What do arrows on a map, wobbly polynomial curves, and audio signals have in common?
On the surface, nothing. One is a direction, one is a shape, one is a sound. But mathematicians throw all three into the same bucket and call them "vectors." Why?
Because you can do the same two things to all of them: add two together, and scale one up or down. That's it. That's the entire idea. Everything else in this post is just working out the consequences.
Start with something you already know
Picture an arrow. Not a fancy one, just "walk 3 steps east, 1 step north." That's a vector. It has a direction and a length, and you can draw it as an arrow from the origin to a point.
Now here's the thing you can do with two arrows: follow one, then follow the other. Walk your first arrow, then walk your second arrow starting from where the first one left off. You end up somewhere, and that somewhere is also describable as an arrow from the origin. We call this vector addition.
And here's the other thing you can do: walk an arrow twice as far. Or half as far. Or backwards. That's scalar multiplication, multiplying a vector by an ordinary number.
Two moves. Addition, and scaling. Everything about vector spaces comes back to these two.
The result always lands back in the plane
Here's the quietly important part. When you add two arrows in the plane, the result is also an arrow in the plane. It never pops out into 3D, never turns into something that isn't an arrow anymore. It just... stays.
Addition Never Leaves the Plane
Drag the orange arrow w anywhere you like. The result v + w (green) always lands on some other point in the same plane, never outside it.
Drag w around. No matter where you put it, v + w lands somewhere else in the same flat plane, never off it. Try dragging w far away, try dragging it right on top of v. The sum never escapes.
This "never escapes" property has a name: closure. The plane is closed under addition. We write vectors as , and addition just adds the matching pieces:
Scalar multiplication is closed too. Stretch v by any number c, and is still a plain old arrow in the same plane:
Simple enough that it barely feels like a rule. But watch what happens when we try it somewhere else.
Not every collection plays along
Here's a question worth asking: does closure always hold? Or can you build a "space" where adding or scaling kicks you out of it?
Does Closure Survive Outside the Plane?
Pick a candidate space, then drag the scalar c. Watch whether c·v (orange) stays inside the shaded region, or escapes it.
Start with "Whole Plane" selected. Slide c anywhere, c·v always checks out. Now switch to "First Quadrant", the set of arrows pointing up-and-right only. Slide c negative, and watch c·v turn red: it leaves the quadrant. A negative scalar flips the arrow, and the quadrant has no room for arrows pointing the wrong way.
Now try "Unit Circle", just the arrows of length exactly 1. Slide c away from 1 or -1, and c·v falls off the circle almost immediately, because scaling changes length, and the circle only tolerates one length. Even addition fails here, v + w usually isn't length 1 either.
This is the whole game. A vector space isn't just "a bunch of things." It's a bunch of things where addition and scaling always, provably, land you back inside the same bunch. The quadrant and the circle are collections of arrows, but they're not vector spaces. The full plane is.
Mathematicians write this out as a checklist, eight rules in total: closure under addition, closure under scaling, addition that's commutative and associative, a zero vector that does nothing, an inverse that cancels every vector out, and scaling that distributes sensibly. Almost every one of them is really just formalizing "yeah, that's obviously true for arrows." The two that actually do work, and the two that just broke above, are closure.
The same rules, a different costume
So closure is the real test. But here's where it gets interesting: the test doesn't care what the objects are. Arrows were just the easiest example to draw.
Take polynomials, like . Can you add two polynomials? Sure, just add the matching coefficients. Can you scale one by a number? Sure, multiply every coefficient. Same two moves, same closure. Add two degree-2 polynomials and you get another degree-2-or-less polynomial. Scale one and you still get a polynomial. Nothing escapes.
The Same Rules, a Different Costume
Drag t. The orange curve is p1 + t·p2, an ordinary polynomial. But its coefficients are just a point moving through coordinate space, exactly like the arrows above.
Drag t. You're watching morph from one curve into another. But look at the readout: three numbers, the coefficients, sliding smoothly as t changes. That triple of numbers is the vector. The curve is just a picture of it, the same way an arrow is a picture of a pair of numbers.
This is the move that makes vector spaces worth naming. Once you see "add the pieces, scale the pieces, never escape the collection" as the actual definition, it stops mattering whether the pieces are coordinates of an arrow or coefficients of a polynomial. Audio signals, images, quantum states, they all qualify too, for the exact same reason.
You don't need every vector, just a few
Last question: if a vector space can be infinite, containing every possible arrow or every possible polynomial, do you have to describe it one vector at a time?
No. And this is the payoff. A small handful of vectors, combined with addition and scaling, can generate the entire space.
Two Building Blocks, an Entire Plane
e1 and e2 are just two fixed arrows. Drag c1 and c2, or press play, and watch every point p = c1·e1 + c2·e2 sweep across the whole plane.
e1 and e2 are just two fixed arrows, not even pointing in tidy directions. Drag c1 and c2, and watch p = c1·e1 + c2·e2 move. Press play and let it sweep. Two vectors, combined with nothing but the two moves you already know, reach every single point in the plane.
A set like that can reach everything this way is called a basis. The number of vectors you need, here just 2, is the dimension of the space. Polynomials up to degree 2 need a basis of 3 (, , ), so that space has dimension 3. The pattern is the same, only the count changes.
The short version
A vector space is any collection of objects where you can always add two of them and always scale one of them, and never fall outside the collection. That's closure, and it's the one property that actually matters, the rest of the axiom list is just making "addition and scaling behave the way you'd expect" precise.
Arrows satisfy it. Polynomials satisfy it. The first quadrant and the unit circle don't. And once a collection qualifies, a small basis of vectors, as few as the dimension of the space, is enough to reach every point in it by mixing addition and scaling.
Once you stop picturing "vector" as "arrow" and start picturing it as "anything you can add and scale without breaking," a huge chunk of math, from polynomials to signals to quantum mechanics, turns out to be the same two moves in different clothing.
All visualizations are interactive React components running entirely in your browser, built with plain SVG and no external libraries. Drag the arrows directly, or use the sliders, everything updates instantly.