Inside a memory, cells are arranged in rows. Each row has a word line, one output of the address decoder, running along it.
- When a word line is 1, every cell on that row is connected to its column's bit line, so the row can be read or written.
- When it is 0, the row's cells are isolated and keep their values.
Because the decoder's outputs are one hot, exactly one word line is active for any address, and so exactly one row takes part in each access.
In a simple memory, one row is one word, so word line Wᵢ selects word i. In a large chip with two dimensional decoding, a word line selects a whole row of the grid, and a column decoder then picks the bits you want from it.
In a ROM, a stored 1 is a connection from a word line to a data bit's OR gate. When the word line goes high, exactly the bits with connections become 1.