Physicians’ reasons for initiating antibiotic therapy without culture and sensitivity information were studied. A 370-bed, university-affiliated VA hospital was the site of this three-month study. After each empiric initiation of antibiotic therapy, the prescriber was interviewed to determine the rationales for: (1) starting therapy empirically, (2) selecting the specific antibiotic, and (3) choosing the dosage regimen. All expressed rationales were grouped according to common themes. An infectious disease specialist evaluated the prescribers’ decisions by reviewing patient data summaries. Culture and sensitivity reports were recorded as they were received. Thirty-three physicians made 82 empiric antibiotic initiations; a total of 118 rationale themes were mentioned. The most frequently expressed rationale theme was “feel Confident regarding type of infection present, so it is unnecessary to wait for culture and sensitivity report”; 72% of these initiations were confirmed by expert review. The most common prescriber rationale for selecting a particular antibiotic was “past clinical experience” and “antibiotic has broad spectrum,” which was confirmed by expert review 29% and 12% of the time, respectively. The most common rationale themes for choosing a dosage regimen were “literature reference” and “clinical experience.” Crosstabulations of initiation rationale frequencies by infectious disease severity showed that 17 of the 22 “confidence” theme rationales occurred in patients whose underlying disease was considered “least severe.” The rationale of “infection exacerbating underlying disease” appeared mbSt often for the patients judged to have “moderately severe” disease, and the “high-risk infection site” rationale occurred most frequently in patients in the “most severe” disease category. Culture and sensitivity reports were received for 49 of the,82 initiations. Thirty-four (69%) of the expert reviewer’s evaluations agreed with the laboratory reports, while 15 disagreed. These results support cognitively based psychological models to describe physicians prescribing decisions. These models suggest that attempts to influence prescribing should be directed at changing the prescribers’ response to the stimuli to prescribe and beliefs regarding the perceived outcome of drug therapy.