a measured network or nutrient budget. Connections between trees do not establish that they deliberately help one another. A 2023 review questioned broad claims about the benefits of shared fungal networks; this remains an active research question.
The fire & cone
For some cones, heat opens the future.
Some lodgepole pine cones remain sealed by resin. Sufficient heat can release that seal, allowing the scales to open and seeds to disperse. This delayed release is called serotiny.
After the Maple Fire in Yellowstone, photographer Jennifer Jerrett documented a cone opened by the heat. Fire-adapted plants can exploit conditions after a fire, but a fire is not automatically beneficial to every plant or forest.
Lodgepole pines can produce both cones that open without fire and cones that hold their seeds for years. Not every cone needs a fire.
What this model leaves out
Heat is a relative control, not a species-specific temperature. Opening is irreversible in this simplified sequence. Fire severity and the interval between fires can determine whether regeneration succeeds.
The river & canyon
The water shapes the route that shapes the water.
Rivers carry sediment and wear away their beds and banks. The route they cut depends on the rock, the slope and the water flowing through. A canyon also records uplift, weathering and other changes to the landscape.
The Colorado River cut through the Colorado Plateau to help form the Grand Canyon. Its exposed rocks tell a far older story than the canyon itself. Scientists continue to investigate how older drainage systems became today’s river.
The age of a rock wall and the age of the canyon cut into it are two very different questions.
What this model leaves out
The slider moves through an illustrated sequence, not a percentage of erosion or a measured span of years. Floods, sediment supply, uplift and differences in rock strength are not modelled.
The sun & soil
New growth. Old material. Fresh energy.
Photosynthesis brings energy into many ecosystems. Decomposers break down dead organic material and help recycle nutrients. Matter moves through cycles; usable energy passes through and is eventually dispersed as heat.
Soil is a habitat as well as a place to grow. Its organisms, minerals, water and air all affect the plants above it. Decomposition can carry on underground whether or not the sun is shining.
At deep-sea hydrothermal vents, some microbes use energy from chemical reactions to make food without light. This process, chemosynthesis, supports communities far below the sunlit water.
What this model leaves out
The sunlight control is relative, not a measured light level or growth rate. The scene leaves out soil chemistry, microbial communities and decomposition times. It does not show sunlight driving decay.
The hunter & soil
The remains become part of something else.
Predators eat other organisms. Scavengers consume remains, while decomposers break down remains and waste, releasing nutrients that plants can use again. Energy disperses as heat along the way; it does not cycle back like matter.
After wolves returned to Yellowstone, some willow and aspen stands grew taller. Researchers have debated how much of that growth came from wolves changing elk numbers or behavior. Other predators and water availability also matter.
“Wolves changed the rivers” compresses a complex, debated history. Water availability, other predators, hunting and changing elk numbers also deserve a place in the account.
What this model leaves out
The slider follows an illustrated nutrient cycle, not population changes or a measured rate of recycling. Real food webs have many overlapping routes; remains from any stage can reach decomposers.
Question & answer
An answer can change the question that comes next.
A question opens a space, but the answer can reshape it. “More time” and “a hand” invite different follow-up questions. In a conversation, listening can change what someone asks next rather than simply fill a blank in a script.
Herbert Clark and Susan Brennan described understanding as something people establish together in Grounding in Communication (1991). A contribution can be accepted, checked, or repaired by what follows. The listener’s response helps shape the next turn.
An answer you did not expect may reveal a better question than the one you planned to ask.
What this model leaves out
This is a small authored branching conversation, not a person responding to you, a diagnosis, or a model of your needs. Real replies are not limited to these buttons. People can also ask and answer their own questions, and some questions remain unanswered.
Speaker & listener
The words arrived. Did the meaning?
A listener brings expectations to a sentence. The speaker brings an intention that may not fit them. Checking a paraphrase gives the speaker a chance to confirm it or put it right. Understanding someone does not require agreeing with them.
In their 1957 essay Active Listening, Carl Rogers and Richard Farson proposed restating another person’s meaning in your own words, then letting that person judge whether it fits. They described an attitude of attention, not a clever phrase that guarantees a response.
“I understand” ends the check. “Do you mean this?” leaves it open.
What this model leaves out
These are possible readings and one authored reply, not access to a real person’s private intention. Clarification can still fail, and someone may not wish to explain. Listening here includes reading and other forms of receiving communication, not only hearing.
Teacher & student
The example was clear. The rule was not.
A teacher can know a rule while a learner sees only its examples. One example may fit several explanations. Asking a learner to try a different case can make that gap visible, giving both people something specific to work with.
In The Story of My Life (1903), Helen Keller recalls Anne Sullivan spelling water into one hand while water flowed over the other. Keller describes making the connection between the spelling and the thing. Her account puts the learner’s discovery, not just the teacher’s demonstration, at the center of the moment.
A correct answer can come from a different rule. It is worth asking how it was reached.
What this model leaves out
This small number machine has a fixed rule: multiply by two. It does not simulate a student or measure learning. Real teaching involves more than choosing examples, and learning can happen without a teacher. Keller’s memoir is a personal recollection, not a universal account of learning.
Promise & person
Someone may already be arranging their day around your words.
A promise does more than describe an intention. Someone may rely on it before the promised action happens. Keeping it, changing it by agreement, and silently missing it leave that person with different information and different options.
In A Treatise of Human Nature, David Hume imagined two people whose harvests ripen on different days. Each could help the other, but the exchange needs a way to bridge the delay. He used the example to argue that promising depends on shared human conventions. That is his philosophical account, not the last word on obligation.
An early warning does not deliver the promised thing. It can give the other person time to choose another plan.
What this model leaves out
The book and its recipient are fictional. The outcomes describe this small situation, not how a real person must feel or whether a promise is legally enforceable. Promises to oneself also exist. No trust score is calculated.
Language & listener
The sentence stays. The meeting moves.
Words such as “me”, “here”, and “tomorrow” depend on the situation in which they are used. Knowing the vocabulary is not enough to identify their references. A listener may need the speaker, place, and time to be clear too.
Herbert Clark and Susan Brennan described communication as a joint activity in their 1991 paper Grounding in Communication. People work toward enough shared understanding for the task at hand. A nod, a correction, or a question can help establish that shared footing.
A perfectly preserved sentence can become a very poor set of directions when its context is lost.
What this model leaves out
This demonstration isolates three context-dependent words. It is not a theory of all language. Signed, written, spoken, and other forms of language can provide context differently. People also use language privately, and shared context does not guarantee agreement.
Trust & truth
A familiar voice cannot change what is inside the envelope.
Trust in a source can affect whether you accept a claim. It does not make the claim true. A trusted person can be mistaken; a stranger can be right. Evidence can change a belief without changing the fact that the belief was about.
In his 1877 essay The Ethics of Belief, William Clifford imagined a shipowner who quieted doubts about a ship instead of inspecting it. Clifford then asked whether a safe voyage would excuse that neglect. His argument separates arriving at a true belief from having good grounds for it.
An incorrect statement does not by itself prove a lie. To call it a lie is also to make a claim about the speaker’s intention.
What this model leaves out
The four counters are fixed before you open the envelope. This is a narrow example of checking an observable fact, not a lie detector or a measure of someone’s character. Truth does not require another person, and trust can include self-trust. In ordinary life we often have to rely on evidence gathered by others.
Path & interference
One dot tells you very little. A thousand tell a different story.
With two coherent alternatives, probability amplitudes combine. Their relative phase produces bright and dark bands in the distribution of detection events. If an ideal detector leaves a distinguishable record of the path, the interference term disappears. A person does not have to read that record.
Akira Tonomura and colleagues published a striking single-electron buildup experiment in 1989. Individual arrivals accumulated into an interference pattern. Their apparatus used an electron biprism, not two holes cut into a wall. The link below shows the experiment and its history.
Turning the detector off changes the distribution for a new run. It does not reach backward and rearrange dots already recorded.
What this model leaves out
The dots are browser-generated samples from an illustrative two-path distribution with a Gaussian envelope. Detector off uses an interference term; ideal path detection removes it. No actual particle trajectory, detector dynamics or quantum hardware is simulated. This is not evidence that consciousness changes physics.
Position & momentum
A sharper position comes with a broader momentum.
For a quantum state, the standard deviations of position and momentum obey Δx Δp ≥ ℏ/2. These are spreads in measurement outcomes across identically prepared systems, not two errors in a badly made ruler. A Gaussian minimum-uncertainty state reaches the lower bound.
The relationship belongs to the wave description. A localized wave packet requires a spread of wavelengths; momentum is related to wavelength. Feynman develops this connection through familiar wave packets before applying it to matter.
The two graphs describe the same prepared state in different ways. They are not simultaneous measurements of one particle.
What this model leaves out
This is a one-dimensional, zero-mean, minimum-uncertainty Gaussian family in units where ℏ = 1. The slider changes the prepared state. Both axes keep fixed scales. Other quantum states can have a larger uncertainty product. These are calculated probability densities, not a live measurement.
Entangled measurements
Neither side gets to choose its result.
In an ideal spin singlet, either observer gets either result with equal probability. But the joint distribution depends on the angle between their measurement axes. Matching axes give opposite results; at a right angle, agreement and disagreement are equally likely.
John Bell showed how different measurement settings could test limits on local hidden-variable explanations. Alain Aspect’s account follows the experimental work that put this distinction to the test. Simply showing two opposite outcomes is not enough to demonstrate entanglement.
To see the joint pattern, the observers must compare their records through an ordinary communication channel. Entanglement does not provide a faster-than-light message.
What this model leaves out
The browser samples the ideal singlet distribution P(a,b) = [1 − ab cos(θ)]/4, with a and b equal to +1 or −1. This is a classical simulation of predicted quantum statistics, not a Bell experiment or a source of entangled particles. Changing the angle clears the old run instead of mixing settings.