The study investigates the applicability of duality theory and flexible functional forms to small scale Javanese agriculture. Multi-product restricted translog profit functions and cost functions are empirically tested against 1980 household survey data. The focus is given to the majority of farmers growing rice and corn;Profit function models turn out to be less appropriate, which persistently give negative supply elasticities for rice output. Several plausable reasons include, among others, (i) the existence of non-decreasing returns to scale of the underlying technology, (ii) the dominance of regulated rice and fertilizer markets, and (iii) the prevalence of objectives other than economic profit among the farm households. In addition, the use of implicit price data inherently commit some conceptual weaknesses and measurement errors;The cost function, which embodies weaker assumptions than the profit function, produces plausible results. Several hypothesis tests reveal (i) the rejection of homotheticity, and which also preclude the use of more restrictive models of homothetic and homogeneous functions, including Cobb-Douglas forms, (ii) the existence of output jointness for rice and corn, lending no legitimate support for single-product model for the multi-crop farms, and (iii) some supports for the homothetically separable rice and corn outputs, indicating the validity of an aggregate output model;The econometric estimation for the multi-product non-homothetic translog cost function indicates that (i) the farmers are significantly responsive to input price changes, though quantitatively modest and inelastic, (ii) the major inputs of fertilizer, hired labor, bullock and seed are net substitutes (except seed and bullock), and (iii) there exists some unutilized production capacity, implying that outputs can still be expanded using less than proportional variable costs;Several directions for model improvement and empirical estimation are suggested.