Purpose: This study explored relationships among perceptual, ultrasound, and acoustic measurements of children's correct and misarticulated /r/ sounds. Longitudinal data documenting changes across these parameters were collected from 2 children who acquired /r/ over a period of intervention and were compared with data from children with typical speech. Method: Participants were 3 children with typical speech, recorded once, and 2 children with /r/ misarticulation, recorded over 7-8 months. The following data from /r/ produced in nonwords were collected: perceptually rated accuracy, ultrasound measures of tongue shape, and F3-F2 distance. Results: Regression models revealed significant associations among perceptual, ultrasound, and acoustic measures of /r/ accuracy. The inclusion of quantitative tongue-shape measurements improved the match between the ultrasound and perceptual/acoustic data. Perceptually incorrect /r/ productions were found to feature posteriorly located peaked tongue shapes. Of the children who were seen longitudinally, 1 developed a bunched /r/ and 1 demonstrated retroflexion. The children with typical speech also differed in their tongue shapes. Conclusion: Results support the validity of using qualitative and quantitative ultrasound measures to characterize the accuracy of children's /r/ sounds. Clinically, findings suggest that it is important to encourage pharyngeal constriction while allowing children to find the /r/ tongue shape that best fits their individual vocal tract. (Contains 12 figures.)