In western societies like the Netherlands and United States the incidence of dog bites is in the range of 8 to 18 per 1,000 people with about 3 of these victims seeking medical help. Behavior tests are used to identify aggressive dogs and expel these from the breeding population or society, but there are uncertainties regarding the usefulness of such tests. The Dutch Socially Acceptable Behavior (SAB)-test has been used since 2001 to select against unwanted aggression and fear in specific dog breeds. To evaluate the usefulness of testing on human directed aggression SAB test results of 345 dogs (479 indoor and outdoor test records), were analyzed to assess the reliability, validity and feasibility of the test. Dogs were labelled aggressive when they had at least one owner-reported bite to human, regardless of the context and the severity of inflicted wounds, and if the dogs were considered to be aggressive by the judges of the Dutch kennel club as a result of displaying a lunge, snap or bite at least once during the SAB-test. Aggressive dogs showed significantly more threats and attacks than did non-aggressive controls. A principal components analysis of detailed observations on 76 dogs grouped bare teeth, snap, bite, growl in one dimension, confirming the test’s capacity to measure an aggressive response. Analysis of 479 test records revealed a sensitivity, specificity and accuracy of 0.33, 0.81 and 0.64, respectively. The low sensitivity may be explained, in part, by our decision to classify dogs as aggressive on the basis of one bite incident only, as reported by a dog’s owner, and by a relative weak capacity of the test to detect specific forms of aggression. Different subtests contributed differently to the provocation of aggression. Typically, attacks by dogs occurred during subtests that involved approach and petting with a dummy hand or doll, and in the second half of the SAB test, i.e., in the absence of the owner. The robustness of the test was investigated primarily by comparing the results on 133 dogs when they were tested outdoors, which is the common practice, with those obtained indoors. The accuracy decreased from 0.67 to 0.62, but overall the indoor test outcomes were similar to those found outdoors. Scores for aggression and fear were significantly higher when dogs were tested for the first time in the morning than the second time in the afternoon, suggesting desensitization. Salivary cortisol concentrations in 20 dogs were not different in samples taken before and after the test, which suggests that the dogs were not under severe levels of stress. The SAB test allows one to evaluate aggression in dogs, but present findings indicate that a considerable portion of aggressive dogs remain undetected, e.g., those that in practice behave aggressively in the absence of fear. Recommendations to increase the test’s usefulness include adding test components that target different forms of aggression and exploiting more detailed information on a dog’s behavior during testing. Formulating a risk assessment based on detailed information instead of simply producing a pass-fail judgement will facilitate a purpose specific us of the SAB-test.