What Is the Intuition Behind a Basis in a Vector Space?

A basis is a way to describe every vector in a space using a set of reference vectors. The essential idea is that the reference vectors are enough to reach everything in the space, but contain no unnecessary vector. This gives you a coordinate system, even when the reference vectors do not look like the familiar horizontal and vertical axes.

The two requirements

Let $V$ be a vector space over a field such as the real numbers. A set of vectors is a basis for $V$ if it satisfies both of these conditions:

1. It spans $V$. Every vector in $V$ can be written as a linear combination of the set’s vectors.
2. It is linearly independent. The only way to combine the set’s vectors to get the zero vector is to use zero for every coefficient.

The spanning condition says the vectors can describe the whole space. Independence says that the description has no redundancy. Together, these conditions guarantee that each vector has exactly one representation as a linear combination of the basis vectors.

A basis as a coordinate system

In the plane, the standard basis is $(1,0)$ and $(0,1)$. It describes a vector $(x,y)$ as

$$(x,y)=x(1,0)+y(0,1).$$

The numbers $x$ and $y$ are the vector’s coordinates in that basis. But those two particular directions are not the only possible basis. You can choose two different, nonparallel vectors and use them as new reference directions. The vector itself stays the same; only its coordinate description changes.

A useful analogy is choosing directions on a map. If you use north and east as reference directions, you describe a location one way. If you choose two other independent directions, you can still locate every point, but the coordinates will differ. In a vector space, the basis vectors play the role of those reference directions.

Worked example: a different basis for the plane

Consider the vectors

$$v_1=(1,1),\qquad v_2=(1,-1).$$

They are not the standard horizontal and vertical directions. To see whether they form a basis for $\mathbb{R}^2$, check that they are independent and span the plane. Put them into the columns of a matrix:

$$A=\begin{pmatrix}1&1\\1&-1\end{pmatrix},\qquad \det(A)=-2\ne 0.$$

A nonzero determinant means the columns are independent and span $\mathbb{R}^2$, so they form a basis. Now express $(3,1)$ using this basis. Find $a$ and $b$ such that

$$a(1,1)+b(1,-1)=(3,1).$$

Matching the two components gives $a+b=3$ and $a-b=1$. Adding the equations gives $2a=4$, so $a=2$; then $b=1$. Thus

$$(3,1)=2(1,1)+1(1,-1).$$

The coordinates of $(3,1)$ relative to the ordered basis $(v_1,v_2)$ are $(2,1)$. These differ from its standard coordinates $(3,1)$. The basis coordinates are the coefficients used with the chosen reference vectors, not necessarily the vector’s usual components.

Why independence gives unique coordinates

Suppose the same vector had two representations using a basis $v_1,\ldots,v_n$:

$$a_1v_1+\cdots+a_nv_n=b_1v_1+\cdots+b_nv_n.$$

Subtract the two sides. This produces a linear combination equal to zero:

$$(a_1-b_1)v_1+\cdots+(a_n-b_n)v_n=0.$$

Because the basis vectors are independent, every coefficient in this combination must be zero. Therefore $a_i=b_i$ for every $i$: the representation is unique. Without independence, you could have different coefficient lists describing the same vector.

What “no redundancy” looks like

The set $(1,0)$, $(0,1)$, and $(1,1)$ spans $\mathbb{R}^2$, but it is not a basis. The third vector is the sum of the first two, so it adds no new direction. In fact, the same vector can be represented in more than one way:

$$(1,1)=1(1,1)=1(1,0)+1(0,1).$$

Removing $(1,1)$ leaves the standard basis, which still spans the plane. A basis is therefore a minimal spanning set: remove one of its vectors and it no longer spans the whole space. It is also a maximal independent set: adding any vector not already in the basis makes the set dependent.

Common mistakes

• Checking only that vectors span. A spanning set can contain redundant vectors, so you must also check independence.
• Checking only that vectors are independent. Independent vectors may not reach every vector in the space. For example, one nonzero vector in the plane is independent by itself, but does not span the plane.
• Assuming basis vectors must be perpendicular or unit length. Those are extra properties of an orthonormal basis, not requirements for a basis.
• Thinking coordinates are fixed. A vector has different coordinates in different bases. The vector is unchanged; the reference system changes.

For a finite-dimensional space, every basis has the same number of vectors. That number is the dimension of the space. In the plane, every basis has two vectors; in three-dimensional space, every basis has three. The particular vectors can vary, but the number needed to describe the space does not.