Over a year ago, Robert Martin and I wrote a paper on exploring and disambiguating requirements with acceptance tests. In it we formulate the equivalence hypothesis - as formality increases, acceptance tests and requirements become indistinguishable. At the limit, acceptance tests and requirements are equivalent. IEEE Software magazine published it in the January/February’08 issue (http://www.computer.org/portal/cms_docs_software/software/homepage/2008/s1mel.pdf).
As usual, comments are welcomed.