With the aid of a scale model of a substomatal cavity it has been shown that the rates of evaporation inside such a cavity are greater from loci near the pore than from further away. Implications for the process of evaporation inside a real substomatal cavity are discussed, especially the contribution towards vapour loss from inner epidermal walls. The hypothesis that the water relations of the epidermis should be treated as distinct from those of the leaf as a whole was originally based on evidence of remark ably rapid changes in leaf thickness (mainly epidermal) and stomatal conductance following changes in rates of water supply to, or vapour loss from a leaf (Meidner, 1951, 1965). Added to this was evidence of relatively heavy accumulations in epidermal walls, especially near open stomatal pores, of solutes fed into the transpiration stream (Tanton and Crowdy, 1972 ; Sheriff and Byott, 1976) ; comparatively high hydraulic conductivities of epidermal tissues (Sheriff and Meidner, 1974, 1975); and evapor ation rates from epidermal strips adequate to account for moderate transpiration rates (Meidner, 1975, 1976). As a further test of the hypothesis a physical model of a substomatal cavity was constructed and vapour losses from specific sites within this cavity were determined. MATERIALS AND METHODS Anatomical measurements of substomatal cavities in different leaf species show that in a generalized model of such cavities the dimensions of the different parts might be related to each other as shown in Fig. 1. In this model it has been assumed that the enlarged area of mesophyll cell walls, due to their convoluted shapes, is offset by the long gaseous diffusion paths between the cells so that the inner surface of a hemisphere surmounting a circular flat area can be taken as representative of the cell surfaces supplying water vapour to the substomatal cavity. It should be noted that the flat area, corresponding to inner epidermal walls, accounts for one-third of the total cavity surface in this model. The flat area was made from a 4-mm-thick piece of perspex with a 3-mm-diameter hole drilled in its centre. The plastic hemisphere was sealed on to this plate with a mastic seal. The interior of the assembly was uniformly coated with perspex cement in order to obtain equal properties of ail surfaces. Pieces of filter paper were fitted into the hemisphere and on to the plate ; they were painted on one side and along their edges with na.il varnish so as to make them non absorbent on one side and along their different length margins. One junction of a copper