I am in principle very attracted by a model where payment attaches to the papers submitted and/or published, rather than the journal subscriptions/licences etc purchased; it would have the great advantage of scaling up with any increase in the amount of research seeking publication.
However, there does seem to be a flaw in Stevan's economics (I'm no economist - perhaps Hal Varian and others can set me right if I am mistaken). In my experience (11 years running one journals list, and 4 years closely associated with another), although the majority of learned journals have relatively low circulations, they nevertheless have significantly more subscribers than papers. The following figures are, I think, plausible averages for this type of journal Subscribers 500 6 issues of 10 papers = 60 papers per year Rejection rate 50% = 120 papers submitted Thus, even if Stevan's estimate is correct that the costs of organising peer review amount to no more than 1/3 of current journal costs (I'm unconvinced - see Bernard Donovan's paper in Shaw, Dennis, and Roger Elliott, Economics, Real Costs and Benefits of Electronic Publishing in Science. Proceedings of the ICSU Press Workshop. Keble College, University of Oxford, UK, 31 March to 2 April 1998., http://associnst.ox.ac.uk/~icsuinfo/confproc.htm, September 1998; and Don King, passim - maybe Don can respond?), then the costs would be as follows (I have assumed that authors of accepted papers pay twice as much as those whose papers are rejected - the figures still stand, however, even if they all pay the same) Cost to existing subscriber = x Total cost = 500x Cost for peer review only = 500x/3 = 167x Cost per rejected paper = 167x/180 = .93x Cost per accepted paper = 1.86x This suggests that the institution of an author whose paper was published would pay more than at present, and that of a rejected author would pay only slightly less. Have I missed something? Sally Morris