Written picture descriptions distinguish variants of primary progressive aphasia

Abstract

Background

Few studies have examined written discourse in primary progressive aphasia (PPA), a clinical syndrome due to frontotemporal lobar degeneration (FTLD) or Alzheimer's disease (AD).

Objective

We aim to: (1) determine differences in written discourse in PPA variants and controls using three approaches for analyzing narratives, and (2) make recommendations regarding the clinical utility of these approaches.

Methods

Individuals with PPA and healthy controls wrote descriptions of the Boston Diagnostic Aphasia Examination Cookie Theft Picture (CTP). We hypothesized that written narratives would be characterized by: (1) fewer total words, lower percentages of CTP content units (CUs) and Core Lexicon Words, and lower communication efficiency in all PPA variants compared to controls; (2) fewer content words in semantic variant PPA (svPPA) than in logopenic variant PPA (lvPPA), nonfluent variant PPA (nfvPPA), and controls; and (3) fewer function words in nfvPPA than in lvPPA and controls.

Results

Participants with svPPA had significantly lower percentages of total CTP CUs, CTP noun and verb phrase CUs, and Core Lexicon Words than lvPPA. Overall content unit (CU) profile of lvPPA was more similar to controls than to the other two variants. Written narratives of participants with nfvPPA and svPPA had significantly lower percentages of particles, a class of function words, than lvPPA participants.

Conclusions

Part of speech analysis showed a deficit in function words in nfvPPA. Content unit analysis distinguished svPPA and lvPPA, and is easily incorporated into the clinical environment when spoken data are scarce or difficult to obtain because of speech production impairments.

Keywords

Alzheimer's disease connected speech content units discourse frontotemporal dementia primary progressive aphasia spontaneous speech

Introduction

Primary progressive aphasia (PPA) is a neurodegenerative clinical syndrome, characterized by the insidious onset of language impairments with relative sparing of other cognitive domains in the early stages of disease progression.^1–3 Three variants are commonly recognized: semantic variant PPA (svPPA), nonfluent variant PPA (nfvPPA), and logopenic variant PPA (lvPPA).⁴ PPA is diagnosed based on clinical features, patterns of atrophy in the left temporal, parietal, and/or frontal cortices, and underlying neuropathologies, typically frontotemporal lobar degeneration (FTLD) in svPPA and nfvPPA, and Alzheimer's disease (AD) in lvPPA.^5–12

Spoken discourse has been investigated extensively in individuals with all PPA variants using various elicitation techniques, including picture description,^13–22 semi-structured interviews,^14,15,19,23 and wordless picture books.^{16,19,24–30} Analysis of spoken descriptions of the Cookie Theft Picture (CTP) from the Boston Diagnostic Aphasia Examination³¹ revealed grammatical impairment in individuals with nfvPPA compared to other PPA variants, with shorter, less complex, and fewer well-formed sentences in nfvPPA than in lvPPA and svPPA.^16,18 Similarly, in semi-structured interviews, there were more frequent grammatical errors in nfvPPA than in lvPPA and svPPA.¹⁵ Themistocleous et al.²⁰ performed an automated analysis of spoken CTP descriptions, which revealed that participants with nfvPPA produced proportionally more nouns than participants with lvPPA and svPPA, and fewer verbs than svPPA, although the latter was not significant—however, the content-to-function word ratio distinguished individuals with nfvPPA from the other variants. Analysis of spoken descriptions of the Picnic Scene from the Western Aphasia Battery³² revealed that slow rate of speech, frequent phonemic errors, syntactic errors, reduced embeddings, and short mean length of utterance in nfvPPA. In contrast, there was normal rate of speech and few speech or syntactic errors, but increased proportions of closed class [versus open class (e.g., nouns)] words, and use of high frequency nouns in svPPA. In lvPPA, speech rate was intermediate between the other two variants, but distortions and syntactic errors were less common than in nfvPPA. The greatest numbers of false starts, filled pauses, and repaired sequences occurred in lvPPA compared to nfvPPA and svPPA.¹³ Specifically, reduced total speech duration, fewer utterances, and a higher use of open class words (e.g., nouns) were the most effective measures for distinguishing nfvPPA from lvPPA. In spoken descriptions of the Picnic Scene by Italian-speaking individuals, nfvPPA was distinguished from lvPPA by reduced total speech duration, fewer utterances, and a higher use of open class words.³³

Much less attention has been directed to written discourse (see Kim and colleagues³⁴ for review), even though written communication is increasingly vital in everyday life with the expanded use of email, text messaging, social media, and online shopping and banking, and despite the current diagnostic criteria specifying associated speech-language characteristics that are captured through written language. Specifically, surface dyslexia/dysgraphia (regularization of irregular words results in phonologically plausible errors, such as “ocean” spelled “oshun”) in svPPA.⁴ Although beyond the scope of this paper, researchers have investigated whether spelling can characterize the PPA variants and have found that nfvPPA and svPPA have distinctive spelling profiles, with some exceptions and overlap of the spelling error types emerging with disease progression.^35,36 Surface dysgraphia (reliance on sublexical phonology-to-orthography conversion, phonologically plausible errors, more difficulty spelling words with low phoneme-to-grapheme conversion probabilities, such as irregular words like “yacht”) is typical of svPPA; phonological dysgraphia (phonologically implausible errors, more difficulty spelling pseudowords than words, better performance on regular than exception phoneme-to-grapheme mapping) is characteristic of nfvPPA; a mixture of both spelling patterns is seen in lvPPA.^35–38 Neophytou et al.³⁹ employed an automated analysis of spelling using pairwise comparisons to accurately classify 64% of their participants with PPA based on spelling profiles in a real-life scenario in which there was no prior knowledge of PPA variant type.

Few studies have examined written discourse in PPA.^40–48 An advantage of studying written discourse is that writing is not confounded by motor speech impairments [i.e., apraxia of speech (AOS), dysarthria] that can be encountered in PPA. While comprehensive assessment of PPA requires evaluation in both spoken and written domains, evaluation of written discourse can enable identification of language impairments that may be masked in spoken discourse by speech production deficits. Apraxia of speech and agrammatism are core features of nfvPPA.^3,4,26,49 Although agrammatism (progressive agrammatic aphasia^50,51) and apraxia may present in isolation [primary progressive apraxia of speech (PPAOS^52,53)], the most common presentation is concomitant AOS and agrammatism,⁵⁴ with AOS being more frequent or prominent than agrammatism and one of the first symptoms that contributes to timely diagnosis.^55–59 Studies of nfvPPA, or PPA unspecified as to type, reported a median AOS prevalence of 78% across 162 aggregated cases, and in about 20% of cases, AOS was the primary and sometimes only deficit with little or no evidence of aphasia.⁶⁰ Mutism, the absence of speech or articulatory movement due to AOS, can present early in the disease course in nfvPPA while other cognitive functions are spared.⁵⁶ AOS can also occasionally present in lvPPA.^13,59,61 Less commonly, dysarthria (typically spastic and hypokinetic subtypes) occurs in PPA.^13,61–63 Motor speech impairment can complicate language-based analysis of spoken discourse because distinguishing phonetic (AOS) and phonemic speech sound errors can be challenging even for experienced clinicians,⁶⁴ and because speech intelligibility can be compromised due to dysarthria.

Most studies of written discourse employed picture scene descriptions. For example, Sitek et al.⁴¹ compared descriptive writing using three picture scenes (the CTP, the picture scene from the Frenchay Aphasia Screening Test,⁶⁵ and the Warrington beach scene⁶⁶) in 10 Polish-speaking individuals with nfvPPA and 17 individuals with progressive supranuclear palsy (PSP). Sitek et al.⁴² used this same approach to analyze written discourse in Polish-speaking patients (nine with lvPPA, 13 with AD, and 13 with MCI).

In three other studies, oral and written discourse were analyzed.^38,41,42 Graham et al.⁴⁰ compared spoken and written CTP descriptions in 14 individuals with nfvPPA (relative to controls). Tetzloff et al.⁴³ compared spoken and written descriptions of the Picnic Scene from the Western Aphasia Battery (WAB)³² in eight individuals with agrammatic primary progressive aphasia (agPPA) defined as predominant agrammatism, 21 with agrammatism in the context of dominant apraxia of speech (DAOS), and 13 with primary progressive apraxia of speech (PPAOS). Tetzloff et al.⁴⁸ subsequently compared written descriptions of the WAB Picnic Scene from 24 patients with AOS+ primary agrammatic aphasia (PAA), 24 patients with PPAOS, and 24 healthy controls. Josephy-Hernandez et al.⁴⁴ analyzed total units (i.e., words, non-words, false starts) and content units in spoken and written descriptions of the WAB Picnic Scene in healthy controls and individuals with PPA (28 nfvPPA, 30 lvPPA, 17 svPPA). Code et al.⁴⁵ conducted a longitudinal analysis of written language in an individual with nfvPPA. Heitkamp et al.⁴⁶ evaluated diary entries written over time by an individual of Swiss origin with svPPA. Hwang et al.⁴⁷ examined two works of historical fiction authored by an individual with svPPA.

The studies above yielded the following observations. With regard to nfvPPA, analysis of written picture scene descriptions reveals manifestations of agrammatism without the confounding factor of speech production deficits, such as apraxia of speech and dysarthria. For example, Graham et al.⁴⁰ reported that their patients had no evidence of agrammatism or reduced verb production in spoken discourse. However, they showed agrammatism, reduced verb production, and overall telegraphic output in written narratives. Sitek et al.⁴¹ found that there were no differences in the distributions of nouns and verbs in the written picture descriptions in nfvPPA and PSP, but that participants with PSP wrote longer narratives and had more letter and diacritic mark omissions than participants with nfvPPA. Tetzloff et al.⁴³ found a lower mean length of sentence/utterance, fewer grammatical utterances, more non-utterances, more syntactic and semantic errors, and fewer complex sentences/utterances in nfvPPA (agPPA and DAOS) than PPAOS in written and spoken narratives, as well as fewer correct verbs and nouns in spoken than written narratives. Tetzloff et al.⁴⁸ found that AOS + PAA group showed significantly reduced number of words compared to PPAOS patients and controls, reduced number of sentences compared to controls, and grammatical deficits. Longitudinal analysis of picture description revealed a decline in the total number of words, verbs, and function words over time in nfvPPA.⁴⁵

With regard to the other PPA variants, Sitek et al.⁴² found that written picture descriptions were comparable across lvPPA, AD, and MCI, but letter insertion errors distinguished lvPPA from AD and MCI, and frequent verb use distinguished lvPPA from AD. Analysis of text from authors’ books and their diary entries over time showed less varied vocabulary, increased use of high-frequency words, and the emergence of simplified sentence structure over time in svPPA.^53,54

When comparing spoken and written discourse, Josephy-Hernandez et al.⁴⁴ analyzed content units (words that are intelligible in context, accurate concerning the picture, and relevant to and informative about the content of the picture) rather than parts of speech, and compared spoken and written discourse. Participants with lvPPA and svPPA wrote fewer content units than controls. There were fewer total content units in written versus spoken narratives in all PPA variants. Participants with lvPPA and svPPA produced fewer written content units than spoken content units (see Table 1 for a summary).

Table 1.

Summary of findings for oral and written discourse in PPA.

	Oral discourse	Written discourse	Tasks
	Oral discourse	Written discourse	Picture scenes	Other	Population	Oral discourse findings	Written discourse findings
Graham et al.⁴⁰	√	√	√		nfvPPA, healthy controls	no evidence of agrammatism or reduced verb production	agrammatism, reduced verb production, and overall telegraphic output
Sitek et al.⁴¹		√	√		nfvPPA, PSP		no differences in the percentages of nouns and verbs in nfvPPA and PSP; longer narratives and more letter and diacritic mark omissions in PSP than nfvPPA
Sitek et al.⁴²		√	√		lvPPA, AD, MCI		written picture descriptions were comparable across the three groups; letter insertion errors distinguished lvPPA from AD and MCI; frequent verb use distinguished lvPPA from AD
Tetzloff et al.⁴³	√	√	√		agPPA, DAOS, PPAOS	lower mean length of utterance, fewer grammatical utterances, more non-utterances, more syntactic and semantic errors, and fewer complex utterances in agPPA and DAOS than PPAOS	shorter sentence length, more syntactic and semantic errors, fewer complex sentences in agPPA and DAOS than PPAOS
Tetzloff et al.⁴⁸		√	√		AOS + PAA, PPAOS, healthy controls		reduced number of words compared to PPAOS patients and controls, reduced number of sentences compared to controls, and grammatical deficits in AOS + PAA
Josephy- Hernandez et al.⁴⁴	√	√	√		PPA, healthy controls	overall number of total units less in written versus spoken narratives in PPA; fewer written content units than spoken content units in lvPPA and svPPA; content unit/total unit ratio greater in written than spoken language in PPA	fewer written total units in PPA than controls; fewer content units in lvPPA and svPPA than controls; written content unit/total unit ratio higher in nfvPPA and svPPA than controls and higher in nfvPPA and svPPA than lvPPA
Code et al.⁴⁵		√	√	sentence-writing tasks	nfvPPA		decline in the total number of words, verbs, and function words over time
Heitkamp et al.⁴⁶		√		diary entries	svPPA		decreased vocabulary variety, increased use of high-frequency words, emergence of semantic paragraphias, increasing rate of simple connectives, and simpler syntax emerged over time
Hwang et al.⁴⁷		√		historical fiction	svPPA, text from novels of other authors		increased use of adverbs, decreased use of nouns, increased use of familiar words, simpler narrative structure over time

PPA: primary progressive aphasia; lvPPA: logopenic variant PPA; nfvPPA: nonfluent variant PPA; svPPA: semantic variant PPA; PSP: progressive supranuclear palsy; agPPA: agrammatic primary progressive aphasia; DAOS: agrammatism in the context of dominant apraxia of speech; PPAOS: primary progressive apraxia of speech; PAA: primary agrammatic aphasia.

In summary, only Josephy-Hernandez et al.,⁴⁴ Heitkamp et al.,⁴⁶ and Hwang et al.⁴⁷ included individuals with svPPA in their analyses (total n = 19). While Josephy-Hernandez et al.⁴⁴ compared total units and content units in spoken and written discourse in healthy controls and individuals with all PPA variants, they did not analyze parts of speech. Thus, little is known about written discourse in svPPA. No studies compare different written discourse measures (e.g., parts of speech, content units) in their capacity to distinguish between all PPA variants. Most previous studies have concentrated on nouns and verbs, leaving the analysis of other word classes relatively underexplored (e.g., particles that are relevant in this paper). Furthermore, it is not known whether spoken language deficits that characterize the variants (i.e., fewer function words in nfvPPA, fewer nouns in lvPPA and svPPA) are modality-specific. Findings regarding spoken discourse cannot be generalized to written discourse because these modalities, while coexisting and complementary, are developmentally and evolutionarily independent.^67,68

In the present study, we aimed to: (1) address the gap in knowledge regarding deficits in parts of speech in all PPA variants manifested in written discourse, (2) expand the analysis to include commonly used approaches, such as the CTP content units (CUs),⁶⁹ and (3) make recommendations regarding the clinical utility of different analysis approaches. We examined several written discourse measures in PPA variants and neurotypical controls using three tools:

(i) lexical and morphological measures, including total words, content and function words, and part of speech; (ii) CTP CUs and communication efficiency [defined as syllables per content unit (CU), Syll/CU]^69,70 and; (iii) Core Lexicon Words.⁷¹ We chose these measures because: (i) lexical and morphological measures will allow us to investigate how known differences among the PPA variants (i.e., fewer function words in nfvPPA, fewer nouns in lvPPA and svPPA) are manifested in written discourse; (ii) CTP CUs and communication efficiency will allow us to investigate the proficiency of written discourse, which, unlike proficiency in the spoken domain, is not compromised by motor speech impairment; and (iii) Core Lexicon Words will allow us to compare the PPA variants using a clinician-friendly tool that captures word retrieval, an impairment common to all PPA variants, albeit with differences in the nature and severity of word retrieval difficulty (i.e., prominent anomia in svPPA due to impaired semantic knowledge⁴; greater verb naming deficits in nfvPPA^72–77 and greater noun naming deficits in svPPA and lvPPA^74–77).⁷⁸ Regarding aims 1 and 2, based on previous literature describing the speech and language profiles of the PPA variants,⁴ we hypothesized that written picture narratives will be characterized by: (1) fewer total words (content and function words), lower percentages of CTP CUs and Core Lexicon Words, and lower communication efficiency (Syll/CU) in all PPA variants compared to controls who do not have speech and language impairments; (2) fewer content words in svPPA than in lvPPA, nfvPPA, and controls because anomia and associated impaired semantic knowledge are the predominant features of svPPA^79,80; and (3) fewer function words in nfvPPA than in lvPPA and controls due to greater impairment for verbs versus nouns^72–76,81 and agrammatism in nfvPPA.⁴

Methods

Participants

Participants were individuals with PPA (n = 49) who were included in previous research studies or seen in an outpatient neurology clinic. Twenty-one participants (16 with lvPPA, five with nfvPPA) were enrolled in clinical trials (tDCS Intervention in Primary Progressive Aphasia, ClinicalTrials.gov Identifier NCT02606422; Targeting Language-specific and Executive-control Networks with Transcranial Direct Current Stimulation in Logopenic Variant PPA, ClinicalTrial.gov Identifier NCT03887481). Sixteen individuals with svPPA and 12 individuals with nfvPPA were seen in an outpatient neurology clinic between June 2012 and December 2017. Eighteen healthy controls participated in a longitudinal study of stroke recovery.

The research was planned, conducted, and recorded according to the principles of the Declaration of Helsinki, seventh edition. The participants in the clinical trials or their legally authorized representatives provided written informed consent. Ethical approval was obtained from the Johns Hopkins University School of Medicine Institutional Review Board (NA_00071337; NA_00042097; IRB00201027) and the University of South Florida Institutional Review Board (IRB005068).

For clinical patients, ethical approval for chart review, data extraction, and data entry into a secured, de-identified spreadsheet from the electronic medical record was obtained from the Johns Hopkins University School of Medicine Institutional Review Board which approved this study as exempt research for which consent is not required (IRB00241582). Overall, 49 individuals with PPA (mean age = 68.53 years ± 7.90; 51% women) completed written narratives of the CTP. This PPA group included 16 with lvPPA, 17 with nfvPPA, and 16 with svPPA. PPA was determined by their medical history, comprehensive neurological examination, neuroimaging (if available), and a battery of cognitive/language tests. Criteria from the international consensus were used to classify PPA.⁴ The particular variant was diagnosed based on expressive and receptive language characteristics, including the key characteristics of impaired repetition in lvPPA, agrammatism and/or apraxia of speech in nfvPPA, and impaired word comprehension in svPPA. Written discourse, our measure of interest, was not used to classify the variants. Eighteen healthy controls (mean age = 58.83 years ± 10.34; 39% women) also completed CTP written narratives. These individuals had no history of cognitive changes, no subjective memory complaints, and no pathological findings on neuroimaging (if available). Performance was within normal limits on the Mini-Mental State Examination (mean = 29.22 ± 0.81; ≥27.6/30 indicates normal cognition).⁸² There were no sex differences across PPA variants and healthy controls. After correcting for multiple comparisons, the PPA variants were not significantly different from one another in terms of age and years of education; however, control participants were significantly younger than individuals with nfvPPA and had significantly more years of education than participants with svPPA (Table 2; see Supplemental Table 1 for medians and mean ranks for age and education for PPA variants and participants overall).

Table 2.

Age, education, and sex for PPA variants and participants overall.

	PPA Variants			Controlsn = 18	Overalln = 67	H, 95% CI [0, .8147]	Pearson Chi-square	p*	Dunn's Post HocComparisons**
	lvPPAn = 16	nfvPPAn = 17	svPPAn = 16	Controlsn = 18	Overalln = 67	H, 95% CI [0, .8147]	Pearson Chi-square	p*	Dunn's Post HocComparisons**
Age (y)
Mean	68.06	70.47	66.94	58.83	65.93	12.713	—	0.005	nfv > c : <0.001
SD	7.82	9.11	6.88	10.34	9.63	12.713	—	0.005	nfv > c : <0.001
Education (y)
mean	15.80	16.47	15.00	18.61	16.59	13.147	—	0.004	sv < c : <0.001
SD	1.94§	2.90	2.45‡	3.03	2.93	13.147	—	0.004	sv < c : <0.001
Sex (F)
N (%)	8 (50)	6 (35)	11 (69)	7 (39)	32 (48)	—	4.484	0.214

F: female; SD: standard deviation; y: years; PPA: primary progressive aphasia; lvPPA: logopenic primary progressive aphasia; nfvPPA: nonfluent variant primary progressive aphasia; svPPA: semantic variant primary progressive aphasia; *p values were calculated using Kruskal-Wallis Test with Dunn's Post Hoc Comparisons for age and education and using Pearson chi-square for sex. **p values were adjusted for six comparisons (controls-lv, controls-nfv, controls-sv, lv-nfv, lv-sv, nfv-sv); p_bonf = 0.05/6 = 0.008. § missing data for 1 subject; ‡ missing data for 2 subjects.

The PPA variant groups were compared on tests of confrontation naming, important for performance on the written discourse, and a measure of overall function. Individuals with svPPA scored lower on the short version of the Boston Naming Test (BNT),^83,84 the Hopkins Action Naming Assessment (HANA),⁸⁵ and the Pyramids and Palms Trees Test (PPTT),^86,87 than individuals with lvPPA and nfvPPA. They also scored lower on the Kissing and Dancing Test⁷⁸ than individuals with nfvPPA. The Global Score on the Clinical Dementia Rating (CDR)^88,89 was significantly higher (reflecting greater impairment) for svPPA than nfvPPA; however, the mean scores for all variants reflected very mild/mild cognitive impairment (Table 3; see Supplemental Table 2 for medians and mean ranks for cognitive/language test scores for PPA variants).

Table 3.

Cognitive/language test scores for PPA variants.

	PPA variants			Overalln = 49	H, 95% CI [0, 0.8147]	p*	Dunn's Post Hoccomparisons**
	lvPPAn = 16	nfvPPAn = 17	svPPAn = 16	Overalln = 49	H, 95% CI [0, 0.8147]	p*	Dunn's Post Hoccomparisons**
Boston Naming Test (of 30)
mean	22.25	21.86	8.56	17.37	20.102	<0.001	sv < lv : <0.001
SD	4.34	7.44†	7.85	9.23	20.102	<0.001	sv < nfv : <0.001
Hopkins Action Naming Assessment (of 35)
mean	23.31	21.50	10.20	18.38	15.179	<0.001	sv < lv : <0.001
SD	5.29	8.22†	8.89§	9.45	15.179	<0.001	sv < nfv : 0.003
Pyramids and Palm Trees Test (of 14)
mean	13.81	13.59	10.25	12.57	19.388	<0.001	sv < lv : <0.001
SD	0.54	0.87	3.24	2.51	19.388	<0.001	sv < nfv : <0.001
Kissing and Dancing Test (of 15)
mean	12.38	13.47	10.50	12.09	9.593	0.008	sv < nfv : 0.002
SD	2.55	2.36‡	2.85	2.83	9.593	0.008	sv < nfv : 0.002
Clinical Dementia Rating (CDR) Scale, Global Score
mean	0.656	0.500	0.969	0.704	15.006	<0.001	sv < nfv : <0.001
SD	0.239	0.433	0.464	0.432	15.006	<0.001	sv < nfv : <0.001

SD: standard deviation; PPA: primary progressive aphasia; lvPPA: logopenic primary progressive aphasia; nfvPPA: nonfluent variant primary progressive aphasia; svPPA: semantic variant primary progressive aphasia; *p values were calculated using Kruskal-Wallis Test with Dunn's Post Hoc Comparisons. **p values were adjusted for three comparisons (lv-nfv, lv-sv, nfv-sv); p_bonf = 0.05/3 = 0.016. § missing data for 1 subject; ‡ missing data for 2 subjects; † missing data for 3 subjects.

Procedures

All participants completed a written description of the CTP. Written picture descriptions were obtained at the initial testing timepoint for those participants who were in clinical trials and at the baseline clinical evaluation for those participants who were seen in the outpatient neurology clinic. Participants were asked to wear their eyeglasses (if needed and available) for the task. They were shown the CTP and were instructed, “Please look at the picture and write a description of everything you see happening in this picture.” There were no time limits.

Handwritten picture descriptions were typed into Word documents by one co-author (KMS). A second co-author (DCT) with experience in language sample collection and analysis independently reviewed the transcriptions to ensure accuracy. Discrepancies were resolved by reviewing handwritten descriptions to achieve consensus. Written discourse was analyzed as described below.

Analysis 1: part of speech

Written narratives were analyzed and scored for measures of lexical, phonological, and morphological characteristics using an online platform (Open Brain AI; https://openbrainai.com).⁹⁰ Open Brain AI uses computational analysis of spoken and written language for research and clinical purposes. Automated lexical (i.e., total number of words, total number of content words, total number of function words), phonological (i.e., total number of syllables), and morphological measures (i.e., number of adjectives, adverbs, articles, conjunctions, nouns, particles, prepositions, pronouns, and verbs; percentage of adjectives, adverbs, articles, conjunctions, nouns, particles, prepositions, pronouns, and verbs of the total number of words) were extracted.

The part of speech tagging from Open Brain AI was verified manually and by analyzing the written discourse samples using a part-of-speech tagger tool from an online website (https://textinspector.com).

Analysis 2: content units and communication efficiency

Written CTP descriptions were also evaluated using the analysis developed by Yorkston and Beukleman⁶⁹ and Craig et al.⁷⁰ These measures included total CTP CUs [concepts mentioned by at least one healthy control describing the CTP (a total of 52)] which captures the relevance of written output to the pictured scene, and communication efficiency (syllables per CU, Syll/CU) which measures digression and irrelevancy in written output.⁹¹ CTP CUs were counted by reviewing each written sample and identifying any of the CUs from the list of 52. Repeated CUs were counted once. Percentages of CUs (out of 52) were calculated.

Analysis 3: core lexicon

Additionally, a Core Lexicon analysis was performed using a predetermined word list developed by Dalton et al.⁷¹ The Core Lexicon Words are single-word lexical items commonly produced by healthy individuals in their descriptions of the CTP. The list consists of 26 words (16 content words and 10 function words). Each written discourse sample was manually analyzed for the presence or absence of each item, and the percentages of the 26 possible items were calculated.

In summary, we measured:

For Analysis 1:

number of total words,

number of total syllables,

number of all content words (i.e., adjectives, adverbs, nouns, verbs),

number of all function words [i.e., articles, conjunctions, particles (a class of function words that includes adverbial particles, the infinitival particle “to”, discourse particles, negative particles), prepositions, pronouns],

percentages of parts of speech (i.e., percentages of adjectives, adverbs, articles, conjunctions, nouns, particles, prepositions, pronouns, verbs), defined as the percentage of each part of speech out of the number of total words,

For Analysis 2:

(vi) percentage of CTP CUs (of 52 total CUs) (i.e., content units relevant to the CPT task according to Yorkston and Beukelman⁶⁹ and Craig et al.,⁷⁰

(vii) percentages of adjective CTP CUs, noun CTP CUs, verb CTP CUs, prepositional phrase CTP CUs, noun phrase CTP CUs, and verb phrase CTP CUs (of 52 total CUs),

(viii) communication efficiency, defined as the number of syllables divided by the number of CTP CUs according to standardized scoring,⁹² with lower syllables/CU (Syll/CU) reflecting more efficient discourse, and

For Analysis 3:

(ix) percentage of Core Lexicon Words (of 26 Core Lexicon Words).⁷¹

The Kruskal-Wallis rank test evaluated the differences in the distribution of age, education, and our measures of interest in the written narratives among the PPA variants and healthy controls. The Kruskal Wallis rank test was used in an analysis of single-word and phrase CUs. Pairwise comparisons were performed using Dunn's Post Hoc Comparisons. A Bonferroni correction was applied to account for multiple comparisons. The adjusted p-values for significance are included in the table legends. Differences in means for sex were tested with Pearson chi-square among the PPA variants and healthy controls. JASP 0.18.3 was used for statistical analysis.

Results

Analysis 1: part of speech

As shown in Table 4, after adjusting for multiple comparisons, the total number of words, the total number of syllables, and the total number of content words were significantly lower for svPPA than lvPPA. There was a trend for fewer total content words for svPPA than controls (p = 0.010; p_bonf = 0.05/6 = 0.008).

Table 4.

Analysis 1 results for total words, total syllables, content words, function words, and part of speech.

	PPA variants			Controlsn = 18	H, 95% CI [0, 7.8147]	p*	Dunn's Post Hoc Comparisons**
	lvPPAn = 16	nfvPPAn = 17	svPPAn = 16	Controlsn = 18	H, 95% CI [0, 7.8147]	p*	Dunn's Post Hoc Comparisons**
Total Words
mean	42.00	28.77	22.81	33.61	10.695	0.013	sv < lv : 0.001
SD	18.58	13.35	12.17	22.53	10.695	0.013	sv < lv : 0.001
Total Syllables
mean	55.50	37.06	28.75	45.89	13.729	0.003	sv < lv : <0.001
SD	24.83	15.61	14.24	28.36	13.729	0.003	sv < lv : <0.001
Content words
mean	21.44	14.59	11.69	18.06	13.221	0.004	sv < lv : <0.001
SD	10.89	5.59	5.07	9.10	13.221	0.004	sv < c : 0.010
Function words
mean	20.56	14.18	11.13	15.56	6.722	0.081
SD	11.20	8.95	7.92	13.69	6.722	0.081
Adjectives (%)
mean	2.97	1.27	1.03	2.24	7.122	0.068
SD	2.52	2.25	1.96	2.49	7.122	0.068
Adverbs (%)
mean	2.02	1.27	1.65	1.17	2.237	0.525
SD	2.65	2.12	3.66	1.85	2.237	0.525
Articles (%)
mean	15.00	17.08	12.90	12.77	2.325	0.508
SD	5.31	7.65	9.42	9.27	2.325	0.508
Conjunctions (%)
mean	2.96	3.98	3.67	1.58	3.801	0.284
SD	2.48	4.29	4.34	2.44	3.801	0.284
Nouns (%)
mean	27.90	35.95	37.42	35.61	7.641	0.054
SD	11.56	13.40	20.64	10.59	7.641	0.054
Particles (%)
mean	2.86	0.50	0.66	2.26	13.023	0.005	nfv < lv : 0.007
SD	2.84	1.20	2.09	2.46	13.023	0.005	sv < lv : 0.006
Prepositions (%)
mean	10.00	8.60	10.41	10.17	1.699	0.637
SD	4.42	4.95	6.99	5.62		0.637
Pronouns (%)
mean	5.77	5.52	6.39	2.14	5.505	0.138
SD	4.95	5.93	7.75	3.10	5.505	0.138
Verbs (%)
mean	18.15	14.56	16.18	19.29	7.713	0.052	nfv < c : 0.011
SD	5.93	4.16	6.48	5.38	7.713	0.052	nfv < lv : 0.90

SD: standard deviation; PPA: primary progressive aphasia; lvPPA: logopenic primary progressive aphasia; nfvPPA: nonfluent variant primary progressive aphasia; svPPA: semantic variant primary progressive aphasia; *p values were calculated using Kruskal-Wallis Test with Dunn's Post Hoc Comparisons. **p values were adjusted for six comparisons (controls-lv, controls-nfv, controls-sv, lv-nfv, lv-sv, nfv-sv); p_bonf = 0.05/6 = 0.008.

The percentages of particles were significantly lower for nfvPPA and svPPA than lvPPA. There was a trend for lower percentages of verbs in nfvPPA than in controls (p = 0.011, p_bonf = 0.05/6 = 0.008) and in lvPPA (p = 0.090, p_bonf = 0.05/6 = 0.008). The percentage of verbs was lowest for nfvPPA (14.56%) followed by svPPA (16.18%), lvPPA (18.15%), and then controls (19.29%) (Table 4; Figure 1(a)–(d); see Supplemental Table 3 for medians and mean ranks for total words, total syllables, content words, function words, and part of speech).

Figure 1.

Examples of written cookie theft picture descriptions by a healthy control individual (a) and by individuals with PPA (b-d). Colors indicate different parts of speech. (a) Example from Healthy Control Participant. (b) Example from a Participant with Logopenic Variant PPA. (c) Example from a Participant with Nonfluent Variant PPA. (d) Example from a Participant with Semantic Variant PPA (Color version available in online).

Analysis 2: content units and communication efficiency

After adjusting for multiple comparisons, percentages of CTP CUs were significantly lower for nfvPPA and svPPA than controls, but that was not the case for lvPPA. There was also a significantly lower percentage of CPT CUs for svPPA than lvPPA. As expected, communication efficiency (Syll/CU) was significantly lower for lvPPA, nfvPPA, and svPPA than controls (Table 5; see Supplemental Table 4 for medians and mean ranks for content units, communication efficiency, and Core Lexicon).

Table 5.

Analysis 2 results for content units and communication efficiency.

	PPA variants			Controlsn = 18	H, 95% CI [0, 7.8147]	p*	Dunn's Post Hoc comparisons**
	lvPPAn = 16	nfvPPAn = 17	svPPAn = 16	Controlsn = 18	H, 95% CI [0, 7.8147]	p*	Dunn's Post Hoc comparisons**
Content Units (% of 52)
mean	22.36	16.29	11.66	24.89	32.563	<0.001	nfv < c : <0.001
SD	7.36	4.77	5.37	6.55	32.563	<0.001	sv < c : <0.001
							sv < lv : <0.001
Adjective Content Units (%)
mean	3.24	0.49	1.32	1.96	5.776	0.123
SD	4.52	2.02	3.61	3.51	5.776	0.123
Noun Content Units (%)
mean	44.66	53.81	55.32	41.23	19.615	<0.001	nfv > c : <0.001
SD	8.26	15.41	12.65	6.20	19.615	<0.001	sv > c : <0.001
Verb Content Units (%)
mean	19.12	21.52	20.70	25.68	2.644	0.450
SD	11.96	11.60	13.63	5.99	2.644	0.450
Noun Phrase Content Units (%)
mean	5.99	2.61	1.67	2.25	9.744	0.021	sv < lv : 0.006
SD	5.17	6.62	4.59	4.62	9.744	0.021	sv < lv : 0.006
Preposition Phrase Content Units (%)
mean	18.46	17.61	17.89	20.24	0.496	0.920
SD	9.34	11.65	11.47	8.16	0.496	0.920
Verb Phrase Content Units (%)
mean	8.61	3.95	3.08	8.64	12.223	0.007	sv < c : 0.007
SD	5.60	6.75	7.20	6.22	12.223	0.007	sv < lv : 0.007
Communication Efficiency (%)
mean	4.77	4.63	5.48	0.34	39.514	<0.001	lv < c : <0.001
SD	1.46	2.10	3.35	0.13	39.514	<0.001	nfv < c : <0.001
							sv < c : <0.001

SD: standard deviation; PPA: primary progressive aphasia; lvPPA: logopenic primary progressive aphasia; nfvPPA: nonfluent variant primary progressive aphasia; svPPA: semantic variant primary progressive aphasia; *p values were calculated using Kruskal-Wallis Test with Dunn's Post Hoc Comparisons. **p values were adjusted for six comparisons (controls-lv, controls-nfv, controls-sv, lv-nfv, lv-sv, nfv-sv); p_bonf = 0.05/6 = 0.008.

We expanded the Cookie Theft analysis to compare PPA variants and controls for single-word CU [percentages of noun CUs (e.g., cookies, boy, girl), verb CUs (e.g., taking, drying, standing), adjective CU (e.g., little, two, three-legged)] and phrase CUs [percentages of noun phrase (e.g., brother standing, feet getting wet), verb phrase (e.g., asking for cookies, reaching up, overflowing onto the floor), and prepositional phrase CUs (e.g., on stool, in the kitchen, with the door open)]. There were significantly higher percentages of noun CUs for nfvPPA and svPPA than controls, but significantly lower percentages of noun phrase CUs for svPPA than lvPPA and significantly lower percentages of verb phrase CUs for svPPA than controls and lvPPA (Table 5; see Supplemental Table 4 for medians and mean ranks for content Units, and communication efficiency).

We also compared the composite distributions, that is, the percentages of each type of CUs (nouns, verbs, adjectives, noun phrases, verb phrases, prepositional phrases) among healthy controls and the PPA variants (Figure 2()–(d)). The composite distributions were not significantly different for controls and all PPA variants (Pearson chi-square = 14.04, df =15, p = 0.523); that is, among healthy controls and PPA variants, nouns comprised the largest percentage of content units in written picture descriptions, followed by verbs, prepositional phrases, verb phrases, noun phrases, and last, adjectives. Upon visual inspection, composite distributions of written CTP descriptions of participants with lvPPA closely resembled the CU type distribution of controls (Figure 2(a) and (b)). Composite distributions of written CTP descriptions of participants with nfvPPA were similar to the svPPA distribution (Figure 2(c) and (d)). There were higher percentages of nouns in written picture descriptions of participants with nfvPPA (53.81%) and svPPA (55.33%) than healthy controls (41.23%) and participants with lvPPA (44.66%), and lower percentages of verb phrases in written picture descriptions of participants with nfvPPA (3.95%) and svPPA (3.08%) than healthy controls (8.64%) and participants with lvPPA (8.61%). Analysis of these data supported visual inspection in part. The percentages of nouns were significantly higher for nfvPPA and svPPA versus healthy controls, and the percentages of verb phrases were significantly lower for svPPA than healthy controls and lvPPA (Table 5).

Figure 2.

Content unit types in cookie theft picture descriptions by healthy controls (a) and by individuals with PPA (b-d).

Analysis 3: core lexicon

Participants with svPPA had a significantly lower percentage of Core Lexicon Words than controls and lvPPA (Table 6; see Supplemental Table 4 for medians and mean ranks for Core Lexicon).

Table 6.

Analysis 3 results for core lexicon.

	PPA variants			Controlsn = 18	H, 95% CI [0, 7.8147]	p*	Dunn's Post Hoc comparisons**
	lvPPAn = 16	nfvPPAn = 17	svPPAn = 16	Controlsn = 18	H, 95% CI [0, 7.8147]	p*	Dunn's Post Hoc comparisons**
Core Lexicon (% of 26)
mean	52.40	43.22	31.97	48.50	18.140	<0.001	sv < c : 0.001
SD	11.74	12.50	13.95	11.66	18.140	<0.001	sv < lv : <0.001

SD: standard deviation; PPA: primary progressive aphasia; lvPPA: logopenic primary progressive aphasia; nfvPPA: nonfluent variant primary progressive aphasia; svPPA: semantic variant primary progressive aphasia; *p values were calculated using Kruskal-Wallis Test with Dunn's Post Hoc Comparisons. **p values were adjusted for six comparisons (controls-lv, controls-nfv, controls-sv, lv-nfv, lv-sv, nfv-sv); p_bonf = 0.05/6 = 0.008.

Discussion

Analysis of written discourse in the three primary progressive aphasia variants

We found that the part of speech analysis showed a deficit in function words in nfvPPA, with significantly lower percentages of particles in the written output of participants with nfvPPA compared to lvPPA, and a trend for fewer verbs than in controls, and that the content unit analysis distinguished svPPA and lvPPA with significantly lower percentages of total CTP CUs, CTP noun phrase, and CTP verb phrase CUs in svPPA than lvPPA. When considering percentages of parts of speech, the percentages of particles were significantly lower for nfvPPA and svPPA than lvPPA, distinguishing lvPPA from the other two variants. Particles “do not fit conveniently into other classes of words,”⁹³ and do not change form through inflection.⁹⁴ Particles include function words (e.g., adverbial particles such as “back” in the phrase “come back,” the infinitival particle “to” in the phrase “to get the cookies,” discourse particles such as “anyway” in the sentence “Anyway he is falling”, negative particles such as “not” in the sentence “She is not paying attention”), and this finding is, in part, consistent with our hypothesis that participants with nfvPPA would generate fewer function words than lvPPA and controls. There was a strong trend for participants with nfvPPA to have the lowest percentage of verbs compared to controls and lvPPA in written output.

Participants with svPPA had significantly fewer total words, total syllables, and content words than participants with lvPPA. This finding may be expected given that participants with svPPA were more impaired in word semantics as evidenced by lower scores on tests of confrontation naming than participants with lvPPA and nfvPPA. Participants with lvPPA generated more total words, total syllables, content words, and function words than controls, although this content was not always relevant to the picture stimulus with significantly lower communication efficiency for lvPPA than controls [communication efficiency (Syll/CU) 4.77 for lvPPA versus 0.34 for controls, see Table 5].

The content unit analysis distinguished nfvPPA and svPPA from controls with significantly lower percentages of CTP CUs (single word and phrase CUs) in nfvPPA and svPPA than controls, but significantly higher percentages of noun CUs in these variants than controls. Percentages of noun and verb phrases were significantly lower in svPPA than in lvPPA; the percentage of verb phrases was significantly lower in svPPA than in controls. Overall, written CTP descriptions of participants with svPPA were most dissimilar to controls whereas those of lvPPA were more closely aligned. As expected, the communication efficiency (Syll/CU) of the PPA variants paled in comparison to controls.

Like the part of speech analysis, the Core Lexicon analysis distinguished svPPA from controls and lvPPA. The significantly lower percentage of Core Lexicon Words in svPPA than controls and lvPPA is consistent with the finding that participants with svPPA generated fewer total content words than controls and fewer total words, total syllables, and content words than participants with lvPPA.

In summary, our results indicate the following key findings for each PPA variant:

The written output in nfvPPA is characterized by significantly fewer particles than lvPPA, and a trend for fewer verbs than controls.

The written output in svPPA is more limited, with significantly fewer total words, content words, CTP total CUs, CTP noun CUs, and Core Lexicon Words than lvPPA, and is syntactically more simplistic, with significantly fewer phrases than lvPPA and controls.

The written narrative composite profile in lvPPA is similar to that of controls but is also characterized by a significantly higher percentage of verb phrase CUs than in svPPA and controls, as well as irrelevant content manifested by a high syllable-to-content ratio.

Difficulty naming, or anomia, is present in all three PPA variants. Some have reported differences when considering word-class deficits. Greater verb naming deficits have been found in nfaPPA^72–77 and greater noun naming deficits have been reported in svPPA and lvPPA.^74–77,78 These conclusions are primarily drawn from performance on confrontation naming tasks rather than connected speech. Performance on written discourse is congruent in part with these core deficits seen in confrontation naming. All the PPA variants were less efficient in their written picture descriptions than controls. We did not compare noun and verb naming; however, there was a trend for fewer verbs in the written narratives of nfvPPA than controls. However, the percentages of nouns were not significantly different among the PPA variants and healthy controls. Anomia is prominent in svPPA due to impaired semantic knowledge.⁴ Impaired semantic knowledge was manifested in the written output of participants with svPPA by participants with this variant writing fewer total words, fewer content words, fewer CTP total CUs, fewer CTP noun CUs, and fewer Core Lexicon Words compared to lvPPA and controls. The written narratives of participants with svPPA were characterized primarily by single words (nouns) and limited phrases. This was somewhat unexpected as individuals with svPPA are thought to have relatively preserved phonological and syntax abilities.⁹⁵ A possible explanation is that impaired semantic knowledge characteristic of svPPA, rather than a pure syntactic deficit, compromised the ability to combine words into informative, meaningful phrases.

Intersections with other studies of written discourse

We found a trend of a lower percentage of verbs in nfvPPA compared to controls. These findings are consistent with Graham,⁴⁰ who found reduced verb production in CTP descriptions in nfvPPA compared to healthy controls, and Code,⁴⁵ who found a decline in the number of verbs over time in sentence-writing tasks. Sitek et al.⁴² reported frequent verb use in lvPPA compared to AD. We found higher percentages of verbs in lvPPA than in nfvPPA and higher percentages of CTP verb phrase CUs in lvPPA compared to svPPA.

Comparison of spoken and written discourse

Analyses of oral CTP descriptions generally have revealed grammatical impairment in nfvPPA compared to lvPPA and svPPA.^16,18 Themistocleous et al.²⁰ found that patients with nfvPPA produced fewer conjunctions, determiners, particles, and prepositions than patients with svPPA and lvPPA in spoken picture descriptions; however, the statistical models for these categories did not show statistically significant differences. Our results that there were fewer particles and verbs in the written output of participants with nfvPPA (significantly fewer particles in the written output of participants with nfvPPA compared to lvPPA, and a trend for fewer verbs than controls) align with the findings of Themistocleous et al.,²⁰ reflecting grammatical impairment in nfvPPA. Grammatical impairment in nfvPPA is characterized by decreased use of verbs which are fundamental to sentence structure,⁹⁶ and the loss of function words, which affects sentence formation and results in telegraphic output.¹³ Limited attention has been given to the use of adjectives in oral and written discourse in PPA. Interestingly, we found the lowest percentage of adjective CUs in the written narratives of participants with nfvPPA (0.49% adjectives, Table 5) and a low percentage of adjectives in the part of speech analysis for nfvPPA (1.27%, Table 4). Our findings that participants with nfvPPA produced few adjectives in written discourse appear consistent with Walenski et al.³⁰ who found significantly fewer attributive adjectives in oral story narrations of individuals with agrammatic PPA and agrammatic post-stroke aphasia than healthy controls. An analysis of contemporaneous spoken and written narratives is needed to discover similarities and differences in these domains.

Clinical implications

Incorporation of analyses of spoken and written connected speech, as a supplement to standard cognitive/language tests commonly used in clinical settings, is generally advocated to capture the impact of aphasia on functional communication despite concerns regarding the lack of standardization of elicitation techniques, data collection, and analysis.^34,97 The time required to transcribe and analyze discourse has been identified as the most significant barrier to the implementation of discourse analysis into the clinical environment.⁹⁸ Incorporation of written discourse may overcome this barrier to clinical implementation because clinicians do not need to transcribe the samples and can review writing samples for specific items from the CTP CU list or from compendiums of core lexicon checklists, perform main concept analyses, or calculate derived communication efficiency scores. These non-transcription-based discourse measures can reduce the amount of time required for discourse analysis, making clinical utilization a reality.⁹⁹ We expanded the CPT CUs analysis to consider single words and phrases separately as a reflection of overall narrative complexity which is also easily incorporated into the clinical setting. In addition, online platforms, such as Open Brain AI,⁹⁰ which we used in this paper, can improve the efficiency of discourse analysis.

The distinction between these variant types can be challenging to determine in clinical practice due to shared language characteristics among variants (e.g., anomia), speech and language features that obscure differential diagnosis, and variability in clinical presentation, particularly in nfvPPA and lvPPA.⁶⁴ Our findings suggest that analysis of written narratives focuses on particles, verbs, total words, content units, and overall complexity of the narratives to aid in diagnosis. Analysis of written narratives focusing on particles and verbs may aid in distinguishing nfvPPA and lvPPA with infrequent use of particles and verbs in nfvPPA and frequent use of verbs in lvPPA. Differential diagnosis may also be facilitated by considering the total number of words with fewer overall words consistent with svPPA and fewer phrases than single words in nfvPPA and svPPA, although for different underlying reasons as indicated above.

Limitations

A limitation of our study may be the elicitation stimulus. The narratives derived from the CTP may be restricted in syntactic variety and spontaneity compared to story narration and structured interviews. However, CTP is the most widely used picture stimulus in studies of discourse in populations with neurogenic communication impairments.¹⁰⁰ Strengths of picture description, in general, include its sensitivity to lexico-semantic abilities in PPA,^14,15 and availability of a standardized administration procedure and scoring system with published content units for the CTP.^69,70 A list of 64 content units is available for the WAB Picnic Scene, although this resource may be less well-known.¹⁰¹

In keeping with the original instructions for spoken descriptions of the CTP, there was not a prescribed time limit for the written CTP task because this task is intended to elicit as complete and accurate a response as possible. However, the unlimited duration of the task might introduce bias in the analysis since impairments in verbal and visual memory, processing speed, executive function, and visuospatial function may be mitigated and this might vary across PPA variants since their neurocognitive profiles differ.¹⁰²

Although we used percentages (except for total words, total syllables, and total content words) in each of our three analyses (part of speech, content units and communication efficiency, and Core Lexicon), another limitation may be the greater cognitive/language impairment in the svPPA group compared to the lvPPA and nfvPPA groups. To address this concern, we conducted a post hoc one-way analysis of covariance (ANCOVA), with CDR scores as a covariate. This allowed us to compare the PPA variants for those findings that showed a statistically significant difference among the PPA variants while removing the contribution of overall cognitive severity in explaining those differences. Specifically, we performed this analysis for: total words, total syllables, total content words, percentage of particles, percentage of content units (out of 52), percentage of noun phrase content units, percentage of verb phrase content units, and percentage of Core Lexicon words (out of 26). CDR scores were significant only for noun phrases. The results using CDR scores to control for differences in cognitive/language impairment in the PPA groups were identical to our original findings (Table 7).

Table 7.

Post hoc ANCOVA analysis of significant findings.

	Cases	Sum of squares	df	Mean square	F	p	η²	Post Hoc comparisons
	Cases	Sum of squares	df	Mean square	F	p	η²	Mean difference		SE	t	p_bonf
Total Words	Variant	3108.276	2	1554.138	6.853	0.003	0.233	lv – nfv	12.830	5.319	2.412	0.060
	CDR	47.759	1	47.759	0.211	0.649	0.004	lv – sv	19.998	5.610	3.565	0.003
	Residuals	10,205.737	45	226.794				nfv – sv	7.168	5.877	1.220	0.687
Total Syllables	Variant	6045.721	2	3022.860	8.452	<0.001	0.272	lv – nfv	17.866	6.680	2.675	0.031
	CDR	96.214	1	96.214	0.269	0.607	0.004	lv – sv	27.901	7.045	3.961	<0.001
	Residuals	16,093.728	45	357.638				nfv – sv	10.035	7.380	1.360	0.450
Content Words	Variant	855.679	2	427.839	7.367	0.002	0.247	lv – nfv	6.425	2.692	2.387	0.064
	CDR	52.238	1	52.238	0.900	0.348	0.015	lv – sv	10.598	2.839	3.733	0.002
	Residuals	2613.255	45	58.072				nfv – sv	4.173	2.974	1.403	0.502
Particles (%)	Variant	0.006	2	0.003	6.242	0.004	0.216	lv – nfv	0.023	0.008	2.993	0.013
	CDR	2.190 × 10⁻⁴	1	2.190 × 10⁻⁴	0.475	0.494	0.008	lv – sv	0.024	0.008	2.967	0.014
	Residuals	0.021	45	4.607 × 10⁻⁴				nfv – sv	0.001	0.008	0.123	1.000
Content Units (% of 52)	Variant	0.083	2	0.041	11.323	<0.001	0.323	lv – nfv	0.059	0.021	2.777	0.024
	CDR	2.151 × 10⁻⁴	1	2.151 × 10⁻⁴	0.059	0.809	8.392 × 10⁻⁴	lv – sv	0.105	0.023	4.676	<0.001
	Residuals	0.164	45	0.004				nfv – sv	0.046	0.024	1.950	0.172
Noun Phrase Content Units (%)	Variant	0.024	2	0.012	4.392	0.018	0.154	lv – nfv	0.025	0.019	1.354	0.548
	CDR	0.018	1	0.018	6.426	0.015	0.112	lv – sv	0.058	0.020	2.962	0.015
	Residuals	0.124	45	0.003				nfv – sv	0.033	0.020	1.602	0.348
Verb Phrase Content Units (%)	Variant	0.032	2	0.016	3.776	0.030	0.143	lv – nfv	0.042	0.023	1.829	0.222
	CDR	0.006	1	0.006	1.298	0.261	0.025	lv – sv	0.064	0.024	2.632	0.035
	Residuals	0.192	45	0.004				nfv – sv	0.022	0.025	0.857	1.000
Core Lexicon (% of 26)	Variant	0.284	2	0.142	8.574	<0.001	0.262	lv – nfv	0.095	0.045	2.094	0.126
	CDR	0.003	1	0.003	0.192	0.663	0.003	lv – sv	0.198	0.048	4.123	<0.001
	Residuals	0.746	45	0.017				nfv – sv	0.102	0.050	2.040	0.142

df:, degrees of freedom; lvPPA: logopenic primary progressive aphasia; nfvPPA: nonfluent variant primary progressive aphasia; svPPA: semantic variant primary progressive aphasia; p values were adjusted for three comparisons (lv-nfv, lv-sv, nfv-sv); p_bonf = 0.05/3 = 0.016.

The bold type indicates significant values for easier reading in this very detailed table.

Future directions

A longitudinal study of written discourse may have the potential to facilitate early detection of PPA and change over time, particularly lvPPA which is typically associated with AD pathology. Analyses of writing samples have been a fruitful avenue of research in nfvPPA⁴⁵ and svPPA.^46,47 Longitudinal analysis of writing samples of famous authors diagnosed with AD uncovered subtle changes that reflected the onset of AD^103–105 and findings from the Nun Study revealed changes in idea density which were strongly related to dementia and AD in later life.¹⁰⁶ Picture description, despite its limitations, is well-suited to the assessment of change over time because the stimulus is readily accessible, repeatable, and easily analyzed using published criteria and automated approaches. In addition, comparison of spoken and written discourse using the same stimulus has received little attention to date. Spoken and written picture descriptions have only been compared in nfvPPA and healthy controls by Graham et al.⁴⁰ and in all PPA variants and healthy controls by Josephy-Hernandez et al.⁴⁴

Analysis of written discourse may serve as a potent measure of the effect of an intervention on functional communication. Written communication is an increasingly important activity of daily living because of the expanded application of technology in everyday life. Treatment success is commonly measured at the impairment level on trained and untrained confrontation naming tasks. However, since the ultimate goal of speech-language treatment is to facilitate successful communication in daily life (e.g., function-based: asking for directions, talking on the phone), assessment of the generalization of treatment at the functional level is vital and consistent with the domains of the International Classification of Functioning Disability and Health—body functions and structure (impairment-based), and activity and participation (function-based).¹⁰⁷ Connected speech samples, such as picture descriptions, are thought to more closely reflect communication as it occurs in everyday life and can provide compelling evidence of treatment effect at a life participation level.

Footnotes

Acknowledgements

We wish to acknowledge funding from NIH to support this work. We extend our sincerest gratitude to the participants, their families, and referring physicians for their dedication and interest in our study.

Ethical considerations

The research was planned, conducted, and recorded according to the principles of the Declaration of Helsinki, seventh edition. Ethical approval was obtained from the Johns Hopkins University School of Medicine Institutional Review Board (NA_00071337; NA_00042097; IRB00201027) and the University of South Florida Institutional Review Board (IRB005068). For clinical patients, ethical approval for chart review, data extraction, and data entry into a secured, de-identified spreadsheet from the electronic medical record was obtained from the Johns Hopkins University School of Medicine Institutional Review Board which approved this study as exempt research for which consent is not required (IRB00241582).

Consent to participate

The participants in the clinical trials or their legally authorized representatives provided written informed consent to participate in the studies.

Consent for publication

Not applicable.

Author contributions

Donna C Tippett: Conceptualization; Funding acquisition; Investigation; Methodology; Validation; Writing – original draft; Writing – review & editing

Katelyn Surrao: Formal analysis; Investigation; Validation; Writing – original draft; Writing – review & editing.

Kyriaki Neophytou: Writing – original draft; Writing – review & editing.

Hana Kim: Data curation; Investigation; Methodology; Writing – review & editing.

Jessica Gallegos: Data curation; Investigation; Writing – review & editing.

Charalambos Themistocleous: Resources; Software; Writing – review & editing.

Brenda Rapp: Methodology; Writing – review & editing.

Argye E Hillis: Data curation; Funding acquisition; Investigation; Methodology; Writing – review & editing.

Kyrana Tsapkini: Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Project administration; Resources; Writing – original draft; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Salary support for authors (DCT, KN, JG, BR, KT) was from awards NIH/NIA R01AG075404, R01AG068881, and R01AG075111 (National Institutes of Health, National Institute of Aging) to KT. For authors AEH and HK salary support was provided from award NIH/NIDCD R01 DC05375 to AEH. No funding was received for this work for KS and CT.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

Supplemental material

Supplemental material for this article is available online.

ORCID iDs

Donna C Tippett

Kyriaki Neophytou

Hana Kim

Charalambos Themistocleous

Brenda Rapp

Argye E Hillis

Kyrana Tsapkini

References

Mesulam

. Primary progressive aphasia. Ann Neurol 2001; 49: 425–432.

Mesulam

M-M

Rogalski

Wieneke

, et al. Primary progressive aphasia and the evolving neurology of the language network. Nat Rev Neurol 2014; 10: 554–569.

Montembeault

Brambati

Gorno-Tempini

, et al. Clinical, anatomical, and pathological features in the three variants of primary progressive aphasia: a review. Front Neurol 2018; 9: 692.

Gorno-Tempini

Hillis

Weintraub

, et al. Classification of primary progressive aphasia and its variants. Neurology 2011; 76: 1006–1014.

Josephs

Whitwell

Duffy

, et al. Progressive aphasia secondary to Alzheimer disease vs FTLD pathology. Neurology 2008; 70: 25–34.

Hodges

Mitchell

Dawson

, et al. Semantic dementia: demography, familial factors and survival in a consecutive series of 100 cases. Brain 2010; 133: 300–306.

Josephs

Hodges

Snowden

, et al. Neuropathological background of phenotypical variability in frontotemporal dementia. Acta Neuropathol 2011; 122: 137–153.

Rohrer

Lashley

Schott

, et al. Clinical and neuroanatomical signatures of tissue pathology in frontotemporal lobar degeneration. Brain 2011; 134: 2565–2581.

Modirrousta

Price

Dickerson

. Neuropsychiatric symptoms in primary progressive aphasia: phenomenology, pathophysiology, and approach to assessment and treatment. Neurodegen Dis Manage 2013; 3: 133–146.

10.

Mesulam

M-M

Weintraub

Rogalski

, et al. Asymmetry and heterogeneity of Alzheimer’s and frontotemporal pathology in primary progressive aphasia. Brain 2014; 137: 1176–1192.

11.

Leyton

Britton

Hodges

, et al. Distinctive pathological mechanisms involved in primary progressive aphasias. Neurobiol Aging 2016; 38: 82–92.

12.

Spinelli

Mandelli

Miller

, et al. Typical and atypical pathology in primary progressive aphasia variants. Ann Neurol 2017; 81: 430–443.

13.

Wilson

Henry

Besbris

, et al. Connected speech production in three variants of primary progressive aphasia. Brain 2010; 133: 2069–2088.

14.

Sajjadi

Patterson

Tomek

, et al. Abnormalities of connected speech in semantic dementia vs Alzheimer’s disease. Aphasiology 2012; 26: 847–866.

15.

Sajjadi

Patterson

Tomek

, et al. Abnormalities of connected speech in the non-semantic variants of primary progressive aphasia. Aphasiology 2012; 26: 1219–1237.

16.

Ash

Evans

O’Shea

, et al. Differentiating primary progressive aphasias in a brief sample of connected speech. Neurology 2013; 81: 329–336.

17.

Hilger

Ramsberger

Gilley

, et al. Analysing speech problems in a longitudinal case study of logopenic variant PPA. Aphasiology 2014; 28: 840–861.

18.

Faroqi-Shah

Treanor

Ratner

, et al. Using narratives in differential diagnosis of neurodegenerative syndromes. J Commun Disord 2020; 85: 105994.

19.

Lavoie

Black

Tang-Wai

, et al. Description of connected speech across different elicitation tasks in the logopenic variant of primary progressive aphasia. Int J Lang Comm Disord 2021; 56: 1074–1085.

20.

Themistocleous

Ficek

Webster

, et al. Automatic subtyping of individuals with primary progressive aphasia. J Alzhehimers Dis 2021; 79: 1185–1194.

21.

Cho

Olm

Ash

, et al. Automatic classification of AD pathology in FTD phenotypes using natural speech. Alzheimers Dement 2024; 20: 3416–3428.

22.

Lukic

Fan

García

, et al. Discriminating nonfluent/agrammatic and logopenic PPA variants with automatically extracted morphosyntactic measures from connected speech. Cortex 2024; 173: 34–48.

23.

Meteyard

Patterson

. The relation between content and structure in language production: an analysis of speech errors in semantic dementia. Brain Lang 2009; 110: 121–134.

24.

Ash

Moore

Antani

, et al. Trying to tell a tale: discourse impairments in progressive aphasia and frontotemporal dementia. Neurology 2006; 66: 1405–1413.

25.

Kavé

Leonard

Cupit

, et al. Structurally well-formed narrative production in the face of severe conceptual deterioration: a longitudinal case study of a woman with semantic dementia. J Neurolinguist 2007; 20: 161–177.

26.

Ash

McMillan

Gunawardena

, et al. Speech errors in progressive non-fluent aphasia. Brain Lang 2010; 113: 13–20.

27.

Thompson

Cho

Hsu

C-J

, et al. Dissociations between fluency and agrammatism in primary progressive aphasia. Aphasiology 2012; 26: 20–43.

28.

Mack

Chandler

Meltzer-Asscher

, et al.

What do pauses in narrative production reveal about the nature of word retrieval deficits in PPA?

Neuropsychologia 2015; 77: 211–222.

29.

Mack

Barbieri

Weintraub

, et al. Quantifying grammatical impairments in primary progressive aphasia: structured language tests and narrative language production. Neuropsychologia 2021; 151: 107713.

30.

Walenski

Sostarics

Mesulam

, et al. The production of adjectives in narratives by individuals with primary progressive aphasia. J Neurolinguist 2024; 69: 101179.

31.

Goodglass

Kaplan

Barresi

. BDAE-3: Boston diagnostic aphasia examination. 3rd ed. Philadelphia, PA: Williams and Wilkins, 2001.

32.

Kertesz

. WAB-R: Western aphasia battery–revised. San Antonio, TX: Pearson Assessments USA, 2006.

33.

Canu

Agosta

Lumaca

, et al. Connected speech alterations and progression in patients with primary progressive aphasia variants. Neurology 2025; 104: e213524.

34.

Kim

Obermeyer

Wiley

. Written discourse in diagnosis for acquired neurogenic communication disorders: current evidence and future directions. Front Hum Neurosci 2024; 17: 1264582.

35.

Sepelyak

Crinion

Molitoris

, et al. Patterns of breakdown in spelling in primary progressive aphasia. Cortex 2011; 47: 342–352.

36.

Shim

Hurley

Rogalski

, et al. Anatomic, clinical, and neuropsychological correlates of spelling errors in primary progressive aphasia. Neuropsychologia 2012; 50: 1929–1935.

37.

Faria

Crinion

Tsapkini

, et al. Patterns of dysgraphia in primary progressive aphasia compared to post-stroke aphasia. Behav Neurol 2013; 26: 21–34.

38.

Graham

. Dysgraphia in primary progressive aphasia: characterisation of impairments and therapy options. Aphasiology 2014; 28: 1092–1111.

39.

Neophytou

Wiley

Rapp

, et al. The use of spelling for variant classification in primary progressive aphasia: theoretical and practical implications. Neuropsychologia 2019; 133: 107157.

40.

Graham

Patterson

Hodges

. When more yields less: speaking and writing deficits in nonfluent progressive aphasia. Neurocase 2004; 10: 141–155.

41.

Sitek

Barczak

Kluj-Kozłowska

, et al. Writing in Richardson variant of progressive supranuclear palsy in comparison to progressive non-fluent aphasia. Neurol Neurochir Pol 2015; 49: 217–222.

42.

Sitek

Barczak

Kluj-Kozłowska

, et al.

Is descriptive writing useful in the differential diagnosis of logopenic variant of primary progressive aphasia, Alzheimer’s disease and mild cognitive impairment?

Neurol Neurochir Pol 2015; 49: 239–244.

43.

Tetzloff

Utianski

Duffy

, et al. Quantitative analysis of agrammatism in agrammatic primary progressive aphasia and dominant apraxia of speech. J Speech Lang Hear Res 2018; 61: 2337–2346.

44.

Josephy-Hernandez

Rezaii

Jones

, et al. Automated analysis of written language in the three variants of primary progressive aphasia. Brain Commun 2023; 5: fcad202.

45.

Code

Muller

Tree

, et al. Syntactic impairments can emerge later: progressive agrammatic agraphia and syntactic comprehension impairment. Aphasiology 2006; 20: 1035–1058.

46.

Heitkamp

Schumacher

Croot

, et al. A longitudinal linguistic analysis of written text production in a case of semantic variant primary progressive aphasia. J Neurolinguist 2016; 39: 26–37.

47.

Hwang

Strikwerda-Brown

El-Omar

, et al. “More than words” – longitudinal linguistic changes in the works of a writer diagnosed with semantic dementia. Neurocase 2021; 27: 243–252.

48.

Tetzloff

Meade

Duffy

, et al. Word count matters: features of written language production in progressive apraxia of speech with and without agrammatism. Folia Phoniatr Logop 2025; 77: 375–383.

49.

Grossman

. The non-fluent/agrammatic variant of primary progressive aphasia. Lancet Neurol 2012; 11: 545–555.

50.

Thompson

Cho

Price

, et al. Semantic interference during object naming in agrammatic and logopenic primary progressive aphasia (PPA). Brain Lang 2012; 120: 237–250.

51.

Tetzloff

Duffy

Clark

, et al. Progressive agrammatic aphasia without apraxia of speech as a distinct syndrome. Brain 2019; 142: 2466–2482.

52.

Josephs

Duffy

Strand

, et al. Characterizing a neurodegenerative syndrome: primary progressive apraxia of speech. Brain 2012; 135: 1522–1536.

53.

Jung

Duffy

Josephs

. Primary progressive aphasia and apraxia of speech. Semin Neurol 2013; 33: 342–347.

54.

Botha

Josephs

. Primary progressive aphasias and apraxia of speech. Continuum (Minneap Minn) 2019; 25: 101–127.

55.

Nestor

. Progressive non-fluent aphasia is associated with hypometabolism centred on the left anterior insula. Brain 2003; 126: 2406–2418.

56.

Gorno-Tempini

Ogar

Brambati

, et al. Anatomical correlates of early mutism in progressive nonfluent aphasia. Neurology 2006; 67: 1849–1851.

57.

Knibb

Woollams

Hodges

, et al. Making sense of progressive non-fluent aphasia: an analysis of conversational speech. Brain 2009; 132: 2734–2746.

58.

Josephs

Duffy

Fossett

, et al. Fluorodeoxyglucose F18 positron emission tomography in progressive apraxia of speech and primary progressive aphasia variants. Arch Neurol 2010; 67: 596–605.

59.

Croot

Ballard

Leyton

, et al. Apraxia of speech and phonological errors in the diagnosis of nonfluent/agrammatic and logopenic variants of primary progressive aphasia. J Speech Lang Hear Res 2012; 55: S1562–S1572.

60.

Duffy

Strand

Clark

, et al. Primary progressive apraxia of speech: clinical features and acoustic and neurologic correlates. Am J Speech Lang Pathol 2015; 24: 88–100.

61.

Tippett

Rapp

, et al. Speech motor planning relies on multiple brain areas: evidence from neurodegenerative disease. In: Annual convention of the association of psychological science, Washington, DC, May 2025.

62.

Ogar

Dronkers

Brambati

, et al. Progressive nonfluent aphasia and its characteristic motor speech deficits. Alzheimer Dis Assoc Disord 2007; 21: S23–S30.

63.

Duffy

Strand

Josephs

. Motor speech disorders associated with primary progressive aphasia. Aphasiology 2014; 28: 1004–1017.

64.

Tippett

. Classification of primary progressive aphasia: challenges and complexities. F1000Res 2020; 9: 64.

65.

Enderby

Palmer

. FDA-2: Frenchay dysarthria assessment. Austin, TX: Pro-Ed, 2008.

66.

Rohrer

Knight

Warren

, et al. Word-finding difficulty: a clinical analysis of the progressive aphasias. Brain 2007; 131: 8–38.

67.

Pulgram

. Graphic and phonic systems: figurae and signs. WORD 1965; 21: 208–224.

68.

Vachek

. Written language revisited. Amsterdam: John Benjamins Publishing Company, 1989.

69.

Yorkston

Beukelman

. An analysis of connected speech samples of aphasic and normal speakers. J Speech Hear Disord 1980; 45: 27–36.

70.

Craig

Hinckley

Winkelseth

, et al. Quantifying connected speech samples of adults with chronic aphasia. Aphasiology 1993; 7: 155–163.

71.

Dalton

Al Harbi

Berube

, et al. Development of main concept and core lexicon checklists for the original and modern Cookie Theft stimuli. Aphasiology 2024; 38: 1975–1999.

72.

Cotelli

Borroni

Manenti

, et al. Action and object naming in frontotemporal dementia, progressive supranuclear palsy, and corticobasal degeneration. Neuropsychology 2006; 20: 558–565.

73.

Hillis

Tuffiash

Caramazza

. Modality-specific deterioration in naming verbs in nonfluent primary progressive aphasia. J Cogn Neurosci 2002; 14: 1099–1108.

74.

Hillis

Ken

. Deterioration of naming nouns versus verbs in primary progressive aphasia. Ann Neurol 2004; 55: 268–275.

75.

Hillis

Heidler-Gary

Newhart

, et al. Naming and comprehension in primary progressive aphasia: the influence of grammatical word class. Aphasiology 2006; 20: 246–256.

76.

Silveri

Ciccarelli

. Naming of grammatical classes in frontotemporal dementias: linguistic and nonlinguistic factors contribute to noun-verb dissociation. Behav Neurol 2007; 18: 197–206.

77.

Thompson

Lukic

King

, et al. Verb and noun deficits in stroke-induced and primary progressive aphasia: the Northwestern naming battery. Aphasiology 2012; 26: 632–655.

78.

Bak

Hodges

. Kissing and dancing—a test to distinguish the lexical and conceptual contributions to noun/verb and action/object dissociation. Preliminary results in patients with frontotemporal dementia. J Neurolinguist 2003; 16: 169–181.

79.

Mesulam

Rogalski

Wieneke

, et al. Neurology of anomia in the semantic variant of primary progressive aphasia. Brain 2009; 132: 2553–2565.

80.

Grossman

. Linguistic aspects of primary progressive aphasia. Annu Rev Linguist 2018; 4: 377–403.

81.

Wang

Walenski

Litcofsky

, et al. Verb production and comprehension in primary progressive aphasia. J Neurolinguist 2022; 64: 101099.

82.

Folstein

McHugh

. Mini-mental state. J Psychiatr Res 1975; 12: 189–198.

83.

Williams

Mack

Henderson

. Boston Naming test in Alzheimer’s disease. Neuropsychologia 1989; 27: 1073–1079.

84.

Mack

Freed

Williams

, et al. Boston naming test: shortened versions for use in Alzheimer’s disease. J Gerontol 1992; 47: P154–P158.

85.

Breining

Faria

Caffo

, et al. Neural regions underlying object and action naming: complementary evidence from acute stroke and primary progressive aphasia. Aphasiology 2022; 36: 732–760.

86.

Howard

Patterson

. The pyramids and palm trees test: a test of semantic access from words and pictures. Cambridge, UK: Pearson, 1992.

87.

Breining

Tippett

Davis

, et al. Assessing dissociations of object and action naming in acute stroke. In: Clinical aphasiology conference proceedings, Monterey, California, May 2015, p.3.

88.

Hughes

Berg

Danziger

, et al. A new clinical scale for the staging of dementia. Br J Psychiatry 1982; 140: 566–572.

89.

Morris

. The clinical dementia rating (CDR): current version and scoring rules. Neurology 1993; 43: 2412.

90.

Themistocleous

. Open brain AI and language assessment. Front Hum Neurosci 2024; 18: 1421435.

91.

Trupe

Hillis

. Paucity vs. verbosity: another analysis of right hemisphere communication deficits. In: Brooksire

(ed.) Clinical aphasiology. Minneapolis, MN: BRK Publishers, 1985, pp.83–96.

92.

Agis

Goggins

Oishi

, et al. Picturing the size and site of stroke with an expanded national institutes of health stroke scale. Stroke 2016; 47: 1459–1465.

93.

Leech

. A glossary of English grammar. Edinburgh: Edinburgh University Press, 2006.

94.

McArthur

. Concise Oxford companion to the English language. Oxford paperback reference. Oxford: Oxford University Press, 2005.

95.

Kertesz

Jesso

Harciarek

, et al. What is semantic dementia?: a cohort study of diagnostic features and clinical boundaries. Arch Neurol 2010; 67: 483–489.

96.

Ash

Moore

Vesely

, et al. Non-fluent speech in frontotemporal lobar degeneration. J Neurolinguist 2009; 22: 370–383.

97.

Stark

Dutta

Murray

, et al. Standardizing assessment of spoken discourse in aphasia: a working group with deliverables. Am J Speech Lang Pathol 2021; 30: 491–502.

98.

Bryant

Spencer

Ferguson

. Clinical use of linguistic discourse analysis for the assessment of language in aphasia. Aphasiology 2017; 31: 1105–1126.

99.

Dalton

SGH

Hubbard

Richardson

. Moving toward non-transcription based discourse analysis in stable and progressive aphasia. Semin Speech Lang 2020; 41: 032–044.

100.

Berube

Nonnemacher

Demsky

, et al. Stealing cookies in the twenty-first century: measures of spoken narrative in healthy versus speakers with aphasia. Am J Speech Lang Pathol 2019; 28: 321–329.

101.

Gallée

Cordella

Fedorenko

, et al. Breakdowns in informativeness of naturalistic speech production in primary progressive aphasia. Brain Sci 2021; 11: 130.

102.

Butts

Machulda

Duffy

, et al. Neuropsychological profiles differ among the three variants of primary progressive aphasia. J Int Neuropsychol Soc 2015; 21: 429–435.

103.

Garrard

. The effects of very early Alzheimer’s disease on the characteristics of writing by a renowned author. Brain 2004; 128: 250–260.

104.

Lancashire

Hirst

, et al. Longitudinal detection of dementia through lexical and syntactic changes in writing: a case study of three British novelists. Lit LingComput 2011; 26: 435–461.

105.

Marckx

Verhoeven

Daelemans

. The Claus case: exploring the use of propositional idea density for Alzheimer detection. Comput Linguistic Netherl J 2018; 8: 66–82.

106.

Riley

Snowdon

Desrosiers

, et al. Early life linguistic ability, late life cognitive function, and neuropathology: findings from the Nun study. Neurobiol Aging 2005; 26: 341–347.

107.

Galletta

Barrett

. Impairment and functional interventions for aphasia: having it all. Curr Phys Med Rehabil Rep 2014; 2: 114–120.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB