Thomas Chamberlin wrote in 1890 of the danger researchers face by developing a “paternal affection” for the first explanation of a phenomenon that gives a reasonable result.  This affection blinds us from considering alternative explanations that may be equally plausible or more appropriate answers to the questions we ask.  Chamberlin’s method of multiple working hypotheses overcomes this problem by urging us to consider various explanations in concert.  
In groundwater modeling, we seek a hydraulic parameter field obtained through calibration of a model to observed data.  Sparse, uncertain data, result in a non-unique solution.  Conventional “unique” solutions are often obtained through parsimony by reducing the parameter field to a reasonable number of homogeneous zones.  The decisions guiding this reduction limit the flexibility of the model, forcing results to conform to the simplified framework.
To consider multiple hypotheses we provide models the flexibility to explore many parameter fields that may imply very different zonations than one might assign a priori. A Bayesian Geostatistical approach is well suited to allow this freedom.  Based on nonconformist Minister Thomas Bayes’ theorem, parameter fields and data are represented by probability distributions.  Prior information is limited to attributes like smoothness or continuity.  The degree of these attributes is informed by the data and changes for different data sets.  As a result, the calibrated parameter field is explicitly conditional on the prior assumptions made about structure and the specific data set used for calibration.  The final result is the most likely solution considering all plausible models given the constraints.  Posterior variance estimates quantify uncertainty and provide a range of likely solutions that are consistent with the data.  Had the Reverand Bayes and Dr. Chamberlin met, we believe they would agree this approach is consistent with both of their goals.