Associations are statistically related groups of anomalies, and the VATER association probably is the best known example. The concept is widely used, but there are difficulties with both definition and delineation. An historical review is used to define implicit commonalities of associations. These commonalities are then used to develop a testable model which defines associations as derivatives of nonspecific teratologic events acting on developmental fields. Findings depend on the timing, location, and magnitude of these events, and reflect developmental constraints. This clarifies the nature of associations and provides guidelines for their study and use.