For speech-language pathologists
How to choose target words for articulation therapy
Published · Updated
The short answer
Start with short words that put the target sound in its easiest contexts, include words the child actually uses, and keep a separate set of untreated probe words to measure generalization. Research on word properties is small and partly conflicting, but word frequency, neighborhood density, age of acquisition and nonwords have all affected generalization in treatment studies. A few words can be enough: in one study of 19 children, three minimal pair exemplars led to generalization in 59% of cases.
Word shape, length and facilitating contexts
Context changes how hard a sound is. ASHA describes contextual utilization: some contexts facilitate correct production, so instruction can begin in the syllable contexts where the child already gets the sound right and build from there. For example, /t/ may be easier before a high front vowel, as in “tea”, and nasals before a low vowel, as in “mad” (ASHA).
- Short before long. An NHS parent booklet starts with short words such as “car”, “cow” and “key” before longer ones such as “caravan” (Oxford Health NHS Foundation Trust). Long words are harder: in 93 preschoolers with speech sound disorders, polysyllables showed substitutions, deletions and changes to timing and phonotactics, and the authors note that routine speech accuracy measures aren’t enough to describe them (Masso et al., 2016).
- Position and complexity. Speech motor chaining, for school-age children, raises difficulty step by step from syllables to words, phrases and sentences, and its probe words vary the target’s position and include multisyllable words (Preston, Leece & Storto, 2019).
- For /ɹ/. A 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 as facilitating contexts, advises avoiding /u/ and /o/ contexts early on, and warns that curriculum word lists with uncontrolled contexts can be counterproductive (Preston et al., 2020). See treating the R sound.
Once the sound is reliable in easy contexts, widen the set to harder vowels, positions and longer words so it doesn’t stay tied to one context. The free articulation word lists are sorted by sound and word position.
Frequency, neighborhood density and phonotactic probability
| Property | What it means | What studies found |
|---|---|---|
| Word frequency | How often a word occurs in speech | Frequent words induced more generalization than infrequent ones in one study (Morrisette & Gierut, 2002, as described in Gierut & Morrisette, 2012), but frequency had no clear effect on its own in a later one (Gierut & Morrisette, 2012) |
| Neighborhood density | How many words differ from the word by one sound (Storkel & Hoover, 2010) | Conflicting: words from sparse neighbourhoods generalized more in one study of 8 children (Morrisette & Gierut, 2002, as described in Gierut & Morrisette, 2012); dense neighbourhoods did in another of 8, with frequent words from dense neighbourhoods best (Gierut & Morrisette, 2012) |
| Phonotactic probability | How likely a sequence of sounds is in the language (Storkel & Hoover, 2010) | Typically developing children learned new words with common sound sequences faster (Storkel, 2001); treatment evidence is limited, and Gierut’s 2016 review discusses it mainly in terms of word learning (Gierut, 2016) |
| Age of acquisition | How early children usually learn the word | In 10 preschoolers, late-acquired words induced greater generalization, with an effect four times that of early-acquired words (Gierut & Morrisette, 2012) |
Treat these as leads rather than rules. The studies are small, density is correlated with word length and phonotactic probability so the effects are hard to separate, and the 2012 density study used an adult word database and applies, its authors say, mainly to children whose main difficulty is a severely reduced sound inventory (Gierut & Morrisette, 2012; Gierut, 2016). Gierut’s review recommends high-frequency and late-acquired words as treatment stimuli (Gierut, 2016).
Word learning is also affected. In 20 preschoolers with phonological delays and 34 typically developing peers, both groups formed initial representations for rare sound sequences from sparse neighbourhoods, but only the typically developing children went on to build a full representation of the new word (Storkel & Hoover, 2010).
Tools. Storkel and Hoover built an online calculator for phonotactic probability and neighborhood density from child corpora of American English (Storkel & Hoover, 2010), and the University of Kansas hosts current calculators for both, including an English child corpus (University of Kansas).
Functional and personally meaningful words
Words the child uses and cares about give practice a purpose. ASHA’s client-specific target selection includes relevance to the child and family, such as a sound in the child’s name. In the core vocabulary approach, for children with inconsistent errors, practice words are those the child uses often in everyday communication, drawn from observation and parent and teacher report, with a few chosen each week (ASHA).
- Ask the family for names, pets, foods, favorite toys and words from school routines that contain the target.
- Frequent words often overlap with functional ones, and high-frequency words are among those Gierut’s review recommends (Gierut, 2016).
- Lists built around pictures can miss these words, because they can’t be drawn: our own S, R, L and K picture lists include 11 of the 134 Dolch service words with those sounds. See articulation word lists vs the words kids say most.
- When words are chosen for their sound rather than their meaning, some may be unfamiliar. Storkel suggests explaining to families that words are chosen to highlight a contrast, not to build vocabulary (Storkel, 2022).
Real words or nonwords?
- Retrospective comparison. Of 60 children with phonological disorders, those treated with nonwords showed greater, faster system-wide generalization, and kept it after treatment ended. Children treated with real words reached comparable levels, but not until 55 days after treatment stopped. The authors call for prospective studies (Gierut, Morrisette & Ziemer, 2010).
- Within-child comparison. In four preschoolers each taught one sound in real words and another in nonwords, accuracy on the treated sound was as good or better with nonwords (Gierut & Morrisette, 2010).
- A mixed result. In four children, the oldest child treated with nonwords learned slightly more and both nonword children generalized, while one of the two children treated with high-frequency real words generalized to untreated sounds (Cummings & Barlow, 2011).
Gierut’s review suggests the nonword advantage may be confined to the treatment period, and that nonwords may suit early intervention with real words to follow (Gierut, 2016). Storkel’s contrast tutorial notes that studies suggest nonwords work at least as well as real words (Storkel, 2022). At home, nonsense syllables are a common bridge: the NHS booklet joins the new sound to vowels (“kee”, “eek”) before moving into real words (Oxford Health NHS Foundation Trust).
How many words in a set
Fewer than you might think, but it varies. In 19 children with phonological impairment taught with minimal word pairs, three exemplars were enough for generalization in 59% of cases, five in 21% and ten in 14%; in 7% generalization didn’t occur even after ten. Variability between children was substantial, and some children seemed to learn later sounds faster after treatment on an earlier one (Elbert, Powell & Swartzlander, 1991).
- Start with a small set the child can succeed with, and add new words as accuracy rises.
- As the sound stabilizes, expand the set and vary positions and contexts (Preston et al., 2020).
- For contrast approaches, see minimal pairs therapy and the free minimal pairs word lists.
Untreated probe words for measuring generalization
Keep a separate list of words you never practice, and use it to test whether learning is spreading. Practice accuracy reflects acquisition, not generalization, so probe data should be the primary outcome. One tutorial uses 25 untreated one- and multisyllable words with the target in varied positions, taking about 3 minutes, probed every 3 to 4 sessions in clinical practice (Preston, Leece & Storto, 2019).
- Untreated words with the treated sound: shows within-sound generalization.
- Untreated sounds: complexity studies measure change in sounds outside the child’s inventory, which is where their predicted gains appear (Gierut et al., 1996).
- Decision points: Williams’ rules, as summarized by Storkel, move to a conversation sample once untreated words reach 90%, and train beyond word level if conversation is below 50% (Storkel, 2022).
For probe schedules, control items and graphing, see progress monitoring; for printable records, see the speech therapy data sheets.
Questions SLPs ask
How many target words should I use in articulation therapy?
There’s no fixed number. In one study of 19 children, three minimal pair exemplars were enough for generalization in 59% of cases, but some children needed five or ten, so start small and add words as accuracy grows.
Should I use nonsense words in speech therapy?
They can help. In small studies, children treated with nonwords generalized as well as or faster than children treated with real words, though the advantage may be limited to the treatment period. One review suggests nonwords early, then real words.
What are facilitating contexts in articulation therapy?
Sound combinations that make a target easier, such as /t/ before a high front vowel as in “tea”. For /ɹ/, a 2020 tutorial lists contexts such as “are”, “ear”, “rat” and /tɹ/ or /dɹ/ clusters.
What is neighborhood density?
The number of words that differ from a word by one sound. Treatment studies on whether dense or sparse neighbourhoods help generalization have conflicting results, and the samples are small.
How do you measure generalization in articulation therapy?
Probe a list of untreated words containing the target every few sessions, without practicing them or giving feedback. One tutorial uses 25 untreated words taking about 3 minutes.
Sources
- ASHA Practice Portal: Speech sound disorders (articulation and phonology).
- Oxford Health NHS Foundation Trust, Speech and Language Therapy: Introducing a new sound (parent booklet).
- Masso, S., McLeod, S., Baker, E., & McCormack, J. (2016). Polysyllable productions in preschool children with speech sound disorders: Error categories and the Framework of Polysyllable Maturity. International Journal of Speech-Language Pathology, 18(3), 272–287.
- Preston, J. L., Leece, M. C., & Storto, J. (2019). Tutorial: Speech motor chaining treatment for school-age children with speech sound disorders. Language, Speech, and Hearing Services in Schools, 50(3), 343–355.
- Preston, J. L., Benway, N. R., Leece, M. C., Hitchcock, E. R., & McAllister, T. (2020). Tutorial: Motor-based treatment strategies for /r/ distortions. Language, Speech, and Hearing Services in Schools, 51(4), 966–980.
- Morrisette, M. L., & Gierut, J. A. (2002). Lexical organization and phonological change in treatment. Journal of Speech, Language, and Hearing Research, 45(1), 143–159.
- Gierut, J. A., & Morrisette, M. L. (2012). Density, frequency and the expressive phonology of children with phonological delay. Journal of Child Language, 39(4), 804–834.
- Storkel, H. L., & Hoover, J. R. (2010). An online calculator to compute phonotactic probability and neighborhood density on the basis of child corpora of spoken American English. Behavior Research Methods, 42(2), 497–506.
- Storkel, H. L. (2001). Learning new words: Phonotactic probability in language development. Journal of Speech, Language, and Hearing Research, 44(6), 1321–1337.
- Gierut, J. A. (2016). Nexus to lexis: Phonological disorders in children. Seminars in Speech and Language, 37(4), 280–290.
- Gierut, J. A., & Morrisette, M. L. (2012). Age of word acquisition effects in treatment of children with phonological delays. Applied Psycholinguistics, 33(1), 121–144.
- Storkel, H. L., & Hoover, J. R. (2010). Word learning by children with phonological delays: Differentiating effects of phonotactic probability and neighborhood density. Journal of Communication Disorders, 43(2), 105–119.
- University of Kansas: Phonotactic Probability and Similarity Neighborhood calculators (English, including a child corpus).
- Storkel, H. L. (2022). Minimal, maximal, or multiple: Which contrastive intervention approach to use with children with speech sound disorders? Language, Speech, and Hearing Services in Schools, 53(3), 632–645.
- Gierut, J. A., Morrisette, M. L., & Ziemer, S. M. (2010). Nonwords and generalization in children with phonological disorders. American Journal of Speech-Language Pathology, 19(2), 167–177.
- Gierut, J. A., & Morrisette, M. L. (2010). Phonological learning and lexicality of treated stimuli. Clinical Linguistics & Phonetics, 24(2), 122–140.
- Cummings, A. E., & Barlow, J. A. (2011). A comparison of word lexicality in the treatment of speech sound disorders. Clinical Linguistics & Phonetics, 25(4), 265–286.
- Elbert, M., Powell, T. W., & Swartzlander, P. (1991). Toward a technology of generalization: How many exemplars are sufficient? Journal of Speech and Hearing Research, 34(1), 81–87.
- Gierut, J. A., Morrisette, M. L., Hughes, M. T., & Rowland, S. (1996). Phonological treatment efficacy and developmental norms. Language, Speech, and Hearing Services in Schools, 27(3), 215–230.