Why a Basis Needs Both Spanning and Independence

In the last post, we saw the core idea:

A basis is a coordinate system — a measuring grid for vectors.

But that raises an obvious question.

If a basis is “just a set of vectors,” why do we need that specific definition? Why insist on spanning and linear independence?

Because coordinates are only useful if they satisfy two requirements:

  1. Existence: every vector you care about must be describable using the grid
  2. Uniqueness: every vector must have exactly one description

Spanning gives you existence.
Independence gives you uniqueness.

And if you drop either one, coordinates stop behaving like coordinates.

Failure mode 1: no spanning → some vectors have no coordinates

Suppose you try to build a coordinate system in using only one vector, say .

Then the only vectors you can describe are multiples of .

That set is the span:

Geometrically, this is just a line through the origin.

So if is not on that line, there is no scalar such that . In that case, has no coordinates in your “system.”

This is why spanning matters:

If your set doesn’t span the space, your “grid” doesn’t cover the space.

Coordinates can’t exist for all vectors.

Failure mode 2: no independence → some vectors have multiple coordinates

Now suppose you try to build a coordinate system in using three vectors:

Here is the problem: these vectors are not linearly independent. And the consequence is immediate:

A vector can have more than one coordinate description.

In fact, the notes give an explicit example for

where can be written in two different ways:

So what are the coordinates of ?

Should they be

There is no correct answer — because “coordinates” are no longer well-defined.

This is why independence matters:

If your set is dependent, your grid assigns multiple addresses to the same point.

Coordinates lose uniqueness.

The definition of basis is forced

The two failures above explain why a basis must satisfy exactly two conditions.

Let be a subspace of . A set is a basis for if:

Spanning ensures every vector in has coordinates.
Independence ensures those coordinates are unique.

So the definition isn’t a random textbook ritual.

It’s the minimal structure needed for “coordinate system” to mean anything.

A one-sentence takeaway

A basis is the kind of spanning set that makes coordinates exist and mean something.

What’s next

Now that coordinates exist and are unique, we can talk about the object that packages them:

the coordinate vector .

That’s where “change of basis” becomes something you can compute.