Its pretty much convention but sometimes we want to distinguish between "row vectors" and "column vectors" because we want to treat them as matrices. We can multiply or but we cannot multiply or .
To multiply matrices the number of columns of the first must equal the number of rows of the second. In the first example above those are both 3 and in the second example they are both 1. In the the third example, the number of columns in the first matrix is 1 and the number of rows in the second is 3. In the fourth example, the number of columens in the first matrix is 1 and the number of rows in the second is 1.