It is just an extension of what you already know, but for multi-dimensional arrays. You can no longer use terms like 'rows' and 'columns' unless you have n ever more creative names for each dimension, but the logic is the same. Take all data from the dimensions that have a : and only the specified data from the dimensions which you give an exact index for.
Personally I rarely if ever use arrays of dimensionality higher than 3 because I like to be able to visualise what indexing means and since I can't visualise beyond a 3d cube I easily get confused. That said, the logic works the same whether it can be visualised intuitively or not!
