Semantic weight and verb retrieval in aphasia

Artificial intelligence classifies primary progressive aphasia from connected speech

Neurodegenerative dementia syndromes, such as primary progressive aphasias (PPA), have traditionally been diagnosed based, in part, on verbal and non-verbal cognitive profiles. Debate continues about whether PPA is best divided into three variants and regarding the most distinctive linguistic features for classifying PPA variants. In this cross-sectional study, we initially harnessed the capabilities of artificial intelligence and natural language processing to perform unsupervised classification of short, connected speech samples from 78 pateints with PPA. We then used natural language processing to identify linguistic features that best dissociate the three PPA variants. Large language models discerned three distinct PPA clusters, with 88.5% agreement with independent clinical diagnoses. Patterns of cortical atrophy of three data-driven clusters corresponded to the localization in the clinical diagnostic criteria. In the subsequent supervised classification, 17 distinctive features emerged, including the observation that separating verbs into high- and low-frequency types significantly improved classification accuracy. Using these linguistic features derived from the analysis of short, connected speech samples, we developed a classifier that achieved 97.9% accuracy in classifying the four groups (three PPA variants and healthy controls). The data-driven section of this study showcases the ability of large language models to find natural partitioning in the speech of patients with PPA consistent with conventional variants. In addition, the work identifies a robust set of language features indicative of each PPA variant, emphasizing the significance of dividing verbs into high- and low-frequency categories. Beyond improving diagnostic accuracy, these findings enhance our understanding of the neurobiology of language processing.

Rational adaptation in word production: Strong conceptual ability reduces the effect of lexical impairments on verb retrieval in aphasia

Language users rely on both linguistic and conceptual processing abilities to efficiently comprehend or produce language. According to the principle of rational adaptation, the degree to which a cognitive system relies on one process vs. another can change under different conditions or disease states with the goal of optimizing behavior. In this study, we investigated rational adaptation in reliance on linguistic versus conceptual processing in aphasia, an acquired disorder of language. In individuals living with aphasia, verb-retrieval impairments are a pervasive deficit that negatively impacts communicative function. As such, we examined evidence of adaptation in verb production, using parallel measures to index impairment in two of verb naming's critical subcomponents: conceptual and linguistic processing. These component processes were evaluated using a standardized assessment battery designed to contrast non-linguistic (picture input) and linguistic (word input) tasks of conceptual action knowledge. The results indicate that non-linguistic conceptual action processing can be impaired in people with aphasia and contributes to verb-retrieval impairments. Furthermore, relatively unimpaired conceptual action processing can ameliorate the influence of linguistic processing deficits on verb-retrieval impairments. These findings are consistent with rational adaptation accounts, indicating that conceptual processing plays a key role in language function and can be leveraged in rehabilitation to improve verb retrieval in adults with chronic aphasia.

Using Generative Artificial Intelligence to Classify Primary Progressive Aphasia from Connected Speech

Neurodegenerative dementia syndromes, such as Primary Progressive Aphasias (PPA), have traditionally been diagnosed based in part on verbal and nonverbal cognitive profiles. Debate continues about whether PPA is best subdivided into three variants and also regarding the most distinctive linguistic features for classifying PPA variants. In this study, we harnessed the capabilities of artificial intelligence (AI) and natural language processing (NLP) to first perform unsupervised classification of concise, connected speech samples from 78 PPA patients. Large Language Models discerned three distinct PPA clusters, with 88.5% agreement with independent clinical diagnoses. Patterns of cortical atrophy of three data-driven clusters corresponded to the localization in the clinical diagnostic criteria. We then used NLP to identify linguistic features that best dissociate the three PPA variants. Seventeen features emerged as most valuable for this purpose, including the observation that separating verbs into high and low-frequency types significantly improves classification accuracy. Using these linguistic features derived from the analysis of brief connected speech samples, we developed a classifier that achieved 97.9% accuracy in predicting PPA subtypes and healthy controls. Our findings provide pivotal insights for refining early-stage dementia diagnosis, deepening our understanding of the characteristics of these neurodegenerative phenotypes and the neurobiology of language processing, and enhancing diagnostic evaluation accuracy.

Semantic Aspects of Verb Production in Various Discourse Tasks in People With Nonfluent Aphasia

PURPOSE

The purpose of this study was to investigate how people with nonfluent aphasia produce semantically weighted verbs compared to people without aphasia, as well as how a discourse elicitation task affects verb production in people with nonfluent aphasia and people without aphasia.

METHOD

This study included 30 people with nonfluent aphasia and 32 age-matched people without aphasia from AphasiaBank. Language samples of five different discourse tasks were obtained and coded for heavy, light, and be-copular verbs. The number of verbs per utterance and the proportion of heavy, light, and be-copular verbs were compared between groups and between tasks.

RESULTS

People with nonfluent aphasia showed a similar proportion of heavy verbs but reduced verbs per utterance and proportion of light verbs compared to people without aphasia. With regard to discourse task effects, we found a trend for a higher proportion of heavy verbs in sequential picture descriptions, and a higher proportion of be-copular verbs and lower proportion of heavy verbs for a recount compared to other tasks in people without aphasia. The discourse task effects were minimally found in people with nonfluent aphasia.

CONCLUSIONS

Our results suggest that people with nonfluent aphasia present with relatively preserved heavy verb production but with impaired production of light verbs in discourse. In addition, it appears that discourse tasks do not significantly influence the type of verbs produced by people with nonfluent aphasia possibly due to the floor effects and wide range of individual variability. This study is a preliminary effort to evaluate methodological factors that impact verb production; future studies are needed to develop a framework for clinical decision making when selecting a discourse elicitation task for people with aphasia.

A reconceptualization of sentence production in post-stroke agrammatic aphasia: the synergistic processing bottleneck model

The language production deficit in post-stroke agrammatic aphasia (PSA-G) tends to result from lesions to the left inferior frontal gyrus (LIFG) and is characterized by a triad of symptoms: fragmented sentences, errors in functional morphology, and a dearth of verbs. Despite decades of research, the mechanisms underlying production patterns in PSA-G have been difficult to characterize. Two major impediments to progress may have been the view that it is a purely morphosyntactic disorder and the (sometimes overzealous) application of linguistic theory without interceding psycholinguistic evidence. In this paper, empirical evidence is examined to present an integrated portrait of language production in PSA-G and to evaluate the assumption of a syntax-specific syndrome. In light of extant evidence, it is proposed that agrammatic language production results from a combination of morphosyntactic, phonomotor, and processing capacity limitations that cause a cumulative processing bottleneck at the point of articulatory planning. This proposed Synergistic Processing Bottleneck model of PSA-G presents a testable framework for future research. The paper ends with recommendations for future research on PSA-G.

Verb Frequency and Density Drive Naming Performance in Primary Progressive Aphasia

Background

Recent work has highlighted the utility of the Boston Naming Test and Hopkins Action Naming Assessment (HANA) for distinguishing between semantic (svPPA), logopenic (lvPPA) and non-fluent agrammatic (nfavPPA) variants of primary progressive aphasia (PPA).

Aims

To determine whether item level differences between variants on when naming verbs on the HANA were able to be accounted for using common variables of lexical interest: word frequency, semantic density, concreteness, or valency. We also examined three specific hypotheses: (1) svPPA and lvPPA may result in increased difficulty with decreased semantic density compared to nfavPPA; (2) svPPA may result in increased difficulty with decreased concreteness; and (3) nfavPPA may result in increased difficulty with high syntactic valency.

Methods & Procedures

268 patients with PPA were evaluated using the HANA. A hierarchical Bayesian regression approach was adopted to account for effects of repeated measurement within participants and items.

Outcomes & Results

The main effects of variant and verb trait were significant in all models, as was the interaction for frequency, semantic density, and valency. Increasing frequency, semantic density, and concreteness led to better performance, while increasing valency led to poorer performance. Low semantic density contributed to greater difficulty in svPPA and lvPPA, but low concreteness did not uniquely impact verb naming in svPPA. Those with nfavPPA had no particular difficulty as a result of valency.

Conclusions

Prior studies have identified the independent effects of frequency and semantic density on verb naming in PPA, which were confirmed by our analyses, and the best predictions of the data were achieved by combining these dimensions. This investigation complements our previous work highlighting the value of the HANA for efficiently demonstrating verb performance in PPA.

Neural evidence for a separation of semantic and phonological control processes

There remain major doubts about the nature and domain specificity of inhibitory control processes, both within and between cognitive domains. This study examined inhibitory processes within the language domain, by contrasting semantic versus phonological inhibitory control. In an fMRI experiment, elderly participants performed phonological and semantic inhibitory control tasks involving resistance to highly or weakly interfering stimuli. In the semantic domain, inhibitory control effects, contrasting high vs. low interference control levels, were observed at univariate and multivariate levels in all fronto-parieto-temporal region-of-interests. In the phonological domain, inhibitory control effects were observed only at multivariate levels, and were restricted to the pars triangularis of the bilateral inferior frontal gyrus and to the left middle temporal gyrus. Critically, no reliable multivariate cross-domain prediction of neural patterns associated with inhibitory control was observed. This study supports a functional dissociation of the neural substrates associated with inhibitory control for phonological vs. semantic domains.

Auditory Verb Generation Performance Patterns Dissociate Variants of Primary Progressive Aphasia

Primary progressive aphasia (PPA) is a clinical syndrome in which patients progressively lose speech and language abilities. Three variants are recognized: logopenic (lvPPA), associated with phonology and/or short-term verbal memory deficits accompanied by left temporo-parietal atrophy; semantic (svPPA), associated with semantic deficits and anterior temporal lobe (ATL) atrophy; non-fluent (nfvPPA) associated with grammar and/or speech-motor deficits and inferior frontal gyrus (IFG) atrophy. Here, we set out to investigate whether the three variants of PPA can be dissociated based on error patterns in a single language task. We recruited 21 lvPPA, 28 svPPA, and 24 nfvPPA patients, together with 31 healthy controls, and analyzed their performance on an auditory noun-to-verb generation task, which requires auditory analysis of the input, access to and selection of relevant lexical and semantic knowledge, as well as preparation and execution of speech. Task accuracy differed across the three variants and controls, with lvPPA and nfvPPA having the lowest and highest accuracy, respectively. Critically, machine learning analysis of the different error types yielded above-chance classification of patients into their corresponding group. An analysis of the error types revealed clear variant-specific effects: lvPPA patients produced the highest percentage of "not-a-verb" responses and the highest number of semantically related nouns (production of baseball instead of throw to noun ball); in contrast, svPPA patients produced the highest percentage of "unrelated verb" responses and the highest number of light verbs (production of take instead of throw to noun ball). Taken together, our findings indicate that error patterns in an auditory verb generation task are associated with the breakdown of different neurocognitive mechanisms across PPA variants. Specifically, they corroborate the link between temporo-parietal regions with lexical processing, as well as ATL with semantic processes. These findings illustrate how the analysis of pattern of responses can help PPA phenotyping and heighten diagnostic sensitivity, while providing insights on the neural correlates of different components of language.

Lack of selectivity for syntax relative to word meanings throughout the language network

To understand what you are reading now, your mind retrieves the meanings of words and constructions from a linguistic knowledge store (lexico-semantic processing) and identifies the relationships among them to construct a complex meaning (syntactic or combinatorial processing). Do these two sets of processes rely on distinct, specialized mechanisms or, rather, share a common pool of resources? Linguistic theorizing, empirical evidence from language acquisition and processing, and computational modeling have jointly painted a picture whereby lexico-semantic and syntactic processing are deeply inter-connected and perhaps not separable. In contrast, many current proposals of the neural architecture of language continue to endorse a view whereby certain brain regions selectively support syntactic/combinatorial processing, although the locus of such "syntactic hub", and its nature, vary across proposals. Here, we searched for selectivity for syntactic over lexico-semantic processing using a powerful individual-subjects fMRI approach across three sentence comprehension paradigms that have been used in prior work to argue for such selectivity: responses to lexico-semantic vs. morpho-syntactic violations (Experiment 1); recovery from neural suppression across pairs of sentences differing in only lexical items vs. only syntactic structure (Experiment 2); and same/different meaning judgments on such sentence pairs (Experiment 3). Across experiments, both lexico-semantic and syntactic conditions elicited robust responses throughout the left fronto-temporal language network. Critically, however, no regions were more strongly engaged by syntactic than lexico-semantic processing, although some regions showed the opposite pattern. Thus, contra many current proposals of the neural architecture of language, syntactic/combinatorial processing is not separable from lexico-semantic processing at the level of brain regions-or even voxel subsets-within the language network, in line with strong integration between these two processes that has been consistently observed in behavioral and computational language research. The results further suggest that the language network may be generally more strongly concerned with meaning than syntactic form, in line with the primary function of language-to share meanings across minds.

Effects of semantic weight on verb retrieval in individuals with aphasia: A different perspective

The majority of people with aphasia have word retrieval difficulty and effective treatment remains elusive. There have been a few studies that have explored the effects of semantic complexity on verb retrieval in individuals with aphasia; each used a variation of Breedin et al.'s (1998) delayed repetition/story completion task. Although each subsequent investigator worked to address potential confounds in order to achieve more valid results that would give rise to a clearer understanding of these deficits, findings and their interpretations have varied. In our replication, groups of individuals with aphasia (9 agrammatic and 9 anomic) plus 12 age-matched controls participated in a story completion task that included novel distracter stories to prevent rehearsal. Additionally, stimuli were developed in strict adherence to novel semantic and syntactic templates to control for relevant factors, and stimuli were prerecorded to ensure uniform delivery. We calculated the number of target verbs produced and overall production of light and heavy verbs, and error analysis was performed with special attention to semantically appropriate substitutions. In contrast to previous studies, we found no significant performance differences on these measures within or between groups. Exploratory analyses were performed. Results are discussed in terms of relevant factors of verb retrieval and implications for future experimental design. Application to much-needed verb retrieval treatment is also considered.

White Matter Disruption and Connected Speech in Non-Fluent and Semantic Variants of Primary Progressive Aphasia

Differential patterns of white matter disruption have recently been reported in the non-fluent (nfvPPA) and semantic (svPPA) variants of primary progressive aphasia (PPA). No single measure is sufficient to distinguish between the PPA variants, but connected speech allows for the quantification of multiple measures. The aim of the present study was to further investigate the white matter correlates associated with connected speech features in PPA. We examined the relationship between white matter metrics and connected speech deficits using an automated analysis of transcriptions of connected speech and diffusion tensor imaging in language-related tracts. Syntactic, lexical, and semantic features were automatically extracted from transcriptions of topic-directed interviews conducted with groups of individuals with nfvPPA or svPPA as well as with a group of healthy controls. A principal component analysis was performed in order to reduce the number of language measures and yielded a five-factor solution. The results indicated that nfvPPA patients differed from healthy controls on a syntactic factor, and svPPA patients differed from controls on two semantic factors. However, the patient groups did not differ on any factor. Moreover, a correlational analysis revealed that the lexical richness factor was significantly correlated with radial diffusivity in the left inferior longitudinal fasciculus, which suggests that semantic deficits in connected speech reflect a disruption of this ventral pathway, and which is largely consistent with the results of previous studies. Using an automated approach for the analysis of connected speech combined with probabilistic tractography, the present findings demonstrate that nfvPPA patients are impaired relative to healthy controls on syntactic measures and have increased radial diffusivity in the left superior longitudinal fasciculus, whereas the svPPA group was impaired on lexico-semantic measures relative to controls and showed increased radial diffusivity in the uncinate and inferior longitudinal fasciculus bilaterally.

Cross-Linguistic Differences in a Picture-Description Task Between Korean- and English-Speaking Individuals With Aphasia

PURPOSE

The purpose of the study was to examine cross-linguistic differences in a picture-description task between Korean- and English-speaking individuals with Broca's and anomic aphasia to determine whether a variation exists in the use of verbs and nouns across the language and aphasia groups.

METHOD

Forty-eight individuals (male = 29; female = 19) participated in the study (n = 28 for aphasic group and n = 20 healthy controls). Data for English speakers were obtained from the Aphasia Bank Project. We used the picture-description task to obtain connected speech samples and analyzed noun- and verb-related linguistic variables.

RESULTS

Korean speakers with aphasia produced more verbs per utterance and a lower noun-to-verb ratio than English speakers with aphasia, whereas English speakers produced more nouns than Korean speakers. The Korean anomic group produced more verbs than Korean speakers with Broca's aphasia, whereas no effects were significant for English speakers with aphasia depending on the type of aphasia.

CONCLUSIONS

Aphasia symptoms vary as a function of linguistic features of languages that individuals with aphasia used premorbidly. Furthermore, the linguistic characteristics of the individual's language influenced aphasia symptoms more strongly than the type of aphasia. It is theoretically and clinically important that this cross-linguistic study provides different perspectives, and that noun-verb deficits may vary depending on linguistic features.

Exploring generalisation processes following lexical retrieval intervention in primary progressive aphasia

PURPOSE

Within the current literature, positive intervention effects demonstrate the significant potential for people with Primary Progressive Aphasia (PPA) to learn/relearn words. Generalisation of intervention effects to other words and/or other contexts, however, remains unclear.

METHOD

This multiple baseline, case-series design investigated the effects of a self-cueing lexical retrieval intervention across word classes (nouns, verbs and adjectives) on four individuals with PPA, three suggestive of the semantic variant and one of the logopenic variant. The intervention integrated semantic, phonological and orthographic levels of language production and drew on autobiographical memory. Changes in accuracy in retrieving treated and untreated items (pre-intervention, post-intervention and 4-weeks maintenance) were determined using the Cochran's Q test, with follow-up McNemar pairwise comparisons.

RESULT

All participants showed significant improvements in naming treated items, across all word classes. Different patterns of generalised improvement to untreated words were found for each participant. In discourse, the semantic variant participants demonstrated a significant increase in correct information units, in contrast to the participant with the logopenic variant who remained stable.

CONCLUSION

This study provides evidence that people with PPA can show improved lexical retrieval following intervention. The findings suggest possible differences in generalisation across word classes and according to underlying deficit.

Language and iconic gesture use in procedural discourse by speakers with aphasia

Background: Conveying instructions is an everyday use of language, and gestures are likely to be a key feature of this. Although co-speech iconic gestures are tightly integrated with language, and people with aphasia (PWA) produce procedural discourses impaired at a linguistic level, no previous studies have investigated how PWA use co-speech iconic gestures in these contexts. Aims: This study investigated how PWA communicated meaning using gesture and language in procedural discourses, compared with neurologically healthy people (NHP). We aimed to identify the relative relationship of gesture and speech, in the context of impaired language, both overall and in individual events. Methods & Procedures: Twenty-nine PWA and 29 NHP produced two procedural discourses. The structure and semantic content of language of the whole discourses were analysed through predicate argument structure and spatial motor terms, and gestures were analysed for frequency and semantic form. Gesture and language were analysed in two key events, to determine the relative information presented in each modality. Outcomes & Results: PWA and NHP used similar frequencies and forms of gestures, although PWA used syntactically simpler language and fewer spatial words. This meant, overall, relatively more information was present in PWA gesture. This finding was also reflected in the key events, where PWA used gestures conveying rich semantic information alongside semantically impoverished language more often than NHP. Conclusions: PWA gestures, containing semantic information omitted from the concurrent speech, may help listeners with meaning when language is impaired. This finding indicates gesture should be included in clinical assessments of meaning-making.

Compound nouns in spoken language production by speakers with aphasia compared to neurologically healthy speakers: an exploratory study

Compounds are words that are made up of at least two other words (lexemes), featuring lexical and syntactic characteristics and thus particularly interesting for the study of language processing. Most studies of compounds and language processing have been based on data from experimental single word production and comprehension tasks. To enhance the ecological validity of morphological processing research, data from other contexts, such as discourse production, need to be considered. This study investigates the production of nominal compounds in semi-spontaneous spoken texts by a group of speakers with fluent types of aphasia compared to a group of neurologically healthy speakers. The speakers with aphasia produce significantly fewer nominal compound types in their texts than the non-aphasic speakers, and the compounds they produce exhibit fewer different types of semantic relations than the compounds produced by the non-aphasic speakers. The results are discussed in relation to theories of language processing.

Treatment of semantic verb classes in aphasia: acquisition and generalization effects

Verb retrieval difficulties are common in aphasia; however, few successful treatments have been documented (e.g. Conroy, P., Sage, K., & Lambon Ralph, M. A. (2006) . Towards theory-driven therapies for aphasic verb impairments: A review of current theory and practice. Aphasiology, 20, 1159-1185). This study investigated the efficacy of a novel verb retrieval treatment in two individuals with aphasia who experience verb retrieval difficulty. It involved training verb classes with large (e.g. cut verbs) and limited (e.g. contact verbs) sets of semantic features. Based on action representation theories, semantically based training of cut verbs was predicted to generalize to retrieval of untrained cut and contact verbs. One participant improved on trained verbs whereas the other participant did not. Neither participant demonstrated within nor across-class generalization to untrained verbs. However, both participants significantly improved in verb naming as measured by An Object and Action Naming Battery, and their predominant error pattern changed from noun to verb substitutions. Therefore, both participants improved in overall verb retrieval strategies despite limited success with verbs trained in this treatment. Implications for the design of future treatments are discussed.

Spoken verb processing in Spanish: An analysis using a new online resource

Verbs are one of the basic building blocks of grammar, yet few studies have examined the grammatical, morphological, and phonological factors contributing to lexical access and production of Spanish verb inflection. This report describes an online data set that incorporates psycholinguistic dimensions for 50 of the most common early-acquired Spanish verbs. Using this data set, predictors of response time (RT) from stimulus onset and mean differences at offset are examined. Native Spanish speakers, randomly assigned to one of two tasks, listened to prerecorded verbs and either repeated the verb (single word shadowing) or produced its corresponding pronoun. Factors such as stimulus duration, number of syllables, syllable stress position, and specific levels of initial phoneme facilitated both shadowing of a verb and production of its pronoun. Higher frequency verbs facilitated faster verb repetition, whereas verbs with alternative pronouns increased RT to pronoun production. Mean differences at offset (stimulus duration is removed) indicated that listeners begin speaking earlier when the verb is longer and multisyllabic compared to shorter, monosyllabic words. These results highlight the association between psycholinguistic factors and RT measures of verb processing, in particular, features unique to languages like Spanish, such as alternative pronoun and tense.

Group effects of instrumentality and name relation on action naming in bilingual anomic aphasia

Verb production in sentences was investigated in two groups of late bilingual Greek-English speakers: individuals with anomic aphasia and a control group. Verb retrieval in sentences was significantly impaired in both languages for the individuals with anomic aphasia. Additional results revealed no effect of instrumentality on action naming in sentences in either language. However, there was a negative effect of verb-noun name relation on instrumental verb production in English only. Results confirm intact verb lemma retrieval for this group of bilingual individuals with anomic aphasia, but a breakdown at the level of accessing the phonological or lexical form.

Saying the right word at the right time: Syntagmatic and paradigmatic interference in sentence production

Retrieving a word in a sentence requires speakers to overcome syntagmatic, as well as paradigmatic interference. When accessing cat in "The cat chased the string," not only are similar competitors such as dog and cap activated, but also other words in the planned sentence, such as chase and string. We hypothesize that both types of interference impact the same stage of lexical access, and review connectionist models of production that use an error-driven learning algorithm to overcome that interference. This learning algorithm creates a mechanism that limits syntagmatic interference, the syntactic "traffic cop," a configuration of excitatory and inhibitory connections from syntactic-sequential states to lexical units. We relate the models to word and sentence production data, from both normal and aphasic speakers.