For speech-language pathologists

How can SLPs improve carryover of speech sounds?

Published 21 October 2026 · Updated 21 October 2026

The short answer

Generalization (carryover) is a planned stage of speech sound treatment, not something that reliably happens on its own. The evidence for specific carryover techniques is limited: self-monitoring outside the clinic led to rapid generalization in small studies from one group, but a replication didn’t find the same effect. Treatment approach may matter too: in a 2026 RCT of 108 children with residual /ɹ/ errors, generalization to untreated words was greater with visual biofeedback.

Where carryover sits in treatment

ASHA’s Practice Portal describes three stages of speech sound treatment:

Establishment: eliciting the target sound and stabilizing it at a voluntary level.
Generalization: carryover of the sound at increasingly challenging levels, from syllables and words to phrases, sentences and conversation.
Maintenance: stabilizing production so it becomes more automatic, with self-monitoring and self-correction.

ASHA frames generalization by linguistic level. Carrying a sound into new settings, partners and spontaneous talk is usually where it’s hardest to see progress.

What the evidence says

In 13 children seen for school therapy twice a week, introducing self-monitoring was followed by correct use of the target sound in spontaneous speech outside the treatment setting (Koegel, Koegel & Ingham, 1986).
In 7 children substituting /θ, ð/ for /s, z/, self-monitoring only inside the clinic didn’t generalize; requiring self-monitoring outside the clinic led to rapid, widespread generalization (Koegel et al., 1988).
A replication with 8 school-age children with /s/ or /r/ errors, who were at least 80% correct in conversation in therapy, did not reproduce those effects (Gray & Shelton, 1992).
In an RCT of 108 children aged 9–15 with residual /ɹ/ errors, both motor-based treatment and treatment with visual biofeedback improved untreated words, but generalization was significantly greater with biofeedback (Preston et al., 2026).

The self-monitoring evidence is small single-case work, mostly from one research group, plus one failed replication. It’s reasonable to try, but it’s fair to call it mixed.

Practical implications

Plan for generalization from the start. For /r/, a 2020 tutorial recommends moving toward a larger, more varied word set and having the child rate their own accuracy once the sound is produced in syllables and words (Preston et al., 2020).
Measure generalization directly with untreated words and conversation samples, rather than inferring it from accuracy on practised items.
If you use self-monitoring, build in practice outside the clinic, where the positive studies found it mattered.
Involve families in short, structured carryover practice. See getting more home practice.

Questions SLPs ask

Is self-monitoring evidence-based for carryover?

There is some evidence from small single-case studies, mainly from one research group, and one replication that didn’t find the effect. Describe it as promising but mixed.

How should I measure carryover?

Probe untreated words and samples of conversation outside the treatment task, so you see generalization rather than performance on practised items.

Does biofeedback help carryover?

In one RCT of 108 children aged 9–15 with residual /ɹ/ errors, generalization to untreated words was significantly greater with visual biofeedback than with motor-based treatment alone. It’s one trial, in older children.