Measured by ultra-slow scanning calorimetry and isothermal circular dichroism, human lung collagen monomers denature at 37 degrees C within a couple of days. Their unfolding rate decreases exponentially at lower temperature, but complete unfolding is observed even below 36 degrees C. Refolding of full-length, native collagen triple helices does occur, but only below 30 degrees C. Thus, contrary to the widely held belief, the energetically preferred conformation of the main protein of bone and skin in physiological solution is a random coil rather than a triple helix. These observations suggest that once secreted from cells collagen helices would begin to unfold. We argue that initial microunfolding of their least stable domains would trigger self-assembly of fibers where the helices are protected from complete unfolding. Our data support an earlier hypothesis that in fibers collagen helices may melt and refold locally when needed, giving fibers their strength and elasticity. Apparently, Nature adjusts collagen hydroxyproline content to ensure that the melting temperature of triple helical monomers is several degrees below rather than above body temperature.