For speech-language pathologists

How do you treat the R sound in speech therapy?

Published · Updated

The short answer

Treat /ɹ/ as a motor skill with a flexible target. Bunched and retroflex tongue shapes can both produce a correct /ɹ/, so aim for the sound rather than one shape: a raised tongue front, a retracted tongue root and braced tongue sides. Start in facilitating contexts, shape from sounds the child already makes, and give frequent, detailed feedback that fades as accuracy grows. For residual errors at ages 9 to 15, a 2026 RCT of 108 children found faster progress with visual biofeedback than motor-based treatment alone.

Bunched and retroflex: both are correct

Textbooks describe two tongue shapes for /ɹ/. In a retroflex /ɹ/, the tongue tip points up toward the alveolar ridge or palate. In a bunched /ɹ/, the front of the tongue body is raised and the tip points forward or angles down behind the lower incisors. Both can sound correct, so can many shapes in between, and bunched shapes are more common among American English speakers (Delattre & Freeman, 1968, and Mielke et al., 2016, as reported in Preston et al., 2020).

Imaging research suggests the shapes form a continuum rather than two categories. In Boyce’s review of MRI and ultrasound studies, typical speakers were judged correct whatever shape they used, many switched toward a more bunched shape depending on phonetic context, and fully retroflex /ɹ/ was the rarest type in larger studies (Boyce, 2015).

What correct productions share matters more than the shape (Boyce, 2015; Preston et al., 2020):

The clinical upshot: don’t decide in advance which shape a child should use. Boyce recommends finding each child’s most workable configuration by trial and error and, if progress stalls, teaching a different shape and trying it across contexts (Boyce, 2015). The tutorial adds that retroflex shapes are uncommon after /i/, so a bunched shape may suit words like “ear” and “hear” (Preston et al., 2020). For a parent-facing overview, see the R sound guide.

Vocalic and prevocalic R

/ɹ/ occurs before vowels (prevocalic), as a syllable nucleus (syllabic) and after vowels (postvocalic). “Vocalic R” is the usual clinical name for the syllabic and postvocalic forms. Boyce found the vocal tract shapes for consonantal and syllabic /ɹ/ similar, with consonantal productions looking like stronger, more extreme versions of syllabic ones, and the 2020 tutorial says the articulatory targets, and so the cues, are similar across positions (Boyce, 2015; Preston et al., 2020).

Contexts of /ɹ/ and what the R literature says about them
ContextExamplesNotes
Prevocalic singletonred, rope, rat/ɹæ/ can facilitate, through anticipation of the low tongue body (Preston et al., 2020).
Prevocalic clustertree, drum, green/tɹ/ and /dɹ/ are listed as facilitating contexts (Preston et al., 2020).
Stressed syllabic /ɝ/bird, her, earthProbed and practiced as its own variant in the C-RESULTS trial (McAllister et al., 2020).
Unstressed /ɚ/butter, father, teacherShort and reduced, so it may need separate training after other variants are learned (Preston et al., 2020).
Postvocalic, back vowelcar, star, door/ɑɹ/ facilitates because /ɑ/ shares tongue-root retraction (Preston et al., 2020).
Postvocalic, front vowelear, deer, chairAfter a high front vowel the tongue sides are already braced and lip rounding is held back (Preston et al., 2020).

C-RESULTS grouped /ɹ/ into five variants for practice: syllabic /ɝ/, postvocalic after front and after back vowels, and prevocalic before front and before back vowels (McAllister et al., 2020). Another study targeted each child’s three least accurate vocalic variants at baseline (Preston et al., 2019). Probing every variant first shows which are emerging and which to treat. Early on, the tutorial advises avoiding contexts with /u/ or /o/, and warns that curriculum word lists with uncontrolled contexts can be counterproductive (Preston et al., 2020).

Eliciting /ɹ/: cues, shaping and contexts

The 2020 motor-based tutorial is written for children who produce distortions and have little or no stimulability for /ɹ/, and its strategies were studied from age 6 into adulthood. None needs technology (Preston et al., 2020):

Expect a wide range of response. In a classic study of 65 children shaped from /l/, 70% produced /ɹ/ in isolation within one session, 10% in a second session and 20% not by the end of the study (Shriberg, 1975, as reported in Preston et al., 2020). Judge attempts by ear: the tutorial’s test is whether the production sounds acceptable. See also traditional articulation therapy and the free articulation word lists.

Practice: motor learning and dose

The tutorial applies principles of motor learning in stages (Preston et al., 2020; Maas et al., 2008):

How much practice? No trial has established an optimal dose for /ɹ/, but research protocols are intensive. C-RESULTS acquisition sessions allowed up to 200 syllables in 90 minutes, and its 16 generalization sessions, twice a week, allowed up to 250 trials in an hour. Practice ran in blocks of 10 trials, with difficulty raised after 8 or more correct and lowered after 5 or fewer (McAllister et al., 2020). In another study, sessions were designed for 162 attempts and children completed a mean of about 133 to 139 (Preston et al., 2019). Speech motor chaining offers another structured format for school-age children (Preston, Leece & Storto, 2019).

Fading feedback has small, mixed support: in 12 children with residual /ɹ/ errors, high-frequency ultrasound feedback followed by low-frequency feedback produced larger gains than the reverse order (Preston et al., 2018). See feedback frequency and motor learning, treatment intensity and dosage and improving carryover.

Scheduling. The tutorial favours frequent sessions with many trials and recommends individual sessions for children who aren’t yet stimulable. Group treatment is common, but the evidence for it comes from homogeneous groups only (Preston et al., 2020). See group therapy and caseloads.

The evidence for residual /ɹ/ errors

The RCT’s authors describe it as an efficacy trial under idealized conditions, with a schedule that’s hard to match in typical school settings, no long-term follow-up yet, and American English /ɹ/ only (McAllister et al., 2026; Preston et al., 2026). Both arms improved, so careful motor-based treatment is still a reasonable starting point where biofeedback isn’t available. For timelines, see how long R therapy takes; for younger children saying [w] for /ɹ/, see gliding in speech therapy.

When to consider, or refer for, biofeedback

There are no published referral criteria: McAllister Byun (2017) concluded that more research is needed on which children are better or poorer candidates for biofeedback. These pointers come from how the studies were designed:

ASHA’s Practice Portal lists ultrasound, spectral (visual-acoustic) and tactile biofeedback among speech sound treatment techniques (ASHA). If you don’t have the equipment, a referral to a university clinic or specialist practice that does is one option. For how the methods work, see biofeedback for the R sound; for tools that score /ɹ/ automatically, see automated speech scoring for SLPs.

Questions SLPs ask

Is bunched or retroflex R better?

Neither. Both shapes, and many in between, can produce a perceptually correct /ɹ/. Imaging research suggests finding the shape that works for each child by trial and error, and trying another if progress stalls.

What is vocalic R?

The clinical name for /ɹ/ after a vowel or as the syllable nucleus, as in car, ear, bird and butter. Its articulation is similar to prevocalic /ɹ/, but variants are usually probed separately, and unstressed /ɚ/ may need its own training.

What are facilitating contexts for R?

A 2020 motor-based tutorial lists /ɑɹ/ (as in “are”), /ɹ/ after high front vowels (as in “ear”), /ɹæ/ (as in “rat”) and /tɹ/ or /dɹ/ clusters. It advises avoiding /u/ and /o/ contexts early on.

How many trials should an R session include?

There’s no proven optimal dose. Research sessions are intensive: the C-RESULTS trial allowed up to 200 to 250 trials per session, and another study averaged about 133 to 139.

Does biofeedback work better than traditional R therapy?

In a 2026 RCT of 108 children aged 9 to 15, children acquired and generalized /ɹ/ faster with biofeedback than with motor-based treatment. Both groups improved, and there’s no long-term follow-up yet.

Sources

Join the Early Adopter Program