What is "dimension," really? The truth about degrees of freedom

We all say "three-dimensional space" without blinking — but what exactly does the "three" mean? The answer is less obvious than it seems, and proving it requires the Steinitz exchange lemma.

Folio Official

March 1, 2026

Everyone knows that space is three-dimensional. But what, precisely, does that mean?

The intuitive answer is: "you need three numbers to specify a point." A point in $R^{3}$ is a triple $(x, y, z)$ , so there are three degrees of freedom. That is certainly true — but it hides a subtlety that most textbooks address only in passing.

1 The hidden difficulty

The catch is that the choice of coordinates is not unique. You can describe points in $R^{2}$ using the standard basis ${(1, 0), (0, 1)}$ , but you can equally well use the basis ${(1, 1), (1, - 1)}$ :

Example 1.

The point

(3, 1)

in standard coordinates becomes

(3, 1) = 2 \cdot (1, 1) + 1 \cdot (1, - 1)

in the new basis. The coordinates have changed — from

(3, 1)

(2, 1)

— but the number of coordinates is still two.

Is this always the case? Could some clever choice of basis for $R^{3}$ use only two vectors? Or could you need four? The answer to both is no, and proving it is the central achievement of the theory of dimension.

2 The Steinitz exchange lemma

The key result is this:

Theorem 2 (Steinitz exchange lemma).

Let

V

be a vector space. If

{v_{1}, \dots, v_{m}}

spans

V

and

{w_{1}, \dots, w_{n}}

is linearly independent, then

n \leq m

In plain language: a linearly independent set can never be larger than a spanning set.

Example 3.

V = R^{3}

, the standard basis

{e_{1}, e_{2}, e_{3}}

spans the space (

m = 3

). If a set

{w_{1}, w_{2}, w_{3}, w_{4}}

were linearly independent, the lemma would give

4 \leq 3

— a contradiction. So any four vectors in

R^{3}

must be linearly dependent.

Example 4 (The exchange procedure).

Let us see the exchange in action. Take the spanning set

{e_{1}, e_{2}}

R^{2}

and the linearly independent set

{w_{1}}

with

w_{1} = (3, 2)

Since

w_{1} = 3 e_{1} + 2 e_{2}

, we can replace

e_{1}

with

w_{1}

to get the new spanning set

{w_{1}, e_{2}}

. Indeed,

e_{1} = \frac{1}{3} (w_{1} - 2 e_{2})

, so anything expressible in the old basis is expressible in the new one.

3 The definition of dimension

The Steinitz lemma has an immediate and powerful corollary:

Theorem 5.

Any two bases of a vector space

V

have the same number of elements.

Proof.

Let

B_{1} = {v_{1}, \dots, v_{m}}

and

B_{2} = {w_{1}, \dots, w_{n}}

be two bases. Since

B_{1}

spans and

B_{2}

is independent,

n \leq m

. Since

B_{2}

spans and

B_{1}

is independent,

m \leq n

. Therefore

m = n

. □

This common value is called the dimension of $V$ , denoted $dim V$ .

Example 6.

$dim R^{n} = n$ (standard basis ${e_{1}, \dots, e_{n}}$ )
$dim P_{n} = n + 1$ (basis ${1, x, x^{2}, \dots, x^{n}}$ for polynomials of degree $\leq n$ )
$dim M_{m \times n} (R) = mn$ (basis: matrices with a single $1$ in each position)
$dim {0} = 0$ (the zero space; its basis is the empty set)

4 Infinite-dimensional spaces

Not every vector space has a finite basis.

Example 7.

The space

R [x]

of all real polynomials has the set

{1, x, x^{2}, x^{3}, \dots}

as a basis, but no finite subset spans the space — generating a polynomial of degree

N

requires at least

N + 1

basis elements. So

dim R [x] = \infty

Example 8.

The space

C [0, 1]

of continuous functions on

[0, 1]

is also infinite-dimensional. Fourier analysis — the idea that a function can be expanded in terms of infinitely many sines and cosines — is a manifestation of this infinite-dimensionality.

5 The dimension formula

Dimensions add, but with a correction term:

Theorem 9 (Dimension formula).

For finite-dimensional subspaces

W_{1}, W_{2}

of a vector space

V

dim (W_{1} + W_{2}) = dim W_{1} + dim W_{2} - dim (W_{1} \cap W_{2}) .

This is the vector-space analogue of the inclusion-exclusion principle for sets.

Example 10.

V = R^{3}

, let

W_{1}

be the

x y

-plane (

dim W_{1} = 2

) and

W_{2}

the

x z

-plane (

dim W_{2} = 2

). Their intersection

W_{1} \cap W_{2}

is the

x

-axis (

dim = 1

), so

dim (W_{1} + W_{2}) = 2 + 2 - 1 = 3 = dim R^{3} .

Two planes through the origin, meeting only along a line, together span all of

R^{3}

6 Why it matters

Dimension is the most fundamental invariant of a vector space. It tells you that $R^{2}$ and $R^{3}$ are genuinely different objects — not just different in appearance, but in structure. That this invariant is well-defined (independent of the choice of basis) is not obvious; it is a theorem, and a deep one. Without the Steinitz exchange lemma, we would have no guarantee that "the number of parameters" is a meaningful concept at all.

Linear Algebra Algebra Between the Lines

Folio Official

Mathematics "between the lines" — exploring the intuition textbooks leave out, written in LaTeX on Folio.

1 followers·107 articles

What is "dimension," really? The truth about degrees of freedom

We all say "three-dimensional space" without blinking — but what exactly does the "three" mean? The answer is less obvious than it seems, and proving it requires the Steinitz exchange lemma.

Folio Official

March 1, 2026

Everyone knows that space is three-dimensional. But what, precisely, does that mean?

1 The hidden difficulty

Example 1.

The point

(3, 1)

in standard coordinates becomes

(3, 1) = 2 \cdot (1, 1) + 1 \cdot (1, - 1)

in the new basis. The coordinates have changed — from

(3, 1)

(2, 1)

— but the number of coordinates is still two.

2 The Steinitz exchange lemma

The key result is this:

Theorem 2 (Steinitz exchange lemma).

Let

V

be a vector space. If

{v_{1}, \dots, v_{m}}

spans

V

and

{w_{1}, \dots, w_{n}}

is linearly independent, then

n \leq m

In plain language: a linearly independent set can never be larger than a spanning set.

Example 3.

V = R^{3}

, the standard basis

{e_{1}, e_{2}, e_{3}}

spans the space (

m = 3

). If a set

{w_{1}, w_{2}, w_{3}, w_{4}}

were linearly independent, the lemma would give

4 \leq 3

— a contradiction. So any four vectors in

R^{3}

must be linearly dependent.

Example 4 (The exchange procedure).

Let us see the exchange in action. Take the spanning set

{e_{1}, e_{2}}

R^{2}

and the linearly independent set

{w_{1}}

with

w_{1} = (3, 2)

Since

w_{1} = 3 e_{1} + 2 e_{2}

, we can replace

e_{1}

with

w_{1}

to get the new spanning set

{w_{1}, e_{2}}

. Indeed,

e_{1} = \frac{1}{3} (w_{1} - 2 e_{2})

, so anything expressible in the old basis is expressible in the new one.

3 The definition of dimension

The Steinitz lemma has an immediate and powerful corollary:

Theorem 5.

Any two bases of a vector space

V

have the same number of elements.

Proof.

Let

B_{1} = {v_{1}, \dots, v_{m}}

and

B_{2} = {w_{1}, \dots, w_{n}}

be two bases. Since

B_{1}

spans and

B_{2}

is independent,

n \leq m

. Since

B_{2}

spans and

B_{1}

is independent,

m \leq n

. Therefore

m = n

. □

This common value is called the dimension of $V$ , denoted $dim V$ .

Example 6.

$dim R^{n} = n$ (standard basis ${e_{1}, \dots, e_{n}}$ )
$dim P_{n} = n + 1$ (basis ${1, x, x^{2}, \dots, x^{n}}$ for polynomials of degree $\leq n$ )
$dim M_{m \times n} (R) = mn$ (basis: matrices with a single $1$ in each position)
$dim {0} = 0$ (the zero space; its basis is the empty set)

4 Infinite-dimensional spaces

Not every vector space has a finite basis.

Example 7.

The space

R [x]

of all real polynomials has the set

{1, x, x^{2}, x^{3}, \dots}

as a basis, but no finite subset spans the space — generating a polynomial of degree

N

requires at least

N + 1

basis elements. So

dim R [x] = \infty

Example 8.

The space

C [0, 1]

of continuous functions on

[0, 1]

is also infinite-dimensional. Fourier analysis — the idea that a function can be expanded in terms of infinitely many sines and cosines — is a manifestation of this infinite-dimensionality.

5 The dimension formula

Dimensions add, but with a correction term:

Theorem 9 (Dimension formula).

For finite-dimensional subspaces

W_{1}, W_{2}

of a vector space

V

dim (W_{1} + W_{2}) = dim W_{1} + dim W_{2} - dim (W_{1} \cap W_{2}) .

This is the vector-space analogue of the inclusion-exclusion principle for sets.

Example 10.

V = R^{3}

, let

W_{1}

be the

x y

-plane (

dim W_{1} = 2

) and

W_{2}

the

x z

-plane (

dim W_{2} = 2

). Their intersection

W_{1} \cap W_{2}

is the

x

-axis (

dim = 1

), so

dim (W_{1} + W_{2}) = 2 + 2 - 1 = 3 = dim R^{3} .

Two planes through the origin, meeting only along a line, together span all of

R^{3}

6 Why it matters

Linear Algebra Algebra Between the Lines

Folio Official

Mathematics "between the lines" — exploring the intuition textbooks leave out, written in LaTeX on Folio.

1 followers·107 articles

What is "dimension," really? The truth about degrees of freedom

1 The hidden difficulty

2 The Steinitz exchange lemma

3 The definition of dimension

4 Infinite-dimensional spaces

5 The dimension formula

6 Why it matters

Share your expertise with the world

More from Folio Official

The geometry that inner products unlock: orthogonality, projection, and least squares

The Spectral Theorem: Orthogonal Diagonalization of Symmetric Matrices

The Singular Value Decomposition: Structure of Arbitrary Matrices

Matrices and Representation of Linear Maps

What is "dimension," really? The truth about degrees of freedom

1 The hidden difficulty

2 The Steinitz exchange lemma

3 The definition of dimension

4 Infinite-dimensional spaces

5 The dimension formula

6 Why it matters

Share your expertise with the world

More from Folio Official

The geometry that inner products unlock: orthogonality, projection, and least squares

The Spectral Theorem: Orthogonal Diagonalization of Symmetric Matrices

The Singular Value Decomposition: Structure of Arbitrary Matrices

Matrices and Representation of Linear Maps