During the intensive observation period of the Tropical Ocean-Global Atmosphere/Coupled Ocean-Atmosphere Response Experiment, a rain-formed barrier layer was observed in a R/V Franklin survey. Currents measured with an acoustic Doppler current profiler were mainly southward. Winds were predominantly westerly with an average speed of 15 knots (7.5 m s-1), but they occasionally reached 25--30 knots (12.5--15 m s-1), during strong tropical storms that yielded 20--80 mm of rain. Wind mixing was active during the survey. The isothermal layer deepened from 25 to 70 m in 5 days. Sea surface water diluted by rainfall penetrated deep under wind forcing through turbulent mixing and entrainment. The diluted water was strongly stratified in salinity with a vertical salinity change of 0.1--0.15 practical salinity units, but it had a temperature change (~0.1 ¿C) close to that of the isothermal layer. As a result, the halocline was shallower than the thermocline and a 10-m-thick barrier layer existed between the two. A barrier - layer is defined as the vertical distance difference between a halocline and a thermocline, in which there is very little temperature change but a large salinity change. Thus the observations suggest that tropical rainfall has a greater impact on salinity than temperature. The descending low-salinity water is slightly warmer during daytime and slightly colder during nighttime, reflecting a link with the diurnal cycle of solar radiation. I propose a mechanism for the formation of a rain-induced barrier layer. When the temperature in the descending dilution water has been mixed to the same level as the environmental temperature, the salinity is mixed more slowly, so that a salinity difference exists between the dilution water and environmental water. Thus more time is required to reduce this rain-induced salinity difference compared to the temperature, which is the cause of the barrier layer. A one-dimensional, time-dependent, rain-formed, barrier layer model is thus developed through integration of a set of one-dimensional equations of temperature, salinity, and turbulent mechanical energy. The model shows that a rain-formed barrier-layer is sensitive to many atmospheric inputs, such as evaporation minus precipitation, surface heat flux and wind forcing at the sea surface. The model proves that with both a small amount of warming but strong freshening and a small amount of cooling but strong freshening, a rain-formed barrier layer can be produced. ¿ 1998 American Geophysical Union