Conventional pumping test analysis methodology assumes that aquifer transmissivity is invariant in space. The ramifications of this assumption are examined for hypothetical units whose variations in transmissivity are considered to be reasonable representations of variations possible in natural systems. The dependence of pumping test transmissivity on spatial (angular and radial) and temporal location of observations and the method of drawdown analysis is assessed. The dependence on the angular position of an observation well appears of little significance. Dependence on radial position, however, can be strong. The dependence on the interval of time used in the pumping test analysis is only important in highly variable systems. Methods of drawdown analysis that yield identical estimates in uniform units yield differing estimates in nonuniform ones as a result of a difference in data-fitting procedures. A reasonable estimate of transmissivity at a regional scale can be obtained from a Cooper-Jacob analysis applied at a large duration of pumpage. In general, conventional approaches for pumping test analysis should be viable in the nonuniform aquifers of the type considered here.