Lipid rafts in the plasma membrane, domains rich in cholesterol and sphingolipids, have been implicated in a number of important membrane functions. Detergent insolubility has been used to define membrane "rafts" biochemically. However, such an approach does not directly contribute to the understanding of the size and the lifetime of rafts, dynamics of the raft-constituent molecules, and the function of rafts in the membrane in situ. To address these issues, we have developed pulse EPR spin labeling and single molecule tracking optical techniques for studies of rafts in both artificial and cell membranes. In this review, we summarize our results and perspectives obtained by using these methods. We emphasize the importance of clearly distinguishing small/unstable rafts (lifetime shorter than a millisecond) in unstimulated cells and stabilized rafts induced by liganded and oligomerized (GPI-anchored) receptor molecules (core receptor rafts, lifetime over a few minutes). We propose that these stabilized rafts further induce temporal, greater rafts (signaling rafts, lifetime on the order of a second) for signaling by coalescing other small/unstable rafts, including those in the inner leaflet of the membrane, each containing perhaps one molecule of the downstream effector molecules. At variance with the general view, we emphasize the importance of cholesterol segregation from the liquid-crystalline unsaturated bulk-phase membrane for formation of the rafts, rather than the affinity of cholesterol and saturated alkyl chains. In the binary mixture of cholesterol and an unsaturated phospholipid, cholesterol is segregated out from the bulk unsaturated liquid-crystalline phase, forming cholesterol-enriched domains or clustered cholesterol domains, probably due to the lateral nonconformability between the rigid planar transfused ring structure of cholesterol and the rigid bend of the unsaturated alkyl chain at C9-C10. However, such cholesterol-rich domains are small, perhaps consisting of only several cholesterol molecules, and are short-lived, on the order of 1-100 ns. We speculate that these cholesterol-enriched domains may be stabilized by the presence of saturated alkyl chains of sphingomyelin or glycosphingolipids, and also by clustered raft proteins. In the influenza viral membrane, one of the simplest forms of a biological membrane, the lifetime of a protein and cholesterol-rich domain was evaluated to be on the order of 100 micro, again showing the short lifetime of rafts in an unstimulated state. Finally, we propose a thermal Lego model for rafts as the basic building blocks for signaling pathways in the plasma membrane.