We describe a multivariate wrapped-Gaussian distribution that can be used to model statistical distributions expressed in periodic angular variables even in the presence of significant angle-angle correlations. Using amino acid and peptide examples, we show how this distribution can be used to characterize accurately conformational fluctuations in torsion-angle space. The method permits more accurate characterization of large-scale peptide and polymer fluctuations than the standard harmonic or quasiharmonic method. This can be used to increase the efficiency of Monte Carlo sampling and simulated annealing in torsion-angle space.