Hydrophobic matching, in which transmembrane proteins cause the surrounding lipid bilayer to adjust its hydrocarbon thickness to match the length of the hydrophobic surface of the protein, is a commonly accepted idea. To test this idea, gramicidin was embedded in dilauroyl phosphatidylcholine (DLPC) and dimyristoyl phosphatidylcholine (DMPC) bilayers at the molar ratio 1:10. The bilayer thickness (PtP) was measured by X-ray lamellar diffraction. In the fluid phase near full hydration, PtP is 30.8 A for pure DLPC, 32.1 A for DLPC/gramicidin mixture, 35.3 A for pure DMPC and 32.7 A for a DMPC/gramicidin mixture. Gramicidin apparently stretches DLPC bilayers and thins DMPC bilayers toward a common thickness as expected by hydrophobic matching. Gramicidin pair correlations were measured by X-ray in-plane scattering. In the fluid phase, the gramicidin-gramicidin nearest-neighbour separation is 26.8 A in DLPC bilayers but shortens to 23.3 A in DMPC bilayers, thus confirming the conjecture that when proteins are embedded in a membrane, hydrophobic matching creates a strain field in the lipid bilayer that in turn gives rise to a membrane-mediated attractive potential between proteins. These results were analysed with an elasticity theory of membrane deformation. The same principle explains the 'concentration-gating' mechanism of pore formation by antimicrobial peptides via the membrane-thinning effect. Concentration-gated pore formation and membrane thinning by alamethicin and magainin have been observed.