Word recognition performance was measured for 18 normal-hearing subjects using the female talker version of the Northwestern University Auditory Test No. 6 (NU-6) in the presence of three background competitors: (1) a meaningful multitalker competing message consisting of three male and three female talkers (forward multitalker competing message [FCM]), (2) the same multitalker competing message recorded in reverse to eliminate semantic content (backward multitalker competing message [BCM]), and (3) an amplitude-modulated speech-spectrum noise (SSN) having the same long-term average spectrum and amplitude fluctuations as the meaningful multitalker competing message. The meaningful competitor had a significantly more deleterious effect on recognition performance compared to performance for the two nonmeaningful competitors. Furthermore, the nonmeaningful speech competitor produced a significantly greater degradation in performance than that for the SSN.