- Email: [email protected]
before and after orthognathic surgery
J Oral Maxillofac Surg 60:364-372, 2002
Acoustic and Perceptual Analysis of the Sibilant Sound /s/ Before and After Orthognathic Surgery Alice S.Y. Lee, BSc,* Tara L. Whitehill, PhD,† Valter Ciocca, PhD,‡ and Nabil Samman, FRCS, FDSRCS§ Purpose:
Orthognathic surgery may have a positive or negative effect on speech. Perceptual evaluation of presurgical and postsurgical articulation is difficult because speech errors, when they occur, are usually fricative distortions, which may be difficult to document reliably. In this study, acoustic analysis was used to supplement perceptual judgment of presurgical and postsurgical productions of /s/. Subjects and Methods: The study population consisted of 9 Cantonese speakers undergoing osteotomy for Class III skeletal deformity and 9 age- and gender-matched adults with normal occlusion and speech. The speech sample consisted of 6 words with the initial sibilant sound /s/. Perceptual analysis included narrow phonetic transcription and classification of error types. Acoustic analysis included measurement of first and second spectral peaks, fricative duration, noise bandwidth, and noise-to-vowel decibel ratio. Results: The results of the perceptual analysis showed a decrease in articulatory errors for the group after surgery, although 5 patients had no perceptual errors before surgery. Acoustic analysis showed significant differences between the experimental and control groups before surgery for 2 variables (spectral peak I and bandwidth). Three months after surgery there were no significant differences between the control group and the experimental group, except for bandwidth. Twelve months after surgery, there were significant differences between the 2 groups in noise bandwidth and spectral peak II. Conclusions: The results suggest a possible relapse at 1 year after surgery, based on spectral peak values. Osteotomy appears to result in a positive change in articulation for most patients, but speech outcome after osteotomy must be evaluated both 1 year and shortly after surgery. © 2002 American Association of Oral and Maxillofacial Surgeons J Oral Maxillofac Surg 60:364-372, 2002 Most previous studies of the effect of orthognathic surgery on speech production have reported that surgery had a positive effect on articulation. Glass et al1 studied the speech, lingual diadochokinetics, and swallowing patterns of 5 patients before and after mandibular osteotomy. All 5 subjects showed a de-
crease in sibilant distortions after the surgery. Ruscello et al2 studied the speech changes of 20 patients who underwent surgery for correction of various skeletal defects. The number of speech errors increased immediately after splint removal compared with presurgical performance. However, the number of errors at the 3- and 6-month postoperative periods was lower than the preoperative numbers. Postsurgical deterioration in articulation was not seen in any patient. Most errors that were detected were sibilant distortions. Vallino3 studied the articulation, voice, resonance, hearing sensitivity, and middle ear function of 34 patients who underwent orthognathic surgery. Thirty of the 34 patients showed speech errors before the surgery. The errors were mainly distortion of the sibilants /s/ and /z/. Articulation improved after surgery without speech therapy. Most of the preoperative articulation errors were eliminated 3 months after the surgery. The decrease in errors continued up to 12 months after surgery, at which time only errors on
Received from the University of Hong Kong, Hong Kong. *Postgraduate Student, Department of Speech and Hearing Sciences. †Associate Professor, Department of Speech and Hearing Sciences. ‡Associate Professor, Department of Speech and Hearing Services. §Associate Professor, Oral and Maxillofacial Surgery. Address correspondence and reprint requests to Dr Whitehill: Department of Speech and Hearing Sciences, University of Hong Kong, 34 Hospital Rd (5th floor), Hong Kong; e-mail: [email protected] © 2002 American Association of Oral and Maxillofacial Surgeons
0278-2391/02/6004-0004$35.00/0 doi:10.1053/joms.2002.31221
364
LEE ET AL
/s/ and /z/ were observed. Witzel et al4 examined the articulation skills of 41 patients before and after osteotomy. Before surgery, 29 patients had retrognathia, of whom 17 also had open bite; 11 patients had prognathia, of whom 7 also had open bite; and 1 patient had only an open bite. Twenty-two of the patients had presurgical articulation errors, and errors in the production of sibilants occurred in all groups except the patient with only an open bite. There was a significant improvement in total articulation score and in sibilant sounds for all groups after surgery. Cases with no improvement in speech after osteotomy have also been reported. Dalston and Vig5 investigated the effects of orthognathic surgery on 40 women, and there was no observable change in judged hypernasality, hyponasality, and articulation proficiency for the group. Approximately 1% of consonants produced by the patients were considered in error both before and after surgery. The major error was distortion of /s/ and /z/. The group did not manifest a significant shift in error rate after surgical intervention Garber et al6 studied the articulation of 6 patients before and after osteotomy for maxillary prognathism. Presurgical articulation was considered within normal limits for all speakers. The number of errors increased immediately after surgery and returned to the presurgical level by 1 month after splint removal. The number of errors further decreased by 1 year after surgery. The authors concluded that the surgery had no long-term effects on speech. A few recent studies have used acoustic analysis to investigate speech errors before and after osteotomy. Acoustic analysis offers an objective and reliable measure, which may supplement perceptual investigation. Acoustic analysis may be particularly useful in investigating speakers with malocclusion whose articulation is characterized by distortion of fricatives, which may be difficult to reliably document perceptually. In addition, acoustic analysis may offer an opportunity to measure changes in articulation that may not be noticeable perceptually. Bowers et al7 investigated 5 patients who were undergoing orthognathic surgery. Speech was recorded at 4 time periods: before treatment, after presurgical orthodontics, after surgery, and after postsurgical orthodontics. The speech of all subjects was perceptually normal before and after surgery. There was a significant change in the second formant frequency (F2) of /i/ after surgical treatment for all patients. The F2 values returned to pretreatment levels (ie, there was no significant difference between the values from the first and fourth time periods) for 4 of the 5 patients Bowers et al7 interpreted this as evidence of “active articulatory accommodation.” As the
365 authors concluded, articulation may change as a result of orthodontic or surgical intervention yet fall within the range of normal. Wakumoto et al8 investigated the extent to which articulatory placement for the lingual consonant /s/ changed after surgery in 5 patients (3 with Class III malocclusion and 2 with Class II malocclusion). Electropalatography was used to evaluate linguapalatal contact for all 5 patients, and acoustic analysis was used for 2 patients. The results indicated that significant articulatory (as measured by electropalatography) and acoustic changes occurred after surgery and that the changes were maintained 6 months after surgery. Acoustic changes were in the expected directions; that is, the patient with mandibular prognathism (skeletal Class III) showed a higher consonant peak energy frequency (CPF) after surgery, whereas the patient with maxillary protrusion (skeletal Class II) had a lower CPF after surgery. There was a high correlation between the electropalatographic measure (center of gravity, a measure of anterior-to-posterior linguapalatal contact) and the acoustic measure (CPF) after surgery. Yamamoto et al9 investigated the acoustic characteristics of /s/ and /ʃ/ produced by 14 surgical Class III patients before surgery and 6 months after surgery. Before surgery, the sound pressure level of the highfrequency range of /s/ and that of the mid-frequency range of /ʃ/ appeared lower than those of normal subjects; the postoperative patterns of patients appeared similar to those of the normal subjects. Unfortunately, no statistical analysis was undertaken. The current study focused on the investigation of the sibilant /s/ in speakers undergoing osteotomy for Class III skeletal deformity. The fricative /s/ was selected because, as previously noted, distortion of /s/ has been reported frequently in speakers with malocclusion.1,6,10-12 This is presumably because the production of /s/ requires very precise placement of the articulators.2,12 Most previous studies have compared presurgical and postsurgical speech performances but have not compared the speech of patients undergoing osteotomy with that of normal speakers. In the current study, a control group of speakers with normal occlusion and normal speech was included. A comparison of the speech performance of patients undergoing osteotomy with that of normal speakers permits an evaluation of not only whether surgery results in change but also whether the presurgical or postsurgical results are comparable to those of normal speakers. Some studies of the effect of orthognathic surgery on speech have sampled speech on only 1 occasion after surgery,1,5,9 but studies that have analyzed
366
EFFECT OF ORTHOGNATHIC SURGERY ON /S/
speech on several occasions after surgery have shown changes over time,2,3,6,7 although the changes have not always been statistically significant. Given these reported changes over time and calls in the literature to monitor speech over a longer time period,8 speech was assessed during a 12-month period in the current study. The research questions of this study were as follows. 1) Is there a difference in the production of /s/ before orthognathic surgery for Class III skeletal deformity and at 3 and 12 months after surgery on the basis of perceptual and acoustic analysis? 2) Is there a significant difference in the production of /s/ between a group of normal speakers and a group of speakers with Class III skeletal deformity at the 3 time periods?
Methods SUBJECTS
The experimental group included 9 patients (4 women and 5 men). Age ranged from 20 to 40 years, with a mean age of 27 years at the time of presurgical evaluation. The subjects were selected from 45 patients who presented consecutively for orthognathic surgical treatment of dentofacial deformity in the Department of Oral and Maxillofacial Surgery, Prince Philip Dental Hospital, University of Hong Kong. All subjects were native Cantonese speakers and had hearing of 25 dB hearing level or better in at least 1 ear as determined by audiometric testing at the octave
frequencies of 250 to 8,000 Hz. Subjects were excluded if they had cleft palate, a syndrome, or moderate-to-severe open bite. To limit the scope of this study, only subjects with Class III skeletal deformity were included. Nine age- and gender-matched adults were also included; these control subjects were native speakers of Cantonese and had normal occlusion and hearing. Table 1 shows the background information on the subjects in the experimental and control groups. DENTOFACIAL EXAMINATION
The dentofacial examination procedures have been previously described.13,14 Dentofacial examination and diagnosis for the experimental group were conducted by the surgical staff before and 6 months after surgery. Diagnosis of skeletal and jaw relationship was made on the basis of cephalometric analysis of bony and soft tissue parameters using a computerized system (Dentofacial Planner Plus Version 2.02; Dentofacial Software Inc, Toronto, Canada). Examination of the control group was done by an experienced fourth-year dental student from the Faculty of Dentistry, University of Hong Kong. Occlusal diagnosis, amount of overjet, and amount of overbite are shown in Table 1. SURGICAL PROCEDURES
All surgery was performed by postgraduate trainees under the supervision of faculty surgeons in the Department of Maxillofacial Surgery, Prince Philip Dental Hospital, University of Hong Kong. Surgical proce-
Table 1. CHARACTERISTICS OF THE STUDY SUBJECTS
Presurgical Measures Subject
Age (yr)/Gender
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
29/M 25/M 34/M 20/M 21/F 21/F 28/F 23/F 40/M 29/M 25/M 34/M 20/M 21/F 21/F 28/F 23/F 40/M
Occlusion
Overbite, Overjet (mm)
Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class
⫺2, ⫺6 ⫺1, ⫺4 ⫺4, ⫺4 ⫺2.1, ⫺3.2 ⫺0.8, ⫺11.1 ⫺2.1, ⫺8.1 ⫺3, ⫺2.9 ⫺2, ⫺3 ⫺2, ⫺8 ⫹3, ⫹4 ⫹2, ⫹2.5 ⫹2, ⫹2.5 ⫹3, ⫹3 ⫹2.5, ⫹4 ⫹2, ⫹2 ⫹7, ⫹4 ⫹1, ⫹1 ⫹4, ⫹3
III III III III III III III III III I I I I I I I I I
Postsurgical Measures Surgical Procedures* 1, 2, 4 1, 3 4, 5 1, 3 1, 4 1, 2 1, 2, 4 1 1, 3 — — — — — — — — —
Occlusion
Class Class Class Class Class Class Class Class Class — — — — — — — — —
I I I I I I I I I
Overbite, Overjet (mm)
⫹2, ⫹5 ⫹2, ⫹3 ⫹3, ⫹3 ⫹4, ⫹3 ⫹4, ⫹3 ⫹2, ⫹4 ⫹3, ⫹3 ⫹4, ⫹2.5 ⫹4, ⫹4 — — — — — — — — —
*Surgical procedures: 1, Le Fort 1 maxillary osteotomy; 2, Hofer anterior subapical osteotomy; 3, body step osteotomy; 4, vertical subsigmoid ramus osteotomy; 5, mandibular genioplasty osteotomy.
367
LEE ET AL
dures included Le Fort I maxillary, Hofer anterior subapical, body step, vertical subsigmoid ramus, and mandibular genioplasty osteotomies. The selection of surgical method was determined by the type and degree of dentofacial deformity. The patients underwent different combinations of surgical procedures, as shown in Table 1. SPEECH EVALUATION
The speech assessment was conducted in a quiet room. Speech samples were recorded using a Sony (Tokyo, Japan) TCD-D3 Digital (DAT) tape recorder and a Sony ECM-909 microphone that was maintained at a mouth-to-microphone distance of 10 cm. The assessment was also videorecorded using a JVC GRAX7E video camera (Tokyo, Japan). The video camera was positioned to allow maximum view of the mouth. Speech was assessed the day before surgery and at 3, 6, and 12 months after surgery. In this study, only the presurgical, 3-month postsurgical, and 12-month postsurgical data were evaluated. One patient failed to appear for the 12-month follow-up; for this patient, the 6-month follow-up was used as the final measurement period. The speech evaluation included assessment of voice, resonance, nasal emission, nasalance, and articulation as part of a larger study.13 This study focused on 6 words from the Cantonese Osteotomy Deep Test (CODT [Hong Kong, Department of Speech and Hearing Services, University of Hong Kong, 1995]). The CODT is a 120-item test that assesses 6 phonemes known to be vulnerable in speakers with dentofacial abnormalities: /s/, /ts/, /tsh/, /f/, /p/, and /ph/. Each phoneme was sampled 20 times, in consonant-vowel and consonant-vowel-consonant structures and in varying phonetic contexts. The 6 words investigated in the current study were all consonant-vowel in structure: /sa1/ (sand), /s⑀4/ (snake), /si4/ (key), /sɔ1/ (comb), /sy6/ (tree), and /s⑀6/ (shoot). (Subscript numbers refer to lexical tones of Cantonese: Subscript 1 indicates a high-level tone; subscript 4, a low falling tone; and subscript 6, a low-level tone.) The subjects read the word list aloud. Transcription procedures have been described previously13 and were as follows. Articulation was rated live independently by 2 native Cantonese speakers trained in IPA phonetic transcription. Transcriptions and coding were compared on a point-by-point basis. Any discrepancies between the 2 evaluations were subsequently resolved by viewing the audio and video recordings together with the second author, a speech-language pathologist with more than 10 years of experience with speech disorders associated with orofacial abnormalities. All discrepancies were discussed and resolved using a consensus model to ensure inclusion of all of the data.
Phonetic transcriptions of /s/ productions were coded as correct or incorrect. Error productions were classified as substitution, distortion, omission, or addition. Distortions were fully described. Distortions that were visual only3 were not considered errors for the purpose of this study. The number of auditorily correct /s/ productions was calculated for each subject at each sampling period (before surgery and 3 and 12 months after surgery). ACOUSTIC ANALYSIS
SoundScope version 1.2 (GW Instruments, Somerville, MA, 1993) was used for the acoustic analysis. Each acoustic signal was digitized at a sampling frequency of 44 kHz. Five parameters were included in the acoustic analysis of /s/: spectral peak I, spectral peak II, duration, noise bandwidth, and /s/-vowel dB ratio. These measures were selected due to their sensitivity to /s/ misarticulation.15-17 For example, dentalized /s/ sounds (produced with the tongue tip close to the upper teeth) result in spectra that have a broader bandwidth and higher frequency peaks than normal productions. Lateralized productions (air escaping from the sides of the tongue) are characterized by /s/ spectra of wider bandwidth and extension into lower frequencies than are normally produced /s/ sounds.15 The fricative and vowel components were visually identified. The duration of frication noise was measured by using both waveform displays and a spectrogram (Fig 1). The start of frication noise was determined by the presence of high-frequency energy above 2.4 kHz, and the end was determined by the sudden disappearance of the high-frequency energy.
FIGURE 1. Spectrogram and waveform display of the word /sɔ1/ spoken by a normal woman. The 2 cursors mark the beginning (220 ms) and the end (414 ms) of the frication noise. The period between these 2 points is the duration of the frication noise (194 ms).
368
EFFECT OF ORTHOGNATHIC SURGERY ON /S/
control group and the experimental group (before and 3 and 12 months after surgery) for each of the 5 acoustic variables. Post-hoc Dunnett’s tests were carried out to compare the acoustic measurements for the presurgical and postsurgical conditions of the experimental group with those of the control group.18 RELIABILITY
Intrajudge and interjudge reliability values were obtained for the acoustic data by measuring 10% of the data again. Reliability was calculated using Pearson’s product-moment correlation. Interjudge reliability was r ⫽ 0.96 (P ⬍ .05), and intrajudge reliability was r ⫽ 0.94 (P ⬍ .05).
Results FIGURE 2. Frequency measurement of the same word, /sɔ1/, using average FFT spectrum. Arrow 1, Spectral peak I (8,268.75 Hz). Arrow 2, Spectral peak II (4,823.44 Hz). Arrows a (4,306.64 Hz) and b (11,541.8 Hz) mark the bandwidth (7,235.16 Hz), which is the frequency range within which the noise energy level was not less than 12 dB lower than the largest spectral peak.
The middle 100 ms of frication noise was then selected to measure the 2 most intense spectral peaks and the bandwidth by using the average spectrum (450-Hz bandwidth, with preemphasis) (Fig 2). Spectral peak I referred to the frequency of the most intense spectral peak, and spectral peak II referred to the frequency of the second most intense spectral peak. Bandwidth was defined as frequency range at which the noise energy level was not less than 12 dB lower than the largest spectral peak.15 This criterion for specifying the bandwidth of fricatives is relatively liberal compared with the ⫺3 dB (half-power) points used to define the bandwidth of vowel formants and filters. Such a choice is necessary to avoid selecting the bandwidth of the (often narrow) spectral peaks as the bandwidth of the (usually a few kilohertz wide) fricative spectra. The /s/-vowel ratio was measured by envelope (Fig 3). The ratio was obtained by comparing the rootmean-square amplitude difference (rms dB) of the highest amplitude of the frication noise with that of the vowel using a 40-ms temporal window. Mean values were calculated across the 6 words for each acoustic measure for each subject. Mean and SD values were calculated for the control group (once) and for the experimental group before and 3 and 12 months after surgery. STATISTICAL ANALYSIS
A 1-way repeated measures analysis of variance (ANOVA) was used to assess differences between the
PERCEPTUAL
The results of the perceptual analysis are shown in Table 2. Five of the 9 speakers in the experimental group showed no articulation errors before the surgery. Of these 5 patients, 4 also showed no speech errors during the 3- and 12-month postsurgical evaluations; that is, there was no deterioration in speech performance on the basis of perceptual judgment. One of the 4, subject 8, was the speaker whose final evaluation was at 6 months rather than at 12 months. The other speaker in this group (subject 9) demonstrated 66.6% accuracy (4 of 6) for the 6 target words
FIGURE 3. The /s/- vowel ratio of /sɔ1/ using an energy envelope (40-ms temporal window). Arrow 1, Highest amplitude of the frication noise (4.333 V). Arrow 2, Peak of the vowel (10 V). The /s/- vowel ratio (⫺7.264 dB) was obtained by calculating the difference between these 2 amplitude values.
369
LEE ET AL
Table 2. PERCEPTUAL ACCURACY BEFORE AND AFTER SURGERY
Evaluation Period Presurgery
3 mo
12 mo
Subject
Accuracy
Error Pattern
Accuracy
Error Pattern
Accuracy
Error Pattern
1 2 3 4 5 6 7 8 9 Total No. of errors
5/6 6/6 6/6 4/6 4/6 6/6 5/6 6/6 6/6 6
/s/ 3 [ts] — — /s/ 3 [ts] /s/ 3 [sj] — /s/ 3 [sj] — —
6/6 6/6 6/6 6/6 6/6 6/6 6/6 6/6 4/6 2
— — — — — — — — /s/ 3 [sj]
6/6 6/6 6/6 5/6 6/6 6/6 6/6 6/6 4/6 3
— — — /s/ 3 [sj] — — — — /s/ 3 [sj]
during both the 3- and 12-month postsurgical assessments. For the 4 patients who manifested articulation errors before the operation, all showed 100% (6 of 6) accuracy at the time of the first (3-month) postsurgical evaluation. Three of these speakers maintained this accuracy at 1 year after surgery; the other speaker (subject 4) showed 1 error (5 of 6) at the 12-month postsurgical evaluation. Before surgery, 2 patients had substitution errors (/s/ 3 [ts] for both) and 2 patients had distortion errors (/s/ 3 [sj] for both). After surgery, all errors were palatalization (/s/ 3 [sj]). ACOUSTIC
Table 3 shows the mean and SD values for each of the 5 acoustic measures for the control group and for the experimental group at 3 time periods (before and
3 and 12 months after surgery). The results for each of the 5 acoustic variables (spectral peaks I and II, duration, noise bandwidth, and /s/-vowel dB ratio) are presented here in separate sections. Spectral Peak I A 1-way ANOVA with repeated measures for the spectral peak I data produced a statistically significant main effect [F(2,14) ⫽ 3.45, P ⬍ .05]. The mean presurgical spectral peak I (9.09 kHz) was significantly higher than that of the control group (7.36 kHz; Dunnett’s test, critical value ⫽ 1,551.41 Hz, P ⬍ .05). At 3 months after surgery, the mean spectral peak I of the experimental group (7.71 kHz) was not significantly different from the mean of the control group (Dunnett’s test, P ⬎ .05). By 12 months after surgery, the mean spectral peak I for the experimental group had increased to 8.7 kHz, moving toward
Table 3. ACOUSTIC VARIABLES FOR THE EXPERIMENTAL AND THE CONTROL GROUP
Experimental Group Variable Spectral peak I (kHz) Mean SD Spectral peak II (kHz) Mean SD Duration (ms) Mean SD Noise bandwidth (kHz) Mean SD /s/-vowel ratio (rms dB) Mean SD
Before Surgery
3 mo After Surgery
12 mo After Surgery
Control Group
9.09* 1.96
7.71 1.34
8.7 1.44
7.36 1.01
9.26 1.89
8.79 1.46
9.39* 1.31
7.91 1.55
163 43.19
159 25.19
7.01* 1.36 ⫺4.7 1.55
7.26* 1.74 ⫺4.9 2.77
Abbreviation: rms, root-mean-square. *Statistically significant difference between the experimental group and the control group.
159 34.41 7.07* 1.89 ⫺5.6 1.72
186 29.77 5.35 1.08 ⫺3.7 4.45
370 the presurgical value. However, the difference between the 12-month postsurgical and the control groups was not statistically significant (Dunnett’s test, P ⬎ .05). Spectral Peak II The pattern of the mean values for this acoustic variable mirrored that of spectral peak I; that is, the mean value for the experimental group appeared higher than that for the control group before surgery. It approached that of the control group 3 months after surgery, but it returned to close to presurgical values at 12 months after surgery. However, the mean value of the experimental group was significantly higher than that of the control group (7.91 kHz) in the 12-month postsurgical condition (9.39 kHz; Dunnett’s test, critical value ⫽ 1,453.17 Hz, P ⬍ .05) but not in the presurgical (9.26 kHz) and 3-month postsurgical (8.79 kHz) conditions (Dunnett’s test, P ⬎ .05). The overall 1-way repeated measures ANOVA failed to show a statistically significant effect [F(3,24) ⫽ 2.66, P ⬎ .05]. Duration The mean /s/ duration of the control group was 186 ms, whereas the mean for the experimental group before surgery was 162 ms. The surgery had virtually no effect on the duration of frication, at both 3 and 12 months after surgery (both 159 ms). None of the mean values of the experimental group were significantly different from the mean value of the control group (Dunnett’s test, critical value ⫽ 30.86 ms, P ⬎ .05). The main effect of duration (1-way ANOVA with repeated measures) was not statistically significant [F(3,24) ⫽ 2.31, P ⬎ .05]. Noise Bandwidth Although the mean noise bandwidth of the experimental group varied little after surgery (3 months, 7.26 kHz; 12 months, 7.07 kHz) relative to the presurgical condition (7.01 kHz), all of these mean bandwidths were significantly wider than those of the control group (5.35 kHz; Dunnett’s test, critical value ⫽ 1,649.03 Hz, P ⬍ .05). A repeated measures 1-way ANOVA was carried out on the noise bandwidth data; the main effect was statistically significant [F(3,24) ⫽ 3.65, P ⬍ .05]. /s/-Vowel dB Ratio None of the mean values of the /s/-vowel dB ratio for the experimental group were significantly different from the mean values of the control group (Dunnett’s test, critical value ⫽ 2.91 dB, P ⬎ .05). A 1-way ANOVA with repeated measures carried out on these data did not produce a statistically significant main effect [F(3,24) ⫽ 0.92, P ⬎ .05].
EFFECT OF ORTHOGNATHIC SURGERY ON /S/
Discussion Estimates of the number of patients with dentofacial deformity who experience articulation impairment are around 50%.2,4,13 The results of the current study, in which 5 of the 9 speakers had 100% accuracy (perceptually based) for the production of /s/ before surgery, were consistent with these findings. All 4 patients who demonstrated articulation errors before surgery on the basis of perceptual judgment showed a decrease in the number of errors after surgery. One of these patients showed a slight deterioration between 3 and 12 months after surgery, but the performance 1 year after surgery was still better than the presurgical performance. These results were in agreement with previous studies,1,3,4 which also found a reduction in errors after orthognathic surgery for some speakers. Clearly, orthognathic surgery brings an improvement in articulation for some patients. One patient with accurate production before surgery showed a decrease in performance after surgery on the basis of perceptual evaluation (palatalization of /s/, for 2 of 6 trials, at 3 and 12 months after surgery). A deterioration in speech after orthognathic surgery has also been previously reported.2,6 This may be the result of a failure to adjust to the new relationships between the oral structures.3,7 Most previous investigations of speakers with dentofacial deformity have reported only distortion errors.12 Although distortion was the most common type of error in the current study, substitution (/s/ 3 [ts]) also occurred. The 9 speakers in the current study were included in the group of 33 speakers investigated by Whitehill et al.13 The [s 3 ts] substitution pattern was reported in that study and has been attributed to phonologic differences between Cantonese and English.13 The results of the acoustic analysis indicated that before surgery, the experimental group had a significantly higher first spectral peak frequency and wider noise bandwidth than the control group. Three months after surgery, there were no significant differences between the control group and the experimental group except for bandwidth, for which the experimental group maintained a wider mean value. Twelve months after surgery, there were significant differences between the control group and the experimental group in 2 acoustic variables: noise bandwidth and spectral peak II. One of the most interesting findings in this study was that for 2 of the acoustic variables (spectral peaks I and II), there appeared to be a positive change (ie, values approached those for the control group) 3 months after surgery, but there was a “relapse” (ie, a movement away from the normal values and toward
LEE ET AL
the presurgical values) at 12 months after surgery. This may be related to structural relapse. Unfortunately, postsurgical dentofacial measures were taken only at 6 months. It is recommended that detailed dentofacial measures be taken at several postsurgical periods, including 12 months after surgery, to document the possible relationship between relapse in speech and dentofacial measures. Another possible explanation for the apparent improvement and then deterioration in articulation (at least as measured acoustically) is adaptation. It is possible that after an initial positive reaction to the new structural relationship between articulators brought about as a result of orthognathic surgery, patients revert to old, habituated patterns of articulation. Bowers et al7 reported F2 vowel patterns that appeared to differ from normal patterns immediately after surgery but that returned to presurgical (and presumably normal) patterns some time after surgery, citing this as evidence of active articulatory adjustment. This pattern is different from that observed in the current study, where the patterns appeared abnormal before surgery, were closer to normal immediately after surgery, and returned to abnormal some time after surgery. Wakumoto et al8 raise the hypothetical possibility of compensation or articulatory reorganization after orthognathic surgery. The patients in their study did not show evidence of such compensation and instead showed a more direct or expected change in articulation consistent with the surgical changes. However, as stated by Wakumoto et al,8 “Further research is needed to establish. . . whether the changes in place of articulation reported here are permanent in the longer term, ie, beyond a 6-month period.” The results of the current study suggest not only that observed changes may not be permanent for all patients but also that the articulatory reorganization hypothesized by Wakumoto et al8 may take place in some patients. Orthognathic surgery involves not only the establishment of a more normal occlusal relationship but also more normal relationships between the speech structures such as (in the case of the sibilant sound /s/) the tongue tip and the alveolar ridge. Many patients undergoing orthognathic surgery show improved speech after being provided a more normal speech anatomy. However, the results of this study indicate that some speakers may revert to their habituated, abnormal speech patterns some time later. As Wakumoto et al8 pointed out, experimental studies have shown that when articulatory structures are temporarily altered (eg, using a bite-block paradigm), speakers are able to compensate, producing unaltered speech. This may be viewed as “positive” compensation whereby speakers overcome a temporarily abnormal speech mechanism to produce normal speech.
371 “Negative” compensation, as hypothesized by Wakumoto et al8 and supported by the current study, involves compensation to a newly normal speech mechanism by a return to abnormal speech patterns. This long-term response to altered speech structures has not been previously well investigated. Whitehill et al13 found no significant difference in articulatory accuracy (percentage of phonemes transcribed as correct) for speakers with malocclusion with and without open bite or with and without abnormal overjet and no significant correlation between speech accuracy and the amount of overjet or degree of open bite. Therefore, the relationship between speech accuracy and these orofacial variables was not investigated in the current study. To summarize, the results of perceptual analysis showed that orthognathic surgery for the Class III skeletal deformity brought about either improvement in articulation or, for those with adequate articulation before surgery, no deterioration (for most patients, 8 of 9). The results of acoustic analysis were mixed. For noise duration, there was virtually no change after surgery. For /s/-vowel ratio, abnormal presurgical values appeared to worsen after surgery. For noise bandwidth, values appeared to worsen after surgery and to return to close to presurgical values at 12 months after surgery. Finally, the results for 2 variables (spectral peaks I and II) suggested that the experimental group became more normal after surgery, with a relapse toward presurgical performance 1 year after surgery. However, not all of these patterns were statistically significant. The lack of significant difference between the experimental and control groups may be related to the small number of subjects included in the study. Speech errors in patients with malocclusion, when they occur, are most commonly distortions of fricatives. Such errors can be difficult to document reliably using perceptual analysis. This may limit analysis of surgical efficacy for this group of speakers. Acoustic procedures offer promise for the evaluation of the speech of this group of speakers because they offer an objective measure that may supplement perceptual judgment. Acoustic measures also permit the investigation of subtle changes in articulation that may not be perceptible to the listener.7-9 This may allow an increased understanding of articulatory changes associated with surgical changes in the articulators and their relationship. However, the precise relationship among acoustic variables, perceptual judgment, and movement of the articulators remains a complex one. For example, it is not clear what extent of change or difference in acoustic measures leads to clinically or perceptually significant differences. Finally, several previous studies have suggested the need to monitor speech performance after orthog-
372 nathic surgery at several time intervals, including up to 12 months after surgery.3,6,8 The results of this study strongly support the need to document speech up to 1 year after surgery, not only to allow for improvements in speech that may occur later for some patients,2,3,6 but also to document any possible deterioration in speech. Although none of the speakers in the current study had deterioration in speech between 3 months and 1 year based on perceptual judgment, acoustic analysis suggested a deterioration, possibly due to relapse or to adaptation, based on 2 measures.
8.
9.
10.
11.
12.
References 1. Glass L, Knapp J, Bloomer H: Speech and lingual behavior before and after mandibular osteotomy. J Oral Surg 35:104, 1977 2. Ruscello DM, Tekieli ME, Jakomis T, et al: The effects of orthognathic surgery on speech production. Am J Orthod 89: 237, 1986 3. Vallino LD: Speech, velopharyngeal function, and hearing before and after orthognathic surgery. J Oral Maxillofac Surg 48:1274, 1990 4. Witzel MA, Ross RB, Munro IR: Articulation before and after facial osteotomy. J Maxillofac Surg 8:195, 1980 5. Dalston RM, Vig PS: Effects of orthognathic surgery on speech: A prospective study. Am J Orthod 86:291, 1984 6. Garber SR, Speidel TM, Marse G: The effects on speech of surgical premaxillary osteotomy. Am J Orthod 79:54, 1981 7. Bowers J, Tobey EA, Shaye R: An acoustic-speech study of
13.
14.
15.
16.
17.
18.
patients who received orthognathic surgery. Am J Orthod 88: 373, 1985 Wakumoto M, Isaacson KG, Friel S, et al: Preliminary study of articulatory reorganization of fricative consonants following osteotomy. Folia Phoniatr Logop 48:275, 1996 Yamamoto T, Imai T, Umeda K: Acoustic characteristics of the fricatives in the surgical class III patients, in Morimoto T, Matsuya T, Takada K (eds): Brain and Oral Functions: Oral Motor Function and Dysfunction. Amsterdam, The Netherlands, Elsevier, 1995, pp 611-615 Ruscello DM, Tekieli ME, Van Sickels JE: Speech production before and after orthognathic surgery: A review. Oral Surg 59:10, 1985 Turvey TA, Journot V, Epker BN: Correction of anterior open bite deformity: A study of tongue function, speech changes, and stability. J Maxillofac Surg 4:93, 1976 Vallino LD, Tompson B: Perceptual characteristics of consonant errors associated with malocclusion. J Oral Maxillofac Surg 51:850, 1993 Whitehill TL, Samman N, Wong LLN, et al: Speech errors associated with dentofacial abnormalities in Cantonese speakers. J Med Speech-Lang Pathol 9:177, 2001 Wong LLN, Samman N, Whitehill TL: The effect of orthognathic surgery on hearing sensitivity and middle ear function. Acta Otolaryngol (submitted for publication) Daniloff RG, Wilcox K, Stephens MI: An acoustic-articulatory description of children’s defective /s/ productions. J Commun Dis 13:347, 1980 Lau P: Temporal and spectral variability of /s/ production by Cantonese-speaking cleft palate children. Undergraduate dissertation, University of Hong Kong, Department of Speech and Hearing Sciences, 1994 Weismer G, Elbert M: Temporal characteristics of “functionally” misarticulated /s/ in 4- to 6-year-old children. J Speech Hear Res 25:275, 1982 Howell DC: Statistical Methods for Psychology (ed 3). Boston, MA, PSW-Kent, 1992, p 365
Acoustic and Perceptual Analysis of the Sibilant Sound /s/ Before and After Orthognathic Surgery Alice S.Y. Lee, BSc,* Tara L. Whitehill, PhD,† Valter Ciocca, PhD,‡ and Nabil Samman, FRCS, FDSRCS§ Purpose:
Orthognathic surgery may have a positive or negative effect on speech. Perceptual evaluation of presurgical and postsurgical articulation is difficult because speech errors, when they occur, are usually fricative distortions, which may be difficult to document reliably. In this study, acoustic analysis was used to supplement perceptual judgment of presurgical and postsurgical productions of /s/. Subjects and Methods: The study population consisted of 9 Cantonese speakers undergoing osteotomy for Class III skeletal deformity and 9 age- and gender-matched adults with normal occlusion and speech. The speech sample consisted of 6 words with the initial sibilant sound /s/. Perceptual analysis included narrow phonetic transcription and classification of error types. Acoustic analysis included measurement of first and second spectral peaks, fricative duration, noise bandwidth, and noise-to-vowel decibel ratio. Results: The results of the perceptual analysis showed a decrease in articulatory errors for the group after surgery, although 5 patients had no perceptual errors before surgery. Acoustic analysis showed significant differences between the experimental and control groups before surgery for 2 variables (spectral peak I and bandwidth). Three months after surgery there were no significant differences between the control group and the experimental group, except for bandwidth. Twelve months after surgery, there were significant differences between the 2 groups in noise bandwidth and spectral peak II. Conclusions: The results suggest a possible relapse at 1 year after surgery, based on spectral peak values. Osteotomy appears to result in a positive change in articulation for most patients, but speech outcome after osteotomy must be evaluated both 1 year and shortly after surgery. © 2002 American Association of Oral and Maxillofacial Surgeons J Oral Maxillofac Surg 60:364-372, 2002 Most previous studies of the effect of orthognathic surgery on speech production have reported that surgery had a positive effect on articulation. Glass et al1 studied the speech, lingual diadochokinetics, and swallowing patterns of 5 patients before and after mandibular osteotomy. All 5 subjects showed a de-
crease in sibilant distortions after the surgery. Ruscello et al2 studied the speech changes of 20 patients who underwent surgery for correction of various skeletal defects. The number of speech errors increased immediately after splint removal compared with presurgical performance. However, the number of errors at the 3- and 6-month postoperative periods was lower than the preoperative numbers. Postsurgical deterioration in articulation was not seen in any patient. Most errors that were detected were sibilant distortions. Vallino3 studied the articulation, voice, resonance, hearing sensitivity, and middle ear function of 34 patients who underwent orthognathic surgery. Thirty of the 34 patients showed speech errors before the surgery. The errors were mainly distortion of the sibilants /s/ and /z/. Articulation improved after surgery without speech therapy. Most of the preoperative articulation errors were eliminated 3 months after the surgery. The decrease in errors continued up to 12 months after surgery, at which time only errors on
Received from the University of Hong Kong, Hong Kong. *Postgraduate Student, Department of Speech and Hearing Sciences. †Associate Professor, Department of Speech and Hearing Sciences. ‡Associate Professor, Department of Speech and Hearing Services. §Associate Professor, Oral and Maxillofacial Surgery. Address correspondence and reprint requests to Dr Whitehill: Department of Speech and Hearing Sciences, University of Hong Kong, 34 Hospital Rd (5th floor), Hong Kong; e-mail: [email protected] © 2002 American Association of Oral and Maxillofacial Surgeons
0278-2391/02/6004-0004$35.00/0 doi:10.1053/joms.2002.31221
364
LEE ET AL
/s/ and /z/ were observed. Witzel et al4 examined the articulation skills of 41 patients before and after osteotomy. Before surgery, 29 patients had retrognathia, of whom 17 also had open bite; 11 patients had prognathia, of whom 7 also had open bite; and 1 patient had only an open bite. Twenty-two of the patients had presurgical articulation errors, and errors in the production of sibilants occurred in all groups except the patient with only an open bite. There was a significant improvement in total articulation score and in sibilant sounds for all groups after surgery. Cases with no improvement in speech after osteotomy have also been reported. Dalston and Vig5 investigated the effects of orthognathic surgery on 40 women, and there was no observable change in judged hypernasality, hyponasality, and articulation proficiency for the group. Approximately 1% of consonants produced by the patients were considered in error both before and after surgery. The major error was distortion of /s/ and /z/. The group did not manifest a significant shift in error rate after surgical intervention Garber et al6 studied the articulation of 6 patients before and after osteotomy for maxillary prognathism. Presurgical articulation was considered within normal limits for all speakers. The number of errors increased immediately after surgery and returned to the presurgical level by 1 month after splint removal. The number of errors further decreased by 1 year after surgery. The authors concluded that the surgery had no long-term effects on speech. A few recent studies have used acoustic analysis to investigate speech errors before and after osteotomy. Acoustic analysis offers an objective and reliable measure, which may supplement perceptual investigation. Acoustic analysis may be particularly useful in investigating speakers with malocclusion whose articulation is characterized by distortion of fricatives, which may be difficult to reliably document perceptually. In addition, acoustic analysis may offer an opportunity to measure changes in articulation that may not be noticeable perceptually. Bowers et al7 investigated 5 patients who were undergoing orthognathic surgery. Speech was recorded at 4 time periods: before treatment, after presurgical orthodontics, after surgery, and after postsurgical orthodontics. The speech of all subjects was perceptually normal before and after surgery. There was a significant change in the second formant frequency (F2) of /i/ after surgical treatment for all patients. The F2 values returned to pretreatment levels (ie, there was no significant difference between the values from the first and fourth time periods) for 4 of the 5 patients Bowers et al7 interpreted this as evidence of “active articulatory accommodation.” As the
365 authors concluded, articulation may change as a result of orthodontic or surgical intervention yet fall within the range of normal. Wakumoto et al8 investigated the extent to which articulatory placement for the lingual consonant /s/ changed after surgery in 5 patients (3 with Class III malocclusion and 2 with Class II malocclusion). Electropalatography was used to evaluate linguapalatal contact for all 5 patients, and acoustic analysis was used for 2 patients. The results indicated that significant articulatory (as measured by electropalatography) and acoustic changes occurred after surgery and that the changes were maintained 6 months after surgery. Acoustic changes were in the expected directions; that is, the patient with mandibular prognathism (skeletal Class III) showed a higher consonant peak energy frequency (CPF) after surgery, whereas the patient with maxillary protrusion (skeletal Class II) had a lower CPF after surgery. There was a high correlation between the electropalatographic measure (center of gravity, a measure of anterior-to-posterior linguapalatal contact) and the acoustic measure (CPF) after surgery. Yamamoto et al9 investigated the acoustic characteristics of /s/ and /ʃ/ produced by 14 surgical Class III patients before surgery and 6 months after surgery. Before surgery, the sound pressure level of the highfrequency range of /s/ and that of the mid-frequency range of /ʃ/ appeared lower than those of normal subjects; the postoperative patterns of patients appeared similar to those of the normal subjects. Unfortunately, no statistical analysis was undertaken. The current study focused on the investigation of the sibilant /s/ in speakers undergoing osteotomy for Class III skeletal deformity. The fricative /s/ was selected because, as previously noted, distortion of /s/ has been reported frequently in speakers with malocclusion.1,6,10-12 This is presumably because the production of /s/ requires very precise placement of the articulators.2,12 Most previous studies have compared presurgical and postsurgical speech performances but have not compared the speech of patients undergoing osteotomy with that of normal speakers. In the current study, a control group of speakers with normal occlusion and normal speech was included. A comparison of the speech performance of patients undergoing osteotomy with that of normal speakers permits an evaluation of not only whether surgery results in change but also whether the presurgical or postsurgical results are comparable to those of normal speakers. Some studies of the effect of orthognathic surgery on speech have sampled speech on only 1 occasion after surgery,1,5,9 but studies that have analyzed
366
EFFECT OF ORTHOGNATHIC SURGERY ON /S/
speech on several occasions after surgery have shown changes over time,2,3,6,7 although the changes have not always been statistically significant. Given these reported changes over time and calls in the literature to monitor speech over a longer time period,8 speech was assessed during a 12-month period in the current study. The research questions of this study were as follows. 1) Is there a difference in the production of /s/ before orthognathic surgery for Class III skeletal deformity and at 3 and 12 months after surgery on the basis of perceptual and acoustic analysis? 2) Is there a significant difference in the production of /s/ between a group of normal speakers and a group of speakers with Class III skeletal deformity at the 3 time periods?
Methods SUBJECTS
The experimental group included 9 patients (4 women and 5 men). Age ranged from 20 to 40 years, with a mean age of 27 years at the time of presurgical evaluation. The subjects were selected from 45 patients who presented consecutively for orthognathic surgical treatment of dentofacial deformity in the Department of Oral and Maxillofacial Surgery, Prince Philip Dental Hospital, University of Hong Kong. All subjects were native Cantonese speakers and had hearing of 25 dB hearing level or better in at least 1 ear as determined by audiometric testing at the octave
frequencies of 250 to 8,000 Hz. Subjects were excluded if they had cleft palate, a syndrome, or moderate-to-severe open bite. To limit the scope of this study, only subjects with Class III skeletal deformity were included. Nine age- and gender-matched adults were also included; these control subjects were native speakers of Cantonese and had normal occlusion and hearing. Table 1 shows the background information on the subjects in the experimental and control groups. DENTOFACIAL EXAMINATION
The dentofacial examination procedures have been previously described.13,14 Dentofacial examination and diagnosis for the experimental group were conducted by the surgical staff before and 6 months after surgery. Diagnosis of skeletal and jaw relationship was made on the basis of cephalometric analysis of bony and soft tissue parameters using a computerized system (Dentofacial Planner Plus Version 2.02; Dentofacial Software Inc, Toronto, Canada). Examination of the control group was done by an experienced fourth-year dental student from the Faculty of Dentistry, University of Hong Kong. Occlusal diagnosis, amount of overjet, and amount of overbite are shown in Table 1. SURGICAL PROCEDURES
All surgery was performed by postgraduate trainees under the supervision of faculty surgeons in the Department of Maxillofacial Surgery, Prince Philip Dental Hospital, University of Hong Kong. Surgical proce-
Table 1. CHARACTERISTICS OF THE STUDY SUBJECTS
Presurgical Measures Subject
Age (yr)/Gender
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
29/M 25/M 34/M 20/M 21/F 21/F 28/F 23/F 40/M 29/M 25/M 34/M 20/M 21/F 21/F 28/F 23/F 40/M
Occlusion
Overbite, Overjet (mm)
Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class Class
⫺2, ⫺6 ⫺1, ⫺4 ⫺4, ⫺4 ⫺2.1, ⫺3.2 ⫺0.8, ⫺11.1 ⫺2.1, ⫺8.1 ⫺3, ⫺2.9 ⫺2, ⫺3 ⫺2, ⫺8 ⫹3, ⫹4 ⫹2, ⫹2.5 ⫹2, ⫹2.5 ⫹3, ⫹3 ⫹2.5, ⫹4 ⫹2, ⫹2 ⫹7, ⫹4 ⫹1, ⫹1 ⫹4, ⫹3
III III III III III III III III III I I I I I I I I I
Postsurgical Measures Surgical Procedures* 1, 2, 4 1, 3 4, 5 1, 3 1, 4 1, 2 1, 2, 4 1 1, 3 — — — — — — — — —
Occlusion
Class Class Class Class Class Class Class Class Class — — — — — — — — —
I I I I I I I I I
Overbite, Overjet (mm)
⫹2, ⫹5 ⫹2, ⫹3 ⫹3, ⫹3 ⫹4, ⫹3 ⫹4, ⫹3 ⫹2, ⫹4 ⫹3, ⫹3 ⫹4, ⫹2.5 ⫹4, ⫹4 — — — — — — — — —
*Surgical procedures: 1, Le Fort 1 maxillary osteotomy; 2, Hofer anterior subapical osteotomy; 3, body step osteotomy; 4, vertical subsigmoid ramus osteotomy; 5, mandibular genioplasty osteotomy.
367
LEE ET AL
dures included Le Fort I maxillary, Hofer anterior subapical, body step, vertical subsigmoid ramus, and mandibular genioplasty osteotomies. The selection of surgical method was determined by the type and degree of dentofacial deformity. The patients underwent different combinations of surgical procedures, as shown in Table 1. SPEECH EVALUATION
The speech assessment was conducted in a quiet room. Speech samples were recorded using a Sony (Tokyo, Japan) TCD-D3 Digital (DAT) tape recorder and a Sony ECM-909 microphone that was maintained at a mouth-to-microphone distance of 10 cm. The assessment was also videorecorded using a JVC GRAX7E video camera (Tokyo, Japan). The video camera was positioned to allow maximum view of the mouth. Speech was assessed the day before surgery and at 3, 6, and 12 months after surgery. In this study, only the presurgical, 3-month postsurgical, and 12-month postsurgical data were evaluated. One patient failed to appear for the 12-month follow-up; for this patient, the 6-month follow-up was used as the final measurement period. The speech evaluation included assessment of voice, resonance, nasal emission, nasalance, and articulation as part of a larger study.13 This study focused on 6 words from the Cantonese Osteotomy Deep Test (CODT [Hong Kong, Department of Speech and Hearing Services, University of Hong Kong, 1995]). The CODT is a 120-item test that assesses 6 phonemes known to be vulnerable in speakers with dentofacial abnormalities: /s/, /ts/, /tsh/, /f/, /p/, and /ph/. Each phoneme was sampled 20 times, in consonant-vowel and consonant-vowel-consonant structures and in varying phonetic contexts. The 6 words investigated in the current study were all consonant-vowel in structure: /sa1/ (sand), /s⑀4/ (snake), /si4/ (key), /sɔ1/ (comb), /sy6/ (tree), and /s⑀6/ (shoot). (Subscript numbers refer to lexical tones of Cantonese: Subscript 1 indicates a high-level tone; subscript 4, a low falling tone; and subscript 6, a low-level tone.) The subjects read the word list aloud. Transcription procedures have been described previously13 and were as follows. Articulation was rated live independently by 2 native Cantonese speakers trained in IPA phonetic transcription. Transcriptions and coding were compared on a point-by-point basis. Any discrepancies between the 2 evaluations were subsequently resolved by viewing the audio and video recordings together with the second author, a speech-language pathologist with more than 10 years of experience with speech disorders associated with orofacial abnormalities. All discrepancies were discussed and resolved using a consensus model to ensure inclusion of all of the data.
Phonetic transcriptions of /s/ productions were coded as correct or incorrect. Error productions were classified as substitution, distortion, omission, or addition. Distortions were fully described. Distortions that were visual only3 were not considered errors for the purpose of this study. The number of auditorily correct /s/ productions was calculated for each subject at each sampling period (before surgery and 3 and 12 months after surgery). ACOUSTIC ANALYSIS
SoundScope version 1.2 (GW Instruments, Somerville, MA, 1993) was used for the acoustic analysis. Each acoustic signal was digitized at a sampling frequency of 44 kHz. Five parameters were included in the acoustic analysis of /s/: spectral peak I, spectral peak II, duration, noise bandwidth, and /s/-vowel dB ratio. These measures were selected due to their sensitivity to /s/ misarticulation.15-17 For example, dentalized /s/ sounds (produced with the tongue tip close to the upper teeth) result in spectra that have a broader bandwidth and higher frequency peaks than normal productions. Lateralized productions (air escaping from the sides of the tongue) are characterized by /s/ spectra of wider bandwidth and extension into lower frequencies than are normally produced /s/ sounds.15 The fricative and vowel components were visually identified. The duration of frication noise was measured by using both waveform displays and a spectrogram (Fig 1). The start of frication noise was determined by the presence of high-frequency energy above 2.4 kHz, and the end was determined by the sudden disappearance of the high-frequency energy.
FIGURE 1. Spectrogram and waveform display of the word /sɔ1/ spoken by a normal woman. The 2 cursors mark the beginning (220 ms) and the end (414 ms) of the frication noise. The period between these 2 points is the duration of the frication noise (194 ms).
368
EFFECT OF ORTHOGNATHIC SURGERY ON /S/
control group and the experimental group (before and 3 and 12 months after surgery) for each of the 5 acoustic variables. Post-hoc Dunnett’s tests were carried out to compare the acoustic measurements for the presurgical and postsurgical conditions of the experimental group with those of the control group.18 RELIABILITY
Intrajudge and interjudge reliability values were obtained for the acoustic data by measuring 10% of the data again. Reliability was calculated using Pearson’s product-moment correlation. Interjudge reliability was r ⫽ 0.96 (P ⬍ .05), and intrajudge reliability was r ⫽ 0.94 (P ⬍ .05).
Results FIGURE 2. Frequency measurement of the same word, /sɔ1/, using average FFT spectrum. Arrow 1, Spectral peak I (8,268.75 Hz). Arrow 2, Spectral peak II (4,823.44 Hz). Arrows a (4,306.64 Hz) and b (11,541.8 Hz) mark the bandwidth (7,235.16 Hz), which is the frequency range within which the noise energy level was not less than 12 dB lower than the largest spectral peak.
The middle 100 ms of frication noise was then selected to measure the 2 most intense spectral peaks and the bandwidth by using the average spectrum (450-Hz bandwidth, with preemphasis) (Fig 2). Spectral peak I referred to the frequency of the most intense spectral peak, and spectral peak II referred to the frequency of the second most intense spectral peak. Bandwidth was defined as frequency range at which the noise energy level was not less than 12 dB lower than the largest spectral peak.15 This criterion for specifying the bandwidth of fricatives is relatively liberal compared with the ⫺3 dB (half-power) points used to define the bandwidth of vowel formants and filters. Such a choice is necessary to avoid selecting the bandwidth of the (often narrow) spectral peaks as the bandwidth of the (usually a few kilohertz wide) fricative spectra. The /s/-vowel ratio was measured by envelope (Fig 3). The ratio was obtained by comparing the rootmean-square amplitude difference (rms dB) of the highest amplitude of the frication noise with that of the vowel using a 40-ms temporal window. Mean values were calculated across the 6 words for each acoustic measure for each subject. Mean and SD values were calculated for the control group (once) and for the experimental group before and 3 and 12 months after surgery. STATISTICAL ANALYSIS
A 1-way repeated measures analysis of variance (ANOVA) was used to assess differences between the
PERCEPTUAL
The results of the perceptual analysis are shown in Table 2. Five of the 9 speakers in the experimental group showed no articulation errors before the surgery. Of these 5 patients, 4 also showed no speech errors during the 3- and 12-month postsurgical evaluations; that is, there was no deterioration in speech performance on the basis of perceptual judgment. One of the 4, subject 8, was the speaker whose final evaluation was at 6 months rather than at 12 months. The other speaker in this group (subject 9) demonstrated 66.6% accuracy (4 of 6) for the 6 target words
FIGURE 3. The /s/- vowel ratio of /sɔ1/ using an energy envelope (40-ms temporal window). Arrow 1, Highest amplitude of the frication noise (4.333 V). Arrow 2, Peak of the vowel (10 V). The /s/- vowel ratio (⫺7.264 dB) was obtained by calculating the difference between these 2 amplitude values.
369
LEE ET AL
Table 2. PERCEPTUAL ACCURACY BEFORE AND AFTER SURGERY
Evaluation Period Presurgery
3 mo
12 mo
Subject
Accuracy
Error Pattern
Accuracy
Error Pattern
Accuracy
Error Pattern
1 2 3 4 5 6 7 8 9 Total No. of errors
5/6 6/6 6/6 4/6 4/6 6/6 5/6 6/6 6/6 6
/s/ 3 [ts] — — /s/ 3 [ts] /s/ 3 [sj] — /s/ 3 [sj] — —
6/6 6/6 6/6 6/6 6/6 6/6 6/6 6/6 4/6 2
— — — — — — — — /s/ 3 [sj]
6/6 6/6 6/6 5/6 6/6 6/6 6/6 6/6 4/6 3
— — — /s/ 3 [sj] — — — — /s/ 3 [sj]
during both the 3- and 12-month postsurgical assessments. For the 4 patients who manifested articulation errors before the operation, all showed 100% (6 of 6) accuracy at the time of the first (3-month) postsurgical evaluation. Three of these speakers maintained this accuracy at 1 year after surgery; the other speaker (subject 4) showed 1 error (5 of 6) at the 12-month postsurgical evaluation. Before surgery, 2 patients had substitution errors (/s/ 3 [ts] for both) and 2 patients had distortion errors (/s/ 3 [sj] for both). After surgery, all errors were palatalization (/s/ 3 [sj]). ACOUSTIC
Table 3 shows the mean and SD values for each of the 5 acoustic measures for the control group and for the experimental group at 3 time periods (before and
3 and 12 months after surgery). The results for each of the 5 acoustic variables (spectral peaks I and II, duration, noise bandwidth, and /s/-vowel dB ratio) are presented here in separate sections. Spectral Peak I A 1-way ANOVA with repeated measures for the spectral peak I data produced a statistically significant main effect [F(2,14) ⫽ 3.45, P ⬍ .05]. The mean presurgical spectral peak I (9.09 kHz) was significantly higher than that of the control group (7.36 kHz; Dunnett’s test, critical value ⫽ 1,551.41 Hz, P ⬍ .05). At 3 months after surgery, the mean spectral peak I of the experimental group (7.71 kHz) was not significantly different from the mean of the control group (Dunnett’s test, P ⬎ .05). By 12 months after surgery, the mean spectral peak I for the experimental group had increased to 8.7 kHz, moving toward
Table 3. ACOUSTIC VARIABLES FOR THE EXPERIMENTAL AND THE CONTROL GROUP
Experimental Group Variable Spectral peak I (kHz) Mean SD Spectral peak II (kHz) Mean SD Duration (ms) Mean SD Noise bandwidth (kHz) Mean SD /s/-vowel ratio (rms dB) Mean SD
Before Surgery
3 mo After Surgery
12 mo After Surgery
Control Group
9.09* 1.96
7.71 1.34
8.7 1.44
7.36 1.01
9.26 1.89
8.79 1.46
9.39* 1.31
7.91 1.55
163 43.19
159 25.19
7.01* 1.36 ⫺4.7 1.55
7.26* 1.74 ⫺4.9 2.77
Abbreviation: rms, root-mean-square. *Statistically significant difference between the experimental group and the control group.
159 34.41 7.07* 1.89 ⫺5.6 1.72
186 29.77 5.35 1.08 ⫺3.7 4.45
370 the presurgical value. However, the difference between the 12-month postsurgical and the control groups was not statistically significant (Dunnett’s test, P ⬎ .05). Spectral Peak II The pattern of the mean values for this acoustic variable mirrored that of spectral peak I; that is, the mean value for the experimental group appeared higher than that for the control group before surgery. It approached that of the control group 3 months after surgery, but it returned to close to presurgical values at 12 months after surgery. However, the mean value of the experimental group was significantly higher than that of the control group (7.91 kHz) in the 12-month postsurgical condition (9.39 kHz; Dunnett’s test, critical value ⫽ 1,453.17 Hz, P ⬍ .05) but not in the presurgical (9.26 kHz) and 3-month postsurgical (8.79 kHz) conditions (Dunnett’s test, P ⬎ .05). The overall 1-way repeated measures ANOVA failed to show a statistically significant effect [F(3,24) ⫽ 2.66, P ⬎ .05]. Duration The mean /s/ duration of the control group was 186 ms, whereas the mean for the experimental group before surgery was 162 ms. The surgery had virtually no effect on the duration of frication, at both 3 and 12 months after surgery (both 159 ms). None of the mean values of the experimental group were significantly different from the mean value of the control group (Dunnett’s test, critical value ⫽ 30.86 ms, P ⬎ .05). The main effect of duration (1-way ANOVA with repeated measures) was not statistically significant [F(3,24) ⫽ 2.31, P ⬎ .05]. Noise Bandwidth Although the mean noise bandwidth of the experimental group varied little after surgery (3 months, 7.26 kHz; 12 months, 7.07 kHz) relative to the presurgical condition (7.01 kHz), all of these mean bandwidths were significantly wider than those of the control group (5.35 kHz; Dunnett’s test, critical value ⫽ 1,649.03 Hz, P ⬍ .05). A repeated measures 1-way ANOVA was carried out on the noise bandwidth data; the main effect was statistically significant [F(3,24) ⫽ 3.65, P ⬍ .05]. /s/-Vowel dB Ratio None of the mean values of the /s/-vowel dB ratio for the experimental group were significantly different from the mean values of the control group (Dunnett’s test, critical value ⫽ 2.91 dB, P ⬎ .05). A 1-way ANOVA with repeated measures carried out on these data did not produce a statistically significant main effect [F(3,24) ⫽ 0.92, P ⬎ .05].
EFFECT OF ORTHOGNATHIC SURGERY ON /S/
Discussion Estimates of the number of patients with dentofacial deformity who experience articulation impairment are around 50%.2,4,13 The results of the current study, in which 5 of the 9 speakers had 100% accuracy (perceptually based) for the production of /s/ before surgery, were consistent with these findings. All 4 patients who demonstrated articulation errors before surgery on the basis of perceptual judgment showed a decrease in the number of errors after surgery. One of these patients showed a slight deterioration between 3 and 12 months after surgery, but the performance 1 year after surgery was still better than the presurgical performance. These results were in agreement with previous studies,1,3,4 which also found a reduction in errors after orthognathic surgery for some speakers. Clearly, orthognathic surgery brings an improvement in articulation for some patients. One patient with accurate production before surgery showed a decrease in performance after surgery on the basis of perceptual evaluation (palatalization of /s/, for 2 of 6 trials, at 3 and 12 months after surgery). A deterioration in speech after orthognathic surgery has also been previously reported.2,6 This may be the result of a failure to adjust to the new relationships between the oral structures.3,7 Most previous investigations of speakers with dentofacial deformity have reported only distortion errors.12 Although distortion was the most common type of error in the current study, substitution (/s/ 3 [ts]) also occurred. The 9 speakers in the current study were included in the group of 33 speakers investigated by Whitehill et al.13 The [s 3 ts] substitution pattern was reported in that study and has been attributed to phonologic differences between Cantonese and English.13 The results of the acoustic analysis indicated that before surgery, the experimental group had a significantly higher first spectral peak frequency and wider noise bandwidth than the control group. Three months after surgery, there were no significant differences between the control group and the experimental group except for bandwidth, for which the experimental group maintained a wider mean value. Twelve months after surgery, there were significant differences between the control group and the experimental group in 2 acoustic variables: noise bandwidth and spectral peak II. One of the most interesting findings in this study was that for 2 of the acoustic variables (spectral peaks I and II), there appeared to be a positive change (ie, values approached those for the control group) 3 months after surgery, but there was a “relapse” (ie, a movement away from the normal values and toward
LEE ET AL
the presurgical values) at 12 months after surgery. This may be related to structural relapse. Unfortunately, postsurgical dentofacial measures were taken only at 6 months. It is recommended that detailed dentofacial measures be taken at several postsurgical periods, including 12 months after surgery, to document the possible relationship between relapse in speech and dentofacial measures. Another possible explanation for the apparent improvement and then deterioration in articulation (at least as measured acoustically) is adaptation. It is possible that after an initial positive reaction to the new structural relationship between articulators brought about as a result of orthognathic surgery, patients revert to old, habituated patterns of articulation. Bowers et al7 reported F2 vowel patterns that appeared to differ from normal patterns immediately after surgery but that returned to presurgical (and presumably normal) patterns some time after surgery, citing this as evidence of active articulatory adjustment. This pattern is different from that observed in the current study, where the patterns appeared abnormal before surgery, were closer to normal immediately after surgery, and returned to abnormal some time after surgery. Wakumoto et al8 raise the hypothetical possibility of compensation or articulatory reorganization after orthognathic surgery. The patients in their study did not show evidence of such compensation and instead showed a more direct or expected change in articulation consistent with the surgical changes. However, as stated by Wakumoto et al,8 “Further research is needed to establish. . . whether the changes in place of articulation reported here are permanent in the longer term, ie, beyond a 6-month period.” The results of the current study suggest not only that observed changes may not be permanent for all patients but also that the articulatory reorganization hypothesized by Wakumoto et al8 may take place in some patients. Orthognathic surgery involves not only the establishment of a more normal occlusal relationship but also more normal relationships between the speech structures such as (in the case of the sibilant sound /s/) the tongue tip and the alveolar ridge. Many patients undergoing orthognathic surgery show improved speech after being provided a more normal speech anatomy. However, the results of this study indicate that some speakers may revert to their habituated, abnormal speech patterns some time later. As Wakumoto et al8 pointed out, experimental studies have shown that when articulatory structures are temporarily altered (eg, using a bite-block paradigm), speakers are able to compensate, producing unaltered speech. This may be viewed as “positive” compensation whereby speakers overcome a temporarily abnormal speech mechanism to produce normal speech.
371 “Negative” compensation, as hypothesized by Wakumoto et al8 and supported by the current study, involves compensation to a newly normal speech mechanism by a return to abnormal speech patterns. This long-term response to altered speech structures has not been previously well investigated. Whitehill et al13 found no significant difference in articulatory accuracy (percentage of phonemes transcribed as correct) for speakers with malocclusion with and without open bite or with and without abnormal overjet and no significant correlation between speech accuracy and the amount of overjet or degree of open bite. Therefore, the relationship between speech accuracy and these orofacial variables was not investigated in the current study. To summarize, the results of perceptual analysis showed that orthognathic surgery for the Class III skeletal deformity brought about either improvement in articulation or, for those with adequate articulation before surgery, no deterioration (for most patients, 8 of 9). The results of acoustic analysis were mixed. For noise duration, there was virtually no change after surgery. For /s/-vowel ratio, abnormal presurgical values appeared to worsen after surgery. For noise bandwidth, values appeared to worsen after surgery and to return to close to presurgical values at 12 months after surgery. Finally, the results for 2 variables (spectral peaks I and II) suggested that the experimental group became more normal after surgery, with a relapse toward presurgical performance 1 year after surgery. However, not all of these patterns were statistically significant. The lack of significant difference between the experimental and control groups may be related to the small number of subjects included in the study. Speech errors in patients with malocclusion, when they occur, are most commonly distortions of fricatives. Such errors can be difficult to document reliably using perceptual analysis. This may limit analysis of surgical efficacy for this group of speakers. Acoustic procedures offer promise for the evaluation of the speech of this group of speakers because they offer an objective measure that may supplement perceptual judgment. Acoustic measures also permit the investigation of subtle changes in articulation that may not be perceptible to the listener.7-9 This may allow an increased understanding of articulatory changes associated with surgical changes in the articulators and their relationship. However, the precise relationship among acoustic variables, perceptual judgment, and movement of the articulators remains a complex one. For example, it is not clear what extent of change or difference in acoustic measures leads to clinically or perceptually significant differences. Finally, several previous studies have suggested the need to monitor speech performance after orthog-
372 nathic surgery at several time intervals, including up to 12 months after surgery.3,6,8 The results of this study strongly support the need to document speech up to 1 year after surgery, not only to allow for improvements in speech that may occur later for some patients,2,3,6 but also to document any possible deterioration in speech. Although none of the speakers in the current study had deterioration in speech between 3 months and 1 year based on perceptual judgment, acoustic analysis suggested a deterioration, possibly due to relapse or to adaptation, based on 2 measures.
8.
9.
10.
11.
12.
References 1. Glass L, Knapp J, Bloomer H: Speech and lingual behavior before and after mandibular osteotomy. J Oral Surg 35:104, 1977 2. Ruscello DM, Tekieli ME, Jakomis T, et al: The effects of orthognathic surgery on speech production. Am J Orthod 89: 237, 1986 3. Vallino LD: Speech, velopharyngeal function, and hearing before and after orthognathic surgery. J Oral Maxillofac Surg 48:1274, 1990 4. Witzel MA, Ross RB, Munro IR: Articulation before and after facial osteotomy. J Maxillofac Surg 8:195, 1980 5. Dalston RM, Vig PS: Effects of orthognathic surgery on speech: A prospective study. Am J Orthod 86:291, 1984 6. Garber SR, Speidel TM, Marse G: The effects on speech of surgical premaxillary osteotomy. Am J Orthod 79:54, 1981 7. Bowers J, Tobey EA, Shaye R: An acoustic-speech study of
13.
14.
15.
16.
17.
18.
patients who received orthognathic surgery. Am J Orthod 88: 373, 1985 Wakumoto M, Isaacson KG, Friel S, et al: Preliminary study of articulatory reorganization of fricative consonants following osteotomy. Folia Phoniatr Logop 48:275, 1996 Yamamoto T, Imai T, Umeda K: Acoustic characteristics of the fricatives in the surgical class III patients, in Morimoto T, Matsuya T, Takada K (eds): Brain and Oral Functions: Oral Motor Function and Dysfunction. Amsterdam, The Netherlands, Elsevier, 1995, pp 611-615 Ruscello DM, Tekieli ME, Van Sickels JE: Speech production before and after orthognathic surgery: A review. Oral Surg 59:10, 1985 Turvey TA, Journot V, Epker BN: Correction of anterior open bite deformity: A study of tongue function, speech changes, and stability. J Maxillofac Surg 4:93, 1976 Vallino LD, Tompson B: Perceptual characteristics of consonant errors associated with malocclusion. J Oral Maxillofac Surg 51:850, 1993 Whitehill TL, Samman N, Wong LLN, et al: Speech errors associated with dentofacial abnormalities in Cantonese speakers. J Med Speech-Lang Pathol 9:177, 2001 Wong LLN, Samman N, Whitehill TL: The effect of orthognathic surgery on hearing sensitivity and middle ear function. Acta Otolaryngol (submitted for publication) Daniloff RG, Wilcox K, Stephens MI: An acoustic-articulatory description of children’s defective /s/ productions. J Commun Dis 13:347, 1980 Lau P: Temporal and spectral variability of /s/ production by Cantonese-speaking cleft palate children. Undergraduate dissertation, University of Hong Kong, Department of Speech and Hearing Sciences, 1994 Weismer G, Elbert M: Temporal characteristics of “functionally” misarticulated /s/ in 4- to 6-year-old children. J Speech Hear Res 25:275, 1982 Howell DC: Statistical Methods for Psychology (ed 3). Boston, MA, PSW-Kent, 1992, p 365