Large relaxation effects have been observed in many high Tc superconductors and the time dependence of the magnetization is usually found to follow a logarithmic time decay. This is in agreement with the results of simple Monte Carlo simulations which lead to M(t, T, B) = M(O, T, B) [1 - (kT/U)In(1 + t/τ)], where M(O, T, B) is the magnetization at time t = 0, temperature T and magnetic field B, and τ is a relaxation time related to the attempt frequency for hopping of flux lines from one pinning region to an adjacent one. The activation energy, U, depends on both T and B. The purpose of this paper is to discuss some of the difficulties encountered in measurements of magnetization relaxation effects in high Tc superconductors in general and thin films in particular. First, how information about the energy barriers for flux motion may be extracted from experimental data on M(t, T, B) is discussed and experimental evidence is given for the necessity to introduce a distribution of activation energies to reproduce the observed temperature dependence of the relaxation rate, S = -dlnM/dlnt. Second, by means of Monte Carlo simulations, the influence of an initial partially critical state (where the critical current, Jc, flows only in a thick surface layer of the sample) on the time dependence of the magnetization relaxation is studied. The results of these simulations are compared with the predictions of an approximate analytical treatment of the relaxation problem. Third, it is shown that the initial conditions have a crucial influence on the magnetization relaxation. As an example, the effect is considered of a small overshoot during the application of a magnetic field before measuring M(t, T, B), which leads to remarkable micro-creep behaviour. Finally, some data are presented which imply that magnetization relaxation is still large at low temperatures (down to 1.5 K) in good quality epitaxial thin films and the possibility of flux line tunnelling at low temperatures is discussed.@en