The evidence behind these guides

21 sources · last updated 13 September 2026

Every claim in our guides traces to one of the studies below. Each entry says what the study actually found, so you can judge it yourself or take it to your speech-language pathologist. Where the evidence is weak, we say so.

Practice and retention in PPA

• Henry et al. 2019 — Retraining lexical retrieval in primary progressive aphasia

  18 people with semantic and logopenic PPA. One clinician session a week plus under 15 minutes of daily home practice produced large gains on trained words (mean d = 7.2), still above baseline at 12 months. Generalisation to untrained words was small (d = 1.06). Twice-weekly clinician sessions were no better than once.

• Meyer et al. 2018 — Prophylaxis and remediation of anomia in PPA

  Gains on words the person could still name at baseline persisted up to 15 months, longer than gains on words already lost. The orthographic condition (read and copy the written word) was more durable than the phonological one. This is the source of the two-thirds keep, one-third recover split.

• Suárez-González et al. 2021 — Semantic variant PPA: a consensus review of home practice

  20 to 60 minutes of daily practice on 15 to 30 items produced gains within a month, with 65 to 82% retention at 3 to 6 months in good responders. Also states plainly that when practice becomes stressful it may not be desirable to continue.

• Middleton et al. 2016 — Retrieval practice and spacing in naming treatment

  Attempting the word before being told it beats pure repetition for durability, and spaced practice beats massed at both one week and one month. The basis for the app's review schedule.

• Savage et al. 2014 — Time for a quick word? Home practice in semantic dementia

  Five people with semantic dementia ran a two-month home programme. All improved naming of trained pictures, four of five improved on a video description task, and the milder participants improved comprehension of spoken instructions.

• Beeson et al. 2011 — Positive effects of language treatment for logopenic PPA

  Semantic elaboration plus generative naming in lvPPA. Gains generalised to untrained items and lasted six months.

Errorless practice, and why nothing is marked wrong

• Jokel & Anderson 2012 — Quest for the best: effects of errorless and active encoding on word re-learning in semantic dementia

  Errorless learning produced larger gains than errorful learning. Items with residual meaning relearned better than items whose meaning was gone.

• Nunn & Middleton 2023 — Errorless and errorful learning in aphasia: a scoping review

  Reports repeated concerns that errorful conditions damage motivation and are less preferred by patients. Why the cue cascade always ends in the correct word and the app never shows a failure state.

The partner is part of the treatment

• Rogalski et al. 2025 — Communication Bridge-2: a randomised controlled trial

  95 dyads across four countries, all three PPA variants, delivered by telehealth. The arm that trained both the person and their partner, used personal photographs and set personal goals beat the impairment-only arm on goal attainment and communicative participation.

• Simmons-Mackie et al. 2010 — Communication partner training in aphasia: a systematic review

  31 studies, updated in 2016 with 25 more. All 56 reported positive outcomes across severity levels and partner types, with benefits maintained up to 23 months in some studies.

• Kagan et al. 2001 — Training volunteers as conversation partners using Supported Conversation

  The controlled trial behind Supported Conversation for Adults with Aphasia, built on the principle that people with aphasia know more than they can say and a skilled partner can reveal it.

• Volkmer et al. 2020 — Speech and language therapy for PPA: referral patterns and practice

  Specialist therapists prioritise communication skills training above impairment-based work, and gains degrade quickly without ongoing practice. Consistency matters more than intensity.

• Grasso et al. 2017 — Caregiver-administered treatment in logopenic PPA

  Caregiver-trained word sets did as well as clinician-trained sets when the caregiver followed the clinician's protocol. A family member can run the practice; the guardrails are about how.

• Siegle et al. 2026 — Care partners as informal therapists: an interpretative phenomenological analysis

  Care partners describe becoming informal therapists, a role that conflicts with spousal connection by turning ordinary interactions into exercises. The reason the guides argue for coaching inside a fixed slot rather than all day.

Stress, anxiety and the case for no timers

• Folia Phoniatrica 2020 — Physiological stress responses during naming in aphasia

  People with aphasia perceive naming tasks as stressful, with heart rate rising during language tasks. Anticipating an error while producing language is described as a threat response.

• Counselling and care partner training in PPA — Anxiety across the variants

  Anxiety affects 27 to 46% of people with PPA depending on variant. Care-partner training teaches partners to prompt self-cueing rather than supply answers.

How strong is this evidence, honestly

• Roheger et al. 2024 — Cochrane review: interventions for PPA

  Rates the certainty of evidence for behavioural treatment in PPA as very low. Worth reading before trusting any single number on this page, including ours.

Speech recognition and aphasic speech

• Sanguedolce et al. 2024 — When Whisper listens to aphasia

  General-purpose speech recognition degrades substantially on aphasic speech. The reason this app verifies a known target word rather than attempting open transcription.

• Barbera et al. 2021 — NUVA: a naming utterance verifier for aphasia treatment

  Verifying whether a known target word was said is a far easier problem than transcribing free speech, and can reach useful agreement with human therapists.

• Koenecke et al. 2024 — Careless Whisper: speech-to-text hallucination harms

  Speech-to-text systems can invent content that was never said. Directly relevant to why a machine must never tell someone they said a word they did not.

• Davudova et al. 2025 — Whisper in clinical practice: post-stroke speech assessment during naming

  Applies current speech recognition to a naming task with post-stroke speakers and reports where it succeeds and fails.

What this page is not

This is a reading list, not medical advice, and Words at Home has no trial of its own. Nothing here says this app works; the studies describe methods the app is built to follow. Decisions about care belong with a qualified clinician who knows the person.