Social science researchers often formulate hypotheses (or theories) using order constraints between the parameters of interest. In a regression model for example one may expect that the relative effect of the first explanatory variable on the dependent variable is larger than the relative effect of the second explanatory variable on the dependent variable, which, in turn, is larger than the relative effect of the third explanatory variable on the dependent variable. Testing such order-constrained hypotheses using classical p-values or classical model selection criteria can be problematic. In this talk I show that the Bayes factor, a Bayesian criterion for model selection and hypothesis testing, is very useful for this testing problem. This is discussed when testing order constraints on regression coefficients in a linear regression model and when testing order constraints on bivariate correlations in an unstructured correlation matrix.