A map of faults in a 60 km 2 area of the southern North Sea has been produced from three-dimensional seismic data. The faults shown on the map obey power-law cumulative-frequency distributions for throw (power-law exponent, D, nearly equal 2.7) and length (D nearly equal 1.1). Simulations have been carried out to correct for sampling biases in the data and to make predictions of the throw and length scaling characteristics of the faults. The most important bias is caused by poor resolution of the small displacement tip zones of faults. The raw data show considerable scatter in their length:throw ratios, but they more closely fit a linear relationship if a length of 500 m is added to each fault, thereby making up for the zones near the fault tips with throws ( nearly equal 15 m) below seismic resolution. Further variability in the data may be caused by such geological factors as fault interaction. Tip lengths have been extended to simulate the actual fault pattern in the study area. Maps produced by this procedure can be used to estimate the true connectivity of the fault network. Extending the faults results in greater connectivity than shown by the raw data, which may cause greater compartmentalization of the rock mass. This greater compartmentalization has implications for hydrocarbon exploitation if the faults are sealing. A problem with the model, however, is that it does not deal effectively with the interaction of subparallel, noncoplanar faults. To test the reliability of the procedure, we analyzed exposure-scale faults in Somerset, United Kingdom, where the tips are well constrained. Both length-throw relationships and map-pattern connectivity for the simulated fault networks agree closely with the actual data.