Laboratory studies with flows driven by both temperature and salinity differences illustrate the role of partial mixing in such flows. The results are compared to fundamental box model results. In the classic Stommel experiment, an experimental chamber is heated from below and exposed to a steady flux of salt water from above. The chamber is laterally connected to a reservoir of fresh isothermal water. This duplicates in part the well-known box models of deep ocean and estuarine circulation where both temperature and salinity determine the density in the chamber that drives a steady exchange flow between chamber and reservoir. Two locally stable modes of circulation are found within a well-known range of the temperature and salinity forcing: one mode is primarily temperature driven (the T-mode) and one mode is primarily salinity driven (the S-mode), with hysteresis and Stommel transitions between the two modes. New experiments with limited turbulent mixing and a narrow vertical slot in the sidewall between a laboratory test chamber and the reservoir produce a three-layer structure for the S-mode and a two-layer structure for the T-mode. Hysteresis and the Stommel transitions are considerably smaller than in the fully mixed box models. Results are consistent with the notion that the Arctic Ocean is presently in the layered S-mode.