Memory is notated as Length * Width.
So, would mean memory that is 4M long and each location has 8bits.
Memory can be thought of as a matrix with rows and columns. The row number represents the address of the word. The word is just one full row.
Addressing
A word can have more than one byte in it. So, when we say byte addressable, we mean that each byte has its own address, and the starting byte's address being the address of the word.
Word addressable just means that each word has an address, and the address increments by one from word to word, and not from byte to byte.