Fifty patients with suspected intraabdominal abscess were investigated prospectively with ultrasound and with 991Tc-hexamethylpropylene-amine oxime (HMPAO) isotope labelled mixed leucocytes, using "'-In tropolonate granulocyte scanning as the reference standard. Twenty five patients had inflammatory bowel disease (three were postoperative): 21 of these had Crohn's disease and four had ulcerative colitis. The remainder comprised nine with postoperative fever and 16 with fever and abdominal pain. An abscess was diagnosed when focal activity on serial 111-In tropolonate and 99m-Tc-HMPOA images at one, three, and 24 hours resulted in activity at least equal to liver activity at 24 hours. Thirteen abscesses were diagnosed using each type ofwhite cell scanning, resulting in 100% sensitivity for 99m-Tc-HMPAO compared with "'-In tropolonate. Bowel inflammation was easily distinguished from abscess on serial images. Eight of these 13 abscesses were detected by ultrasound. Altogether 17 abscesses were found. Ultrasound detected 12, including four liver abscesses which were not purulent and had not been detected by white cell scanning. Ultrasound had a sensitivity of 71% (12 of 17) and a specificity of 87% (33 of 38) using all confirmed abscesses as the reference standard. White cell scanning showed a sensitivity of 76% (13 of 17: as a result of the four non-purulent liver abscesses) and a specificity of 100%. 99m-Tc-HMPAO scanning is as accurate as "'-In tropolonate scanning, and has several advantages including simplicity, availability, superior image quality, and reduced radiation dose. Both methods are more sensitive and specific than ultrasound for intra-abdominal abscess detection but ultrasound is advisable if a neutrophil infiltrate is not suspected. (Gut 1995; 37: 557-564)