Linear algebra feels hard after the first few weeks because the course switches from calculation to structure. Row reduction, determinants and matrix products are procedures you can drill, but questions about subspaces, linear independence, bases and linear maps ask whether you understand what the words mean, and a memorised procedure cannot answer them. The fix is to learn each definition with an example and a non-example, check definitions line by line in short proofs, and say in words what every calculation tells you.
This guide shows where the switch happens, works five typical questions from the "hard" part of the course, and ends with practice and a weekly method.
Where the course changes direction
Most first courses spend a few weeks on systems of equations and matrices. Then the questions change shape.
| The question says | What it is really asking | First line to write |
|---|---|---|
| Is W a subspace? | does W contain 0 and stay closed under + and scalar ×? | "Check 0 ∈ W. Let u, v ∈ W and c a scalar…" |
| Are these vectors independent? | does c₁v₁ + … + cₖvₖ = 0 force every cᵢ = 0? | "Suppose c₁v₁ + … + cₖvₖ = 0." |
| Find a basis for the kernel | which inputs map to 0, written as a spanning set with no spare vectors? | "Row reduce and solve Ax = 0." |
| Show T is linear | does T(u + v) = T(u) + T(v) and T(cu) = cT(u) for all inputs? | "Let u, v be arbitrary and c a scalar." |
| What do the eigenvalues tell you? | which directions T only stretches, and by how much? | "On the line spanned by v, T multiplies by λ." |
Every row in that table needs a definition, not a formula. That is the moment many students who were doing well start to struggle. If this is where you are, our one-to-one linear algebra help online page explains how sessions with Ajay Vatsyayan follow your own course.
Example 1: a subspace and two non-examples
Let W = {(x, y, z) : x − 2y + z = 0} in ℝ³. Show W is a subspace.
- Zero vector. 0 − 0 + 0 = 0, so (0, 0, 0) ∈ W.
- Addition. Let u = (x₁, y₁, z₁) and v = (x₂, y₂, z₂) be in W. Then (x₁ + x₂) − 2(y₁ + y₂) + (z₁ + z₂) = (x₁ − 2y₁ + z₁) + (x₂ − 2y₂ + z₂) = 0 + 0 = 0, so u + v ∈ W.
- Scalars. For a scalar c, cx₁ − 2cy₁ + cz₁ = c(x₁ − 2y₁ + z₁) = 0, so cu ∈ W.
Geometrically W is a plane through the origin. Now two sets that fail.
- U = {(x, y, z) : x + y + z = 1} does not contain (0, 0, 0), so it is not a subspace. It is a plane, but it misses the origin.
- Q = {(x, y) : xy ≥ 0}, the first and third quadrants with the axes, is closed under scalar multiplication, since (cx)(cy) = c²xy ≥ 0. But (1, 0) and (0, −1) are in Q while their sum (1, −1) has product −1, so Q is not closed under addition.
The second non-example is the useful one. It shows you have to check every condition: passing one test says nothing about the others.
Example 2: three vectors that are pairwise non-parallel but dependent
Are v₁ = (1, 2, 0), v₂ = (0, 1, 3), v₃ = (2, 5, 3) linearly independent?
No two of them are multiples of each other, and many students stop there and say "independent". That is wrong. Suppose c₁v₁ + c₂v₂ + c₃v₃ = 0 and row reduce, or compute the determinant of the matrix with these columns: it is 0. In fact
2v₁ + v₂ = (2, 4, 0) + (0, 1, 3) = (2, 5, 3) = v₃.
So the three vectors are dependent and span only a plane. To find that plane, take the cross product v₁ × v₂ = (6, −3, 1), giving 6x − 3y + z = 0. Check: v₁ gives 6 − 6 + 0 = 0 and v₂ gives 0 − 3 + 3 = 0.
What it means: {v₁, v₂} is a basis of that plane, and v₃ adds nothing new. The dimension of the span is 2, not 3.
Example 3: kernel and image from one row reduction
Let A be the 3 × 4 matrix with rows (1, 2, 0, −1), (2, 4, 1, 1), (1, 2, 1, 2). Find bases for the kernel and the image.
Row reduction gives the reduced form with rows (1, 2, 0, −1), (0, 0, 1, 3), (0, 0, 0, 0). Pivots are in columns 1 and 3, so the rank is 2.
Kernel. The free variables are x₂ = s and x₄ = t. Then x₃ = −3t and x₁ = −2s + t, so
x = s(−2, 1, 0, 0) + t(1, 0, −3, 1).
A basis for the kernel is {(−2, 1, 0, 0), (1, 0, −3, 1)}. Check by multiplying: A times each vector is the zero vector.
Image. Take the pivot columns of the original matrix, not the reduced one: {(1, 2, 1), (0, 1, 1)}. Their cross product is (1, −1, 1), so the image is the plane x − y + z = 0 in ℝ³. Every column of A satisfies that equation, which is a quick check.
What it means: the kernel lives in ℝ⁴ (the inputs) and has dimension 2; the image lives in ℝ³ (the outputs) and has dimension 2. The two dimensions add to 4, the number of columns. A frequent slip is reading the image off the reduced matrix, whose columns span a different space.
Example 4: proving a map is linear, and its kernel
Let P₂ be the polynomials of degree at most 2, and define T(p) = (p(1), p′(0)) from P₂ to ℝ². Show T is linear and find its kernel.
Linearity. For polynomials p, q and a scalar c, (p + q)(1) = p(1) + q(1) and (p + q)′(0) = p′(0) + q′(0), so T(p + q) = T(p) + T(q). Similarly (cp)(1) = c·p(1) and (cp)′(0) = c·p′(0), so T(cp) = cT(p).
Kernel. Write p = a + bx + cx². Then p(1) = a + b + c and p′(0) = b. Setting both to 0 gives b = 0 and a = −c, so p = c(x² − 1). The kernel is spanned by x² − 1 and has dimension 1.
Since P₂ has dimension 3, the image has dimension 2, which is all of ℝ². So T is onto but not one-to-one.
For contrast, S(x, y) = (x + 1, y) is not linear, because S(0, 0) = (1, 0) ≠ (0, 0). A linear map must send 0 to 0, and checking that first saves time.
Example 5: saying what eigenvalues mean
Let M be the 2 × 2 matrix with rows (2, 1) and (1, 2). Its eigenvalues are 1 and 3, with eigenvectors (1, −1) and (1, 1).
What it means: M stretches every vector on the line y = x by a factor 3, and leaves every vector on the line y = −x unchanged. Any other vector is a mix of these two, so its direction changes. The determinant is 1 × 3 = 3, so areas are multiplied by 3 and M is invertible, since no eigenvalue is 0. Our step-by-step guide to finding eigenvalues and eigenvectors covers the calculation; the sentence of interpretation is the part students leave out.
Where students lose marks
- Checking closure but not the zero vector, or checking one closure property and assuming the other, as with Q in example 1.
- Pairwise checks for independence. Three or more vectors can be dependent even when no two are parallel. Use the full equation c₁v₁ + … = 0.
- Image from the reduced matrix. Use the pivot columns of the original matrix.
- Mixing up kernel and image. The kernel is a set of inputs, the image a set of outputs. Write which space each lives in.
- Proving linearity with numbers. Showing T(1, 2) + T(3, 4) = T(4, 6) proves nothing general. Use arbitrary u, v and c.
- Calling a basis a span. A spanning set can have spare vectors; a basis cannot.
Practice set
- Is {(x, y, z) : z = xy} a subspace of ℝ³?
- Are (1, 1, 0), (1, 0, 1), (0, 1, 1) linearly independent?
- Find a basis for the kernel of the matrix with rows (1, −1, 2) and (2, −2, 4).
- Is T(x, y) = (2x − y, 3y, x + y) linear? What is its kernel?
- What are the eigenvalues of the matrix with rows (4, 0) and (1, 2), and why can you read them off?
Answers. (1) No: (1, 1, 1) is in the set but 2(1, 1, 1) = (2, 2, 2) is not, since 2 × 2 ≠ 2. (2) Yes: the determinant is −2 ≠ 0, so they form a basis of ℝ³. (3) Rank 1; x = y − 2z, basis {(1, 1, 0), (−2, 0, 1)}. (4) Yes, it is multiplication by a matrix; the kernel is {(0, 0)} because the rank is 2. (5) 4 and 2, because the matrix is triangular, so the eigenvalues are the diagonal entries.
A study method that works
For every definition, write it out, then give one example and one non-example in your own words, as in example 1.
For every theorem, note its conditions and one situation where it fails if a condition is dropped.
For every calculation, add one sentence saying what the answer means: "rank 2, so the image is a plane", "eigenvalue 0, so the matrix isn't invertible".
For proofs, start with two lines: "We need to show…" and "We know…". Most first-course proofs follow from writing those clearly. The same habit is the core of starting real analysis proofs, where the switch from computing to proving is even sharper.
If you studied matrices as transformations at school, the IB matrix transformations guide is a good picture-first refresher. US students can see where the course sits in the US college maths sequence, and UK students can read about the jump from A Level to university maths.
Why it matters beyond the course
Linear algebra returns in differential equations (systems are solved with eigenvectors), multivariable calculus (the derivative is a linear map, and our advanced calculus sessions lean on it), numerical methods, statistics, machine learning, quantum mechanics and control systems. Students who rush through it often meet it again, harder, a year later.
How Maths Bodhi helps
We work through your own problem sheets, check each definition step in your proofs, and set extra subspace, basis and kernel questions until they feel routine. Other courses are listed on our university maths hub. We explain, practise and review your own work; we don't complete graded assignments or exams. Home tuition is available in Gurgaon (Gurugram). Students elsewhere in India and abroad learn with us online. You can book a free demo class.
Written and checked by the Maths Bodhi team (November 2026).
Frequently asked questions
- Why is linear algebra so hard?
- Most courses move from row reduction and determinants to abstract ideas such as subspaces, bases and linear maps, which need exact definitions and short proofs.
- How do I check if a set is a subspace?
- Show it contains the zero vector and is closed under addition and scalar multiplication, using arbitrary vectors rather than numbers.
- Can three vectors be dependent if no two are parallel?
- Yes. For example (1, 2, 0), (0, 1, 3) and (2, 5, 3) are dependent because the third is 2 times the first plus the second.
- How should I study vector spaces?
- For each definition, write an example and a non-example, and draw the two- and three-dimensional picture where you can.
- How do I start a linear algebra proof?
- Write down exactly what you need to show and what you already know; most first-course proofs follow from checking a definition step by step.
- Is linear algebra harder than calculus?
- It is different rather than harder: calculus rewards technique, while linear algebra rewards understanding definitions and structure.
