Abstract
INTRODUCTION
Parkinson’s disease (PD) is a common, slowly progressive idiopathic neurologic disease, generally believed to be caused by the degeneration of dopaminergic neurons in the substantia nigra. In the past, PD was traditionally considered a movement disorder, but nowadays, there is a growing body of evidence demonstrating that the disease also affects other aspects of life including communication. In particular, the vast majority of individuals with PD develop speech symptoms as the disease progresses [1, 2]. PD is prototypically associated with hypokinetic dysarthria, a perceptually distinctive motor speech disorder, the characteristics of which are most evident in articulation, voice, and prosody [3].
Prosody pertains to the intonation, stress, and rhythm of speech [4]. Dysprosody in parkinsonian dysarthria is characterized by variable speech rate, short phrases, short rushes of speech, inappropriate silences, reduced stress, monoloudness, and monopitch [3]. Such prosody breakdowns have a potentially detrimental impact on speech intelligibility [4, 5] and speech naturalness [4], which may in turn lead to a reduced ability to communicate properly and thus function fully in society [6]. When addressing prosody in this population, we naturally think of assessing and training productive skills, because dysarthria is, after all, a motor speech disorder. It is, however, important to remember that adequate receptive skills are important for successful therapy of productive skills. As speech therapy relies on modelling and auditory-perceptual feedback, adequate auditory perception skills are required to ensure adequate response evaluation by the speaker [7]. However, the literature provides us with a few additional reasons why assessing and training receptive prosodic skills in individuals with hypokinetic dysarthria due to PD is relevant.
First, there is evidence that reception of prosody goes downhill as a result of ageing. Elderly adults, when compared to young adults, demonstrate a significant deficit in the reception of emotional prosody, independently of age-related factors such as hearing loss or cognitive decline [8–10]. Interestingly, Paulmann, Pell, and Kotz [11] compared middle-aged and young adults and concluded that the decline in emotional prosody recognition already starts well before elderliness. Other reports indicate that such a deficit may also extend to linguistic prosody: older adults, when compared to younger adults, seem less able to identify the mode (statement, question, or command) of a sentence based upon the intonation contour [12–14], and to make efficient use of prosodic information to interpret syntactically ambiguous sentences [15]. As PD usually begins in mid-to-later life [3]), it seems advisable to take into account such ageing effects when assessing prosody in middle-aged or elderly individuals with hypokinetic dysarthria due to PD.
Second, and perhaps more importantly, there is evidence for a disturbed reception of prosody in individuals with PD. The bulk of the accumulated knowledge in this area pertains to the reception of one specific communicative function of prosody, namely emotional prosody (i.e. noticing emotion as conveyed through the voice). This literature has recently been reviewed by three independent research teams [16–18], who all concluded that there are numerous studies supporting the notion of a deficit of reception in emotional prosody in PD. Gray and Tickle-Degnen [16] also found that this deficit extended across various task types (identification, discrimination, and rating tasks). Kwan and Whitehill [17] argued that there is fairly strong evidence in favour of the involvement of working memory as a possible explanation for the deficit. Péron et al. [18] related the deficit to a disruption in amygdala function and impairment of the dopaminergic pathways and/or the basal ganglia. Gray and Tickle-Degnen [16] argued that the deficit may contribute to interpersonal difficulties and thus increase social stress, which may in its turn accelerate disease progression. In order to interrupt this vicious cycle, the authors suggested training individuals with PD in emotion recognition.
Comparatively few studies have documented the reception of other communicative functions of prosody in PD, such as lexical stress, i.e. noticing stress on a syllable in a word [19–22], boundary marking, i.e. noticing syntactic breaks in a sentence as conveyed through the voice [21, 24], focus, i.e. noticing stress on a word in a sentence as conveyed through the voice [21, 24], and sentence mode, i.e. noticing the statement/question contrast as conveyed through the voice [19, 21–25]. The aforementioned studies predominantly used discrimination tasks to assess perception of prosodic forms, and/or identification tasks to assess comprehension of prosodic meanings, but none of them systematically investigated both perception and comprehension of all communicative functions of prosody discussed so far. Perception is reported to be intact across prosodic functions, while for comprehension, there are reports of deficits for lexical stress [19] and sentence mode [21, 22], but not for the other prosodic functions.
The present study is the first one to systematically assess both perception and comprehension of all five aforementioned communicative functions of prosody in individuals with PD, and to investigate the influence of age on reception of prosody using an assessment that incorporates various communicative functions of prosody. As such, this study aims to answer the following research questions: (1) what is the difference in performance on the receptive prosody assessment between unimpaired individuals and individuals with hypokinetic dysarthria due to PD; (2) what is the difference between performance on the perception subtest and performance on the comprehension subtest within both groups; (3) what is the relationship between age and performance within both groups; and (4) what is the effect of subtest sequence on performance within both groups?
MATERIALS AND METHODS
Participants
The recruitment and assessment of participants was supervised by a certified and experienced speech-language pathologist, and assisted by two master’s level speech-language pathology students who had received comprehensive instructions and training regarding the entire procedure prior to the study. Two groups of participants were recruited: a group of 22 individuals with dysarthria due to PD (from here on referred to as the PD group) and a healthy control group of 22 individuals with no history of dysarthria (from here on referred to as the HC group). Upon entering the study, all participants signed an informed consent form (Belgian registration number B300201112083), approved by the Ethics Committees of the St-Rembert Hospital (Torhout, Belgium), that recruited the individuals with PD, and the Antwerp University Hospital (Edegem, Belgium), that coordinated the study. All but two participants of the PD group were assessed in a quiet office room at the recruiting hospital, whereas the other two PD group participants and all of the HC group participants were assessed in a quiet room at their homes.
The two groups were precision matched for gender and age, thus having a comparable male/female distribution (16 males and 6 females), age range (50 to 83 years) and mean age (HC group: M = 66.18 years, SD = 9.69 years; PD group: M = 66.41 years, SD = 9.35 years). We included participants whose cognition was well within normal limits (i.e. scoring 27 or above out of 30) as assessed by a Dutch version of the Mini-Mental State Examination [26], called S-MMSE (HC group: M = 29.18, SD = 0.96; PD group: M = 28.77, SD = 1.02). We also included participants who did not exhibit signs of depression (i.e. scoring 14 or below out of 63) as evaluated using a Dutch version of the Beck Depression Inventory [27], called BDI-II-NL (HC group: M = 4.14, SD = 3.31; PD group: M = 7.36, SD = 4.48). All participants spoke Dutch as their mother tongue and had sufficient vision and reading skills as established by clinical judgement. Auditory skills of the HC group were deemed sufficient by clinical judgement, while auditory skills of the PD group were formally tested (see next paragraph).
Concerning the PD group, all participants were recruited at the same hospital (see Table 1 for individual participant information). They had all been officially diagnosed with idiopathic PD (years post-diagnosis: M = 4.09, SD = 2.35) and were confirmed to have no other neurological problems by the recruiting hospital’s attending neurologist. They showed mild to moderate motor disturbances according to the criteria of the motor examination part of the Unified Parkinson’s Disease Rating Scale or UPDRS-III (M = 16.77, SD = 6.24) and the modified Hoehn and Yahr scale (M = 1.95, SD = 0.49). All of them took antiparkinson medication (Amantan, Azilect, Mirapexin, Prolopa, Requip, Sinemet, Stalevo or combinations) and were in the on-stage of their medication cycle at the time of the present study. To ensure sufficient auditory skills, PD group participants underwent a pure-tone audiometric evaluation administered by an otorhinolaryngologist affiliated to the recruiting hospital. Participants were included when a threshold below 25 dB HL could be demonstrated in at least one ear by means of the Indice de Perte Auditive or IPA (M = 12.13 dB HL, SD = 6.98 dB HL). This index takes into account hearing loss at three frequencies critical for speech reception, according to the formula (500 Hz + (2 × 1000 Hz) + 2000 Hz)/4. Unfortunately, no audiogram could be obtained for participants 14 and 16, who had to be tested at home, but as both of them demonstrated adequate hearing skills in the course of the study, they were nonetheless included in the PD group. Dysarthria severity (M = 1.23, SD = 0.43) was judged by the master students during an elicited conversation prior to the actual prosodic assessment on a four-point scale (0 = no dysarthria, 1 = mild dysarthria, 2 = moderate dysarthria, 3 = severe dysarthria). Prior to this evaluation, the students had trained themselves in judging severity of 51 dysarthric speech samples from a database that had been collected for previous research and judged by the first author of the present study [28]. An inter-rater reliability analysis using a two-way random single measures intraclass correlation coefficient (ICC = 0.76; p < 0.001) revealed sufficient consistency between the (joint) evaluation of the students and the evaluation of the first author to ensure a reliable judgement by the students. At the time of this study, only two out of the 22 individuals with PD (participants 1 and 16, see Table 1) received speech therapy that included training of prosodic skills.
Materials
Taking into account the particulars of Dutch prosody as explained in [29], we designed a test battery for assessing reception of five important communicative functions of prosody, which are listed and exemplified in Table 2. The assessment consisted of two subtests: a perception subtest and a comprehension subtest. Perception of prosody was assessed by means of a discrimination task, which involved listening to a pair of model audio samples and deciding whether the two samples sounded similar or different. Comprehension of prosody was assessed by means of an identification task, which involved listening to a model audio sample and labelling the prosodic meaning expressed by the model speaker. Testing both perception and comprehension makes it possible to detect whether a deficit, if any, can be situated on the lower-order auditory processing level or on the higher-order cognitive level respectively [30]. In both subtests, all five prosodic functions were tested by means of six test items, resulting in a maximal raw score of 30 for both subtests and thus a maximal raw score of 60 for the entire test. As can be seen from the examples of test items in Table 2, all 60 test items were a constituent element of a prosodic minimal set.
Model audio samples, all of them sentences, were recorded by one female voice, namely of the first author (who is a speech and language therapist and linguist). The recordings took place in a quiet office room with an AKG (C555 L) headset microphone connected to a Dell Vostro laptop computer, an external sound card (E-MU 0404 USB), and Audacity software (freely available, sampling rate 44.1 kHz, 24 bit, mono) [31], which was also used to save each audio sample as a separate wav-file. In order to standardize sample presentation throughout the perception subtest and to spare participants’ auditory memory, we inserted a short silent interval of approximately one second between the two constituent audio samples of each sample pair and then saved each pair as one wav-file.
During the perception subtest, assessment of the five prosodic functions was similar throughout all functions. Participants were first presented with two printed labels (‘similar’ or ‘different’). They listened to six utterance pairs per prosodic function, either prosodically identical (e.g. ‘Hij heeft een UITstekende neus. Hij heeft een UITstekende neus.’) or prosodically different (e.g. ‘Hij heeft een UITstekende neus. Hij heeft een uitSTEkende neus.’), and decided for each pair whether the paired utterances sounded similar or different, by indicating or phrasing the appropriate label. For reasons of balance, each function was tested by means of three identical pairs and three different pairs (presented in a random order).
During the comprehension subtest, assessment of the five prosodic functions was different for each function, as explained in the rest of this section. The ‘Lexical Stress’ function refers to the reception of homophones. To test this ability, we used model samplesof six sentences containing homophones, which sound phonetically identical but differ in meaning due to a different word stress pattern. In Dutch, this phenomenon occurs in word pairs (Lexical stress A in Table 2) such as ‘UITstekend’ (‘protruding’) versus ‘uitSTEkend’ (‘excellent’), and in noun compound and noun phrase pairs (Lexical stress B in Table 2) such as ‘VIERhoeken’ (‘rectangles’) versus ‘vier HOEken’ (‘four angles’). During the identification task, participants were shown a multiple-choice array of three numbered pictures for each utterance, a procedure comparable to the one used in [22]: the target response (e.g. picture of a professional wine-taster smelling wine, corresponding to ‘uitSTEkende neus’), the response for the other minimal pair member (e.g. picture of a man with a prominently protruding nose, corresponding to ‘UITstekende neus’) and a foil (e.g. picture of a man sticking out his tongue, corresponding to ‘UITstekende tong’). Participants listened to the utterances containing one homophone and were asked to identify the picture corresponding to the auditory stimulus by pointing to it or naming its number.
The ‘Boundary Marking’ function refers to the reception of syntactic boundaries as conveyed through the voice. To test this ability, we used model samples of three complete and three incomplete sentences. The complete sentences ended in a final syntactic boundary and a phrase-final intonation fall indicating that the speaker had finished the sentence. The incomplete sentences ended in a nonfinal syntactic boundary and a levelling or slightly rising intonation pattern indicating that the speaker had not yet finished the sentence and might continue to finish it off by adding a list of items or a clause. During the identification task, participants were presented with two printed labels (‘finished’ or ‘not finished’), listened to the six utterances, and each time decided whether or not the speaker had finished his sentence by pointing out or phrasing the label corresponding to the auditory stimulus.
The ‘Focus’ function refers to the reception of highlighted information in an utterance as conveyed through the voice. To test this ability, we used model samples of six sentences, each of which containing three words that could potentially be highlighted through sentence stress, but only one particular word that was actually highlighted. During the identification task, participants were presented with a printed multiple-choice array of the three potentially highlighted words for each sentence. Participants listened to each utterance and then were required to indicate or phrase the word which was most highlighted by the model speaker.
The ‘Sentence Mode’ function refers to the reception of sentence mode as conveyed through the voice. To test this ability, we used samples of three statements and three declarative questions. Declarative questions (e.g. ‘Karen plays tennis?’) syntactically look like statements, but prosodically sound like questions due to a phrase-final intonation rise. During the identification task, participants were presented with two printed labels (‘statement’ or ‘question’), they listened to each utterance, and then decided whether they heard a statement or a question by indicating or phrasing the appropriate label.
Finally, the ‘Emotional Prosody’ function refers to the reception of different emotional states conveyed through the voice, such as anger, happiness, sadness, and neutrality. To test this ability, we used six samples of one sentence that was uttered in a number of different emotional states. During the identification task, participants were presented with a printed multiple-choice array of four numbered labels and corresponding smileys, they listened to each utterance and were asked to identify the conveyed emotional state by pointing out the smiley or phrasing the label corresponding to the auditory stimulus.
Procedure
As the flow chart in Fig. 1 illustrates, a cross-over design was used to determine whether the perception subtest, if administered first, might affect performance on the comprehension subtest, or vice versa. Participants were non-randomly assigned to the perception-comprehension (PC) or the comprehension-perception (CP) sequence groups (see Table 1). During the assessment, the 60 pre-recorded model audio samples were presented at a comfortable loudness level through Hercules XPS 2.0 35 USB speakers connected to a Lenovo Thinkpad Edge E320 NWY3RGE notebook, or alternatively through Puro Music Ball speakers connected to a Packard Bell EasyNote TS11HR notebook, depending on the master student administering the assessment. No practice items were included in the assessment procedure, but participants received clear and concise standardized instructions before each task. Participants listened to the samples and responded by indicating or phrasing the appropriate picture or printed label as explained in the Materials section. By default, auditory stimuli were presented only once, but a sample replay was granted whenever participants wished. The entire assessment procedure typically lasted about 20 minutes. Afterwards, raw scores were calculated by marking each response as correct (1 point) or incorrect (0 points) and adding up the number of correct responses, and were then expressed as percentages. This was done for each individual prosodic function (function scores), for both subtests (subtest scores), and for the test as a whole (global test score).
Data analysis
Data were statistically analysed using SPSS software version 20. Non-parametric tests were chosen due to the relatively small number of participants and the non-normal distribution of the data. An alpha level of .05 was used for all tests. In order to compare HC group versus PD group scores, a one-tailed Mann-Whitney U test was used, and an additional one-tailed Wilcoxon signed-rank test was used for a paired comparison. For the comparison of subtest scores within groups, a two-tailed Wilcoxon signed-rank test was performed. The relationship between age and global test score was assessed with a Spearman rank-order correlation coefficient. Finally, a two-way ANOVA was conducted to examine the effect of subtest sequence (perception-comprehension versus comprehension-perception) and group (HC group versus PD group) on the global test score.
RESULTS
When comparing scores of the HC group versus the PD group (see Table 3), no significant differences could be demonstrated for the global test score, the subtest scores, or the function scores, except for the Focus function during the Comprehension subtest. An additional paired test analysis (taking into account the fact that the HC and PD groups were precision matched for gender and age) did not reveal any significant differences.
When comparing subtest scores within groups (see Table 4), both groups scored significantly higher on the perception subtest compared to the comprehension subtest. In the HC group, perception of the Lexical Stress and Focus functions was significantly better than comprehension, whereas for the other prosodic functions, no significant differences could be demonstrated. In the PD group, perception of all prosodic functions except the Boundary Marking function was significantly better than comprehension.
There was a moderate, negative correlation between age and global test score, which was statistically significant in the HC group (r s = –0.44, p = 0.043) and the PD group (r s = –0.49, p = 0.022).
Concerning the influence of subtest sequence on the global test score, there were no significant main effects for subtest sequence (F(1, 40) = 0.44, p = 0.511) or group (F(1, 40) = 0.60, p = 0.443), and there was no significant interaction effect between subtest sequence and group (F(1, 40) = 0.17, p = 0.687).
DISCUSSION
The present study was the first one to investigate both perception and comprehension of five crucial communicative functions of Dutch prosody in individuals with hypokinetic dysarthria due to PD, and to investigate the effect of age on reception of prosody in this population using an assessment that includes a variety of communicative functions of prosody. The study nuances the literature concerning reception of prosody in PD, and supports the literature concerning the role of age in reception of prosody in general.
A first finding was that virtually no difference in receptive prosodic abilities could be shown between the HC group and the PD group. This may result from the absence of a robust receptive prosody deficit in individuals with PD, as has previously been argued in [20] and [23]. Alternatively, this may result from the fairly strict inclusionary criteria for cognition, depression, and hearing, and the test characteristics, all of which will be discussed next.
Cognition has already been associated with a deficit in the reception of prosody in PD [32, 33]. In their review of the literature on reception of speech by individuals with PD, Kwan and Whitehill [17] concluded that a cognitive deficit (and more specifically impaired working memory) may be involved in the decreased accuracy of reception of emotional prosody in PD. In the present study, only participants with an MMSE score of minimally 27 were included, which was stricter than in comparable studies mentioning MMSE cut-off scores of 23 [21], 24 [22], or 25 [20, 25]. As speech perception tasks involve some degree of executive control and working memory capacity [34], the high MMSE score of all participants involved in the present study may help to explain why no differences in global test scores, subtest scores, or function scores were found between the HC group and the PD group. We should, however, bear in mind that a single screening test such as the MMSE does not warrant a fully-fledged picture of intactness of executive control and more specifically working memory.
Depression, a substantially prevalent condition in PD [35], is associated with a deficit in the reception of emotional prosody [36, 37]. A meta-analysis by Gray and Tickle-Degnen [16], however, revealed that comorbid depression in PD is not likely to cause a deficit in the reception of emotional prosody. Unfortunately, no literature could be found concerning the impact of depression on reception of other types of prosody in unimpaired individuals or individuals with PD. Further study is required to confirm our hypothesis that inclusion of participants with signs of depression would not have altered the results significantly.
Whereas cognition and depression were screened using an identical protocol for all participants, this was not the case for hearing, which was formally screened in 20 out of the 22 participants with PD, and clinically judged in all other participants. As a result, we cannot be sure whether hearing thresholds of the HC and PD groups were fully comparable. However, perception of prosodic speech characteristics seems to remain relatively robust in mild to moderate peripheral hearing loss [10] and even in severe sensorineural hearing loss [38]. Therefore, we are inclined to believe that a potential discrepancy in hearing capacities of both participant groups was not a decisive factor in the obtained results.
The test characteristics of the receptive prosody assessment used in this study constitute a final potentially confounding factor. First, as the assessment was designed for use in clinical practice, it contained a limited number of test items (6 per prosodic function; 30 per subtest; 60 in total) in order to ensure a reasonable administration time. As a result, the assessment may not have been sensitive enough to detect mild impairments on the detailed level of the different prosodic functions (compare for instance [21]). However, although the number of test items for the two subtests and for the entire test may have been sufficient to detect impairment of receptive prosody on a more global level, still no differences could be found between both participant groups. Second, the present assessment predominantly covered linguistic prosody (four out of the five assessed prosodic functions), whereas emotional prosody represented a minor part of the assessment (one out of the five assessed prosodic functions). A possible explanation for the absence of differences between both participant groups is that reception of linguistic prosody may be more intact than reception of emotional prosody in PD, and that the assessment, covering in majority linguistic prosody, may thus have failed to expose any deficit. This resonates with the numerical abundance of reports of a deficit in reception of emotional prosody versus the rather limited number of reports of a deficit in reception of linguistic prosody in PD.
A second finding, with regard to subtest scores, was that prosody comprehension (as assessed through an identification task) appeared more difficult than prosody perception (as assessed through a discrimination task) for both the HC group and the PD group. This result seems in agreement with Witteman, van IJzendoorn, van de Velde, van Heuven, and Schiller [39], who advanced that discrimination tasks tap the early stages of prosody reception, in which the speech signal is acoustically analysed and the emotional or linguistic information is identified, whereas identification tasks additionally tap the later stages, in which the information is made available to higher-order cognitive processes. A look at the function scores further reveals that PD participants experienced greater trouble in identifying than in discriminating prosody in four out of the five assessed functions. This result seems to corroborate in a more general sense Gray and Tickle-Degnen’s [16] conclusion specifically with regard to emotional prosody that identification tasks yield a greater deficit than discrimination tasks.
A third finding was that an older age was moderately associated with a lower global test score, regardless of group affiliation. This association is consistent with previous reports about the negative effect of ageing on the reception of emotional prosody, independently of age-related factors such as hearing loss and cognitive decline [8–11]. Moreover, considering the variety of communicative functions of prosody covered by the present study, the result indicates that reception of prosody at large may be affected by ageing, which is in keeping with reports that also aspects of linguistic prosody are affected in older adults [12–14]. Based on these insights, we suggest that clinicians should not presume intactness of receptive prosodic abilities in older adults, even if those adults present with apparently adequate hearing abilities and cognition.
Finally, subtest sequence apparently did not influence global test scores. This finding ran counter to our speculation that taking the perception subtest first might enhance sensitivity to prosodic differences, and thus possibly lead to a better performance during the subsequent comprehension subtest than would be the case in the reverse sequence. It could be concluded that the order in which discrimination and identification tasks are administered may be irrelevant.
There were some limitations to the present study. A first limitation was that hearing capacity of the PD group participants was formally tested through pure-tone audiometry, while hearing capacity of the HC group participants was merely evaluated through clinical judgement. Consequently, we cannot be sure that hearing capacity of both groups were fully comparable.
Another weakness is the fact that only one (female) model voice was used for the model samples. With regard to the vocal expression of emotion, acoustic cues such as intonation contour type and F0 range may be speaker-independent [40], but as there can nevertheless be important inter-speaker differences in the vocal encoding of emotion [41], it may have been preferable to provide more than one model voice to assess reception of emotional prosody, for instance both a male and a female voice [13].
A final limitation is related to the fact that the assessment was not validated prior to its use during the present study. For instance, face validity of the recorded test items, in terms of authenticity and recognizability of the intended prosodic meanings expressed by the model speaker, was not checked with an independent group of unimpaired listeners before the actual study took place (compare e.g. [13, 41]). A look at the medians and interquartile ranges in Table 3 reveals that (even in the absence of training items) a significant proportion of the participants of both groups managed to obtain the maximum score for various communicative functions of prosody in both subtests, or, in other words, the assessment seems to suffer from a ceiling effect. Taken together with the fact that non-parametric tests had to be performed due to the relatively small sample size and the non-normal distribution of the data, this may have led to a lack of statistical power to detect differences between the healthy control group and the Parkinson’s disease group.
Future research should be directed towards enhancing representativeness of the PD population by assessing individuals with and without intact cognition, with and without signs of depression, and with varying degrees of dysarthria. Apart from this, test item validity and test reliability need to be investigated. An evaluation of the number of test items per prosodic function is warranted in order to achieve an acceptable balance between reasonable administration time on the one hand and sufficient sensitivity on the other hand. To further eliminate the ceiling effect, test item difficulty level may be adjusted by introducing samples with more subtle prosody during the comprehension subtest and sample pairs with more subtle prosodic differences during the perception subtest.
In sum, considering that the older speakers with hypokinetic dysarthria due to Parkinson’s disease had receptive prosodic skills inferior to those of the younger speakers, notwithstanding apparently intact cognition and hearing, the findings suggest that age is a factor to be reckoned with in prosody assessment and management in this population.
CONFLICT OF INTEREST
The authors have no conflict of interest to report.
Footnotes
ACKNOWLEDGMENTS
This research is part of a larger project called ‘Computerized Assessment and Treatment of Rate, Intonation, and Stress’, which ran from 2009 until 2013 and was supported by fund TBM-080662 from the Flemish Agency for Innovation by Science and Technology (IWT), awarded to Marc De Bodt, Gwen Van Nuffelen, and Heidi Martens. We thank all participants who made this study possible. We are also indebted to Lotte De Vos en Karen Godderis-Coene (graduating speech and language therapists at the time) for their contribution to the recruiting and testing of the participants. We thank Charlotte Libbrecht (speech and language therapist), Jen Maes (neurologist), and Mieke Naessens (otorhinolaryngologist), all affiliated with the St-Rembert Hospital (Torhout, Belgium) for their help in recruiting the participants with Parkinson’s disease and in providing access to the participants’ medical records.