The in vitro activities of ticarcillin (T) and carbenicillin (C) against 253 strains of Pseudomonas aeruginosa (108 randomly selected clinical isolates and 145 isolates determined to be C resistant by 10 referring hospitals) were evaluated using both broth microdilution and disk diffusion techniques. T activity was usually twofold or greater than C activity for individual strains in both sample populations. The minimal inhibitory concentration (MIC) values of T and C were </=0.4 to >400 (median, 25) and 0.8 to >400 (median, 50) mug/ml, respectively, for the random sample and 1.6 to >400 (median, 50) and 12.5 to >400 (median, 200) mug/ml, respectively, for the select sample. For predicting susceptibility (MIC, </=50 mug of T and </=100 mug of C per ml), 75-mug T and 100-mug C disks were equivalent in accuracy for the random sample (T/T = 92.6%, T/C = 92.0%, C/T = 91.7%, C/C = 89.8%). Regression-line analyses of zones of inhibition versus MIC values for the four combinations showed coefficients of correlation to be -0.81, -0.76, -0.77, and -0.75, respectively. With the sample selected as being resistant to C, T was better than C for predicting T susceptibility (P < 0.05), and T was equivalent to C for predicting C susceptibility (T/T = 84.3%, T/C = 81.0%, C/T = 73.6%, C/C = 80.4%). Coefficients of correlation were -0.70, -0.74, -0.55, and -0.61, respectively. T and C disks were, therefore, equivalent for determining T and C susceptibilities, but the T disk was more reliable for determining T susceptibility in a population skewed with C-resistant strains.