We have used a novel, largely automated, calculation method to refine the NMR solution structure of the pleckstrin homology domain of beta-spectrin. The method is called ARIA for Ambiguous Restraints for Iterative Assignment. The starting point for ARIA is an almost complete assignment of the proton chemical shifts, and a list of partially assigned NOEs, mostly sequential and secondary structure NOEs. The restraint list is then augmented by automatically interpreting peak lists generated by automated peak-picking. The central task of ARIA is the assignment of ambiguous NOEs during the structure calculation using a combination of ambiguous distance restraints and an iterative assignment strategy. In addition, ARIA calibrates ambiguous NOEs to derive distance restraints, merges overlapping data sets to remove duplicate information, and uses empirical rules to identify erroneous peaks. While the distance restraints for the structure calculations were exclusively extracted from homonuclear 2D experiments, ARIA is especially suited for the analysis of multidimensional spectra. Applied to the pleckstrin homology domain, ARIA generated structures of good quality, and of sufficiently high accuracy to solve the X-ray crystal structure of the same domain by molecular replacement. The comparison of the free NMR solution structure to the X-ray structure, which is complexed to D-myo-inositol-1,4,5-triphosphate, shows that the ligand primarily induces a disorder-order transition in the binding loops, which are disordered in the NMR ensemble but well ordered in the crystal. The structural core of the protein is unaffected, as evidenced by a backbone root-mean-square difference between the average NMR coordinates and the X-ray crystal structure for the secondary structure elements of less than 0.6 A.