Speech inversion holds much potential to describe speech errors in childhood speech sound disorders. However, the clinical interpretability of speech inversion tract variables is unknown. This study is the first to show, through linear mixed modeling, that acoustic-to-articulatory speech inversion can quantify statistically significant articulatory differences between several perceptually salient subtypes of /ɹ/ and /s/ speech sound errors, and correct /ɹ/ and /s/ targets, in American English.