Study area and experimental design: This study was carried out in nine farmers’ rice plots with different sizes ranging from 12.5 to 25.0 decimal in Sadar Upazila (sub district) of Kushtia district (23° 55′ 11.48″ N latitude, 89° 13′ 12.11″ E longitude), Bangladesh for a period of 4 months from August to November, 2011. Nine plots were grouped into three different ditch sizes in triplicate: 5%, 10% and 15% of the total cultivable land area, herein designated as Ditch- 5%, Ditch-10% and Ditch-15% respectively. Mono-sex tilapia fingerlings were stocked in all plots with a same density of 5000 ha-1. Pre-stocking preparations: After harvesting of rice, the rice plots were left fallow for a period of 10–15 days during the interval of two subsequent crops. During this period, embankments with a 1.5 m height and 1.0 m width were constructed for all the rice fields used for the experiment. During ditch construction, 5%, 10% and 15% area of the cultivable land for Ditch- 5%, Ditch-10% and Ditch-15%, respectively, were converted with a common depth of 1.0 m for all the plots. The ditches were constructed at the lower part of the fields to provide refuge to fish during high water temperature and low water depth with a temporary dyke with the height 0.3 m. After 5 days of water filling by rainfall and minor irrigation, liming was performed in all the ditches at a rate of 125 kg ha-1. Seven days later, the ditches were fertilized with cow manure at the rate of 500 kg ha-1, urea and triple super phosphate (TSP) at the rate of 25 kg ha-1 both. The rice plots were ploughed for three times before transplanting of rice seedling, and then levelled properly by a ladder and stubbles from the previous crop were incorporated into the soil. An initial dose of TSP, muriate of potash (MP) and Zipsam were applied at the rate of 52, 82, and 60 kg ha-1, respectively, and mixed with the soil during the time of last ploughed. After 3 days of fertilization, the rice seedlings were transplanted in alternate row spacing and the row to row distances were maintained at 25 cm, whereas the cluster to cluster distances were maintained at 20 cm for the seedlings. After preparing all the plots properly selected aman rice (BR-11) seedlings were planted. The rice fields were fertilized with urea, TSP and MP with a common dose of 200, 100 and 20 kg ha-1, respectively, in two instalments during 25 days after transplantation (DAT) and 55 DAT. Besides this tilapia fingerlings were stocked during 15 DAT. Stocking of fish fingerling and sampling: Mono-sex tilapia fingerlings were purchased from a private nursery and transported to the study site by oxygenated polyethylene bags. The bags were kept in ditch of rice plots about 30 min for temperature adjustment and then fingerlings were gradually released. Initial weights (g) were recorded for 10% of the stocked fingerlings. Monthly random sampling was done using a fine-meshed seine net to check fish growth and health by measuring weight of 10% of total number of stocked fish. Monitoring of water and sediment quality parameters: The water quality parameters such as temperature, water transparency (Secchi depth), pH, dissolved oxygen (DO), ortho-phosphate, ammonia- nitrogen and alkalinity were monitored fortnightly within 10:00 hours to 11:00 hours in the ditch of rice field. Water temperature was recorded with a Celsius thermometer (°C), water transparency was recorded with a Secchi disk (cm), pH and DO (mg L-1) were measured using a digital pH (PSH-J4; company: Rex, Suzhou, China) and DO meter (DO-5509; Lutron, Electronic Enterprise Co. Ltd., Taipei, Taiwan) respectively. Other parameters were measured using a HACH kit (model FF-2, No. 2430-01; Loveland, CO, USA). Sediment samples were collected twice (i.e. before stocking and after final harvest) from five locations (four corners and centre of the rice field) from each of the rice field following the procedures described by Bangladesh Soil Research Institute (BSRI). Sediment samples were analysed for pH, organic matter total nitrogen and available phosphorous. Fish and rice harvesting: Fish were harvested followed by rice harvesting after 4 months of DAT. Rice harvest was performed manually with the use of hand harvesting tools using sickles (kanchi), consisting of a wooden handle and a knife blade. The paddies were carried to yard and rice was separated with a power thresher and the amount (kg) of rice and straw were recorded. Fish were harvested by repeated netting and finally by dewatering the ditches. At harvest, fish from each plot were counted and weighed individually. During fish harvesting, gross yield was calculated as the sum of individual weights of harvested fish; net yield was calculated as gross yield minus the biomass of fish stocked. Specific growth rate (SGR) of fish was calculated. Benefit–cost analysis: At the end of the experiment, fish were sold to a local market. Benefit–cost analysis of different treatments was calculated on the basis of the prices of used fertilizer, fish seed and the income from the sale of tilapia fish, rice and straw. Statistical analysis: Water quality, yield parameters (growth, yield, SGR and survival) and economic performance were analysed by one-way ANOVA. Sediment quality parameters were compared between before and after rice–fish culture by paired t-test and sediment quality parameters were compared by one-way ANOVA with ditch size of rice plots as main factor. When a main effect was significant, the ANOVA was followed by Duncan Multiple Range Test (DMRT) at 5% level of significance. The percentages and ratio data were analysed using arcsine transformed data. All analyses were performed using SPSS (Statistical Package for Social Science) version 15.0 (IBM Corporation, Armonk, NY, USA).