Linguistic Markers of Subtle Cognitive Impairment in Connected Speech: A Systematic Review
Abstract
Purpose:
Method:
Results:
Conclusions:
Investigating mNCDs With Markers in Connected Speech
Objectives of This Review
Terminology Used on This Review
Organization of the Review
Method
Eligibility Criteria
Population
Connected Speech Tasks and Neuropsychological Tests
Search Strategy
Selection Process
Data Extraction Process
Category | Data extracted |
---|---|
Population | Number, age, language, for both case and control populations Pathology for case population |
Connected speech task | Name of the connected speech task Type of elicited discourse according to the authors Type of elicited discourse according to our categorization |
Tests | Name of the neuropsychological tests used Cognitive domains assessed according to the tests' authors |
Methods | Name of the linguistic features extracted Feature extraction mode (manual vs. automatic) Statistical tests used for discriminating case population from control and their p value Machine learning models used |
Synthesis Methods
Feature Synthesis
Linguistic domain Category | Linguistic features | Definition | Examples |
---|---|---|---|
Phonetics-prosody Disfluencies | Silent pauses | Absence of intensity and F0, any interval where the amplitude is undistinguishable from that of the background noise above 200 ms (see for instance, Duez, 1982) | A: and this woman drives her kids to # school |
Filled pauses | Any type of filler particle that delays speech production: “A phonetic exponent which is segmentally structured, semantically empty, syntactically unconstrained, and does not show an interjectional function” (Belz, 2023). | A: and this woman drives her kids to uh school | |
Breaks | Self-interruption from the speaker involving an incomplete delivery of a phonetic, syntactic, or morphological unit as a result of its abandonment Can lead to the resumption of a new unit as part of self-repair At the lexical level, includes deletions, substitutions, insertions, and articulation errors linked to missed phonological targets (Pallaud et al., 2019) | With self-repair: A: and this woman drives her daugh- her kids to school Without self-repair: A: and this woman drives her kids to # | |
Repetitions | Duplication(s) of a phoneme, syllable, word, or phrase (Fox Tree, 1995) | A: and this woman drives her k- kids school | |
Speed of speech | Speech rate | Measure for speed and density per time unit, as the number of syllables produced per minute of speech (De Jong & Wempe, 2009) For studies relying on a fully automated data extraction, a robust definition can be found in He et al. (2023): “number of nuclei / (total nuclei duration + total internuclei duration).” | |
Articulation rate | Measure for speed and density per time unit, as the number of syllables produced per minute of speech, without silent pause or laughter time (Miller et al., 1984) | ||
Length of speech units | IPU duration | Interpausal units are speech segments separated by a 250-ms silent pause, with a minimum duration of 300 ms (Bigi & Priego-Valverde, 2022) | A: and this woman drives her kids to school (2390 ms) # (250 ms) I didn't know she had kids (1770 ms) |
Turn duration | Conversational turn-taking system. A speech turn (turn constructional unit) is a stretch of speech by one speaker during which other participants assume the role of listeners, until a point of “projected completion” (Sacks et al., 1974), called a transition relevance place (TRP), which can include a silent pause. A turn shift can occur (but does not have to) at a TRP. If no turn shift occurs at the TRP, the turn continues. | A: and this woman drives her kids to school (2390 ms) # B: Did you know she had kids? | |
Speech duration | Total speech time duration. Whether this includes silent pauses and other vocal phenomena such as laughter has to be indicated. | A: and this woman drives her kids to school # I didn't know she had kids (4410 ms) | |
Lexicon-semantics Words | Lexical items | Refers to content (i.e., concepts, actions, beings or objects with clear mental representation) and open-class words (i.e., nouns, [non-auxiliary] verbs, adjectives, and [some] adverbs). | A: and this woman drives her kids to school |
Grammatical items | Refers to function and closed-class words No independent meaning on its own or no full-fledged semantic features Can be prepositions, pronouns, conjunctions, auxiliary verbs, or (some) adverbs | A: and this woman drives her kids to school | |
Lexical richness | Also referred as lexical diversity and density, is used to assess the number produced by a speaker, showing the proficiency of a discourse Often measured with a type–token ratio (number of different words divided by the total number of words) or the Shannon's Entropy | ||
Syntax–morphosyntax | Clause | A subject and a predicate (verb + any complement) Sometimes labeled as sentence, utterance, phrase, or proposition | A: and this woman drives her kids to school. |
Clause completion | Refers to the ability for a participant to finish a clause in accordance with grammatical rules | ||
Clause length | Number of lexical and grammatical items in a clause Whether this excludes or includes filler particles has to be specified. | ||
Mean length of utterance (MLU) | In the studies we reviewed, the definition was the following: MLU is calculated by dividing the total number of words by the number of utterances (Galetto et al., 2013; Lowit et al., 2022). | ||
Syntactic errors | Clauses with inconsistent syntax, grammar, or morphology (for instance using a noun in a verb slot or using an incorrect inflection for tense) Dependent on context of use and language variety (e.g., African American Vernacular English, General American, Southern American English) | A: and this woman have driven her kids to school. | |
Item class | Also called word classes or lexical categories POS tagset: syntactic class of the item This list includes nouns, verbs, adjectives, adverbs, pronouns, prepositions, conjunctions, articles, and interjections. | ||
Syntactic complexity | No consensus yet This is what we propose based on the studies we reviewed. Number of items that are additional to the obligatory constituents in a clause Includes subordinators and coordinators Can also be number of complex sentences (made up of several clauses) as opposed to simple sentences (made up of one clause) This has to be made explicit. | ||
Discourse | Informativeness | Notion of relevance (Grice's maxim Be Relevant) Proportion of discourse segments on topic vs. off topic in the case of a thematic task (i.e., description or narrative). Main discourse structure vs. side discourse structure (Van Kuppevelt, 1995) | A: and this woman drives her kids to school # I didn't know she had kids # then she goes to the office |
Coherence | How different segments of discourse are made to relate to one another (Kehler, 2006; Labov, 2001; Wright et al., 2014) | Coherence: A: and this woman drives her kids to school # I didn't know she had kids # then she goes to the office Incoherence: A: and this woman drives her kids to school # I didn't know she had kids # we were only two at the office | |
Cohesion | Consistent use of reference in speech Consistent use of cohesive devices such as anaphora (e.g., through pronouns and noun phrases), connectives, and simultaneity markers Can also include temporal cohesion (McNamara et al., 2010) | A: and this woman drives her kids to school # but she [another woman] doesn't know that | |
Epistemic and pragmatic events | Epistemic events: use of any hedging expression (Schiffrin, 1987) A speaker's lexical reference to their own uncertainty or inability; use of modal auxiliaries; use of hedging discourse markers (see Cuenca & Crible, 2019, for the rhetorical and interpersonal domains in the classification of discourse markers) Pragmatic event: speech overlapping with another participant's turn; turn change (can be forced turn-taking as in interruptions); turn retention; use of conversational feedbacks (also called backchannels) Conversational feedbacks usually consist of brief signals produced by the interlocutor during the main speaker's speech and can be verbal (e.g., yes), vocal (e.g., mhm), and/or gestural (head movements, smiles). They are mandatory to update the shared knowledge (common ground) and promote the alignment between participants, which is necessary for mutual comprehension and success of the interaction (Boudin et al., 2021). | Epistemic events: A: and this woman drives her kids to school # I can't remember her name # then she goes to the office Pragmatic events: A: and this woman drives her kids to school # B: mhm # A: then she goes to the office | |
Theory of mind (ToM) | Items showing the speaker is aware of others' different representations and mental states. Includes items that refer to the acknowledgment/consciousness of shared knowledge between participants, and those that manage it In the reviewed studies, ToM was measured with emotional tone and words semantically related to feelings (Baron-Cohen et al., 1985). | A: and this woman drives her kids to school (i.e., no needs to name the woman because the speaker knows that their co-speaker knows the woman in question) |
Discourse Types
Results

Population Overview


Linguistic Features


Study Quality Assessment


Statistical Significance of Group Comparisons Via Linguistic Features
Summary
Discussion
High Variability in Results
Different Labels for Similar Features
Multiplicity of Methods
Assessing a Feature's Reliability: How Can a Feature Become a Marker?
Discrimination Power
Replicability
Linguistic Characteristics of Discourse Types
Marker Layout in Function of Discourse Type
Exploring Cognition Through Speech Analysis
Hypotheses on Underlying Mechanisms of Discourse Type
Neuropsychological and Neural Correlates of Linguistic Markers
Limitations of This Review and Future Research
Contributions of This Review This review covers different fields of interest. In this section, we state what this review adds to previous literature depending on the reader's background and interests. 1. Interest in analyzing connected speech: • A list of currently examined linguistic markers is available in the Results section, as well as our clustering approach for categorizing linguistic features in the Method section. • A detailed overview of the features and their definitions are proposed in Table 2. • The Exploring Cognition Through Speech Analysis section in the Discussion section presents an opinion on current challenges in the domain of speech analysis. 2. Interest in assessing cognition using a discourse task: • The Method section draws up a list of the various pathologies and cognitive-behavioral syndromes we included. • The variability in feature outcomes is revealed in the Results section. • The Discussion section provides an overview of the association between linguistic markers and cognitive impairment, as this will provide guidance when choosing features and discourse tasks. |
Conclusions
Author Contributions
Data Availability Statement
Acknowledgments
References
Information & Authors
Information
Published In
Copyright

History
- Received: Apr 29, 2024
- Revised: Jun 24, 2024
- Accepted: Aug 26, 2024
- Published online: Nov 15, 2024
- Published in issue: Dec 9, 2024
Authors
Metrics & Citations
Metrics
Article Metrics
View all metricsCitations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.
For more information or tips please see 'Downloading to a citation manager' in the Help menu.
View Options
View options
Sign In Options
ASHA member? If so, log in with your ASHA website credentials for full access.
Member Login