Motor Voice Learning Electromyographic study of motor voice learning Edwin M-L YIU Voice Research Laboratory Division of Speech and Hearing Sciences The University of Hong Kong Katherine VERDOLINI Communication Sciences and Disorders University of Pittsburgh and Linda P.Y. CHOW Voice Research Laboratory Division of Speech and Hearing Sciences The University of Hong Kong Correspondence concerning this article should be addressed to Dr. Edwin Yiu (PhD), Associate Professor; Director, Voice Research Laboratory, Division of Speech and Hearing Sciences, The University of Hong Kong, 5/F Prince Philip Dental Hospital, 34 Hospital Road, Hong Kong. Email: eyiu@hku.hk. Tel: +852 - 28590599. Fax: +852- 25590060 1 Motor Voice Learning ABSTRACT The principle of motor learning has been applied to the study of voicing control in recent years. Clinicians and researchers are keen to find out what factors would facilitate good voice production. One of the many factors includes the performance feedback system used during learning. Performance feedback provides information to learners during motor skills acquisition. The feedback can be facilitative or inhibitory to learning. Concurrent and terminal feedbacks are two common feedback systems that are often used in motor learning tasks. These two types of feedbacks influence learning differently. In general, concurrent feedback given during the task would improve motor performance during practice. However, it depresses long term learning measured at a later (retention) stage after the training (Schmidt & Lee, 1999). On the contrary, the use of terminal feedback upon the completion of each trial in a motor learning task tends to degrade the performance during practice but it would enhance learning in the long run. Voice clinicians are often not aware of these principles and use both types of feedback invariantly in voice therapy tasks as there is no empirical evidence available to show whether these principles hold true in learning voice motor task. The effect of concurrent and terminal electromyographic (EMG) biofeedback on the learning of laryngeal muscle relaxation during voicing in a reading task was investigated in the present study. Twenty-two speakers with normal voices were randomly assigned to two groups. One group was given concurrent online EMG waveform display during each reading trial (CONCURRENT feedback group). The other group was given a terminal static EMG waveform display upon the completion of each trial (TERMINAL feedback group). Both groups demonstrated no significant differences or changes in their laryngeal muscle performance over time. However, motor learning appeared to be evident in the orofacial muscles, of which no biofeedback was provided. The TERMINAL group demonstrated improved muscle relaxation in the orofacial muscles over time, whereas the CONCURRENT showed no such improvement. These results show that the provision of terminal feedback for a particular muscle group facilitated learning retention in another muscle group. Keywords: Voice therapy, motor learning, knowledge performance 2 Motor Voice Learning Electromyographic study of motor voice learning Knowledge of performance enables the learners to monitor the quality of the movement they produce during motor learning. Traditionally, feedback is viewed as an reinforcement, reward or motivation for the learner in a motor learning task (Adams, 1987). However, it is now better understood that feedback affects the learning process. Motor performance feedback provides critical information to learners so that they can modify their performance and learning process. Different types of feedback information and their frequency schedules have been shown to affect the learning process differently (Park, Shea, & Wright, 2000; Schmidt & Wulf, 1997; Schmidt, Young, Swinnen, & Shapiro, 1989; Steinhauer & Grayhack, 2000). Two types of augmented feedback have been identified in the literature and they have received much attention in the motor learning research. They are concurrent feedback and terminal feedback. Concurrent feedback refers to knowledge of performance provided to the learner during motor activity, while terminal feedback refers to performance information delivered after the completion of the movement (Schmidt & Lee, 1999). Biofeedback is a widely used clinical tool which provides information on the features or patterns of motor activities which are not usually perceived consciously by the learners. One commonly used biofeedback technique in motor learning is surface electromyography (EMG). Surface EMG records collective electrical activities of muscles lying underneath the surface electrodes (Hocevar-Boltezar, Janko, & Zargi, 1998). EMG biofeedback allows subjects to use visual or auditory monitoring of the electrical activities of their muscles and to attempt exerting some control over these motor activities (Stemple, Weiler, Whitehead, & Komray, 1980). EMG biofeedback has been reported with success in the diagnosis of 3 Motor Voice Learning excessive thyrohyoid muscle tension in the hyperfunctional dysphonic subjects (HocevarBoltezar et al., 1998; Redenbaugh & Reich, 1989). EMG has also been shown to be effective in treating hyperfunctional voice disorders. In a study by Stemple et al. (1980), seven subjects diagnosed with vocal nodules learned to reduce their laryngeal muscle tensions significantly using EMG biofeedback following eight training sessions. In another study, Andrews et al. (1986) reported five women with hyperfunctional voice disorders who learned to improve their voice quality using EMG biofeedback. The improvement lasted for up to three months after the training. Allen, Berstein, and Chait (1991) also demonstrated success in treating vocal nodules in children using EMG as a biofeedback. However, there is little information on how the presentation of these biofeedbacks can be varied to influence the learning process. Feedback can be manipulated in terms of frequency of provision (e.g. more frequent or less frequent), timing of provision (e.g., concurrently or terminally), and type (e.g. performance feedback or results feedback) (Verdolini & Lee, 2001). Generally, it has been found that motor learning improved with reduced frequency of feedback (see, for examples, Park et al., 2000; Schmidt et al., 1989). This general principle has been shown recently by Steinhauer and Grayhack (2000) to be applicable to the area of motor voice learning. Steinhauer and Grayhack (2000) examined the relationship between the provisions of feedback at different frequencies during the learning of a vowel nasalization task. Their results showed that the more frequent the feedback was given, the less learning was shown at the retention period (Steinhauer & Grayhack, 2000). In terms of the timing of feedback, it has been shown that, in general, concurrent biofeedback would improve motor performance during practice but it would often depress learning measured at a later stage after the training – the retention stage (Schmidt & Lee, 1999). On 4 Motor Voice Learning the contrary, the use of terminal feedback upon the completion of each trial in a motor learning task would degrade the performance during the acquisition stage but would enhance learning at the retention stage. Schmidt and his colleagues (Schmidt & Wulf, 1997; Schmidt et al., 1989) first proposed the guidance hypothesis to explain the effect of the concurrent feedback on motor performance and learning. The hypothesis considers concurrent feedback as a strong guidance for the learner to achieve the target behaviour during training. However, too much reliance on this feedback does not allow the learner to consolidate the skill for long-term retention. Therefore, the strong guidance effect of the concurrent feedback improves immediate performance but when the feedback is removed and the guidance disappears, it will result in performance degradation. Moreover, attention and cognitive effort are two additional features put forward by Schmidt and Wulf (1997) and Verdolini and Lee (2001) to account for the different effects of concurrent and terminal feedback on motor performance and learning. Provision of concurrent feedback is considered to reduce a learner’s attention and cognitive effort required for developing learning capabilities. With the same explanation, the increased attention and cognitive effort that a learner puts into the learning process with the provision of terminal feedback results in poorer immediate performance, but in the long run, it will enhance better learning. However, such principle has not been tested in voice motor learning tasks yet. In the present study, the effect of concurrent EMG feedback presented continuously to a group of learners during a voice motor learning task was investigated. The performance of the learners during training and after training was compared with those of another group of learners who received static terminal EMG waveform feedback upon the completion of each voicing practice trial. 5 Motor Voice Learning The objectives of the present study were to investigate, first, whether concurrent feedback facilitated a better motor performance during the training or acquisition phase, and second, whether terminal feedback facilitated a better motor learning as assessed during the retention stage. It was hypothesised that the feedback paradigms used in voice motor task would follow the motor learning principles in general. In other words, terminal feedback is the ultimate feedback paradigm that could improve learning in voice task. These findings help to inform voice clinicians on how to design a training paradigm that would maximize learning in a voice motor task during voice therapy. PILOT STUDY A pilot experiment was carried out to determine the most suitable anatomical sites for electrode placement which could yield stable EMG signals from the laryngeal and neck muscles during a voice motor task. Hocevar-Boltezar et al. (1998) had identified demonstrable muscle tension in patents with dysphonia over the muscles involved in speech and vocal mechanism. These include the orofacial, lower facial, suprahyoid and thyrohyoid muscles. The stability of the EMG signals can be affected by the thickness of the skin, amount of fatty tissue underneath the skin, the size and properties of underlying muscles (Hocevar-Boltezar et al., 1998) as well as the articulatory gesture involved with different phonetic contexts (personal observation), it was therefore necessary to investigate, using a pilot study, which of these four sites could give rise to relatively stable EMG signals with low inter-and intra-subject variability. 6 Motor Voice Learning METHOD OF PILOT STUDY Participants Five females (mean age = 22.6 years, SD = 2.51, range = 21-27) were recruited from among the students studying at the University of Hong Kong. All the subjects had no hearing problems, no history of voice disorders and were medically healthy. None of them received any voice training or voice therapy before. Procedures The electromyography system from the ADInstrument (PowerLab Unit, model ML 780 with an eight-channel Dual Bio Amp model ML135) was used in the study. Each EMG recording electrode was 10mm wide (see Figure 1). The PowerLab Scope software program was used for displaying and analysing the EMG signals. A sample of the EMG signal display is given in Figure 2. Each pair of electrodes was placed symmetrically on either side of the midline of the body for each pair of target site as shown in Figure 3. The sites for electrode placement (see Figure 3) included the: 1) Orofacial area - cheek muscles associated with the Buccinator and Risorius muscles. Each electrode was placed at 2 cm away from the corner of the mouth. 2) Lower facial area - chin muscles associated with the Mentalis. Each electrode was placed at 0.5 cm from the midline of the chin. 7 Motor Voice Learning 3) Suprahyoid area - submandibular muscles associated with the Digastric, Mylohyoid, Stylohyoid, Platysma. Each electrode was place just above the hyoid bone and at 1 cm from the midline in the submandibular area. 4) Thyrohyoid area – extrinsic laryngeal muscles associated with the Thyrohyoid, Omohyoid, Sternohyoid and Platysma. Each electrode was placed over the thyrohyoid membrane at 0.5 cm from the midline. Put Figures 1 to 3 about here An earth strap was attached firmly around the wrist of each subject. In order to ensure there were no loose wirings or loose electrode-skin contact, each subject was asked to rotate her head to confirm that the movement did not produce any artifact in the EMG signal display. Once the set up was determined to be intact, each subject was asked to use her most comfortable pitch and loudness to phonate each of the three vowels (/a/, /i/ and /u/) for four seconds long. Each vowel phonation was repeated three times. The EMG signals were bandpass filtered at 10-500 Hz and the central two-second portion of each vowel phonation was extracted, and then computed to obtain the root-mean-square (RMS) voltage. The RMS voltage represents the effective amplitude of EMG (Baken & Orlikoff, 2000) RESULTS AND SUMMARY FOR PILOT STUDY An inter-subject variability measure was obtained for each vowel by determining the standard deviation of the EMG amplitude of the three phonation trials for each vowel across all five subjects. An intra-subject variability measure for each subject was obtained by determining the standard deviation of the EMG amplitude across the three trials of each vowel phonation 8 Motor Voice Learning by each subject. These intra-subject measures were then averaged to give rise to the mean intra-subject variability. Table 1 list the mean RMS EMG signals in microvolt (V) and the variability measures for the three sustained phonation tasks across the four target sites. The orofacial (cheek) muscles and thyrohyoid area (extrinsic laryngeal muscles) demonstrated two relatively lower overall inter-subject variability measures (6.56 and 5.16 respectively). These two sites also demonstrated relatively lower mean intra-subject variability measures when compared to the lower facial (chin) and suprahyoid (submadibular) areas. The thyrohyoid site was therefore considered to be a suitable site for obtaining stable EMG signals from the laryngeal areas. Since the present study was concerned with laryngeal muscle relaxation training, electrode placement at the thyrohyoid site was therefore selected. As the orofacial muscles also demonstrated relatively stable EMG recording, placement of electrodes at the orofacial (cheek) area was selected as a control site to examine whether laryngeal muscle relaxation training would transfer to the non-laryngeal, orofacial muscles (i.e. the cheek area). Put Table 1 about here MAIN STUDY The Main Study was carried out to determine the effect of concurrent and terminal feedback on the learning of relaxed voice production. EMG activities of the thyrohyoid area were provided to the subjects as feedback since this site demonstrated relatively stable measurements in different phonetic contexts in the pilot study. The objectives of the Main Study were to determine whether: 1) the voice motor learning retention was shown after training (main learning retention effect); 9 Motor Voice Learning 2) the feedback type (concurrent versus terminal) affected learning (main feedback type effect); 3) the learning retention depended on the feedback type (interaction effect). METHOD Participants Eighteen female and four male subjects (mean age = 22.41 years, SD = 1.62, range = 19-27), were recruited from the student community of the University of Hong Kong. They all had normal hearing, were medically healthy, had no history of speech or voice disorders and had no prior experience with electromyography (EMG). The subjects were not informed of the investigator’s hypotheses of the experiment. They were only instructed to undertake a series of reading tasks with minimal speaking effort. Instrumental set up The same instrumental set up as described in the pilot study was used in this Main Study. One pair of electrodes was placed 0.5cm on either side from the midline of thyrohyoid membrane (thyrohyoid site). The other pair was placed 1cm away from the lip corner on either side of the face (orofacial site). A dry earth strap was attached firmly around the wrist. Stimuli Twenty four Cantonese words, that covered all the sounds (19 consonants, eight vowels, 10 dipthongs) and lexical tones in Cantonese, were selected from among the first 750 most 10 Motor Voice Learning frequently occurring Chinese Characters in Hong Kong according to Ho (1993) (see Table 2). They were used as the training and testing stimuli in the study. Each word was embedded in a Cantonese carrier phrase /ji55 k33 hi22/ (meaning “this one is…”) to form a sentence. These 24 sentences constituted one block of stimuli. There was a total five blocks of sentences. Each block of sentences was different from the other blocks in the presentation order of the sentences. Each subject was given the five blocks of stimuli for training. Each subject was also assessed on three other occasions (pre-training - baseline; five minutes after the completion of the training task - immediate retention test; and one week after the training delayed retention test), each using two blocks of stimuli. Motor learning was determined by comparing the assessment at the baseline with those at the immediate and delayed retention stages. Put Table 2 about here Procedures The 22 subjects were randomly assigned to either a group which received concurrent (CONCURRENT group) or terminal (TERMINAL group) biofeedback during the tasks. Concurrent biofeedback refers to the real-time visual display of EMG wave of each production during the training phase. Terminal biofeedback refers to a static EMG visual display after each production was completed during the training. Each subject participated in two sessions which were one week apart. The first session consisted of the baseline, training and immediate retention testing while the second session constituted the delayed retention testing. 11 Motor Voice Learning Each subject was seated approximately one meter in front of two computer monitors, each with a 32cm wide screen. One was used for presenting the stimulus words and the other was used to present the EMG waveform from the thyrohyoid area. EMG activities from the orofacial area (cheek) were not displayed at all. The monitor for displaying the EMG waveform was covered by a movable cardboard when the subjects in the TERMINAL group were carrying out the task. They were only able to view the static EMG waveform display while the cardboard was removed after each production. The subject in the CONCURRENT group, however, were able to view both the monitor screens, one for stimulus presentation and one for EMG feedback, at all time during the training phase. After the electrodes were placed over the thyrohyoid, orofacial and the wrist areas, each subject was asked to rotate his/her head to ensure that there was no movement artifact in the EMG recording. Baseline measurement. Pre-training baseline was measured with no EMG feedback given. No information about the aim of the study (i.e. muscle relaxation) was given to the subjects at this stage so that the EMG recordings were meant to represent the natural phonation of the subjects. Each subject was asked to read aloud two blocks of sentences presented on the monitor screen with their most comfortable pitch and loudness. Each sentence was presented on a screen page with a flashing marker above the characters to prompt the subject to speak at the rate of approximately two words per second. The inter-sentence duration was set at four seconds. The EMG activities were recorded for each sentence and saved for later analysis. Relaxed phonation training. After the baseline was taken, each subject was given a brief explanation on the function of the EMG waveform display. In general, the larger the amplitude the waveform, the more muscle activities and tension there were. Therefore, in a relaxed state, the EMG waveform should be small in amplitudes. Each subject was also informed of the aim of the training was to reduce the EMG amplitude display by relaxing the 12 Motor Voice Learning neck muscles during the reading task. The training involved reading aloud five blocks of 24 sentences. The EMG activity for each sentence was recorded for later analysis. The subjects in the CONCURRENT group were given real time thyrohyoid EMG display of each production superimposed on the previous production (except for the first production). The subjects in the TERMINAL group were only given a static EMG display after each production. Both the static EMGs of the current and previous items were displayed side by side for comparison for all sentences except the beginning sentence of each block. Before the actual training began, each subject was given three trial practices to ensure that they understood how to interpret the biofeedback. Immediate and delayed retention tests. A retention test was carried out immediately after the relaxed phonation training to assess the learning demonstrated by the subject. A similar delayed retention test was carried out one week after the phonation training and the immediate retention test. In each of these tests, the subjects were required to read aloud two blocks of 24 sentences. The EMG activities were recorded during the reading but no EMG feedback was given to the subjects. Data analysis The EMG signals collected from the thyrohyoid and orofacial areas were band-pass filtered at 10-500 Hz. The central two-second portion of each target word was extracted and then computed to obtain the amplitude, which was the RMS voltage in V. The mean amplitude of each block of stimuli was calculated by averaging the amplitude of each of the 24 stimuli. 13 Motor Voice Learning RESULTS EMG muscle activities The mean and standard deviation of EMG amplitude for each block of stimuli measured at the thyrohyoid and the orofacial areas under different measurement phases are shown in Figures 4 and 5 respectively. Repeated measures ANOVA with the 11 measurement phases as dependent (within-subject) variable and the subject group (i.e. feedback type) as independent (between-subject) variable was used for each set of data (i.e. thyrohyoid or orofacial). Multivariate Pillai’s Trace test of significance, which is considered to be a robust test against violation of assumptions in multivariate tests (Coakes & Steed, 2001), was used to determine if there were significant main learning retention effect, main feedback type effect, and interaction effects. Thyrohyoid area. The mean EMG amplitudes of the two baseline stimuli blocks obtained at the thyrohyoid area were 14.80 V and 14.89 V for the CONCURRENT group and 20.39 V and 22.09 V for the TERMINAL group (see Figure 4). Indeed the CONCURRENT group showed a generally lower mean and smaller standard deviation EMG activities than the TERMINAL group. However, Pillai’s Trace test of significance showed that there was no significant main feedback type effect, i.e. no difference between the two subject groups (F=2.80, df=1, p=0.11); no main learning retention effect (i.e. no significant changes within the subjects over the 11 measurement phases (F=1.59, df=10, p=0.23); and no significant interaction effect (F=2.02, df=10, p=0.13). Orofacial area. The muscle activities at the orofacial area, at which the subjects received no feedback on their muscle performance, showed similar pattern for the 14 Motor Voice Learning TERMINAL and CONCURRENT groups (Figure 5). At the baseline measures, mean EMG activities were between 22.78 to 27.59 V (see Figure 5). The two subject groups showed similar mean EMG activities with a decreasing trend across the measurement phases. This is supported by the results of the repeated measures ANOVA (Pillai’s Trace test of significance), which showed a significant main within-subject (i.e. learning retention) effect (F=11.55, df=10, p<0.0001). No significant main between-subject (i.e. feedback type) effect (F=39.74, df=1, p=0.25) or interaction effect (F=0.59, df=10, p=0.23) were found. In order to determine the phase at which significant learning retention effect was demonstrated, planned contrasts were carried out to determine if the mean EMG activity of each block of stimulus measure was significantly different from that of the previous block of measure. Since 10 contrasts were carried out, Bonferroni adjustment (0.05/10=0.005) was used as the new alpha level. The muscle activities within each of the five training blocks (T1 to T5 in Figure 5) showed significant reduction when compared to their previous measurement block (p< 0.0001). The first training block (T1) was also significantly lower than the second block of the baseline measurement (B2; p<0.0001). Extent of learning In order to quantify the amount of learning at each of the training, immediate retention and delayed retention phase, the measurement blocks of each measurement phase were pooled together. Percentage changes in EMG amplitude relative to the baseline measurement phase were calculated for each of the TERMINAL and CONCURRENT group. They are given in Figures 6 (thyrohyoid area) and 7 (Orofacial area). 15 Motor Voice Learning Thyrohyoid area. With the muscle activities at the thyrohyoid area, only the CONCURRENT group showed a reduction of 8.3% EMG activities during training (see Figure 6). Increased muscle activities were found in the TERMINAL group under all three phases and also in the CONCURRENT group during the two retention phases. A repeated measures ANOVA, with the three phases (training, immediate retention and delayed retention) as within-subject variables (main learning retention effect) and the feedback type group as the between-subject variable (main feedback type effect), was carried out to determine if these changes of muscle activities over time were significantly different. Multivariate Phillai's Trace statistics, which is considered to be the most robust statistics against violation of assumptions was used (Bryman & Cramer, 1997). No significant main learning retention effect (F=1.64, df=2, p=0.22), main feedback type effect (F=0.33, df=1, p=0.57) or interaction effect (F=1.62, df=2, p=0.22) were found. Orofacial area. Both the CONCURRENT and TERMINAL groups showed reduced EMG activities relative to the baseline measures (see Figure 7). A repeated measures ANOVA was also carried out to determine if significant changes occurred over time or across the two feedback type groups with different feedback types. No significant main learning retention effect (F=0.23, df=2, p=0.63) or main feedback type effect (F=0.28, df=1, p=0.76) were found. However, a significant interaction effect (F=3.85, df=2, p=0.03) was found. This means that learning was dependent upon the feedback type. Figure 7 shows clearly that the CONCURRENT group demonstrated changes in EMG activities over time while the TERMINAL group maintained relatively stable EMG activities over time. 16 Motor Voice Learning DISCUSSION The present study set out to determine the effect of concurrent and terminal feedback on the learning of relaxed voice production. Electromyograph (EMG) was used in the study as an outcome measure and also as the feedback for the learner as knowledge of performance. The pilot study established that EMG electrodes placed at the thyrohyoid and the orofacial areas relative to the lower facial and the suprahyoid areas, provided the most stable EMG signals during voice production. We propose that these two areas are suitable sites for obtaining relatively stable muscle EMG activities in the study of voice production. The Main Study aimed to investigate the effect of feedback type on vice motor learning. There were two research questions. The first was whether the concurrent feedback facilitated a better motor performance during training. The second was whether terminal feedback facilitated a better long term learning. Previous motor learning studies showed that concurrent feedback often enhanced immediate performance during practice but it is detrimental to performance in longer term retention or transfer tests (Schmidt & Lee, 1999). Schmidt and his colleagues (Schmidt & Wulf, 1997; Schmidt et al., 1989) suggested that concurrent real-time feedback on the muscle activities during motor learning provides strong guiding information which allows the subjects to modify their behavior immediately to yield a better result. However, the reliance on the guiding information prevents the subjects to attend to the natural intrinsic feedback which therefore affects long term learning. On the other hand, subjects receiving terminal feedback require more attention and cognitive effort to recall their performances and associate such performance to the delayed feedback. Therefore, the immediate performance was often degraded but long term learning is facilitated. The present study showed that the two groups of subjects receiving different types 17 Motor Voice Learning of feedback (terminal versus concurrent) demonstrated no significant difference in their performance at the thyrohyoid area. Such finding does not seem to support the general principle of motor learning. However, a closer examination of the results (see Figure 4) showed that the CONCURRENT group demonstrated less variability (as indicated by the SD) than the TERMINAL group in all measurement blocks. Furthermore, the means and the dispersions demonstrated by the CONCURRENT group were all within the dispersion range of the TERMINAL group. This therefore accounted for the non-significant results. A larger sample size with a better power may provide a different result. Nevertheless, the muscle activities recorded at the orofacial area provided some interesting findings which appeared to support the guidance hypothesis. When individual measurement blocks were compared, a significant decreasing trend in the muscle activities in both feedback type subject groups was noticed (see Figure 5). Figure 7 shows clearly the muscle activities at the orofacial area decreased (negative percentage change) over time while those at the thyrohyoid area increased. Since the subjects did not receive any feedback regarding their orofacial muscle activity, the significant improvement in orofacial muscle relaxation relative to the baseline measures, could be attributed to the benefit of not having any “guidance”. As the subjects were not given specific instruction on how to lower the EMG activities in response to the visual waveform feedback from the thyrohyoid area, the subjects would have to develop their own motor strategies to achieve this. Such strategies might have included, for example, gentle initiation of the first syllable, reduced articulatory movement of the articulators (jaw and lips), and conscious effort in reducing the muscle tension at the head and neck areas. Without the augmented feedback from the orofacial area, the subjects would have to attend to the natural intrinsic feedback that may be present during the practice (Schmidt & Wulf, 1997). Since electrodes were also placed on the orofacial area, these might have 18 Motor Voice Learning prompted the subjects to attend to the muscle activities around the orofacial area as well. This Hawthorne effect might have further contributed to the improved relaxation at the orofacial area. On the other hand, the augmented feedback from the thyrohyoid area might have “distracted” the subjects. Although the data from the thyrohyoid area did not show any difference inn the performance or learning between the CONCURRENT and TERMINAL groups, the pooled data from the orofacial area (Figure 7) showed that the feedback types had a significant effect on the learning retention. The group who received the terminal feedback was able to maintain the learning effect much better than the group who received concurrent feedback over time. Results from the present study showed that subjects with normal voices learned to phonate with relatively more relaxed orofacial muscles when terminal feedback of the thyrohyoid muscle activities was provided. The feedback from the thyrohyoid area appeared to have relatively little effect on learning to relax the thyrohyoid muscles themselves in phonation. It may well be that the control of the orofacial muscle was easier to control than the thyrohyoid muscles. These findings provide some preliminary supports for the application of the motor learning principle in voice therapy. There is therefore a need to empirically test the other wellestablished principles of motor learning, such as distribution of practice, role of demonstrations and the role of feedback types and frequency, in voice motor tasks, so that we can develop evident-based practice for voice therapy. 19 Motor Voice Learning References Adams, J. A. (1987). Historical Review and Appraisal of Research on the Learning, Retention, and Transfer of Human Motor-Skills. Psychological Bulletin, 101(1), 4174. Allen, K. D., Bernstein, B., & Chait, D. H. (1991). EMG biofeedback treatment of pediatric hyperfunctional dysphonia. Journal of Behavior Therapy and Experimental Psychiatry, 22(2), 97-101. Andrews, S., Warner, J., & Stewart, R. (1986). EMG biofeedback and relaxation in the efficacy treatment of hyperfunctional dysphonia. Journal of Disorders of Communication, 21, 353-369. Baken, R. J., & Orlikoff, R. F. (2000). Clinical measurement of speech and voice. San Diego: Singular Publishing Group. Bryman, A., & Cramer, D. (1997). Concepts and their measurement, Quantitative data analysis, with SPSS for Windows. London: Routledge. Coakes, S. J., & Steed, L. G. (2001). SPSS: Analysis without anguish. Version 10.0 for Windows. Brisbane: John Wiley & Sons. Ho, K. C. (1993). A Comparison of the 2,000 Most Frequently Used Chinese Characters Found in Three Frequency Counts Carried Out in Chinese, Taiwan, and Hong Kong. Report of Institute of Language in Education. Hong Kong: Institute of Language in Education. Hocevar-Boltezar, I., Janko, M., & Zargi, M. (1998). Role of surface EMG in diagnostics and treatment of muscle tension dysphonia. Acta Otolaryngologica, 118(5), 739-743. Park, J. H., Shea, C. H., & Wright, D. L. (2000). Reduced-frequency concurrent and terminal feedback: A test of the guidance hypothesis. Journal of Motor Behavior, 32(3), 287296. Redenbaugh, M. A., & Reich, A. R. (1989). Surface EMG and Related Measures in Normal and Vocally Hyperfunctional Speakers. Journal of Speech and Hearing Disorders, 54(1), 68-73. Schmidt, R. A., & Lee, T. D. (1999). Motor control and learning: A behavioral emphasis. Champaign, Illinois: Human Kinetics. Schmidt, R. A., & Wulf, G. (1997). Continuous concurrent feedback degrades skill learning: implications for training and simulation. Human Factors, 39(4), 509-525. Schmidt, R. A., Young, D. E., Swinnen, S., & Shapiro, D. C. (1989). Summary knowledge of results for skill acquisition: support for the guidance hypothesis. Journal of Experimental Psychology. Learning, Memory, and Cognition, 15(2), 352-359. Steinhauer, K., & Grayhack, J. P. (2000). The role of knowledge of results in performance and learning of a voice motor task. Journal of Voice, 14(2), 137-145. 20 Motor Voice Learning Stemple, G. J., Weiler, E., Whitehead, W., & Komray, R. (1980). EMG biofeedback training with patients exhibiting a hyperfunctional voice disorder. Laryngoscope, 90, 471-476. Verdolini, K., & Lee, T. D. (2001). Optimizing motor learning in speech intervention: Theory and practice. In C. Sapienza & J. Casper (Eds.), For clinician by clinician: Vocal rehabilitation in medical speech-language pathology. Austin, Texas.: Pro-Ed. Authors’ Note The first author (EY) would like to gratefully acknowledge the influence and inspiration of Professor Bill Hardcastle on his research career. Without the enthusiastic help and encouragement from Professor Hardcastle, the first author would not have appreciated the importance and usefulness of clinical instrumentation in speech science. The authors would like to acknowledge that this project was supported in part by a grant from Leung Kau Kui/Run Run Shaw Research and Teaching Endowment Fund. 2001. We would like to acknowledge Dr Estella Ma for her generous advice on the data analysis procedures. 21 Motor Voice Learning Table 1. Mean EMG values, inter-subject variability and mean intra-subject variability across the four sites of electrode placement. Vowels Mean EMG (V) Inter-subject Mean intra-subject variability variability Orofacial (Cheek) area /a/ 15.35 4.55 2.21 /i/ 24.09 10.21 5.73 /u/ 18.07 4.92 2.30 Overall average 19.17 6.56 3.41 Suprahyoid (Submandible) area /a/ 36.36 18.04 6.67 /i/ 46.25 34.72 7.29 /u/ 31.38 7.84 7.28 Overall average 38.00 20.20 7.08 /a/ 37.60 27.01 2.31 /i/ 54.13 30.54 7.30 /u/ 30.74 10.24 3.36 Overall average 40.82 22.60 4.32 /a/ 17.08 4.20 4.11 /i/ 18.95 6.49 3.34 /u/ 17.04 4.79 5.06 Overall average 17.69 5.16 4.17 Lower facial (chin) area Thyrohyoid (larynx) area 22 Motor Voice Learning Table 2. The 24 target words used in the reading task. Target IPA Order of Target IPA Order of stimuli symbol frequency stimuli symbol frequency based on Ho based on Ho (1993) (1993) 1 的 tik55 1 13 情 tshi21 176 2 不 pt55 4 14 每 mui23 196 3 有 ju23 5 15 月 jyt22 216 4 在 tsi22 6 16 教 kau33 231 5 了 liu23 7 17 老 lou23 239 6 我 23 9 18 片 phin33 246 7 為 wi21 10 19 給 khp55 259 8 這 ts35 11 20 男 nam21 328 9 水 sy35 75 21 父 fu22 332 10 起 hei35 104 22 卻 khk33 461 11 解 kai35 117 23 談 tham21 464 12 果 kw35 171 24 群 kwhn21 716 23 Motor Voice Learning Figure captions Figure 1. Electrodes for the electromyography. Figure 2. Electromyographic display. Figure 3. Sites for surface electrode placement. Figure 4. Amplitude mean and standard deviation of thyrohyoid electromyography. Figure 5. Amplitude mean and standard deviation of orofacial electromyograph. Figure 6. Mean percentage change of electromyographic activities relative to baseline measurements at the thyrohyoid area. Figure 7. Mean percentage change of electromyographic activities relative to baseline measurements at the orofacial area. 24 Motor Voice Learning Figure 1. Electrodes for the electromyography 25 Motor Voice Learning Figure 2. Electromyographic display as biofeedback for the subjects 26 Motor Voice Learning Orofacial area Lower facial area Suprahyoid area Thyrohyoid area Figure 3. Sites for surface electrode placement 27 Motor Voice Learning Baseline (B1-B2) blocks Training (T1-T5) blocks Immediate Retention (IR1-IR2) blocks Delayed Retention (DR1-DR2) blocks Figure 4. Amplitude mean and standard deviation of thyrohyoid electromyograph 28 Motor Voice Learning Baseline (B1-B2) blocks Training (T1-T5) blocks Immediate Retention (IR1-IR2) blocks Delayed Retention (DR1-DR2) blocks Figure 5. Amplitude mean and standard deviation of orofacial electromyograph 29 Motor Voice Learning Thyrohyoid site Percentage change relative to baseline-- 50 40 30 20 Concurrent Terminal 10 0 -10 Training Immediate Retention Delayed Retention -20 Figure 6. Mean percentage change of electromyographic activities relative to baseline measurements at the thyrohyoid area. 30 Motor Voice Learning Orofacial site Percentage change relative to baseline-- 0 Training Immediate Retention Delayed Retention -10 -20 Concurrent Terminal -30 -40 -50 Figure 7. Mean percentage change of electromyographic activities relative to baseline measurements at the orofacial area. 31 Motor Voice Learning Figures and tables from this page onward are for own references only…. Surface electrode placement over the oral facial and neck regions (Incorporate these into the figure) Details of the Electrode Placements and Their Associated Muscles for Investigation Area Site of electrodes Associated muscles Orofacial area On the cheek, 2 cm away from the Buccinator, Risorius corner of the mouth Lower facial area On the chin near the midline, 1 cm Mentalis apart between the 2 elecctrodes Suprahyoid area Over the submandible region and Digastric, Mylohyoid, above the hyoid bone, 2 cm apart Stylohyoid, Platysma between the 2 electrodes Thyrohyoid area Over the thyrohyoid membrane near Thyrohyoid, Omohyoid, the midline, 1 cm apart between the Sternohyoid, Platysma 2 electrodes Note. The sites of the lower facial, suprahyoid and thyrohyoid areas are adapted from Hocevar-Boltezar et al., 1998. 32 Motor Voice Learning Table ?. Mean (standard deviation) microvolts of laryngeal muscle tension at the thyrohyoid site. Baseline Acquisition Block CONC TERM Immediate Delayed Retention Retention Block Block Block 1 2 14.80 14.89 14.38 13.04 12.11 12.16 11.90 (7.49) (6.82) (7.52) (5.41) (6.13) (5.66) (5.12) 20.39 22.09 24.70 25.89 25.38 23.07 24.14 (12.45) (13.38) 1 2 3 4 5 (15.63) (16.81) (17.42) (17.81) (22.41) CONC - concurrent feedback group TERM - terminal feedback group. 33 1 2 17.19 21.22 (11.41) (18.00) 26.59 25.76 (25.62) (21.77) 1 2 13.87 14.09 (6.62) (5.16) 17.77 15.94 (12.40) (9.92) Motor Voice Learning Table ?. Mean and (standard deviation) microvolts of orofacial muscle tension at the cheek. Baseline Acquisition Block CONC TERM 1 2 23.90 22.78 Immediate Delayed Retention Retention Block Block Block 1 2 1 2 13.21 14.08 14.06 15.76 15.50 12.97 (9.98) (10.78) (8.92) (7.29) (4.53) (6.05) (6.64) (6.32) (7.13) (6.33) (6.39) 27.59 25.62 20.21 18.16 16.30 15.49 15.41 15.81 17.06 15.89 (6.88) (6.33) (5.23) (4.63) (4.24) (3.70) (4.45) (4.27) (4.99) (7.11) (6.10) 15.30 14.19 3 13.90 17.69 4 CONC - concurrent feedback group TERM - terminal feedback group. 34 5 1 2