A long-term field trial, with Boro – GM – T.Aman cropping system, was conducted at the Bangladesh Rice Research Institute (BRRI) Farm, Gazipur (lat. 238590N, long. 908240E, 30 m above mean sea level) during the period 1993 – 1999. The average temperature ranges from 7.28C in winter to 36.78C in summer. The mean annual rainfall is about 2000 mm. The soil of the experimental field has a silty clay loam texture (sand 21%, silt45%, and clay 35%) and a slightly acidic pH (6.6). The other soil parameters were as follows: CEC 23 cm ol kg21soil,exchangeable calcium (Ca) 7.16 cmol kg21 soil, exchangeable magnesium (Mg)1.99 cmol kg21soil, exchangeable potassium (K) 0.16 cmol kg21 soil, organic carbon (C) 12.8 g kg21, total nitrogen (N) 1.0 g kg21, available phosphorus (P)(modified Olsen’s) 6 mg kg21, available sulfur (S) [0.01 M CaH2(PO4)2extraction]14 mg kg21, and available zinc (Zn) (DTPA extraction) 2.69 mg kg21.The experiment was laid out in a modified split-plot design with three replications. The treatments for the first crop of the cropping pattern (Boro, BRRIdhan29) were no fertilizer (T1); fertilization following the Bangladesh Agricultural Research Council (BARC) fertilizer-recommendation guide (Anonymous1987) for medium-yield goal (MYG) for the particular area (T2); soil-test-based (STB) fertilizer recommendation for high-yield goal (HYG) (T3); T3þCD at the rate of 5 t ha21on an oven-dry basis (T4); and local farmers’ practice(T5). Dhaincha (S. aculeata) was the second crop (Kharif I) grown as a GM.In T. Aman, the third crop in the cropping sequence, each original plot under treatments T2,T3, and T4 was divided into four subplots. The treatments for these subplots were full inorganic fertilizer doses as for the first crop without GM (T2a,T3a,T4a), full inorganic fertilizer doses plus GM (T2b,T3b,T4b),60% N and 50% other nutrient rates of full inorganic fertilizer doses plus GM (T2c,T3c,T4c), and 60% N only plus GM (T2d,T3d,T4d). The sources of N, P, K, S, and Zn were urea, triple superphosphate, muriate of potash, gypsum, and zinc sulfate, respectively. Fertilizers were applied to each crop (except dhaincha). In treatment T4, CD at 5 tha21 (oven-dry basis) was applied once a year before transplanting Boro rice. One third of the N and all of the P, K, S, and Zn were applied at the time of final land preparation in both Boro and T. Aman seasons. The remaining two-thirds of the N was applied in two equal installments: 25 – 30 days (Boro) and20 – 25 days (T. Aman) after transplanting and 7 days before the panicle initiation stage in both seasons. The variety of Boro was BRRI dhan29 in all the years of the experiment; for T. Aman, BR 11 was used in 1993 – 1998 and BRRI dhan31was used in 1999. Three or four 45-day-old (Boro) and 30-day-old (T. Aman) seedlings were transplanted in hills 20 cm apart and in rows 20 cm apart.The dhaincha (S. aculeata) was grown as GM in appropriate plots between Boro and T. Aman seasons. Seeds were sown by broadcasting at a rate of 50 kgha21in the first week of May. Fifty-five-day-old dhaincha plants (10 – 12 t ha21on a fresh-weight basis) were incorporated 7 – 8 days prior to planting T. Aman. Appropriate cultural and management practices including plant-protection measures were followed during each growing season. The plot size was4m4 m. The crops were harvested at maturity from a 2.5 m2 m area.Plant height from 20 randomly selected plants in each plot and panicle number from 16 random hills per plot were recorded at maturity in the first3 years (1993 – 1995) of the experiment. Grain yields (14% moisture) and straw yields (oven-dry basis) were recorded in each growing season of the experiment. In the last four years (1996 – 1999) of the experiment, a portion of straw and grain samples in each growing season were analyzed. Straw and grain samples were oven-dried at 708C+58C for 3 days and then ground in a Willey Mill. These samples were analyzed for P, K, S, and Zn content by digesting with a di-acid mixture of nitric and perchloric acid at the ratio 5:2 following the method described by Yoshida et al. (1976) and N by micro-Kjeldahl distillation method (Yoshida et al. 1976).Initial composite soil samples from the two layers (0 – 15 cm and 16 – 30 cm deep) from the 30 spots of the main field were collected prior to fertilizer application in the first crop (Boro) of 1993. After harvesting the 21st crop (T. Aman) of 1999, composite soil samples from the two layers (0 – 15 cm and 16 – 30 cm deep) from the six spots of each of the experimental plots were collected. The soil samples were air-dried, ground to pass through a 2-mm sieve, and analyzed for texture (Day 1965), pH (1:2.5) (Jackson 1962), organic carbon (C) by the black and Walkley method (Walkley and Black 1965), total N by the micro-Kjeldahl distillation method (Bremner 1960), available P by the modified Olsen’s method (Watanabe and Olsen 1965), exchangeable K by the 1 Nammonium acetate (pH 7.0) method (Pratt 1965), available S by 0.01 M Ca(H2PO4)2extraction (Hunter 1984) and available Zn, available Fe, available Cu, and available Mn by DTPA extraction (Pratt 1965). Statistical analyses were performed as a randomized complete block design, because it was not a full split-plot (the control and the farmers practice plots were not split), and means were compared by a least significant difference (LSD) test. Economic analyses were done for net benefit and benefit–cost ratio for different treatment combinations (Saha et al. 1998).The “partial” nutrient balance, including only major inputs (fertilizer, nutrient content in irrigation water, biological nitrogen fixation, etc.) and major outputs (nutrient removal by crops) were considered.