Are the following statements true for elementary row operations?

I was wondering if the following statements about ERO's (elementary row operations) are mathematically correct. Any help and insight is appreciated. Any additional facts to the table is also welcomed.

ERO's performed on a matrix $A$ will preserve

1) The row space of the matrix A

2) The null space of the matrix A

3) Any linear dependence or independence of the columns of $A$

ERO's performed on a matrix $A$ will NOT preserve

1) The column space of the matrix $A$

2) Any linear dependence or independence of the rows of $A$

note that the preservation of linear dependence / independence refers to preserving not just if the set of vectors is linearly dependent / independent - but also the exact relationships between the vectors themselves (e.g. vector #1 is exactly 2x that of vector #3)

I would also add: ERO's preserve the dimension of the columns space (and nullspace of $A^T$), even if they don't preserve the spaces precisely. Also, "the linear dependencies of the columns of $A$" is precisely the information conveyed by the nullspace of $A$. That is: for any $x_1,\dots,x_n$ such that $x_1 \mathbf a_1 + \cdots + x_n \mathbf a_n = 0$, we have $$A \pmatrix{x_1\\ \vdots \\ x_n} = \pmatrix{0 \\ \vdots \\ 0}$$ Similarly, the nullspace of $A^T$ conveys the linear dependencies of the rows of $A$.
• I mean the second thing. That second thing is exactly the information that is conveyed by the nullspaces of $A$ and of $A^T$, as I try to clarify in my answer. – Ben Grossmann May 31 '16 at 8:47