Two-factor field experiments during 1993 and 1994 with sharply contrasting wet seasons were carried out for in situ water conservation and nutrient use in rainfed lowland rice (Oryza sativa L.) grown on a sandy soil. The main plots compared subsoil compaction with shallow dry tillage (C1), subsoil compaction with deep dry tillage (C2), and shallow dry tillage without compaction (C0). Five mineral fertilization techniques were used as subplot treatments: no fertilizer (control, F0), 40-13-25 kg NPK/ha in two applications of conventional fertilizer (F1), 40-13-25 kg NPK/ha using slow-release fertilizer (F2), 80-26-50 kg NPK/ha in two applications of conventional fertilizer (F3), and 80-26-50 kg NPK/ha using slow-release fertilizer (F4). Soil was compacted with 10 passes of a vibrating road roller on 15 May 1993. Rice seedlings (cv. RD6) were transplanted in all plots. Compaction increased total weeks with surface water accumulation from 3.7 and 2.4 wk in C0 to 11.0 and 14.3 wk in C1, and 11.7 and 14.9 wk in C2 for the 1993 and 1994 wet seasons, respectively. Because three irrigations were applied during the 1993 WS, drought stress was far more severe during the nonirrigated 1994 WS. With the single exception of the effect on rice grain weight in the 1994 WS, no significant interaction between the main factors (compaction) and subfactors (fertilization) was observed. Although the highest grain production was also harvested in compacted plots in the 1993 WS, grain yield increase because of subsoil compaction was found to be significant (P <5%) in the 1994 WS only. Grain yield in C2 plots was 2.2 t/ha versus 1.8 t/ha in C1 plots and 0.8 t/ha in C0 plots. The difference in grain yield between C1 and C2 was also statistically significant. Effects of mineral fertilization were significant in the 1993 WS only. The F4 slow-release fertilizer-based treatment gave a grain yield of 3.2 t/ha compared with 1.4 t/ha in the F0 plots. Corresponding values for F1, F2, and F3 were 1