The sensitivity of asset fixity conclusions, input adjustment rates, and elasticities to choice of functional form is examined using a dynamic dual model of U.S. agriculture. A very general initial specification allows tests of instantaneous adjustment to be performed for every input. Test results are mixed across functional forms for all inputs except real estate, which is consistently found to be quasi-fixed. Important differences in estimated adjustment rates and elasticities are also found among the functional forms. The translog has higher likelihood support than either the generalized Leontief or normalized quadratic functional forms for this dynamic model specification and data set.