To characterize conductive fracture networks, geologists use judgement to categorize their observations of geological features into sets. To test this judgement, we propose mathematical models which relate, via parameters, tallies of sets of observations in core to transmissivity measurements made in boreholes. We show that if these models are applied to aquifers in which groundwater flow is predominantly horizontal, the major sources of error are the misalignment of core relative to hydraulic test intervals and the erroneous categorization of observations. We tallied up (1) core observations that are categorized on the basis of a descriptive code given at time of drilling and (2) breaks in core that were later categorized on the basis of their probability of being permeable, and we use these tallies to find least squares parameter estimates from the models applied to a set of measured transmissivities. We evaluated the success of each method to distinguish the most transmissive fractures from the least from the goodness of fits as well as from the consistency of the parameter estimates with the understood hydrogeologic role of the constituent members of each set. It was found that highly transmissive fractures in bedrock could be identified by inspection of core and that the skilled judgment used by geologists for this purpose was better encapsulated in permeability rankings than in descriptive codes and written comments.