The conference version of the paper ``A new rank technique for formula
size lower bounds'' incorrectly claimed to obtain a lower bound of
n^2 + k2^l-k^2 on the formula size of parity on n=2^l +k bits. The error
is in Proposition 2, assuming that an optimal selection function will select
each s_i to be a power of 2. This error was pointed out to me by Dmitry
Cherukhin, who also pointed out the reference to the current best lower
bounds for formula size of parity by K. Rychkov, ``On lower bounds of
complexity of series-parallel contact networks which realize Boolean
functions'', Siberian Journal of Operations Research, Vol. 1, 33--52, 1994.
Rychkov shows a lower bound of n^2 +3 when n>=5 is odd and n^2+2 for even
n>=6 not a power of 2.