The recent serial section electron microscopic studies by Adamson and Michel (1993) on microves gels of frog mesentery have revealed that the large pores in the junction strand of the interendothelial cleft are widely separated 150 nm wide orifice-like breaks whose gap height 20 nm is the same as the wide part of the cleft. In this paper a modified version of the model in Weinbaum et al. (1992) is first developed in which this orifice structure is explored in combination with a random or ordered fiber matrix layer that is at the luminal surface and/or occupies a fraction of the wide part of the cleft. This basic orifice model predicts that for the measured Lp to be achieved the fiber layer must be confined to a relatively narrow region at the entrance to the cleft where it serves as the primary molecular filter. The model provides a much better fit of the permeability P for intermediate size solutes between 1 and 2 nm radius than the previous model in Weinbaum et al., where the junction strand breaks were treated as finite depth circular or rectangular pores, but like the previous model significantly underestimates P for small ions. However, it is shown that if a small frequent pore of 1.5 nm radius with characteristic spacing comparable to the diameter of the junction proteins or a continuous narrow slit of approximately 1.5 to 2.3 nm gap height is also present in the continuous part of the junction strand, small ion permeability can also be satisfied. The 1.5 nm radius pore does not significantly change Lp, whereas the continuous narrow slit provides a contribution to Lp that is comparable to, or in the case of the 2.3 nm slit greater than, the widely spaced 150 nm orifices. Thus, for the narrow slit the contribution to Lp from the orifices can be as low as 1.0 x 10(-7) cm/s/cm H2O and it is also possible to satisfy the 2.5 fold increase in permeability that occurs when the matrix is enzymatically removed from the luminal side of the cleft, Adamson (1990). The likelihood of each of these cleft structures is discussed.