Rigorous back analysis plays a major role in providing information to engineers for better decision-making. Most research on this topic has focused on optimization techniques. Comparative studies of data interpretation methodologies have seldom been reported. Data interpretation methodologies are important since they affect how the results of model prediction are interpreted when field measurements are available. In this paper, two data-interpretation methodologies are described and compared for a hypothetical excavation task. The first approach is error-domain model falsification (EDMF). EDMF is a population-based approach that falsifies parameter values based on error bounds, which are defined to account for modelling uncertainties and measurement uncertainties. The second approach is traditional Bayesian model updating, which is a probabilistic approach that evaluates the posterior distributions of parameter values. Both methodologies are compared under two conditions: (i) known correlations between uncertainties and (ii) unknown correlations between uncertainties. Both methodologies are able to identify the true parameter values when correlation information is known. EDMF yields comparable results as the traditional Bayesian model updating does at a much lower computational cost. Bayesian model updating produces biased results when correlation information is not accurately estimated while EDMF is robust against incomplete information. The study shows that the concept of falsification provides a practical approach to interpret model predictions with field measurements, and it is robust in the absence of information.