How much daily practice actually works
The 15-minute finding, in detail
Henry et al. 2019 (Journal of Speech, Language, and Hearing Research) is the study most often cited on home dose in progressive aphasia. Eighteen participants with semantic or logopenic variant PPA worked through a fixed cueing sequence on 20 to 40 personally chosen words over four to eight weeks.
The trained-word effect was large, with a mean effect size of d = 7.2, and it was maintained at 3, 6 and 12 months. Generalisation to untrained words was small, d = 1.06, and held only to 6 months.
Once-weekly clinician sessions plus daily home practice produced the same outcome as twice-weekly clinician sessions. That comparison is the reason the home component gets the credit: the variable that doubled was clinic time, and doubling it changed nothing. The daily home component was Copy and Recall Treatment, or CART: look at a picture with the written word, say and copy the word about 10 times, then cover it and produce it from memory, spoken and written. Under 15 minutes. Dial et al. 2019 found in-person and teletherapy delivery of the same protocol gave equivalent outcomes out to one year.
An honest caveat on all of this: the 2024 Cochrane review of non-pharmacological interventions in PPA (Roheger et al.) rates the randomised-trial evidence in this area as very-low certainty, because the trials are small. The trained-item effect is consistent across studies; the certainty rating is about how confidently it can be generalised.
What other studies used as a dose
- Meyer et al. 2018 (all variants): 10 to 15 minutes, three times a week, over five months, with a caregiver running the cards.
- Suárez-González et al. 2021, the semantic-variant consensus review drawing on 20 years of behavioural research: 20 to 60 minutes daily on 15 to 30 items, with gains appearing within a month.
- Evans et al. 2016: one person with semantic variant PPA using Anki flashcards for about 30 minutes, 3 to 4 days a week, roughly 35 cards a session, sustained over 20 months.
- Communication Bridge-2 (Rogalski et al. 2025, randomised controlled trial, n = 95, all three variants): 30 minutes a day, 5 days a week of app exercises.
- Modified VISTA (Grasso et al. 2024, n = 13): 30-minute clinician sessions twice weekly for 8 weeks, plus home video practice 3 days a week.
- A 2025 logopenic-variant teletherapy study of phonological short-term memory: three 45-minute sessions a week for five weeks. Trained-item gains held at 3 months, with little transfer to naming.
The pattern across all of them is short and frequent rather than long and occasional. No PPA study in this set used a long weekly block as its home dose.
How many words at a time
Fifteen to thirty items is the range the literature converges on. Suárez-González et al. 2021 recommend sets of 15 to 30. Henry et al. 2019 trained 20 to 40 words across a four-to-eight-week block, working through them in small sets. Evans et al. 2016 ran about 35 cards per session.
Which words matters as much as how many. Meyer et al. 2018 split their items into two kinds. "Prophylaxis" items were words the person could still name at baseline; "remediation" items were words already lost. Gains on the still-retrievable words persisted up to 15 months. Gains on the already-lost words faded by 8 months. Their own conclusion: when clinician time is short, spend it on words that are still retrievable and personally relevant.
Evans et al. 2016 found the same asymmetry across time rather than across item type. Over 20 months, 139 of 591 trained words were retained, and that participant was 5.6 times more likely to name a trained than an untrained item on the Boston Naming Test. Words added in the first 10 months stuck considerably better than words added later.
Massed versus spaced practice
Massed practice means repeating a word many times in one sitting. Spaced practice means returning to it after a gap, and then after a longer gap. The two can involve exactly the same number of repetitions and produce different durability.
Middleton et al. 2016 tested this directly in naming rehabilitation and found spaced practice beat massed practice at both 1 week and 1 month. The same work found that retrieval practice, where the person attempts the word after a familiarisation trial rather than simply repeating it, produced more durable naming than repetition alone.
Nunn and Middleton's 2023 scoping review of errorless, errorful and retrieval practice in stroke aphasia reached a compatible conclusion: errorless and errorful learning came out broadly comparable, while retrieval practice spaced over time was the more durable option. The review also records repeated concerns that error-heavy methods damage motivation. In semantic dementia, Jokel and Anderson 2012 found errorless learning outperformed errorful.
The practical combination those findings point to is a first attempt made without help, followed by a guaranteed correct production whatever happens, with the word returning on a lengthening schedule. Guidance on errorless learning and spaced retrieval in dementia rehabilitation is to keep success rates in the 80 to 90 percent range and to shorten the interval after a miss rather than dwell on it.
Why review matters more than adding volume
This is the finding families most often miss. Suárez-González et al. 2021 report retention of 65 to 82 percent of trained items at 3 to 6 months in good responders, and state that maintenance revisions, aimed at holding at least 80 percent of therapy items, require less practice than the original learning did.
They put numbers on it: fewer than 10 revision sessions across 6 months for people who are mild to moderate, weekly for people who are more severely affected. They also report that performance at 2 months after treatment predicts how much revision a person will need.
So the effort curve is front-loaded. Learning a set is expensive; holding it is cheap. Stopping altogether is what costs you the set. Evans et al. 2016 is the clearest illustration of the alternative: 20 months of continued, lighter review is what kept 139 words alive.
How long the gains last
- Henry et al. 2019: trained-word gains above baseline at 12 months, in people whose cognition measurably declined over that same year.
- Meyer et al. 2018: still-retrievable items held up to 15 months; already-lost items faded by 8.
- Suárez-González et al. 2021: 65 to 82 percent retention at 3 to 6 months in good responders.
- Grasso et al. 2024 (modified VISTA): script gains held at 3 and 6 months, with participant satisfaction rated 4.7 to 4.9 out of 5.
- Henry et al. 2018 (VISTA, non-fluent variant, n = 10): d = 11 on trained scripts, maintained at 12 months, with intelligibility generalising to untrained topics.
- The 2025 logopenic phonological short-term memory study: trained-item gains held at 3 months.
In every one of these, the durable gains are on the specific items practised. Generalisation to untrained words is consistently the weaker result. That is the single most important thing to understand before choosing a word list.
What else the dose does not explain
Communication Bridge-2 (Rogalski et al. 2025) is worth reading carefully, because both of its arms got the same dose. Both did 30 minutes a day, 5 days a week, and both improved similarly on word retrieval. What separated them was everything around the drill: the arm that also trained the communication partner, used the participant's own photographs, and set personal goals improved 66.7 percent of its goals against 49.1 percent in the impairment-only control, and did not show the 12-month decline in communication participation that the control arm showed.
Dose is not the only variable, in other words. A follow-up report on Communication Bridge-2 app usage found participants averaged 13.7 exercises across 4 days a week, sustained over a year, which is a useful benchmark for what real adherence looks like outside a lab.
Two findings from stroke aphasia are also relevant here, with the caveat that stroke aphasia and progressive aphasia are different conditions. Brady et al.'s 2016 Cochrane review and the RELEASE collaboration's 2022 analysis both favour higher total therapy dose, while noting dropout at very high intensity. And an analysis of Constant Therapy data found home users mastered tasks in about 6 days against 12 days for clinic users, with individual adherence ranging from 0 to 104 hours: practice frequency, rather than where the practice happened, tracked mastery.
When practice stops being useful
Two limits are stated explicitly in the literature. Evans et al. 2016 found that words added later in the disease consolidated less well, which argues for shifting effort toward maintaining an existing list once new words stop sticking. And the semantic-variant consensus review says plainly that when practice becomes stressful, it may not be desirable to continue. Anxiety affects 27 to 46 percent of people with PPA depending on variant, and naming tasks are measurably stressful, with heart rate rising during language tasks.
One methodological note that affects every number above: Henry et al. 2019 could not analyse the effect of post-treatment practice because only 8 of 18 participants returned usable logs, all of it self-reported on paper. Modified VISTA did not objectively track practice intensity at all. The dose figures in this literature are what studies asked for, not always what participants did.
Sources
- Henry ML et al. 2019. Treatment for word retrieval in semantic and logopenic variants of PPA: immediate and long-term outcomes. JSLHR. PMC6802912
- Meyer AM et al. 2018. Long-term maintenance of anomia treatment effects in PPA. Aphasiology. PMC6066454
- Suárez-González A et al. 2021. Semantic variant PPA: practical recommendations for treatment from 20 years of behavioural research. Brain Sciences. PMC8699306
- Evans WS et al. 2016. Relearning and retaining personally relevant words using computer-based flashcard software in PPA. Frontiers in Human Neuroscience. Frontiers
- Rogalski E et al. 2025. Efficacy of Communication Bridge-2 for PPA: randomised controlled trial. Alzheimer's & Dementia. PMC11947768
- Grasso S et al. 2024. Effects of modified VISTA in the three variants of PPA. JSLHR. PMC11482574
- Henry ML et al. 2018. Retraining speech production and fluency in non-fluent/agrammatic PPA (VISTA). Brain. PMC5972572
- Roheger M et al. 2024. Non-pharmacological interventions for improving language and communication in people with PPA. Cochrane. PMC11134511
- Nunn K, Middleton EL. 2023. Errorless, errorful, and retrieval practice for naming treatment in aphasia: scoping review. JSLHR. PMC10023178
- Middleton EL et al. 2016. Towards a theory of learning for naming rehabilitation: retrieval practice and spacing effects. JSLHR. PubMed 27716858
- Dial H et al. 2019. Investigating the utility of teletherapy in individuals with PPA. Clinical Interventions in Aging. PubMed 30880927
- Phonological short-term memory teletherapy in lvPPA, 2025. PMC12756969
- Brady MC et al. 2016. Speech and language therapy for aphasia following stroke. Cochrane. Cochrane
- RELEASE Collaboration 2022. Dosage, intensity, and frequency of language therapy for aphasia. Stroke. Stroke
- Comparison of therapy practice at home and in the clinic: Constant Therapy data set. PMC6398324