Journals with 90% rejection rates, like Nature, Science and Cell have
considerably higher editorial costs (per published paper) than those with
rejection rates of 40%-60%, which is an average value for middle-of-the-road
biomedical journals. Nearly the same effort goes into peer reviewing a
The Science web site is run by HighWire Press.
HighWire lists its free online full text journals at
http://www.highwire.org/lists/freeart.dtl
I do not recollect their announcing the implementation of the free after
one year policy, but it does seem to be in effect for certain sections.