Although the self-hearing of a child with defective articulation often has been altered in traditional methods of therapy (Van Riper & Irwin, 1958) , such practices have received little empirical verification. The present experiment was designed to investigate through an operant conditioning paradigm the differences between normal and modified auditory feedback in a sound-learning task for children with defective articulation of the / r / sound. The use of the stimulus-response method within an operant conditioning paradigm permitted consideration of the latency period between stimulus and response. Forty children with the consistent misarticulation of the / r / sound were assigned on the basis of age to one of four cypes of training sessions based on normal or modified feedback; 0to 3-sec. stimulus-response latency period. Each training procedure included 1 0 Ss equally divided between younger children ( 6 through 8 ) and older children (9 through 1 1 ) . During the training sessions a Uher recorder was used to present the taped stimuli consisting of the / r / sound in the prevocalic and posrvocalic positions in the following words: work, worry, far, and run. Each child completed 42 stimulus-response trials in approximately 25 min. During the normal feedback condition, the child heard his own responses without alteration in auditory feedback. During the modified feedback condition, the child heard his own sound production through earphones at a slightly increased intensity level. A lightbulb representing a clown's nose signalled the child's response immediately or 3 sec. following the stimulus. Reinforcement, based on the method of successive approximation, consisted of cereal bits delivered through the mouth of the clown whenever the investigator perceived a change in the child's response closer to that of the stimulus. A panel of 5 clinicians (reliability. 3 9 ) evaluated the recorded responses of each child before and after training on a three-point equal-appearing interval scale. Analysis of variance was used to examine effects of feedback, age, and latency. Mean difference scores show (Fig. 1 ) modified feedback (combined mean difference score, 9.7; SD, 9.3) was more successful than normal feedback (combined mean difference score, -.04; SD. 11.3) in eliciting standard productions of the /r / sound. N o significant differences were found between the conditions of 0-sec. latency and 3-sec. latency, nor bew e e n rhe older and younger children. The results agree with those of Arnold ( 1956) who