Membrane proteins are much more variable than lipid molecules in their mobility and some appear to move freely within the lipid bi-layer whereas others are constrained, often because they are anchored to protein complexes located adjacent to one side of the membrane or the other. Particularly convincing evidence for the mobility of at least some membrane proteins has come from cell fusion experiments. 
In these studies, David Frye and Michael Edidin took advantage of two powerful techniques, one that enabled them to fuse cells from two different species and another that made it possible for them to label, or tag, the surfaces of cells with fluorescent dye molecules. Their approach was to fuse human and mouse cells and identify human-specific and mouse-specific membrane proteins on the surface of the fused cells by tagging them with fluorescent antibodies that had been complexed with either of two different kinds of fluorescent dyes. They then observed the fused cells with a fluorescence microscope to see what happened to the plasma membrane proteins that served as markers for the two different parent cell types.
The mobility of membrane proteins can be shown experimentally by the mixing of membrane proteins that occurs when cells from two different species (mouse and human) are fused and the membrane proteins are tagged with two sets of fluorescent antibodies, one specific for mouse membrane proteins and linked to a green dye, and the other specific for human membrane proteins and linked to a red dye. Proteins begin to mix within a few minutes and are almost completely mixed after 40 minutes.
The antibodies needed for this kind of experiment were prepared by injecting small quantities of purified membrane proteins from either human or mouse cells into separate experimental animals. (Rabbits or goats are commonly used for this purpose.) The animals responded immunologically to the foreign protein by producing antibodies that reacted specifically with the membrane proteins used in the injections. After the antibodies had been isolated from the blood of the animal, fluorescent dyes were covalently linked to them so that the protein/antibody/dye complexes could be visualised with a microscope.
Frye and Edidin tagged human and mouse plasma membrane proteins with antibodies that had dyes of two different colours linked to them so that the two types of proteins could be distinguished by the colour of their fluorescence. The mouse cells were reacted with mouse-specific antibodies linked to a green fluorescent dye called fluorescein, whereas the human cells were reacted with human-specific antibodies linked to a red fluorescent dye, rhodamine.
To fuse the human and mouse cells, they treated them with Sendai virus, an agent known to cause fusion of eukaryotic cells, even those from different species. The fused cells were then exposed to the red and green fluorescent antibodies and observed by fluorescence microscopy. When the fused cells were first exposed to the antibodies, the green fluorescent membrane proteins from the mouse cell were localised on one-half of the hybrid cell surface, whereas the red fluorescent membrane proteins derived from the human cell were restricted to the other half. 
In a few minutes, however, the proteins from the two parent cells began to intermix and after 40 minutes the separate regions of green and red fluorescence were completely intermingled. If the fluidity of the membrane was depressed by lowering the temperature below the transition temperature of the lipid bi-layer, this intermixing could be prevented. Frye and Edidin therefore concluded that the intermingling of the fluorescent proteins had been caused by lateral diffusion of the human and mouse proteins through the fluid lipid bi-layer of the plasma membrane. Compared to most membrane lipids, however, membrane proteins diffuse through the lipid bilayer much more slowly.
If proteins are completely free to diffuse within the plane of the membrane, then they should eventually become randomly distributed. Support for the idea that at least some membrane proteins behave in this way has emerged from freeze-fracture microscopy, which directly visualises proteins embedded within the lipid bi-layer. When plasma membranes are examined in freeze-fracture micrographs, their embedded protein particles often tend to be randomly distributed. Such evidence for protein mobility is not restricted to the plasma membrane. It has also been found, for example, that the protein particles of the inner mitochondrial membranes are randomly arranged. If isolated mitochondrial membrane vesicles are exposed to an electrical potential, the protein particles, which bear a net negative charge, all move to one end of the vesicle. Removing the electrical potential causes the particles to become randomly distributed again, indicating that these proteins are free to move within the lipid bi-layer. The transfer of electrons within the inner mitochondrial membrane depends on random collisions between mobile complexes of membrane proteins.
Although many types of membrane proteins have been shown to diffuse through the lipid bi-layer, their rates of movement vary. A widely used approach for quantifying the rates at which membrane proteins diffuse is fluorescence photo-bleaching recovery. The rate at which unbleached molecules from adjacent parts of the membrane move back into the bleached area can be used to calculate the diffusion rates of various kinds of fluorescent lipid or protein molecules.