High mobility group box 1 (HMGB1) has been implicated as a late mediator in sepsis. We here sought to determine the extent of HMGB1 release in patients with sepsis stratified to the three most common infectious sources and to determine HMGB1 concentrations at the site of infection during peritonitis or pneumonia. Observational studies in patients and healthy humans challenged with lipopolysaccharide. Three intensive care units and one clinical research unit. Three patient populations were studied: 1) 51 patients with sepsis due to pneumonia (n = 29), peritonitis (n = 12), or urinary tract infection (n = 10); 2) 17 patients with peritonitis; and 3) four patients with community-acquired pneumonia. In addition, eight healthy subjects were studied after intravenous injection of lipopolysaccharide (4 ng/kg). One population of healthy volunteers received lipopolysaccharide intravenously. Patients with severe sepsis due to pneumonia displayed elevated circulating HMGB1 concentrations at both days 0 and 3 after inclusion. Patients with sepsis due to peritonitis had elevated HMGB1 levels at day 0 but not at day 3, whereas urinary tract infection was associated with a delayed HMGB1 response, with elevated levels only at day 3. HMGB1 concentrations did not differ between survivors and nonsurvivors and were not correlated to either disease severity or concurrently measured cytokine levels. In line with these observations, although intravenous lipopolysaccharide injection clearly elevated plasma cytokine levels, HMGB1 remained undetectable. In patients with peritonitis, HMGB1 concentrations in abdominal fluid were more than ten-fold higher than in concurrently obtained plasma. In pneumonia patients, HMGB1 levels were higher in bronchoalveolar lavage fluid obtained from the site of infection than in lavage fluid from healthy controls. In severe sepsis, the kinetics of HMGB1 release may differ depending on the primary source of infection. In patients with severe infection, HMGB1 release may predominantly occur at the site of infection.