We used population-based data to evaluate how often groups of randomly selected clinical laboratories accurately estimated the prevalence of resistant pneumococci and captured trends in resistance over time. Surveillance for invasive pneumococcal disease was conducted in eight states from 1996 to 1998. Within each surveillance area, we evaluated the proportion of all groups of three, four, and five laboratories that estimated the prevalence of penicillin-nonsusceptible pneumococci (%PNSP) and the change in %PNSP over time. We assessed whether sentinel groups detected emerging fluoroquinolone resistance. Groups of five performed best. Sentinel groups accurately predicted %PNSP in five states; states where they performed poorly had high between-laboratory variation in %PNSP. Sentinel groups detected large changes in prevalence of nonsusceptibility over time but rarely detected emerging fluoroquinolone resistance. Characteristics of hospital-affiliated laboratories were not useful predictors of a laboratory's %PNSP. Sentinel surveillance for resistant pneumococci can detect important trends over time but rarely detects newly emerging resistance profiles.