Time fluctuations of steps are investigated as a function of temperature on Cu(11n) surfaces with n = 13, 19, and 79. The time dependence of the step position correlation function is characterized by a t(alpha) power law. The exponent alpha decreases from about 0.4 to 0.25 as the magnitude of the fluctuations increase with the temperature. Monte Carlo simulations demonstrate that this change in the exponent cu marks a transition from uncorrelated kink motion to correlated kink motion. The correlation is a consequence of the restriction of the mass transport to the step edge. The correlation becomes effective when more than a single kink passes the scan line during the time span of observation. The temperature dependence of the correlation function is evaluated using the macroscopic Langevin model. The diffusion coefficient for the mass transport along the step edge is calculated. The activation energy of this diffusion coefficient is quantitatively consistent with the microscopic activation energies which have been obtained recently by an analysis of the probability for the intervals between step jumps.