Two single-shot localization techniques, STEAM and PRESS, are analyzed with regard to specifications for in vivo localized proton NMR. In particular, attention is paid to optimum signal intensity per unit volume, sensitivity to motion and diffusion, shortest attainable echo time, water suppression and editing possibilities. Experimental results are shown for cat brain at 4.7 T and human brain at 1.5 T. Both STEAM and PRESS are highly effective localization methods. For long echo times, PRESS is the method of choice, because it offers a factor of two gain in signal intensity. In addition, the method is less sensitive to motion and diffusion, and not susceptible to multiple-quantum effects. STEAM offers advantages for observation of (coupled) metabolites with short T2, because (a) shorter TEs can be attained and (b) effective water suppression sequences can be implemented without penalty in echo time. Differences relating to editing possibilities and B1 dependence, possibly important in choosing a method, are discussed.