Intergenerational choral singing to improve communication outcomes in Parkinson’s disease: Development of a theoretical framework and an integrated measurement tool

Abstract Purpose: This study presented an initial step towards developing the evidence base for intergenerational choral singing as a communication-focussed rehabilitative approach for Parkinson’s disease (PD). Method: A theoretical framework was established to conceptualise the rehabilitative effect of intergenerational choral singing on four domains of communication impairments – motor drive, timing mechanism, sensorimotor integration, higher-level cognitive and affective functions – as well as activity/participation, and quality of life. A computer-assisted multidimensional acoustic analysis was developed to objectively assess the targeted domains of communication impairments. Voice Handicap Index and the World Health Organization’s Quality of Life assessment-abbreviated version were used to obtain patient-reported outcomes at the activity/participation and quality of life levels. As a proof of concept, a single subject with PD was recruited to participate in 9 weekly 1-h intergenerational choir rehearsals. The subject was assessed before, 1 week post, and 8 weeks post-choir. Result: Notable trends of improvement were observed in multiple domains of communication impairments at 1 week post-choir. Some improvements were maintained at 8 weeks post-choir. Patient-reported outcomes exhibited limited pre-post changes. Conclusion: This study provided the theoretical groundwork and an empirical measurement tool for future validation of intergenerational choral singing as a novel rehabilitation for PD.


Introduction
As the second most common neurodegenerative condition, Parkinson's disease (PD) occurs in 1% of the population aged over 60 and its prevalence increases with age (Reeve, Simcox, & Turnbull, 2014).The primary neuropathological underpinning of PD is the degeneration of dopaminergic neurons of substantia nigra, leading to a dysfunction of the basal ganglia (BG) pathways.Although clinically categorised as a movement disorder, PD leads to both motor (e.g.bradykinesia, hypokinesia, rigidity, resting tremor, postural instability, gait disturbance, dystonia, dysarthria, dysphagia) and non-motor (e.g.cognitive and sensory disturbances, depression, anxiety, sleep disorder, constipation) symptoms (Sveinbjornsdottir, 2016).Voice and speech impairments are among the earliest and most prevalent symptoms of PD, affecting close to 90% of patients over the disease course (Dashtipour, Tafreshi, Lee, to reduced activation and poor control of skeletal muscles of all physiological speech subsystems (i.e.phonatory, respiratory, articulatory, resonatory) (Duffy, 2013;Tjaden, 2008).Such a neuromotor deficit has been associated with reduced vocal loudness and intonation, abnormal voice quality, and imprecise articulation, which are collectively referred to as hypokinetic dysarthria -the most prevalent form of motor speech disorder in PD (Darley, Aronson, & Brown, 1969a;Darley et al., 1969b).Furthermore, the internal timing system, which relies largely on the functioning of BG, specifically putamen (Grahn, 2009;Harrington, Haaland, & Hermanowicz, 1998), has been shown to be disrupted in individuals with PD, resulting in difficulty initiating, sequencing, and coordinating movements (Ackermann, Konczak, & Hertrich, 1997).Such timing deficits have been suggested to underlie the tempo, rhythm, and coordination abnormalities in PD (Ackermann et al., 1997;Puyjarinet et al., 2019;Skodda & Schlegel, 2008).
In addition to motor deficits, the impairment of sensory processing (e.g.auditory, somatosensory) and sensorimotor integration and adaptation are well documented in PD (Abbruzzese & Berardelli, 2003;Brown, Schneider, & Lidsky, 1997;Hammer & Barlow, 2010;Lewis & Byblow, 2002;Mollaei, Shiller, & Gracco, 2013).These deficits lead to abnormal perception of sensory inputs and/or impaired integration of sensory inputs with motor programs, both of which can downregulate the outputs of the speech production system, globally impacting voice and speech performance.Moreover, the impairment of higher-level cognitive, affective, and psychomotor functions, commonly reported in individuals with PD, could exert modulatory influences on speech production, further disrupting voice and speech features such as intonation, tempo, and rhythm (Flint, Black, Campbell-Taylor, Gailey, & Levinton, 1992;Middleton & Strick, 2000;Ramig, Pawlas, & Countryman, 1995).
Taken together, a variety of motor and non-motor factors can contribute to voice and speech symptoms in PD.Besides these putative contributing factors, the neural substrates of some voice and speech abnormalities in PD are yet unclear.For example, vocal tremor -a common voice abnormality in PDappears to have a separate but uncertain neural origin from the other voice and speech symptoms.It has been argued to reflect abnormal involuntary neural oscillatory activities in the brain, possibly involving both the BG-thalamo-cortical and cerebello-thalamo-cortical circuits (Helmich, Hallett, Deuschl, Toni, & Bloem, 2012).

Current standard treatments and limitations
Levodopa-based pharmacotherapy is the mainstay of treatment targeting general motor symptoms in PD (Pezzoli & Zini, 2010).Additionally, deep brain stimulation of BG structures such as subthalamic nucleus or globus pallidus has emerged as a surgical treatment alternative, which can be integrated with the traditional pharmacotherapy to improve the efficacy in motor symptom management (Benabid, 2003;Deuschl et al., 2006;Volkmann, 2004).Despite the demonstrated rehabilitative effect on general motor functions, there is a lack of solid evidence for the effectiveness of these pharmacological and surgical treatments in improving voice and speech impairments.Behavioural therapies therefore remain the gold standard for managing communication impairments in PD.
The Lee Silverman Voice Treatment (LSVT/ LOUD) is the most widely used behavioural therapy for managing voice/speech deficits in PD (Ramig et al., 1995).LSVT/LOUD utilises intensive, higheffort exercises emphasising vocal loudness and selfperception of vocal loudness, with the aim to globally increase neural drive to the speech production system through recalibrating the underlying sensorimotor processes, thereby achieving a system-wide treatment effect (Fox et al., 2006;Ramig et al., 1995).LSVT/ LOUD has proven to upscale the magnitude of motor outputs across multiple physiological speech subsystems, improving vocal loudness, intonation, and articulatory precision, effectively alleviating hypokinetic dysarthria-related symptoms (Ramig, Countryman, O'Brien, Hoehn, & Thompson, 1996;Ramig et al., 2001;Sapir, Spielman, Ramig, Story, & Fox, 2007).Another behavioural therapy targeting vocal loudness is the "SPEAK OUT! and LOUD Crowd" program developed by the Parkinson Voice Project.This therapy consists of (1) an individualised program (i.e.SPEAK OUT!) aimed at enhancing the cognitive focus on increasing attention to speech production to compensate for the control deficits related to BG dysfunction, and (2) a maintenance group program (i.e.LOUD Crowd) for retaining the therapeutic gains from the individualised program (Behrman, Cody, Elandary, Flom, & Chitnis, 2020;Levitt, 2014;Levitt, Chitnis, & Walker-Batson, 2015).Yet, because LSVT/LOUD and SPEAK OUT! and LOUD Crowd both have a targeted focus on increasing the magnitude of motor drive, the timing and rhythm abnormalities in PD remain largely unattended.These abnormalities, as noted above, can be attributed to defective internal timing, sensorimotor disintegration, and higher-order brain dysfunctions.Alternative or supplementary treatments targeting these areas of impairments are needed.

Music as a rehabilitative modality
Music has gained increasing interest as a low-cost, safe, and engaging rehabilitative modality, demonstrating a variety of therapeutic benefits to motor, sensory, cognitive, and affective dysfunctions (Barnish & Barran, 2020;Monroe, Halaki, Kumfor, & Ballard, 2020;Wan, R€ uber, Hohmann, & Schlaug, 2010).Among the variety of music elements, rhythm is of particular interest from a neurorehabilitation standpoint, owing to its central role in neural entrainment.Here entrainment is defined as a temporal locking process, wherein the frequency of one system entrains the frequency of another system (Thaut, 2015).For neural systems, the oscillations in the cerebral cortex can be entrained by the inherent rhythms in music, providing a means of regulating cortical activities.Building upon the concept of rhythmic entrainment, Thaut developed neurologic music therapy, utilising the rhythm of music to shape auditory cortex oscillations and link them with motor functions through entrainment of the auditory-motor circuit (Thaut, McIntosh, & Hoemberg, 2014;Thaut & Hoemberg, 2014).A desirable feature of rhythmic entrainment as a rehabilitative mechanism is that it works robustly in a wide range of neurological conditions.Because of the rich rhythmic organisation of music and the robustness of the rhythmic entrainment mechanism, music is a natural rehabilitative modality, demonstrating strong promise for improving general motor functions such as gait in PD (Ashoori, Eagleman, & Jankovic, 2015).
Compared with the vast body of work on the rehabilitative effect of music on general motor functions, there is less research evidence on the relation of music to communicative function.This is somewhat surprising given the commonalities between music and speech in many neurological, physical, and acoustic attributes.Most existing studies relating music to communication disorders are narrowly focussed on a sparse set of subjective or objective features reflecting specific aspects of communication breakdown and/or the overall functional communication outcomes (Barnish & Barran, 2020;Di Benedetto et al., 2009;Elefant, Baker, Lotan, Lagesen, & Skeie, 2012;Evans, Canavan, Foy, Langford, & Proctor, 2012;Stegem€ oller, Hibbing, Radig, & Wingate, 2017;Tamplin, Morris, Marigliani, Baker, & Vogel, 2019;Tanner, Rammage, & Liu, 2016;Yinger & Lapointe, 2012).Common features assessed in these studies include the mean, range, and variability of vocal fundamental frequency and intensity, as derived acoustically, maximum inspiration/expiration pressure and lung volumes, as measured by respiratory instruments such as pressure metre and plethymography, orofacial muscle activities, as measured by electromyography (EMG), as well as voice/speech quality ratings, as evaluated by auditory-perceptual means.These features, focussing largely on the hypokinetic dysarthria-related voice/ speech deficits, provide limited insights into the other important aspects of communication such as timing, rhythm, and coordination as well as the global functioning of the sensorimotor and higher-level brain networks.An efficacious and quantifiable measurement tool is yet lacking to comprehensively evaluate the rehabilitative effect of music on all aspects of communication impairments in PD.
The current literature presents a clear gap between the strong rehabilitative potential of music and the scant and narrowly scoped theoretical and empirical evidence on the relation of music to communication disorders.This gap highlights two areas of need: (1) a theoretical framework to link music with the neurological and physiological underpinnings of communication impairments and the resulting restrictions in activity, participation, and quality of life, in order to guide the design and implementation of music-based rehabilitation protocols for communication disorders, and (2) an integrative measurement tool for evaluating the rehabilitative effect of music on the targeted areas of communication function, activity, participation, and quality of life.

Purpose of the study
This study was aimed at presenting a novel perspective on intergenerational choral singing as a rehabilitative approach to improving communication impairments, activity limitation/participation restriction, and quality of life in individuals with PD.As an active group-based performing art, intergenerational choral singing embraces all potential therapeutic benefits of traditional music therapies and provides additional psychosocial benefits pertaining to engagement, interaction, joint action, and peer models (Barnish & Barran, 2020).In the following sections, we first conceptualise a theoretical framework to link the potential therapeutic benefits of intergenerational choral singing with different areas of communicative function, activity/participation, and quality of life in individuals with PD, integrating the collective body of neurological and behavioural evidence in the extant literature with the International Classification of Function, Disability, and Health (ICF) model (World Health Organization, 2001).We then introduce (1) a novel computer-assisted multidimensional acoustic measurement tool to objectively evaluate the effects of intergenerational choral singing on all targeted aspects of communication impairments as outlined in the theoretical framework, and (2) two standard selfassessments to obtain patient-reported outcomes at the activity/participation and quality of life levels.

Theoretical framework
To conceptualise the rehabilitative effect of intergenerational choral singing on PD-related communication impairments, we focussed on four primary domains: (1) motor drive, (2) timing mechanism, (3) sensorimotor integration, and (4) higher-level cognitive and affective functions.An overview of this theoretical framework is provided in Figure 1.Specifically, from the physiological perspective, singing is a vocal motor behaviour that shares the same musculature as speaking but demands a greater respiratory-laryngeal effort.In this view, singing could provide similar therapeutic benefits as loudness-focussed voice therapies such as LSVT/LOUD and SPEAK OUT! and LOUD Crowd.Choral singing can further augment such therapeutic benefits due to the Lombard effect (Stathopoulos et al., 2014;Tonkinson, 1994).Lombard effect refers to the spontaneous increase of vocal effort as ambient noise increases to enhance the audibility of voice (Lane & Tranel, 1971).Through high-effort vocal exercises pertaining to the Lombard effect, choral singing could conceivably recalibrate the speech production system to increase the magnitude of neural drive to all physiological subsystems, thereby improving hypokinetic dysarthria-related symptoms.
Moreover, the immersive sound environment created by choral singing provides a natural and effective means of attenuating the auditory feedback of one's own voice.Individuals with PD are known to misperceive their vocal loudness and effort as being greater than they actually are, and such abnormal auditory feedback has been suggested to contribute to the downscaled vocal activities related to hypokinetic dysarthria (Kwan & Whitehill, 2011;Sapir, 2014).Masking the self-monitored auditory feedback, e.g. by immersing an individual in a noisy environment, has been shown to not only increase vocal intensity but also regulate articulatory and prosodic behaviours in ways as to enhance the functional communication outcomes (e.g.speech intelligibility) against the ambient noise (Garnier, Henrich, & Dubois, 2010).As such, the sound immersion appears to elicit a goaldirected action paradigm shift through reducing selfmonitoring of auditory feedback and directing the attention focus externally towards the goal of functional communication.Such a putative paradigm shift, leading towards enhanced coupling between the goal and the activities of the physiological subsystems underlying speech production, can globally modulate the speech production mechanism to mitigate the negative impact of abnormal auditory feedback and, in turn, improve the hypokinetic dysarthria-related symptoms in individuals with PD.
In addition to the improvement of hypokinetic dysarthria, singing can also enhance the timing mechanism of speech -an area that remains largely unattended in loudness-focussed programs such as LSVT/LOUD and SPEAK OUT! and LOUD Crowd.Rhythm plays an important role in regulating speech timing (Skodda & Schlegel, 2008;Turk & Shattuck-Hufnagel, 2013).The inherent rhythms in music provide temporally structured external acoustic cues, which can facilitate the sequencing and coordination of speech motor programs (Turk & Shattuck-Hufnagel, 2014).Such an effect of external acoustic cues on speech motor control can be attributed to two possible neural adaptations in individuals with PD, through (1) enhancing the activation of the cerebello-thalamo-cortical pathway to bypass the defective BG-thalamo-cortical pathway or (2) using the external acoustic rhythms as a stimulation to facilitate the impaired BG-thalamo-cortical pathway (Leuk, Low, & Teo, 2020;McFerran et al., 2020).These putative changes would mitigate the impact of the deficient internal timing mechanism related to BG dysfunction on speech production and in turn improve the timing, rhythm, and coordination deficits in PD.
As it integrates music perception and production, choral singing provides a natural paradigm for sensorimotor rehabilitation.As an auditory stimulus, music serves as a trigger to shape the firing pattern of auditory neurons, which can subsequently be used to entrain the firing of motor neurons (Borrie & Liss, 2014;Cummins, 2009;Todd, Lee, & O'Boyle, 2002).Such entrainment would enhance the functional coupling between the auditory and motor cortical areas, allowing auditory information to be better integrated with motor programs.Prior research has identified two primary benefits related to enhanced auditory-motor coupling: (1) auditory stimulation can prime the motor system towards a state of readiness; (2) auditory rhythm can create anticipatory temporal templates to regulate motor programming and execution and thus improve movement quality (e.g.displacement, velocity, timing) (Thaut et al., 2014).Such changes, when occurring in the speech production system, are expected to globally improve the system performance on all aspects of concern.
Importantly, choir participation yields a more engaging, motivating, enjoyable, and relaxing experience compared with traditional behavioural therapies, creating a stimulating atmosphere for learning and practise, which can arguably improve higher-level cognitive and affective functions.Individuals with PD view speaking as an active process of communication relying not only on physical effort, but also on cognitive and emotional resources that take up significant attentional capacity (Yorkston, Baylor, & Britton, 2017).The cognitive burden of speaking, despite being reported by individuals with PD as a significant barrier to successful communication, is largely overlooked in traditional voice/speech therapies.Music training, on the other hand, can sharpen one's skill in perceiving and comprehending complex, hierarchical sound structures and subsequently improve cognitive prediction of upcoming auditory events (Fiveash, Bedoin, Gordon, & Tillmann, 2021;Koelsch, 2009Koelsch, , 2011)).Given the similarity between music and speech, both as temporally structured sound sequences, the music-elicited enhancement of cognitive prediction can conceivably translate into speech (Arvaniti, 2009;Beier & Ferreira, 2018).Improved cognitive prediction of speech could free up the attentional capacity taken up by the "thinking" process to allow the speaker to focus more on the physical production process, thereby facilitating the learning of production strategies for improving the quality of the speech outcomes.
A last and mostly speculative therapeutic benefit of choral singing on communication impairments pertains to its enjoyable and relaxing nature (Elefant et al., 2012), which may regulate abnormal neural oscillatory activities related to vocal tremor.Given the demonstrated link between tremor and mental stress (Buhmann, Jungnickel, & Lehmann, 2018;Schlesinger, Benyakov, Erikh, Suraiya, & Schiller, 2009), mental relaxation following choir participation has a tentative effect on reducing vocal tremor.
In addition to the rehabilitative effects on the aforementioned areas of communication impairments, the psychosocial impacts of intergenerational choir participation, which have been recognised as critical but underattended components of therapeutic activities in the current clinical practise for PD (Yorkston et al., 2017), represent a unique strength to alleviate the activity limitation/participation restriction resulting from communication impairments.Specifically, intergenerational choir, which takes place in a natural social environment with a mixed age group, may help individuals with PD to (re)establish a positive social identity and adapt to the social dynamics, thereby enhancing their social and emotional wellness (Harwood, Giles, & Ryan, 1995;Kuehne, 2003;Tajfel & Turner, 1986).Intergenerational interactions have long recognised psychosocial benefits in promoting healthy ageing (Gaggioli et al., 2014;Glass et al., 2004;Wong et al., 2021;Zhong, Lee, Foster, & Bian, 2020).As an intervention approach, the therapeutic effects of intergenerational programs have been well documented in the dementia literature (Galbraith, Larkin, Moorhouse, & Oomen, 2015;George, 2011;Wallach, Kelley, & Abrahams, 1979).For individuals with PD, psychosocial wellbeing plays a central role in their mental and physical health, as well as their quality of life (Vescovelli, Sarti, & Ruini, 2018).In a survey of the subjective experience of speech-language therapy in patients with PD, psychosocial impacts have been identified as important contributing factors to patient's perspective towards the therapy (Spurgeon, Clarke, & Sackley, 2015).As such, the psychosocial benefits of intergenerational choral singing would likely positively shape the patient's view towards the related activities.It is important to note that the impression of the therapeutic activities being repetitive, boring, and socially irrelevant has been identified as the leading cause of declining and/ or discontinuing traditional voice/speech therapies in individuals with PD (Yorkston et al., 2017).As a well-known motor learning principle, continued practice is the key to long-term changes in neural structures and pathways, which are necessary for maintaining the benefits of therapeutic activities and translating these benefits into daily activities (Fox et al., 2006).To fulfil this need, it has been recommended that therapeutic activities should be implemented in a socially relevant, group-based context for individuals with PD (Yorkston et al., 2017), which can increase their motivation and likelihood of continued practise to facilitate the maintenance and translation of the therapeutic benefits to daily communicative activities and in turn improve the quality of life.

A novel multidimensional objective measurement tool for assessing changes in communicative function elicited by intergenerational choral singing
Based on the theoretical framework above, intergenerational choral singing exhibits a variety of putative therapeutic benefits to communication impairments in PD, including hypokinetic dysarthria and the previously underexplored timing, coordination, and rhythm deficits, as well as global sensorimotor, cognitive, and affective dysfunctions.There is, however, a lack of assessment tool to comprehensively evaluate these putative therapeutic benefits.To address this gap, we propose a multidimensional acoustic analysis to objectively assess the rehabilitative effect of intergenerational choral singing on the targeted areas of communication impairments in PD as outlined in the theoretical framework.This analysis was constructed based on a variety of standard and custom-developed and validated algorithms to semi-automatically extract a multitude of evidence-based objective outcome measures from various vocal tasks for evaluating the targeted areas of communication impairments in PD.An overview of these objective outcome measures is provided in Table I, and the technical details about the measurement procedures are elaborated in the Methods section.
The impairment of the timing mechanism was assessed by a set of novel measures targeting the tempo (e.g.variable rate), coordination (e.g.dyscoordinated voicing and articulatory activities), and rhythm (e.g.abnormal rhythm of syllable stress) abnormalities (Ackermann et al., 1997;Harel et al., 2004;Kent et al., 1999;Liss, LeGendre, & Lotto, 2010;Rusz et al., 2011;Skodda & Schlegel, 2008), along with the aforementioned measures of pause, which represented an area influenced by both motor drive and timing mechanism.During choral singing, the inherent rhythms in music can provide structured timing cues to facilitate the temporal organisation of speech motor programs, which is expected to improve the tempo (e.g. more stable rate), coordination (e.g. more stable coordination between laryngeal and supralaryngeal activities), rhythm (e.g.improved syllable stress), and pause (e.g.improved breath timing allowing for more efficient use of breath groups to compensate for reduced respiratory capacity) of speech production.
In addition to motor drive and timing mechanism, both directly related to the physical process of speech production, sensorimotor, cognitive, and affective functions, which constitute central components of the higher-level speech production processes but remain largely underexplored in previous intervention studies, were assessed by novel nonlinear measures derived by approximate entropy (Richman & Moorman, 2000) and recurrence quantification analysis (Eckmann, Kamphorst, & Ruelle, 1987a;Marwan & Kurths, 2002).Nonlinear analyses are types of data analysis techniques developed upon the concept of nonlinear dynamical system, in which the relationship between the time-varying changes in the output and the input is nonlinear and nonstationary (Aubin & Ekeland, 2006).As far as is known, the vast majority of biological systems are nonlinear and nonstationary in nature, generating complex and chaotic behaviours that are hard to predict by traditional linear analyses.Hence, nonlinear analyses have gained growing interest in biomedical research due to their robustness to different signal types, owing to no a priori assumptions about linearity and stationarity.Among the variety of nonlinear analyses, entropy and recurrence dynamics-based measures have been increasingly used to quantify the nonlinear complexity of neurological, physiological and behavioural signals to inform the underlying sensorimotor, cognitive, and affective processes in a range of biomedical applications (Jackson, Tiede, Beal, & Whalen, 2016;Richman & Moorman, 2000;Shafer, Solomon, Newell, Lewis, & Bodfish, 2019;Tzinis, Paraskevopoulos, Baziotis, & Potamianos, 2018;Wang et al., 2018;Young & Benton, 2015).The rationale for these applications lies in that complexity is a central descriptor of biological systems, manifesting their ability in integrating information related to different underlying sources into a unified and coherent scene as well as their adaptability/flexibility in coping with the changing environment (Tononi, Edelman, & Sporns, 1998).Disorders and/or degeneration of biological systems, however, often lead to decreased complexity of their outputs, reflecting impaired functional integration and/or adaptation.
Entropy is a key concept in information theory, representing a quantify of information content carried by a signal (Wehrl, 1978).A complex signal with high information content, reflected by greater randomness/unpredictability of the signal, is characterised by greater entropy.For biological systems, a highly adaptable system is usually characterised by a high level of information (i.e.randomness) in its output, reflecting the ability of the system in coping with unpredictable changes in the environment (Stergiou, Harbourne, & Cavanaugh, 2006;Stergiou & Decker, 2011).The adaptability of a disordered or degenerated biological system can, however, be compromised, as manifested by reduced entropy of its output (i.e.reduced amount of information content carried), resulting in rigid, stereotyped, and inflexible behaviours as reported in both developmental disorders such as autism (Shafer, Newell, Lewis, & Bodfish, 2017) and ageing-related degeneration (Bronson-Lowe, Loucks, Ofori, & Sosnoff, 2013).For individuals with PD, converging evidence from the limb motor literature has shown reduced entropy of limb motor performance (Gil, Nunes, Silva, Faria, & Melo, 2010;Pansera et al., 2009;Pelykh, Klein, B€ otzel, Kosutzka, & Ilmberger, 2015).Related to speech, entropy-based nonlinear characterisation of speech signals has demonstrated success in differentiating pathological speech from normal speech (Fabris, De Colle, & Sparacino, 2013;Henriquez et al., 2009;Travieso et al., 2017).
Recurrence dynamics, on the other hand, portray the recurrence (e.g.periodicity) of nonlinear patterns in a signal (Marwan, Carmen Romano, Thiel, & Kurths, 2007).Higher frequency of recurrence corresponds to a structurally simpler signal with more periodic contents.A speech signal, as the integrated output of the speech production system, is underpinned by the interplay of many sensorimotor, cognitive-linguistic, and affective processes.The better the speech production system integrates these underlying processes into its output, the more complex and less periodic the output would be.This is supported by behavioural evidence showing that the structural complexity of a speech signal effectively captures the emotional state of the speaker, demonstrating high accuracy for speech emotion recognition (Tzinis et al., 2018).For a disordered biological system, its ability of functional integration is often impaired, which would in turn result in disintegration of the underlying processes and reduced structural complexity of its output (Tononi et al., 1998).In individuals with PD, reduced structural complexity of limb motor performance has been well documented (Afsar, Tirnakli, & Marwan, 2018;Flood, Jensen, Malling, & Lowery, 2019;Meigal et al., 2009;Schmit et al., 2006).Related to communication, reduced adaptation to social and cognitive stress in people who stutter has been associated with decreased structural complexity of their speech outputs, pointing towards a more restrictive and less flexible speaking style compared with that of people who do not stutter (Jackson et al., 2016).Likewise, reduced structural complexity of articulatory performance has also been reported in individuals with neuromotor speech disorder secondary to amyotrophic lateral sclerosis (Rong, 2021).
The centrality of complexity is further corroborated by neurological evidence linking the complexity of electrophysiological and neuroimaging signals with the functional connectivity across brain networks (Wang et al., 2018).The collective body of behavioural and neurological evidence as presented above provides the impetus to use entropy and recurrence dynamics-based measures, which provide complementary measures of nonlinear complexity based on the randomness/unpredictability and periodicity of the signal, respectively, to assess sensorimotor, cognitive, and affective integration and adaptation in individuals with PD.According to the theoretical framework, it is anticipated that impaired sensorimotor, cognitive, and affective integration and adaptation underlying speech production would reduce the complexity of the speech outputs in individuals with PD.Participating in intergenerational choral singing, on the other hand, would (1) increase sensorimotor coupling through entrainment of the auditory-motor network to music rhythms, (2) enhance cognitive prediction of temporally organised sound structures, and (3) improve emotional resilience and adaptation.These putative improvements in highlevel brain functions are expected to be integrated into and reflected by increased complexity of the speech output.
Lastly, vocal tremor, resulting from involuntary oscillations of the muscles controlling voice generation (Gillivan-Murphy & Miller, 2011;Richards, 2017), was assessed by the frequency and depth of vocal fundamental frequency (F0) modulation.Voice modulation analysis during sustained phonation has demonstrated success in objective characterisation of vocal tremor in neurological diseases (Hlavni� cka et al., 2020;Peplinski et al., 2019).In accordance with these existing findings, two voice modulation features -the frequency and depth of the dominant modulation pattern of F0 -were adopted to quantify the frequency and extent of vocal tremor in individuals with PD.The relaxing and immersive experience through participating in intergenerational choral singing is expected to alleviate mental stress and in turn reduce vocal tremor.
Taken together, the proposed multidimensional acoustic analysis provides a comprehensive assessment of all areas of communication impairments in PD as outlined in the theoretical framework, using a set of objective outcome measures based solidly on scientific evidence.Some of these measures reflect empirically well-established outcomes (e.g.prosody, voice quality, pause) that have been commonly used in assessment and intervention studies for PD, whereas other measures represent novel outcomes targeting important but previously underattended aspects of communication impairments in PD (e.g.tempo, coordination, rhythm, complexity, vocal tremor).

Tools for assessing changes in activity, participation, and quality of life related to intergenerational choral singing
In addition to the objective outcomes targeting communication impairments, two standard questionnaires -Voice Handicap Index (VHI; Jacobson et al., 1997) and the World Health Organization's Quality of Life assessment-abbreviated version [WHOQOL-BREF, The WHOQOL Group (1998)] -were used to obtain patient-reported outcomes at the activity/ participation and quality of life levels.VHI is a validated and widely used self-assessment of voice disabilities and their impacts on daily lives.It consists of 30 items on a 0-4 Likert scale (0: never, 1: almost never, 2: sometimes, 3: almost always, 4: always).These items are categorised into three subscalesphysical, emotional, and functional -assessing patients' self-perceived voice abnormalities, affective response to their voice disorder, and impact of the voice disorder on their daily activities, respectively.The psychometric properties of VHI demonstrate high test-retest reliability (Pearson's r ¼ 0.92) and internal consistency (Cronbach's alpha ¼ 0.95) (Jacobson et al., 1997).The total score on all subscales of VHI provides an integrative index of voice handicap, with lower scores denoting less voice handicap.In this study, the total VHI score was used as an outcome measure to assess the effect of intergenerational choral singing at the activity/participation level.
The WHOQOL-BREF is the abbreviated version of the original 100-item WHOQOL assessment, providing an efficient and integrated instrument for self-assessment of quality of life (not restricted to communication).The WHOQOL-BREF consists of 26 items on a 1-5 Likert scale, categorised into four domains -physical health, psychological, social relationships, and environment.The assessment demonstrates good psychometric properties of internal consistency (Cronbach's alpha > 0.7), discriminant validity (significant for discriminating between most sick and well individuals), and construct validity (stronger correlations for within-domain items than for across-domain items) (Skevington, Lotfy, & O'Connell, 2004).The total score in all four domains was used as an outcome measure to assess the overall quality of life, with higher scores indicating better quality of life.

Study design & hypothesis
To provide a proof of concept for the proposed theoretical framework and measurement tool, we recruited a single subject with PD to participate in 9 weekly 1-h intergenerational choir rehearsals and assessed her communicative function, activity/participation, and quality of life at three time points -before, 1 week post-, and 8 weeks post-choir participation, using combined objective (i.e.multidimensional acoustic analysis) and subjective (i.e.VHI, WHOQOL-BREF) assessment tools as described above.The results of this preliminary study would provide an initial step towards developing the evidence base for intergenerational choral singing as a rehabilitative approach to improving communication and quality of life in individuals with PD, which aligns with the goal of Phase I research in the five-phase model for clinical-outcome research (Robey, 2004).
In accordance with the theoretical framework, we hypothesised that intergenerational choir participation would improve the communicative function, activity/participation, and quality of life of the participant, which would be reflected by immediate and longer-term improvements in the objective outcome measures reflecting prosody, voice quality, pause, tempo, coordination, rhythm, complexity, and vocal tremor, as well as in the subjective outcome measures pertaining to activity/participation (i.e.reduced VHI score) and quality of life (i.e.increased WHOQOL-BREF score).

Method
The study protocol was approved by the Institutional Review Board of the University Medical Centre.Written informed consent was obtained from the participant.The intergenerational choir consisted of 9 weekly rehearsals, which occurred at a local senior activity centre, followed by a performance in the University Music Hall.The second author observed all rehearsals to take notes and provide support to the participant as needed.Data collection took place in the lab of the first author at three time points.The first session occurred before choir participation (pre) to obtain baseline performance; the second and third sessions happened 1 week (post1) and 8 weeks (post2) after the participant completed the 9 weekly choir rehearsals to assess the short-and long-term changes elicited by the choir participation, respectively.

Participant
The participant was an 82 year old woman who was diagnosed with idiopathic PD by a certified neurologist approximately two years prior to enrolling into this study.The participant was recruited from the local PD support group and met the following inclusionary criteria: (1) self-reported speech/voice changes since the disease onset, (2) had no cognitive impairment as indicated by standardised measures of cognition, (3) had no neurological diseases or injury other than PD, and (4) did not have voice/speech treatment within 6 months prior to study participation.
The participant was on levodopa medication at the time of study participation.Her baseline speech characteristics were consistent with hypokinetic dysarthria, as determined by the second author.In addition, she presented with speech timing abnormality, as reflected by her faster-than-normal speaking rate [210.90words per minute (WPM), exceeding the norm of healthy adult speakers (approximately 180 WPM)].She also had a mild impairment of speech intelligibility (90.90%).Both intelligibility and speaking rate were obtained by a standard functional speech assessment -the Sentence Intelligibility Test (SIT; Yorkston, Beukelman, Hakel, & Dorsey, 2007).In this test, the participant read 11 randomly generated sentences of 5-15 words.Speech intelligibility and speaking rate were derived as the percentage of intelligible words and number of words per minute, respectively, from the SIT recording.

Intergenerational choir
The participant participated in 9 weekly choir rehearsals, each lasting 1 h.The choir was conducted by students in the music therapy and music education departments of the university.At each rehearsal, there were 20-25 choir members, consisting of a mixed age group including college students and senior-aged individuals from the local community.
Twenty-one songs with varying rhythmic complexities (see Appendix A) were taught and rehearsed throughout the choir.Each rehearsal typically started with a non-singing warm-up such as a name game to create a welcoming social atmosphere.Following the warm-up exercise, the selected songs were taught by the leading student modelling in a phrase-by-phrase "echoing" fashion (i.e.listening-repeating) and accompanied by musical instruments (e.g.kazoos, shakers, triangles, handbells, guitar, and keyboard) in the latter half of the rehearsals.Choir members were provided with a copy of written lyrics.Later rehearsals included additional exercises in harmonisation and counterpoint, with subgroups of choir participants learning separate rhythms or notes to improve their general musicality.

Speech and behavioural data collection
Within each data collection session, the participant performed various vocal tasks, including sustained phonation (i.e. say "ah" with normal pitch for as long as possible), oral diadochokinesis (DDK: repetitions of "puh tuh kuh" as fast as possible on one breath), reading of the Caterpillar passage (Patel et al., 2013), in a quiet lab environment.All tasks except passage reading were repeated three times.Speech was recorded by a head-mounted microphone (Crown CM311A) placed approximately 5 cm away from the right lip corner, processed by the Behringer Xenyx 802 sound conditioner, and acquired at 48 kHz by the Audacity software on a Dell Precision Tower desktop with the Windows 10 operating system.In addition, the VHI and WHOQOL-BREF were administrated to the participant.

Outcome measures
Twenty-four outcome measures including 22 objective outcomes and two subjective outcomes were derived from the acoustic and behavioural data collected during each session.All measures were descriptively compared across sessions to evaluate the impacts of intergenerational choir participation on communication and overall quality of life.The outcome measures are described below.

Objective outcome measures
Twenty-two linear and nonlinear features were derived (semi-)automatically by a multidimensional acoustic analysis implemented in MATLAB (Mathworks, Inc.).These features, as summarised in Table I, assessed the communication impairments in the following domains -motor drive, timing mechanism, higher-level sensorimotor, cognitive, and affective integration and adaptation, and involuntary oscillations.Each domain included one or more subdimensions in line with the areas of communication impairments as specified in the theoretical framework.These subdimensions were categorised as follows: (1) motor drive -prosody, voice quality, pause, (2) timing mechanism -pause, tempo, coordination, rhythm, (3) higher-level brain functions -complexity, and (4) involuntary oscillations -vocal tremor.The features were extracted in a task-specific manner, wherein the task-feature combinations were determined based on the evidence from prior work (Rong & Heidrick, 2021) to optimise the efficacy in capturing the specific deficits along each subdimension.These task-specific features were averaged across trials and served as the outcome measures for objective evaluation of the singing-elicited changes in communicative function.The descriptions and evidence base of these objective outcome measures are listed in Table I and elaborated below.

Intergenerational Choir and Speech Rehabilitation 731
Prosody Four prosodic measures were extracted automatically from the passage reading task.F0 and intensity traces were first derived using a 1-ms moving window.The interquartile range (IQR) and standard deviation of F0 and intensity were then calculated to measure pitch and loudness variations throughout the passage.
Voice quality Mean F0, jitter, shimmer, noise-to-harmonic ratio (NHR), subharmonic-to-harmonic ratio (SHR), and cepstral peak prominence (CPP) were extracted from the sustained phonation task to characterise voice quality.The first four features were derived by the PRAAT Voice report function (Boersma & Weenink, 2014).SHR and CPP were calculated by custom functions in MATLAB based on the algorithms developed by Sun (2002) and Hillenbrand, Cleveland, and Erickson (1994), respectively.
Pause Three pause measures were derived from the passage reading task.Pauses were automatically detected as silence intervals longer than 200 ms (Rochester, 1973), using an algorithm adopted from Green, Beukelman, and Ball (2004).The total number of pauses, percentage of pause time, and total pause time were calculated to characterise the pause pattern during passage reading.
Tempo Three temporal features, including sequential motion rate (SMR), cross-unit temporal variability, and unit-to-unit temporal variability, were derived from the oral DDK task by a custom semi-automatic acoustic algorithm developed and validated by Rong (2020).These features targeted different tempo deficits, with SMR targeting abnormal rate, and crossunit temporal variability and unit-to-unit temporal variability targeting temporal instability (Rong, 2020;Rong, Yunusova, Richburg, & Green, 2018).All features were extracted from the amplitude modulation envelope of the acoustic waveform (see an example in Figure 2), excluding the first and last trisyllabic units.The specific procedures of the analysis are provided in Appendix B.
Coordination One feature reflecting the stability of temporal coordination between laryngeal (i.e.voicing) and supralaryngeal (i.e. release of initial [p] in the trisyllabic unit) activities, that is, the voice onset time (VOT) variability across the trisyllabic units, was extracted semi-automatically from the oral DDK task (see an example in Figure 2), excluding the first and last units.The specific procedures of the analysis are again specified in Appendix B.
Rhythm One feature related to the rhythm of syllable stress was extracted automatically from the passage reading task.This feature, denoted as E_deltatheta in Table I, was derived following similar procedures as in Liss et al. (2010), by calculating the energy within the delta (1-3 Hz) and theta (4-8 Hz) bands of the power spectrum of the wideband modulation envelope and normalising it by the total spectral energy and its amplitude modulation envelope (red line), as well as the peaks (green stars) and troughs (green dots) of the amplitude modulation envelope.The peaks correspond to the location of vowel nucleus of each syllable (i.e."puh," "tuh," "kuh") and the troughs correspond to the location of syllable onset/offset.The black line is the smoothed peak envelope (i.e.spline-interpolated local maxima separated by at least 1000 points), providing finer-grained spatiotemporal information about the waveform to serve as a secondary reference for manual peak/trough correction when the peaks and troughs on the amplitude modulation envelope are visually not prominent.The lower panel shows the vocal fundamental frequency (F0) trace (blue dots), along with the onset and offset of each trisyllabic unit (vertical green lines) and the voice onset for the initial syllable within the trisyllabic unit (vertical red lines).Note that the first and last trisyllabic units are excluded from analysis.
within the 0-10 Hz range (i.e. the temporal frequency range containing most suprasegmental information).For details about the rhythm analysis, refer to Appendix C.
Delta and theta rhythms reflect the timescales of two critical linguistic events -prosodic stress and syllable (Leong & Goswami, 2015).During auditory processing, the auditory cortex tracks the stress and syllable rhythms of speech through entraining its delta and theta oscillations to the modulation envelope of the speech acoustic signal, allowing the acoustic signal to be parsed into stress and syllable units to enable hierarchical speech processing and comprehension (Bosker & Ghitza, 2018;Doelling, Arnal, Ghitza, & Poeppel, 2014;Giraud & Poeppel, 2012;Hyafil, Fontolan, Kabdebon, Gutkin, & Giraud, 2015;Riecke, Formisano, Sorger, Bas ¸kent, & Gaudrain, 2018).Hence, the delta-theta rhythms of the modulation envelope encode critical time information reflecting the intervals at which stress and syllable units recur during running speech.Disrupted envelope modulation in the delta-theta frequency range, reflective of impaired control over stress and syllable rhythms, has shown a detrimental effect on speech intelligibility (Doelling et al., 2014).As such, the extent of rhythmic modulation in the delta-theta band, as measured by the normalised delta-theta energy in the envelope modulation spectrum, provides a means of assessing rhythm impairment pertaining to syllable stress.
Complexity Two nonlinear features were derived algorithmically from the passage reading task to measure the complexity of speech performance.First, 13 Melfrequency cepstral coefficients (MFCCs) were extracted from the acoustic signal.MFCCs are the cepstral representations of an audio signal, which are calculated by taking the Fourier transform of the logarithmic spectrum (Rabiner & Schafer, 2010).For speech signals, the acoustic information related to the resonance structure of the vocal tract is mostly encoded in the lower order MFCCs (e.g.12-13 MFCCs).These lower-order MFCCs have been widely used in acoustic-to-articulatory inverse mapping in speech studies (Chartier, Anumanchipalli, Johnson, & Chang, 2018;Goldstein, 2019;Mitra, Nam, Espy-Wilson, Saltzman, & Goldstein, 2011).Based on these lower-order MFCCs, Goldstein (2019) has proposed a modulation function, defined as: where MFCCði, kÞ represents the i th MFCC at the k th time frame.This acoustic-based modulation function has been demonstrated be correlate strongly with an articulatory-based modulation function reflecting the total kinetic energy of all primary oral articulators including tongue tip, tongue blade, tongue body, tongue dorsum, upper and lower lips, and jaw (Goldstein, 2019).This MFCC modulation function was adopted in the present study as an acoustic representation of articulatory performance, based on which two nonlinear features were extracted to quantify the complexity of articulatory performance during passage reading.
The first feature was the approximate entropy (ApEn) of the MFCC modulation function.Approximate entropy is a nonlinear measure of complexity based on the amount of randomness/unpredictability in a time series (Pincus, Gladstone, & Ehrenkranz, 1991).As noted earlier in the Introduction section, the randomness/unpredictability of the outputs of biological systems is associated with the adaptability of these systems to cope with the changes in the external and internal environments.The approximate entropy of the MFCC modulation function thus provides a proxy to inform the adaptability of the speech production system.For mathematical details of approximate entropy, refer to Appendix D.
The second feature was the determinism of the MFCC modulation function, which was derived by recurrence quantification analysis (RQA), using the CRPtool MATLAB toolbox (Marwan, 2020;Marwan et al., 2007).RQA is a nonlinear technique to quantifying the recurrence of nonlinear patterns in a time series based on the dynamics of the phase space trajectory of the signal (Eckmann, Kamphorst, & Ruelle, 1987b;Marwan et al., 2007).Determinism measures the overall periodic contents in the phase space trajectory, providing a quantitative metric of structural complexity.According to the background as laid out earlier, the structural complexity of a biological time series is associated with the functional integration of the underlying biological processes.The determinism of the MFCC modulation function is therefore used as a proxy to inform the functional integration of the speech production system.Mathematical formula and technical details for RQA are provided in Appendix E.
Vocal tremor Two features were derived from the F0 modulation spectrum for the sustained phonation task to characterise vocal tremor.First, F0 was calculated in 1-ms intervals, using the cross-correlation method.The modulation spectrum of F0 was then derived by Welch's method.The frequency at which the maximum spectral amplitude occurred was identified as the primary modulation frequency of F0.In addition, the depth of F0 modulation was derived as the maximum spectral amplitude normalised by the total spectral energy.These features jointly characterised the dominant frequency and extent of oscillatory fluctuations in vocal fundamental frequency (i.e.vocal tremor).

Intergenerational Choir and Speech Rehabilitation 733
Subjective outcome measures In addition to the objective outcomes above, two patient-reported subjective outcomes were obtained.The first was the VHI score, assessing patient's selfperceived activity limitation/participation restriction related to voice/speech impairments.The second was the WHOQOL-BREF score, reflecting patient's selfperceived quality of life.

Objective outcomes
Figure 3 displays the bar plots for all objective outcome measures by session.To illustrate the extent of pre-post changes, the percentage of short-and longterm changes in these measures before and after intergenerational choir participation are provided in Table II.Based on the comparisons between the pre and post1 sessions in Figure 3 and Table II, notable changes were observed in various objective outcomes before and immediately after intergenerational choir participation, including (1) decreased cross-unit and unit-to-unit temporal variability, reflecting a trend towards improved temporal control, (2) decreased VOT variability, indicating a trend towards improved stability of laryngeal-supralaryngeal temporal coordination, (3) decreased determinism of the MFCC modulation function, reflecting a trend towards increased complexity of speech performance, (4) increased F0 IQR, F0 standard deviation, and intensity standard deviation, manifesting a global trend towards improved prosody, (5) increased number of pauses with no substantial change in the percentage of pause time and total pause time, which together implied a trend towards more frequent, shorter pauses, (6) decreased jitter, shimmer, NHR, and SHR, collectively revealing a trend towards improved voice quality, and (7) increased F0 modulation frequency, denoting an upward shift of vocal tremor frequency.Other objective outcomes, including SMR, ApEn, intensity IQR, E_deltatheta, mean F0, CPP, and F0 modulation depth, showed relatively minor changes between the pre and post1 sessions.
Further comparisons between the pre and post2 sessions revealed the maintenance of the trends of change in cross-unit and unit-to-unit temporal variability, VOT variability, determinism of the MFCC modulation function, F0 IQR, intensity standard deviation, and F0 modulation frequency.The changes in F0 standard deviation, pause pattern (#pause, %pause, pause time), and voice quality measures (jitter, shimmer, NHR, SHR) were, however, diminished in the post2 session.
Taken together, most of the objective outcome measures exhibited notable trends of change immediately following intergenerational choir participation, in directions consistent with the predictions of the theoretical framework.Some, but not all, of these changes were maintained in the longer term (i.e. 8 weeks post-choir participation).

Subjective outcomes
The VHI score was 74, 75, and 59, and the WHOQOL-BREF score was 96, 95, and 97, during pre, post1, and post2, respectively.The percentage of pre-post changes in these subjective outcomes are shown in Table II.Compared with the objective outcomes, the changes in the subjective outcomes before and after choir participation were overall smaller.The only notable change in the subjective outcomes was the VHI score during the post2 session, which revealed a decrease relative to the pre session, implying a trend towards reduced voice handicap and hence improved activity limitation/participation restriction related to communication impairments.Such a change was, however, not observed in the post1 session.

Discussion
In this study, we established a theoretical framework conceptualising the therapeutic benefits of intergenerational choral singing to different aspects of communication impairments, including hypokinetic dysarthria and various previously underexplored deficits associated with timing, coordination, rhythm, sensorimotor and higher-level brain functions, and vocal tremor, as well as activity limitation/participation restriction and quality of life, in individuals with PD.In line with this framework, we further proposed a semi-automatic multidimensional acoustic analysis to derive a set of evidence-based objective outcome measures for assessing the targeted aspects of communication impairments.These objective outcomes, along with two subjective, patient-reported outcomes (i.e.VHI, WHOQOL-BREF) assessing activity limitation/participation restriction and quality of life, comprised an integrated measurement tool for evaluating the therapeutic benefits of intergenerational choral singing as conceptualised in the theoretical framework.To provide a proof of concept, we recruited a single participant with PD to participate in 9 weekly 1-h intergenerational choir rehearsals and assessed the changes in the targeted aspects of communication and quality of life before, 1 week post-, and 8 weeks post-choir participation, using the integrated measurement tool.Notable trends of change were observed in most objective outcome measures at 1 week post-choir participation, in directions consistent with the predictions of the theoretical framework.Some, but not all, of these changes were maintained at 8 weeks post-choir participation.The subjective outcome measures showed overall smaller changes before and after choir participation.This preliminary exploratory study provided the theoretical groundwork and an integrated measurement tool for future validation of intergenerational choral singing as a novel rehabilitation approach to improving communication impairments, activity limitation/participation restriction, and quality of life in individuals with PD.

Changes in speech performance at one week post intergenerational choir participation
Among different dimensions of communication impairments in PD, tempo, coordination, complexity, prosody, pause, and voice quality revealed a trend of improvement at 1 week post-choir participation.The changes in tempo were reflected by reduced crossunit and unit-to-unit temporal variability during the oral DDK task (percentage of change: À 10.87% -À 41.28%).These changes, in line with the predictions of the theoretical framework, can be attributed to the inherent rhythms in music providing external acoustic cues to enhance the temporal control of orofacial muscles, thereby improving the temporal stability of oral DDK rate.As a maximum performance task, oral DDK exerts a high demand on temporal control of orofacial muscles in generating rapid, coordinative contractions in a rhythmic manner (Kent, Kent, & Rosenbek, 1987).Impaired temporal control of orofacial muscles can disrupt the rhythmic structure of oral DDK and in turn increase the temporal variability of DDK rate.Cross-unit and unit-to-unit temporal variability have been previously demonstrated to be highly sensitive in detecting temporal control deficits in both individuals with amyotrophic lateral sclerosis and cerebral palsy (Allison, Nip, Rong, & Green, 2019;Rong, 2020;Rong, Yunusova, Eshghi, Rowe, & Green, 2020;Rong et al., 2018).The current study further indicates that, in addition to detecting disease-related changes, cross-unit and unit-to-unit temporal variability can also capture the changes in temporal control of orofacial muscles related to a potential rehabilitation approach (i.e.intergenerational choral singing) in an individual with PD. Between the two measures, cross-unit temporal variability assesses global temporal deficits during the DDK task, which is susceptible to fatigue (i.e. an incremental increase in syllabic unit duration due to fatigue would inflate cross-unit temporal variability), whereas unit-to-unit temporal variability focuses more on local temporal deficits and is less susceptible to fatigue.The finding that the singing-related changes in temporal control are manifested by both cross-unit and unit-to-unit temporal variability accentuates the need for both measures in the assessment.The change in coordination was manifested by reduced VOT variability during the oral DDK task (percentage of change: À 9.75%).As a measure of temporal coordination between the laryngeal and supralaryngeal systems, VOT has proven to be susceptible to PD (Harel et al., 2004;Kent et al., 1999;Rusz et al., 2011).The observed decrease in VOT variability post-choir participation can be interpreted as increased stability of laryngeal-supralaryngeal temporal coordination, which resonates with the prediction of the theoretical framework, suggesting that the inherent rhythms in music might facilitate the temporal coordination between different physiological processes underlying speech production.
The change in the complexity of speech performance was reflected by decreased determinism of the MFCC modulation function during the passage reading task (percentage of change: À 14.58%).The MFCC modulation function, derived from the lower level of the speech production hierarchy (i.e.acoustic), has been previously demonstrated to encode the underlying articulatory activities (Goldstein, 2019).Articulatory activities themselves are further underpinned by a variety of sensorimotor, cognitive, and affective factors at the higher levels of the speech production hierarchy (Tononi et al., 1998).Therefore, in view of hierarchical control of speech production, the MFCC modulation function can be interpreted as the integrated outcome of the functioning of a complex sensorimotor and high-order brain network.The observed decrease in the determinism of the MFCC modulation function, implying increased structural complexity of the signal, is consistent with the prediction of the theoretical framework, revealing a trend of improvement in functional integration of the sensorimotor and higher-order brain processes related to speech production.Such a change may be attributed, in part, to the music-elicited enhancement of entrainment between the auditory and motor cortical areas, which has proven to facilitate the integration of auditory feedback with speech motor programming (Thaut, 2015).Additionally, enhanced social engagement and mood following intergenerational choir participation may improve emotional resilience of the participants (Unwin, Kenny, & Davis, 2002).Such an emotional response would be integrated with and reflected in vocal activities (Mozziconacci & Hermes, 1997;Scherer, Sundberg, Fantini, Trznadel, & Eyben, 2017;Vroomen, Collier, & Mozziconacci, 1993).These sensorimotor and higher-order cerebral interactions can conceivably facilitate functional integration of different sources underlying speech production, thereby increasing the complexity of speech performance.
The change in prosody was manifested by increased F0 IQR, F0 standard deviation, and intensity standard deviation during the passage reading task (percentage of change: 11.17% À 41.46%).These changes are in keeping with the predictions of the theoretical framework as well as the findings of prior studies, which have reported a similar therapeutic effect of music on prosody in individuals with PD (Di Benedetto et al., 2009;Tanner et al., 2016).The resulting prosodic changes reflect improved control of respiratory-laryngeal muscles, which, according to the theoretical framework, could be attributed to the high-effort vocal exercises during choral singing resulting in increased neural drive to the respiratorylaryngeal muscles.
The changes in voice quality were manifested by reduced jitter, shimmer, NHR, and SHR during the sustained phonation task (percentage of change: À 23.70% -À 74.34%).These changes are consistent with the predictions of the theoretical framework, pointing towards increased neural drive to the respiratory-laryngeal system owing to the high-effort vocal exercises during choral singing.These changes in voice quality, along with the prosodic changes above, tend to result in an improvement in hypokinetic dysarthria.This observation is corroborated by the findings of prior studies reporting a similar effect of singing on hypokinetic dysarthria-related symptoms (Di Benedetto et al., 2009;Evans et al., 2012;Tanner et al., 2016).
Lastly, the change in pause pattern, as indicated by increased number of pauses (percentage of change: 12.50%) with no substantial change in the percentage of pause time and total pause time (percentage of change: 1.66% À 4.03%) during the passage reading task, implied more frequent, shorter pauses following intergenerational choir participation.Abnormal pause patterns characterised by reduced number of pauses and inappropriate pause locations have been commonly reported in individuals with PD, reflecting an interplay of respiratory, cognitive-linguistic, and timing deficits (Darling-White & Huber, 2020; Skodda & Schlegel, 2008).The finding that more frequent, shorter pauses were inserted during passage reading post-choir participation revealed a trend towards improved management of breath timing by utilising smaller breath groups to compensate for respiratory impairment (e.g.reduced lung volume).Such a compensatory strategy is common in older adults as a behavioural adaptation to the age-related changes in the respiratory physiological system (Huber, 2008;Huber & Darling, 2011).During singing, a well-known respiratory strategy, referred to as "catch breath", is characterised by brief and fast rhythmic breathing, resembling the compensatory strategy during speech in older adults.The finding of this study suggests that the catch breath strategy during singing may be adopted and carried over to speech, enabling the participant to insert more frequent, shorter pauses to mitigate the effect of respiratory impairment on speech pause pattern.

Changes in speech performance at eight weeks post intergenerational choir participation
Among different changes in speech performance as observed at 1 week post-choir participation, only those related to tempo, coordination, complexity, and prosody were maintained at 8 weeks post-choir participation.The changes related to pause and voice quality were diminished.Given the narrow scope of the data based on a single participant and the focus of this study on piloting the feasibility of the theoretical framework and the measurement tool rather than assessing the efficacy/effectiveness of intergenerational choral singing, we restrained from reading too much into the findings in terms of why some changes were maintained while other were not.One speculative interpretation of the diminished effects of intergenerational choir participation on pause and voice quality, both associated with the function of the respiratory-laryngeal system, is that the retention of the therapeutic benefits of intergenerational choral singing on the respiratory-laryngeal system may require a maintenance program -a critical component of the existing voice treatments such as LSVT/ LOUD and SPEAK OUT! and LOUD Crowd.Although the participant self-reported practising singing and other vocal exercises from her previous LSVT/LOUD program 1 during the post-choir participation study period (i.e. between post1 and post2), none of these exercises were practised in a systematic way.Further research is needed to delineate the need for maintenance practises to retain the therapeutic benefits of intergenerational choral singing.

Changes in subjective outcomes before and after intergenerational choir participation
Overall, intergenerational choral singing did not show an interpretable positive effect on the subjective outcomes for the participant.The most notable prepost change was the decrease in the VHI score at 8 weeks post-choir participation, reflecting a self-perceived improvement in activity limitation/participation restriction related to communicative impairments.However, since this improvement was not observed at 1 week post-choir participation, it should not be over-interpreted.Although the subjective outcome measures did not exhibit the expected changes as predicted by the theoretical framework, it should be noted that the relatively high baseline functional performance of the participant might have limited the extent of change in the outcomes at the activity/participation and quality of life levels.The finding that various objective outcomes exhibited detectible changes before and after choir participation despite the overall lack of meaningful change in the subjective outcomes resonates with the existing studies of other clinical populations (Rong, Yunusova, Wang, & Green, 2015;Rong et al., 2018), suggesting that an objective measurement tool like the multidimensional acoustic analysis in this study may provide more sensitive outcome measures for detecting and monitoring subtle but important changes in the speech production system.Such changes, despite not exerting a notable impact on the patient-reported, holistic outcomes in a mildly impaired individual like the participant in this study, might add up and lead to more significant and meaningful changes in these holistic outcomes as the disease progresses to more severe stages.

Limitations and implications for future studies
It must be emphasised that this study, which fits within the scope of Phase I research in the five-phase model for clinical-outcome research (Robey, 2004), is by no means an effectiveness or efficacy study.The findings should, therefore, not be interpreted in terms of effectiveness or efficacy but rather be treated as an initial effort in conceptualising and piloting the empirical feasibility of intergenerational choral singing as a rehabilitative approach to improving communication and quality of life in individuals with PD.Because of the exploratory nature, this study (1) only recruited a single case, (2) did not employ a control group, and (3) did not blind the investigators to the participant.Due to these limitations, it is important to not over-interpret the findings.Furthermore, since this study did not employ a case-control design, we did not comprehensively collect the clinical data of the participant.Only the demographic and clinical information directly related to this study, which focussed on communication, was collected as part of the screening and interviewing processes.In future larger-scale case-control studies, the clinical characteristics of the participants, such as disease presentation (e.g. as per Movement Disorder Society clinical diagnostic criteria for PD), staging (e.g. as per Hohen and Yahr scale), severity and progression (e.g. as per Unified Parkinson's Disease Rating Scale), and treatment (e.g.dosage), should be collected and calibrated in a more systematic manner.
Despite the preliminary nature of the findings, this study provides the theoretical groundwork and an empirical measurement tool for future validation of intergenerational choral singing as a novel rehabilitation approach for individuals with PD.Future work should focus on systematic analytical and clinical validation of the protocol developed in this study for assessing and quantifying changes in the targeted areas of communication impairments, activity limitation/participation restriction, and quality of life in larger cohorts of individuals with PD with carefully controlled clinical characteristics.This protocol has several notable strengths, including (1) non-invasiveness, (2) simple experimental procedures with minimal equipment requirement, and (3) efficient data analyses using semi-automatic computational algorithms.Moreover, the protocol is scalable in that the experimental procedures (e.g.adding in new tasks such conversational speech) and implementation (e.g.translating into clinical or home environments) can be easily modified to accommodate emerging needs in new applications.
Looking into the future, a logical next step is to convert the protocol of this study into a virtual form (e.g. by implementing the recording procedures on an online platform or mobile app), which may provide a remote storage and monitoring platform to allow patients with PD to self-record vocal and communicative activities, report disease status, and store and transfer data through the platform.This, as a result, would enable researchers and clinicians to remotely access patient data, and analyse and evaluate the outcomes of interest, such as those related to disease progression and treatment response.Another future direction is to design randomised controlled trials to evaluate the effectiveness and efficacy of intergenerational choral singing as a rehabilitative approach for individuals with PD.

Conclusion
Music as a widely accepted rehabilitative modality has shown strong therapeutic potential for many motor, sensory, cognitive, and affective disorders.For individuals with PD, singing has been a popular remedy for communication impairments.The scientific basis for such a remedy, however, remains poorly understood.The extant empirical evidence for the therapeutic benefits of singing to communication impairments in PD is narrowly focussed on hypokinetic dysarthria and patient/caregiver/clinicianreported functional outcomes.This study, for the first time, conceptualised a theoretical framework outlining the rehabilitative effect of intergenerational choral singing -an active group-based performing art -on PD-related communication impairments across multiple domains, including motor drive, timing mechanism, sensorimotor and higher-level cognitive and affective functions, and vocal tremor, as well as activity limitation/participation restriction, and quality of life.Furthermore, a semi-automatic multidimensional acoustic analysis was developed to derive a set of evidence-based objective outcome measures for assessing the targeted domains of communication impairments in the theoretical framework.To provide preliminary feasibility data, we used the multidimensional acoustic analysis, along with two standard selfassessments (i.e.VHI, WHOQOL-BREF) focussing on activity/participation and overall quality of life, to evaluate the targeted aspects of change as outlined in the theoretical framework in a single participant with PD, at 1 week and 8 weeks post-participating in a 9-week intergenerational choir.Various changes were identified across multiple domains at 1 week postchoir participation, in directions consistent with the predictions of the theoretical framework.These changes collectively revealed a trend towards improved tempo, coordination, complexity, prosody, pause, and voice quality.Some, but not all, of these changes were maintained at 8 weeks post-choir participation.The outcomes at the activity/participation and quality of life levels did not exhibit meaningful changes before and after choir participation, possibly due to the relatively high baseline functional performance of the participant yielding a ceiling effect.Collectively, the findings shed new light on intergenerational choral singing as a low-cost, safe, and engaging communication treatment alternative for individuals with PD.

Notes
1.The participant received the LSVT/LOUD about one year prior to her enrollment in this study.During the period of choir participation, she was restrained from all voice/ speech treatments.
x j ð Þ ðj � N À m þ 1Þ being similar to xðiÞ within the allowance as determined by the threshold r (set to 0.2) is calculated as: where d½x i ð Þ, xðjÞ� is the maximum distance between the scalar components of x i ð Þ and xðjÞ: The logarithmic transformation of C m i r ð Þ is averaged across all data vectors as / m r ð Þ ¼ ðNÀ m þ 1Þ À 1 P NÀ mþ1 i¼1 logðC m i ðrÞÞ, based on which approximate entropy is calculated as: ApEn ¼ / mþ1 ðrÞ À / m ðrÞ Mathematically, approximate entropy measures the logarithmic likelihood of similar patterns observed in m-length vectors being no longer similar in vectors of an incremental length.Practically, a signal with a greater extent of randomness would have higher approximate entropy.

Appendix E. Mathematics for recurrence quantification analysis (RQA)
RQA is a nonlinear technique for quantifying the structural complexity of a signal based on the recurrence plot.Recurrence plot is a visualization tool for displaying the dynamics of the phase space trajectory of a signal.Let xðtÞ be the phase space trajectory of the signal, its inherent dynamics can be represented as: u iþs , . . .u iþ mÀ 1 ð Þs �, where T is the sampling period, u is a time-series of length m, m is the embedding dimension, and s is the time delay.The values of m and s are set to 4 and 15, as determined by standard methods (i.e.false nearest neighbors and mutual information).
Based on the dynamics of the phase space trajectory, recurrence plot is generated as: R i, j ¼ HðeÀ k x !i À x !j kÞ, where e is a threshold set to 0.2, k � k is the distance, and H is the Heaviside function that has a value of 1 when k x !i À x !j k � e and a value of 0 when k x !i À x !j k > e: A recurrence point P i, j is defined as an occurrence when the phase space trajectory at state j returns to a previous state i so that the distance between x !j and x !i is below the threshold e: The recurrence plot displays all recurrence points throughout the time-series.
RQA identifies all diagonal structures in the recurrence plot, which characterize the periodicity of the signal.Based on these diagonal structures, determinism (DET) is derived as the ratio of recurrence points forming diagonal structures to all recurrence points.DET quantifies the overall periodic content in the signal, which is inversely related to the structural complexity of the signal.

Figure 1 .
Figure 1.Theoretical framework conceptualising the rehabilitative effect of intergenerational choral singing on impairments to communicative function, activity limitation/participation restriction, and quality of life in individuals with Parkinson's disease.Different aspects of impairments to communicative function (bullet points) are categorised into four domains (bubbles) based on their neural underpinnings (underlined texts).The arrows denote how intergenerational choral singing putatively improves each domain of communication impairments, activity limitation/participation restriction, and quality of life.Note that the effect of intergenerational choral singing on vocal tremor is tentative, as marked by the striped bubble and arrow to distinguish it from the other domains of communication impairments.

Intergenerational
Choir and Speech Rehabilitation 727Table I. Summary of objective outcome measures derived by the multidimensional acoustic analysis. .(2019) Hlavni� cka et al. (2020) Note.DDK: diadochokinesis; F0: vocal fundamental frequency; IQR: interquartile range; SMR: sequential motion rate; MFCC: Mel-frequency cepstral coefficient.Pause is a subdimension of both domains of motor drive and timing mechanism.

Figure 2 .
Figure 2. Example of semi-automatic acoustic diadochokinetic (DDK) analysis.The upper panel shows the acoustic waveform (blue)and its amplitude modulation envelope (red line), as well as the peaks (green stars) and troughs (green dots) of the amplitude modulation envelope.The peaks correspond to the location of vowel nucleus of each syllable (i.e."puh," "tuh," "kuh") and the troughs correspond to the location of syllable onset/offset.The black line is the smoothed peak envelope (i.e.spline-interpolated local maxima separated by at least 1000 points), providing finer-grained spatiotemporal information about the waveform to serve as a secondary reference for manual peak/trough correction when the peaks and troughs on the amplitude modulation envelope are visually not prominent.The lower panel shows the vocal fundamental frequency (F0) trace (blue dots), along with the onset and offset of each trisyllabic unit (vertical green lines) and the voice onset for the initial syllable within the trisyllabic unit (vertical red lines).Note that the first and last trisyllabic units are excluded from analysis.

Table II .
Percentage of change in the outcome measures before and after intergenerational choir participation.