Elsevier

Brain and Language

Volume 162, November 2016, Pages 19-28
Brain and Language

How language flows when movements don’t: An automated analysis of spontaneous discourse in Parkinson’s disease

https://doi.org/10.1016/j.bandl.2016.07.008Get rights and content

Highlights

  • We assessed spontaneous discourse in Parkinson’s disease (PD) with automatized tools.

  • Compared to controls, patients used fewer action concepts and more subordinators.

  • Analysis of grammar choices allowed classifying patients and controls above chance.

  • The incidence of word repetitions predicted the patients’ level of motor impairment.

  • Naturalistic discourse features may index the integrity of specific neural networks.

Abstract

To assess the impact of Parkinson’s disease (PD) on spontaneous discourse, we conducted computerized analyses of brief monologues produced by 51 patients and 50 controls. We explored differences in semantic fields (via latent semantic analysis), grammatical choices (using part-of-speech tagging), and word-level repetitions (with graph embedding tools). Although overall output was quantitatively similar between groups, patients relied less heavily on action-related concepts and used more subordinate structures. Also, a classification tool operating on grammatical patterns identified monologues as pertaining to patients or controls with 75% accuracy. Finally, while the incidence of dysfluent word repetitions was similar between groups, it allowed inferring the patients’ level of motor impairment with 77% accuracy. Our results highlight the relevance of studying naturalistic discourse features to tap the integrity of neural (and, particularly, motor) networks, beyond the possibilities of standard token-level instruments.

Introduction

Affecting more than 1% of individuals above age 60, Parkinson’s disease (PD) is the second most prevalent neurodegenerative disease worldwide (de Rijk et al., 2000, Samii et al., 2004). It is characterized by progressive basal ganglia degeneration and dopamine depletion, which disrupts corticostriatal circuits involved in motor function and multiple high-level cognitive domains (Fearnley and Lees, 1991, Mattay et al., 2002, McKinlay et al., 2010, Muslimovic et al., 2005, Rodriguez-Oroz et al., 2009). Thus, the impact of PD goes well beyond the presence of movement disorders (Mattay et al., 2002, Svenningsson et al., 2012).

This is particularly evident in linguistic performance. Indeed, articulatory disorders in PD (Goberman and Blomgren, 2003, Goberman et al., 2010) are often accompanied by impairments in grammar (Bocanegra et al., 2015, Hochstadt et al., 2006, Lieberman et al., 1992), pragmatics (Holtgraves and McNamara, 2010, Monetta and Pell, 2007), verbal fluency (Raskin, Sliwinski, & Borod, 1992), and action-verb semantics (García and Ibáñez, 2014a, Bak, 2013, Bocanegra et al., 2015, Cardona et al., 2013). While these findings are quite revealing about the physiopathology of PD, it is hard to assess their impact in real life, since they stem from highly artificial tasks in which disconnected stimuli are processed in random or arbitrary succession. Also, the active demands of such often exhausting tasks render them limited as tools for prospective diagnosis criteria.

Our aim was to address both issues using automated tools. Specifically, we examined whether PD patients exhibit distinguishing discourse-level features as they produce naturally unfolding texts. This process, termed logogenesis, is based on the accumulation of interrelated lexico-grammatical selections which create distributed patterns above the word and sentence levels (Halliday & Matthiessen, 2004). Insights into this dynamic process could afford a more ecological understanding of how this disease impacts verbal communication.

Section snippets

Background and hypotheses

Discourse production involves construing supra-sentential textual relations and deploying diverse communicative strategies (Halliday & Matthiessen, 2004). Emergent distributed patterns can be detected by considering semantic fields, lexicogrammatical choices, and relations between adjacent or neighboring words (Bedi et al., 2014, Bedi et al., 2015, Mota et al., 2014, Mota et al., 2012). Analyses of these and other text-level variables have revealed population-specific patterns in various

Participants

The study included 51 non-demented PD patients (25 female) and 50 healthy controls (25 female) from the PC-GITA database (Orozco-Arroyave, Arias-Londoño, Vargas-Bonilla, González-Rátiva, & Nöth, 2014). All participants were monolingual Spanish speakers from Colombia. The patients had a mean age of 61.45 (SD = 9.77), with 10.71 (SD = 4.2) years of education. Mean values for these variables in the control sample were 60.9 (SD = 9.47) and 10.98 (SD = 4.54), respectively. Both groups were matched for age [t

Quantitative output

The monologues of each group were compared for basic linguistic attributes, namely, total word count, and number of content words, nouns, verbs, action verbs, non-action verbs, and type/token ratio, indicating that quantitative output was similar for both groups (Table 2).

Semantic fields

LSA measures revealed significant between-group differences in the second [Wilcoxon: Z = 2.12, p = 0.03; t-test: t (99) = 2.7, p = 0.008] and third [Wilcoxon: Z = 1.94, p = 0.05; t-test: t (99) = 2.5, p = 0.01] semantic components. As shown

Discourse patterns at the semantic level

Despite moderate predictive success, analysis of semantic fields robustly discriminated between groups. Action-related domains figured less strongly in patients than controls, whereas non-action concepts showed the opposite pattern. Greater impairments in action- than non-action language have been repeatedly observed in PD through controlled, atomistic tasks, including picture naming (Rodriguez-Ferreiro, Menendez, Ribacoba, & Cuetos, 2009), related-word production (Peran et al., 2003), and

Conclusion

The study of discourse phenomena in spontaneous speech offers an ecological window into language mechanisms. Here, using automatized tools, we found evidence for distinctive patterns in PD patients, as manifested in semantic fields, grammatical features, and word-level repetitions. Thus, the basal ganglia seem crucial for the deployment of multidimensional textual relationships in naturalistic discourse production. These insights could hardly have been reached via decontextualized atomistic

Conflict of interest

None to declare.

Acknowledgments

This work was partially supported by grants from CONICET, UBA, CONICYT/FONDECYT Regular (1130920), COLCIENCIAS 1115-569-33858, COLCIENCIAS 1115-545-31374 (contract: 392), FONCyT-PICT 2012-0412, FONCyT-PICT 2012-1309, FONCyT-PICT 2013-1794, FONDAP 15150012, and the INECO Foundation. The authors would like to thank Fundalianza Parkinson Colombia for its generous help in the data collection process. DFS is sponsored by the Microsoft Faculty Fellowship 2014.

References (86)

  • E. Herrera et al.

    Verbal fluency in Parkinson’s disease patients on/off dopamine medication

    Neuropsychologia

    (2012)
  • J. Hochstadt et al.

    The roles of sequencing and verbal working memory in sentence comprehension deficits in Parkinson’s disease

    Brain and Language

    (2006)
  • A. Ibáñez et al.

    Motor-language coupling: Direct evidence from early Parkinson’s disease and intracranial cortical recordings

    Cortex

    (2013)
  • J. Illes

    Neurolinguistic features of spontaneous language production dissociate three forms of neurodegenerative disease: Alzheimer’s, Huntington’s, and Parkinson’s

    Brain and Language

    (1989)
  • J. Illes et al.

    Language production in Parkinson’s disease: Acoustic and linguistic considerations

    Brain and Language

    (1988)
  • S.A. Kotz et al.

    Non-motor basal ganglia functions: A review and proposal for a model of sensory predictability in auditory language perception

    Cortex

    (2009)
  • Y. Lebrun

    Cluttering after brain damage

    Journal of Fluency Disorders

    (1996)
  • C. Lee et al.

    Attentional resource and processing speed limitations during sentence processing in Parkinson’s disease

    Brain and Language

    (2003)
  • P. Lieberman et al.

    Speech production, syntax comprehension, and cognitive deficits in Parkinson’s disease

    Brain and Language

    (1992)
  • P. McNamara et al.

    Pragmatic communication skills in patients with Parkinson’s disease

    Brain and Language

    (2003)
  • P. McNamara et al.

    Speech monitoring skills in Alzheimer’s disease, Parkinson’s disease, and normal aging

    Brain and Language

    (1992)
  • L. Monetta et al.

    Effects of verbal working memory deficits on metaphor comprehension in patients with Parkinson’s disease

    Brain and Language

    (2007)
  • L.L. Murray et al.

    Productive syntax abilities in Huntington’s and Parkinson’s diseases

    Brain and Cognition

    (2001)
  • S.A. Raskin et al.

    Clustering strategies on tasks of verbal fluency in Parkinson’s disease

    Neuropsychologia

    (1992)
  • J. Rispens et al.

    Negation in agrammatism: A cross-linguistic comparison

    Journal of Neurolinguistics

    (2001)
  • J. Rodriguez-Ferreiro et al.

    Action naming is impaired in Parkinson disease patients

    Neuropsychologia

    (2009)
  • M.C. Rodriguez-Oroz et al.

    Initial clinical manifestations of Parkinson’s disease: Features and pathophysiological mechanisms

    The Lancet Neurology

    (2009)
  • A. Samii et al.

    Parkinson’s disease

    The Lancet

    (2004)
  • P. Svenningsson et al.

    Cognitive impairment in patients with Parkinson’s disease: Diagnosis, biomarkers, and treatment

    The Lancet Neurology

    (2012)
  • C. Thiriez et al.

    Improvement in developmental stuttering following deep brain stimulation for Parkinson’s disease

    Parkinsonism & Related Disorders

    (2013)
  • B. Tomasino et al.

    To move or not to move: Subthalamic deep brain stimulation effects on implicit motor simulation

    Brain Research

    (2014)
  • M.T. Ullman

    Contributions of memory circuits to language: The declarative/procedural model

    Cognition

    (2004)
  • M.T. Ullman

    Is Broca’s area part of a basal ganglia thalamocortical circuit?

    Cortex

    (2006)
  • S. Zanini et al.

    Grammar improvement following deep brain stimulation of the subthalamic and the pedunculopontine nuclei in advanced Parkinson’s disease: A pilot study

    Parkinsonism & Related Disorders

    (2009)
  • H. Ackermann et al.

    Palilalia as a symptom of levodopa induced hyperkinesia in Parkinson’s disease

    Journal of Neurology, Neurosurgery and Psychiatry

    (1989)
  • A.J. Angwin et al.

    The influence of dopamine on semantic activation in Parkinson’s disease: Evidence from a multipriming task

    Neuropsychology

    (2006)
  • A. Ardila et al.

    How localized are language brain areas? A review of Brodmann areas involvement in oral language

    Archives of Clinical Neuropsychology.

    (2015)
  • S. Ash et al.

    Trying to tell a tale: Discourse impairments in progressive aphasia and frontotemporal dementia

    Neurology

    (2006)
  • T.H. Bak

    The neuroscience of action semantics in neurodegenerative brain diseases

    Current Opinion in Neurology

    (2013)
  • L.W. Barsalou

    Perceptual symbol systems

    Behavioral and Brain Sciences

    (1999)
  • L. Bebout

    Processing of negative morphemes in aphasia: An example of the complexities of the closed class/open class concept

    Clinical Linguistics & Phonetics

    (1993)
  • G. Bedi et al.

    Automated analysis of free speech predicts psychosis onset in high-risk youths

    Npj Schizophrenia

    (2015)
  • G. Bedi et al.

    A window into the intoxicated mind? Speech as an index of psychoactive drug effects

    Neuropsychopharmacology

    (2014)
  • Cited by (75)

    • Depression assessment in people with Parkinson's disease: The combination of acoustic features and natural language processing

      2022, Speech Communication
      Citation Excerpt :

      A review about topics related to depression and suicide risk assessment using speech analysis is performed in Cummins et al. (2015a). On the one hand, there are some studies that consider natural language analysis tools to evaluate PD patients (García et al., 2016; Pérez-Toro et al., 2019). In García et al. (2016), the authors extracted different components from transcriptions of spontaneous speech recordings using Latent Semantic Analysis (LSA), part-of-speech tagging and word-level repetitions via graph embedding tools.

    View all citing articles on Scopus
    View full text