- Email: [email protected]
Sensorimotor actions in the control of multi-movement speech gestures
391
TINS - September 1983
Sensorimotor actions in the control of multi-movement speech gestures James H. Abbs and Vincent L. Gracco The generation o f speech is perhaps the most characteristic human motor act. Historically, owing to the complexity o f speech motor behavior, most analyses have provided only general information on the underlying neural control mechanisms. However, recent work offers some intriguing neurophysiological insights. These studies indicate that: (1) sensorimotor actions are active in minimizing speech movement errors; (2) both intra- and intermovement sensorimotor mechanisms are involved; (3) speech motor-afferent-dependent actions are not stereotypic but have control characteristics that vary, moment-to-moment, in relation to motor execution and task variables; and (4) co-ordination among multiple speech movements is not wholly prespecified, but is subject to dynamic refinement via intermovement sensorimotor adjustments. These observations appear to offer some useful hypotheses f o r the neural control o f other complex motor behaviors.
Most physiological investigations on the While some research groups have focused neural mechanisms of voluntary motor con- on the sensorimotor actions underlying trol have focused on actions around a single maintenance of three-dimensional postural joint. Analyses of sensorimotor responses equilibrium, parallel efforts aimed at 'less to load perturbations similarly have been automatic', multijoint limb movement are confined primarily to single-joint control just beginning to emerge 4'~o.~4,2o. processes". This extensive work has In the last few years, scientists at several yielded numerous insights into the sen~ri- laboratories have utilized perturbation motor control of movement. However, techniques to study sensorimotor control of many motor behaviors in nature are not the multiple co-ordinated actions of human restricted to movements around a single speech 2.3,8`9`11.1z,19,zt. Results of these joint, but require complex actions among studies appear to offer some interesting and multiple movements and multiple muscles. new perspectives on the role of sensori-
~
n
motor mechanisms in the control of complex movements. Even simple speech gestures require co-ordinated movements of the lips, tongue, jaw, pharynx, larynx and respiratory system. Fig. 1 illustrates the lip movements and associated facial muscle contractions (EMG) for the speech gesture 'aba'. As shown, there are several muscles involved in producing these two coordinated movements. The lower-lip muscles, orbicularis oris inferior and mentalis, act to elevate the lower lip while the upperlip muscles, orbicularis otis superior and depressor anguli otis, act to depress the upper lip. The depressor labii inferior acts to depress the lower lip. While it has been suggested that the orbicularis otis muscle is a single sphinctor, electromyographic measures indicate that the inferior and superior portions are controlled independently ~,17. As such, one might consider this speech pattern as analogous to the co-ordinated actions of the elbow and wrist in a multi-joint limb gesture. Perturbation analyses of speech m o v e m e n t control
As noted, perhaps the critical test of whether motor actions are under the influence of ascending afferent signals is to introduce an unanticipated error and examine the adjustments that occur in response to that error. The importance of afferent mechanisms in the control of multimovement speech gestures is apparent from experiments in which unanticipated perturbation analysis has been used.
~
i
Otis
UPPER-LIP DEPRESSORS
UPPER-LIP
'a
b
a' SUPERIOR
MOVEMENTS
LOWER-LIP
~
INFERIOR
lnf.
LOW R-L ' DEPRESSOR
Fig. 1. Oscillographicrecordof upper-lip and lower-lipmovement and associatedfacialmuscleactivityfor the speechgesture 'aba'. Thejaw rnovementis eliminated in this example by having the subject bite on a block of rubber materialplaced between the molars. EMG signalsare rectifiedand smoothed. ~) 1983, Elsevier Science Publishers B.V., Amslerdafa
0378 - 5912/83/$01,00
'I'INA" o 3 c p w m b e r
392 Results from perturbations applied to the two-movement gesture shown in Fig. I provide a useful illustration. Fig. 2 indicates how a d.c. brushless torque motor has been interfaced to the lower lip to introduce unanticipated perturbations during speech. Rotation of the torque-motor sector arm is transmitted to the lower lip via a steel wire and a low-friction lever. A similar configuration has been used to apply loads to the upper lip or jaw. As noted, the torque motor was under force-feedback control, permitting the motor to follow lower- lip, upper- lip or jaw movements at a low, constant tracking force of approximately 4 g. In the experiments involving lower-lip perturbations, inferiorly directed loads ranging from 10 to 40 g were superimposed upon the constant tracking force. These forces, and the displacement perturbations they produced, were within the physiological range of forces and displacements generated during normal lip movements for speech. In these particular experiments, loads were applied to the lower lip in the lO0 ms time window ranging from 50 ms before orbicularis oris inferior contraction to 50 ms into that contraction. To ensure that perturbations were unanticipated, loads were applied randomly on only 10-15% of the movement trials. More frequent perturbations appeared to cause adaptation in the form of a general 'stiffening' of the lips. The major observations from these lower-lip perturbation experiments (illustrated in Figs 3 and 4) indicate the operation of sensorimotor actions. Fig. 3 provides representative examples of upper-lip and lower-lip movement for three control and three loaded conditions. The magnitude and velocity of both upper-lip and lower-lip movements are increased in response to the induced lower-lip perturbation. Fig. 4 shows EMG and upper-lip movement recordings superimposed for a pair of control and loaded trials. The load magnitude for this example was 40 g, and the load was introduced approximately 20 ms prior to the onset of the phasic burst of EMG in the orbicular& oris inferior muscle. Both upper-lip and lower-lip muscles show relatively short latency changes (30-40 ms) in response to the unanticipated load. That is, while the perturbation was applied to the lower lip, compensations were seen in both upper-lip and lower-lip muscles:L The latencies of these compensatory responses are not stereotypic, but range from 25 to 75 ms, depending on the muscle observed and the timing of the load onseP 2. In the various studies where unanticipated perturbations have been applied to speech movements, several remarkable findings have been consistently reported. First, despite some sizeable movement per-
I t/.~j
Adjustable Head M o u n t
Low-
Lip Paddle
Friction Pivot
Torque.
Lood
i
Under Force Feedback Fig. 2. The instrumental configuration employed in applying unanticipated loads to the lower lip during speech. A similar apparatus has been employed to perturb upper-lip and jaw movements.
turbations (> 15 mm) in some studies, the intended speech motor objective is not disrupted in a discernible way (i.e., a listener cannot distinguish the acoustic speech patterns of loaded trials from those generated during unloaded trials). Second, compensations such as those shown in Figs 3 and 4 are observed the first time a load is introduced. These observations would appear to indicate that these sensorimotor compensations reflect 'normal' processes of speech movement control, Other results indicating that these sensorimotor processes are utilized in natural situations (outside the laboratory) are from studies where load magnitudes were limited to 10-15 g, producing displacement perturbations as small as 1 mm (Ref. 2). Even with these small perturbations, the pattern of EMG and movement responses was comparable to those illustrated in Figs 3 and 4. The interpretation that these sensorimotor actions are indicative of 'normal,, on-line control processes is also reflected in verbal reports of the subjects. For example, subjects have typically noted that the loads did not bother them - they simply 'ignored them'. Also, if subjects were instructed to stop all lip movement when they sensed a load, cessation of activity was only accomplished after compensation had occurred and lip closure was achieved. As noted, these results are representative of those from other comparable studies of
speech sensorimotor actions. For example~ when .law loads are introduced during a combined jaw-lip movement gesture (a "p' sound), compensatory responses are observed in both the upper and lower lipss'9. Similarly, upper-lip loads yield both upper-lip and lower-lip compensatory responses. In other experiments, jaw loads introduced during a jaw-tongue gesture (a 'z' sound) yielded compensatory responses in the tongue elevation muscles~. As noted, in all of t ~ experinaents sensorimotor compensations were manifest the first time a load was introduced: the intended speech patterns were never disrupted. These results are consistent with the seemingly unremarkable observations that lip, tongue, and pharyngeal movements are reorganized for adequate speech when the jaw is fixed, as in pipe smoking 21
Neural control processes underlying multlmovement sltet~h gestures In interpreting the results from these various perturbation experiments, one must be cautious in the use of concepts like closedloop or feedback. That is. while an error is introduced in the intended movement of one structure (for example the lower lip), significant compensations are consistently observed in synergistic movements of other structures (for example the upper lip). As such, the compensatory responses cannot be ascribed to a closed-loop, feedback pro-
393
T I N S - September 1 983
cess, especially as these multiple structures are controlled independently. More specifically, the compensatory upper-lip muscle actions do not reflect the comparator-based reactive adjustments characteristic of a feedback system; these adjustments do not correct the induced error in the lower lip. It is possible to demonstrate that the muscles of the upper and lower lips are independently controlled by observing responses to lower-lip loads applied during different speech movement patterns (for example during generation of the sounds 'aba' and 'afa'). For 'aba', both upper-lip and lowerlip movements normally are employed to achieve lip closure, and upper-lip compensation occurs when the lower lip is loaded. By contrast, for 'afa', upper-lip movement is not required, and there is no compensatory upper-lip movement when the lower lip is loaded. In both of these gestures, however, lower-lip compensatory responses to the applied loads are clearly discernible. Given these latter results, it is apparent that when an error is introduced in the lower-lip movement and compensations are observed in both the upper-lip and lower-lip muscles, at least two different sensorimotor processes are involved. As represented in Fig. 5, the compensations in the lower lip appear to be indicative of a classical closed-loop, feedback process. However, by contrast, based upon the obvious dissociation between the site of the induced error and the site of the compensation, the changes in the upper-lip muscles and movements indicate an open-loop adjustment. This open-loop adjustment apparently is based upon a pre-established sensorimotor translation between lower-lip afferent signals and upper-lip motor actions. This intermovement sensorimotor translation has been referred to as a predictive or feedforward processl.~'t~6; i.e., these adjustments are made without a reactive feedback comparator to ensure their corrective accuracy. Similar results have been obtained in studies of eye--head movement co-ordination 5.2a. In those experiments, sensorimotor interactions from neck muscle afferents and the vestibular system were shown to influence eye movements during co-ordinated eye-head movements, reflecting feedforward, open-loop sen~orimotor pathways 7.~a.22.26. From these recent studies of speech, it appears that these predictive adjustments, variously observed to be operating among the jaw, upper lip, lower lip, and tongue, likewise are afferent dependent. As with the results of eye-head movements, these sensorimotor processes may reflect neural mechanisms underlying temporo-spatial movement co-ordination. These interpretations are augmented by earlier observations, indicating a consider-
able degree of temporo-spatial covariance among the upper-lip, lower-lip, jaw, and tongue movements for speech gestures ~3.'7. For example, observation of movements for multiple speech repetitions requiring the same lip--jaw goal indicated that when jaw movements were large, corresponding lip movements were relatively small; when jaw movements were small, corresponding lip movements were relatively large. These observations of intermovement covariance suggest that the sensorimotor actions among independent structures, as revealed by perturbation analysis, may be utilized in a predictive, open-loop manner to generate normal compensatory adjustments and ensure that common multimovement goals are achieved 1. Yet other studies indicate functional differences between the closed-loop and open-loop sensorimotor responses to loads applied during speech gestures. These differences were discernible in the characteristics of load-induced responses in lower-lip muscles versus upper-lip muscles TM. For loads introduced after the onset of the orbicularis oris agonist burst, the open-loop (upper lip) compensations were more robust than the closed-loop (lower lip) compensations. Likewise, it was found that the latencies of the open-loop responses were
generally faster than those of the closedloop responses 12. One might expect classical limitations on closed-loop sensorimotor contributions (for example neural transport delays, consequent instability, etc.) to be minimized in open-loop control because the latter adjustments are predictive rather than reactive. These differences between openloop and closed-loop responses are thus consistent with the relatively greater speed and stability hypothesized for open-loop, feedforward processes 1". Moreover, differences between intra- and intermovement sensorimotor processes are consistent with the impression that the neural control mechanisms underlying co-ordination may be different from the mechanisms involved in controlling single movements. This distinction is most apparent in certain movement disorders where dyscoordination among multijoint actions appears as a disproportionately salient manifestation '5. Regarding the potential neural pathways underlying these sensorimotor control processes, the absence of a consistent, timelocked response pattern appears to mitigate against classical reflex mechanisms TM. Particularly notable is the absence of responses with latencies corresponding to lower brainstem reflexes (for example the perioral reflex has a latency of between I0
CONTROL i
WEN
LOADED
T
5 mm
2
°7V Fig. 3. Examples of upper-lip and lower-lip movements for unloaded (upper panels) and loaded (lower panels) trials.
11N3 - .~epwmber i 9,~5
394
Upper-Up (4
'~t
Depressors
is Superior E !
Upper-Up
~ 'a
b
a'
De~esNon~
Load T
~
_L LOAD ONSET
"
Control
200 ms ~
Lower_Up----------~~ ~~hm~-~--Inferior Elevators \
Mentalis
Fig. 4. Example of EMG and upper-lip movement for a pair of unloaded (light lines) and loaded (heavy lines) trials. Load onset is indicated by the vertical dashed line.
and 18 ms). These observations thus appear to implicate suprabulbar sensorimotor pathways, perhaps involving the extensive sensory and motor connections of the orofacial region with the cerebral cortex and the cerebellum. The monosynaptic connections from pyramidal tract cells to orofacial motoneurons also are consistent with this interpretationL Additionally, the dominance of suprabuibar motorsensory mechanisms in speech control fits the hypothesis of Phillips and Porterz" who argue that there are greater degrees of motorsensory cortex involvement in 'least automatic' behaviors.
Sununary Experiments on sensorimotor mecha, nisms in the control of speech movements
or via actions of single agonistic muscles. Rather, the results observed for these more complex, natural motor gestures indicate that intended nervous system outputs are achieved through a flexible interaction of several sensorimotor control processes, including dosed-loop correction and openloop predictionz~. It is tempting to hypothesize that analysis of co-ordination among the multiple joint movements for
~..~ ~ . ~
~
t
have yielded results that augment those reported in comparable analyses of move-
I
ments around a single joint. For example, earlier observations suggest that sensorimotor processes provide only moderate resistance to induced movement errors. However, when the motor task investigated has multiple degrees of compensatory freedom, the nervous system control variables appear very resistant to such disturbances, This resistance is not manifest simply through one-dimensional control processes
I
Reading list i Abbs. J. H and Cole. K. J (t982) m Speech Motor Control (Gdllner. S,. Lindblom. B.. Lubker. I and Persson. A., eds), pp 159-186,
MUSCLES
I
y
Upper-Lip Movement
(
I
OPEN-LOOP (Feedforward)
limb control would reveal comparable seu~ orimotor actions if investigated in a similar manner. Experiments by Polit and Bizzi~:' indicated that while single-joint lbrearm movements could be executed more or le~s successfully following deafferentatlon, even minor static adjustments of the shoulder (namely, involving a second joint) seriously disrupted forearm movement control in the absence of afferent information. Patterns of intermovement motor equivalence between elbow and wrist movements ~'' also are consistent with the hypothesis that limb co-ordination may be under the influence of open-loop, predictive, sensorimotor actions. From a more abstract perspective, these recent observations of speech indicate that multiple degrees of freedom in complex motor systems may facilitate the neural control process. That is, the freedom to accomplish the same intended motor objective in many different ways relieves the nervous system of the burden of having to prespecify all the complex details of the motor subgestures. Such details apparently are determined via sensonmotor processes operating dynamically and flexibly among the multiple subactions. From this point of view, it might even be argued that movements experimentally constrained around a single joint present a more difficult control problem to the nervous system than their apparently' more complex' multijoint counterparts. The learning and neural programming of a skilled motor behavior certainly would be more demanding if the criterion for successful performance was restricted to a single stereotyped pattern of muscle contraction and movement.
CORRECTIVE ADJI~TMENTS
k
t
INDUCED ERROR
I., !
Lower-Lip Movement
/ I II I t
I I
/
I It------ CLOSED -LOOPlFeedback)
I I !
Fig. 5. Schematic illustration of the hypothesized sensorimotor control pathways underlying compensatzons to lip perturbations.
T I N S - S e p t e m b e r 1 983
PergamonPress, London 2 Abbs, J. H. and Gracco, V. L. (1981)£ Acoust. Soc. Am. 70, $78 (Abstract) 3 Abbs, J. H. and Gracco, V. L. (1982) Soc. Neurosci. Abstr. 8,282 4 Abend, W., Bizzi, E. and Morasso, P. (1982) Brain 105, 331-348 5 Bizzi, E., Kalil, R. E. and Tagliasco, V. (1971) Science 173,452--454 6 Desmedt, J. E. (ed.) (1978) Cerebral Motor Control in Man: Long Loop Mechanisms, Vol. 4, Karger, Basel 7 Evarts, E. V. (1982) in Speech Motor Control (Grillner, S., Lindblom, B., Lubker, J. and Persson, A., ers), pp. t9--42, Pergamon Press, London 8 Folkins, J. W. and Abbs, J. H. (1975)J. Speech Hear. Bes. 19, 207-220 9 Folkins, J. W. and Abbs, J. H. (1976) J. Speech Hear. Res. 20, 820-821 10 Georgopoulos,A. P., Kalaska, J. F. and Massey,
Biological Timekeeping (Society for ExperimentalBiology, Seminar Series 14) edited by J. Brady, C a m b r i d g e University Press, 1982. £ 2 2 . 5 0 ( H / C ) / £ 9 . 9 5 ( P / B ) (xvii + 1 9 7 pages) I S B N 0 521 2 3 3 0 7 0 ( H / C ) / 1 S B N 0 521 2 9 8 9 9 7 (P/B)
The study of 'biological clocks' in living organisms has become one of the most exciting and rapidly growing fields within the life sciences. Indeed, over ten books on this subject have appeared during the past few years. Nevertheless, there is still a great need for a comprehensive introductory textbook which can be used to introduce advanced undergraduate or graduate students to the field o f biological clocks. In addition, for many new and established investigators, an understanding of biological rhythms has become essential for the design o f experiments and the interpretation of their data, and a general introductory text is needed for these individuals. (Although a recent book, The Clocks That T i m e Us by Moore-Ede, Sulzman and Fuller, Harvard University Press, 1982, does an excellent job of introducing the basic concepts and physiology of biological rhythms in mammals.) As stated in the preface, Biological T i m e k e e p i n g was written to serve as a textbook for students as well as an introductory text for investigators about to embark on research into biological clocks. Unfortunately, Biological T i m e k e e p i n g does not fill that niche. In writing a comprehensive textbook on biological clocks, essentially all levels of biology must be covered, from the biochemical to the physiological to the ecological. The fields of genetics,
395 J. T. (1981)£ Neurophysiol. 4, 725-743 It Gracco, V. L. and Abbs, J. H. (1982)J. Acoust. Soc. Am. 71, SI01 (Abstract) 12 Gracco, V. L. and Abbs, J. H. (1982) Soc. Neurosci. Abstr. 8,282 13 Hasewaga,A., McCutcheon, M. J., Wolf, M. B. and Fletcher, S. G. J. Acoust. Soc. Am. 59, SI (Abstract) 14 Hollerbach, J. M. and Hash, T. (1982) Biol. Cybern. 44, 67-77 15 Holmes,G. (1939) Brain 62, 1-30 16 Houk, J. C. and Rymer, W. Z. (1981) in Handbook of Physiology, Sect. 1 (Vol. 11: Motor Control, Part 1) (Brooks, V.B., ed.), pp. 257-323, American Physiological Society, Bethesda, MD 17 Hughes,O. M. and Abbs, J. H. (1976)Phonetica 33, 19%221 18 lto, M. (1975) in Central Processing of Sensory Input Leading to Motor Output (Evarts, E. V., ed.), pp. 293-304, MIT Press, Cambridge,MA
development, neurobiology and behavior (to name just a few) must be integrated in the book. In addition, the writer(s) of such a book must have sufficient breadth to discuss experiments in single-celled organisms, plants, insects, and vertebrates, particularly birds and mammals. Such an author(s) must be able to write comfortably about the nervous and endocrine systems of mammals and insects, as well as the transport o f information in plants. There are two general approaches that can be taken in writing such a book. First, an expert in one area of biological clocks can devote a substantial portion of his life for a few years attempting to integrate the entire field into a single book. The second, and of course, easy way is to bring together a group of experts and ask each one of them to write a separate chapter. This is exactly what Brady has done, and the resultant book has the same major shortcoming of many multi-authored books: there is little co-ordination between the various chapters. This lack of co-ordination results in two serious problems with Biological Timekeeping. First, much of the same material is covered in different chapters. In the beginning of Chapter I0 (there are 11 chapters in the book), Kluge writes 'Evidently such rhythms are endogenously determined, but in natural conditions respond to the regular change of external stimuli by synchronizing (i.e. entraining) to the periodicity o f the environment'. Since this basic concept is covered in the introductory chapters, as well as elsewhere in the book, it is annoying to find it repeated at the end of the book. There are similar duplications in both concepts and facts throughout the book, and it is not surprising to find the same point being made two or
19 Kelso, J. A. S., Tuller, B. and Fowler, C. A. (1982)J. Acoust. Soc. Am. 72, S103 (Abstract) 20 Lacquaniti, F. and Soechting, J. F. (1982) J. Neurosci. 2,399--408 21 Lindblom, B., Lubker, J. and Gay, T. (1979) J. Phonetics 7, 147-161 22 Miles, F. A. and Evarts, E. V. (1979)Annu. Rev. Psychol. 30, 327-362 23 Morasso, P., Bizzi, E. and Dichgans, J. (1973) Exp. Brain Res. 16,492-500 24 Phillips, C. G. and Porter, R. (1977) Corticospinal Neurones - Their Role in Movement, AcademicPress, London 25 Polit, A. and Bizzi, E. (1979)J. Neurophysiol. 39, 11%142 26 Robinson, D. A. (1974) Brain Res. 71,549 James H. Abbs and Vincent L. Gracco are at the Speech Motor Control Laboratories, Waisman Center, University of Wisconsin-Madison, Madison, W153706, USA.
three times in different chapters. A second problem that arises out of a lack of co-ordination between the various writers is that complex material is presented before adequate introductory information has been supplied (presumably because one author thought it would be covered by another). To give just a few examples: the complexities of models that have been advanced to explain the role of circadian clocks in photoperiodic time measurement are discussed before the reader has a clear understanding of circadian rhythms or the multi-oscillatory nature of circadian systems; incomplete references to the cellular and physiological basis for the generation of circadian rhythms are made in many different chapters before a comprehensive discussion of this subject appears in the f'mal chapter. The only visible attempt to link the various chapters to one another is the use of a cross-referencing system. This enables the reader to know in which other chapters related, or the same, material is covered. I found this to be counter-productive and quite irritating. In reading a single page I was often directed to two or three other chapters, and I was left with the impression that in order to understand what was really being said on a particular page I would first need to read later chapters. While the editor has brought together a group of distinguished scientists who are all established investigators of biological time, little attempt has been made to integrate their writings. As a result, the chapters do not build upon one another and there is no development of a comprehensive story. Rather than being an introductory textbook, Biological T i m e k e e p i n g represents a collection of introductory reviews on some aspects o f biological rhythms. FRED W. TUREK Associate Professor, Department of Neurobiology and Physiology, Northwestern University, Evanston, 1L 60201, USA.
TINS - September 1983
Sensorimotor actions in the control of multi-movement speech gestures James H. Abbs and Vincent L. Gracco The generation o f speech is perhaps the most characteristic human motor act. Historically, owing to the complexity o f speech motor behavior, most analyses have provided only general information on the underlying neural control mechanisms. However, recent work offers some intriguing neurophysiological insights. These studies indicate that: (1) sensorimotor actions are active in minimizing speech movement errors; (2) both intra- and intermovement sensorimotor mechanisms are involved; (3) speech motor-afferent-dependent actions are not stereotypic but have control characteristics that vary, moment-to-moment, in relation to motor execution and task variables; and (4) co-ordination among multiple speech movements is not wholly prespecified, but is subject to dynamic refinement via intermovement sensorimotor adjustments. These observations appear to offer some useful hypotheses f o r the neural control o f other complex motor behaviors.
Most physiological investigations on the While some research groups have focused neural mechanisms of voluntary motor con- on the sensorimotor actions underlying trol have focused on actions around a single maintenance of three-dimensional postural joint. Analyses of sensorimotor responses equilibrium, parallel efforts aimed at 'less to load perturbations similarly have been automatic', multijoint limb movement are confined primarily to single-joint control just beginning to emerge 4'~o.~4,2o. processes". This extensive work has In the last few years, scientists at several yielded numerous insights into the sen~ri- laboratories have utilized perturbation motor control of movement. However, techniques to study sensorimotor control of many motor behaviors in nature are not the multiple co-ordinated actions of human restricted to movements around a single speech 2.3,8`9`11.1z,19,zt. Results of these joint, but require complex actions among studies appear to offer some interesting and multiple movements and multiple muscles. new perspectives on the role of sensori-
~
n
motor mechanisms in the control of complex movements. Even simple speech gestures require co-ordinated movements of the lips, tongue, jaw, pharynx, larynx and respiratory system. Fig. 1 illustrates the lip movements and associated facial muscle contractions (EMG) for the speech gesture 'aba'. As shown, there are several muscles involved in producing these two coordinated movements. The lower-lip muscles, orbicularis oris inferior and mentalis, act to elevate the lower lip while the upperlip muscles, orbicularis otis superior and depressor anguli otis, act to depress the upper lip. The depressor labii inferior acts to depress the lower lip. While it has been suggested that the orbicularis otis muscle is a single sphinctor, electromyographic measures indicate that the inferior and superior portions are controlled independently ~,17. As such, one might consider this speech pattern as analogous to the co-ordinated actions of the elbow and wrist in a multi-joint limb gesture. Perturbation analyses of speech m o v e m e n t control
As noted, perhaps the critical test of whether motor actions are under the influence of ascending afferent signals is to introduce an unanticipated error and examine the adjustments that occur in response to that error. The importance of afferent mechanisms in the control of multimovement speech gestures is apparent from experiments in which unanticipated perturbation analysis has been used.
~
i
Otis
UPPER-LIP DEPRESSORS
UPPER-LIP
'a
b
a' SUPERIOR
MOVEMENTS
LOWER-LIP
~
INFERIOR
lnf.
LOW R-L ' DEPRESSOR
Fig. 1. Oscillographicrecordof upper-lip and lower-lipmovement and associatedfacialmuscleactivityfor the speechgesture 'aba'. Thejaw rnovementis eliminated in this example by having the subject bite on a block of rubber materialplaced between the molars. EMG signalsare rectifiedand smoothed. ~) 1983, Elsevier Science Publishers B.V., Amslerdafa
0378 - 5912/83/$01,00
'I'INA" o 3 c p w m b e r
392 Results from perturbations applied to the two-movement gesture shown in Fig. I provide a useful illustration. Fig. 2 indicates how a d.c. brushless torque motor has been interfaced to the lower lip to introduce unanticipated perturbations during speech. Rotation of the torque-motor sector arm is transmitted to the lower lip via a steel wire and a low-friction lever. A similar configuration has been used to apply loads to the upper lip or jaw. As noted, the torque motor was under force-feedback control, permitting the motor to follow lower- lip, upper- lip or jaw movements at a low, constant tracking force of approximately 4 g. In the experiments involving lower-lip perturbations, inferiorly directed loads ranging from 10 to 40 g were superimposed upon the constant tracking force. These forces, and the displacement perturbations they produced, were within the physiological range of forces and displacements generated during normal lip movements for speech. In these particular experiments, loads were applied to the lower lip in the lO0 ms time window ranging from 50 ms before orbicularis oris inferior contraction to 50 ms into that contraction. To ensure that perturbations were unanticipated, loads were applied randomly on only 10-15% of the movement trials. More frequent perturbations appeared to cause adaptation in the form of a general 'stiffening' of the lips. The major observations from these lower-lip perturbation experiments (illustrated in Figs 3 and 4) indicate the operation of sensorimotor actions. Fig. 3 provides representative examples of upper-lip and lower-lip movement for three control and three loaded conditions. The magnitude and velocity of both upper-lip and lower-lip movements are increased in response to the induced lower-lip perturbation. Fig. 4 shows EMG and upper-lip movement recordings superimposed for a pair of control and loaded trials. The load magnitude for this example was 40 g, and the load was introduced approximately 20 ms prior to the onset of the phasic burst of EMG in the orbicular& oris inferior muscle. Both upper-lip and lower-lip muscles show relatively short latency changes (30-40 ms) in response to the unanticipated load. That is, while the perturbation was applied to the lower lip, compensations were seen in both upper-lip and lower-lip muscles:L The latencies of these compensatory responses are not stereotypic, but range from 25 to 75 ms, depending on the muscle observed and the timing of the load onseP 2. In the various studies where unanticipated perturbations have been applied to speech movements, several remarkable findings have been consistently reported. First, despite some sizeable movement per-
I t/.~j
Adjustable Head M o u n t
Low-
Lip Paddle
Friction Pivot
Torque.
Lood
i
Under Force Feedback Fig. 2. The instrumental configuration employed in applying unanticipated loads to the lower lip during speech. A similar apparatus has been employed to perturb upper-lip and jaw movements.
turbations (> 15 mm) in some studies, the intended speech motor objective is not disrupted in a discernible way (i.e., a listener cannot distinguish the acoustic speech patterns of loaded trials from those generated during unloaded trials). Second, compensations such as those shown in Figs 3 and 4 are observed the first time a load is introduced. These observations would appear to indicate that these sensorimotor compensations reflect 'normal' processes of speech movement control, Other results indicating that these sensorimotor processes are utilized in natural situations (outside the laboratory) are from studies where load magnitudes were limited to 10-15 g, producing displacement perturbations as small as 1 mm (Ref. 2). Even with these small perturbations, the pattern of EMG and movement responses was comparable to those illustrated in Figs 3 and 4. The interpretation that these sensorimotor actions are indicative of 'normal,, on-line control processes is also reflected in verbal reports of the subjects. For example, subjects have typically noted that the loads did not bother them - they simply 'ignored them'. Also, if subjects were instructed to stop all lip movement when they sensed a load, cessation of activity was only accomplished after compensation had occurred and lip closure was achieved. As noted, these results are representative of those from other comparable studies of
speech sensorimotor actions. For example~ when .law loads are introduced during a combined jaw-lip movement gesture (a "p' sound), compensatory responses are observed in both the upper and lower lipss'9. Similarly, upper-lip loads yield both upper-lip and lower-lip compensatory responses. In other experiments, jaw loads introduced during a jaw-tongue gesture (a 'z' sound) yielded compensatory responses in the tongue elevation muscles~. As noted, in all of t ~ experinaents sensorimotor compensations were manifest the first time a load was introduced: the intended speech patterns were never disrupted. These results are consistent with the seemingly unremarkable observations that lip, tongue, and pharyngeal movements are reorganized for adequate speech when the jaw is fixed, as in pipe smoking 21
Neural control processes underlying multlmovement sltet~h gestures In interpreting the results from these various perturbation experiments, one must be cautious in the use of concepts like closedloop or feedback. That is. while an error is introduced in the intended movement of one structure (for example the lower lip), significant compensations are consistently observed in synergistic movements of other structures (for example the upper lip). As such, the compensatory responses cannot be ascribed to a closed-loop, feedback pro-
393
T I N S - September 1 983
cess, especially as these multiple structures are controlled independently. More specifically, the compensatory upper-lip muscle actions do not reflect the comparator-based reactive adjustments characteristic of a feedback system; these adjustments do not correct the induced error in the lower lip. It is possible to demonstrate that the muscles of the upper and lower lips are independently controlled by observing responses to lower-lip loads applied during different speech movement patterns (for example during generation of the sounds 'aba' and 'afa'). For 'aba', both upper-lip and lowerlip movements normally are employed to achieve lip closure, and upper-lip compensation occurs when the lower lip is loaded. By contrast, for 'afa', upper-lip movement is not required, and there is no compensatory upper-lip movement when the lower lip is loaded. In both of these gestures, however, lower-lip compensatory responses to the applied loads are clearly discernible. Given these latter results, it is apparent that when an error is introduced in the lower-lip movement and compensations are observed in both the upper-lip and lower-lip muscles, at least two different sensorimotor processes are involved. As represented in Fig. 5, the compensations in the lower lip appear to be indicative of a classical closed-loop, feedback process. However, by contrast, based upon the obvious dissociation between the site of the induced error and the site of the compensation, the changes in the upper-lip muscles and movements indicate an open-loop adjustment. This open-loop adjustment apparently is based upon a pre-established sensorimotor translation between lower-lip afferent signals and upper-lip motor actions. This intermovement sensorimotor translation has been referred to as a predictive or feedforward processl.~'t~6; i.e., these adjustments are made without a reactive feedback comparator to ensure their corrective accuracy. Similar results have been obtained in studies of eye--head movement co-ordination 5.2a. In those experiments, sensorimotor interactions from neck muscle afferents and the vestibular system were shown to influence eye movements during co-ordinated eye-head movements, reflecting feedforward, open-loop sen~orimotor pathways 7.~a.22.26. From these recent studies of speech, it appears that these predictive adjustments, variously observed to be operating among the jaw, upper lip, lower lip, and tongue, likewise are afferent dependent. As with the results of eye-head movements, these sensorimotor processes may reflect neural mechanisms underlying temporo-spatial movement co-ordination. These interpretations are augmented by earlier observations, indicating a consider-
able degree of temporo-spatial covariance among the upper-lip, lower-lip, jaw, and tongue movements for speech gestures ~3.'7. For example, observation of movements for multiple speech repetitions requiring the same lip--jaw goal indicated that when jaw movements were large, corresponding lip movements were relatively small; when jaw movements were small, corresponding lip movements were relatively large. These observations of intermovement covariance suggest that the sensorimotor actions among independent structures, as revealed by perturbation analysis, may be utilized in a predictive, open-loop manner to generate normal compensatory adjustments and ensure that common multimovement goals are achieved 1. Yet other studies indicate functional differences between the closed-loop and open-loop sensorimotor responses to loads applied during speech gestures. These differences were discernible in the characteristics of load-induced responses in lower-lip muscles versus upper-lip muscles TM. For loads introduced after the onset of the orbicularis oris agonist burst, the open-loop (upper lip) compensations were more robust than the closed-loop (lower lip) compensations. Likewise, it was found that the latencies of the open-loop responses were
generally faster than those of the closedloop responses 12. One might expect classical limitations on closed-loop sensorimotor contributions (for example neural transport delays, consequent instability, etc.) to be minimized in open-loop control because the latter adjustments are predictive rather than reactive. These differences between openloop and closed-loop responses are thus consistent with the relatively greater speed and stability hypothesized for open-loop, feedforward processes 1". Moreover, differences between intra- and intermovement sensorimotor processes are consistent with the impression that the neural control mechanisms underlying co-ordination may be different from the mechanisms involved in controlling single movements. This distinction is most apparent in certain movement disorders where dyscoordination among multijoint actions appears as a disproportionately salient manifestation '5. Regarding the potential neural pathways underlying these sensorimotor control processes, the absence of a consistent, timelocked response pattern appears to mitigate against classical reflex mechanisms TM. Particularly notable is the absence of responses with latencies corresponding to lower brainstem reflexes (for example the perioral reflex has a latency of between I0
CONTROL i
WEN
LOADED
T
5 mm
2
°7V Fig. 3. Examples of upper-lip and lower-lip movements for unloaded (upper panels) and loaded (lower panels) trials.
11N3 - .~epwmber i 9,~5
394
Upper-Up (4
'~t
Depressors
is Superior E !
Upper-Up
~ 'a
b
a'
De~esNon~
Load T
~
_L LOAD ONSET
"
Control
200 ms ~
Lower_Up----------~~ ~~hm~-~--Inferior Elevators \
Mentalis
Fig. 4. Example of EMG and upper-lip movement for a pair of unloaded (light lines) and loaded (heavy lines) trials. Load onset is indicated by the vertical dashed line.
and 18 ms). These observations thus appear to implicate suprabulbar sensorimotor pathways, perhaps involving the extensive sensory and motor connections of the orofacial region with the cerebral cortex and the cerebellum. The monosynaptic connections from pyramidal tract cells to orofacial motoneurons also are consistent with this interpretationL Additionally, the dominance of suprabuibar motorsensory mechanisms in speech control fits the hypothesis of Phillips and Porterz" who argue that there are greater degrees of motorsensory cortex involvement in 'least automatic' behaviors.
Sununary Experiments on sensorimotor mecha, nisms in the control of speech movements
or via actions of single agonistic muscles. Rather, the results observed for these more complex, natural motor gestures indicate that intended nervous system outputs are achieved through a flexible interaction of several sensorimotor control processes, including dosed-loop correction and openloop predictionz~. It is tempting to hypothesize that analysis of co-ordination among the multiple joint movements for
~..~ ~ . ~
~
t
have yielded results that augment those reported in comparable analyses of move-
I
ments around a single joint. For example, earlier observations suggest that sensorimotor processes provide only moderate resistance to induced movement errors. However, when the motor task investigated has multiple degrees of compensatory freedom, the nervous system control variables appear very resistant to such disturbances, This resistance is not manifest simply through one-dimensional control processes
I
Reading list i Abbs. J. H and Cole. K. J (t982) m Speech Motor Control (Gdllner. S,. Lindblom. B.. Lubker. I and Persson. A., eds), pp 159-186,
MUSCLES
I
y
Upper-Lip Movement
(
I
OPEN-LOOP (Feedforward)
limb control would reveal comparable seu~ orimotor actions if investigated in a similar manner. Experiments by Polit and Bizzi~:' indicated that while single-joint lbrearm movements could be executed more or le~s successfully following deafferentatlon, even minor static adjustments of the shoulder (namely, involving a second joint) seriously disrupted forearm movement control in the absence of afferent information. Patterns of intermovement motor equivalence between elbow and wrist movements ~'' also are consistent with the hypothesis that limb co-ordination may be under the influence of open-loop, predictive, sensorimotor actions. From a more abstract perspective, these recent observations of speech indicate that multiple degrees of freedom in complex motor systems may facilitate the neural control process. That is, the freedom to accomplish the same intended motor objective in many different ways relieves the nervous system of the burden of having to prespecify all the complex details of the motor subgestures. Such details apparently are determined via sensonmotor processes operating dynamically and flexibly among the multiple subactions. From this point of view, it might even be argued that movements experimentally constrained around a single joint present a more difficult control problem to the nervous system than their apparently' more complex' multijoint counterparts. The learning and neural programming of a skilled motor behavior certainly would be more demanding if the criterion for successful performance was restricted to a single stereotyped pattern of muscle contraction and movement.
CORRECTIVE ADJI~TMENTS
k
t
INDUCED ERROR
I., !
Lower-Lip Movement
/ I II I t
I I
/
I It------ CLOSED -LOOPlFeedback)
I I !
Fig. 5. Schematic illustration of the hypothesized sensorimotor control pathways underlying compensatzons to lip perturbations.
T I N S - S e p t e m b e r 1 983
PergamonPress, London 2 Abbs, J. H. and Gracco, V. L. (1981)£ Acoust. Soc. Am. 70, $78 (Abstract) 3 Abbs, J. H. and Gracco, V. L. (1982) Soc. Neurosci. Abstr. 8,282 4 Abend, W., Bizzi, E. and Morasso, P. (1982) Brain 105, 331-348 5 Bizzi, E., Kalil, R. E. and Tagliasco, V. (1971) Science 173,452--454 6 Desmedt, J. E. (ed.) (1978) Cerebral Motor Control in Man: Long Loop Mechanisms, Vol. 4, Karger, Basel 7 Evarts, E. V. (1982) in Speech Motor Control (Grillner, S., Lindblom, B., Lubker, J. and Persson, A., ers), pp. t9--42, Pergamon Press, London 8 Folkins, J. W. and Abbs, J. H. (1975)J. Speech Hear. Bes. 19, 207-220 9 Folkins, J. W. and Abbs, J. H. (1976) J. Speech Hear. Res. 20, 820-821 10 Georgopoulos,A. P., Kalaska, J. F. and Massey,
Biological Timekeeping (Society for ExperimentalBiology, Seminar Series 14) edited by J. Brady, C a m b r i d g e University Press, 1982. £ 2 2 . 5 0 ( H / C ) / £ 9 . 9 5 ( P / B ) (xvii + 1 9 7 pages) I S B N 0 521 2 3 3 0 7 0 ( H / C ) / 1 S B N 0 521 2 9 8 9 9 7 (P/B)
The study of 'biological clocks' in living organisms has become one of the most exciting and rapidly growing fields within the life sciences. Indeed, over ten books on this subject have appeared during the past few years. Nevertheless, there is still a great need for a comprehensive introductory textbook which can be used to introduce advanced undergraduate or graduate students to the field o f biological clocks. In addition, for many new and established investigators, an understanding of biological rhythms has become essential for the design o f experiments and the interpretation of their data, and a general introductory text is needed for these individuals. (Although a recent book, The Clocks That T i m e Us by Moore-Ede, Sulzman and Fuller, Harvard University Press, 1982, does an excellent job of introducing the basic concepts and physiology of biological rhythms in mammals.) As stated in the preface, Biological T i m e k e e p i n g was written to serve as a textbook for students as well as an introductory text for investigators about to embark on research into biological clocks. Unfortunately, Biological T i m e k e e p i n g does not fill that niche. In writing a comprehensive textbook on biological clocks, essentially all levels of biology must be covered, from the biochemical to the physiological to the ecological. The fields of genetics,
395 J. T. (1981)£ Neurophysiol. 4, 725-743 It Gracco, V. L. and Abbs, J. H. (1982)J. Acoust. Soc. Am. 71, SI01 (Abstract) 12 Gracco, V. L. and Abbs, J. H. (1982) Soc. Neurosci. Abstr. 8,282 13 Hasewaga,A., McCutcheon, M. J., Wolf, M. B. and Fletcher, S. G. J. Acoust. Soc. Am. 59, SI (Abstract) 14 Hollerbach, J. M. and Hash, T. (1982) Biol. Cybern. 44, 67-77 15 Holmes,G. (1939) Brain 62, 1-30 16 Houk, J. C. and Rymer, W. Z. (1981) in Handbook of Physiology, Sect. 1 (Vol. 11: Motor Control, Part 1) (Brooks, V.B., ed.), pp. 257-323, American Physiological Society, Bethesda, MD 17 Hughes,O. M. and Abbs, J. H. (1976)Phonetica 33, 19%221 18 lto, M. (1975) in Central Processing of Sensory Input Leading to Motor Output (Evarts, E. V., ed.), pp. 293-304, MIT Press, Cambridge,MA
development, neurobiology and behavior (to name just a few) must be integrated in the book. In addition, the writer(s) of such a book must have sufficient breadth to discuss experiments in single-celled organisms, plants, insects, and vertebrates, particularly birds and mammals. Such an author(s) must be able to write comfortably about the nervous and endocrine systems of mammals and insects, as well as the transport o f information in plants. There are two general approaches that can be taken in writing such a book. First, an expert in one area of biological clocks can devote a substantial portion of his life for a few years attempting to integrate the entire field into a single book. The second, and of course, easy way is to bring together a group of experts and ask each one of them to write a separate chapter. This is exactly what Brady has done, and the resultant book has the same major shortcoming of many multi-authored books: there is little co-ordination between the various chapters. This lack of co-ordination results in two serious problems with Biological Timekeeping. First, much of the same material is covered in different chapters. In the beginning of Chapter I0 (there are 11 chapters in the book), Kluge writes 'Evidently such rhythms are endogenously determined, but in natural conditions respond to the regular change of external stimuli by synchronizing (i.e. entraining) to the periodicity o f the environment'. Since this basic concept is covered in the introductory chapters, as well as elsewhere in the book, it is annoying to find it repeated at the end of the book. There are similar duplications in both concepts and facts throughout the book, and it is not surprising to find the same point being made two or
19 Kelso, J. A. S., Tuller, B. and Fowler, C. A. (1982)J. Acoust. Soc. Am. 72, S103 (Abstract) 20 Lacquaniti, F. and Soechting, J. F. (1982) J. Neurosci. 2,399--408 21 Lindblom, B., Lubker, J. and Gay, T. (1979) J. Phonetics 7, 147-161 22 Miles, F. A. and Evarts, E. V. (1979)Annu. Rev. Psychol. 30, 327-362 23 Morasso, P., Bizzi, E. and Dichgans, J. (1973) Exp. Brain Res. 16,492-500 24 Phillips, C. G. and Porter, R. (1977) Corticospinal Neurones - Their Role in Movement, AcademicPress, London 25 Polit, A. and Bizzi, E. (1979)J. Neurophysiol. 39, 11%142 26 Robinson, D. A. (1974) Brain Res. 71,549 James H. Abbs and Vincent L. Gracco are at the Speech Motor Control Laboratories, Waisman Center, University of Wisconsin-Madison, Madison, W153706, USA.
three times in different chapters. A second problem that arises out of a lack of co-ordination between the various writers is that complex material is presented before adequate introductory information has been supplied (presumably because one author thought it would be covered by another). To give just a few examples: the complexities of models that have been advanced to explain the role of circadian clocks in photoperiodic time measurement are discussed before the reader has a clear understanding of circadian rhythms or the multi-oscillatory nature of circadian systems; incomplete references to the cellular and physiological basis for the generation of circadian rhythms are made in many different chapters before a comprehensive discussion of this subject appears in the f'mal chapter. The only visible attempt to link the various chapters to one another is the use of a cross-referencing system. This enables the reader to know in which other chapters related, or the same, material is covered. I found this to be counter-productive and quite irritating. In reading a single page I was often directed to two or three other chapters, and I was left with the impression that in order to understand what was really being said on a particular page I would first need to read later chapters. While the editor has brought together a group of distinguished scientists who are all established investigators of biological time, little attempt has been made to integrate their writings. As a result, the chapters do not build upon one another and there is no development of a comprehensive story. Rather than being an introductory textbook, Biological T i m e k e e p i n g represents a collection of introductory reviews on some aspects o f biological rhythms. FRED W. TUREK Associate Professor, Department of Neurobiology and Physiology, Northwestern University, Evanston, 1L 60201, USA.