Influence of Oral Morphology on Speech Production

Original Paper Folia Phoniatr Logop 2018;70:138–148 DOI: 10.1159/000491789

Published online: August 23, 2018

Influence of Oral Morphology on Speech Production in Subjects Wearing Maxillary Removable Partial Dentures with Major Connectors Junichiro Wada a Masayuki Hideshima b Shusuke Inukai a Azusa Katsuki a Hiroshi Matsuura c Noriyuki Wakabayashi a

a Section

of Removable Partial Prosthodontics, Department of Masticatory Function Rehabilitation, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University (TMDU), Tokyo, Japan; b Dental Clinic for Sleep Disorders (Apnea and Snoring), Oral and Maxillofacial Rehabilitation, University Hospital of Dentistry, Tokyo Medical and Dental University (TMDU), Tokyo, Japan; c Graduate School of Management and Information of Innovation, University of Shizuoka, Shizuoka, Japan

Abstract Background/Aims: Speech impairment during the initial phase of removable partial denture (RPD) treatment can prevent patient adaptation to RPDs. This study was undertaken to investigate the influence of oral morphology on speech production in subjects wearing RPDs with major connectors. Methods: Two types of connectors were fabricated for 17 subjects with normal dentitions: covering the middle palate (M-bar) and the anterior/posterior palate (AP-bar). Four target sounds ([∫i], [t∫i], [çi], and [ki]) were evaluated under 3 recording conditions: no connector, M-bar, and AP-bar. The mean appearance ratios of correct labels (MARCs) were calculated as parameters representing speech production accuracy with the speech evaluation system. Subgroup analysis was conducted based on palate height, dental arch width, and front space volume of the oral cavity. Results: Based on

© 2018 S. Karger AG, Basel E-Mail [email protected] www.karger.com/fpl

the multiple linear regression test, a significant association was found between the MARCs of [∫i] with M-bar and front space (p = 0.036). In the subgroup analysis, the AP-bar had a significant effect on the MARCs of [∫i] among subjects with high palate (p = 0.026), narrow arch (p = 0.004), and small front space (p = 0.014). Conclusion: RPDs with major connectors could disturb speech production among patients with high palates, narrow arches, and small front spaces. © 2018 S. Karger AG, Basel

Introduction

Improving speech production is an important goal of prosthetic dentistry in partially edentulous patients. A removable partial denture (RPD) is an appropriate treatment to restore missing teeth for partially edentulous patients whose remaining teeth have been weakened by severe tooth wear or periodontitis [1–3]. Recently, the combination of RPDs and implants has become a common method for restoring missing teeth with heavy reJunichiro Wada Department of Masticatory Function Rehabilitation, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University (TMDU) 1-5-45 Yushima, Bunkyo-ku, 113–8549 Tokyo (Japan) E-Mail wadajun.rpro @ tmd.ac.jp

Downloaded by: W. Junichiro - 239864 Tokyo Ikashika Daigaku 202.242.187.110 - 8/27/2018 1:37:54 AM

Keywords Dental consonant misarticulation · Morphology · Speech production · Removable partial denture

sorption of partially edentulous ridges [4–6], and for recovery after failure of implant-supported fixed prostheses [7], due to the adaptability of RPDs with respect to changes in oral conditions [8, 9]. However, it has been reported that RPDs can disturb speech production, in comparison to other prostheses [10, 11]. Speech impairment significantly affects the quality of life of RPD wearers [12]. Many studies have investigated the influence of RPD design on speech production [13–15]. According to these reports, RPDs designed to cover a large area of the palate, especially the anterior palate, tend to disturb speech production; thus, a palatal bar covering the middle area of the palate is recommended to serve as a major connector of RPDs. Sound spectrography evaluates the influence of removable dentures on speech production by analyzing the energy spectrum and formant frequency of recorded acoustic data [16, 17]. Palatography, which evaluates the contact patterns of the tongue on the palate during pronunciation, has also been used as a standard method [18– 22]. Although these methods have demonstrated effectiveness in evaluating structural and physiological mechanisms, they are limited to the assessment of a single aspect of speech production. More specifically, these methods cannot directly evaluate how listeners recognize the sound produced by the patient. Auditory evaluation by speech pathologists can directly assess the produced sounds. However, this involves limitations in terms of reliability and reproducibility [23, 24]. We previously developed an evaluation system with incorporated speech recognition. This system could specifically and objectively evaluate the accuracy of speech production [25] and report the influence of RPD designs on speech production [15, 26–28]. Nevertheless, speech impairment remains an intractable problem in clinical practice, as RPD design is influenced by the condition of the remaining teeth and alveolar ridge, rather than the phonetic function of patients. Frank et al. [29] reported that design/fabrication standards were not related to patient satisfaction, including speech production. Clinically, RPDs with similar designs could disturb patients’ speech production by varying degrees; thus, speech impairment after RPD insertion is dependent on both RPD design and patient factors. Although the relationship between speech production and deformation of the oral cavity in patients after tumor excision has been evaluated with and without prostheses, few studies have investigated the relationship between RPDs and speech production, focusing on morphological characteristics of the patients’ oral cavities [30–32]. Brun-

ner et al. [33, 34] reported that the shape of the palate affects acoustic variability. Tanaka [35] reported that palate morphology affected the speech patterns of complete denture (CD) wearers. The shapes of the palate and dental arch are important factors in RPD design, especially when choosing the major connector. Predicting the influence of the major connector on speech production, specifically on the basis of oral morphology, would substantially facilitate RPD design. With regard to long-term adaptation, Hamlet and Stone [36] reported that successful adaptation to speech production occurred in one-half of subjects at 2 weeks after placement of the palatal plate. Knipfer et al. [37] reported that disorders of speech production with CDs disappeared 6 months after placement. Although those studies indicated that adaptation for a certain period after placement of RPD could help patients to overcome disorders of speech production, speech disorders that occur soon after RPD insertion may prevent patients from continuing to use their new RPDs before sufficient adaptation to wearing their RPDs [38]. Additionally, for clinicians, the ability to predict whether patients will experience speech problems, and to determine the severity of those potential problems, will facilitate decisions regarding whether particular patients must be monitored and referred for speech therapy. We previously evaluated the influence of placement area and cross-sectional shape of major connectors on speech production [15, 28]. In this study, we inserted experimental major connectors into subjects with various palatal shapes and dental arch morphologies. The accuracy of phonetic output and changes in Japanese consonants were evaluated by a speech evaluation system. Our goal was to investigate the influence of oral morphology on speech production in subjects wearing major connectors in the initial phase of RPD treatment. The null hypothesis was that oral morphology would not influence speech production in subjects wearing major connectors.

Influence of Oral Morphology on Speech with Partial Dentures

Folia Phoniatr Logop 2018;70:138–148 DOI: 10.1159/000491789

Materials and Methods

139


Subjects Adult subjects were recruited from a pool of patients and staff members at Tokyo Medical and Dental University Dental Hospital (Tokyo, Japan) in 2016–2017. Only volunteers who spoke standard Japanese and had class I occlusion were included in this study. Exclusion criteria comprised the following: (1) suspicion of speech impairment (either reported by the participants or based on the clinical impression of the experimenter [J.W.] at the time of intake), (2) acute symptoms of oral and/or craniofacial diseases, (3) current dental treatment, and (4) missing teeth, except third mo-

10 mm 10 mm

10 mm 0.8 mm

a

0.8 mm

b

0.8 mm

Fig. 1. Experimental major connectors. Photograph depicting experimental major connectors covering the middle area of the palate (M-bar) (a) and both anterior and posterior areas of the palate (AP-bar) (b). Schematic illustrations represent cross-sectional views of the connectors.

Design of the Experimental Procedure As shown in Figure 1, the experimental major connectors were fabricated from cobalt-chromium (Cobaltan; Shofu Co., Kyoto, Japan) by using casts made from silicone impressions of the subjects’ maxillae. Two different designs were fabricated: a connector covering the middle area of the palate (M-bar) and a connector covering both the anterior and posterior areas (AP-bar). The connectors were fabricated with a maximum thickness of 0.8 mm and a width of 10 mm at the narrowest part. The cross-sectional shape of the connecter in the sagittal plane was moderately curvilinear. Recording and Evaluation of Speech Data In speech, consonants whose production requires contact between tongue, palate, and alveolus were reported to be affected by the placement of dentures [13–17, 20, 22, 26–28, 39, 40]. In addition, disorders of pronunciation have been reported to appear readily on the second mora [39, 40]. Accordingly, the second mo-

140


rae (/shi/: [∫i], /chi/: [t∫i], /hi/: [çi], and /ki/: [ki]) of 4 test words (/i/shi/kawa/, /i/chi/ro/, /e/hi/meken/, and /o/ki/nawa/) were chosen as the target sounds (Table 1). These target sounds can be classified into 3 groups on the basis of articulation position: (1) alveolar: [∫i] and [t∫i]; (2) palatal: [çi]; and (3) velar: [ki]. The sounds can also be classified into 3 groups on the basis of the manner of articulation: (1) fricative: [∫i] and [çi]; (2) affricate: [t∫i]; and (3) plosive: [ki]. The target sounds included the vowel [i], which requires the highest tongue position among Japanese vowels. These test words are meaningful Japanese words, which have both target sounds and accents on the second morae. Speech data of the test words were recorded under the following conditions: without a connector (control); with M-bar; and with AP-bar. Each subject was required to pronounce a test word 5 times under the above-mentioned conditions. The order of the recording conditions was randomly arranged. Each subject was instructed not to vocalize the test words before recording, and the recordings were initiated immediately after the connectors were stably inserted. During the measurements, the subjects were asked to maintain their pronunciation at a constant speed and volume, as in their daily life. Speech data were recorded with a headset-type microphone and a laptop computer (PAC9214LDEW; Toshiba Co., Tokyo, Japan) with a speech evaluation system. Subjects were given rests after each recording and were not allowed to speak during the breaks.

Wada/Hideshima/Inukai/Katsuki/ Matsuura/Wakabayashi


lars. Subjects received written information regarding the study purpose and protocol before they consented to participation. A total of 17 subjects (7 female, 10 male; mean age, 35.0 years; range, 21–58 years) were eligible to participate. All experimental procedures upheld the principles of the 1975/1983 Helsinki declaration and were approved by the Ethics Committee of the University (Authorization No. 286).

23 ms

Speech waveform of test word

Analytical frames 79 ms

8 ms

Power spectrum FFT 64 ms

8 ms MAFP 48 ms

Reference patterns

GG

Segment label 8 ms

a

$$

$I

## Segment label

$I

b

Fig. 2. Flow process of the speech evaluation system. a Acoustic analysis was performed using a fast Fourier transform (FFT) that transformed the speech data of 8 analytical frames (79 ms) into 64 ms of information (FFT parameters). Then, the multiple acoustic feature plane (MAFP) pattern, emphasizing the local changes in frequency and time frame, was extracted. Finally, 6 MAFP patterns

Pattern matching

(48 ms) were quantized with statistical pattern-matching processing and the integrated phonetic segment was labeled every 8 ms. b The pattern is extracted from the power spectrum every 8 ms. The extracted pattern is matched to a phonetic segment in the reference patterns and labeled as a combination of 2 letters.

Table 1. Target sounds and test words

Test words

IPA

sound

articulation point

articulation manner

correct label(s)

word

meaning

[∫i] [t∫i] [çi] [ki]

/shi/ /chi/ /hi/ /ki/

alveolar alveolar palatal velar

fricative affricate fricative plosive

$$, SS, $I CC, CI ##, HH, HI KI

i/shi/kawa i/chi/ro e/hi/meken o/ki/nawa

name of a Japanese golf player name of a Japanese baseball player one of the prefectures in Japan one of the prefectures in Japan

The speech evaluation system (Voice Analyzer; Toshiba Digital Media Engineering Co., Tokyo, Japan) used speech recognition based on 213 types of integrated phonetic segments that represent phonetic features of transitions from one phoneme to another [27]. The flow process of this system is shown in Figure 2. The acoustic analysis, which had an analytical frame of 23 ms and a frame shift range of 8 ms, was performed by using a fast Fourier transform) that transformed the speech data of 8 analytical frames (79 ms) into 64 ms of information (fast Fourier transform parameters). Then, multiple acoustic feature plane patterns, emphasizing local changes in frequency and time frame, were extracted. Finally, the 6 multiple

acoustic feature plane patterns (48 ms) were quantized with statistical pattern-matching processing and the integrated phonetic segment was labeled every 8 ms. Thus, this system performed microperiod (8 ms) evaluations on a frame-by-frame basis and was able to record time-scale information (Fig. 2a). Integrated phonetic segment labels which indicated the correct pronunciation of the target part of the target sound were defined as “correct labels” (Table 1). The number of correct labels varied with the type of test sound. This system also performed macro-period quantitative sound evaluations by using the ratio between the numbers of correct and incorrect integrated phonetic segment labels during the evaluation pe-



141


Target sounds

I I I I I $ C $ $ $ $ C C $ C I I I I I I I K K K K K K K A A A A A A A WW S S S S S $ C $ I $ $ C C $ Y 4 4 Q Q Q 4 Q A A A A A A A A A A W O WW A A

I

I

I

$

C

$

$

$

$

C

C

$

C

I

I

I

S

S

S

$

C

$

I

$

$

C

C

$

Y

4

4

Q

Transition from [i] to [∫i]

Consonant part of [∫i] (Target part) MARC = 60%

Vowel part of [∫i]

Fig. 3. Method of calculating the mean appearance ratio of correct labels (MARC) (sample [∫i]). In this record-

ing, there are 10 segment labels in the target part. There are 6 correct labels of [∫i] ($$ and $I) and 4 incorrect labels (CC and CY). Thus, the MARC of [∫i] is calculated to be 60%.

142


target sounds were calculated. The mean appearance ratio of correct labels (MARC) was regarded as a representative value reflecting the accuracy of phonetic output of the target sounds. The mean appearance ratio of incorrect labels (MARIC) was also calculated for statistical analysis. For other target sounds, calculations of MARCs and MARICs were performed in the same manner. Morphometry of Oral Cavity and Subgrouping Maxillary stone models of the subjects were digitized with a three-dimensional (3D) scanner (ARCTICA Scan; KaVo, Biberach, Germany). The digital data were saved in a surface tessellation language (STL) format. Three parameters – representing the shape of the palate and dental arch; palate heights, narrowing of dental arches, and front spaces of the oral cavities – were then computed for each STL datum by using 3D evaluation software (Geomagic Studio 2014; 3D Systems, Rock Hill, SC, USA). Calculation methods for these parameters are shown in Figure 4. The palate height was defined as the distance from the midpoint between the mesiopalatal cusps of both first molars to the palate (Fig. 4a). The ratio

Wada/Hideshima/Inukai/Katsuki/ Matsuura/Wakabayashi


riod. Matsuura et al. [25] conducted a recall test to verify the reliability of this system with sound subjects. The recall and precision averages were 95.4 and 95.9%, respectively. The F-measure was 95.7, which indicated that the speech evaluation system was a reliable method [41]. The results of our studies with this system were consistent with the results of previous studies with conventional methods, such as sound spectrography, palatography, and auditory evaluations by speech pathologists [15–28]. As shown in Figure 2b, the pattern was extracted from the power spectrum and matched to a phonetic segment of the reference pattern. These results were labeled and displayed. As an example, the method to extract the consonant part of the target sound [∫i] from the speech data of the test word /i/shi/kawa/ is shown in Figure 3. The head border of the range of the [∫i] sound was the transition from [i] to [∫i], and the foot was the part containing the transition from [∫i] to [ka]. The vowel part of [∫i] was then removed, and the remaining section was defined as the target part. Extractions of other target sounds were performed in the same manner. The ratios of correct labels to all labels appearing in the consonant parts of the

CR DIP CL

WMD

Palate height

a

b

Reference plane (RP)

RP

PI

PI PR

BP PL DP

c

d

Fig. 4. Calculation of the 3 morphological parameters. a Palate height was calculated as the length of the line (perpendicular to the reference plane [RP] from the midpoint between the mesiopalatal cusps of the left (CL) and right first molars (CR) to the palate. b Narrowing of the dental arch was calculated as the ratio of the interpremolar distance (DIP) to the mesiodistal width of the dental arch (WMD). c The RP included the incisive papilla (PI) as well as the palatal interdental papillae between the first and second molar

of the left side (PL) and right side (PR). d The front space was calculated as the ratio of front volume to total volume. The total volume was defined as the space enclosed by the palate, RP, and distal plane (DP; the perpendicular plane to RP including the distal ends of the left and right second molars). The bounding plane (BP) between the front and rear spaces included the midpoint of the perpendicular line from PI to DP.

of the interpremolar distance to the mesiodistal width of the dental arch was defined as narrowing of the dental arch (Fig. 4b). The total volume of the oral cavity was divided into front volume and rear volume. The ratio of front volume to total volume was defined as the front space (Fig. 4c, d). The study group was divided into 2 subgroups for each parameter (Table 2). The boundary between the 2 subgroups was the median of all subjects.

performed for each subgroup by using the Friedman test and pairwise comparisons. A multiple linear regression test was applied to evaluate the association of the 3 morphometric parameters to MARCs with each recording condition. For the comparison of MARICs, the Mann-Whitney U test was performed among subgroups for the 3 parameters (SPSS version 22.0; SPSS Japan Inc., Tokyo, Japan); p < 0.05 after Bonferroni correction was considered to be statistically significant.

Statistical Analysis The Kolmogorov-Smirnov test was used for normality analysis. Due to the absence of normality of some data, such as MARCs of [∫i] and [ki], nonparametric statistical analyses were conducted. Spearman’s rank correlation with Bonferroni correction was performed to evaluate associations between morphometric parameters. For each parameter, comparisons of MARCs among the 3 recording conditions (no connector, M-bar, and AP-bar) were

Correlations between Morphometric Parameters Front space was significantly correlated with palate height (ρ = –0.591; p = 0.039). Narrowing of the dental arch



143


Results

Table 2. Morphological characteristics of subjects

Subject No.

sex

age, years

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

M F F F M M M F M M F M M F M M F

28 22 27 28 27 46 54 23 27 39 58 29 37 29 58 21 42

HP, mm

HP subgroup

NDA

NDA subgroup

FS, %

FS subgroup

13.17 15.32 16.60 16.79 16.97 17.16 17.28 17.99 18.09 19.27 19.48 19.54 20.05 20.35 20.50 20.66 21.18

low low low low low low low low low1 high high high high high high high high

1.09 1.12 1.11 1.08 1.14 1.08 1.10 1.01 1.08 1.04 1.14 1.03 1.22 1.08 1.17 0.96 1.08

narrow wide wide narrow wide narrow wide narrow narrow narrow wide narrow wide narrow1 wide narrow narrow

24.90 20.64 20.12 27.90 23.22 31.37 27.40 27.22 25.59 22.26 20.40 20.63 16.57 15.61 18.88 21.92 15.58

large small small large large large large large large large small small small small small small1 small

HP, height of palate; NDA, narrowing of dental arch; FS, front space. 1 Median of all subjects.

Table 3. Partial regression coefficients (β) in multiple regression analysis among morphometric parameter and MARC with each recording condition

Recording condition: M-bar Target sound: [∫i] Height of palate Narrowing of dental arch Front space

0.512* 0.143 –0.188

AP-bar [t∫i]

[çi]

[ki]

[∫i]

[t∫i]

[çi]

[ki]

0.422 –0.061 0.118

0.244 –0.124 –0.09

–0.299 –0.089 –0.506

–0.076 0.427 –0.343

0.101 0.42 0.108

0.167 0.056 0.001

–0.062 0.134 –0.37

* p < 0.05.

Correlations between MARCs and Parameters The results of multiple linear regression tests are shown in Table 3. A significant correlation was found only between the change in MARCs of [∫i] with M-bar and “front space” (β = 0.512, p = 0.036). Subgroup Comparison of MARCs among 3 Recording Conditions The MARCs under the 3 recording conditions, according to morphological subgroup, are shown in Figure 5. The placement of an M-bar had no significant effect on 144


the MARCs of all target sounds in all subjects; however, placement of an AP-bar had a significant effect on the MARCs of [t∫i] and [ki] (Fig. 5c). No significant effect was found under any recording condition with respect to the MARCs of [çi] (Fig. 5b). In contrast, the placement of an AP-bar had a significant effect on the MARCs of [∫i] among high-palate (p = 0.026), narrow dental arch (p = 0.004), and small front space subgroups (p = 0.014) (Fig. 5a). Comparison of MARICs Regardless of the recording condition, type of target sound, or subgroup, various incorrect labels appeared among all subjects. Regarding the MARICs of [ki], incorWada/Hideshima/Inukai/Katsuki/ Matsuura/Wakabayashi


was not significantly correlated with palate height (ρ = –0.184; p = 1.00) or front space (ρ = –0.328; p = 0.594).

■ Ctrl

■M

*

■ AP

** **

*

100

MARC of [∫i], %

80 60 40 20 0

Low

a

High

Narrow

Palate

Wide

Small

Dental arch

Large

Front space

100

MARC of [çi], %

80 60 40 20 0

Low

b

100

High

Narrow

Palate

*

**

**

High

Low

Wide

Small

Dental arch

**

** *

High

Narrow

Large

Front space

**

**

Wide

Narrow

**

**

Wide

Small

*

**

Large

Small

**

MARC, %

80 60 40 20 0

c

Low

[t∫i]

[ki]

Palate

[t∫i]

[ki]

[t∫i]

Dental arch

[ki]

Large

Front space

ing conditions (the white bar, grey bar, and dark bar represent control [Ctrl], with M-bar, and with AP-bar, respectively) are shown by the subgroups. The MARCs of [∫i], [çi], and both of [t∫i] and [ki] are shown in a, b, and c, respectively. The asterisks indicate a significant difference between recording conditions (* p < 0.05, ** p < 0.01). Influence of Oral Morphology on Speech with Partial Dentures


145


Fig. 5. MARCs under 3 recording conditions, according to morphological subgroup. The MARCs under 3 record-

■ High-palate

*

*

MARIC, %

16 12 8 4 0

M-bar

AP-bar [gi]

M-bar

AP-bar [çi]

Fig. 6. Appearance ratios of incorrect labels of [ki]. The mean ap-

pearance ratios of incorrect labels (MARICs) of 2 labels of [ki] ([gi] and [çi]) are shown by the recording conditions. The white bars represent the low-palate group, while the dark bars represent the high-palate group. The asterisk indicates a significant difference (* p < 0.05) between groups.

rect labels representing the [gi] sound were significantly higher with the placement of M-bars in low-palate subjects (p = 0.027). The number of incorrect labels representing the [çi] sound significantly increased with the placement of an AP-bar in a high-palate subject (p = 0.027) (Fig. 6). Discussion

In this study, 2 alveolar ([∫i] and [t∫i]), 1 palatal ([çi]), and 1 velar ([ki]) sounds were considered the target sounds. During articulation, the [∫i] and [çi] sounds do not require contact between the tongue and palate, whereas [t∫i] and [ki] sounds do require contact. To minimize the influence of individual variations on the speech data, the selected test words were proper familiar nouns for these study subjects. In contrast to the auditory evaluation by speech pathologists, this selection did not affect evaluation by the speech recognition system. The results of this study showed that [t∫i] and [ki] were significantly disturbed by the placement of an APbar, with no significant effect on [çi] among all subjects. This was consistent with the findings of previous studies [13–17, 20–22, 26–28, 39, 40]. In contrast, there were distinctive trends in subgroup analysis only in the MARCs of [∫i], which was readily disturbed by dentures (in a 146


manner similar to [t∫i] and [ki]). Thus, the findings of this study suggested that a palatal bar covering the middle area of the palate was not detrimental for speech production, regardless of the shapes of the dental arch and palate. Furthermore, the major connectors of RPDs exhibited diverse categories of sound that were readily disturbed among high-palate, narrow dental arch, and small front space patients. Previous palatographical studies reported that vocal tract leaks-in and closures in the oral cavity, both caused by changes in the tongue contact, as well as disorders of speech production, were typically encountered following the placement of dentures [18–21]. The speech evaluation system used in this study could assess leaks-in or closure of the vocal tracts by comparing tongue contact areas between the target and emerging sounds, thus identifying the incorrect labels [15, 27, 28]. In this study, the appearance of the [gi] sound as an incorrect label for [ki] significantly increased with the placement of an M-bar in low-palate subjects, compared with high-palate subjects. Since the voice onset time of the voiced plosive [gi] is shorter than that of the voiceless plosive [ki], the appearance of [gi] indicated that the voice onset time of [ki] was incidentally shortened [42, 43]. This suggests that the placement of a palatal bar covering the middle area in low-palate patients could readily cause closures of the vocal tracts. In contrast, the appearance of the [çi] sound as the incorrect label for [ki] significantly increased with the placement of an AP-bar in high-palate subjects, in comparison to low-palate subjects. The tongue does not block airflow through the vocal tract when the fricative [çi] is produced. However, the vocal tract is blocked when the plosive [ki] is produced. Thus, the appearance of [çi] indicated that the vocal tract for [ki] was incidentally leaked by the placement of an AP-bar. This suggests that the placement of a palatal bar covering both anterior and posterior areas for high-palate patients could readily cause a leak in the vocal tract. Despite various reported methods for evaluating the shape of the palate and dental arch, there is no consensus standard [44, 45]. In this study, palate height, narrowing of the dental arch, and the ratio of front area space were regarded as “palate and dental arch parameters,” which could be easily measured on maxillary stone models. Tanaka [35] reported that disorders of speech production with CDs were more severe in small front space patients than in large front space patients. Although we identified a significant association between MARC of [∫i] with Mbar and front space on the basis of multiple linear regression, our findings that placement of the experimental bar Wada/Hideshima/Inukai/Katsuki/ Matsuura/Wakabayashi


■ Low-palate


tation to structural modification of the oral cavity could occur relatively quickly by using intensive, object-specific practice with a palatal plate. Therefore, we randomized the order of the 3 recording conditions for each subject to avoid the possibility of short-term adaptations. Although several reports regarding palatal plates or CDs have focused on long-term adaptations, there have been no studies investigating adaptations to speech production in RPD wearers with bar-type major connectors [36, 37]. Celebić and Knezović-Zlatarić [11] reported that disorders in speech production were more prevalent among RPD wearers than among CD wearers. The findings of the present study indicated that oral morphology could amplify disorders of speech production in RPD wearers in the initial phase of RPD treatments. For clinicians, it is important to elaborate sensational disturbances associated with initial RPD treatment phases until sufficient patient adaptation occurs, and to determine which patients will require care for a speech disorder. Our findings showed that speech impairments tended to appear immediately after RPD placement in patients with high palates and/or narrow dental arches. The relationship between oral morphology and adaptation to speech production in RPD wearers would be an interesting topic for future studies. Conclusion

We evaluated the influence of oral morphology on speech production in subjects with experimental major connectors. Notably, patients with high palates and/or narrow dental arches appear more prone to speech disorders during the initial phase of RPD treatment. Thus, treating patients with such characteristics requires further consideration regarding shapes and designs of major connectors. Acknowledgements This work was partially supported by grants (No. 16K20485 to J.W. and No. 15K20430 to A.K.) offered by the Ministry of Education, Culture, Sports, Science, and Technology of Japan. The authors would like to thank the staff of Toshiba Digital Media Engineering, who were involved in the development of the Voice Analyzer used in this study. We would also like to thank Kazuyuki Handa and Hironari Hayama for their valuable comments.

Disclosure Statement The authors have no conflict of interest to declare.


147


significantly disturbed speech production in small front space subjects, compared with large front space subjects, suggested that speech production in small front space patients was readily disturbed by RPDs. However, the ratio of front volume to total volume is difficult to calculate at the chairside. Conversely, palate height and narrowing of the dental arch can be easily calculated by dental clinicians. The results of correlation analysis among the 3 parameters revealed a negative correlation between the ratio of front volume to total volume and palate height, as the palate height of small front space subjects was high. The results of MARC analysis in high-palate subjects were similar to that of small front space subjects in this study. Brunner et al. [33] reported that acoustic variability was lower among speakers with low palates than among speakers with high palates. Our findings suggested that oral morphology affected the speech production of RPD wearers, as mentioned in previous studies regarding sound dentitions; moreover, prediction of speech disorders in RPD wearers, based on their palate heights or narrowing of the dental arches, was effective. It was appropriate to recruit RPD wearers to evaluate the influence of major connectors of RPDs. However, other factors, including history of RPD use or the condition of opposing dentitions, could affect speech production in RPD wearers [10]. In this study, we recruited subjects with normal dentitions to clarify the influence of connector placement, palate shape, and dental arch shape. In this study, MARC evaluations were consistent with the findings of previous studies that investigated speech production among RPD wearers [13–16, 26–28]. Therefore, subjects chosen for this study were appropriate, regardless of whether they were RPD wearers. However, post hoc analysis yielded a power