Which PPA variant, and why it changes practice

Last updated 12 September 2026

If you take one question to the next appointment, take this one: which variant is it? The answer changes which words you practise, what help you offer when a word will not come, and whether the comprehension work should use single words or whole sentences. The cue that unlocks a word for one person is wasted breath for another.

What primary progressive aphasia is

Primary progressive aphasia (PPA) is a neurodegenerative condition in which language goes first, while memory and other thinking stay relatively intact in the early years. Three variants are recognised. Diagnosis belongs with the neurologist and the speech-language pathologist. This page is here to help you understand the answer you were given, or to help you ask if nobody has said one out loud.

The three variants

Volkmer, Rogalski, Henry and colleagues set out these variant profiles for clinicians in Practical Neurology (2020), which is a readable summary if you want the clinical version.

Everyday signs worth writing down

None of this is a diagnosis. It is the kind of observation that makes an appointment useful, so note the actual sentence and take it with you.

• They say "cup" clearly, but call the dog "the thing", and when you ask what a whisk is, they have no idea. Meaning is going.
• They plainly know what it is, the word arrives slowly with sound errors, and they cannot repeat a long sentence back to you. The sound form is the problem.
• The words are right, the grammar is missing, and every sentence is physical work. Production is the problem.
• Hardy et al. 2024 built a symptom-led staging description for the semantic and nonfluent variants: semantic PPA reaches stage 3 when common words become incomprehensible and multi-step tasks fail, while nonfluent PPA shows difficulty in stressful conversations and yes/no confusions as early as stages 1 to 2. PMC10917001
• Anxiety affects between 27% and 46% of people with PPA depending on variant. If conversation is getting harder in crowds specifically, that is worth saying to the clinician too.

The cue that helps one person is the wrong cue for another

Here is the flip, with one word. The photo is a coffee cup and the word will not come.

Semantic variant. "It starts with kuh" hands them the sound of a word whose meaning has faded. They may well produce "cup" and still not know what to do with it. What helps is the picture, the written word, and your voice saying it, together and repeatedly. Feature choices work better than feature questions, because generating features is hard and recognising them is not: "you drink from it" or "you sit on it" as two options, rather than "tell me about it."

Logopenic variant. "You drink coffee out of it" is information they already have. It costs the moment and adds a sentence to hold in exactly the memory system that is failing. What helps is the first sound, then the first letter, then the written word, then your model, then saying it together. The cueing order for logopenic PPA is its own page.

Nonfluent variant. The word is available and the mouth will not do it. What helps is something to speak along with, slowly, and rehearsed phrases rather than single words.

How home practice differs by variant

Semantic variant. Choose words that still have some meaning attached, even partial meaning ("it's food" for egg). Items with no residual meaning relearn poorly, which the 2021 consensus review on the semantic variant states directly. Keep the set to 15 to 30 personal, frequently used items with photographs of the person's actual belongings. Run it errorless: picture, spoken word, written word, they repeat. Add written production, because Taylor-Rubin and Nickels 2020 saw bigger gains when writing was added to speaking. Put comprehension work in every session, since single-word understanding is the thing being lost: word-to-picture matching, yes/no verification, sorting photos into two labelled bins. And introduce a photo board or communication book early. The consensus review recommends doing this as early as possible in the disease process, so that the person is well practised before the skills to acquire its use are lost.

Logopenic variant. Lead with sound and spelling: first sound, first letter, written word, model, unison. Keep any meaning cue at the end or switch it off. Word length is the difficulty dial, so start with short common words and let the clinician raise the ceiling. Repetition and reading aloud carry the weight here. Make the comprehension work sentence-length instead of single words: one-step then two-step instructions, yes/no questions about a short sentence, matching a spoken sentence to one of three photos. In conversation, repeat the same short sentence rather than rephrasing it, since a new phrasing is a fresh string of sounds to hold. Scripts are worth asking about.

Nonfluent/agrammatic variant. Script training with speech entrainment is the best-evidenced approach here: the person speaks in unison with a video of a familiar speaker, then with the sound muted, then from memory. Use scripts they would actually say, like ordering coffee or phoning their daughter. Sentence-level grammar protocols exist but come from stroke aphasia, so only run them if the clinician asks for them by name.

What stays the same for all three

• Personal photos and personal words. Twenty to forty items chosen by the person, the family, and the clinician. Photographs of their own objects and people, not stock pictures.
• Protect the words they can still say. Meyer et al. 2018 found gains on still-retrievable words lasted up to 15 months, while gains on words already lost faded by 8. Roughly two-thirds of the active list should be words they still manage.
• Short and daily beats long and occasional. Henry's home practice was under 15 minutes a day. Volkmer et al. 2020 note that gains degrade quickly without ongoing practice, so consistency is the win.
• Practice helps the words you practise. Carryover to unpractised words was small in every study that measured it. Choose the words accordingly, and choose the ones they need at dinner.
• Errorless floor. Every trial ends with the person saying the word. No timers, no scores on their screen, no "wrong".
• Train the partner, not only the person. Communication partner training has the broadest evidence base in aphasia care: Simmons-Mackie and colleagues' 2010 review and 2016 update covered 56 studies, all of which reported positive outcomes, with benefits maintained up to 23 months in some.
• The protocol travels; the details change. Henry et al. 2019 found no difference in response between semantic and logopenic participants on the same protocol. What shifts with variant is the word list and the cue order.

What the research says

• Henry et al. 2019, 18 people with semantic and logopenic PPA. Weekly clinician sessions plus daily home practice under 15 minutes gave large gains on trained words (d = 7.2) that held above baseline at 12 months, with small carryover to untrained words (d = 1.06) and no benefit from doubling clinician time. Response did not differ by variant. PMC6802912
• Suárez-González et al. 2021, a consensus review of 20 years of behavioural research in the semantic variant. Twenty to sixty minutes of daily practice on 15 to 30 items produces gains within a month, with retention of 65% to 82% at three to six months in good responders. Their words on the maintenance phase: maintenance revisions, to maintain at least 80% of therapy items, require less practice than the original learning. PMC8699306
• Savage et al. 2014: five people with semantic dementia did a two-month home programme of picture, spoken word and repetition. All five improved naming of trained pictures, four of five improved on a video description task, and the milder participants improved their understanding of verbal instructions. PubMed 24413622
• Evans et al. 2016: one person with semantic-variant PPA using flashcard software for 30 minutes, three or four days a week, over 20 months retained 139 of 591 words and was 5.6 times more likely to name trained than untrained Boston Naming Test items. Words added in the first 10 months stuck far better than those added later. Frontiers in Human Neuroscience
• Beeson et al. 2011, logopenic variant: semantic elaboration combined with generative naming produced gains that carried over to untrained items and lasted six months. PubMed 21710364
• A 2025 logopenic teletherapy study trained phonological short-term memory using immediate and five-second-delayed repetition of words of increasing syllable length, three 45-minute sessions a week for five weeks. Trained-item gains held at three months, with little transfer to naming. PMC12756969
• Henry et al. 2018, video-implemented script training (VISTA) in nonfluent/agrammatic PPA, n = 10: an effect size of d = 11 on trained scripts, maintained at 12 months, with improved intelligibility carrying over to untrained topics. PMC5972572
• Grasso et al. 2024 ran a modified version of the same script training across all three variants (n = 13). Logopenic and nonfluent participants improved most, which the authors put down to scripts lowering the demands associated with word retrieval. Gains held at three and six months. PMC11482574
• Meyer et al. 2018, across variants: reading and copying the written word was more durable than listening and repeating, and gains on still-retrievable words lasted up to 15 months against 8 for words already lost. PMC6066454
• Rogalski et al. 2025, Communication Bridge-2, a randomised trial of 95 people across all three variants. The arm that trained the person and their partner together, with personal photos and personal goals, improved 66.7% of functional goals against 49.1% in the drill-only arm, and did not show the 12-month decline in communication participation that the control arm showed. Word retrieval improved similarly in both arms. PMC11947768
• Croot et al. 2019 describe the realistic goal of all of this as preserving retrieval of a core vocabulary for as long as possible. Macquarie University
• The caveat to hold onto. Roheger et al.'s 2024 Cochrane review rates the randomised evidence for behavioural PPA treatment as very low certainty, because the trials are small. The effect on practised words shows up consistently. Nothing here slows the disease, and nobody can promise a given person a given result. Cochrane CD015067

What to ask the speech-language pathologist

• Which variant, how confident are you, and what would change your mind?
• Given that variant, what cue order should we use at home, and in what order exactly?
• Which words should be on the list, and how many of them should be words they can still say?
• Should our comprehension practice use single words or whole sentences?
• Are scripts worth the time for this variant, and will you write one with us?
• Is Communication Bridge (Northwestern) or Better Conversations with PPA (UCL) available to us, and will you write partner-training goals?