The Backus-Gilbert method is used to evaluate the quality of estimates of sound speed anomaly and (two) current components obtained from acoustic travel time data in a horizontal ocean slice. The approach is more flexible than previous ones in that reciprocal transmissions are not required and, if present, are not rigidly coupled by addition and subtraction. Nine categories of bias error are identified. They are associated with imperfect resolution of the three unknowns and with cross contamination between them. Measures of these errors are developed that are system properties, as are stochastic error and resolution length. Bias error is calculated in numerical experiments and is combined with stochastic error to obtain mean error. The various quantities contributing to the mean error can be selectively reduced and brought into balance by trading off resolution. It is shown that it is possible in principle to estimate currents from one-way transmission data, and results suggest that further study is merited. It is shown that array design can incorporate a mixture of reciprocal and nonreciprocal transmissions and that equal weights for reciprocal data pairs are not always optimal.