Matrix definition — rows, columns, order, elements: Visual walkthrough

The parent note told you a matrix is a "rectangular array of numbers." That sentence hides a lot of quiet decisions: why a rectangle? why two labels on every number? why rows before columns? This page rebuilds the whole idea in pictures, one decision at a time, so that by the end nothing is a convention you had to memorise — everything is a thing you can see.

We assume you know only what a number is and how to count. That's the starting line.

Step 1 — Why a single row is not enough

WHY. A line works when there is only one thing changing (the day). To find any number I only need to answer one question: "how far along the line?" That's a single coordinate — one address, like a house number on a straight street.

PICTURE. In the figure below, the three amber boxes sit in one horizontal strip. The white arrow underneath is the only direction of travel: "across." One number, the position, names each box completely.

Figure — Matrix definition — rows, columns, order, elements

Step 2 — Why a second direction forces a second label

WHY. The moment two things change independently (the fruit and the day), a single "how far along" answer is ambiguous. "Position 5" — is that the 5th box reading across, or down? We need two questions answered at once: which strip (which fruit) and which spot along it (which day). Two independent directions demand two addresses.

PICTURE. The figure shows the grid. The cyan arrow points down (choose the fruit-strip), the white arrow points across (choose the day). Every amber box now sits where one down-choice and one across-choice cross. No box can be named by a single number anymore.

Step 3 — Naming the two directions: rows and columns

WHY these names and this shape? A rectangle (rows all the same length, columns all the same height) is the only shape where "3rd spot across" means the same thing in every row. Ragged rows would break addressing — the 3rd box of a short row and a long row wouldn't line up. Regularity is what makes an address reusable.

PICTURE. The figure highlights one row in cyan (a horizontal band) and one column in amber (a vertical band). Where the cyan band and amber band overlap sits exactly one box — that single overlap is the whole point of the grid.

This matches Types of Matrices, where the shape of that rectangle (square, single row, single column) gives each matrix its family name.

Step 4 — The address of one box: the double subscript $a_{ij}$

WHY row first? Because we agreed once and stick to it forever — otherwise $a_{23}$ would mean two different boxes depending on who's reading. The chosen order is Row then Column, i.e. alphabetical R–C, matching how we read a page: first pick the line (go down), then the spot on it (go across).

PICTURE. Look at the highlighted box. A cyan number on the left edge counts rows down $= i$ . An amber number on the top edge counts columns across $= j$ . The label $a_{ij}$ floats on the box, its two subscripts colour-matched to the two edges.

Step 5 — The order $m \times n$ : measuring the whole rectangle

WHY this and not just "big" or "small"? Two matrices can only be compared, added, or checked for equality if their shapes match exactly. "Order" is the precise word for shape. It also instantly tells you the box count: $m$ rows, each holding $n$ boxes, gives $m \times n$ boxes total.

PICTURE. The figure brackets the grid. Down the left side, a cyan bar counts $m$ rows; along the top, an amber bar counts $n$ columns. The caption reads " $m \times n$ ", the cyan factor first (rows), amber second (columns) — same R-before-C order as the subscripts.

Step 6 — Degenerate cases: when the rectangle collapses

Every idea must survive its extreme cases. What happens when the grid shrinks to its thinnest possible forms?

WHY show these? They are not exceptions to the rules — they obey every rule from Steps 4–5, just with one count squeezed to $1$ . Confirming this proves the addressing scheme never breaks. A row matrix's boxes are still $a_{11}, a_{12}, \dots$ ; the first subscript simply can never be anything but $1$ .

PICTURE. Three mini-grids side by side: a flat cyan strip (row matrix), a tall amber strip (column matrix), and a square with a highlighted top-left-to-bottom-right diagonal (square matrix — the diagonal $a_{11}, a_{22}, \dots$ only exists when $m = n$ ).

Step 7 — Putting it to work: build a matrix from a formula

WHY this is the payoff. This is Steps 4–5 working together. The address is no longer just a label — it becomes an input. Feed each $(i, j)$ into the rule, get the number that lives there. This is exactly how Linear Transformations and System of Linear Equations pack their information into a grid.

PICTURE. The final $3 \times 3$ grid, each box showing the substitution $i + 2 j \to$ value. Row index $i$ (cyan) runs down the left; column index $j$ (amber) runs across the top; each box's amber number is what the rule produced.

The one-picture summary

The single figure below compresses all seven steps: a line becomes a grid (why two addresses appear), the grid gets its row/column names, one box shows its $a_{ij}$ address, the whole rectangle carries its $m \times n$ order, and the thin degenerate shapes hang off the side.

Recall Feynman retelling (plain words)

Start with one shelf of boxes in a row — I only need "how far along?" to find any box. Now I add more shelves stacked downward: suddenly "how far along?" isn't enough, because I could be talking about any shelf. So I need two answers — which shelf (row) and which spot on it (column). We always say the shelf first, then the spot, like R comes before C in the alphabet: that's the little pair of numbers $a_{ij}$ . To describe the whole stack in one go, I count the shelves ( $m$ ) and the spots per shelf ( $n$ ) and call it " $m$ by $n$ " — that's the order, and it also tells me the total number of boxes. If the stack is only one shelf tall it's a row matrix; only one spot wide, a column matrix; as many shelves as spots, a square. And once every box has a clean two-number address, I can even build the whole grid from a rule that turns each address into a number — like $a_{ij} = i + 2 j$ , where the box's own position tells it what to hold.

Recall

What are the two addresses a box in a matrix needs, and in which order? ::: The row number $i$ first, then the column number $j$ — written $a_{ij}$ (R before C). Why can't a single row of numbers name each entry with one index once you add a second direction? ::: Because two independent directions (down and across) mean "position 5" is ambiguous — you need both a row and a column answer. What does the order $m \times n$ tell you besides shape? ::: The total number of elements, since $m$ rows each hold $n$ entries, giving $m \times n$ boxes. When does the diagonal $a_{11}, a_{22}, \dots$ actually exist? ::: Only for a square matrix, where $m = n$ .