natural selection cannot build any feature in which intermediate steps do not confer a net benefit on the organism.
as reported in Hopless Matzke
contrast this with what Allan Miller said at The Skeptical Zone
It is sufficient that NS does not act too strongly against, not that it must act for, a particular change.
TSZ Allan Miller says natural selection has to fail for evolution to work
I actually agree with Miller to the extent that Miller agrees with Michael Lynch and Mae Wan Ho
many genomic features could not have emerged without a near-complete disengagement of the power of natural selection
opening, The Origins of Genome Architecture
a relative lack of natural selection may be the prerequisite for major evolutionary advance
Mae Wan Ho
What does this all mean? The fact that a complex feature may be selectively advantaged does not imply that natural selection was the mechanism that evolved the feature. It may well be that the intermediate transitional forms were selectively disfavored, hence “luck despite the odds” enabled the feature to finally emerge. But in such case, this is an appeal to pure luck not Darwinian mechanisms!
Once such a feature emerges, it could be selectively favored, but this does not mean at all that the steps leading to the final product were selectively beneficial. I highlighted a fallacy that is common in evolutionary biology: Selection is falsely called a mechanism when instead it should be labeled an outcome.
I found Coyne’s quote in Hopless Matzke, and because I still hear some swearing by the power of Darwinian selection, I felt I occasionally have to contest such devoted belief to an insufficient cause for complex designs in biology.
Allan Miller wins the battle with Coyne, but loses the war with Behe’s irreducible complexity since Allan’s claim is essentially an appeal to blind luck despite the odds.
HT David Berlinski and Tyler Hampton