Journalof Speech and HearingResearch, Volume 35, 1216-1229, December 1992
Gradations in a Pattern of Neuromuscular Activity Associated With Stuttering Margaret Denny Anne Smith Department of Audiology and Speech Sciences Purdue University West Lafayette, IN EMGs were recorded from muscles of the lip, jaw, and neck during conversational speech of 17 stuttering subjects. Averaged power spectra and coherence between pairs of EMGs were computed. Results indicate that tremorlike oscillations in the range of 5-15 Hz and high amplitudes of EMGs occupy a common continuum of motor patterns that may occur instuttering. In subjects whose results fell at the strong end of this continuum, stuttered speech was distinguished by widely distributed, high-amplitude oscillations and relatively high coherence at the frequency of oscillation. At the other extreme, neither oscillatory activity nor amplitude was greater for stuttered speech; in fact stuttered and fluent speech were often associated with approximately equal EMG amplitude. These results suggest that there is not a single set of physiological events that uniformly characterize stuttering in all individuals; rather, events such as the occurrence of high-amplitude oscillations occur at different strengths in different individuals. KEY WORDS: stuttering, motor control, tremor, EMG, speech
A basic tenet of research on the neural control of movement is that in order to investigate and ultimately identify the neural sources that produce a motor output, a careful description of what motor output actually occurs (e.g., movements and muscle activity) is necessary. Numerous investigators who have applied this logic to experimental inquiry on the physiological correlates of stuttering have reported that stuttered speech is associated with excessive levels of muscle activity (Freeman & Ushijima, 1978; Kalotkin, Manschreck, & O'Brien, 1979; Shapiro, 1980; Williams, 1955) and tremorlike movements and/or modulations of EMG activity in muscles of the face, tongue, lip, jaw, larynx, and respiratory system (Fibiger, 1971; Freeman & Ushijima, 1975; McClean, Goldsmith, & Cerf, 1984; Morley, 1937; Platt & Basili, 1973; Smith, 1989; Travis, 1927, Zimmermann, 1980a). These tremorlike phenomena have been demonstrated insubjects exhibiting a broad range of disfluent behaviors (Smith, 1989), and have been reported to occur within a frequency band (5-15 Hz) overlapping that of normal physiological tremor. Because the contractile elements of muscles and their loads behave like low-pass filters (for review see Partridge & Benton, 1981), it is possible that oscillatory modulations of EMGs within the 5-15-Hz band often would not be visible as movements. Such oscillations may be obvious with visual inspection of EMG recordings, but only if the firing rates of a high proportion of active motor units were modulated at a common frequency. Oscillatory modulation of activity of smaller subpopulations of motor units might not be visually apparent. For these reasons, spectral analyses of EMGs are necessary to characterize adequately the nature of tremorlike oscillations in stuttering. Smith (1989) reported results of spectral analyses of EMGs recorded from a variety of muscles of the lip, jaw, and neck in normally fluent and stuttering subjects. © 1992, American Speech-Language-Heanng Association
1216
0022-4685/2/3506-1216$01.00/0
Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929152/ by a Carleton University Library User on 07/21/2017 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Denny & Smith: NeuromuscularActivity in Stuttering
Averaged power spectra of the amplitude envelopes of EMGs were computed from records of the fluent speech of normal subjects and the disfluent speech of stutterers. The EMG spectra for the normally fluent subjects always had peak amplitudes at or below 4 Hz, probably reflecting EMG modulation at the speaker's rate of syllable production. In contrast, for 6 of the 10 stuttering subjects peak amplitudes of at least one of the EMG spectra occurred within the 5-15-Hz band. A striking finding of the Smith (1989) study was that, for subjects who showed these higher-frequency peaks in EMG spectra from more than one muscle, the peak amplitudes often occurred at the same frequency for the different muscles. Smith (1989) suggested that during disfluency, some common oscillatory source could be driving muscles that normally served a variety of functions (e.g., lip protrusion, jaw-opening, and jaw-closing). Inspection of the data gathered by Smith suggested that, inat least some subjects, these spatially distributed oscillations and high-amplitude EMG activity tended to co-occur over a consistent time course. Figure 1a and b show EMGs, audio, and lip position traces recorded from 2 stuttering subjects before, during, and after episodes of perceptibly disfluent speech. The beginning and end of each perceptibly disfluent episode are marked by vertical lines. Averaged EMG spectra computed by Smith (1989) for these 2 subjects are shown to the right of the time-domain traces in panels C and D. For the subject shown in Figure 1a, stuttering was usually accompanied by the pattern of muscle activation shown here. The elements of this pattern are (a) tremorlike oscillations in EMG activity (the appearance of oscillations is marked by solid arrows in Figure 1a); (b) common frequency of oscillation across different muscles, as shown in the spectra of Figure 1c; and (c) a common time course across muscles for the development of high-amplitude, oscillatory EMG (oscillations in EMGs occur between the vertical arrows positioned over the top EMG trace in Figure la). When tremorlike oscillations first appear, the amplitudes of the EMGs are relatively small. As long as the oscillations continue to be present, the amplitudes of the EMGs increase dramatically. These high-amplitude oscillations cease abruptly in all EMG channels shortly before fluent speech is resumed. None of the features of this pattern are visually apparent in Figure b, which shows data from a different stuttering subject. For this subject, EMGs appeared quite similar during fluent and disfluent speech. However, as shown in Figure 1d, the averaged spectra from two of these EMG channels (LLIP and ABD) had relatively large peaks in the 5-15-Hz band. Unlike the data shown in Figure la, these oscillations did not occur at a common frequency. The possibility that other elements of the pattern discussed above (e.g., timing or amplitude) may have been present during this subject's disfluencies could not be addressed without further analyses. Observations such as these suggest that tremorlike oscillations may not be best described as simply present or absent during stuttering; rather, the study of such oscillations may reveal a continuum of aberrant motor output patterns occurring during stuttered speech. Thus in some people who stutter, oscillations in the 5-15-Hz band are accompanied by all elements of the pattern illustrated in Figure la; for others, only some elements of the pattern are present, whereas still others show no evidence of any aspect of the pattern.
1217
Support for this view has come from the results of Smith, Denny, and Wood (1991). Smith et al. examined the time course of changes in the amplitude of spectral power in the 5-15-Hz band during disfluent speech. The hypothesis tested by Smith et al. was as follows: If the time course of tremorlike oscillation illustrated in Figure 1a were common in people who stutter, overall amplitude and/or power within the 5-15-Hz band should be greater during the disfluency than during temporally adjacent intervals, and greater in the final 0.5 sec of disfluency than in the initial 0.5 sec of disfluency. Thus, four 0.5-sec intervals were selected for analysis: the intervals immediately preceding and following episodes of perceptible disfluency and the initial and final intervals within disfluent episodes. Spectral power in the 5-15-Hz band and overall EMG amplitude were measured. Smith et al. reported that whereas a minority of stuttering subjects consistently conformed to the predicted time course, there was no tendency for stuttering subjects as a group to display the predicted differences in amplitude over time. Thus, the predicted time course appears to be an optional aspect of the pattern. The results reported by Smith et al. (1991) included two surprising findings. First, within-subject statistical analyses revealed that power in the 5-15-Hz band tended to be greater in the 0.5-sec interval immediately following a disfluency. This finding raised the possibility that tremorlike oscillations may reflect some ever-present instability in the speech motor systems of people who stutter, and thus their amplitude and time course may not be tied to the occurrence of perceptible disfluency. If this is the case, EMG spectra of stutterers' perceptually fluent speech, even intervals that are temporally distant from disfluent intervals, should show peaks in the 5-15-Hz band similar to those observed by Smith (1989) for disfluent speech of people who stutter. Alternatively, tremorlike oscillations may be exclusively associated with perceptible disfluency, but may not be strictly limited to the perceptibly disfluent interval. If this is the case, the co-occurrence of oscillations and perceptually fluent speech would be limited to intervals immediately surrounding disfluency. The second unexpected finding reported by Smith et al. was that there were no consistent differences in overall amplitude between disfluent intervals and the adjacent fluent intervals. In light of previous reports that stuttered speech is associated with relatively high levels of EMG activity (Freeman & Ushijima, 1978; Kalotkin et al., 1979; Shapiro, 1980; Williams, 1955), it would be expected that the intervals within the disfluency would be of consistently greater amplitude than the intervals prior to and following the disfluency. This was not the case. It may be that, as hypothesized for oscillations in the 5-15-Hz band, high levels of EMG activity are present during both the perceptible disfluency and its immediate temporal surround. The present study was designed to address questions raised by the results of Smith et al. (1991) by making within-subject comparisons between the perceptually fluent and disfluent speech of people who stutter. Two specific questions to be addressed are (a) Is stuttered speech associated with high levels of EMG activity relative to perceptually fluent speech? (b) Are tremorlike oscillations in EMGs present in perceptually fluent as well as stuttered speech?
Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929152/ by a Carleton University Library User on 07/21/2017 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1218 Journalof Speech and Hearing Research
35
1216-1229
December 992
C
A
MASS
10tov
0
ABD 10 AV
I
LLIP 50 VvI
POS 5 mr
I
AUD
0
2
4
6
8
10
12
14
16
18
20
14
16
18
20
Frequency (Hz)
D
B MP 25 AV
*14IX
Q) 4-
JniM1h~VMOU 50A
I
0-J~~
POS
AUD
E
Gradations in a Pattern of Neuromuscular Activity Associated With Stuttering Margaret Denny Anne Smith Department of Audiology and Speech Sciences Purdue University West Lafayette, IN EMGs were recorded from muscles of the lip, jaw, and neck during conversational speech of 17 stuttering subjects. Averaged power spectra and coherence between pairs of EMGs were computed. Results indicate that tremorlike oscillations in the range of 5-15 Hz and high amplitudes of EMGs occupy a common continuum of motor patterns that may occur instuttering. In subjects whose results fell at the strong end of this continuum, stuttered speech was distinguished by widely distributed, high-amplitude oscillations and relatively high coherence at the frequency of oscillation. At the other extreme, neither oscillatory activity nor amplitude was greater for stuttered speech; in fact stuttered and fluent speech were often associated with approximately equal EMG amplitude. These results suggest that there is not a single set of physiological events that uniformly characterize stuttering in all individuals; rather, events such as the occurrence of high-amplitude oscillations occur at different strengths in different individuals. KEY WORDS: stuttering, motor control, tremor, EMG, speech
A basic tenet of research on the neural control of movement is that in order to investigate and ultimately identify the neural sources that produce a motor output, a careful description of what motor output actually occurs (e.g., movements and muscle activity) is necessary. Numerous investigators who have applied this logic to experimental inquiry on the physiological correlates of stuttering have reported that stuttered speech is associated with excessive levels of muscle activity (Freeman & Ushijima, 1978; Kalotkin, Manschreck, & O'Brien, 1979; Shapiro, 1980; Williams, 1955) and tremorlike movements and/or modulations of EMG activity in muscles of the face, tongue, lip, jaw, larynx, and respiratory system (Fibiger, 1971; Freeman & Ushijima, 1975; McClean, Goldsmith, & Cerf, 1984; Morley, 1937; Platt & Basili, 1973; Smith, 1989; Travis, 1927, Zimmermann, 1980a). These tremorlike phenomena have been demonstrated insubjects exhibiting a broad range of disfluent behaviors (Smith, 1989), and have been reported to occur within a frequency band (5-15 Hz) overlapping that of normal physiological tremor. Because the contractile elements of muscles and their loads behave like low-pass filters (for review see Partridge & Benton, 1981), it is possible that oscillatory modulations of EMGs within the 5-15-Hz band often would not be visible as movements. Such oscillations may be obvious with visual inspection of EMG recordings, but only if the firing rates of a high proportion of active motor units were modulated at a common frequency. Oscillatory modulation of activity of smaller subpopulations of motor units might not be visually apparent. For these reasons, spectral analyses of EMGs are necessary to characterize adequately the nature of tremorlike oscillations in stuttering. Smith (1989) reported results of spectral analyses of EMGs recorded from a variety of muscles of the lip, jaw, and neck in normally fluent and stuttering subjects. © 1992, American Speech-Language-Heanng Association
1216
0022-4685/2/3506-1216$01.00/0
Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929152/ by a Carleton University Library User on 07/21/2017 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Denny & Smith: NeuromuscularActivity in Stuttering
Averaged power spectra of the amplitude envelopes of EMGs were computed from records of the fluent speech of normal subjects and the disfluent speech of stutterers. The EMG spectra for the normally fluent subjects always had peak amplitudes at or below 4 Hz, probably reflecting EMG modulation at the speaker's rate of syllable production. In contrast, for 6 of the 10 stuttering subjects peak amplitudes of at least one of the EMG spectra occurred within the 5-15-Hz band. A striking finding of the Smith (1989) study was that, for subjects who showed these higher-frequency peaks in EMG spectra from more than one muscle, the peak amplitudes often occurred at the same frequency for the different muscles. Smith (1989) suggested that during disfluency, some common oscillatory source could be driving muscles that normally served a variety of functions (e.g., lip protrusion, jaw-opening, and jaw-closing). Inspection of the data gathered by Smith suggested that, inat least some subjects, these spatially distributed oscillations and high-amplitude EMG activity tended to co-occur over a consistent time course. Figure 1a and b show EMGs, audio, and lip position traces recorded from 2 stuttering subjects before, during, and after episodes of perceptibly disfluent speech. The beginning and end of each perceptibly disfluent episode are marked by vertical lines. Averaged EMG spectra computed by Smith (1989) for these 2 subjects are shown to the right of the time-domain traces in panels C and D. For the subject shown in Figure 1a, stuttering was usually accompanied by the pattern of muscle activation shown here. The elements of this pattern are (a) tremorlike oscillations in EMG activity (the appearance of oscillations is marked by solid arrows in Figure 1a); (b) common frequency of oscillation across different muscles, as shown in the spectra of Figure 1c; and (c) a common time course across muscles for the development of high-amplitude, oscillatory EMG (oscillations in EMGs occur between the vertical arrows positioned over the top EMG trace in Figure la). When tremorlike oscillations first appear, the amplitudes of the EMGs are relatively small. As long as the oscillations continue to be present, the amplitudes of the EMGs increase dramatically. These high-amplitude oscillations cease abruptly in all EMG channels shortly before fluent speech is resumed. None of the features of this pattern are visually apparent in Figure b, which shows data from a different stuttering subject. For this subject, EMGs appeared quite similar during fluent and disfluent speech. However, as shown in Figure 1d, the averaged spectra from two of these EMG channels (LLIP and ABD) had relatively large peaks in the 5-15-Hz band. Unlike the data shown in Figure la, these oscillations did not occur at a common frequency. The possibility that other elements of the pattern discussed above (e.g., timing or amplitude) may have been present during this subject's disfluencies could not be addressed without further analyses. Observations such as these suggest that tremorlike oscillations may not be best described as simply present or absent during stuttering; rather, the study of such oscillations may reveal a continuum of aberrant motor output patterns occurring during stuttered speech. Thus in some people who stutter, oscillations in the 5-15-Hz band are accompanied by all elements of the pattern illustrated in Figure la; for others, only some elements of the pattern are present, whereas still others show no evidence of any aspect of the pattern.
1217
Support for this view has come from the results of Smith, Denny, and Wood (1991). Smith et al. examined the time course of changes in the amplitude of spectral power in the 5-15-Hz band during disfluent speech. The hypothesis tested by Smith et al. was as follows: If the time course of tremorlike oscillation illustrated in Figure 1a were common in people who stutter, overall amplitude and/or power within the 5-15-Hz band should be greater during the disfluency than during temporally adjacent intervals, and greater in the final 0.5 sec of disfluency than in the initial 0.5 sec of disfluency. Thus, four 0.5-sec intervals were selected for analysis: the intervals immediately preceding and following episodes of perceptible disfluency and the initial and final intervals within disfluent episodes. Spectral power in the 5-15-Hz band and overall EMG amplitude were measured. Smith et al. reported that whereas a minority of stuttering subjects consistently conformed to the predicted time course, there was no tendency for stuttering subjects as a group to display the predicted differences in amplitude over time. Thus, the predicted time course appears to be an optional aspect of the pattern. The results reported by Smith et al. (1991) included two surprising findings. First, within-subject statistical analyses revealed that power in the 5-15-Hz band tended to be greater in the 0.5-sec interval immediately following a disfluency. This finding raised the possibility that tremorlike oscillations may reflect some ever-present instability in the speech motor systems of people who stutter, and thus their amplitude and time course may not be tied to the occurrence of perceptible disfluency. If this is the case, EMG spectra of stutterers' perceptually fluent speech, even intervals that are temporally distant from disfluent intervals, should show peaks in the 5-15-Hz band similar to those observed by Smith (1989) for disfluent speech of people who stutter. Alternatively, tremorlike oscillations may be exclusively associated with perceptible disfluency, but may not be strictly limited to the perceptibly disfluent interval. If this is the case, the co-occurrence of oscillations and perceptually fluent speech would be limited to intervals immediately surrounding disfluency. The second unexpected finding reported by Smith et al. was that there were no consistent differences in overall amplitude between disfluent intervals and the adjacent fluent intervals. In light of previous reports that stuttered speech is associated with relatively high levels of EMG activity (Freeman & Ushijima, 1978; Kalotkin et al., 1979; Shapiro, 1980; Williams, 1955), it would be expected that the intervals within the disfluency would be of consistently greater amplitude than the intervals prior to and following the disfluency. This was not the case. It may be that, as hypothesized for oscillations in the 5-15-Hz band, high levels of EMG activity are present during both the perceptible disfluency and its immediate temporal surround. The present study was designed to address questions raised by the results of Smith et al. (1991) by making within-subject comparisons between the perceptually fluent and disfluent speech of people who stutter. Two specific questions to be addressed are (a) Is stuttered speech associated with high levels of EMG activity relative to perceptually fluent speech? (b) Are tremorlike oscillations in EMGs present in perceptually fluent as well as stuttered speech?
Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929152/ by a Carleton University Library User on 07/21/2017 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1218 Journalof Speech and Hearing Research
35
1216-1229
December 992
C
A
MASS
10tov
0
ABD 10 AV
I
LLIP 50 VvI
POS 5 mr
I
AUD
0
2
4
6
8
10
12
14
16
18
20
14
16
18
20
Frequency (Hz)
D
B MP 25 AV
*14IX
Q) 4-
JniM1h~VMOU 50A
I
0-J~~
POS
AUD
E
