logo
ResearchBunny Logo
A dependency distance approach to the syntactic complexity variation in the connected speech of Alzheimer's disease

Linguistics and Languages

A dependency distance approach to the syntactic complexity variation in the connected speech of Alzheimer's disease

N. Gao and Q. He

This groundbreaking study by Nan Gao and Qingshun He explores how Alzheimer's disease affects syntactic complexity, revealing that patients struggle with complex structures and prefer simpler syntactic forms. With working memory deficits in focus, the research highlights critical insights into the cognitive challenges faced by individuals with AD.

00:00
00:00
Playback language: English
Introduction
Alzheimer's Disease (AD) is a neurodegenerative condition causing memory and cognitive deficits, impacting language use. While previous research consistently demonstrates lexical and discourse differences in AD patients' connected speech (shorter lexical output, less lexical richness, reduced semantic units, inappropriate cohesive devices), the nature of syntactic impairments remains debated. Some studies suggest syntactic abilities are preserved, characterized as "grammatic and coherent," while others highlight changes in syntactic complexity (shorter utterances, reduction in complex structures, difficulties with specific clause types). However, there's a lack of consensus on the distinctive marker of syntactic complexity in AD, with studies varying in their approaches (sentence length, syntactic structures, rating scales). Working memory plays a crucial role in syntactic processing; its deficits in AD may negatively affect complex syntax production. This study aims to explore syntactic complexity variation in AD by considering working memory load and fine-grained syntactic features from a dependency syntax perspective, using dependency distance as a holistic marker of syntactic complexity and working memory load. The research questions are: 1) Can dependency distance and direction detect syntactic complexity variation in AD? 2) What are the reasons for this variation in terms of fine-grained syntactic features?
Literature Review
Existing research on the syntactic complexity of Alzheimer's disease patients' speech is inconsistent. Some studies find evidence for preserved syntax, while others report syntactic simplification, shorter utterances, and reduced complex structures, such as embedding and subordinate clauses. There's no agreement on the best measure of syntactic complexity, with studies employing different metrics, such as mean sentence length, which yield conflicting results. The impact of working memory deficits is also implicated, with some studies suggesting that these deficits directly affect the production of complex syntax in AD patients. This research aims to address these inconsistencies and clarify the nature of syntactic impairments by focusing on dependency distance, which offers a more holistic and comprehensive measure of syntactic complexity and its relation to working memory.
Methodology
This study utilized the DementiaBank clinical dataset, creating comparable treebanks of 65 individuals with probable AD and a demographically matched healthy control (HC) group. Only baseline recordings were used, and transcripts were cleaned to remove symbols, interjections, and extraneous information. Utterances were segmented based on conversational boundaries and intonation. Dependency distance, a measure of the linear distance between syntactically linked words, was used as the primary metric of syntactic complexity. Mean dependency distance (MDD) was calculated for each transcript. Dependency direction (head-initial vs. head-final) was also analyzed. Fine-grained syntactic features included percentage of adjacent dependency (1dd%), mean sentence length (MSL), and MDDs for major grammatical relations (subject, object, attributive, adverbial). The spaCy (3.7.0) natural language parsing tool was used for dependency annotation. Two linguists reviewed the annotations, achieving a 95% agreement rate. Statistical analyses included independent t-tests, linear regression, and the Benjamini-Hochberg procedure for multiple comparisons.
Key Findings
The study found that AD patients had significantly lower MDDs and a higher percentage of head-final dependencies than the HC group. Although both groups exhibited the dependency distance minimization (DDM) tendency (a preference for adjacent dependencies), the AD group showed a stronger preference for simpler structures. AD patients produced significantly shorter sentences (MSL). Regression analyses revealed a negative correlation between 1dd% and MDD in both groups, but with different intercepts, indicating that even without the influence of adjacent dependencies, AD patients have lower MDD. The correlation between MDD and MSL was positive in both groups, but the slope was steeper for the AD group, implying that AD patients experience greater difficulty processing syntactic complexity as sentence length increases. Analysis of major grammatical relations showed lower MDDs in AD patients for several dependency types, particularly nsubj and advcl, indicating difficulty in extending these structures. AD patients showed significantly fewer prepositional objects (pobj) in head-initial Object relations and more advmod in head-final Adverbial relations. The increased use of sentence-initial adjuncts (advmod) and inverted constructions (nsubj in head-initial Subject relations) in AD speech may be influenced by non-syntactic factors, such as the need for clear contextual orientation and reduced attentional focus.
Discussion
The significantly lower MDD in AD patients indicates impaired working memory, leading to a preference for syntactically simpler structures that minimize cognitive load. This contrasts with some research showing increased low-frequency syntactic structures in a specific type of primary progressive aphasia, highlighting the importance of considering different measures of syntactic complexity. The increased head-final dependency preference in AD might be due to a "short-before-long" preference, prioritizing shorter, more accessible phrases. The similar DDM tendency in both groups but with different intercepts and slopes implies that while AD patients minimize dependency distance, their capacity for complex structures is significantly reduced. The shorter MSL and steeper slope in the MDD-MSL correlation for AD patients confirm their greater difficulty processing longer and more complex sentences. The observed variations in specific dependency types suggest that while fundamental syntactic abilities are preserved, the capacity to extend structures and handle higher working memory demands is impaired. The findings support a hierarchical decline model of syntactic impairments in AD, where more complex structures are more vulnerable.
Conclusion
This study demonstrates that dependency distance provides a valuable framework for investigating syntactic complexity variations in AD, revealing impaired working memory and a preference for simpler sentence structures. The findings support a hierarchical decline model of syntactic impairment, where higher working memory demands lead to greater vulnerability. Future research should consider other metrics and explore non-linguistic factors influencing syntactic choices in AD.
Limitations
The study used a specific dataset (DementiaBank) and focused on English speakers. Generalizability to other languages and populations might be limited. The interpretation of some findings, particularly the increased use of sentence-initial adjuncts and inverted constructions, involves speculative connections to non-syntactic factors that need further investigation. While the study employed sophisticated methods, the interpretations are based on statistical correlations and do not definitively establish causal relationships. Future studies should explore these aspects more deeply.
Listen, Learn & Level Up
Over 10,000 hours of research content in 25+ fields, available in 12+ languages.
No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.
listen to research audio papers with researchbunny