Successive inventories of a silvicultural experiment in terra firme rain forest within the Tapajós National Forest in the Brazilian Amazon are examined to provide guidelines for operational forest management on a sustainable basis. The experiment was logged in 1979 without additional silvicultural treatment, but included protection from further logging and encroachment (`log and leave'). Thirty six permanent plots established in 1981 were remeasured in 1987 and 1992. Logging changed the canopy structure and altered the composition of the stand, reducing the number of shade tolerant species and stimulating light demanding species. There was a net increase in stem number and stand basal area during the 11 year observation period, and this trend also holds for most of the individual species. The stand basal area 13 years after logging was about 75% of that in a comparable unlogged forest. Logging stimulated growth, but this effect was short lived, lasting only about 3 years, and current growth rates are similar to those in the unlogged forest. Between the first and second remeasures, average diameter increment decreased from 0.4 to 0.2 cm year-1, mortality remained relatively constant at 2.5% year-1, while recruitment (at 5 cm diameter at breast height) decreased from 5 to 2%. Total volume production declined from approximately 6 to 4 m3 ha-1 year-1, while commercial production remained about 0.8 m3 ha-1 year-1. New commercial species increased the commercial volume in 1992 from 18 to 54 m3 ha-1, and the increment to 1.8 m3 ha-1 year-1. Results from this experiment provide the first quantitative information for management planning in the Tapajós Forest, and may guide the choice of cutting cycle and annual allowable cut. Silvicultural treatment to stimulate growth rates in forest areas zoned for timber production should be considered as a viable management option. Extrapolations of these results to an anticipated 30-35 year cutting cycle must be interpreted with caution. Ongoing remeasurement and analysis of these and other plots over the next 30 years or more are necessary to provide a stronger basis for management inferences.